id
stringlengths 27
33
| source
stringclasses 1
value | format
stringclasses 1
value | text
stringlengths 13
1.81M
|
|---|---|---|---|
warning/0003/nucl-th0003064.html
|
ar5iv
|
text
|
# A two-pion exchange three-nucleon force based on a realistic 𝜋𝑁 interaction
## I Introduction
Recently T.-Y. Saito and I. R. Afnan (SA) have calculated the contribution of a $`\pi \pi `$-exchange three-body force (see Fig. 1) to the three-nucleon binding energy in terms of the $`\pi N`$ amplitude using perturbation theory. Their approach determines the contributions of the different $`\pi N`$ partial waves and (via the division of the $`\pi N`$ amplitude into a pole and nonpole term) allows for a consistent determination of the $`\pi NN`$ form factor. The calculations are based on phenomenological separable $`\pi N`$ potentials . The total contribution of this three-body force (TBF) to the binding energy of the triton has been found to be very small. It is typically of the order of a few keV. This result falls short of calculations based on the Tucson-Melbourne (TM) and the Brazilian $`\pi \pi `$-exchange three-nucleon potentials . The latter potentials make a contribution to the binding energy of the triton that is of the order of the discrepancy between experiment and calculations with realistic nucleon-nucleon potentials (i. e. about 1 MeV).
The origin of this surprisingly large difference of two orders of magnitude in the contribution of the three-body force has been the topic of two subsequent very detailed investigations by Saito and Afnan and Murphy and Coon . SA concluded that the total contribution of the TBF to the binding energy of the triton found in their approach is so small as a result of the energy dependence of the $`\pi N`$ amplitudes, cancellations between the contributions from the S- and P-wave $`\pi N`$ partial waves, and in particular, the soft $`\pi NN`$ form factor. Indeed the form factors extracted from their $`\pi N`$ interaction models correspond to monopole cutoff masses of around or even less than 0.4 GeV - which have to be compared to values of about 0.7 - 0.8 GeV suggested from other information and to the value of 0.8 GeV used in calculations with the TM potential.
Murphy and Coon carried out a thorough comparison of the $`\pi N`$ amplitudes that underlie the TM TBF and the calculations of SA . They criticized that the amplitudes used by SA do not fulfil the low-energy theorems of the $`\pi N`$ interaction as imposed by chiral symmetry. But still they attested that these amplitudes are qualitatively similar to the one used in the TM potential. The prime reason for the two-orders-of-magnitude discrepancy found in the contributions of the TBF was suspected to be likewise the soft $`\pi NN`$ form factor emerging from the phenomenological separable $`\pi N`$ interactions employed by SA.
In the present paper we want to re-investigate the origin of this large discrepancy in the predicted contribution of the TBF. We follow the same approach as SA. However, we start out from a meson-theoretical $`\pi N`$ model developed recently by the Jülich group . This model, besides being conceptionally much better founded than the simple separable potentials employed in Refs. , has the important advantage that it does not exhibit those deficiencies which led to a criticism on the work by Saito and Afnan. Firstly, the Jülich model is in agreement with empirical information on the $`\pi N`$ amplitude in the subthreshold region . In particular its prediction for the amplitude at the so-called Cheng-Dashen point is close to the empirical value. Second, and most importantly it yields $`\pi NN`$ form factors which are comparable to those of the TM potential, with monopole cutoff masses of around 0.7 GeV . Therefore this model provides an ideal starting point for re-assessing the role of the $`\pi \pi `$-exchange TBF in the binding of the three-nucleon system.
The paper is structured in the following way: In Section II we review the salient features of the Jülich $`\pi N`$ model. In particular we concentrate on those properties that are relevant for the present study. The specific structure of our three-nucleon code makes it necessary to represent the meson-theoretical $`\pi N`$ amplitude in separable form. For this purpose we applied the so-called EST method which allows to generate separable representations that agree exactly (on- and half-off-shell) with the original interaction at specific predetermined energies. This method is briefly described also in Section II. Furthermore, we discuss the reliability of the separable representation by comparing (off-shell) amplitudes obtained from it to the ones of the original Jülich model for various $`\pi N`$ partial waves.
The formulation of the TBF is given in Section III together with a short outline of the formalism. Results for the contribution of the $`\pi \pi `$-exchange TBF to the triton binding energy, based on the $`\pi N`$ amplitude of the Jülich model are presented in Section IV. Anticipating our results we find again that the TBF is very small. Therefore, in Section V, we embark on a detailed discussion of the $`\pi N`$ amplitudes on which the TM TBF and our calculation are based. We focus specifically on the off-shell properties of these amplitudes since they are determined quite differently in the two approaches. Indeed, we will argue that much of the observed two-orders-of-magnitude discrepancy in the binding energy is due to off-shell effects and we will present numerical evidence for this conjecture. Finally, a summary is given in Section VI.
## II The $`\pi N`$ interaction
The $`\pi N`$ models employed in the present investigation are based on meson exchange and have been developed by the Jülich Group . They include the s-channel and u-channel nucleon and delta-isobar pole diagrams together with correlated $`\pi \pi `$ exchange in the $`J^P=0^+`$ ($`\sigma `$) and $`1^{}`$ ($`\rho `$) channels as shown in Fig. 2. The interaction potentials are derived in time ordered perturbation theory and then unitarized by means of a relativistic (Lippmann-Schwinger type) scattering equation
$$T=V+VG_0T.$$
(1)
The resulting models account for the scattering data in the elastic region as well as for the low-energy parameters . Furthermore they also satisfy chiral symmetry constraints. In particular, the resulting values for the so-called $`\pi N`$ $`\mathrm{\Sigma }`$ term (the contribution of the isoscalar forward scattering amplitude at the Cheng-Dashen point) - $`\mathrm{\Sigma }`$ = 66.4 (65.6) MeV for model 1 (2) of Ref. \- are in good agreement with the empirical value of $`\mathrm{\Sigma }`$ = 60 MeV .
Since in the discussion of Murphy and Coon special attention was given to the subthreshold behaviour of the $`\pi N`$ amplitudes, we want to present here the corresponding results for the Jülich $`\pi N`$ model. In the continuation of the T-matrix for the potential models to the subthreshold region, we follow the procedure outlined in section 4 of Ref. . We calculated the (on-shell) background isoscalar amplitude $`\overline{F}^+(\nu ,t)`$ (conventionally called $`\overline{D}^+`$; for definition see, e.g., Refs. ) from the subthreshold point $`\nu =0`$, $`t=0`$ to the Cheng-Dashen point ($`\nu =0`$, $`t=2m_\pi ^2`$). The predictions of the Jülich models 1 and 2 are shown in Fig. 3 in comparison to the ones of the $`\pi N`$ amplitudes that form the basis of the TM and Brazil $`\pi \pi `$-exchange TBF. One can see that the results of the meson-exchange models more or less coincide with the $`\pi N`$ amplitude employed in the TM TBF. Furthermore, they are also in rather nice agreement with the empirical subthreshold amplitude given by Höhler .
In a recent paper C. Schütz et al. have determined the $`\pi NN`$ vertex functions resulting from the models of Ref. (cf. Ref. and the Appendix for definitions and relevant formulue). From the vertex functions at the nucleon pole, the decrease in the $`\pi NN`$ form factors $`F_{\pi NN}(p^2)`$ from the pion pole $`p^2=m_\pi ^2`$ to $`p^2=0`$, has been extracted following the procedure proposed by Mizutani et al. . This quantity is a measure for the softness of the $`\pi NN`$ form factor. It has been found that the $`\pi NN`$ form factors implied by the models considered in Ref. are, in general, significantly harder than the ones used by SA in their study of the contribution of the $`\pi \pi `$-exchange three-nucleon force to the $`3N`$ binding energy. In particular, the model 2’ of Ref. yields a decrease in the form factor from the pion pole to $`p^2=0`$ of slightly less then 4% - which is quite close to the value of 3% implied by the form factor introduced in the TM three-nucleon force and also in good agreement with a recent lattice QCD calculation and other independent information . Accordingly, the major concern raised by Murphy and Coon against the work of SA does not apply for this model and therefore it provides an excellent starting point for re-analyzing the contributions of the $`\pi \pi `$-exchange three-nucleon force to the $`3N`$ binding energy in the approach of Saito and Afnan.
In this work we will also present results for the other models considered in Ref. and it is appropriate to say some words about the basic differences between these models. All the models are based on the same dynamical input (cf. Fig. 2). They differ, however, in the (phenomenological) parametrization of the (bare) vertex form factors. Details and explicit formulae can be found in sect. III of Ref. . Both models provide a similarly good description of the $`\pi N`$ scattering data. However, the differences in the parametrization of the vertex form factors lead to different vertex functions and in turn to different (dressed) $`\pi NN`$ form factors.
Two further models (1’ and 2’) have been presented in Ref. for the following reason: The models 1 and 2 are constructed by assuming $`\pi NN`$ pseudovector ($`pv`$) coupling. However, the $`\pi NN`$ form factor derived from lattice QCD calculations , and the values of $`F_{\pi NN}(0)`$ from other independent information (and also the values given for the models used by SA), are based on pseudoscalar ($`ps`$) coupling. In order to allow for a meaningful comparison for the different couplings the authors of Ref. have constructed variants of models 1 and 2 (labelled 1’ and 2’) where $`ps`$ coupling is used in the nucleon s-channel pole terms.
Values for $`F_{\pi NN}(0)`$ for the various models are compiled in Table I. (Note that $`F_{\pi NN}`$ is normalized to $`F_{\pi NN}(m_\pi ^2)=1`$.) For the ease of comparison we also include here the result for the separable $`\pi N`$ model $`PJ`$ by McLeod and Afnan which has been used (amongst others) in the investigation by SA and on which most of the concerns and criticism of Murphy and Coon are based. It is evident that for this model the decrease of the form factor from the pion pole to $`p^2=0`$ is much larger - almost 20%. We want to emphasize here, however, that it is the $`\pi NN`$ vertex function which enters into the calculation of the TBF and not the form factor (cf. Section III). Therefore we also show this quantity (cf. Fig. 4). Note that the $`\pi NN`$ vertex functions for models 1 and 1’ are practically the same. Accordingly we expect these models to give the same result for the contributions of the $`\pi \pi `$-exchange TBF to the $`3N`$ binding energy, and therefore we will consider only model 1 in the following analysis. Furthermore, the vertex function for model 2 is obviously harder than the one resulting from model 2’ (for small and intermediate momenta) - in contrast the form factors for model 2’ looks harder (cf. Table I). This seemingly paradoxical situation has been discussed thoroughly in Ref. .
Before carrying out the actual three-nucleon calculations the meson-theoretical $`\pi N`$ models have to be expanded in separable form. This is necessitated by the specific structure of our three-nucleon code which can only deal with interaction models given in separable form. For this purpose we apply the so-called Ernst-Shakin-Thaler (EST) method which allows us to generate separable representations of arbitrary rank $`N`$ that agree exactly (on- and half-off-shell) with the original reaction matrix at $`N`$ specific predetermined energies.
Let us begin with the (partial wave projected) Lippmann-Schwinger equation for the radial wave function
$$|\psi _E=|k_E+G_0(E)V|\psi _E,$$
(2)
where $`|k_E`$ is the incoming wave and $`G_0(E)`$ the two-body Green’s function. (In Eq. (2) and in the following the partial wave index is suppressed for convenience.) For proper scattering solutions (on-shell) $`k_E`$ and $`E`$ are related by
$$E=\sqrt{m_N^2+k_E^2}+\sqrt{m_\pi ^2+k_E^2}.$$
(3)
According to the EST method a rank-$`N`$ separable representation for the potential $`V`$ is given by
$$\stackrel{~}{V}=\underset{i,j=1}{\overset{N}{}}V|\psi _{E_i}\lambda _{ij}\psi _{E_j}|V,$$
(4)
where $`E_i`$, ($`i=1,\mathrm{},N`$), are $`N`$ freely chosen energies. The coupling strengths $`\lambda _{ij}`$ are determined by the condition
$$\underset{j=1}{\overset{N}{}}\lambda _{ij}\psi _{E_j}|V|\psi _{E_k}=\delta _{ik}.$$
(5)
It is evident from Eq. (4) that the ”form factors” of the separable potential $`\stackrel{~}{V}`$ consist of the objects $`V|\psi _{E_i}`$, where $`|\psi _{E_i}`$ are solutions of Eq. (2) for the potential $`V`$ at the energies $`E_i`$. Therefore, by virtue of Eq. (5), the following relation holds
$$\stackrel{~}{V}|\psi _{E_i}=V|\psi _{E_i}=T(E_i)|k_{E_i}=\stackrel{~}{T}(E_i)|k_{E_i}$$
(6)
at the $`N`$ energies $`E_i`$, where $`\stackrel{~}{T}`$ is the solution of the Lippmann-Schwinger equation (1) for the separable representation $`\stackrel{~}{V}`$. This means that the on-shell as well as the half-off-shell t-matrix for both interactions $`V`$ and $`\stackrel{~}{V}`$ are exactly the same at the energies $`E_i`$.
In the present case the interaction models $`V`$ are energy-dependent and therefore a modification of this scheme proposed by B. Pearce is employed. According to it the condition
$$\psi _{E_l}|V(E)|\psi _{E_k}=\psi _{E_l}|\stackrel{~}{V}(E)|\psi _{E_k}=\underset{i,j=1}{\overset{N}{}}\psi _{E_l}|V(E_i)|\psi _{E_i}\lambda _{ij}(E)\psi _{E_j}|V(E_j)|\psi _{E_k}$$
(7)
has to be used for determining the coupling strengths $`\lambda _{ij}(E)`$ instead of Eq. (5). As expected also the separable representation becomes now energy-dependent.
A special treatment is required for the $`P_{11}`$ partial wave which contains the (s-channel) nucleon pole. It must be possible to clearly separate the contribution of this pole term from the total $`P_{11}`$ amplitude. Its contribution to the three-nucleon binding energy is already taken into account by solving the standard bound-state Faddeev equations. Therefore, in order to avoid double counting, only the non-pole part of the $`P_{11}`$ must be considered in the present investigation. Furthermore in the consistent approach of SA the $`\pi NN`$ vertex function is extracted from this pole term and is then used for the vertices where the pions are emitted (absorbed) by (at) the outer nucleons. Consequently a separable representation for the $`P_{11}`$ partial wave must guarantee that (a) the non-pole amplitude is reliably reproduced and (b) the $`\pi NN`$ vertex function extracted from the pole term agrees exactly with the one obtained for the original interaction model. This can be achieved and we summarize details of the construction procedure in the Appendix.
Of course if one relies on such separable representations one has to ensure that they incorporate all the relevant properties of the original interaction models. From extensive tests, we find that a rank-1 separable representation is sufficient for the present purpose provided that the expansion energy is choosen in the relevant energy domain (i. e. around the triton binding energy which corresponds to a center-of-mass energy of roughly 930 MeV in the $`\pi N`$ system). Thus we have selected this particular energy for the separable representation to be applied in the present study. Since this energy is below the elastic $`\pi N`$ threshold, $`k_E`$ in Eq. (2) can no longer be fixed by the on-shell condition. Following our previous work we choose $`k_E`$ in such a way that $`ik_E`$ fulfils Eq. (3). For $`E_1=930`$ MeV this implies $`k_{E_1}138`$ MeV/c.
The quality of the separable represention is demonstrated in the Figs. 5 and 6 for model 2. Fig. 5 shows the off-shell transition amplitude $`T_\alpha (q,q^{};Z)`$ for fixed off-shell momenta $`q=q^{}=130`$ MeV as a function of the total energy $`Z`$. Note that in case of the $`P_{11}`$ partial wave only the non-pole part of the t-matrix is shown since, as explained above, this is the part relevant for the present study. Fig. 6 shows the transition amplitude $`T_\alpha (q,q^{};Z)`$ for fixed $`Z`$ and $`q^{}`$ as a function of the other off-shell momentum. In the latter figure we display only results for those partial waves that are expected to give the dominant contribution of the $`\pi \pi `$-exchange three-nucleon force to the $`3N`$ binding energy , namely $`S_{31}`$, $`P_{11}`$, and $`P_{33}`$. We want to point out, however, that the agreement in the other partial waves is of similar quality. Likewise we want to refrain from displaying corresponding results for the model 1 here since the quality of its separable representation is pretty much the same. Furthermore, we do not show the $`\pi NN`$ vertex functions resulting from the separable representation because - by construction - they are identical to the ones of the original model (cf. the Appendix).
Note that we have also constructed rank-1 separable representations where the expansion energies are near the $`\pi N`$ threshold ($`E_1=1077`$ MeV for $`S_{11}`$; $`E_1=1000`$ MeV for $`S_{31}`$; $`E_1=1100`$ MeV for $`P_{11}`$, $`P_{31}`$, $`P_{13}`$, $`P_{33}`$). (In case of the $`S_{31}`$ partial wave the amplitude around 1077 MeV has a peculiar energy dependence which would require, in principle, a rank-2 representation. In order to avoid this complication we have chosen a somewhat lower value for $`E_1`$.) These representations will be also employed in our investigations. They will serve as a term of reference for how strongly the resulting $`3N`$ binding energies depend on the specific choice of the separable representation.
## III Formulation of the three-body force
In this section we formulate the TBF using the $`\pi N`$ amplitude and $`\pi NN`$ vertex function which have been derived in the previous section.
We follow the prescription by SA. The TBF is schematically shown in Fig. 7, i.e., (i) a pion is emitted from the first nucleon, (ii) the pion is scattered off the second nucleon, (iii) the pion is absorbed by the third nucleon. The strength function of pion emission and absorption is given by the $`\pi NN`$ vertex function which depends on the energy of the $`\pi N`$ system as explained in the previous section. Also the $`\pi N`$ amplitude corresponding to the scattering of the pion on the second nucleon is energy dependent and includes only the non-pole contribution in the $`P_{11}`$ channel. We introduce Jacobi variables in the $`\pi NNN`$ system so that we define the $`\pi N`$ relative momenta and energies at the stages (i), (ii) and (iii).
The momenta of the three nucleons before the three-body interaction in Fig. 7 are given by $`𝐤_1`$, $`𝐤_2`$ and $`𝐤_3`$ and the momenta after the interaction are $`𝐤_1^{}`$, $`𝐤_2^{}`$ and $`𝐤_3^{}`$, respectively. We define the relative momenta between the third nucleon and the system of first and second nucleons and pion ($`𝐪_3`$), between the second nucleon and the system of first nucleon and pion ($`𝐩_3`$) and between the first nucleon and pion ($`𝐐_3`$) at the stage (i)
$`𝐪_3`$ $`=`$ $`𝐤_3,`$ (8)
$`𝐩_3`$ $`=`$ $`{\displaystyle \frac{m_N(𝐤_\pi +𝐤_1^{})(m_N+m_\pi )𝐤_2}{(2m_N+m_\pi )}},`$ (9)
$`𝐐_3`$ $`=`$ $`{\displaystyle \frac{m_N𝐤_\pi m_\pi 𝐤_1^{}}{(m_N+m_\pi )}},`$ (10)
where $`𝐤_\pi `$ is the pion momentum. Then the center-of-mass energy in the system of first nucleon and pion, $`E_3`$, is obtained using those relative momenta
$$E_3=E+m_N\frac{q_3^2}{2\mu _2}\frac{p_3^2}{2\mu _1}$$
(11)
where $`E=E_T`$ is the total energy of the whole system not including rest masses, and the reduced masses $`\mu _1`$ and $`\mu _2`$ are defined, respectively, by the relations
$$\frac{1}{\mu _1}=\frac{1}{m_N}+\frac{1}{m_N+m_\pi },$$
(12)
and
$$\frac{1}{\mu _2}=\frac{1}{m_N}+\frac{1}{2m_N+m_\pi }.$$
(13)
In the same way we define the relative momenta and energies of the system of second nucleon and pion, $`𝐐_3^{}`$ (before scattering), $`𝐐_1^{}`$ (after scattering) and $`E_2`$ at the stage (ii) and $`𝐐_1`$ and $`E_1`$ of the system of third nucleon and pion at the stage (iii), respectively.
Using these $`\pi N`$ relative momenta and energies the TBF, $`W(E)`$, can be symbolically written as
$$W(E)=v_{\pi N}^R(Q_1;E_1)G_{\pi NNN}(E)T_{\pi N}(Q_1^{},Q_3^{};E_2)G_{\pi NNN}(E)v_{\pi N}^R(Q_3;E_3).$$
(14)
Here, $`v_{\pi N}^R(Q_i;E_i)`$ ($`i=`$ 1, 3) is the renormalized $`\pi NN`$ vertex function (cf. the Appendix) which gives the strength of the pion emission and absorption on the nucleon. $`G_{\pi NNN}(E)`$ is the propagator of the $`\pi NNN`$ system and $`T_{\pi N}(Q_1^{},Q_3^{};E_2)`$ is the non-pole part of the $`\pi N`$ scattering amplitude. Eq. (14) represents how the pion is emitted from the first nucleon, scattered off the second nucleon, and then absorbed by the third nucleon. Note that the $`\pi N`$ amplitude and the $`\pi NN`$ vertex function are determined in the same framework, i. e. they are obtained from the same $`\pi N`$ interaction model. A detailed discussion of Eq. (14) can be found in Ref. . Since in the Jülich model the energy is defined fully relativistic whereas a semi-relativistic form is employed by SA, we have to change the $`\pi NNN`$ propagator of Ref. to be
$$G_{\pi NNN}(E)=\left(E_i\sqrt{Q_i^2+m_N^2}\sqrt{Q_i^2+m_\pi ^2}\right)^1,$$
(15)
with $`i=`$ 1 or 3 (cf. Eq. (4.9) of Ref. ).
## IV Results
The contribution of the TBF to the binding energy of the three-nucleon system, $`\mathrm{\Delta }E^{(3)}`$, is calculated in first order perturbation theory, i. e.
$$\mathrm{\Delta }E^{(3)}=\mathrm{\Psi }|W(E_T)|\mathrm{\Psi },$$
(16)
where $`|\mathrm{\Psi }`$ is the triton wave function. This wave function is obtained from solving the Faddeev equations for the so-called PEST potential which is a separable representation of the Paris $`NN`$ potential derived by the EST method. All nucleon-nucleon partial waves with total angular momentum less than or equal two are employed in the calculation. The triton properties obtained by the PEST potential are comparable with those by the original Paris potential as shown in Refs. .
Evidently, like in the work by SA there is no consistency between the $`NN`$ interaction which is used to generate the triton wave function and the TBF. However, we are making use of the Born approximation and therefore the triton wave function and the TBF are obtained separately, anyway. Thus, as argued already by SA , we do not expect that this inconsistency has an influence on the qualitative features of our results. One should also keep in mind that a similar inconsistency is involved in standard $`3N`$ calculations employing the TM TBF.
The contributions of the TBF generated by the $`\pi N`$ models 1, 2 and 2’ of Ref. to the $`3N`$ binding energy are listed in Table I. These results are based on the separable representations of the original interaction models described in the preceeding section. Note that the value of 930 MeV is used for the expansion energy since we expect the average $`\pi N`$ energy in the three-nucleon system to be near this value. In Table I also the result for model $`PJ`$ of Ref. is listed for the sake of comparison. Thus, it is easy to see that our results are qualitatively very similar to those of SA. Again the contibutions of the individual $`\pi N`$ partial waves to the $`3N`$ binding energy are very small (only in the order of $`keV`$) and there is again a strong cancellation between the attractive $`P_{11}`$ and $`P_{33}`$ partial waves and the repulsive $`S_{31}`$ partial wave.
In Ref. Murphy and Coon have presented a thorough comparison of the $`\pi N`$ amplitudes used in the TM and Brazilian TBF and the one applied in the calculations by SA. Their main conclusion was that the two-orders-of-magnitude smaller value for the binding energy resulting from the $`\pi \pi `$-exchange TBF based upon the separable models of SA is most likely due to the very soft form factor resulting from these potentials. Therefore we want to look now at the $`\pi N`$ form factors extracted from the models applied in the present study. The values for $`F_{\pi NN}(0)`$ are also given in Table I. Note that model 2’ yields the hardest $`\pi NN`$ form factor which corresponds to a monopole form factor with a cutoff mass of about 708 MeV whereas the model $`PJ`$ of SA corresponds to a monopole form factor with a cutoff mass of just 317 MeV. From comparing different columns of Table I one might conclude that there is some influence of the form factor on the magnitude of the TBF. However, the variations in the individual partial wave contributions are only around a factor of two or three and definitely not two orders of magnitude. Furthermore, the cancellation effects are independent of the softness of the $`\pi NN`$ form factor and they tend to reduce the variations in the total contribution. In fact, the results for the $`\pi N`$ models considered in the present paper lie all within a range of 15 $`keV`$, as can be seen from Table I.
At this point one may wonder how reliable results based on a rank 1 separable approximation of the Jülich models are. In order to estimate the uncertainty due to the simplicity of the representation we constructed another rank 1 represention where the expansion energy was chosen at $`\pi N`$ threshold (for the S-waves) or slightly above. Specifically we chose 1077 MeV for $`S_{11}`$, 1000 MeV for $`S_{31}`$ and 1100 MeV for the other partial waves. The results for these alternative separable representations are compared with the ones obtained for our ’standard’ choice in Table II. From this Table we see that there are variations of the order of $`2030\%`$ in some partial waves - but qualitatively there is no change in the results. Therefore we are confident that the used rank-1 separable representations are sufficiently accurate for the aim of the present investigation.
## V Discussion
We conclude from the previous section that the softness of the $`\pi NN`$ form factor is not responsible for the smallness of the $`\pi \pi `$-exchange TBF based on $`\pi N`$ potential models. If so, what makes the contribution to the three-body binding energy so small? In order to shed some light on this let us examine the basic two differences between the TM TBF and the one derived from a $`\pi N`$ potential model. The first difference concerns the energy dependence. The original $`\pi N`$ amplitude on which the TM TBF is based, is given in a covariant form and therefore depends on the energy (of the pion). However, in order to make this TBF suitable for application in standard (non-relativistic) 3N calculations the $`\pi N`$ amplitude is expanded in powers of $`k_\pi /m_N`$ and only the lowest order terms are kept. As a consequene the initial energy dependence drops completely out of the resulting TBF. In the approach of SA the TBF is energy dependent in a two-fold way, namely via the $`\pi N`$ amplitude but also via the $`\pi NN`$ form factor. The effect of switching off this energy dependence in the TBF has been analyzed thoroughly in Ref. where it was found that it leads to a sizeable increase in the resulting binding energy. However, it was concluded by SA that the approximation of fixing the energy in the $`\pi N`$ amplitude and the $`\pi NN`$ form factor does not lead to a sufficiently large change in the contributions to make them comparable with the result for the TM TBF. None the less we would like to look at this point again because (unlike the potentials employed by SA) now the $`\pi N`$ interaction model itself is energy dependent and therefore the effects from the energy dependence might be stronger. Note that in the present case the energy dependence of the TBF enters at three levels: (a) the energy dependence of the $`\pi N`$ potential itself; (b) the energy dependence of the $`\pi N`$ t-matrix; (c) the energy dependence of the $`\pi NN`$ vertex function.
We investigate the role of the energy dependence by fixing the energy at $`E=930`$ MeV. This is the value chosen for constructing the separable interaction via the EST method and, accordingly, where the $`\pi N`$ amplitudes generated by the Jülich model and its separable representation agree almost exactly. Thus, the results for our rank-1 separable interaction should be practically identical to the one for the original Jülich model in the particular case where the whole energy dependence (a)-(c) is fixed. Corresponding values are given in the last column of Table III. The numbers in the third column of Table III are obtained by fixing only the energy dependence of the $`\pi N`$ interaction. For the ease of comparison results without restricting the energy dependence are also shown in the Table. We see that the contribution of each $`\pi N`$ partial wave is enhanced by about 50% after fixing the energy dependence of the potential. Fixing the energy dependence also in the $`\pi N`$ t-matrix and the $`\pi NN`$ form factor leads to a further increase in the contributions from the $`P_{11}`$ and $`P_{33}`$ waves and to a slight suppression in the $`S_{31}`$. However, those approximations definitely do not provide any really substantial enhancement of the resulting triton binding - in line with the findings of SA. All results shown in Table III are obtained by using model 2’. The other models behave qualitatively very similar and therefore we don’t give the corresponding numbers here.
The other major difference between the TM force and the TBF based on a $`\pi N`$ potential concerns the off-shell extrapolation of the $`\pi N`$ amplitude. In the Jülich model the off-shell properties of the $`\pi N`$ amplitude are completely determined by the dynamical ingredients of the $`\pi N`$ model and the fit to the $`\pi N`$ data. Since also for the potential models employed by SA the off-shell properties are, in principle, constrained by a fit to $`\pi N`$ data let us emphasize the main difference here. In case of phenomenological separable interactions the off-shell behavior is, to a large extend, determined by the specific choice of the form-factor function. Moreover, separable interactions act only in single partial waves and accordingly the free parameters are determined by fitting only a single partial wave. On the other hand, in a meson-exchange potential like the Jülich model the dynamical ingredients give, in general, contributions to all partial waves and therefore the free parameters in this model - which are essentially the cutoff masses in the (baryon-baryon-meson) vertex form factors \- are much better constrained by a fit to the $`\pi N`$ data. Clearly also here differences in the dynamics and/or differences in the parametrization of the vertex form factors will lead to variations in the off-shell properties of the resulting $`\pi N`$ amplitude. Indeed such differences exist between the models 1 and 2 (or 2’) considered here. But, as we have already seen in the last section, they to not lead to any significant variations in the results for the TBF.
The $`\pi N`$ amplitude used in the TM force is given by (cf., e.g., Ref. )
$$T_{\pi N}^{ij}(𝐤_\pi ,𝐤_\pi ^{})=F_{\pi NN}(𝐤_\pi ^2)F_{\pi NN}(𝐤_\pi ^2)\left\{\delta ^{ij}\left[a+b𝐤_\pi 𝐤_\pi ^{}+c(𝐤_\pi ^2+𝐤_\pi ^2)\right]dϵ^{ijk}\tau ^k\sigma 𝐤_\pi \times 𝐤_\pi ^{}\right\},$$
(17)
where $`i,j`$ are pion (cartesian) indices, $`𝐤_\pi `$ and $`𝐤_\pi ^{}`$ are the momenta of the incoming and outgoing (off-shell) pions, and $`a`$, $`b`$, $`c`$, and $`d`$ are constants defined, e.g., in Ref. . Evidently the off-shell extrapolation is provided by the form factor function $`F_{\pi NN}`$ . In the standard version of the TM force this form factors are assumed to be of monopole type,
$$F_{\pi NN}(𝐤_\pi ^2)=\frac{\mathrm{\Lambda }_\pi ^2m_\pi ^2}{\mathrm{\Lambda }_\pi ^2+𝐤_\pi ^2},$$
(18)
with a cutoff mass $`\mathrm{\Lambda }_\pi =5.8m_\pi 800`$ MeV. We would like to emphasize at this point that, in principle, there is no connection between the $`\pi NN`$ vertex (with an off-shell pion) and the $`\pi N`$ amplitude entering into the TBF. It would appear only in the contribution of the direct nucleon pole diagram to the $`\pi N`$ amplitude, which, however, is omitted in order to avoid double counting (cf. the discussion in section II). Therefore the prescription for the off-shell extrapolation used in the TM force must be considered as rather arbitrary.
In the discussion above we have tacitly ignored a conceptional subtlety when we talk about ”off-shell”. In case of the Jülich model the $`\pi N`$ amplitude is obtained off-energy-shell, while an off-mass-shell $`\pi N`$ amplitude is used in the TM potential. This means, that we can not make a simple comparison between them. The off-pion-mass-shell $`\pi N`$ amplitude of the TM force depends on the pion momenta $`𝐤_\pi `$ and $`𝐤_\pi ^{}`$. The off-energy-shell value of the $`\pi N`$ amplitude for the potential model (at a certain energy) is given as a function of the relative momentum between the pion and the nucleon, $`𝐐`$. Within non-relativistic kinematics $`𝐐`$ is obtained by (cf. Eq. (10))
$$𝐐=\frac{𝐤_\pi \frac{m_\pi }{m_N}𝐤_N}{1+\frac{m_\pi }{m_N}},$$
(19)
where $`𝐤_𝐍`$ is the nucleon momentum. One can see from this relation that $`𝐐`$ becomes equivalent to $`𝐤_\pi `$ only in the limit of $`m_\pi /m_N0`$.
In the following we want to discuss the off-shell properties entering into the calculations with the TM force and into the results presented in this paper. In view of the aforementioned difficulties it should be clear that any comparision can be only of qualitative nature. None the less, as we will see below such an analysis is useful and we believe that it indicates the source of the large discrepancy found in the contributions to the binding energy from the two approaches.
For this purpose let us introduce a half-off-shell function in the following way,
$$f_\alpha (Q)=\frac{T_\alpha (Q,Q^{};Z)}{Q^l}$$
(20)
where $`T_\alpha (Q,Q^{};Z)`$ is the off-shell $`\pi N`$ t-matrix (projected on the partial wave $`\alpha `$ with angular momentum $`l`$) at a fixed energy $`Z`$ and a fixed momentum $`Q^{}`$. The factor $`Q^l`$ is taken out for convenience because then we can normalize these half-off-shell functions to $`1`$ at $`Q=0`$ for s- as well p-waves and we can easily compare them with each other. Corresponding results for the Jülich $`\pi N`$ potential 2’ are shown in Fig. 8, where $`Z`$ and $`Q^{}`$ in Eq. (20) have been fixed to 930 MeV and 130 MeV/c, respectively.
Let us first take a look at the p-waves and in particular at the $`P_{11}`$ and $`P_{33}`$ partial waves which provide the main attractive contributions to the TBF (cf. Table I). In this case the momentum dependence of the TM $`\pi N`$ amplitudes is roughly given by a monopole type function $`F(𝐐^2)=\mathrm{\Lambda }^2/(\mathrm{\Lambda }^2+𝐐^2)`$ with $`\mathrm{\Lambda }=800`$ MeV, cf. Eqs. (17-18), which is shown by the dashed curve in Fig. 8. We observe that the corresponding half-off-shell functions of the Jülich $`\pi N`$ model fall off much faster with increasing (off-shell) momentum than this function. Accordingly we expect that a TBF based on the potential model will yield a much smaller attractive contribution to the three-body binding than one with off-shell properties similar to the monopole type function.
In case of the s-waves we get large repulsive contributions from the $`S_{31}`$ partial waves and small attractive contributions from $`S_{11}`$ (cf. Table I). Please note that also the corresponding half-off-shell functions are radically different. The one for $`S_{31}`$ exhibits a strong enhancement whereas the one for $`S_{11}`$ falls off very strongly with increasing (off-shell) momentum. The s-wave part of the TM force ($`a`$\- and $`c`$-terms), on the other hand, has no isospin dependence, cf. Eq. (17). This means that here the $`S_{31}`$ and $`S_{11}`$ partial waves have exactly the same momentum dependence. Therefore we suspect that a strong cancellation between the contributions from those two s-waves takes place. Indeed, actual triton calculations employing the TM force confirm that the total s-wave contributions are comparably small . Evidently, such a cancellation does not occur with the TBF based on the Jülich $`\pi N`$ model because of the differences in the off-shell properties. As a consequence, the (large) repulsive contribution of the $`S_{31}`$ partial wave to the three-body binding survives (cf. Table I).
Summarizing this phenomenological discussion of the off-shell properties we expect that a calculation based on the off-shell extrapolation used in the TM force should lead to a strong enhancement of the attractive contributions and at the same time reduce the repulsive contributions. We believe that this is the basic mechanism which makes the binding energy obtained with the TM TBF so large. We would like to substantiate this claim quantitatively with a model calculation. We can do this by substituting the off-shell properties of our $`\pi N`$ model by the ones used in the TM force. This can be easily done for the separable representations that we are using. We only need to replace the Jülich off-shell behavior by defining the form factor of the separable potential, $`g_\alpha (Q):=Q|V_\alpha |\psi _{E_1}`$, as follows:
$$g_\alpha (Q)\left(\underset{Q0}{lim}\frac{g_\alpha (Q)}{Q^l}\right)\frac{Q^l}{1+Q^2/\mathrm{\Lambda }^2}.$$
(21)
Furthermore we fix the energy dependence in the $`\pi N`$ amplitude and the $`\pi NN`$ vertex function again.
We demonstrate the effect on the binding energy for several values of $`\mathrm{\Lambda }`$ in Table IV. From this Table it is clear that we can get a substantial increase in the triton binding energy by choosing $`\mathrm{\Lambda }600800`$ MeV. Specifically one can see that the repulsion of the $`S_{31}`$ partial wave is suppressed and, moreover, cancels to a large extent with the $`S_{11}`$. At the same time the attraction provided by the $`P_{11}`$ and $`P_{33}`$ partial waves is strongly enhanced - as we expected from analyzing Fig. 8. In fact, if we take into account that our results are based on first-order perturbation theory and therefore may underestimate the correct values by a factor two or even three then our simulation with the choice $`\mathrm{\Lambda }=800`$ MeV practically reproduces the TM result, which is likewise based on $`\mathrm{\Lambda }_\pi =800`$ MeV.
## VI Summary
In the present paper we have re-investigated the origin of the large discrepancy in the contribution of a $`\pi \pi `$-exchange TBF found by Saito and Afnan to the commenly accepted values obtained with the Tucson-Melbourne or Brazil TBF. Unlike SA, who employed phenomenological separable potentials, we started out from a meson-theoretical $`\pi N`$ interaction model developed recently by the Jülich group. This model provides a good description of elastic $`\pi N`$ scattering data. It is also in agreement with empirical information on the $`\pi N`$ amplitude in the subthreshold region. In particular, it predicts the $`\pi N`$ $`\mathrm{\Sigma }`$ term close to the empirical value. Furthermore, the decrease in the $`\pi NN`$ form factors $`F_{\pi NN}(q^2)`$ from $`q^2=m_\pi ^2`$ to $`q^2=0`$ of about 4-7% and is comparable to that of the TM potential. Thus the form factors are much harder than those used by SA (which show a decrease of up to 20 %). Accordingly, the Jülich $`\pi N`$ model does not show the deficiencies which, so far, have been thought to be the main reason for the small contribution of the $`\pi \pi `$-exchange TBF in SA’s work.
None the less it turned out that also the $`\pi \pi `$-exchange TBF based on the $`\pi N`$ amplitude of the Jülich model is very small. The contributions to the triton binding energy are in the order of a few $`keV`$, which means comparable to the results obtained by Saito and Afnan.
A detailed analysis of the main differences between the TM TBF and that derived from the Jülich $`\pi N`$ potential model suggests that the differences in the contribution of the TBF to the 3N binding energy is due to the off-shell behaviour of the non-pole $`\pi N`$ amplitude. In the Jülich model the off-shell properties of the $`\pi N`$ amplitude are determined by the dynamical ingredients of the model and the fit to the $`\pi N`$ data. As a result, the off-shell properties of the amplitude are different in the individual $`\pi N`$ partial waves. In particular, the $`\pi N`$ amplitudes that provide attractive contributions to the three-body binding ($`P_{11}`$, $`P_{33}`$) fall off relatively fast with increasing off-shell momentum while the repulsive $`S_{31}`$ partial wave is enhanced. As a consequence, the total contribution of the TBF is very small as a result of the cancellation effects. On the other hand, in the $`\pi N`$ amplitude underlying the TM TBF, the off-shell extrapolation is done in terms of a monopole form factor with a cut-off mass of 800 MeV for all partial waves. Since this monopole form factor falls off much slower then the $`P_{11}`$ and $`P_{33}`$ amplitudes of the Jülich $`\pi N`$ model, the corresponding attraction provided by the TM force is considerably enhanced. At the same time, the repulsion in the $`S_{31}`$ partial wave (and therefore any cancellation effects) is strongly suppressed. These combined effects do indeed explain the two-orders-of-magnitude discrepancy in the resulting contribution to the triton binding energy as we have demonstrated in a numerical model study.
Naturally the question arises how realistic and well-defined the off-shell properties of the Jülich $`\pi N`$ model are, especially in view of the so-called quasi-potential ambiguity . This is a topic which needs to be further investigated in the future, but it is certainly beyond the scope of the present paper. Here we only want to point to the fact that rather different ansatzes for the $`\pi N`$ interaction (meson-exchange and simple separable forms, respectively) lead to qualitatively similar off-shell features and, in consequence, to similar results for the TBF. Qualitatively similar off-shell properties seem to be also predicted by other $`\pi N`$ models - at least as far as we can judge from corresponding publications .
With regard to the off-shell extrapolation used in the $`\pi N`$ amplitude on which the TM TBF is based the situation is, in our opinion, much less clear. First of all, we do not see any stringent physical reason for adopting the $`\pi NN`$ form factor for this purpose, especially because no $`\pi NN`$ vertex is present at all in the non-nucleon-pole part of the $`\pi N`$ amplitude that enters into the derivation of the TBF. Furthermore, the choice of having the same off-shell properties in all $`\pi N`$ partial waves is also hard to justify. Therefore the large contribution of the TM (and also the Brazil) TBF to the triton binding energy of around 1 MeV - though certainly desired by phenomenology - must be interpreted with caution.
Acknowledgements
We would like to thank Prof. I.R. Afnan for valuable discussions and for a careful reading of the manuscript. We acknowledge the hospitality of the RCNP in Osaka, Japan, where most of the numerical calculations were carried out. This work was financially supported by the Deutsche Forschungsgemeinschaft (Grant no. 447 AUS-113/3/0) and by the Japanese Society for the Promotion of Science.
Appendix: Separable expansion of a potential with two terms
We consider to represent a potential $`V`$ which consists of two terms, $`V=V_1+V_2`$, in separable form. At first we expand $`V_2`$ by means of the standard EST-method, cf. Eqs. (2) to (6). Then the t-matrix $`\stackrel{~}{T}_2(E)`$ obtained from the separable representation $`\stackrel{~}{V}_2`$ agrees (on- as well as half-off-shell) with $`T_2(E)`$ corresponding to $`V_2`$ at the choosen expansion energies $`E=E_i`$, $`i=1,\mathrm{},N`$.
Next we expand the potential $`V`$ by assuming the following form
$$\stackrel{~}{V}(E)=|\stackrel{~}{v}_0\lambda _1(E)\stackrel{~}{v}_0|+\stackrel{~}{V}_2.$$
(22)
We determine the ”form factor” $`|\stackrel{~}{v}_0`$ in such a way that $`\stackrel{~}{V}`$ satisfies
$$V|\psi _\epsilon =\stackrel{~}{V}(\epsilon )|\psi _\epsilon ,$$
(23)
where $`|\psi _\epsilon `$ is a solution of the scattering equation,
$$|\psi _\epsilon =|k_\epsilon +G_0(\epsilon )V|\psi _\epsilon ,$$
(24)
at a fixed predetermined energy $`\epsilon `$. We want to emphasize that we may choose the energy $`\epsilon `$ at which the potential $`V`$ is expanded to be different from any of the energies $`E_i`$ chosen for the separable expansion of $`V_2`$ . We see easily that Eq. (23) is satisfied if we choose $`|\stackrel{~}{v}_0`$ to be
$$|\stackrel{~}{v}_0=(V\stackrel{~}{V}_2)|\psi _\epsilon ,$$
(25)
and $`\lambda _1(\epsilon )`$ to be
$`\lambda _1(\epsilon )`$ $`=`$ $`{\displaystyle \frac{1}{\psi _\epsilon |V\stackrel{~}{V}_2|\psi _\epsilon }}`$ (26)
$`=`$ $`{\displaystyle \frac{1}{\stackrel{~}{v}_0|\psi _\epsilon }}`$ (27)
Note that the (half-off-shell) t-matrix $`\stackrel{~}{T}(E)`$ obtained from $`\stackrel{~}{V}(\epsilon )`$ is identical to $`T(E)`$ obtained from $`V`$ at $`E=\epsilon `$ because of condition (23).
If $`T(E)`$ has a pole at $`E_p`$ we may choose the expansion energy $`\epsilon `$ for the separable representation to be $`\epsilon =E_p`$. The wave function at the pole, $`|\psi _p`$, is a solution of the equation
$$|\psi _p=G_0(E_p)V|\psi _p.$$
(28)
The t-matrix $`\stackrel{~}{T}(E)`$ for the separable potential $`\stackrel{~}{V}(E)`$ is given by
$$\stackrel{~}{T}(E)=|\stackrel{~}{v}(E)\frac{1}{1/\lambda _1(E)\stackrel{~}{\mathrm{\Sigma }}(E)}\stackrel{~}{v}(E)|+\stackrel{~}{T}_2(E),$$
(29)
where
$$\stackrel{~}{\mathrm{\Sigma }}(E)=\stackrel{~}{v}_0|G_0(E)|\stackrel{~}{v}(E),$$
(30)
and
$$|\stackrel{~}{v}(E)=(1+\stackrel{~}{T}_2(E)G_0(E))|\stackrel{~}{v}_0.$$
(31)
The ”form factor” $`|\stackrel{~}{v}_0`$ is defined by
$$|\stackrel{~}{v}_0=(V\stackrel{~}{V}_2)|\psi _p.$$
(32)
Substituting Eqs. (31) and (32) into Eq. (30) leads to
$$\stackrel{~}{\mathrm{\Sigma }}(E)=\stackrel{~}{v}_0|(1+G_0(E)\stackrel{~}{T}_2(E))G_0(E)VG_0(E)\stackrel{~}{T}_2(E)|\psi _p.$$
(33)
We can evaluate $`\stackrel{~}{\mathrm{\Sigma }}(E)`$ at $`E=E_p`$ by using Eq. (28),
$`\stackrel{~}{\mathrm{\Sigma }}(E_p)`$ $`=`$ $`\stackrel{~}{f}_0|\psi _p`$ (34)
$`=`$ $`{\displaystyle \frac{1}{\lambda _1(E_p)}},`$ (35)
where we have utilized, in addition, Eq. (27). By substituting this result into Eq. (29), we see that the t-matrix $`\stackrel{~}{T}(E)`$ obtained for the separable potential $`\stackrel{~}{V}`$ has the same pole position as the t-matrix of the original potentil $`V`$.
Let us now assume that the first term of the original potential $`V`$ has a separable form,
$$V_1(E)=|v_0\mathrm{\Lambda }(E)v_0|,$$
(36)
where $`\mathrm{\Lambda }(E)`$ is
$$\mathrm{\Lambda }(E)=\frac{1}{Em_0}.$$
(37)
This is exactly the case for the (direct) nucleon pole contribution, Fig. 2(a). Then $`|v_0`$ corresponds to the bare $`\pi NN`$ vertex function and $`m_0`$ is the bare nucleon mass. The solution of the Lippmann-Schwinger equation for $`V=V_1+V_2`$ can then be written as
$$T(E)=|v(E)\frac{1}{1/\mathrm{\Lambda }(E)\mathrm{\Sigma }(E)}v(E)|+T_2(E),$$
(38)
where $`T_2(E)`$ is a solution of the scattering equation for the potential $`V_2`$. The self-energy $`\mathrm{\Sigma }(E)`$ is given by
$$\mathrm{\Sigma }(E)=v_0|G_0(E)|v(E),$$
(39)
and the dressed $`\pi NN`$ vertex function, $`|v(E)`$, by
$$|v(E)=(1+T_2(E)G_0(E))|v_0.$$
(40)
Assuming that the t-matrix of Eq. (38) has a pole at $`E=m_N`$, $`\mathrm{\Sigma }(m_N)`$ is evaluated as
$`\mathrm{\Sigma }(m_N)`$ $`=`$ $`{\displaystyle \frac{1}{\mathrm{\Lambda }(m_N)}}`$ (41)
$`=`$ $`m_Nm_0.`$ (42)
Also, the pole part of the t-matrix can be written in the form,
$$T(E)=|v(E)\frac{1}{1/\mathrm{\Lambda }(E)\mathrm{\Sigma }(E)}v(E)|=|v^R(E)\frac{1}{Em_N}v^R(E)|,$$
(43)
which defines the renormalized $`\pi NN`$ vertex function $`|v^R(E)`$. At the pole $`|v^R(E)`$ is given by (cf., e.g., Ref. )
$$|v^R(m_N)=\frac{|v(m_N)}{(1\mathrm{\Sigma }^{}(m_N))^{1/2}},$$
(44)
where
$$\mathrm{\Sigma }^{}(m_N)=v(m_N)|G_0^2(m_N)|v(m_N).$$
(45)
In the last step we have assumed that $`V_2`$ does not dependent on the energy $`E`$. If it does (like in our case) then $`\mathrm{\Sigma }^{}`$ is determined by
$$\mathrm{\Sigma }^{}(m_N)=\frac{}{E}v_0|G_0(E)|v(E)|_{E=m_N}.$$
(46)
Furthermore, we note that the renormalized vertex function is related to the $`\pi NN`$ coupling constant $`f_{\pi NN}`$ and the $`\pi NN`$ form factor $`F_{\pi NN}`$ by
$$f_{\pi NN}F_{\pi NN}(𝐪^2)a(q)=q|v^R(m_N).$$
(47)
where $`a(q)`$ is a kinematical factor depending on the particular ($`ps`$ or $`pv`$) $`\pi NN`$ coupling (cf. Eqs. (3) and (16) of Ref. ).
The wave function at the pole energy is obtained by solving Eq. (28)
$$|\psi _p=G_0(m_N)|v(m_N).$$
(48)
By substituing Eq. (48) into the $`lhs`$ of Eq. (28), we get another relation between $`|v(m_N)`$ and $`|\psi _p`$,
$$|v(m_N)=V(m_N)|\psi _p.$$
(49)
Now we expand the potential model $`V(E)`$ using the method described above. Then the resulting t-matrix obtained from the separable representation has the same pole position as the original interaction model. We will examine whether also the $`\pi NN`$ coupling constant and form factor determined from the separable representation agree with the ones of the original interaction.
At first, we derive a relation between $`|\stackrel{~}{v}(m_N)`$ and $`|v(m_N)`$. Using Eqs. (31), (32), and finally (49) we obtain
$`|\stackrel{~}{v}(m_N)`$ $`=`$ $`(1+\stackrel{~}{T}_2(m_N)G_0(m_N))(V(m_N)\stackrel{~}{V}_2)|\psi _p`$ (50)
$`=`$ $`V(m_N)|\psi _p`$ (51)
$`=`$ $`|v(m_N).`$ (52)
Eq. (52) implies that the momentum dependence of the $`\pi NN`$ vertex function obtained from the separable representation is identical to the one of the original potential.
The renormalized $`\pi NN`$ vertex function at the pole, $`|\stackrel{~}{v}^R(m_N)`$, for the separable representation $`\stackrel{~}{V}`$ can be calculated in the same way as for the original potential. It is obtained by placing tildes over all quantities in Eq. (44) and (45):
$`|\stackrel{~}{v}^R(m_N)`$ $`=`$ $`{\displaystyle \frac{|\stackrel{~}{v}(m_N)}{(1\stackrel{~}{\mathrm{\Sigma }}^{}(m_N))^{1/2}}},`$ (53)
$`\stackrel{~}{\mathrm{\Sigma }}^{}(m_N)`$ $`=`$ $`\stackrel{~}{v}(m_N)|G_0^2(m_N)|\stackrel{~}{v}(m_N).`$ (54)
Because of Eq. (52) we get
$$\stackrel{~}{\mathrm{\Sigma }}^{}(m_N)=\mathrm{\Sigma }^{}(m_N).$$
(55)
Finally, by substituting Eqs. (52) and (55) into Eq. (53) it follows that
$$|\stackrel{~}{v}^R(m_N)=|v^R(m_N),$$
(56)
and consequently that $`\stackrel{~}{f}_{\pi NN}=f_{\pi NN}`$ and $`\stackrel{~}{F}_{\pi NN}(𝐪^2)=F_{\pi NN}(𝐪^2)`$.
In the above discussion we have not specified the form of $`\lambda _1(E)`$, since we needed only the value of $`\lambda _1(E)`$ at the energy $`E=m_N`$ to get the correct half-off shell t-matrix. If we assume $`\lambda _1(E)`$ to be of the form $`1/\lambda _1(E)=E\stackrel{~}{m}_0`$, then $`\stackrel{~}{m}_0`$ can be determined by Eq. (27) (or Eq. (35)) to be
$$\stackrel{~}{m}_0=m_N\psi _p|V\stackrel{~}{V}_2|\psi _p.$$
(57)
Therefore, by following the outlined procedure it is possible to construct a separable representation where (i) the half-off-shell behavior of the non-pole part of the t-matrix (at selected energies), (ii) the pole position, (iii) the $`\pi NN`$ form factor, and (iv) the value of the $`\pi NN`$ coupling constant are the same as in the original interaction model.
|
warning/0003/cond-mat0003187.html
|
ar5iv
|
text
|
# Quantum spin hydrodynamics and a new spin-current mode in ferromagnetic metals
\[
## Abstract
We derive the quantum spin hydrodynamic equations in a ferromagnetic metal. From these equations we show the existence of a new massive spin-current mode. This mode can be observed in neutron scattering experiments and we discuss the difficulties in seeing it. At the end we discuss the existence of this mode in localized ferromagnets.
\] Propagating collective modes in condensed matter systems are often a consequence of spontaneously broken symmetry. In ferromagnetic systems, the spin rotational symmetry is spontaneously broken leading to the existence of spin waves. They were first observed in iron using neutron scattering experiments in the nineteen fifties. Theoretically spin waves were predicted in lattice models with a ferromagnetic ground state as well as in ferromagnetic Fermi liquids. In paramagnetic metals, spin waves propagate in an external magnetic field. Not all propagating modes are a consequence of spontaneously broken symmetry and an example is zero sound in liquid helium.
In this paper we will demonstrate the existence of a new collective spin mode in ferromagnetic metals. We start from a quasi-classical kinetic equation, the Landau kinetic equation, for the momentum distribution function in a ferromagnetic metal. From this we can derive the dynamic equations for the evolution of the spin density and spin current. In the low temperature limit the spin dynamics is collisionless and dominated by spin precession, and we will call this the quantum spin hydrodynamic (QSH) regime. At higher temperature this dynamics crosses over to the classical spin hydrodynamics (CSH) of Halperin and Hohenberg. In the QSH regime we show that a new massive mode exists along with the usual magnon. While gaped, it exists outside of the particle-hole, Stoner, continuum and therefore it can propagate. From a macroscopic point of view, this mode is induced by the collective oscillations of the spin current and it should be observable in neutron scattering experiments.
In the long wave length limit one can derive the existence of the spin waves from the conservation law for the magnetization in the CSH limit. However, the spin current is not a conserved quantity and therefore one cannot derive propagating modes related to the fluctuations of the spin current from the conservation laws. Here the Landau kinetic equation first used by Abrikosov and Dzyaloshinksii for weak ferromagnetic metals is used to derive the full QSH theory.
We consider a three dimensional, weak itinerant ferromagnet below its Curie temperature. To avoid complications we ignore the effects related to the lattice. The phenomena which we will describe is independent of the lattice, although lattice effects can add more features to the ones that we study. We assume that the itinerant ferromagnet is well described by ferromagnetic Fermi liquid theory. The kinetic equations for the magnetization and the spin current can be derived from the equation for the Fourier transform of the Wigner distribution function $`n_\stackrel{}{p}^{\alpha \beta }(\stackrel{}{r},t)`$
$$\frac{n_\stackrel{}{p}^{\alpha \beta }(\stackrel{}{r},t)}{t}\frac{1}{i\mathrm{}}F_\stackrel{}{p}^{\alpha \beta }[n;ϵ]=I_Q[n_\stackrel{}{p}^{}],$$
(1)
where
$`F_\stackrel{}{p}^{\alpha \beta }[n;ϵ]`$ $`=`$ $`{\displaystyle \underset{\gamma }{}}{\displaystyle Dp^{}Dqd^3r^{}d^3\rho e^{i\left[(\stackrel{}{p}\stackrel{}{p}^{})\stackrel{}{\rho }+\stackrel{}{q}(\stackrel{}{r}\stackrel{}{r}^{})\right]}}`$ (4)
$`\times (ϵ_{\stackrel{}{p}^{}+\mathrm{}\stackrel{}{q}/2}^{\alpha \gamma }(\stackrel{}{r}^{}+{\displaystyle \frac{\mathrm{}\stackrel{}{\rho }}{2}},t)n_\stackrel{}{p}^{}^{\gamma \beta }(\stackrel{}{r}^{},t)`$
$`n_\stackrel{}{p}^{}^{\alpha \gamma }(\stackrel{}{r}^{},t)ϵ_{\stackrel{}{p}^{}\mathrm{}\stackrel{}{q}/2}^{\gamma \beta }(\stackrel{}{r}^{}{\displaystyle \frac{\mathrm{}\stackrel{}{\rho }}{2}},t)),`$
$$n_\stackrel{}{p}^{\alpha \beta }(\stackrel{}{r},t)=Dqe^{i\stackrel{}{q}\stackrel{}{r}}a_{\stackrel{}{p}\stackrel{}{q}/2,\alpha }^{}a_{\stackrel{}{p}+\stackrel{}{q}/2,\beta },$$
(5)
with $`a_{\stackrel{}{p}+\stackrel{}{q}/2,\beta }`$ the single fermion operator for a particle with spin $`\beta `$. Here $`Dq=\frac{d^3q}{(2\pi \mathrm{})^3}`$ and $`ϵ_\stackrel{}{p}^{\alpha \beta }(\stackrel{}{r},t)`$ is the effective quasiparticle Hamiltonian. Also $`I_Q`$ is the quantum collision integral which contains contributions from two body collisions and from the rate of change due to the spin precession.
Following Ref. we expand all single body operators in the basis set spanned by the Pauli matrices
$$ϵ_\stackrel{}{p}^{\alpha \beta }(\stackrel{}{r},t)=ϵ_\stackrel{}{p}\delta ^{\alpha \beta }(\stackrel{}{r},t)+\stackrel{}{h}_\stackrel{}{p}(\stackrel{}{r},t)\stackrel{}{\tau }^{\alpha \beta },$$
(6)
$$n_\stackrel{}{p}^{\alpha \beta }(\stackrel{}{r},t)=n_\stackrel{}{p}\delta ^{\alpha \beta }(\stackrel{}{r},t)+\stackrel{}{\sigma }_\stackrel{}{p}(\stackrel{}{r},t)\stackrel{}{\tau }^{\alpha \beta }$$
(7)
at every point $`(\stackrel{}{r},t)`$ of space and time with $`\stackrel{}{\sigma }_\stackrel{}{p}`$ the magnetization and
$$\stackrel{}{h}_\stackrel{}{p}=\gamma \frac{\mathrm{}}{2}\stackrel{}{}+2Dp^{}f_{\stackrel{}{p}\stackrel{}{p}^{}}^a\stackrel{}{\sigma }_p^{}$$
(8)
the effective magnetic field which is a sum of an external magnetic field $`\stackrel{}{}`$ and an internal magnetic field generated by the quasiparticle interactions $`f_{\stackrel{}{p}\stackrel{}{p}^{}}^a`$. We expand the quasiparticle interactions in spherical harmonics and separate the spin symmetric ($`f_l^s`$) and spin antisymmetric ($`f_l^a`$) parts. The result is
$$N(0)f_{\stackrel{}{p}\stackrel{}{p}^{}}^{\tau \tau ^{}}=\underset{l}{}(F_l^s+F_l^a\stackrel{}{\tau }\stackrel{}{\tau }^{})P_l(\widehat{p}\widehat{p}^{}),$$
(9)
with
$$f_{\stackrel{}{p}\stackrel{}{p}^{}}^{\tau \tau ^{}}f_{\stackrel{}{p}\stackrel{}{p}^{}}^s+f_{\stackrel{}{p}\stackrel{}{p}^{}}^a\stackrel{}{\tau }\stackrel{}{\tau }^{}.$$
(10)
Here we treat the quasiparticle interaction as spin rotation invariant. This is possible in the case of small magnetizations and external magnetic fields as explained in. Here $`N(0)`$ is the average density of states, averaged over the two Fermi surfaces. We must note that the condition that the ground state of the metal is ferromagnetic means that $`N(0)f_0^a=F_0^a<1`$.
For an inhomogeneous system the spin quantization axis is position dependent and therefore the spin density
$$\stackrel{}{\sigma }(\stackrel{}{r},t)=2Dp\stackrel{}{\sigma }_\stackrel{}{p}(\stackrel{}{r},t)$$
(11)
is position dependent. From the kinetic equation one can obtain the continuity equation for the spin density which reduces to the usual spin conservation law
$$\frac{\stackrel{}{\sigma }(\stackrel{}{r},t)}{t}+\frac{}{x_i}j_{\stackrel{}{\sigma },i}(\stackrel{}{r},t)=\gamma \stackrel{}{\sigma }(\stackrel{}{r},t)\times \stackrel{}{},$$
(12)
where the spin current is
$$j_{\stackrel{}{\sigma },i}(\stackrel{}{r},t)=2Dp\left[\frac{ϵ_\stackrel{}{p}}{p_i}\stackrel{}{\sigma }_\stackrel{}{p}(\stackrel{}{r},t)+\frac{\stackrel{}{h}_\stackrel{}{p}}{p_i}n_\stackrel{}{p}(\stackrel{}{r},t)\right]$$
(13)
and a summation over repeated indices is used. The precession term in the $`r.h.s.`$ of Eq.(10) follows from the non-commutativity of the particle density operator and the quasiparticle Hamiltonian. Assuming that the magnetization is small and that we are close to equilibrium we can linearize the spin current which leads to
$$j_{\stackrel{}{\sigma },i}(\stackrel{}{r},t)=2Dpv_{\stackrel{}{p}_i}\stackrel{}{\sigma }_\stackrel{}{p}\left(1+\frac{F_1^a}{3}\right),$$
(14)
because in the Fermi liquid only momenta in the vicinity of the Fermi surface contributes to the magnetization. Here $`\stackrel{}{v}_p=\stackrel{}{}_pϵ_p^0`$, $`n_p^0`$ is the equilibrium distribution function, and $`F_1^a`$ is a Fermi liquid parameter.
Starting from the kinetic equation for the spin density distribution function
$`{\displaystyle \frac{\stackrel{}{\sigma }_\stackrel{}{p}}{t}}+v_{\stackrel{}{p}_i}{\displaystyle \frac{}{r_i}}\left(\stackrel{}{\sigma }_\stackrel{}{p}{\displaystyle \frac{n_p^0}{ϵ_p}}\stackrel{}{h}_\stackrel{}{p}\right)`$ $`=`$ (15)
$`{\displaystyle \frac{2}{\mathrm{}}}\stackrel{}{\sigma }_\stackrel{}{p}\times \stackrel{}{h}_\stackrel{}{p}+\left({\displaystyle \frac{\stackrel{}{\sigma }_\stackrel{}{p}}{t}}\right)_{coll.}`$ (16)
then multiplying by the quasiparticle velocity and integrating over the quasiparticle momentum the linearized equation for the spin current takes the form
$$\frac{}{t}j_{\stackrel{}{\sigma },i}(\stackrel{}{r},t)+\frac{}{x_m}\mathrm{\Pi }_{\stackrel{}{\sigma }im}=\left[\frac{}{t}j_{\stackrel{}{\sigma },i}\right]_{prec.}+\left[\frac{}{t}j_{\stackrel{}{\sigma },i}\right]_{coll.},$$
(17)
where the spin stress tensor is
$$\mathrm{\Pi }_{\stackrel{}{\sigma }im}=2\left(1+\frac{F_1^a}{3}\right)Dpv_{p_i}v_{p_m}\left(\stackrel{}{\sigma }_\stackrel{}{p}\frac{n_\stackrel{}{p}^0}{ϵ_p}\stackrel{}{h}_\stackrel{}{p}\right).$$
(18)
Next we apply a small, transverse (to the external field and therefore to the magnetization) magnetic field $`\delta \stackrel{}{}=\delta _x\widehat{e}_x+\delta _y\widehat{e}_y`$ which induces a change in the magnetization $`\delta \stackrel{}{\sigma }_\stackrel{}{p}=\delta \sigma _{\stackrel{}{p}x}\widehat{e}_x+\delta \sigma _{\stackrel{}{p}y}\widehat{e}_y`$, where $`\widehat{e}_i`$ are the unit vectors and $`\delta \stackrel{}{\sigma }_\stackrel{}{p}=\stackrel{}{\sigma }_\stackrel{}{p}\stackrel{}{\sigma }_\stackrel{}{p}^0`$. To make the expansion in spherical harmonics more transparent we define
$$\delta \sigma _\stackrel{}{p}^\pm =\frac{n_\stackrel{}{p}^0}{ϵ_p^0}\nu _\stackrel{}{p}^\pm $$
(19)
and
$$\sigma _\stackrel{}{p}^0=\frac{n_\stackrel{}{p}^0}{ϵ_p^0}\frac{m}{N(0)},$$
(20)
where $`\delta \sigma _\stackrel{}{p}^\pm =\delta \sigma _{\stackrel{}{p}x}\pm i\delta \sigma _{\stackrel{}{p}y}`$, $`\sigma _\stackrel{}{p}^0`$ is the magnitude of the magnetization in the absence of the perturbing field, and $`m`$ is the equilibrium magnetization. We expand in spherical harmonics the quantities
$$\nu _\stackrel{}{p}^\pm =\underset{l}{}\nu _l^\pm P_l(\widehat{p}\widehat{k})$$
(21)
and substitute the corresponding expressions in Eq.(13). Here $`\widehat{k}`$ is the unit vector in the direction of the spin quantization axis. Now Eq.(13) involves terms grouped according to the value of the integer number $`l`$. Next we neglect the terms with $`l>1`$. Although the term $`l=2`$ is very important as we discuss later, the derivation which we present below is clearer and contains all the steps needed for the derivation of the dispersion including higher spherical harmonics. Then using the expansion in spherical harmonics, Eq.(18), and in the presence of an external magnetic field, Eq.(14) reduces to
$`{\displaystyle \frac{}{t}}\stackrel{}{j}_{\stackrel{}{\sigma },i}+\left(1+{\displaystyle \frac{F_1^a}{3}}\right)\left|1+F_0^a\right|{\displaystyle \frac{v_F^2}{3}}{\displaystyle \frac{}{x_i}}\stackrel{}{\sigma }(\stackrel{}{r},t)`$ (22)
$`=`$ $`\gamma \stackrel{}{j}_{\stackrel{}{\sigma },i}\times {\displaystyle \frac{2}{\mathrm{}}}\left(f_0^a{\displaystyle \frac{f_1^a}{3}}\right)\stackrel{}{j}_{\stackrel{}{\sigma },i}\times \stackrel{}{\sigma }(\stackrel{}{r},t)\left(1+{\displaystyle \frac{F_1^a}{3}}\right){\displaystyle \frac{\stackrel{}{j}_{\stackrel{}{\sigma },i}}{\tau _D}},`$ (23)
Here we have used a relaxation time approximation for the collision contribution to the spin current.
Before we examine the full spin dynamics coming from Eqs.(10) and (19) it is useful to contrast this with the behavior of a paramagnetic metal. In a paramagnetic Fermi liquid the first two terms on the $`r.h.s.`$ of Eq.(19) are proportional to $`\omega _L=2\gamma `$, where $``$ is the external magnetic field. At high temperature the CSH regime is realized and at low temperature the system crosses over to the precession dominated regime. The cross over scale is set by $`\omega _L\tau _D1`$. This behavior is similar to that in paramagnetic metals. In the ferromagnetic metal the cross over scale is nearly independent of the external magnetic field. To see this we first note that at high temperature, Eq.(19) in the uniform case reduces to
$$\frac{}{t}\stackrel{}{j}_{\stackrel{}{\sigma },i}=\left(1+\frac{F_1^a}{3}\right)\frac{\stackrel{}{j}_{\stackrel{}{\sigma },i}}{\tau _D}$$
(25)
which means that $`\stackrel{}{j}_{\stackrel{}{\sigma },i}`$ exponentially decays and its explicit time dependence can be ignored. In this limit, which is called the classical spin hydrodynamic limit with $`\frac{\stackrel{}{j}}{t}=0`$ the equations reduce to those used by Halperin and Hohenberg and it is easy to derive the dispersion, $`\omega q^2`$ for the spin waves. The quadratic dependence for the Goldstone mode follows from the fact that the order parameter $`m`$ is a conserved charge of the spin rotation algebra. As the system crosses over to low temperature it enters into the QSH regime in which the spin current changes in time due to the precession around the effective field with $`\omega _1^+=2m(f_1^a/3f_0^a)`$. For zero external field the cross over from QSH to CSH occurs when $`\omega _1^+\tau _D1`$. The range of temperatures where $`\omega _1^+\tau _D>>1`$ is what we have referred to as the QSH regime. In this QSH regime we can not ignore the $`\frac{}{t}\stackrel{}{j}_{\stackrel{}{\sigma },i}`$ term in Eq.(19). This as we will see is the source of the new spin-current mode.
There is another way to see the difference between spin waves in a ferromagnetic and a paramagnetic metal. Consider a ferromagnetic metal in an external magnetic field. The magnetization of the metal will align with the external field. If we apply a small pulse in a direction not collinear with the external magnetic field, the magnetization will start to precess around the external field. Next we reduce to zero the external magnetic field. In a paramagnetic metal the polarization will also go to zero and the spin wave will become damped. However, in a ferromagnetic metal the magnetization will continue to precess about the same axis, although no external magnetic field is present to define that axis. Now one can imagine looking at the system from a frame of reference rotating with the magnetization. In this frame of reference there will be no magnetization precession, but there will be a precession of the spin current, because the spin current is not a conserved quantity and as seen from Eqn.(19) it precesses with the frequency $`\omega _1^+`$.
It is instructive to look first at the modes of the system in the simpler case when $`F_l^a=0`$, for $`l>1`$ and the external field is $`=0`$. In that case the equations simplify to
$$\frac{\delta \stackrel{}{\sigma }(\stackrel{}{r},t)}{t}+\frac{}{x_i}j_{\stackrel{}{\sigma },i}(\stackrel{}{r},t)=0,$$
(26)
$`{\displaystyle \frac{}{t}}\stackrel{}{j}_{\stackrel{}{\sigma },i}+c_s^2{\displaystyle \frac{}{x_i}}\delta \stackrel{}{\sigma }`$ $`={\displaystyle \frac{2}{\mathrm{}}}(f_0^a{\displaystyle \frac{f_1^a}{3}})\stackrel{}{j}_{\stackrel{}{\sigma },i}\times \left(\stackrel{}{\sigma }^0+\delta \stackrel{}{\sigma }\right)`$ (28)
$`(1+{\displaystyle \frac{F_1^a}{3}}){\displaystyle \frac{\stackrel{}{j}_{\stackrel{}{\sigma },i}}{\tau _D}},`$
where $`c_s^2=\left|1+F_0^a\right|(1+\frac{F_1^a}{3})\frac{v_F^2}{3}`$ is the spin wave velocity, with $`v_F`$ the Fermi velocity. Here we have introduced $`\delta \stackrel{}{\sigma }=\stackrel{}{\sigma }(\stackrel{}{r},t)\stackrel{}{\sigma }^0`$, with $`\stackrel{}{\sigma }^0=m\widehat{k}`$, where $`m`$ is the magnetization. If the system is in the QSH regime and $`\delta \stackrel{}{\sigma }\stackrel{}{\sigma }^0`$ one can write for the spin current the following hydrodynamic equation
$$\frac{}{t}\stackrel{}{j}_{\stackrel{}{\sigma },i}+c_s^2\frac{}{x_i}\delta \stackrel{}{\sigma }=\frac{2}{\mathrm{}}(f_0^a\frac{f_1^a}{3})\stackrel{}{j}_{\stackrel{}{\sigma },i}\times \stackrel{}{\sigma }^0.$$
(29)
Differentiating this equation with respect to $`x_i`$ and using the equation for the spin density we obtain
$$\frac{^2}{t^2}\delta \stackrel{}{\sigma }+c_s^2\frac{^2}{x_i^2}\delta \stackrel{}{\sigma }+\frac{2}{\mathrm{}}(f_0^a\frac{f_1^a}{3})\frac{\delta \stackrel{}{\sigma }}{t}\times \stackrel{}{\sigma }^0=0.$$
(30)
Below we will set $`\mathrm{}=1`$. Applying again a transverse magnetic field as above Eq.(16) we can write
$$\frac{^2\delta \sigma ^+}{t^2}+c_s^2\frac{^2\delta \sigma ^+}{x_i^2}2im(f_0^a\frac{f_1^a}{3})\frac{\delta \sigma ^+}{t}=0.$$
(31)
Assuming that the solution is of the form
$$\delta \sigma ^+=m^+e^{i(\stackrel{}{q}\stackrel{}{r}\omega t)}$$
(32)
with $`\stackrel{}{q}=q\widehat{e}_x`$ we obtain the dispersion equation
$$\omega ^2+c_s^2q^2+2m(f_0^a\frac{f_1^a}{3})\omega =0.$$
(33)
The two modes have the dispersion
$$\omega _l=\left|m(f_0^a\frac{f_1^a}{3})\right|\left(1\pm \sqrt{1\frac{c_s^2q^2}{\left|m(f_0^a\frac{f_1^a}{3})\right|^2}}\right).$$
(34)
where the $`l=0(+)`$ solution corresponds to the spin waves and the $`l=1()`$ solution corresponds to a new collective spin mode. For small $`q`$ we have the usual Goldstone mode
$$\omega _0(q)=\frac{c_s^2}{\omega _1^+}q^2$$
(35)
and the new mode has the dispersion
$$\omega _1(q)=\omega _1^+\frac{c_s^2}{\omega _1^+}q^2.$$
(36)
Under close examination one finds that to the $`q^2`$ term there is a contribution from the $`l=2`$ spherical harmonics in Eq.(18). Taking that into account gives the full expression for this mode
$$\omega _1(q)=\omega _1^+\left[\frac{c_s^2}{\omega _1^+}+\frac{4N(0)v_F^2}{30m}\left(\frac{3}{F_1^a}+1\right)\right]q^2.$$
(37)
The above equation is valid for $`qv_F<<m|f_1^a|`$. In the limit $`f_1^a0`$ the mode merges with the Stoner continuum and is Landau damped. For $`f_1^a0`$ the mode propagates with a dispersion that depends on the sign of $`f_1^a`$.
It is useful to look at the contribution of this mode to the $`f`$-sum rule. It is known that the spin waves do not satisfy this sum rule. However the spin waves plus the new mode exhaust the $`f`$-sum rule and the spectral weight is shifted from the Stoner excitations to this new mode. To see this we first introduce the dynamic structure function
$$S^{\tau \tau ^{}}(\stackrel{}{q},\omega )=\frac{1}{\pi }Im\chi ^{\tau \tau ^{}}(\stackrel{}{q},\omega )\theta (\omega ),$$
(38)
where $`\chi (\stackrel{}{q},\omega )`$ is the dynamic spin susceptibility obtained from the kinetic equation by noting that $`\chi =\frac{\delta \sigma }{\delta }`$. Here $`\theta (\omega )`$ is the step function. From the expression for the dynamic structure function, calculated from the kinetic equation Eq.(13), we can derive the f-sum rule for the transverse dynamic spin response function, $`S^+(\stackrel{}{q},\omega )`$. This is given by
$$_0^{\mathrm{}}𝑑\omega \omega S^+(\stackrel{}{q},\omega )=(1+\frac{F_1^a}{3})\frac{nq^2}{2m^{}},$$
(39)
where $`n`$ is the total density and $`m^{}`$ is the effective mass. This has the same form as the one obtained for the spin dynamic structure function in the paramagnetic phase.
This sum rule is not exhausted by the Goldstone mode alone. If we now include the new mode at zero temperature we have
$$S^+(\stackrel{}{q},\omega )=\alpha _\stackrel{}{q}^+\delta (\omega \omega _0(q))+\beta _\stackrel{}{q}^+\delta (\omega \omega _1(q))$$
(40)
and this exhausts the $`f`$-sum rule, Eq.(33). This is significant, because the spectral weight of the Stoner excitations has been transfered to the two modes, leaving
$$_0^{\mathrm{}}𝑑\omega \omega S_{Stoner}^+(\stackrel{}{q},\omega )q^4.$$
(41)
This might be a plausible explanation for the difficulty of the direct observation of the Stoner excitations in a neutron scattering experiments in Fe and Ni since the oscillator strength of these single excitations has been reduced.
The observation of the new mode described in this paper is difficult because it has a small spectral weight at small $`q`$. This follows from the $`q^2`$ dependence of $`\beta _\stackrel{}{q}^+`$ term in Eq.(34). Another obstacle in observing the mode is that it originates from the oscillations of the spin current, which is not a conserved quantity and therefore it is easily damped.
In conclusion, in this paper we described a new collective mode in weak ferromagnetic metals in the quantum spin hydrodynamic regime. While our calculations are for small moment itinerant ferromagnets this mode will exist in other ferromagnetic metals in which the moment is not small. In that case the calculations are significantly more involved. This new mode should exist as well in local moment ferromagnets. Mathematically this can be seen if one relaxes the condition used by Halperin and Hohenberg that the partial time derivative of the spin current is zero. This as well as a calculation of this mode at finite temperature we leave for a future publication.
We would like to thank A. Balatsky, A. Bishop, R. Cowley, P.B. Littlewood, R. McQeeney, J.L. Smith, and S. Trugman, for the fruitful discussions. P. Petkova participated at the early developments of this paper and we would like to thank her for the discussions during that period. K.S. Bedell would like to thank the Aspen Center for Theoretical Physics where some of the ideas developed in this paper were born. This work was supported by the DOE Grant DEFG0297ER45636.
|
warning/0003/nucl-th0003048.html
|
ar5iv
|
text
|
# Reduction of superfluid gap by scattering
## I Introduction
Nuclear systems at low temperatures undergo a superfluid phase transition. This is observed in finite nuclei as the even-odd staggering of masses and is expected to occur in infinite nuclear matter inside neutron stars. Calculations in finite nuclei usually assume an effective pairing interaction fitted to the available data. Calculations in nuclear matter are using predominately the bare nucleon-nucleon interaction in the gap equation.
Due to the short range repulsive core in the nuclear potentials Brueckner type ladder resummation of the interaction in medium is necessary. On the other hand, the effective interaction in the gap equation should be two-particle irreducible, i.e. without resummation of the particle-particle (and hole-hole) ladder migdal . Thus a good starting point could be the bare nucleon-nucleon interaction, the same as used in the ladder approximation for the two-particle correlation.
The value of the pairing gap depends on nucleon single particle energies. Using single-particle energies obtained from Hartree-Fock or Brueckner-Hartree-Fock (BHF) calculation a different value of the paring gap is obtained. It can be understood as a modification of the effective mass, and consequently a change in the density of states at the Fermi energy. Obviously the value the effective mass is an inherent part of effective parameterizations of the pairing interactions used in calculations in finite systems.
Only relatively few works discuss possible in medium modifications of the pairing interactions beyond a change in the effective mass. Polarization corrections to the bare nucleon-nucleon interactions were discussed in Refs. wambach . It was found that the screening reduces the pairing gap by a factor $`3`$. Another in medium modification was analyzed in the framework of self-consistent nuclear matter calculations ja2 . Self-consistent nuclear matter calculations use off-shell nucleon propagators in the resummation of ladder diagrams for the self-energy. In this way a self-consistent spectral function can be obtained. It was found that the use of full spectral functions in the gap equation leads to a strong reduction of the critical temperature and of the superfluid gap in comparison to the quasi-particle approximation. It is the goal of the present paper to identify the main cause of this modification of superfluid properties of nuclear matter and to propose a renormalization of the quasi-particle approximation for the gap equation. In Sect. II we discuss modifications of the two-particle correlations in the normal phase. This enables us to calculate the critical temperature, at which long range two-particle pairing correlations appear. The critical temperature is estimated using full spectral functions and in the renormalized quasi-particle approximation. In Sect. III we present a discussion of the gap equation with full spectral function and compare it to its quasi-particle limit and to the usual mean-field gap equation. In the concluding section IV we identify the most important renormalization of the pairing interaction due to off-shell propagation and indicate its consequences for realistic nuclear matter calculations.
## II Quasi-particle limit of self-consistent ladder resummation
Nuclear medium is a relatively dense system of particles strongly interacting on short distances. Brueckner resummation of particle-particle ladder diagrams defines the so called in medium G-matrix
$`<𝐩|G(𝐏,\mathrm{\Omega })|𝐩^{^{}}>`$ $`=`$ $`V(𝐩,𝐩^{^{}})+{\displaystyle \frac{d^3q}{(2\pi )^3}V(𝐩,𝐪)\frac{(1f(\omega _{p_1}))(1f(\omega _{p_2}))}{\mathrm{\Omega }\omega _{p_1}\omega _{p_2}}<𝐪|G(𝐏,\mathrm{\Omega })|𝐩^{^{}}>},`$ (1)
where $`𝐩_{\mathrm{𝟏},\mathrm{𝟐}}=𝐏/2\pm 𝐪`$. G-matrix resummation allows to define single particle energies and gives relatively good results for the saturation properties of nuclear matter. In the above equation and in the following we skip the spin, isospin indices which are implicitly summed over. The above equation corresponds to a resummation of particle-particle ladders, with medium effects entering through the Pauli blocking factors $`1f(\omega _p)`$ in the numerator and single-particle energies $`\omega _p`$ in the denominator. Most advanced calculations in the Brueckner scheme use the so called continuous choice for the single particle energies $`\omega _p`$, self-consistently defined by the G-matrix bhf .
Another approach starts from the T-matrix approximation for the two-particle correlations vo ; roepke . In this scheme the ladder diagrams include both particle particle and hole-hole propagation. The Pauli blocking factor $`(1f(\omega _{p_1}))(1f(\omega _{p_2}))`$ in the G-matrix equation is replaced by $`1f(\omega _{p_1})f(\omega _{p_2})`$ in the equation for the retarded T-matrix
$`<𝐩|T(𝐏,\mathrm{\Omega })|𝐩^{^{}}>`$ $`=`$ $`V(𝐩,𝐩^{^{}})+{\displaystyle \frac{d^3q}{(2\pi )^3}V(𝐩,𝐤)\frac{(1f(\omega _{p_1})f(\omega _{p_2}))}{\mathrm{\Omega }\omega _{p_1}\omega _{p_2}+iϵ}<𝐪|T(𝐏,\mathrm{\Omega })|𝐩^{^{}}>}.`$ (2)
The imaginary part of the retarded self-energy in the T-matrix approximation is
$`Im\mathrm{\Sigma }(p,\omega )={\displaystyle \frac{d^3k}{(2\pi )^3}<(𝐩𝐤)/2|ImT(|𝐩+𝐤|,\omega _k+\omega )|(𝐩𝐤)/2>_A\left(f(\omega _k)+b(\omega +\omega _k)\right)},`$ (3)
where $`b(\omega )`$ is the Bose distribution. $`<\mathrm{}>_A`$ denotes antisymmetrization of the T-matrix (also in the spin, isospin indices not explicitly show). The real part of the self energy consists of the Hartree Fock self-energy and a dispersive contribution obtained from $`Im\mathrm{\Sigma }`$
$$Re\mathrm{\Sigma }(p,\omega )=\mathrm{\Sigma }_{HF}(p)+𝒫\frac{d\omega ^{^{}}}{\pi }\frac{Im\mathrm{\Sigma }(p,\omega ^{^{}})}{\omega ^{^{}}\omega }.$$
(4)
The imaginary part of the self-energy is usually neglected leading to the quasi-particle approximation for the two-nucleon propagator in the T-matrix (Eq. 2).
Allowing for off-shell propagation of nucleons and taking the self-energy self-consistently (also its imaginary part) requires the use of full spectral functions in the calculation resulting in a more complicated expressions for the T-matrix and the self-energy d1 ; ja
$`<𝐩|T(𝐏,\mathrm{\Omega })|𝐩^{^{}}>`$ $`=`$ $`V(𝐩,𝐩^{^{}})`$ (5)
$`+{\displaystyle \frac{d\omega _1}{2\pi }\frac{d\omega _2}{2\pi }\frac{d^3q}{(2\pi )^3}V(𝐩,𝐪)\frac{\left(1f(\omega _1)f(\omega _2)\right)}{\mathrm{\Omega }\omega _1\omega _2+iϵ}A(p_1,\omega _1)A(p_2,\omega _2)<𝐪|T(𝐏,\mathrm{\Omega })|𝐩^{^{}}>}`$
and
$`\mathrm{Im}\mathrm{\Sigma }^+(p,\omega )`$ $`=`$ $`{\displaystyle \frac{d\omega _1}{2\pi }\frac{d^3k}{(2\pi )^3}A(k,\omega _1)<(𝐩𝐤)/2|\mathrm{Im}T(𝐩+𝐤,\omega +\omega _1)|(𝐩𝐤)/2>_A\left(f(\omega _1)+g(\omega +\omega _1)\right)}.`$ (6)
Equations (5), (6) and (4) have to be solved iteratively with the constraint on the assumed density $`\rho `$ ja ; ja2
$$\rho =\frac{d\omega }{2\pi }\frac{d^3p}{(2\pi )^3}A(p,\omega )f(\omega ).$$
(7)
Results of the self-consistent calculation and of the quasi-particle approximation are very different. In medium cross sections for nucleon-nucleon scattering are smaller when using off-shell nucleons d1 . Also the critical temperature is strongly reduced when using full spectral functions ja2 .
The T-matrix approximation for the two-particle propagator is directly related to the superfluid gap properties by the Thouless criterion thouless and the condition for long range order km ; hau2 . On the other BHF calculation using the G-matrix are are much developed for realistic interactions. In this work we derive an improved gap equation which could use the results of advanced G-matrix calculations as an input.
### II.1 Thouless criterion for superfluidity
The critical temperature can be obtained from the Thouless criterion for superconductivity thouless . This is the temperature where a singularity in the T-matrix appears at twice the Fermi energy ($`\mathrm{\Omega }=0`$) and zero total momentum of the pair ($`P=0`$). It means that the real part of the inverse T-matrix develops a zero-eigenvalue at the critical temperature
$$\frac{d^3p^{^{}}}{(2\pi )^3}<𝐩|\mathrm{Re}T^1(𝐏=\mathrm{𝟎},\omega =0)|𝐩^{^{}}>\mathrm{\Delta }(𝐩^{^{}})=0.$$
(8)
It is equivalent to the existence of a nontrivial solution of the gap equation at $`T_c`$
$`\mathrm{\Delta }(𝐩)=`$
$`{\displaystyle \frac{d^3k}{(2\pi )^3}\frac{d\omega }{2\pi }\frac{d\omega ^{^{}}}{2\pi }V(𝐩,𝐤)\frac{A(k,\omega \omega ^{^{}})A(k,\omega )\left(1f(\omega \omega ^{^{}})f(\omega ^{^{}})\right)}{\omega }\mathrm{\Delta }(𝐤)}.`$ (9)
The two propagators in the gap equation above enter with the full spectral function. On the other hand the BCS quasi-particle gap equation is
$`\mathrm{\Delta }(𝐩)+{\displaystyle \frac{d^3k}{(2\pi )^3}V(𝐩,𝐤)\frac{\left(12f(\zeta _k)\right)}{2\zeta _k}\mathrm{\Delta }(𝐤)}=0`$ (10)
with $`\zeta _p=p^2/2m+\mathrm{\Sigma }_{HF}(p)\mu `$, and it corresponds to the Thouless criterion for the quasi-particle T-matrix, i.e. the appearance of a singularity in the T-matrix given by Eq. (2). Because of additional averaging over spectral functions in the gap equation with off-shell propagators, a different value of critical temperature comes out.
The self-consistent T-matrix calculation was done for a simple, S-wave, interaction and compared to the quasi-particle approximation for the gap equation ja2 . At a density of $`0.45`$ of normal nuclear density $`\rho _0`$ it was found that the critical temperature was reduced from $`T_c=5`$MeV in the mean-field gap equation with Hartree-Fock single particle energies to $`T_c=1.6`$MeV in the T-matrix approximation with off-shell propagation ja2 . Below $`T_c`$ a modified T-matrix resummation was used in Ref. ja2 . The resulting superfluid gap is significantly smaller than the one obtained from the usual gap equation with quasiparticles. Modifications to the gap equation below $`T_c`$ coming from the use of full spectral functions will be discussed in Sect. III. Here we concentrate on the T-matrix equations in the normal phase, which is sufficient for the calculation of the critical temperature. The simple Yamaguchi S wave interaction ja2 does not permit calculations at low temperature for normal density nuclear matter.
### II.2 Renormalized quasi-particle interactions
Excitations in the Fermi liquid close to the Fermi energy can be described by quasi-particles. Quasi-particles are propagating on shell with dispersion relation modified by the presence of the medium. Also the scattering amplitudes between quasi-particles are modified by the medium. One of these modification comes from the quasi-particle limit in the propagator of two nucleons migdal ; pn . Two particle Green’s function can be formally written as a resummation of particle-particle (and hole-hole) ladder diagrams starting from in medium two-particle irreducible vertex migdal ; abrikosov . For short range interactions the two-particle irreducible vertex can be approximated in the lowest order by the bare nucleon-nucleon interaction, leading to the T-matrix equation (5). It should be pointed out that due to the instantaneous form of the interaction in the T-matrix equation, the full T-matrix depends only on the total energy, and its equation takes a simple form for the retarded T-matrix (Eq. 5), also at finite temperature.
Close to the Fermi energy the spectral function becomes peaked around the quasi-particle pole
$$A(p,\omega )=Z_p2\pi \delta (\omega \omega _p)+R(p,\omega ),$$
(11)
where
$$Z_p=\left(1\frac{Re\mathrm{\Sigma }(p,\omega )}{\omega }|_{\omega =\omega _p}\right)^1$$
(12)
and $`R(p,\omega )`$ is the regular part, smooth in the vicinity of the quasi-particle pole. The retarded propagator of two nucleons appearing in the T-matrix ladder can be written as
$`{\displaystyle \frac{\left(1f(\omega _1)f(\omega _2)\right)}{\mathrm{\Omega }\omega _1\omega _2+iϵ}}A(p_1,\omega _1)A(p_2,\omega _2)`$ $`=`$ $`B^{reg}(p_1,\omega _1,p_2,\omega _2,\mathrm{\Omega })`$ (13)
$`+(2\pi )^2\delta (\omega _1\omega _{p_1})\delta (\omega _2\omega _{p_2}){\displaystyle \frac{Z_{p_1}Z_{p_2}\left(1f(\omega _{p_1})f(\omega _{p_2})\right)}{\mathrm{\Omega }\omega _{p_1}\omega _{p_2}}}.`$
$`B^{reg}`$ describes the part of the propagator of two nucleons which cannot be written as propagation of two quasi-particles, it originates from the background part $`R(p,\omega )`$ of the spectral function. Besides this contribution, the propagator of two-nucleons differs from the one used in the quasi-particle T-matrix (2) by the presence of renormalization factors $`Z_p`$ (Eq. 12). The T-matrix with quasi-particle propagators takes the following renormalized form
$`<𝐩|T(𝐏,\mathrm{\Omega })|𝐩^{^{}}>`$ $`=`$ $`<𝐩|V^{ren}(𝐏,\mathrm{\Omega })|𝐩^{^{}}>`$ (14)
$`+{\displaystyle \frac{d^3q}{(2\pi )^3}<𝐩|V^{ren}(𝐏,\mathrm{\Omega })|𝐪>Z_{p_1}Z_{p_2}\frac{(1f(\omega _{p_1})f(\omega _{p_2}))}{\mathrm{\Omega }\omega _{p_1}\omega _{p_2}+iϵ}<𝐪|T(𝐏,\mathrm{\Omega })|𝐩^{^{}}>},`$
with the renormalized interaction $`V^{ren}`$ given by
$`<𝐩|V^{ren}(𝐏,\mathrm{\Omega })|𝐩^{^{}}>=V(𝐩,𝐩^{^{}})+{\displaystyle \frac{d\omega _1}{2\pi }\frac{d\omega _2}{2\pi }\frac{d^3q}{(2\pi )^3}V(𝐩,𝐪)B(p_1,\omega _1,p_2,\omega _2,\mathrm{\Omega })<𝐪|V^{ren}(𝐏,\mathrm{\Omega })|𝐩^{^{}}>}.`$ (15)
The bare interaction is renormalized by contributions from background parts of the spectral functions, this involves integration over energies far from the quasi-particle pole. The background part of the spectral function is of course necessary to recover sum rules for the particle strength. However, the full treatment of the renormalized interaction is difficult and in the following we replace it by the bare interaction $`V^{ren}V`$ in the gap equation. The only remnant of the dressing of nucleons in medium are the single particle energies
$$\omega _p=\xi _p+Re\mathrm{\Sigma }(p,\omega _p)$$
(16)
and the $`Z_p`$ factors in the homogeneous term in the T-matrix equation. It will turn out that these are the dominant modifications responsible for shifting the critical temperature. In the vicinity of the pole the T-matrix equation is dominated by the homogeneous term. In this region the interaction for quasi-particles is effectively renormalized by a factor $`Z_{p_1}Z_{p_2}`$.
### II.3 Two-particle pole with renormalized interactions
According to the Thouless criterion, a pole in the T-matrix at the Fermi energy ($`\mathrm{\Omega }=0`$) means that superfluid long range order sets in (Sect. II.1). As mentioned earlier, standard mean-field gap equation uses the following kernel
$$V(p,k)\frac{\left(12f(\zeta _k)\right)}{2\zeta _k}$$
(17)
with mean-field single particle energies $`\zeta _k=k^2/2m+\mathrm{\Sigma }_{HF}(k)\mu `$. It leads to a critical temperature $`T_c=5`$MeV. One could take single particle energies $`\omega _p`$ beyond Hartree-Fock, e.g. from BHF or T-matrix calculations. However, the difference between the resulting effective masses is of the order $`Z_p`$. Hence we stay at the order of Hartree-Fock effective mass in the standard gap equation, including the modified single particle energies (16) only together with $`Z_p`$ factor renormalization.
The second estimate for the critical temperature is obtained from the condition of appearance of the pole in the T-matrix with full self-consistent spectral functions. It is equivalent to the following kernel in the gap equation
$$\frac{d\omega }{2\pi }\frac{d\omega ^{^{}}}{2\pi }V(𝐩,𝐤)\frac{A(k,\omega \omega ^{^{}})A(k,\omega )\left(1f(\omega \omega ^{^{}})f(\omega ^{^{}})\right)}{\omega },$$
(18)
where $`A(p,\omega )`$ is obtained from full self-consistent calculation of normal nuclear matter at finite temperature with off-shell propagators in the T-matrix ladder ja . It gives a very different value for $`T_c=1.6`$MeV at $`\rho =.45\rho _0`$.
Finally we can use the renormalized interaction strength in the homogeneous term of the T-matrix equation (Eq. 14). It is equivalent to using a renormalized interaction in the kernel of the quasi-particle gap equation
$$V(p,k)Z_p^2\frac{\left(12f(\omega _k)\right)}{2\omega _k},$$
(19)
with the single particle energies $`\omega _p`$ and $`Z_p`$ factors obtained from the full self-consistent T-matrix calculation (Eq. 16). It gives a value of $`T_c=2.2`$MeV, much closer to the result of the calculation with full spectral functions. The $`Z`$ factor obtained from the full self-energy is not close to $`1`$, at the Fermi energy we find in our model calculation ja2 $`Z_{p_f}0.7`$. Such a small value of the renormalization factor can explain the large difference in critical temperatures found in the standard quasi-particle gap equation and in the one with full spectral function.
## III Gap equation with dressed propagators
### III.1 Gap equation with full spectral function
Below the critical temperature a nonzero solution of the gap equation is possible. The kernel of the gap equation is very similar to the two-particle propagator in the T-matrix resummation (Eq. II.1). However one of the propagators includes the normal self-energy as well as the off-diagonal one $`\mathrm{\Delta }(p)`$ ja2 . The full retarded propagator can be expressed using the normal propagator $`G(p,\omega )`$ (which includes only the normal self-energy)
$$G_s(p,\omega )=\frac{1}{G(p,\omega )^1+|\mathrm{\Delta }|^2(p)G^{}(p,\omega )}.$$
(20)
As a result in the kernel of the gap equation
$`\mathrm{\Delta }(𝐩)={\displaystyle \frac{d^3k}{(2\pi )^3}\frac{d\omega }{2\pi }\frac{d\omega ^{^{}}}{2\pi }V(𝐩,𝐤)\frac{A(k,\omega \omega ^{^{}})A_s(k,\omega )\left(1f(\omega \omega ^{^{}})f(\omega ^{^{}})\right)}{\omega }\mathrm{\Delta }(𝐤)}.`$ (21)
two spectral functions $`A(p,\omega )=2ImG(p,\omega )`$ and $`A_s(p,\omega )=2ImG_s(p,\omega )`$ appear. Both spectral functions can be calculated knowing the two self-energies $`\mathrm{\Sigma }(p,\omega )`$ and $`\mathrm{\Delta }(p)`$. The off-diagonal self-energy, ie. the superfluid gap can be obtained from the gap equation (21). On the other hand the normal self-energy cannot be calculated in the usual T-matrix approximation as above $`T_c`$. This is related to the appearance of the Cooper instability in the two-particle propagator in the ladder approximation.
The T-matrix equation can be modified by introducing also anomalous propagators in the ladder hau2 . In Ref. ja2 we used a simpler modification of the ladder diagrams, using a mixed ladder with one full propagator and one with only normal self-energy included. This means that the singularity of the T-matrix which appears at $`T_c`$ at the Fermi energy $`\mathrm{\Omega }=0`$ and zero total momentum of the pair, stays there also below $`T_c`$. This reflects the presence of long range order km ; hau2 . In such a way we were able to obtain self-consistent solutions for the normal self-energy in the ladder approximation and for the superfluid gap in the mean-field approximation, using at all stages full spectral functions $`A(p,\omega )`$ and $`A_s(p,\omega )`$, without quasi-particle approximation. The results for the superfluid gap at several temperatures below $`T_c`$ are indicated by triangles in Fig. 1. This calculation cannot be extended straightforwardly to zero temperature because at the Fermi energy the imaginary part of the self-energy vanishes and the discretization of the spectral functions is not possible close to the Fermi energy. At the Fermi momentum quasi-particles appear with small width.
By comparing the imaginary part of the self-energy at two temperatures $`T=1.7`$MeV and $`1.2`$MeV one notices that the only modification is the reduction of single-particle width close to the Fermi energy with temperature, in agreement with general properties of Fermi liquids. Thus we can take as an approximation a temperature independent imaginary part of the self-energy. We proceed by taking the imaginary part of the self-energy $`Im\mathrm{\Sigma }(p,\omega )`$ as calculated at $`T=1.63`$MeV, slightly above $`T_c`$. Obviously the dispersive contribution to the real part of the self-energy is also fixed (4). The Hartree-Fock energy is obtained from
$$\mathrm{\Sigma }_{HF}(p)=\frac{d^3k}{(2\pi )^3}V(|𝐩𝐤|/2,|𝐩𝐤|/2)n(p)$$
(22)
where the momentum distribution is given by
$$n(p)=\frac{d\omega }{2\pi }A_s(p,\omega )f(\omega ).$$
(23)
The shift of the Fermi energy from the value $`\omega =0`$ at $`T=1.63`$MeV to keep the density
$$\rho =.45\rho _0=\frac{d^3p}{(2\pi )^3}n(p)$$
(24)
constant when decreasing the temperature to zero is only $`0.2`$MeV and has negligible effect on the Hartree-Fock energy. Thus the main change in the self-energy when decreasing the temperature occurs in the off-diagonal part $`\mathrm{\Delta }(p)`$. At any given temperature the superfluid gap is obtained from the gap equation with full spectral function (21). Where the dependence of the kernel of the gap equation on $`\mathrm{\Delta }(p)`$ enters through the spectral function
$`A_s(p,\omega )`$ $`=`$ $`2((\omega +\xi _p+\mathrm{Re}\mathrm{\Sigma }^+(p,\omega ))^2\mathrm{Im}\mathrm{\Sigma }^+(p,\omega )+\mathrm{Im}\mathrm{\Sigma }^+(p,\omega )\mathrm{\Delta }^2(p)+\left(\mathrm{Im}\mathrm{\Sigma }^+(p,\omega )\right)^2\mathrm{Im}\mathrm{\Sigma }(p,\omega ))/`$ (25)
$`(((\omega \xi _p\mathrm{Re}\mathrm{\Sigma }^+(p,\omega ))(\omega +\xi _p+\mathrm{Re}\mathrm{\Sigma }^+(p,\omega ))\mathrm{Im}\mathrm{\Sigma }^+(p,\omega )\mathrm{Im}\mathrm{\Sigma }^+(p,\omega )\mathrm{\Delta }^2(p))^2`$
$`+(\mathrm{Im}\mathrm{\Sigma }^+(p,\omega )(\omega +\xi _p+\mathrm{Re}\mathrm{\Sigma }^+(p,\omega ))+\mathrm{Im}\mathrm{\Sigma }^+(p,\omega )(\omega \xi _p\mathrm{Re}\mathrm{\Sigma }^+(p,\omega )))^2).`$
In the above equation $`Im\mathrm{\Sigma }`$ is fixed and temperature independent, according to our approximation, and $`Re\mathrm{\Sigma }`$ depends only very weakly on the temperature and the superfluid gap through the Hartree-Fock energy. The solution of the gap equation using a fixed single particle width down to zero temperature is represented by the solid line in Fig. 1. As expected close to the temperature where the imaginary part of the self-energy was fixed ($`T=1.63`$MeV) the result is close to the fully self-consistent solution denoted by the triangles on the figure. As the temperature is lowered some deviations appear. It is due to the decrease of the single particle width close to the Fermi energy in the fully self-consistent solution (Fig. 2). In the following we will compare the solution of the gap equation with full spectral function to the corresponding quasi-particle approximation. Since the imaginary part of the self-energy is fixed at $`T=1.63`$MeV, the properties of the quasi-particle pole ($`\omega _p`$ and $`Z_p`$) are taken from the self-consistent solution at the same temperature. Our aim is to explain the big difference between the mean-field solution with on-shell propagators (long-dashed line in Fig. 1) and the gap equation with full spectral function below $`T_c`$ (solid line). For this purpose it is sufficient to compare the approximation with nontrivial but fixed spectral properties with its quasi-particle limit.
The mean-field BCS gap equation is
$`\mathrm{\Delta }(𝐩)+{\displaystyle \frac{d^3k}{(2\pi )^3}V(𝐩,𝐤)\frac{\left(12f(E_k)\right)}{2E_k}\mathrm{\Delta }(𝐤)}=0,`$ (26)
where $`E_k=\sqrt{\zeta _k^2+\mathrm{\Delta }(k)^2}`$. It gives significantly larger values for the superfluid gap (long-dashed line in Fig. 1) than the full solution (solid line) in all ranges of temperatures below $`T_c`$.
### III.2 Quasi-particle approximation for the superfluid spectral function
Above $`T_c`$ we have used the standard form of the quasi-particle approximation for the spectral function $`A(p,\omega )=2\pi Z_p\delta (\omega \omega _p)`$. Using it we have found a significant reduction of the critical temperature with respect to the mean-field approximation. To obtain a quasi-particle approximation for the kernel of the gap equation (21) with full spectral function we have to construct an approximation for the superfluid spectral function $`A_s`$. We can write the spectral function (25) putting an infinitesimally small imaginary part of the self-energy ($`Im\mathrm{\Sigma }(p,\omega )=i\eta `$)
$`A_s(p,\omega )=2Im((\omega +\xi _p+Re\mathrm{\Sigma }(p,\omega )+i\eta )/((\omega \xi _pRe\mathrm{\Sigma }(p,\omega ))`$
$`(\omega +\xi _p+Re\mathrm{\Sigma }(p,\omega ))+\mathrm{\Delta }^2(p)+i\eta (2\omega Re\mathrm{\Sigma }(p,\omega )+Re\mathrm{\Sigma }(p,\omega )))),`$ (27)
with $`\xi _p=p^2/2m\mu `$. It is useful to define the even $`S(p,\omega )=Re\mathrm{\Sigma }(p,\omega )+Re\mathrm{\Sigma }(p,\omega )`$ and odd $`\omega O(p,\omega )=Re\mathrm{\Sigma }(p,\omega )Re\mathrm{\Sigma }(p,\omega )`$ parts of the real part of the self-energy with respect to the Fermi energy mahan . Notice that we can take $`S(p,0)=0`$ redefining $`\mu `$. We have
$`A_s(p,\omega )=2Im\left({\displaystyle \frac{\omega (1O(p,\omega ))+\xi _p+S(p,\omega )}{\left((1O(p,\omega ))E(p,\omega )\right)\left(\omega (1O(p,\omega ))+E(p,\omega )\right)+i\eta \omega (1O(p,\omega ))}}\right),`$ (28)
where
$$E(p,\omega )=\sqrt{(\xi _p+S(p,\omega ))^2+\mathrm{\Delta }^2(p)}.$$
The above expression has poles at
$$\pm \omega =ϵ_p=\frac{E(p,ϵ_p)}{1O(p,ϵ_p)}=\frac{E_p}{1O_p},$$
which gives two quasi-particle contributions to the spectral function on both sides of the Fermi energy
$$A_s(p,\omega )=2\pi Z_p^{^{}}\left(\frac{E_p+\xi _p+S_p}{2E_p}\delta (\omega ϵ_p)+\frac{E_p\xi _pS_p}{2E_p}\delta (\omega ϵ_p)\right),$$
(29)
with $`S_p=S(p,ϵ_p)`$ and
$$Z_p^{}_{}{}^{}1=\left(1\frac{\left(E(p,\omega )/(1O(p,\omega ))\right)}{\omega }|_{\omega =ϵ_p}\right)(1O_p).$$
(30)
It will be useful to relate this new pole renormalization strength to the usual renormalization factor $`Z_p`$ (Eq. 12). We can write
$$Z_p^{}_{}{}^{}1=\left(1\frac{\mathrm{\Sigma }(p,\omega )}{\omega }|_{\omega =\mathrm{sign}(pp_f)ϵ_p}\right)+\frac{E_p\mathrm{sign}(pp_f)(\xi _p+S_p)}{E_p}\frac{S(p,\omega )}{\omega }|_{\omega =ϵ_p},$$
(31)
with $`\mathrm{sign}(x)=\mathrm{\Theta }(x)\mathrm{\Theta }(x)`$ . For small values of the superfluid gap the superfluid quasiparticle position $`\mathrm{sign}(pp_f)ϵ_p`$ is very close to the position of the pole in the normal propagator $`\omega _p=\xi _p+Re\mathrm{\Sigma }(p,\omega _p)`$. Substituting $`\omega _p`$ for the energy argument in the first term in Eq. (31) we get
$$Z_p^{}_{}{}^{}1Z_p^1=\frac{E_p\mathrm{sign}(pp_f)(\xi _p+S_p)}{E_p}\frac{S(p,\omega )}{\omega }|_{\omega =ϵ_p}.$$
(32)
The first factor on the right hand side of the above equation is small except close to the Fermi energy where the second factor precisely vanishes. Thus we expect that the new quasi-particle pole renormalization factor $`Z_p^{^{}}`$ can be approximated by the renormalization factor of the normal spectral function $`Z_p`$ corresponding to the same momentum. A numerical calculation confirms this to a very good accuracy (Fig. 3).
The spectral function takes the form
$$A_s(p,\omega )=2\pi Z_p\left(u_p^2\delta (\omega ϵ_p)+v_p^2\delta (\omega ϵ_p)\right),$$
(33)
with coherence factors
$$u_p^2(v_p^2)=\frac{ϵ_p+()(\xi _p+S_p)/(1O_p)}{2ϵ_p}.$$
(34)
These coherence factors can be very well approximated by an expression similar to the one used in the usual mean-field gap equation
$$u_p^2(v_p^2)=\frac{ϵ_p+()\omega _p}{2ϵ_p}.$$
(35)
The approximation works to within $`1\%`$ for $`\mathrm{\Delta }(p_f)<10`$MeV.
Substituting the quasi-particle expressions for the spectral functions $`A`$ and $`A_s`$ into the kernel of the gap equation one obtains
$$\mathrm{\Delta }(p)=\frac{d^3k}{(2\pi )^3}Z_k^2V(p,k)\frac{(12f(ϵ_p))}{2ϵ_p}.$$
(36)
We have obtained an expression for the gap equation with a kernel very similar to the mean-field BCS one (Eq. 17), but with the interaction renormalized by $`Z_p^2`$ and a different quasi-particle energy $`ϵ_p`$.
In the limit of vanishing gap $`\mathrm{\Delta }(p)0`$ all quasi-particle approximations for the gap equation reduce to the Thouless condition for the renormalized T-matrix equation (Eq. 19). Accordingly we have the same condition for $`T_c`$ as discussed in Sect. II.3.
### III.3 Quasi-particle energies
In order to relate our quasi-particle gap equation (36) to usual approaches we have to obtain an approximation for the quasi-particle energies
$$ϵ_p=\frac{\sqrt{(\xi _p+S_p)^2+\mathrm{\Delta }^2(p)}}{1O_p}.$$
(37)
To extract the positions $`\pm ϵ_p`$ of the poles of the spectral function one has to know the real part of the self energy for energies on both sides of the Fermi energy. We would like to obtain an expression which could be used as a correction to calculations using quasi-particle approximation like BHF.
In BHF approaches one calculates a single particle energy which includes dispersive corrections to the position of the quasi-particle pole (Eq. 16) in the usual spectral function $`A(p,\omega )`$. A simple approximation would be to use $`\omega _p`$ instead of $`\zeta _p`$ in the expression for the quasi-particle energies in the superfluid
$$ϵ_p=\sqrt{\omega _p^2+\mathrm{\Delta }^2(p)}.$$
(38)
Clearly the above expression does not give the right value for the energy gap at the Fermi momentum (Fig. 4). The energy gap at $`p_F`$ should be
$$ϵ_{p_F}=\frac{\mathrm{\Delta }(p_F)}{1O_{p_F}}=Z_{p_F}\mathrm{\Delta }(p_F).$$
(39)
Let us define a renormalized energy gap
$$\stackrel{~}{\mathrm{\Delta }}(p)=\frac{\mathrm{\Delta }(p)}{1O_p}.$$
(40)
Results for the quasi-particle poles using this renormalization of the superfluid order parameter
$$ϵ_p=\sqrt{\omega _p^2+\stackrel{~}{\mathrm{\Delta }}^2(p)}$$
(41)
are indistinguishable from the solution (37) in Fig. 4. It should be noted that for the calculation of the factor $`(1O_p)`$ we can take $`O_p=O(p,ϵ_p)O(p,\omega _p)`$. So that the approximation for $`ϵ_p`$ can be expressed using the quasi-particle pole energy $`\omega _p`$ and the values of the real part of the self energy at $`\pm \omega _p`$. However, usually in BHF calculations one does not calculate the factor $`1O_p`$, since the value and the derivative of the real part of the self-energy is known only at $`\omega _p`$ and not at $`\omega _p`$. In this case the position of the poles in the superfluid can be approximated by
$$ϵ_p=\sqrt{\omega _p^2+\widehat{\mathrm{\Delta }}^2(p)}$$
(42)
with the renormalization of the superfluid gap taken as
$$\widehat{\mathrm{\Delta }}(p)=\mathrm{\Delta }(p)Z_p.$$
(43)
$`\widehat{\mathrm{\Delta }}(p)`$ is very close to $`\stackrel{~}{\mathrm{\Delta }}(p)`$ around the Fermi energy. Some differences appear only around $`p=0`$ (for the assumed interaction), but this region is less important for the solution of the gap equation. Again the energies obtained using expression (42) are indistinguishable from the solid line in Fig. 4.
### III.4 Results for superfluid gap in quasi-particle approximation
In previous sections we derived a quasi-particle approximation for the kernel of the superfluid gap equation (Eq. 21). The results of the full gap equation with integration of over the energies in the fermion propagators (solid line in Fig. 1) can be compared to quasi-particle results. The dotted line in Fig. 1 represents the solution of the gap equation (36) with renormalization of the energy gap $`\stackrel{~}{\mathrm{\Delta }}(p)`$ for the calculation of the superfluid quasi-particle energies (Eq. 41). The dashed-dotted line represents an analogous calculation but with the energy gap $`\widehat{\mathrm{\Delta }}(p)`$ (Eq. 43). Both quasi-particle approximation are very close to each other with critical temperature $`T_c=2.2`$MeV. The value of $`T_c`$ is the same as obtained in Sect. II.3 using renormalized interactions in the ladder approximation. This value of critical temperature is much closer to the one obtained from the self-consistent solution $`T_c=1.6`$MeV than was the mean-field value $`T_c=5`$MeV. Also the value of the superfluid gap is much closer to the one calculated using full of-shell spectral functions (Table 1). The superfluid gap at zero temperature in the quasi-particle approximation $`\mathrm{\Delta }(p_F)=6`$MeV is much closer to the solution of the off-shell gap equation $`\mathrm{\Delta }(p_F)=4.3`$MeV than the one obtained from the mean-field BCS solution $`\mathrm{\Delta }(p_F)=10.8`$MeV. From the results in Table 1 we can notice that the ratio between the superfluid gap at zero temperature and the critical temperature is larger in the solution of the full gap equation (21) and in quasi-particle approximations (36) with renormalization of the energy gap (41 or 42), than in the mean-field solution (26) and in quasi-particle approximations without renormalization of the superfluid energy gap (36, 38). In the weak coupling limit we have mahan ; fw
$$\frac{\mathrm{\Delta }(p_F)|_{T=0}}{T_c}=\pi e^\gamma 1.76.$$
(44)
If the energy gap $`\stackrel{~}{\mathrm{\Delta }}(p_F)`$ is renormalized with respect to the superfluid gap $`\mathrm{\Delta }(p_F)`$ (the off diagonal self-energy) we have
$$\frac{\mathrm{\Delta }(p_F)|_{T=0}}{T_c}=\frac{\stackrel{~}{\mathrm{\Delta }}(p_F)|_{T=0}(1O_{p_F})}{T_c}=\frac{\stackrel{~}{\mathrm{\Delta }}(p_F)|_{T=0}}{T_cZ_{p_F}}\frac{1.76}{Z_{p_F}}2.5.$$
(45)
Ratios of the superfluid gap and of the critical temperature for the solution of the full gap equation and quasi-particle approximations with renormalized superfluid gap is close to $`1.76/Z_{p_F}`$ (Table 1). On the other hand mean-field gap equation and quasi-particle approximation without renormalization of $`\mathrm{\Delta }(p)`$ give $`\mathrm{\Delta }(p_F)|_{T=0}/T_c`$ closer to $`1.76`$. These relation can be fulfilled only approximately since we are far from the region of applicability of the weak coupling BCS solution.
## IV Conclusions
Numerical solution of the gap equation with full spectral functions (21) shows a strong reduction of the superfluid gap and of the critical temperature with respect to the mean-field BCS solution (26). To understand this effect we constructed a quasi-particle approximation for the full gap equation. The effects of nontrivial spectral functions can be approximated using a renormalized strength of the interaction $`V(p,k)Z_k^2`$. Also the energy gap in the calculation of the quasi-particle poles in the superfluid is renormalized. For the renormalization of the gap we used two expression $`\mathrm{\Delta }(p)/(1O_p)`$ and $`\mathrm{\Delta }(p)Z_p`$. In both cases we obtained results for $`\mathrm{\Delta }(p)`$ and $`T_c`$ much closer to the full solution than the mean-field approximation. Having at one’s disposal only the single particle energies $`\omega _p`$ and the $`Z_p`$ factors obtained from a realistic BHF type calculation, the gap equation can be corrected. A reduced interaction strength $`V(p,k)Z_k^2`$ must be taken and a factor $`Z_p`$ appears between the energy gap $`\widehat{\mathrm{\Delta }}(p)`$ and the off-diagonal self-energy $`\mathrm{\Delta }(p)`$. This is equivalent to solving a gap equation for $`\widehat{\mathrm{\Delta }}(p)`$ with reduced interaction $`V(p,k)Z_pZ_k`$. In the weak coupling limit $`\widehat{\mathrm{\Delta }}(p)|_{T=0}1.76T_c`$ but $`\mathrm{\Delta }(p)|_{T=0}1.76T_c/Z_{p_F}`$ .
In the illustrative model here presented the scattering corrections are very strong ($`Z_{p_F}.7`$) and there are still some differences between the improved quasi-particle approximation and the full solution of the gap equation. We expect that at normal nuclear density where the $`Z`$ factor is closer to $`1`$, quasi-particle approximation with renormalized interaction strength would be much closer to the full solution. This is the case for neutron matter where effects of renormalization of quasi-particle poles and shifts in single-particle energies are smaller baldop . In future work we plan to investigate the effects of the renormalization of the interaction by background corrections (15).
In this investigation we used the bare potential for the two-particle irreducible vertex. However, it is known that polarization effects reduce the superfluid gap by a factor $`3`$ wambach . We must conclude that although there are no ladder corrections to the interaction in the gap equation, other many-body effects modify the effective interaction. Due to the exponential dependence of the gap solution on the strength of the interaction, these usually neglected corrections modify strongly superfluid parameters in nuclear matter.
###### Acknowledgements.
This work was partly supported by the National Science Foundation under Grant PHY-9605207.
|
warning/0003/hep-ph0003209.html
|
ar5iv
|
text
|
# Magnetic moment of Δ⁺⁺ baryon in QCD string approach
## Abstract
Magnetic moments (m.m.) of the $`\mathrm{\Delta }^{++}`$ baryons is computed within the new approach based on the QCD string Hamiltonian. The string tension $`\sigma `$ is only dimensionful quantity forming m.m. of both nucleon and $`\mathrm{\Delta }^{++}`$, but color Coulomb and spin-spin interactions cancel each other in nucleon m.m. while in $`\mathrm{\Delta }^{++}`$ they add coherently. The result $`\mu _{\mathrm{\Delta }^{++}}=4.36\mu _N`$ is in good agreement with experimental data.
Recently in a new approach to evaluation of the baryon m.m. has been proposed. It is based on the Feynman-Schwinger (world-line) representation of the 3q Green’s function and yields remarkably simple expressions for the m.m. through the only fundamental parameter –string tension $`\sigma `$ (and strange quark current mass for strange baryons.) In calculations have been performed for the octet baryons and for $`\mathrm{\Omega }^{}`$. The present letter aims at the calculation of the $`\mathrm{\Delta }^{++}`$ m.m. and some amendments to our previous treatment of $`\mathrm{\Omega }^{}`$.
The case of $`\mathrm{\Delta }^{++}`$ is a particular one both from experimental and theoretical sides. Experimentally m.m. of $`\mathrm{\Delta }^{++}`$ was measured rather recently and is still subject to substantial uncertainties . On the theoretical side the calculation of the $`\mathrm{\Delta }^{++}`$ m.m. causes serious difficulties in various approaches and calls for introduction of several additional parameters (see e.g. recent paper for a brief review of the current status of the sum rules approach to the problem) As will be seen in what follows the treatment of the $`\mathrm{\Delta }^{++}`$ in our approach requires taking into account hyperfine and color Coulomb interactions. This was not the case for octet baryons considered in since for them Coulomb and spin-spin terms extinguished each other . For $`\mathrm{\Omega }^{}`$ baryon also considered in the situation is the following. As for other baryons the m.m. of $`\mathrm{\Omega }^{}`$ is predominantly controlled by string tension $`\sigma `$, next comes the contribution from the strange quark current mass $`m_s`$, and then Coulomb and hyperfine terms contribute. In the first two factors ($`\sigma `$ and $`m_s`$) were considered while the present work deals in addition with the two remaining terms listed above. This does not alter the value of the $`\mathrm{\Omega }^{}`$ m.m. obtained in but allows to take somewhat smaller value of $`m_s`$ than that used in .
We start with a recapitulation of the very few key points of the approach to baryon m.m. developed in . For the $`3q`$ Green’s function one can write the following Feynman–Schwinger (world-line representation -)
$$G^{(3q)}(X,Y)=\underset{i=1}{\overset{3}{}}ds_iDz_\mu ^{(i)}e^KW_3(X,Y),$$
(1)
where $`X;Y=x^{(i)};y^{(i)},i=1,2,3`$, and the integration is along the path $`z_\mu ^{(i)}(s_i)`$ of the $`i`$-th quark with $`s_i`$ playing the role of the proper ”time” parameter along the path, and
$$K=\underset{i=1}{\overset{3}{}}\left(m_i^2s_i+\frac{1}{4}_0^{s_i}\left(\frac{dz_\mu ^{(i)}}{d\tau _i}\right)𝑑\tau _i\right).$$
(2)
Here $`m_i`$ is the current quark mass and the three-lobed Wilson loop is a product of the three parallel transporters -. The standard approximation of the QCD string approach is the minimal area law for $`W_3`$
$$W_3=\mathrm{exp}(\sigma \underset{i=1}{\overset{3}{}}S_i),$$
(3)
where $`S_i`$ is the minimal area of one loop. The next step is to calculate the quark constituent mass $`\nu _i`$ in terms of its current mass $`m_i`$ and string tension $`\sigma `$. To this end one connects the proper and real times via
$$ds_i=\frac{dt}{2\nu _i(t)},$$
(4)
where $`t=z_4^{(i)}(s_i)`$ is a common c.m. time on the hypersurface $`t=const`$ . The new entity, $`\nu _i(t)`$, being determined from the condition of the minimum of the corresponding Hamiltonian (see below) plays the role of the quark constituent mass (see for details). Referring to for the derivation of the Hamiltonian from the above defined Green’s function $`G^{(3q)}`$ we write down the final expression containing $`\nu _i`$ as parameters
$$H=\underset{k=1}{\overset{3}{}}\left(\frac{m_k^2}{2\nu _k}+\frac{\nu _k}{2}\right)+\frac{1}{2m}\left(\frac{^2}{\xi ^2}\frac{^2}{\eta ^2}\right)+\sigma \underset{k=1}{\overset{3}{}}|𝐫^{(k)}|,$$
(5)
where $`𝝃`$ and $`𝜼`$ are three-body Jacoby coordinates defined as in , $`m`$ is an arbitrary mass parameter which ensures correct dimensions and drops out of final expressions, and $`|𝐫^{(k)}|`$ is the distance from the $`k`$-sh quark to the string-junction positions which was for simplicity taken coinciding with the c.m. point. It is at this point worth stressing that the QCD string model outlined above is a fully relativistic string model for light current masses, and the ”nonrelativistic” appearance of the Hamiltonian (5) is a consequence of the rigorous einbein formalism .
Tooled with the Hamiltonian (5) one can use the standard hyperspherical formalism , introduce hyperradius $`\rho ^2=\xi ^2+\eta ^2`$, and write the following eigenwalue equation (quarks with equal masses are considered)
$$\frac{d^2\chi }{d\rho ^2}+2\nu \{E_nW(\rho )\}\chi (\rho )=0,$$
(6)
$$W(\rho )=b\rho +\frac{d}{2\nu \rho ^2},b=\sigma \sqrt{\frac{2}{3}}\frac{32}{5\pi },d=15/4.$$
(7)
The baryon mass $`M_n(\nu )`$ is equal to
$$M_n(\nu )=\frac{3m^2}{2\nu }+\frac{3}{2}\nu +E_n(\nu ).$$
(8)
According to the QCD string prescription - the value of $`\nu `$ is determined as a stationary point of $`M_n(\nu )`$:
$$\frac{M_n}{\nu }=0.$$
(9)
In passing from (4) to (9) we have changed from $`\nu (t)`$ depending on the trajectory to the operator $`\nu `$ and finally to the constant $`\nu `$ to be found from the minimum condition (9) – see - for details.
The perturbative gluon exchanges and spin–dependent terms can be selfconsistently included into the above picture. Including the Coulomb term and passing to dimensionless quantities $`x,\epsilon _n`$ and $`\lambda `$ defined as
$$x=(2\nu b)^{1/3}\rho ,\epsilon _n=\frac{2\nu E_n}{(2\nu b)^{2/3}},\lambda =\alpha _s=\frac{8}{3}\left(\frac{10\sqrt{3}\nu ^2}{\pi ^2\sigma }\right)^{1/3}\stackrel{~}{\lambda }\left(\frac{\nu ^2}{\sigma }\right)^{1/3},$$
(10)
where $`\alpha _s`$ is the strong coupling constant, one arrives at the following reduced equation
$$\left\{\frac{d^2}{dx^2}+x+\frac{d}{x^2}\frac{\lambda }{x}\epsilon _n(\lambda )\right\}=\chi (x)=0.$$
(11)
Then (9) yields the equation defining the quark dynamical mass $`\nu `$
$$\epsilon _n(\lambda )\left(\frac{\sigma }{\nu ^2}\right)^{2/3}\left\{1+\frac{2\lambda }{\epsilon _n(\lambda )}\left|\frac{d\epsilon _n}{d\lambda }\right|\right\}+\frac{9}{16}\left(\frac{75\pi ^2}{2}\right)^{1/3}\left(\frac{m^2}{\nu ^2}1\right)=0.$$
(12)
At this point the essential difference of $`\mathrm{\Delta }^{++}`$ from $`p`$ or $`n`$ arises. It concerns the interplay of the Coulomb and spin-spin interaction (the later is not yet included into (11) and (12)). The spin-spin interaction in baryon made of equal mass quarks results in the shift of $`E_n`$ equal to
$$\delta E_n=\frac{16}{9}\frac{\alpha _s}{\nu ^2}\underset{i>j}{}𝐬_i𝐬_j\delta (𝐫_{ij}).$$
(13)
For $`\mathrm{\Delta }^{++}`$ summation over ($`i,j)`$ yields a factor 3/4 corresponding to positive interaction energy , i.e. in $`\mathrm{\Delta }^{++}`$ spin-spin interaction acts coherently with the Coulomb one. In nucleon the corresponding factor is –4/3 resulting in reverse situation. This distintion realizes in the fact that the masses of quarks forming the nucleon remain ”unrenormalized” due to Coulomb and spin-spin interaction while the masses of quarks forming $`\mathrm{\Delta }^{++}`$ substantially increase as shown below.
To include the term (13) into equation (12) one has to smear the delta functions over small regions. This procedure has been done by two independent methods in with the result $`\delta (𝐫_{ij})=\delta _0\nu ^{3/2},\delta _0=2.6410^2`$ GeV<sup>3/2</sup>. With spin-spin interaction included Eq. (12) for $`\mathrm{\Delta }^{++}`$ takes the form
$$\nu ^2\frac{16}{9}\left(\frac{2}{75\pi ^2}\right)^{1/3}\epsilon (0)\sigma ^{2/3}\nu ^{2/3}\left[1+\frac{\stackrel{~}{\lambda }}{\epsilon (0)}\left|\frac{d\epsilon }{d\lambda }\right|_{\lambda =0}\left(\frac{\nu ^2}{\sigma }\right)^{1/3}\right]\frac{4\pi }{9}\alpha _s\delta _0\nu ^{1/3}=0.$$
(14)
Only terms linear in Coulomb coupling constant $`\stackrel{~}{\lambda }`$ are kept in (14), the corresponding expansion parameter is $`\stackrel{~}{\lambda }/\epsilon (0)1/4`$. Eq(14) has been solved for the following set of parameters
$$\sigma =0.15\mathrm{GeV}^2,\alpha _s=0.39.$$
(15)
The string tension value (15) which is smaller than in meson case is in line with baryon calculations by Capstick and Isgur . A similar smaller value of $`\sigma `$ is implied by recent lattice calculations by Bali . Solving (14) with the above set of parameters one gets
$$\nu _\mathrm{\Delta }=0.43\mathrm{GeV},$$
(16)
which should be compared to $`\nu _N=c\sqrt{\sigma }=0.37`$ GeV, where $`C=0.957`$ is a constant calculated in .
Now we turn directly to $`\mathrm{\Delta }^{++}`$ magnetic moment. The magnetic momet interaction term is included into $`G^{(3q)}`$ in a straightforward way - resulting in
$$\mu _{\mathrm{\Delta }^{++}}=3\psi _{\mathrm{\Delta }^{++}}\left|\frac{e_3\sigma _z^{(3)}}{2\nu _{\mathrm{\Delta }^{++}}}\right|\psi _{\mathrm{\Delta }^{++}},$$
(17)
where $`e_3=e_u=2/3`$. The structure of this matrix element with $`\nu _{\mathrm{\Delta }^{++}}`$ in the denomitor may be considered as an additional evidence that the quantity $`\nu `$ first introduced by (4) has the physical meaning of the quark constituent mass.
The calculation of the matrix element (17) is trivial provident one considers only totally symmetric coordinate wave functions, i.e. the lowest hyperspherical harmonic. It is known that the contribution of higher harmonics into normalization does not exceed few percent . Then (17) yields
$$\mu _{\mathrm{\Delta }^{++}}=\frac{2m_p}{\nu _\mathrm{\Delta }}4.36\mu _N,$$
(18)
which is in good agreement with the experimental value $`\mu _{\mathrm{\Delta }^{++}}=(4.52\pm 0.50\pm 0.45)\mu _N`$ .
The treatment of the $`\mathrm{\Delta }^{++}`$ baryon presented above makes it possible to reexamine the case of $`\mathrm{\Omega }^{}`$ baryon considered in and to improve upon the value of the strange quark current mass used in . Namely in the constituent mass $`\nu _\mathrm{\Omega }`$ was calculated using Eq.(12) above and then m.m. of $`\mathrm{\Omega }^{}`$ was obtained as a matrix element similar to (17). Coulomb and spin-spin terms in $`\mathrm{\Omega }^{}`$ were neglected since their contribution is smaller than that of the strange quark constituent mass $`m_s`$. If following the lines outlined above these hitherto omitted terms are included into the treatment of the $`\mathrm{\Omega }^{}`$ m.m. all the values of the baryon m.m. presented in (see also Table 1 below) remain unchanged but they are reproduced with smaller values of $`m_s`$, namely with $`m_s=0.15`$ GeV instead of $`m_s^{}=0.245`$ GeV used in . The new value of $`m_s`$ agrees with the result $`m_s=0.175\pm 0.025`$ GeV deduced by Leutwyler .
In Table 1 we summarize the results on baryon m.m. obtained in the present work and in . The typical deviations from the experimental values are about 10% which is remarkably successful keeping in mind plentiful possible corrections (meson exchanges, higher harmonics, etc).
The author is especially indebted to Yu.A.Simonov for formulating the problem, discussions and remarks. The author is grateful to Yu.S.Kalashnikova for numerous enlighting discussions and suggestions and to A.M.Badalian and N.O.Agasian for useful remarks.
The financial support of the grant RFFI 00-02-17836 and RFFI 96-15-96740 are gratefully acknowledged.
Table 1. Magnetic moments of baryons (in nuclear magnetons) in comparison with experimental data from PDG
| Baryon | $`p`$ | $`n`$ | $`\mathrm{\Delta }^{++}`$ | $`\mathrm{\Lambda }`$ | $`\mathrm{\Sigma }^{}`$ | $`\mathrm{\Sigma }^0`$ | $`\mathrm{\Sigma }^+`$ | $`\mathrm{\Xi }^{}`$ | $`\mathrm{\Xi }^0`$ | $`\mathrm{\Omega }^{}`$ |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Ref. and | | | | | | | | | | |
| present work | 2.54 | -1.69 | 4.36 | -0.69 | -0.90 | 0.80 | 2.48 | -0.63 | -1.49 | -2.04 |
| Experiment | 2.79 | -1.91 | 4.52 | -0.61 | -1.16 | | 2.46 | -0.65 | -1.25 | -2.02 |
|
warning/0003/hep-th0003156.html
|
ar5iv
|
text
|
# Duality, Equivalence, Mass and The Quest For The Vacuum aafootnote aInvited talk presented at PASCOS 99, Lake Tahoe, CA December 10–16 1999.
## Acknowledgments
I would like to thank Gaetano Bertoldi, Jerry Cleaver, Dimitri Nanopoulos, Joe Walker and especially Marco Matone, for collaboration on part of the work reported in this paper, and Tonnis ter Veldhuis for comments on the manuscript. This work is supported in part by DOE grant No. DE–FG–0287ER40328.
## References
|
warning/0003/physics0003093.html
|
ar5iv
|
text
|
# Untitled Document
Hierarchic Theory of Condensed Matter:
Role of water in protein dynamics, function & cancer emergency
Alex Kaivarainen
JBL, University of Turku, FIN-20520, Turku, Finland
http://www.karelia.ru/~alexk
H2o@karelia.ru
> Materials, presented in this original article are based on following publications:
>
> . A. Kaivarainen. Book: Hierarchic Concept of Matter and Field. Water, biosystems and elementary particles. New York, NY, 1995, ISBN 0-9642557-0-7
>
> . A. Kaivarainen. New Hierarchic Theory of Matter General for Liquids and Solids: dynamics, thermodynamics and mesoscopic structure of water and ice (see: http://www.karelia.ru/~alexk \[New articles\] ).
>
> . A. Kaivarainen. Hierarchic Concept of Condensed Matter and its Interaction with Light: New Theories of Light Refraction, Brillouin Scattering and Mössbauer effect (see: http://www.karelia.ru/~alexk \[New articles\]).
>
> . A. Kaivarainen. Hierarchic Concept of Condensed Matter : Interrelation between mesoscopic and macroscopic properties (see: http://www.karelia.ru/~alexk \[New articles\]).
>
> . A. Kaivarainen. Hierarchic Theory of Complex Systems (see URL: http://www.karelia.ru/~alexk \[New articles\]).
>
> See also papers at Los Alamos archives at: http://arXiv.org/find/physics/1/au:+Kaivarainen\_A/0/1/0/all/0/1
>
> CONTENTS OF ARTICLE:
Introduction to new Hierarchic theory of condensed matter
1. Role of inter-domain water clusters in large-scale dynamics of proteins
2. Description of large-scale dynamics of proteins, based on generalized. Stokes-Einstein and Eyring-Polany equation
3. Dynamic model of protein-ligand complexes formation
4. The life-time of quasiparticles and frequencies of their excitation
5. Mesoscopic mechanism of enzyme catalysis
6. The mechanism of ATP hydrolysis energy utilization in muscle contraction and protein polymerization
7. Water activity as a regulative factor in the intra- and inter-cell processes
8. Water and cancer
> Computerized verification of described here new models and theories has been presented, using special computer program, based on our new Hierarchic Theory of Condensed Matter (copyright, 1997, A. Kaivarainen).
=====================================================================
Introduction to new
Hierarchic theory of condensed matter (http://arXiv.org/abs/physics/00030044)
A basically new hierarchic quantitative theory, general for solids and liquids, has been developed.
It is assumed, that unharmonic oscillations of particles in any condensed matter lead to emergence of three-dimensional (3D) superposition of standing de Broglie waves of molecules, electromagnetic and acoustic waves. Consequently, any condensed matter could be considered as a gas of 3D standing waves of corresponding nature. Our approach unifies and develops strongly the Einstein’s and Debye’s models.
Collective excitations, like 3D standing de Broglie waves of molecules, representing at certain conditions the mesoscopic molecular Bose condensate, were analyzed, as a background of hierarchic model of condensed matter.
The most probable de Broglie wave (wave B) length is determined by the ratio of Plank constant to the most probable impulse of molecules, or by ratio of its most probable phase velocity to frequency. The waves B are related to molecular translations (tr) and librations (lb).
As the quantum dynamics of condensed matter does not follow in general case the classical Maxwell-Boltzmann distribution, the real most probable de Broglie wave length can exceed the classical thermal de Broglie wave length and the distance between centers of molecules many times.
This makes possible the atomic and molecular Bose condensation in solids and liquids at temperatures, below boiling point. It is one of the most important results of new theory, which we have confirmed by computer simulations on examples of water and ice.
Four strongly interrelated new types of quasiparticles (collective excitations) were introduced in our hierarchic model:
1. Effectons (tr and lb), existing in ”acoustic” (a) and ”optic” (b) states represent the coherent clusters in general case;
2. Convertons, corresponding to interconversions between tr and lb types of the effectons (flickering clusters);
3. Transitons are the intermediate $`\left[ab\right]`$ transition states of the tr and lb effectons;
4. Deformons are the 3D superposition of IR electromagnetic or acoustic waves, activated by transitons and convertons.
Primary effectons (tr and lb) are formed by 3D superposition of the most probable standing de Broglie waves of the oscillating ions, atoms or molecules. The volume of effectons (tr and lb) may contain from less than one, to tens and even thousands of molecules. The first condition means validity of classical approximation in description of the subsystems of the effectons. The second one points to quantum properties of coherent clusters due to molecular Bose condensation.
The liquids are semiclassical systems because their primary (tr) effectons contain less than one molecule and primary (lb) effectons - more than one molecule. The solids are quantum systems totally because both kind of their primary effectons (tr and lb) are molecular Bose condensates. These consequences of our theory are confirmed by computer calculations.
The 1st order $`\left[gasliquid\right]`$ transition is accompanied by strong decreasing of rotational (librational) degrees of freedom due to emergence of primary (lb) effectons and $`\left[liquidsolid\right]`$ transition - by decreasing of translational degrees of freedom due to Bose-condensation of primary (tr) effectons.
In the general case the effecton can be approximated by parallelepiped with edges corresponding to de Broglie waves length in three selected directions (1, 2, 3), related to the symmetry of the molecular dynamics. In the case of isotropic molecular motion the effectons’ shape may be approximated by cube.
The edge-length of primary effectons (tr and lb) can be considered as the ”parameter of order”.
The in-phase oscillations of molecules in the effectons correspond to the effecton’s (a) - acoustic state and the counterphase oscillations correspond to their (b) - optic state. States (a) and (b) of the effectons differ in potential energy only, however, their kinetic energies, impulses and spatial dimensions - are the same. The b-state of the effectons has a common feature with Frölich’s polar mode.
The $`(ab)`$ or $`(ba)`$ transition states of the primary effectons (tr and lb), defined as primary transitons, are accompanied by a change in molecule polarizability and dipole moment without density fluctuations. At this case they lead to absorption or radiation of IR photons, respectively.
Superposition (interception) of three internal standing IR photons of different directions (1,2,3) - forms primary electromagnetic deformons (tr and lb).
On the other hand, the \[lb$``$tr\] convertons and secondary transitons are accompanied by the density fluctuations, leading to absorption or radiation of phonons.
Superposition resulting from interception of standing phonons in three directions (1,2,3), forms secondary acoustic deformons (tr and lb).
Correlated collective excitations of primary and secondary effectons and deformons (tr and lb), localized in the volume of primary tr and lb electromagnetic deformons, lead to origination of macroeffectons, macrotransitons and macrodeformons (tr and lb respectively).
Correlated simultaneous excitations of tr and lb macroeffectons in the volume of superimposed tr and lb electromagnetic deformons lead to origination of supereffectons.
In turn, the coherent excitation of both: tr and lb macrodeformons and macroconvertons in the same volume means creation of superdeformons. Superdeformons are the biggest (cavitational) fluctuations, leading to microbubbles in liquids and to local defects in solids.
Total number of quasiparticles of condensed matter equal to 4!=24, reflects all of possible combinations of the four basic ones \[1-4\], introduced above. This set of collective excitations in the form of ”gas” of 3D standing waves of three types: de Broglie, acoustic and electromagnetic - is shown to be able to explain virtually all the properties of all condensed matter.
The important positive feature of our hierarchic model of matter is that it does not need the semi-empiric intermolecular potentials for calculations, which are unavoidable in existing theories of many body systems. The potential energy of intermolecular interaction is involved indirectly in dimensions and stability of quasiparticles, introduced in our model.
The main formulae of theory are the same for liquids and solids and include following experimental parameters, which take into account their different properties:
$`\left[1\right]`$\- Positions of (tr) and (lb) bands in oscillatory spectra;
$`\left[2\right]`$\- Sound velocity;
$`\left[3\right]`$\- Density;
$`\left[4\right]`$\- Refraction index (extrapolated to the infinitive wave length of photon$`)`$.
The knowledge of these four basic parameters at the same temperature and pressure makes it possible using our computer program, to evaluate more than 300 important characteristics of any condensed matter. Among them are such as: total internal energy, kinetic and potential energies, heat-capacity and thermal conductivity, surface tension, vapor pressure, viscosity, coefficient of self-diffusion, osmotic pressure, solvent activity, etc. Most of calculated parameters are hidden, i.e. inaccessible to direct experimental measurement.
The new interpretation and evaluation of Brillouin light scattering and Mössbauer effect parameters may also be done on the basis of hierarchic theory. Mesoscopic scenarios of turbulence, superconductivity and superfluity are elaborated.
Some original aspects of water in organization and large-scale dynamics of biosystems - such as proteins, DNA, microtubules, membranes and regulative role of water in cytoplasm, cancer development, quantum neurodynamics, etc. have been analyzed in the framework of Hierarchic theory.
Computerized verification of our Hierarchic concept of matter on examples of water and ice is performed, using special computer program: Comprehensive Analyzer of Matter Properties (CAMP, copyright, 1997, Kaivarainen). The new optoacoustic device, based on this program, with possibilities much wider, than that of IR, Raman and Brillouin spectrometers, has been proposed ( http://www.karelia.ru/~alexk \[CAMP\]).
This is the first theory able to predict all known experimental temperature anomalies for water and ice. The conformity between theory and experiment is very good even without any adjustable parameters.
The hierarchic concept creates a bridge between micro- and macro- phenomena, dynamics and thermodynamics, liquids and solids in terms of quantum physics.
===============================================================
1. Role of inter-domain water clusters in large-scale dynamics of proteins
The functioning of proteins, namely antibodies, enzymes, is caused by the physicochemical properties, geometry and dynamics of their active sites. The mobility of an active site is related to the dynamics of the residual part of a protein molecule, its hydration shell and the properties of a free solvent.
The dynamic model of a protein proposed in 1975 and supported nowadays with numerous data (Käiväräinen, $`1985,1989b)`$, is based on the following statements:
1. A protein molecule contains one or more cavities or clefts capable to large scale fluctuations - pulsations between two states: ”closed” (A) and ”open” $`(`$B) with lesser and bigger accessibility to water.
The frequency of pulsations $`\left(\nu _{AB}\right)`$:
$$10^4s^1\nu _{AB}10^7s^1$$
depends on the structure of protein, its ligand state, temperature and solvent viscosity. Transitions between A and B states are the result of the relative displacements of protein domains and subunits forming the cavities;
2. The water, interacting with protein, consists of two main fractions.
The 1st major fraction, which solvates the outer surface regions of protein has less apparent cooperative properties than the 2nd minor fraction confined to ”open” cavities. Water molecules, interacting with the cavity in the ”open” (B)-state form a cooperative cluster, whose lifetime $`(10^{10}s^1).`$Properties of clusters are determined by the geometry, mobility and polarity of the cavity, as well as by temperature and pressure.
It is seen from X-ray structural data that the protein cavities: active sites (AS), other interdomain clefts, the space between subunits of oligomeric proteins, have a high nonpolar residues content. In contrast to the small intra domain holes isolated from the outer medium, which sometimes contain several $`H_2O`$ molecules, the interdomain and intersubunit cavities can contain several dozens of molecules (Fig. 1), exchanging with bulk water.
The development of the above dynamic model has lead us to the following classification of dynamics in the native globular proteins.
1. Small-scale (SS) dynamics:
low amplitude ($``$ 1 Å) thermal fluctuations of atoms, aminoacids residues, and displacements of $`\alpha `$-helixes and $`\beta `$-structures within domains and subunits, at which the effective Stokes radius of domains does not change. This type of motion, related to domain stability, can differ in the content of A and B conformers (Fig. 1, dashed line). The range of characteristic times at SS dynamics is $`(10^410^{11})s`$, determined by activation energies of corresponding transitions.
2. Large-scale (LS) dynamics:
is subdivided into LS-pulsations and LS-librations with a character of limited diffusion of domains and subunits of proteins:
LS- pulsations are represented by relative translational-rotational displacements of domains and subunits at distances $`3\AA `$. Thus, the cavities, which are formed by domains, fluctuate between states with less (A) and more (B) water-accessibility. The life-times of these states depending on protein structure and external conditions are in the limits of$`\left(10^410^7\right)`$ s.
In accordance to our model, one of contributions to this time is determined by frequency of excitations of $`[lb/tr]`$ macroconvertons. The frequency of macroconvertons excitation at normal conditions is about $`10^7(1/s)`$.
The pulsation frequency of big multi-subunit oligomeric proteins of about 10$`{}_{}{}^{4}(1/s)`$ could be related to stronger fluctuations of water cluster in their central cavity like macrodeformons or even superdeformons (Fig.3c,d).
The life-times of (A) and (B) conformer markedly exceeds the time of transitions between them $`(10^910^{11})`$ s.
The $`\left(AB\right)`$ pulsations of various cavities in proteins could be correlated. The corresponding A and B conformers have different Stokes radii and effective volume.
The geometrical deformation of the inter-subunits large central cavity of oligomeric proteins and the destabilization of the water cluster located in it lead to relaxational change of $`(AB)`$ equilibrium constant:
$$K_{AB}=\mathrm{exp}\left(\frac{G_AG_B}{RT}\right).$$
The dashed line means that the stability and the small-scale dynamics of domains and subunits in the content of A and B conformers can differ from each other. The $`\left[AB\right]`$ pulsations are accompanied by reversible sorption-desorption of $`\left(2050\right)H_2O`$ molecules from the cavities.
Structural domains are space-separated formations with a mass of $`(12)10^3`$ D. Protein subunits $`(MM210^3D)`$, as a rule, consist of 2 or more domains. The domains can consist only of $`\alpha `$ or only of $`\beta `$-structure or have no like secondary structure at all (Schulz, Schirmer, 1979).
The shift of $`AB`$ equilibrium of central cavity of oligomeric proteins determines their cooperative properties during consecutive ligand binding in the active sites. Signal transmission from the active sites to the remote regions of macromolecules is also dependent on $`\left(AB\right)`$equilibrium.
> Fig. 1. Examples of large-scale (LS) protein dynamics: $`AB`$ pulsations and librations with correlation times $`(\tau _{\text{lb}}^B<\tau _{\text{lb}}^A)(`$Käiväräinen, 1985, 1989):
>
> a) mobility of domains connected by flexible hinge or contact region, like in the light chains of immunoglobulins;
>
> b) mobility of domains that form the active sites of proteins, like in hexokinase, papain, pepsin, lysozyme etc. due to flexibility of contacts;
>
> c) mobility of subunits forming the oligomeric proteins like hemoglobin. Besides transitions of the active sites of each subunit, the $`\left(AB\right)`$ pulsations with frequencies of $`(10^410^6)`$ $`s^1`$ are pertinent to the common central cavity.
>
> b) librations represent the relative rotational - translational motions of domains and subunits in composition of A and B conformers with correlation times $`\tau _M(15)10^8`$s.
LS - librations of domains are accompanied by ”flickering” of water cluster in the open cavity between domains or subunits. The process of water cluster ”flickering”, i.e. \[dissociation $``$association\] is close to the reversible first-order phase transition, when:
$$\mathrm{\Delta }G_{H_2O}=\mathrm{\Delta }H_{H_2O}T\mathrm{\Delta }S_{H_2O}0$$
Such type of transitions in water-macromolecular systems could be responsible for so called ”enthalpy-entropy compensation effects” (Lumry and Biltonen, 1969).
The ”flickering clusters” means excitation of \[$`lb/tr]`$conversions between librational and translational primary water effectons, accompanied by \[association/dissociation\] of coherent water cluster (see difference in dimensions of lb and tr effectons on Fig. $`18a,b`$of $`)`$.
The water cluster (primary lb effecton) association and dissociation in protein cavities in terms of mesoscopic model represent the $`(ac)`$ \- convertons or $`\left(\mathrm{𝑏𝑐}\right)`$ \- convertons. These excitations stimulate the LS- librations of domains in composition of B-conformer. The frequencies of (ac) and (bc) convertons, has the order of about $`10^8c^1`$. This value coincides well with experimental characteristic times for protein domains librations.
The (ac) and (bc) convertons represent transitions between similar states of primary librational and translational effectons: $`[a_{lb}a_{tr}]`$ and $`[b_{lb}b_{tr}]`$ (see Introduction to).
For the other hand, the Macroconvertons, representing simultaneous excitation of $`\left(ac+bc\right)`$ convertons, are responsible for $`\left[BA\right]`$large-scale pulsations of proteins.
The librational mobility of domains and subunits is revealed by the fact that the experimental value of $`\tau _M`$ is less than the theoretical one ($`\tau _M^t`$) calculated on the Stokes-Einstein formula:
$$\tau _M^t=(V/k)\eta /T$$
This formula is based on the assumption that the whole protein can be approximated by a rigid sphere. It means, that the large-scale dynamics can be characterized by the ”flexibility factor”, in the absence of aggregation equal to ratio:
$$fl=(\tau _M/\tau _M^t)1$$
Antonchenko (1986) has demonstrated, using the Monte-Carlo method for simulations, that the disjoining pressure of a liquid in the pores onto the walls changes periodically depending on the distance (L) between the limiting surfaces. If the water molecules are approximated by rigid globes, then the maxima of the wedging pressure lie on the values of distance L$`:9.8;7;`$ and $`3.3\AA `$. It points, that small changes in the geometry of cavities can lead to significant changes in their $`AB`$ equilibrium constant $`(K_{AB})`$.
According to our model the large-scale transition of the protein cavity from the ”open” B-state to the ”closed” A-state consists of the following stages:
1. Small reorientation (libration) of domains or subunits, which form an ”open” cavity (B-state). This process is induced by $`\left(ac\right)`$ or $`\left(bc\right)`$ convertons of water librational effecton, localized in cavity (flickering of water cluster);
2. Cavitational fluctuation of water cluster, containing ($`2050)H_2O`$ molecules and the destabilization of the B-state of cavity as a result of \[lb$``$tr\] macroconverton excitation;
3. Collapsing of a cavity during the time about $`10^{10}s`$, dependent on previous stage and concomitant rapid structural change in the hinge region of interdomain and intersubunit contacts: $`\left[BA\right]`$ transition.
The $`ba`$ transition of one of the protein cavities can be followed by similar or the opposite $`AB`$ transition of the other cavity in the macromolecule.
It should be noted that the collapsing time of a cavitation bubble with the radius: $`r(1015)\AA `$ in bulk water and collapsing time of interdomain cavity are of the same order: $`\mathrm{\Delta }t\mathrm{~}10^{10}s`$ under normal conditions$`(`$Shutilov, 1980).
If configurational changes of macromolecules at $`BA`$ and $`AB`$ transitions are sufficiently quick and occur as a jumps of the effective volume, they accompanied by appearance of the shock acoustic waves in the bulk medium.
When the cavitational fluctuation of water in the ”open” cavity does not occur, then $`(ba)`$ or $`(BA)`$ transitions are slower processes, determined by continuous diffusion of domains and subunits. This happens when $`[lbtr]`$ macroconvertons are not excited.
In their review, Karplus and McCammon (1986) analyzed data on alcoholdehydrogenase, myoglobin and ribonuclease, which have been obtained using molecular dynamics approach. It has been shown that large-scale reorientation of domains occur together with their deformation and motions of $`\alpha `$ and $`\beta `$ structures.
It has been shown also (Karplus and McCammon, 1986) that activation free energies, necessary for $`\left[AB\right]`$ transitions and the reorganization of hinge region between domains, do not exceed (3-4) kcal/mole. Such low values were obtained for proteins with even rather dense interdomain region, as seen from X-ray data. The authors explain such low values of activation energy by the fact that the displacement of atoms, necessary for such transition, does not exceed 0.5 Å, i.e. they are comparable with the usual amplitudes of atomic oscillation at temperatures $`2030^0`$C. It means that they occur very quickly within times of $`10^{12}s`$, i.e. much less than the times of $`\left[AB\right]`$ or $`\left[BA\right]`$ domain displacements $`(10^910^{10}s)`$. Therefore, the high frequency small-scale dynamics of hinge is responsible for the quick adaptation of hinge geometry to the changing distance between the domains and for decreasing the total activation energy of $`\left[AB\right]`$ pulsations of proteins,.
Recent calculations by means of molecular dynamics reveal that the oscillations in proteins are harmonic at the low temperature (T$`<220K)`$ only. At the physiological temperatures the oscillations are strongly unharmonic, collective, global and their amplitude increases with hydration (Steinback et al., 1996). Water is a ”catalyzer” of protein unharmonic dynamics.
It is obvious, that both small-scale (SS) and large-scale (LS) dynamics, introduced in our model, are necessary for protein function. To characterize quantitatively the LS dynamics of proteins, we proposed the unified Stokes-Einstein and Eyring-Polany equation.
2. Description of large-scale dynamics of proteins, based on generalized Stokes-Einstein and Eyring-Polany equation
In the case of the continuous Brownian diffusion of a particle, the rate constant of diffusion is determined by the Stokes-Einstein law:
$$k=\frac{1}{\tau }=\frac{k_BT}{V\eta }$$
(1)
where: $`\tau `$ is correlation time, i.e. the time, necessary for rotation of a particle by the mean angle determined as $`\overline{\phi }0.5`$ of the turn or the characteristic time for the translational movement of a particle with the radius (a) on the distance $`(\overline{\mathrm{\Delta }}_x)^{1/2}0.6a(`$Einstein, 1965);
$`V=4\pi a^3/3`$ is the volume of the spherical particle; $`k_B`$ is the Boltzmann constant, T and $`\eta `$ are the absolute temperature and bulk viscosity of the solvent.
On the other hand, the rate constant of$`\left[AB\right]`$ reaction for a molecule in gas phase, which is related to passing through the activation barrier $`G^{AB}`$, is described with the Eyring-Polany equation:
$$k^{AB}=\frac{kT_B}{h}\mathrm{exp}\left(\frac{G^{AB}}{RT}\right)$$
(2)
To describe the large-scale dynamics of macromolecules in solution related to fluctuations of domains and subunits (librations and pulsations), an equation is needed which takes into account the diffusion and activation processes simultaneously.
The rate constant for the rotational- translational diffusion of the particle $`(k_c)`$ forming a macromolecule (continuous LS-dynamics) is expressed with the generalized Stokes-Einstein and Eyring-Polany equation (Käiväräinen and Goryunov, 1987):
$$K_c=\frac{k_BT}{\eta V}\mathrm{exp}\left(\frac{G_{st}}{RT}\right)=\tau _c^1$$
(3)
where: V is the effective volume of domain or subunit, which are capable to the Brownian mobility independently from the rest part of the macromolecule, with the probability:
$$P_{lb}=\mathrm{exp}\left(\frac{G_{st}}{RT}\right),$$
(4)
where: G<sub>st</sub> is the activation energy of structural change in the contact (hinge) region of a macromolecule, necessary for independent mobility of domain or subunit; $`\tau _c`$ is the effective correlation time for the continuous diffusion of this relatively independent particle.
The effective volume V can be changed under the influence of temperature, perturbants and ligands.
The generalized Stokes-Einstein and Eyring-Polany equation (3) is applicable also to describing the diffusion of the whole (integer) particle, dependent on the surrounding medium fluctuations with activation energy $`(G_a)`$. The ligand diffusion in the active site cavity of proteins is such a type of processes.
To describe noncontinuous process, the formula for rate constant $`(k_{\text{jump}})`$ of the jump-like translations of particle, related to emergency of cavitational fluctuations (holes) near the particle was proposed (Käiväräinen and Goryunov, 1987):
$$k_{\text{jump}}=\frac{1}{\tau _{\text{jump}}^{\mathrm{min}}}\mathrm{exp}\left(\frac{W}{RT}\right)=\frac{1}{\tau _{\text{jump}}},$$
(5)
where:
$$W=\sigma S+n_s(\mu _{\text{out}}\mu _{\text{in}})$$
(6)
is the work of cavitation fluctuation with the cavity surface S, at which $`n_s`$ molecules of the solvent (water) change its effective chemical potential from $`\mu _{\text{in}}`$ to $`\mu _{\text{out}}`$.
The dimensions of cavity fluctuation near particle must be comparable to corresponding particles.
In a homogeneous phase (i.e. pure water) under equilibrium conditions we have: $`\mu _{\text{in}}=\mu _{\text{out}}`$. With an increase of particle sizes, surface of cavitational fluctuation (S) and its work (W), the corresponding probability of cavitation fluctuations:
$$P_{\text{jump}}=\mathrm{exp}(W/RT)$$
will fall.
The notion of the surface energy ($`\sigma `$) retains its meaning even at very small ”holes” because of its molecular nature (see Section 11.4 of and ).
$`\tau _{\text{jump}}^{\mathrm{min}}`$ in eq. (5) is the minimal possible jump-time of a particle with mass (m) over the distance $`\lambda `$ with the mean velocity:
$$v_{\mathrm{max}}=(2kT/m)^{1/2}$$
(7)
Hence, we derive for the maximal jump-rate at W=0:
$$k_{\text{jump}}^{\mathrm{max}}=\frac{1}{\tau _{\text{jump}}^{\mathrm{min}}}=\frac{V_{\mathrm{max}}}{\lambda }=\frac{1}{\lambda }\left(\frac{2kT}{m}\right)^{1/2}$$
(8)
In the case of hinged domains, forming macromolecules their relative $`AB`$ displacements (pulsations) are related not only to possible holes forming in the interdomain (intersubunit) cavities or near their outer surfaces, but to the structural change of hinge regions as well.
If the activation energy of necessary structure changes is equal to G$`{}_{}{}^{AB}{}_{st}{}^{}`$, then eq. (5), with regard for (8), is transformed into
$$k_{\text{jump}}^{AB}=\frac{1}{\lambda }\left(\frac{2kT}{m}\right)^{1/2}\mathrm{exp}\left(\frac{W_{A,B}+G_{st}^{AB}}{RT}\right)$$
(9)
where: W<sub>A,B</sub> is the work required for cavitational fluctuations of water; this work can be different in two directions: $`(W_B)`$ is necessary for nonmonotonic $`BA`$ transition; $`(W_A)`$ is necessary for jump-way $`AB`$ transition.
Under certain conditions $`AB`$ transitions between protein conformers (LS- pulsations) can be realized owing to the jump-way and continuous types of relative diffusion of domains or subunits as two stage reaction. In this case, the resulting rate constant of the process will be expressed through (9) and (3) as:
$$\begin{array}{c}k_{\text{res}}^{AB}=k_{\text{jump}}^{AB}=k_c^{AB}=\frac{1}{\lambda }\left(\frac{2kT}{m}\right)^{1/2}\mathrm{exp}\left(\frac{W_{A,B}+G_{st}^{AB}}{RT}\right)+\hfill \\ \\ +\frac{kT}{\lambda V}\mathrm{exp}\left(\frac{G_{st}^{AB}}{RT}\right)\hfill \end{array}$$
(10)
The interaction between two domains in A-conformer can be described using microscopic Hamaker - de Bour theory. One of the contributions into $`G_{st}^{AB}`$ is the energy of dispersion interactions between domains of the radius (a) (Käiväräinen, 1989b):
$$[G_{st}U_HA^{}a/12H]_{A,B}$$
(11)
where
$$A^{}(A_s^{1/2}A_c^{1/2})^2\frac{3}{2}\pi h\nu _0^s\left[\alpha _sN_s\alpha _cN_c\right]^2$$
(12)
is the complex Hamaker constant; H is the slit thickness between domains in A-state; $`A_c`$ and $`A_s`$ are simple Hamaker constants, characterizing the properties of water in the A-state of the cavity and in the bulk solvent, correspondingly. They depend on the concentration of water molecules $`(N_cN_s)`$ and their polarizability $`(\alpha _c\alpha _s)`$:
$$A_c=\frac{3}{2}\pi h\nu _0^c\alpha _c^2N_c^2\text{ and }A_s=\frac{3}{2}\pi h\nu _0^s\alpha _s^2N_s^2;$$
where: $`h\nu _0^ch\nu _0^s`$ are the ionization potentials of $`H_2O`$ molecules in a cavity and in a free solvent.
In the ”closed” A-state of a cavity the water layer between domains has a more compact packing as compared with the ice-like structure of a water cluster in the B-state of a cavity, or with a free solvent. As far $`H_A<H_B`$ the dispersion interaction (11) between domains in A- state of cavity is stronger, than that in B-state: $`U_H^A>U_H^B`$.
Disjoining pressure of water in the cavities
$$\mathrm{\Pi }=A^{}/6\pi H^3$$
(12a)
decreases with the increase of the complex Hamaker constant $`(A^{})`$ that corresponds to the increase of the attraction energy $`(U_H)`$ between domains.
Cooperative properties of clusters in open (B)-states of the cavities are more pronounced as compared to that in bulk water. That results in the greater changes of $`\alpha _cN_c`$ than that of $`\alpha _sN_s`$ induced by temperature. The elevation of temperature decreasing the dimensions of interdomain water clusters leads to the strengthening of interdomain interaction, while the lowering temperature leads to opposite effect.
We can judge about the changes of $`\alpha _s`$N$`_s`$ in the experiment on measuring the solvent refraction index, as far from our theory of refraction index (eq. 8.14 of or paper ):
$$(n_s^21)/n_s^2=\frac{4}{3}\pi \alpha _sN_s\text{ or: }\alpha _sN_s=\frac{3}{4}\pi \frac{n_s^21}{n_s^2}$$
(12b)
In the closed cavities the effect of temperature on water properties is lower as compared to that in bulk water. It follows that thermoinduced nonmonotonic transition in the solvent refraction index must be accompanied by in-phase nonmonotonic changes of the $`\left[𝐀𝐁\right]`$ equilibrium constant $`(K_{AB})`$. As far (A) and (B) conformers usually have different stability and flexibility, the changes of $`K_{AB}`$ will be manifested in the changes of protein large-scale and small-scale dynamics. It has been shown before that viscosity itself has nonmonotonic temperature dependence due to the nonmonotonic dependence of $`n^2(t)(eq\mathrm{.11.44},11.45`$ and 11.48 of or paper ).
Thus, thermoinduced non-denaturational transitions of macromolecules and supramolecular systems located in the aqueous environment are caused by nonmonotonic changes in solvent properties, including its refraction index.
The influence of $`D_2O`$ and other perturbants on protein dynamics is explained in a similar way. The effect of deuterium oxide ($`D_2O)`$ is a result of substitution of $`H_2O`$ from protein cavities and corresponding change of complex Hamaker constant (12).
Generalized equation (3) is applicable not only for evaluating the frequency of macromolecules transition between A and B conformers but also for the frequency of the dumped librations of domains and subunits within A and B conformers. Judging by various data (Käiväräinen, $`1985,1989b)`$, the interval of $`AB`$ pulsation frequency is:
$$\nu _{AB}=\frac{1}{t_A+t_B}k^{AB}=(10^410^7)\text{ }s^1$$
(13)
where: $`t_A`$ and $`t_B`$ are the lifetimes of A and B conformers.
The corresponding interval of the total activation energy of the jump-way $`AB`$ pulsations can be evaluated from the eq. (9). We assume for this end that the pre-exponential multiplier is about $`10^{10}s^1`$ as a frequency of cavitational fluctuations in water with the radius ~(10-15) $`\stackrel{𝑜}{A}`$.
Taking a logarithm of (9) we derive:
$$G_{\text{res}}^{AB}=(W_{A,B}+G_{st}^{AB})RT(\mathrm{ln}10^{10}\mathrm{ln}\nu _{AB})$$
(14)
At physiological temperatures the following region of energy corresponds to the frequency range of pulsations (13)
$$G_{\text{res}}^{AB}(48)\text{ kcal/mole}$$
(15)
Such a region of energies is pertinent to a wide range of biochemical processes.
The quick jump-way pulsations of macromolecules can cause acoustic shock- waves in the solvent and its structure destabilization. Concomitant increase in water activity leads to distant interaction between different proteins as well as proteins and cells. Such solvent-mediated phenomena were discovered and studied in our laboratory by set of specially elaborated methods (Käiväräinen, 1985, 1986, 1987; Käiväräinen et al., 1990, Käiväräinen et al., 1993).
When the $`\left[AB\right]`$ transitions in proteins are related to continuous diffusion only, then the $`G_{st}`$ values calculated using eq.(3) for the same frequency interval $`(10^410^7)`$ $`s`$, is about (3 - 7) kcal/mole.
The Kramers equation (1940), which has earlier been widely used for describing diffusion processes, has the form:
$$k=\frac{A}{\eta }\mathrm{exp}\left(\frac{H^{}}{RT}\right)$$
(16)
where A is a constant and $`\eta `$ \- solvent viscosity.
The pre-exponential factor in our generalized equation (3) contains not only the viscosity, but also the temperature and the effective volume of a particle. It was shown in our experiments that eq. (3) describes the dynamic processes, which occur in macromolecules, solutions much better than the Kramer’s equation (16).
3. Dynamic model of protein-ligand complexes formation
According to our model of specific complexes formation the following order of events is assumed (Fig. 2):
1. Ligand (L) collides with the active site (AS), formed usually by two domains, in its open (b) state: the structure of water cluster in AS is being perturbed and water is forced out of AS cavity totally or partially;
2. Transition of AS from the open (b) to the closed (a) state occurs due to strong shift of $`[ab]`$ equilibrium to the left, i.e. to the AS domains large scale dynamics;
3. A process of dynamic adaptation of complex \[L+AS\] begins, accompanied by the directed ligand diffusion in AS cavity due to its domains small-scale dynamics and deformation of their tertiary structure;
4. If the protein is oligomeric with few AS, then the above events cause changes in the geometry of the central cavity between subunits in the open state leading to the destabilization of the large central water cluster and the shift of the $`AB,`$ corresponding to $`RT`$ equilibrium of quaternary structure leftward. Water is partially forced out from central cavity.
Due to the feedback mechanism this shift can influence the $`\left[ab\right]`$ equilibrium of the remaining free AS and promotes its reaction with the next ligand. Every new ligand stimulates this process, promoting the positive cooperativity. The negative cooperativity also could be resulted from the interaction between central cavity and active sites;
5. The terminal $`\left[proteinligand\right]`$ complex is formed as a consequence of the relaxation process, representing deformation of domains and subunits tertiary structure. This stage could be much slower than the initial ones \[1-3\]. As a result of it, the stability of the complex grows up.
Dissociation of specific complex is a set of reverse processes to that described above which starts from the $`\left[a^{}b\right]`$ fluctuation of the AS cavity.
In multidomain proteins like antibodies, which consist of 12 domains, and in oligomeric proteins, the cooperative properties of $`H_2O`$ clusters in the cavities can determine the mechanism of signal transmission from AS to the remote effector regions and allosteric protein properties.
The stability of a librational water effecton as coherent cluster strongly depends on its sizes and geometry. This means that very small deformations of protein cavity, which violate the \[cavity-cluster\] complementary condition, induce a cooperative shift of $`\left[AB\right]`$ equilibrium leftward. The clusterphilic interaction, introduced by us (see section 13.3 of or paper ) turns to hydrophobic one due to $`[lb/tr]`$ conversion.
This process can be developed step by step. For example, the reorientation of variable domains, which form the antibodies active site (AS) after reaction with the antigen determinant or hapten deforms the next cavity between pairs of variable and constant domains forming F<sub>ab</sub> subunits (Fig.2). The leftward shift of $`\left[AB\right]`$ equilibrium of this cavity, in turn, changes the geometry of the big central cavity between $`F_{ab}`$ and $`F_c`$ subunits, perturbing the structure of the latter. Therefore, the signal transmission from the AS to the effector sites of $`F_c`$ subunits occurs due to the balance shift between clusterphilic and hydrophobic interactions. This signal is responsible for complement- binding sites activation and triggering the receptors function on the lymphocyte membranes, in accordance to our model.
The leftward shift of $`\left[AB\right]`$ equilibrium in a number of cavities in the elongated multidomain proteins can lead to the significant decrease of their linear size and dehydration. The mechanism of muscular contraction is probably based on such phenomena and clusterphilic interactions (see next section).
For such a nonlinear system the energy is necessary for reorientation of the first couple of domains only. The process then goes on spontaneously with decreasing the averaged protein chemical potential.
The chemical potential of the A- conformer is usually lower than that of B- conformer $`(\overline{G}_A<\overline{G}_B)`$ and the relaxation of protein is accompanied by the leftward $`AB`$ equilibrium shift of cavities.
It is predictable, that hydration of proteins will decrease, when clusterphilic \[water-cavity\] interaction turns to hydrophobic one.
> Fig. 2. The schematic picture of the protein association (Fab subunits of antibody with a ligand), which is accompanied by the destabilization of water clusters in cavities, according to the dynamic model (Käiväräinen, 1985). The dotted line denotes the perturbation of the tertiary structure of the domains forming the active site. Antibodies of IgG type contain usually two such Fab subunit and one Fc subunit, conjugated with 2Fab by flexible hinge, forming the general Y-like structure.
Our dynamic model of protein behavior and signal transmission, described above, is an alternative to solitonic mechanism of non dissipative signal transmission in proteins and in other biosystems proposed by Davidov (1973). Propagation of solitonic wave is a well known nonlinear process in the ordered homogeneous mediums. The solitons can originate, when the nonlinear effects are compensated by the wave dispersion effects. Dispersion is reflected in fact that the longer waves spreads in medium with higher velocity than the shorter ones.
However, biosystems of nonregular, fluctuating structure are not the mediums, good for solitons emergency and propagation.
Our dynamic model takes into account the real multidomain and multiglobular structure of a proteins and properties of their hydration shell fractions. In contrast to Davidov’s solitonic model, the dissipation processes like reversible ”melting” of water clusters, accompanied by large-scale dynamics of proteins, are the necessary stages of our \[hydrophobic $``$clusterphilic\] mechanism of signal transmission in biosystems.
The evolution of the ideas of the protein-ligand complex formation proceeded in the following sequence:
1. ”Key-lock” or the rigid conformity between the geometry of an active site and that of a ligand (Fisher, 1894);
2. ”Hand-glove” or the so-called principle of induced conformity (Koshland, 1962);
3. At the current stage of complex-formation process understanding, the crucial role of protein dynamics gets clearer. Our model allows us to put forward the ”Principle of Stabilized Conformity (PSC)” instead that of ”induced conformity” in protein-ligand specific reaction.
Principle of Stabilized Conformity (PSC) means that the geometry of the active site (AS), optimal from energetic and stereochemical conditions, is already existing BEFORE reaction with ligand. The optimal geometry of AS is to be the only one selected among the number of others and stabilized by ligand, but not induced ”de nova”.
For example, the $`[ab]`$ large-scale pulsations of the active sites due to domain fluctuations and stabilization of the closed (a) state by ligand are necessary for the initial stages of reaction. Such active site pulsations decreases the total activation energy necessary for the terminal complex formation as multistage process.
4. The life-time of quasiparticles and frequencies of their excitation
The set of formula, describing the dynamic properties of quasiparticles, introduced in mesoscopic theory was presented at Chapter 4 of book and paper :
The frequency of c- Macrotransitons or Macroconvertons excitation, representing \[dissociation/association\] of primary librational effectons \- ”flickering clustersas a result of interconversions between primary \[lb\] and \[tr\] effectons is:
$$F_{cM}=\frac{1}{\tau _{Mc}}P_{Mc}/Z$$
(17)
where: $`P_{Mc}=P_{ac}P_{bc}`$ is a probability of macroconverton excitation;
$`Z`$ is a total partition function (see eq.4.2 of );
the life-time of macroconverton is:
$$\tau _{Mc}=(\tau _{ac}\tau _{bc})^{1/2}$$
(18)
The cycle-period of (ac) and (bc) convertons are determined by the sum of life-times of intermediate states of primary translational and librational effectons:
$$\begin{array}{c}\tau _{ac}=(\tau _a)_{tr}+(\tau _a)_{lb};\\ \tau _{bc}=(\tau _b)_{tr}+(\tau _b)_{lb};\end{array}$$
(19)
The life-times of primary and secondary effectons (lb and tr) in a\- and b-states are the reciprocal values of corresponding state frequencies:
$$[\tau _a=1/\nu _a;\text{ }\tau _{\overline{a}}=1/\nu _{\overline{a}}]_{tr,lb};\text{ [}\tau _b=1/\nu _b;\text{ }\tau \overline{_b}=1/\nu _{\overline{b}}]_{tr,lb}$$
(20)
\[$`(\nu _a)`$ and $`(\nu _b)]_{tr,lb}`$ correspond to eqs. 4.8 and 4.9 of ;
\[$`(\nu _{\overline{a}})`$ and $`(\nu _{\overline{b}})]_{tr,lb}`$ could be calculated using eqs.4.16; 4.17 .
The frequency of $`(\mathrm{𝐚𝐜})`$ and $`(\mathrm{𝐛𝐜}`$) convertons excitation \[lb/tr\]:
$$F_{ac}=\frac{1}{\tau _{ac}}P_{ac}/Z$$
(21)
$$F_{bc}=\frac{1}{\tau _{bc}}P_{bc}/Z$$
(22)
where: $`P_{ac}`$ and $`P_{bc}`$ are probabilities of corresponding convertons excitations (see eq.4.29a of ).
The frequency of Supereffectons and Superdeformons (biggest fluctuations) excitation is:
$$F_{SD}=\frac{1}{(\tau _A^{}+\tau _B^{}+\tau _D^{})}P_S^D^{}/Z$$
(23)
It is dependent on cycle-period of Supereffectons: $`\tau _{SD}=\tau _A^{}+\tau _B^{}+\tau _D^{}`$
and probability of Superdeformon activation ($`P_S^D^{}),`$ like the limiting stage of this cycle.
The averaged life-times of Supereffectons in $`A^{}`$ and $`B^{}`$ state are dependent on similar states of translational and librational macroeffectons :
$$\tau _A^{}=[(\tau _A)_{tr}(\tau _A)_{lb}]=[(\tau _a\tau _{\overline{a}})_{tr}(\tau _a\tau _{\overline{a}})_{lb}]^{1/2}$$
(24)
and that in B state:
$$\tau _B^{}=[(\tau _B)_{tr}(\tau _B)_{lb}]=[(\tau _b\tau _{\overline{b}})_{tr}(\tau _b\tau _{\overline{b}})_{lb}]^{1/2}$$
(25)
The life-time of Superdeformons excitation is determined by frequency of beats between A and Bstates of Supereffectons as:
$$\tau _D^{}=1/\left|(1/\tau _A^{})(1/\tau _B^{})\right|$$
(26)
The frequency of translational and librational macroeffectons $`AB`$ cycle excitations could be defined in a similar way:
$$\left[F_M=\frac{1}{(\tau _A+\tau _B+\tau _D)}P_M^D/Z\right]_{tr,lb}$$
(27)
where:
$$(\tau _A)_{tr,lb}=[(\tau _a\tau _{\overline{a}})_{tr,lb}]^{1/2}$$
(28)
and
$$(\tau _B)_{tr,lb}=[(\tau _b\tau _{\overline{b}})_{tr,lb}]^{1/2}$$
(29)
$$(\tau _D)_{tr,lb}=1/\left|(1/\tau _A)(1/\tau _B)\right|_{tr,lb}$$
(30)
The frequency of primary translational effectons $`(a`$$`b)_{tr}`$ transitions could be expressed like:
$$F_{tr}=\frac{1/Z}{(\tau _a+\tau _b+\tau _t)_{tr}}(P_d)_{tr}$$
(31)
where: $`(P_d)_{tr}`$ is a probability of primary translational deformons excitation (eq. 4.25 of );
$`[\tau _a;\tau _b]_{tr}`$ are the life-times of (a) and (b) states of primary translational effectons (eq. 20).
The frequency of primary librational effectons as ($`ab)_{lb}`$ cycles excitations is:
$$F_{lb}=\frac{1/Z}{(\tau _a+\tau _b+\tau _t)_{lb}}(P_d)_{lb}$$
(32)
where: $`(P_d)_{lb}`$is a probability of primary librational deformons excitation; $`\tau _a`$ and $`\tau _b`$ are the life-times of (a) and (b) states of primary librational effectons defined as (20).
The life-time of primary transitons (tr and lb) as a result of quantum beats between (a) and (b) states of primary effectons could be introduced as:
$$[\tau _t=\left|1/\tau _a1/\tau _b\right|^1]_{tr,lb}$$
(33)
For the case of $`(ab)^{1,2,3}`$ transitions of primary and secondary effectons (tr and lb), their life-times in (a) and (b) states are the reciprocal value of corresponding frequencies: $`[\tau _a=1/\nu _a`$ and $`\tau _b=1/\nu _b]_{tr,lb}^{1,2,3}`$. These parameters and the resulting ones could be calculated from eqs.(2.27; 2.28 of ) for primary effectons and (2.54; 2.55 of ) for secondary ones.
The results of calculations, using eqs. (31, 32) for frequency of excitations of primary tr and lb effectons are plotted on Fig. 3a,b.
The frequencies of Macroconvertons and Superdeformons were calculated using eqs.(17 and 23).
> Fig. 3. (a) - Frequency of primary \[tr\] effectons excitations, calculated from eq.(31);
>
> (b) - Frequency of primary \[lb\] effectons excitations, calculated from eq.(32);
>
> (c) - Frequency of $`[lb/tr]`$ Macroconvertons (flickering clusters) excitations, calculated from eq.(17);
>
> (d) - Frequency of Superdeformons excitations, calculated from eq.(28).
At the temperature interval (0-100)$`{}_{}{}^{0}C`$ the frequencies of translational and librational macrodeformons (tr and lb) are in the interval of (1.3-2.8)$`10^9s^1`$ and (0.2-13)$`10^6s^1`$ correspondingly. The frequencies of (ac) and (bc) convertons could be defined also using our software and formulae, presented at the end of Chapter 4 of .
The frequency of primary translational effectons $`[ab]_{tr}`$ excitations at 20$`{}_{}{}^{0}C`$, calculated from eq.(31) is $`\nu 710^{10}(1/s)`$. It corresponds to electromagnetic wave length in water with refraction index $`(n=1.33)`$ of:
$$\lambda =(cn)/\nu 6mm$$
(34)
For the other hand, there are a lot of evidence, that irradiation of very different biological systems with such coherent electromagnetic field exert great influences on their properties (Grundler and Keilman, 1983).
Between the dynamics/function of proteins, membranes, etc. and dynamics of their aqueous environment the strong interrelation is existing.
The frequency of macroconvertons, representing big density fluctuation in the volume of primary librational effecton at 37C is about $`10^7(1/s)`$ $`(`$Fig 3c).
The frequency of librational macrodeformons at the same temperature is about 10<sup>6</sup> s$`^1,`$i.e. coincides with frequency of large-scale protein cavities pulsations between open and closed to water states (see Fig.2). This confirm our hypothesis that the clusterphilic interaction is responsible for stabilization of the proteins cavities open state and that transition from the open state to the closed one is induced by coherent water cluster dissociation.
The frequency of Superdeformons excitation (Fig.3d) is much lower:
$$\nu _s(10^410^5)\text{ }s^1$$
(35)
Superdeformons are responsible for cavitational fluctuations in liquids and origination of defects in solids. Dissociation of oligomeric proteins, like hemoglobin or disassembly (peptization)of actin and microtubules could be also related with such big fluctuations.
5. Mesoscopic mechanism of enzyme catalysis
The mechanism of enzyme catalysis is one of the most intriguing and unresolved yet problems of molecular biology. It becomes clear, that it is interrelated not only with a spatial, but as well with hierarchical complicated dynamic properties of proteins (see book: ”The Fluctuating Enzyme” , Ed. by G.R.Welch, 1986).
The \[proteins + solvent\] system should be considered as a cooperative one with feedback links (Kaivarainen, 1985, 1992). Somogyi and Damjanovich (1986) proposed a similar idea that collective excitations of protein structure are interrelated with surrounded water molecules oscillations.
The enzymatic reaction can be represented in accordance with our dynamic model as a consequence of the following stages (Käiväräinen, 1989; ).
The first stage:
$$(I)E^b+SE^bS$$
(36)
\- the collision of the substrate (S) with the open (b) state of the active site \[AS\] cavity of enzyme (E).
The frequency of collisions between the enzyme and the substrate, whose concentrations are $`[C_E]`$ and $`[C_S]`$, respectively, is expressed with the known formula (Cantor and Schimmel, 1980):
$$\nu _{\text{col}}=4\pi r_0(D_E+D_S)N_0[C_E][C_S]$$
(37)
where: $`r_0=a_E+a_S`$ is the sum of the enzyme’s and substrate’s molecular radii; $`N_0`$ is the Avogadro number;
$$D_E=\frac{kT}{6\pi \eta a_E}\text{ and }D_S=\frac{kT}{6\pi \eta a_S}$$
(38)
\- are the diffusion coefficients of the enzyme and substrate; k is the Boltzmann constant; T is absolute temperature; $`\eta `$ is a solvent viscosity.
The probability of collision of (b) state of the active site with substrate is proportional to the ratio of the b-state outer cross section area to the whole enzyme surface area:
$$P_b=\frac{\mathrm{}^b}{\mathrm{}_E}F_b$$
(39)
where: $`F_b`$ $`=\frac{f_b}{f_b+f_a}`$ is a fraction of time, the active site \[AS\] spend in the open (b) - state.
So, the frequency of collision between the substrate and (b) state of the active site (AS) with account for (40), meaning the first stage of reaction is:
$$\nu _{col}^b=\nu _{col}P_b=k_I$$
(40)
The second stage of enzymatic reaction is a formation of the primary enzyme-substrate complex:
$$(II)E^bS[E^a^{}S]^{(1)}$$
(41)
It corresponds to transition of the active site cavity from the open (b) state to the closed (a) one and stabilization the latter state by a ligand.
The rate constant of the \[$`ba]`$ transitions is derived with the Stokes-Einstein and Eyring -Polany generalized equation (3):
$$k_{\text{II}}^{ba^{}}=\frac{kT}{\eta V}\mathrm{exp}\left(\frac{G_{st}^{ba}}{RT}\right)$$
(42)
where: $`\eta `$ is the solvent viscosity; V is the effective volume of the enzyme domain, whose diffusional reorientation accompanies the $`(ba)`$ transition of the active site \[AS\].
The leftward shift of the \[$`ab]`$ equilibrium between two states of the active site is concomitant with this stage of the reaction. It reflects the principle of stabilized conformity, related to AS domains movements that we have put forward in the previous section.
The third stage:
$$(III)\text{ }[E^aS][E^a^{}S^{}]$$
(43)
represents the formation of the secondary specific complex. This process is related to directed ligand diffusion into the active site cavity and the dynamic adaptation of its geometry to the geometry of the active site.
Here the Principle of Stabilized microscopic Conformity is realized, when the AS change its geometry from (a) to $`(a^{})`$ without domains reorientation. The rate constant of this stage is determined by the rate constant of substrate diffusion in the closed (a) state of the active site cavity. It is also expressed by generalized kinetic equation (42), but with other values of variables:
$$k_S^{12^{}}=\frac{1}{\tau _s^{}}\mathrm{exp}\left(\frac{G_s^a}{RT}\right)=k_{\text{III}}$$
(44)
where:
$$\tau _s=(v_s/k)\eta ^{\text{in}}/T$$
(45)
is the correlation time of substrate of volume ($`v_s)`$ in the (a) state of the active site; $`\eta ^{\text{in}}`$ is the internal effective viscosity; $`G_S^a`$ is the activation energy of thermal fluctuations of groups, representing small-scale dynamics (SS), which determine the directed diffusion of a substrate in the active site \[AS\] closed cavity.
The directed character of ligand diffusion in AS can be determined by the relaxation of a protein structure, due to perturbation of AS domains by ligand. The relaxation changes were observed in many reactions of specific protein-ligand complexes formation (Käiväräinen, 1985, 1989).
The complex formation \[pair of domains forming the AS + substrate\], followed by these domain immobilization can be considered as an emergency of a new enlarged protein primary effecton from two smaller ones, corresponding to less independent AS domains or their compact ”nodes”.
We assume that at this important stage, the waves B of the attacking catalytic atoms $`(\lambda _B^c)`$ and the attacked substrate atoms $`(\lambda _B^S)`$ start to overlap and interfere in such a way that conditions for quantum-mechanical beats between them become possible.
Let us consider these conditions in more detail.
According to classical statistics, every degree of freedom gets the energy, which is equal to $`kT/2`$. This condition corresponds to harmonic approximation when the mean potential and kinetic energies of particles are the same:
$$V=T_k=mv^2/2kT/2$$
(46)
The corresponding to such ideal case the de Broglie wave (wave B) length is equal to:
$$\lambda _B=\frac{h}{mv}=\frac{h}{(\text{mkT})^{1/2}}$$
(47)
For such condition the wave B length of proton at room temperature is nearly 2.5Å, for a carbon atom it is about three times smaller and for oxygen - four times as small.
In the latter two cases, the wave B lengths are comparable and even less than the sizes of the atoms itself. So, their waves can not overlap and, beats between them are not possible.
However, in real condensed systems with quantum properties, including the active sites of enzymes, the harmonic approximation is not valid because $`(T_k/V)1`$ $`(`$see Fig. 5 of ). Consequently, the kinetic energy of atoms of AS:
$$T_k(1/2)kT\text{ and }\lambda _Bh/(mkT)^{1/2}$$
(47a)
It must be taken into account that librations, in a general case are presented by rotational-translational unharmonic oscillations of atoms and molecules, but not by their rotational motions only (Coffey et al., 1984).
The length of waves B of atoms caused by a small translational component of most probable impulse, related to librations is bigger than that related to pure translations:
$$[\lambda _{lb}=(h/P_{lb})]>[\lambda _{tr}=(h/P_{tr})]$$
(47b)
Even in pure water the linear sizes of primary librational effectons are several times bigger than that of translational effectons and the size of one $`H_2O`$ molecule (see Fig. 7 of or Fig.4 of ).
In composition of the active site rigid core the librational waves B of atoms can significantly exceed the sizes of the atoms themselves. In this case their superposition, leading to quantum beats of waves B in the \[active site - substrate\] complex, accelerating the enzyme reaction is quite possible.
In accordance to our model, periodic energy exchange resulting from such beats occurs between waves B of the substrate and active site atoms.
The reaction $`\left[S^{}P\right]`$, accelerating by these quantum beats, is the next 4th stage of enzymatic process - the chemical transformation of a substrate into product:
$$(IV)[E^a^{}S^{}][E^a^{}P]$$
(48)
The angular wave B frequency of the attacked atoms of substrate with mass $`m_S`$ and the amplitude $`A_S`$ can be expressed by eq.(2.20):
$$\omega ^S=\mathrm{}/2m_SA_S^2$$
(49)
The wave B frequency of the attacking catalytic atom (or a group of atoms) in the active site is equal to:
$$\omega ^{\text{cat}}=\frac{\mathrm{}}{2m_cA_c}$$
(50)
The frequency of quantum beats which appear between waves B of catalytic and substrate atoms is:
$$\omega ^{}=\omega ^{\text{cat}}\omega ^S=\mathrm{}\left(\frac{1}{2m_cA_c}\frac{1}{2m_SA_S}\right)$$
(51)
The corresponding energy of beats:
$$E^{}=E^{\text{cat}}E^S=\mathrm{}\omega ^{}$$
(52)
It is seen from these formulae that the smaller the wave B mass of the catalytic atom $`(m_c)`$ and its amplitude (A$`{}_{c}{}^{})`$, the more frequently these beats occur at constant parameters of substrate (m<sub>S</sub> $`and`$ A$`{}_{S}{}^{})`$. The energy of beats is transmitted to the wave B of the attacked substrate atom from the catalytic atom, accelerating the reaction.
According to our model, the perturbations in the region of the active site are accompanied by the appearance of phonons and acoustic deformons in a form of small-scale dynamics of protein structure. They provide the signal transmission in oligomeric proteins to the central cavity and other active sites leading to allosteric effects.
It is known from the theory of oscillations (Grawford, 1973) that the effect of beats is maximal, if the amplitudes of the interacting oscillators are equal:
$$A_c^2A_S^2$$
(53)
The \[substrate $``$ product\] transformation can be considered as a result of the substrate wave B transition from the main \[S\] state to excited \[P\] state. The rate constant of such a reaction in the absence of the catalyst ($`k^{SP})`$ can be presented by the modified Eyring-Polany formulae, leading from eq.(2.27 of ) at condition: $`\mathrm{exp}(h\nu _p/kT)>>1`$
$$\nu _A^S=k^{SP}=\frac{E^P}{h}\mathrm{exp}\left(\frac{E^PE^S}{kT}\right)=\nu _B^P\mathrm{exp}\left[\frac{h(\nu _B^P\nu _B^S)}{kT}\right]$$
(54)
where: $`E^S=h\nu ^S`$ and $`E^P=h\nu ^P`$ are the main and excited - transitional to product energetic states of substrate, correspondingly; $`\nu ^S`$ and $`\nu ^P`$ are the substrate wave B frequencies in the main and excited states, respectively.
If catalyst is present, which acts by the above described mechanism, then the energy of the substrate E<sup>S</sup> is increased by the magnitude $`E^{\text{cat}}`$ with the quantum beats frequency $`\omega ^{}(51)`$ and gets equal to:
$$E^{Sc}=E^S+E^{\text{cat}}$$
(55)
Substituting $`E^P=h\nu _B^P`$ and $`E^{Sc}=h(\nu _B^S+\nu _B^{cat})`$ in (54), we derive the rate constants for the catalytic reaction in the moment of beats ($`k^{ScP}).`$ This corresponds to the 4th stage of enzymatic reaction:
$$\text{(IV): }k^{ScP}=\nu ^P\mathrm{exp}\left[\frac{h(\nu ^P\nu ^S\nu ^{\text{cat}})}{kT}\right]=k_{\text{IV}}$$
(56)
where: $`\nu ^P,\nu ^S`$ and $`\nu ^c`$ are the most probable B wave frequencies of the transition $`[SP]`$ state, of the substrate and of the catalyst atoms, correspondingly.
Hence, in the presence of the catalyst the coefficient of acceleration (q) is equal to:
$$q_{\text{cat}}=\frac{k^{ScP}}{k^{SP}}=\mathrm{exp}\left(\frac{h\nu ^{\text{cat}}}{kT}\right)$$
(57)
For example, at
$$h\nu ^c/kT10;\text{ }q_{\text{cat}}\text{ =}2.210^4$$
(58)
At room temperatures this condition corresponds to
$$E^{\text{cat}}=h\nu ^{\text{cat}}6\text{ kcal/mole. }$$
The beating acts, followed by transitions of a substrate molecule to activated by catalyst excited state $`(SSc)`$, can be accompanied by the absorption of phonons or photons with the frequency $`\omega ^{}(51)`$. In this case, the insolation of a \[substrate - catalyst\] system with ultrasound or electromagnetic field of the frequency $`\omega ^{}`$ should strongly accelerate the reaction when the resonance conditions are satisfied.
It is possible that the resonance effects of this type can account for the experimentally revealed response of various biological systems to electromagnetic field radiation with the frequency about $`\mathrm{\hspace{0.17em}6}10^{10}Hz`$ $`(`$Deviatkov et al., 1973). We have to point out that just such frequency is close to the frequency of (a$`b)_{tr}`$ transitions excitation of primary translational water effectons. The strict correlation between the dynamics of water and that of biosystems should exist on each hierarchic level of time and space.
Changes in the volume, geometry and electronic properties of the substrate molecule, resulting from its transition to the product, change its interaction energy with the active site by the magnitude:
$$\mathrm{\Delta }E_{SP}^a^{}=(E_S^a^{}E_P^a^{})$$
(59)
It must destabilize the closed state of the active site and increase the probability of its reverse \[$`a^{}b^{}]`$ transition.
Such transition promotes the last 5th stage of the catalytic cycle - the dissociation of the enzyme-product complex:
$$(V)[E^a^{}P][E_P^bP]E^b+P$$
(60)
The resulting rate constant of this stage, like stage (II), is described by the generalized Stokes-Einstein and Eyring-Polany equation (42), but with different activation energy $`G_{st}^{a^{}b}`$ valid for the \[$`a^{}b]`$ transition:
$$k_P^{a^{}b}=\frac{kT}{\eta V}\mathrm{exp}\left(\frac{G_{st}^{a^{}b}}{kT}\right)=1/\tau _P^{a^{}b}$$
(61)
If the lifetime of the $`(a^{})`$ state is sufficiently long, then the desorption of the product can occur irrespective of \[$`a^{}b]`$ transition, but with a longer characteristic time as a consequence of its diffusion out of the active site’s ”closed state”. The rate constant of this process $`(k_p^{})`$ is determined by small-scale dynamics. It practically does not depend on solvent viscosity, but can grows up with rising temperature, like at the 3d stage, described by (44):
$$k_P^{}=\frac{kT}{\eta _a^{}^{\text{in}}v_P}\mathrm{exp}\left(\frac{G_P^a^{}}{kT}\right)=1/\tau _P^{}$$
(62)
where $`\eta _a^{}^{\text{in}}`$ is active site interior viscosity in the closed $`(a^{})`$ state; $`v_P`$ is the effective volume of product molecules; $`P_a^{}=\mathrm{exp}(G_P^a^{}/RT)`$ is the probability of small-scale, functionally important motions necessary for the desorption of the product from the $`(a^{})`$ state of the active site; $`G_p^a^{}`$is the free energy of activation of such motions.
Because the processes, described by the eq.(61) and (62), are independent, the resulting product desorption rate constant is equal to:
$$k_V=k_P^{a^_b}+k_P^{}=1/\tau _P^{a^_b}+1/\tau _P^{}$$
(63)
The characteristic time of this final stage of enzymatic reaction is:
$$\tau _V=1/k_V=\frac{\tau _P^{a^_b}\tau _P^{}}{\tau _P^{a^_b}+\tau _P^{}}$$
(64)
This stage is accompanied by the relaxation of the perturbed AS domain and remnant protein structure to the initial state.
After the whole reaction cycle is completed the enzyme gets ready for the next cycle. The number of cycles (catalytic acts) in the majority of enzymes is within the limits of $`(10^210^4)s^1`$. It means that the \[$`ab]`$ pulsations of the active site cavities must occur with higher frequency as far it is only one of the five stages of enzymatic reaction cycle.
In experiments, where various sucrose concentrations were used at constant temperature, the dependence of enzymatic catalysis rate on solvent viscosity $`(T/\eta )`$ was demonstrated $`(`$Gavish and Weber, 1979). The amendment for changing the dielectric penetrability of the solvent by sucrose was taken into account. There are reasons to consider stages (II) and/or (V) in the model described above as the limiting ones of enzyme catalysis. According to eqs.(43) and (61), these stages depend on $`(T/\eta )`$, indeed.
The resulting rate constant of the enzyme reaction could be expressed as the reciprocal sum of life times of all its separate stages (I-V):
$$k_{\text{res}}=1/\tau _{\text{res}}=1/(\tau _I+\tau _{\text{II}}+\tau _{\text{III}}+\tau _{\text{IV}}+\tau _V)$$
(65)
where
$$\tau _I=1/k_I=\frac{1}{\nu _{\text{col}}P_b};\text{ }\tau _{\text{II}}=1/k_{\text{II}}=\frac{\eta V}{kT}\mathrm{exp}\left(\frac{G_{st}^{ba}}{RT}\right);$$
(65a)
$$\tau _{\text{III}}=1/k_{\text{III}}=\frac{\eta ^{\text{in}}v_S}{kT}\mathrm{exp}\left(\frac{G_{SS}^a}{RT}\right);$$
(65b)
$`\tau _{\text{IV}}=1/k_{\text{IV}}=\frac{1}{(\nu ^p)}\mathrm{exp}\left[\frac{h(\nu ^s\nu ^p\nu ^c}{kT}\right];`$
$`\tau _V`$ corresponds to eq.(64).
The slowest stages of the reaction seem to be stages (II), (V), and stage (III). The latter is dependent only on the small- scale dynamics in the region of the active site.
Sometimes product desorption goes on much more slowly than other stages of the enzymatic process, i.e.
$$k_Vk_{\text{III}}k_{\text{II}}$$
Then the resulting rate of the process $`(k_{\text{res}})`$ is represented by its limiting stage (eq. 63):
$$k_{\text{res}}\frac{kT}{\eta V}\mathrm{exp}\left(\frac{G_{st}^{a^{}b}}{RT}\right)+k_P^{}=1/\tau _{\text{res}}$$
(66)
The corresponding period of enzyme turnover: $`\tau _{\text{res}}\tau _V(eq\mathrm{\hspace{0.17em}64})`$. The internal medium viscosity ($`\eta ^{\text{in}}`$) in the protein regions, which are far from the periphery, is 2-3 orders higher than the viscosity of a water-saline solvent (0.001 P) under standard conditions:
$$(\eta ^{\text{in}}/\eta )10^3$$
Therefore, the changes of sucrose concentration in the limits of 0-40% at constant temperature can not influence markedly internal small- scale dynamics in proteins, its activation energy $`(G^a^{})`$ and internal microviscosity (Käiväräinen, 1989b). This fact was revealed using the spin-label method. It is in accordance with viscosity dependencies of tryptophan fluorescence quenching in proteins and model systems related to acrylamide diffusion in protein matrix (Eftink and Hagaman, 1986). In the examples of parvalbumin and ribonuclease $`T_1`$ it has been shown that the dynamics of internal residues is practically insensitive to changing solvent viscosity by glycerol over the range of 0.01 to 1 P.
It follows from the above data that the moderate changes in solvent viscosity ($`\eta `$) at constant temperature do not influence markedly the $`k_P^{}`$ value in eq.(66).
Therefore, the isothermal dependencies of k$`_{\text{res}}`$ on $`(T/\eta )_T`$ with changing sucrose or glycerol concentration must represent straight lines with the slope:
$$tg\alpha =\frac{\mathrm{\Delta }k_{\text{res}}}{\mathrm{\Delta }(T/\eta )_T}=\frac{k}{V}\mathrm{exp}(\frac{G_{st}^{a^{}b}}{RT})_T$$
(67)
The interception of isotherms at extrapolation to $`(T/\eta 0)`$ yields (62):
$$\left(k_{\text{res}}\right)_{(T/\eta )0}=k_P^{}=\frac{kT}{\eta _a^{}^{\text{in}}v_P}\mathrm{exp}\left(\frac{G^a^{}}{RT}\right)$$
(68)
The volume of the Brownian particle (V) in eq.(67) corresponds to the effective volume of one of the domains, which reorientation is responsible for (a$`b)`$ transitions of the enzyme active site.
Under conditions when the activation energy G$`{}_{}{}^{a^{}b}{}_{st}{}^{}`$ weakly depends on temperature, it is possible to investigate the temperature dependence of the effective volume V, using eq.(67), analyzing a slopes of set of isotherms (67).
Our model predicts the increasing of V with temperature rising. This reflects the dumping of the large-scale dynamics of proteins due to water clusters melting and enhancement the Van der Waals interactions between protein domains and subunits (Käiväräinen, 1985; 1989b, Käiväräinen et al., 1993). The contribution of the small-scale dynamics $`(k^{})`$ to $`k_{\text{res}}`$ must grow due to its thermoactivation and the decrease in $`\eta ^{\text{in}}`$ and $`G^a^{}(eq.32)`$.
The diffusion trajectory of ligands, substrates and products of enzyme reactions in ”closed” (a) states of active sites is probably determined by the spatial gradient of minimum wave B length (maximum impulses) values of atoms, forming the active site cavity.
We suppose, that functionally important motions (FIM), introduced by H. Frauenfelder et al., (1985, 1988), are determined by specific geometry of the impulse space characterizing the distribution of small-scale dynamics of domains in the region of protein’s active site.
The analysis of the impulse distribution in the active site area and energy of quantum beats between de Broglie waves of the atoms of substrate and active sites, modulated by solvent-dependent large-scale dynamics, should lead to complete understanding of the physical background of enzyme catalysis.
6. The mechanism of ATP hydrolysis energy utilization in muscle contraction
and protein polymerization
A great number of biochemical reactions are endothermic, i.e. they need additional thermal energy in contrast to exothermic ones. The most universal and common source of this additional energy is a reaction of adenosinetriphosphate (ATP) hydrolysis:
$$\text{ATP }\begin{array}{c}k_1\hfill \\ \hfill \\ k_1\hfill \end{array}\text{ ADP + P}$$
(69)
The reaction products are adenosinediphosphate (ADP) and inorganic phosphate (P).
The equilibrium constant of the reaction depends on the concentration of the substrate \[ATP\] and products \[ADP\] and \[P\] like:
$$K=\frac{k_1}{k_1}=\frac{[\text{ADP]}\text{[P]}}{[\text{ATP]}}$$
(70)
The equilibrium constant and temperature determine the reaction free energy change:
$$\mathrm{\Delta }G=RT\mathrm{ln}K=\mathrm{\Delta }HT\mathrm{\Delta }S$$
(71)
where: $`\mathrm{\Delta }`$H and $`\mathrm{\Delta }`$S are changes in enthalpy and entropy, respectively.
Under the real conditions in cell the reaction of ATP hydrolysis is highly favorable energetically as is accompanied by strong free energy decrease: $`\mathrm{\Delta }G=(11÷13)`$ kcal/M.
It follows from (71) that $`\mathrm{\Delta }G<0`$, when
$$T\mathrm{\Delta }S>\mathrm{\Delta }H$$
(72)
and the entropy and enthalpy changes are positive $`(\mathrm{\Delta }S>0`$ and $`\mathrm{\Delta }H>0).`$ However, the specific molecular mechanism of these changes in different biochemical reactions, including muscle contraction, remains unclear.
Acceleration of actin polymerization and tubulin self-assembly to the microtubules as a result of the ATP and nucleotide GTP splitting, respectively, is still obscure as well.
Using our model of water-macromolecule interaction , we can explain these processes by the ”melting” of the water clusters - librational effectons in cavities between neighboring domains and subunits of proteins. This melting is induced by absorption of energy of ATP or GTP hydrolysis and represents $`[lb/tr]`$ conversion of primary librational effectons to translational ones. It leads to the partial dehydration and rapprochement of domains and subunits. The concomitant transition of interdomain/subunit cavities from the ”open” B-state to the ”closed” A-state should be accompanied by decreasing of linear dimensions of a macromolecule. This process is usually reversible and responsible for the large-scale dynamics.
In the case when disjoining clusterphilic interactions that shift the $`\left[𝐀𝐁\right]`$ equilibrium to the right are stronger than Van der Waals interactions stabilizing A-state, the expansion of the macromolecule can induce a mechanical ”pushing” force.
In accordance to our model, this ”swelling driving force” is responsible for shifting of myosin ”heads” as respect to the actin filaments and muscle contraction.
Such FIRST relaxation ”swelling working step” is accompanied by dissociation of products of ATP hydrolysis from the active sites of myosin heads (heavy meromyosin).
The SECOND stage of reaction, the dissociation of the complex: \[myosin ”head” + actin\], is related to the absorption of ATP at the myosin active site. At this stage the $`AB`$ equilibrium between the heavy meromyosin conformers is strongly shifted to the right, i.e. to an expanded form of the protein.
The THIRD stage is represented by the ATP hydrolysis, (ATP $``$ ADP + P) and expelling of (P) from the active site. The concomitant local enthalpy and entropy jump leads to the melting of the water clusters in the cavities, $`𝐁𝐀`$ transitions and the contraction of free meromyosin heads.
The energy of the clusterphilic interaction at this stage is accumulated in myosin like in a squeezed spring.
After this 3d stage is over the complex \[myosin head + actin\] forms again.
We assume here that the interaction between myosin head and actin induces the releasing of the product (ADP) from myosin active site. It is important to stress that the driving force of ”swelling working stage$`\mathrm{"}:[AB]`$ transition of myosin cavities - is represented by our clusterphilic interactions (see Section 13.3 of and paper ).
A repetition of such a cycle results in the relative shift of myosin filaments with respect to actin ones and finally in muscle contraction.
The mechanism proposed does not need the hypothesis of Davydov’s soliton propagation (Davydov, 1984) along a myosin macromolecule. It seems that this nondissipative process scarcely takes place in strongly fluctuating biological systems. Soliton model does not take into account the real mesoscopic structure of macromolecules and their interaction with water as well.
Polymerization of actin, tubulin and other globular proteins composing cytoplasmic and extracell filaments due to hydrophobic interaction can be accelerated as a result of their selected dehydration due to local temperature jumps in mesoscopic volumes where the ATP and GTP hydrolysis takes place.
The \[assembly $``$ disassembly\] equilibrium is shifted as a result of such mesophase transition to the left in the case when $`\left[proteinprotein\right]`$ interface Van-der-Waals interactions are stronger than a clusterphilic one. The latter is mediated by librational water effecton stabilization in interdomain or intersubunit cavities.
It looks that the clusterphilic interactions play an extremely important role on mesoscale in the self-organization and dynamics of biological systems.
7. Water activity as a regulative factor in the intra- and inter-cell processes
Three most important factors can be responsible for the spatial processes in living cells:
1. Self-organization of supramolecular systems in the form of membranes, oligomeric proteins and filaments. Such processes can be mediated by water-dependent hydrophilic, hydrophobic and introduced by us clusterphilic interactions ;
2. Compartmentalization of cell volume by semipermeable lipid- bilayer membranes and due to different cell’s organelles formation;
3. Changes in the volumes of different cell compartments by osmotic process correlated in space and time. These changes are dependent on water activity regulated, in turn, mainly by \[assembly$``$ disassembly\] equilibrium shift of microtubules and actin’s filaments.
Disassembly of filaments leads to water activity decreasing due to increasing the fraction of vicinal water, representing a developed system of enlarged primary librational effectons near the surface of proteins (see section 13.5 of and ). The thickness of vicinal water layer is about 50 Å, depending on temperature and mobility of intra-cell components.
The vicinal water with more ordered and cooperative structure than that of bulk water represents the dominant fraction of intra-cell water. A lot of experimental evidences, pointing to important role of vicinal water in cell physiology were presented in reviews of Drost-Hansen and Singleton (1992) and Clegg and Drost-Hansen (1991).
The dynamic equilibrium: $`\{`$ I $``$ II $``$ III $`\}`$ between three stages of macromolecular self-organization, discussed in Section 13.5 of (Table 2) and in paper , has to play an important role in biosystems: blood, lymph as well as inter- and intra-cell media. This equilibrium is dependent on the water activity (inorganic ions, pH), temperature, concentration and surface properties of macromolecules.
Large-scale protein dynamics, decreasing the fraction of vicinal water (Käiväräinen, 1986, Käiväräinen et al., 1990) is dependent on the protein’s active site ligand state. These factors may play a regulative role in \[coagulation $``$ peptization\] and \[gel $``$ sol\] transitions in the cytoplasm of mobile cells, necessary for their migration.
A lot of spatial cellular processes such as the increase or decrease in the length of microtubules or actin filaments are dependent also on their self-assembly from corresponding subunits $`(\alpha ,\beta `$ tubulins and actin).
The self-assembly of such superpolymers is dependent on the \[association $`(A)`$ dissociation (B)\] equilibrium constant $`(K_{AB}=K_{BA}^1)`$. In turn, this constant is dependent on water activity $`(a_{H_2O})`$, as was shown earlier $`(eq\mathrm{.13.11}a`$ and 13.12 of and ).
The double helix of actin filaments, responsible for the spatial organization and cell’s shape dynamics, is composed of the monomers of globular protein - actin (MM 42.000). The rate of actin filament polymerization or depolymerization, responsible for cells shape adaptation to environment, is very high and strongly depends on ionic strength (concentration of $`NaCl`$, $`Ca^{++},Mg^{++})`$. For example, the increasing of $`NaCl`$ concentration and corresponding decreasing of $`a_{H_2O}`$ stimulate the actin polymerization. The same is true of $`\alpha `$ and $`\beta `$ tubulin polymerization in the form of microtubules.
The activity of water in cells and cell compartments can be regulated by $`\left[Na^+K^+\right]ATP`$dependent pumps. Even the equal concentrations of $`Na^+`$ and $`K^+`$ decrease water activity $`a_{H_2O}`$ differently due to their different interaction with bulk and, especially with ordered vicinal water (Wiggins 1971, 1973).
Regulation of pH by proton pumps, incorporated in membranes, also can be of great importance for intra-cell $`a_{H_2O}`$ changing.
Cell division is strongly correlated with dynamic equilibrium: \[assembly $``$ disassembly\] of microtubules of centrioles.. Inhibition of tubulin subunits dissociation (disassembly) by addition of $`D_2O`$, or stimulation this process by decreasing temperature or increasing hydrostatic pressure stops cell mitosis - division (Alberts et al., 1983).
The above mentioned factors enable to affect the $`AB`$ equilibrium of cavity between $`\alpha and\beta `$ tubulins, composing microtubules. These factors action confirm our hypothesis, that microtubules assembly are mediated by clusterphilic interaction (see section 17.5 of and paper ).
The decrease in temperature and increase in intra-microtubules pressure lead to the increased dimensions of librational water effectons, clustrons and finally this induce disassembly of microtubules.
Microtubules are responsible for the coordination of intra-cell space organization and movements, including chromosome movement at the mitotic cycle, coordinated by centrioles.
The communication between different cells by means of channels can regulate the ionic concentration and correspondent $`a_{H_2O}`$ gradients in the embryo.
In accordance with our hypothesis, the gradient of water activity, regulated by change of vicinal water fraction in different compartments of cell can play a role of so-called morphogenic factor necessary for differentiation of embryo cells.
8. Water and cancer
We put forward a hypothesis that unlimited cancer cell division is related to partial disassembly of cytoskeleton’s actin-like filaments due to some genetically controlled mistakes in biosynthesis and increasing the osmotic diffusion of water into transformed cell.
Decreasing of the intra-cell concentration of any types of ions $`(Na^+,K^+,H^+`$, $`Mg^{2+}`$etc.), as the result of corresponding ionic pump destruction, incorporated in biomembranes, also may lead to disassembly of filaments.
The shift of equilibrium: \[assembly $``$ disassembly\] of microtubules (MTs) and actin filaments to the right increases the amount of intra-cell water, involved in hydration shells of protein and decreases water activity. As a consequence of concomitant osmotic process, cells tend to swell and acquire a ball-like shape. The number of direct contacts between transformed cells decrease and the water activity in the intercell space increases also.
We suppose that certain decline in the external inter-cell water activity could be a triggering signal for the inhibition of normal cell division. The shape of normal cells under control of cell’s filament is a specific one, providing good dense intercell contacts with limited amount of water, in contrast to transformed cells.
If this idea is true, the absence of contact inhibition in the case of cancer cells is a result of insufficient decreasing of intercell water activity due to loose \[cell-cell\] contacts.
If our model of cancer emergency is correct, then the problem of tumor inhibition is related to the problem of inter - and intra-cell water activity regulation by means of chemical and physical factors.
Another approach for cancer healing we can propose here is the IR laser treatment of transformed cells with IR photons frequencies, stimulating superdeformons excitation and collective disassembly of MTs in composition of centrioles. This will prevent cells division and should have a good therapeutic effect. This approach is based on assumption that stability of MTs in transformed cells is weaker and/or resonant frequency of their superdeformons excitation differs from that of normal cells.
====================================================================
REFERENCES
> Alberts B., Bray D., Lewis J., Ruff M., Roberts K. and Watson J.D. Molecular Biology of Cell. Chapter 10. Garland Publishing, Inc.New York, London, 1983.
>
> Antonchenko V.Ya. Physics of water. Naukova dumka, Kiev, 1986.
>
> Cantor C.R., Schimmel P.R. Biophysical Chemistry. W.H.Freemen and Company, San Francisco, 1980.
>
> Clegg J. S. On the physical properties and potential roles of intracellular water. Proc.NATO Adv.Res.Work Shop. 1985.
>
> Clegg J.S. and Drost-Hansen W. On the biochemistry and cell physiology of water. In: Hochachka and Mommsen (eds.). Biochemistry and molecular biology of fishes. Elsevier Science Publ. vol.1, Ch.1, pp.1-23, 1991.
>
> Coffey W., Evans M., Grigolini P. Molecular diffusion and spectra. A.Wiley Interscience Publication, N.Y., Chichester, Toronto, 1984.
>
> Davydov A.S. Solitons in molecular systems. Phys. Scripta, $`1979,\mathrm{\hspace{0.17em}20},\mathrm{\hspace{0.17em}387}394`$.
>
> Davydov A.S. Solitons in molecular systems. Naukova dumka, Kiev, 1984 (in Russian).
>
> Del Giudice E., Dogulia S., Milani M. and Vitello G. A quantum field theoretical approach to the collective behaviour of biological systems. Nuclear Physics $`1985,`$ $`B251[FS13],375400`$.
>
> Drost-Hansen W. In: Colloid and Interface Science. Ed. Kerker M. Academic Press, New York, 1976, p.267.
>
> Drost-Hansen W., Singleton J. Lin. Our aqueous heritage: evidence for vicinal water in cells. In: Fundamentals of Medical Cell Biology, v.3A, Chemisrty of the living cell, JAI Press Inc.,1992, p.157-180.
>
> Eftink M.R., Hagaman K.A. Biophys.Chem. 1986, 25, 277.
>
> Einstein A. Collection of works. Nauka, Moscow, 1965 (in Russian).
>
> Eisenberg D., Kauzmann W. The structure and properties of water. Oxford University Press, Oxford, 1969.
>
> Gavish B., Weber M. Viscosity-dependent structural fluctuations in enzyme catalysis. Biochemistry 18 (1979) 1269.
>
> Gavish B. in book: The fluctuating enzyme. Ed. by G.R.Welch. Wiley- Interscience Publication, 1986, p.264-339.
>
> Grawford F.S. Waves. Berkley Physics Course. Vol.3. McGraw- Hill Book Co., N.Y., 1973.
>
> Grundler W. and Keilmann F. Sharp resonance in Yeast growth proved nonthermal sensitivity to microwaves. Phys.Rev.Letts., $`1983,51,12141216`$.
>
> Käiväräinen A.I. Solvent-dependent flexibility of proteins and principles of their function. D.Reidel Publ.Co., Dordrecht, Boston, Lancaster, 1985, pp.290.
>
> Käiväräinen A.I. The noncontact interaction between macromolecules revealed by modified spin-label method. Biofizika (USSR$`)\mathrm{\hspace{0.33em}1987},\mathrm{\hspace{0.17em}32},\mathrm{\hspace{0.17em}536}`$.
>
> Käiväräinen A.I. Thermodynamic analysis of the system: water-ions-macromolecules. Biofizika (USSR$`),\mathrm{\hspace{0.17em}1988},\mathrm{\hspace{0.17em}33},\mathrm{\hspace{0.17em}549}`$.
>
> Käiväräinen A.I. Theory of condensed state as a hierarchical system of quasiparticles formed by phonons and three-dimensional de Broglie waves of molecules. Application of theory to thermodynamics of water and ice. J.Mol.Liq. $`1989a,\mathrm{\hspace{0.17em}41},\mathrm{\hspace{0.17em}53}60`$.
>
> Käiväräinen A.I. Mesoscopic theory of matter and its interaction with light. Principles of selforganization in ice, water and biosystems. University of Turku, Finland 1992, pp.275.
>
> Kaivarainen A. Dynamic model of wave-particle duality and Grand unification. University of Joensuu, Finland 1993. pp.118.
>
> Kaivarainen A. Mesoscopic model of elementary act of perception and braining. Abstracts of conference: Toward a Science of Consciousness 1996, p.74. Tucson, USA.
>
> Käiväräinen A., Fradkova L., Korpela T. Separate contributions of large- and small-scale dynamics to the heat capacity of proteins. A new viscosity approach. Acta Chem.Scand. $`1993,47,456460`$.
>
> Karplus M., McCammon J.A. Scientific American, April 1986, p.42.
>
> Koshland D.E. J.Theoret.Biol. 2(1962)75.
>
> Kovacs A.L. Hierarcical processes in biological systems. Math. Comput. Modelling. $`1990,\mathrm{\hspace{0.17em}14},\mathrm{\hspace{0.17em}674}679`$.
>
> Lumry R. and Gregory R.B. Free-energy managment in protein reactions: concepts, complications and compensations. In book: The fluctuating enzyme. A Wiley-Interscience publication. 1986, p. 341- 368.
>
> Schulz G.E., Schirmer R.H. Principles of protein structure. Springer-Verlag, New York, 1979.
>
> Somogyi B. and Damjanovich S. In book: The fluctuating enzyme. Ed. by G.R.Welch. A Wiley-InterScience Publication. 1986, 341-368.
>
> Wiggins P.M. Thermal anomalies in ion distribution in rat kidney slices and in a model system. Clin. Exp. Pharmacol. Physiol. 1972, 2, 171-176.
|
warning/0003/math-ph0003002.html
|
ar5iv
|
text
|
# On Some Additivity Problems in Quantum Information Theory
## 1 Introduction
Quantum information theory is not merely a theoretical basis for physics of information and computation. It is also a source of challenging mathematical problems, often having elementary formulation but still resisting solution. It appears that surprisingly little is known about what may be called the combinatorial geometry of tensor products of Hilbert spaces, even in finite dimensions. One group of open problems concerns the additivity properties of various quantities characterizing quantum channels, notably the capacity for classical information, and the “maximal output purity”, defined below. All known results, including extensive numerical work in the IBM group , the Quantum Information group in the Technical University of Braunschweig, and elsewhere, are consistent with the conjecture that these quantities are indeed additive (resp. multiplicative) with respect to tensor products of channels. A proof of this conjecture would have important consequences in quantum information theory: in particular, according to this conjecture, the classical capacity or the maximal purity of outputs cannot be increased by using entangled inputs of the channel.
In this paper we state the additivity/multiplicativity problems, give some relations between them, and prove some new partial results, which also support the conjecture.
## 2 Statement of the problem
Let us give precise formulation of the additivity problem for the classical capacity (see , ). Let $`()`$ be the $``$ -algebra of all operators in a finite dimensional unitary space $``$. We denote the set of states, i.e. positive unit trace operators in $`()`$ by $`𝒮()`$, the set of all $`m`$-dimensional projections by $`𝒫_m()`$ and the set of all projections by $`𝒫()`$. A quantum channel $`\mathrm{\Phi }`$ is a completely positive trace preserving linear map of $`()`$ (we are in the finite dimensional case and we use the Schrödinger picture). These are the maps admitting the Kraus decomposition (see e. g. , )
$$\mathrm{\Phi }(\rho )=\underset{k}{}A_k\rho A_k^{},$$
(1)
where $`A_k`$ are operators satisfying $`_kA_k^{}A_k=I.`$
Let $`H(\rho )=\text{Tr}\rho \mathrm{log}\rho `$ denote the von Neumann entropy of the state $`\rho `$ and define
$$C(\mathrm{\Phi })=\underset{p_i,\rho _i}{\mathrm{max}}[H(\underset{i}{}p_i\mathrm{\Phi }(\rho _i))\underset{i}{}p_iH(\mathrm{\Phi }(\rho _i))],$$
where the maximum is taken over all finite probability distributions $`\{p_i\}`$ on $`𝒮()`$, ascribing probabilities $`p_i`$ to (arbitrary) states $`\rho _i`$. The quantity $`C(\mathrm{\Phi })`$ appears as the “one-step classical capacity” of the quantum channel $`\mathrm{\Phi }`$ or the capacity with unentangled input states (we refer to for a detailed information-theoretic discussion and the proof of the corresponding coding theorem). A thorough discussion of the properties of $`C(\mathrm{\Phi })`$ is given in .
The additivity problem can be formulated as follows: let $`\mathrm{\Phi }_1,\mathrm{},\mathrm{\Phi }_n`$ be channels in the algebras $`(_1),\mathrm{},(_n)`$ and let $`\mathrm{\Phi }_1\mathrm{}\mathrm{\Phi }_n`$ be their tensor product in $`(_1\mathrm{}_n)`$ . Is it true that
$$C(\mathrm{\Phi }_1\mathrm{}\mathrm{\Phi }_n)=\underset{i=1}{\overset{n}{}}C(\mathrm{\Phi }_i)\mathrm{?}$$
(2)
This obviously holds for reversible unitary channels; in the additivity was established for the so called classical-quantum and quantum-classical channels, which map from or into an Abelian subalgebra of $`()`$.
Another closely related problem is the additivity of a quantity, which can be read as the “maximal output purity” of a channel. In fact, there are several quantities of this kind, depending on the way we measure “purity”. If we just take the von Neumann entropy as a measure of purity, we arrive at the question , whether or not
$$\underset{\rho 𝒮(_1\mathrm{}_n)}{\mathrm{min}}H((\mathrm{\Phi }_1\mathrm{}\mathrm{\Phi }_n)(\rho ))=\underset{i=1}{\overset{n}{}}\underset{\rho 𝒮(_i)}{\mathrm{min}}H(\mathrm{\Phi }_i(\rho ))\mathrm{?}$$
(3)
For a particular class of channels this property implies (2) (see the Lemma in Section 5 below).
We will also consider this problem for other measures of purity, based on the noncommutative $`\mathrm{}_p`$-norms
$$A_p=\left(\mathrm{Tr}|A|^p\right)^{\frac{1}{p}},$$
defined for $`p1`$, and $`A()`$, with the operator norm $`A`$ corresponding naturally to the case $`p=\mathrm{}`$. For an arbitrary quantum channel $`\mathrm{\Phi }`$ let us introduce the following notations for the “highest purity” of outputs of a channel
$`\nu _H(\mathrm{\Phi })`$ $`=`$ $`\underset{\rho }{\mathrm{min}}H(\mathrm{\Phi }(\rho )),`$ (4)
$`\nu _p(\mathrm{\Phi })`$ $`=`$ $`\underset{\rho }{\mathrm{max}}\mathrm{\Phi }(\rho )_p,`$ (5)
$`\nu _{\mathrm{}}(\mathrm{\Phi })`$ $`=`$ $`\underset{\rho }{\mathrm{min}}\mathrm{\Phi }(\rho )^1^1,`$ (6)
where the extrema are taken with respect to all input density matrices $`\rho `$ . By convexity of the norms, the extrema in the above definitions are attained on pure states (in the first (resp. last) case the operator convexity of the function $`xx\mathrm{log}x`$ (resp. $`xx^1`$) is also relevant).
Then the additivity/multiplicativity inequalities
$`\nu _p(\mathrm{\Phi }_1\mathrm{\Phi }_2)`$ $``$ $`\nu _p(\mathrm{\Phi }_1)\nu _p(\mathrm{\Phi }_2)`$ (7)
$`\nu _H(\mathrm{\Phi }_1\mathrm{\Phi }_2)`$ $``$ $`\nu _H(\mathrm{\Phi }_1)+\nu _H(\mathrm{\Phi }_1)`$ (8)
are clear from inserting product density operators into the defining variational expressions. The standing conjecture is that equality always holds in these inequalities, i.e. that choosing entangled input states is never helpful for getting purer output states.
Before proceeding to show some new partial results on this problem, it is helpful to establish the relation between $`\nu _H(\mathrm{\Phi })`$, and $`\nu _p(\mathrm{\Phi })`$ for $`p`$ close to one. Of course, some relationship is expected, as the von Neumann entropy $`H(\rho )`$ can be computed in terms of the derivative of $`\rho _p`$ at $`p=1^+`$. Here we find that if the equality holds in ( 7) for $`p`$ arbitrarily close to $`1`$, then it holds also in (8).
Proof. We shall use the fact that for every $`0<x1`$
$$\frac{1x^p}{p1}x\frac{\mathrm{log}x}{\mathrm{log}e}$$
if $`p1`$. Thus $`\frac{1\mathrm{Tr}\mathrm{\Phi }(\rho )^p}{p1}`$ is a monotonely increasing family of continuous functions of the variable $`\rho `$ which varies in the compact set $`𝒮()`$, converging pointwise to the continuous function $`H(\mathrm{\Phi }(\rho ))`$. By Dini’s Theorem, the convergence is uniform, and
$$\underset{\rho }{\mathrm{min}}H(\mathrm{\Phi }(\rho ))=\underset{p1}{lim}\frac{1\mathrm{max}_\rho \mathrm{Tr}\mathrm{\Phi }(\rho )^p}{p1}.$$
Therefore, if the equality holds in (7) for $`p`$ close to $`1`$,
$$\underset{\rho }{\mathrm{min}}H(\left(\mathrm{\Phi }_1\mathrm{\Phi }_2\right)(\rho ))=\underset{p1}{lim}\frac{1\mathrm{max}_\rho \mathrm{Tr}\left((\mathrm{\Phi }_1\mathrm{\Phi }_2)(\rho )\right)^p}{p1}$$
$$=\underset{p1}{lim}\frac{1\mathrm{max}_\rho \mathrm{Tr}\left(\mathrm{\Phi }_1(\rho )\right)^p\mathrm{max}_\rho \mathrm{Tr}\left(\mathrm{\Phi }_2(\rho )\right)^p}{p1}=\underset{\rho }{\mathrm{min}}H(\mathrm{\Phi }_1(\rho ))+\underset{\rho }{\mathrm{min}}H(\mathrm{\Phi }_2(\rho )).$$
$`\mathrm{}`$
## 3 Tensoring with an ideal channel
The first natural step is to establish the multiplicativity property when one factor is the identity channel.
Lemma. For $`=p,H,\mathrm{}`$
$$\nu _{}(\mathrm{\Phi }\mathrm{Id})=\nu _{}(\mathrm{\Phi }).$$
(9)
Since $`\nu _p(\mathrm{Id})=1`$, and $`\nu _H(\mathrm{Id})=0`$, this is indeed an instance of the additivity/multiplicativity conjecture.
Proof. We shall restrict to the case $`=p,1p\mathrm{}`$ . The argument in the case $`=\mathrm{}`$ is similar and the case $`=H`$ follows by the argument given above.
Let us denote by $`_1,_2`$ the Hilbert spaces of the first and the second system, respectively. Let $`\varphi _{12}`$ be a unit vector in $`_1_2,`$ and write $`\rho _{12}=|\varphi _{12}\varphi _{12}|`$ and $`\rho _1=\mathrm{Tr}_2\rho _{12}`$ for the partial state in $`_1`$. If $`\mathrm{\Phi }`$ is the channel in $`_1`$, we denote $`\rho _{12}^{}=(\mathrm{\Phi }\mathrm{Id})(\rho _{12})`$. Let us dilate the channel $`\mathrm{\Phi }`$ to a unitary evolution $`U_{13}`$ with the environment $`_3`$, initially in a pure state $`\rho _3=|\varphi _3\varphi _3|`$. The the final state of the environment is
$$\rho _3^{}=\mathrm{Tr}_1U_{13}(\rho _1|\varphi _3\varphi _3|)U_{13}^{}\mathrm{\Psi }(\rho _1).$$
Since the state of the composite system $`_1_2_3`$ remains pure after the unitary evolution, its partial states $`\rho _{12}^{},\rho _3^{}`$ are isometric . Therefore
$$(\mathrm{\Phi }\mathrm{Id})(\rho _{12})_p=\rho _{12}^{}_p=\rho _3^{}_p=\mathrm{\Psi }(\rho _1)_p.$$
Now the map $`\rho _1\mathrm{\Psi }(\rho _1)`$ is affine and the norm is convex, therefore the maximum of the quantity above is attained on pure $`\rho _1,`$ whence $`\rho _{12}=\rho _1\rho _2,`$ and the statement follows. $`\mathrm{}`$
We shall specifically need this Lemma in the case $`p=\mathrm{}`$. It is instructive to see an alternative direct proof in this case.
Proof. In what follows we take $`\rho =|\varphi \varphi |`$. We compute the operator norm of the Hermitian operator $`\mathrm{\Phi }(\rho )`$ as $`\mathrm{\Phi }(\rho )=sup_\psi \psi ,\mathrm{\Phi }(\rho )\psi `$, and take $`\mathrm{\Phi }`$ to be given in the Kraus decomposition (1 ). Then
$$\nu _{\mathrm{}}(\mathrm{\Phi })=\underset{\varphi ,\psi }{sup}\underset{k}{}\psi ,A_k\varphi \varphi ,A_k^{}\psi ,$$
where the supremum is over all unit vectors in the appropriate spaces. The expression under the supremum can be read as the $`\mathrm{}^2`$-norm of a vector with components $`\varphi ,A_k^{}\psi `$. We write this norm also as “the largest scalar product with a unit vector” $`\chi `$, i.e.,
$`\nu _{\mathrm{}}(\mathrm{\Phi })`$ $`=`$ $`\left(\underset{\varphi ,\psi ,\chi }{sup}{\displaystyle \underset{k}{}}\overline{\chi _k}\varphi ,A_k^{}\psi \right)^2=\left(\underset{\psi ,\chi }{sup}{\displaystyle \underset{k}{}}\overline{\chi _k}A_k^{}\psi \right)^2`$ (10)
$`=`$ $`\underset{\chi }{sup}{\displaystyle \underset{k}{}}\overline{\chi _k}A_k^{}^2=\underset{\chi }{sup}{\displaystyle \underset{k}{}}\chi _kA_k^2,`$ (11)
where all suprema are over unit vectors. Obviously, the Kraus operators for $`\mathrm{\Phi }\mathrm{Id}`$ are $`A_kI`$, so
$$\nu _{\mathrm{}}(\mathrm{\Phi }\mathrm{Id})=\underset{\chi }{sup}\underset{k}{}\chi _k(A_kI)^2=\underset{\chi }{sup}\left(\underset{k}{}\chi _kA_k\right)I^2=\nu _{\mathrm{}}(\mathrm{\Phi }).$$
$`\mathrm{}`$
We will also need the analogous result for a quantity in which the two vectors $`\varphi ,\psi `$ in the above proof are fixed to be the same: for any channel $`\mathrm{\Phi }`$, let
$$\nu _{\mathrm{}}(\mathrm{\Phi })=\underset{\psi }{sup}\psi ,\mathrm{\Phi }(|\psi \psi |)\psi ,$$
(12)
where the supremum is again over all unit vectors. Note that this expression only makes sense, if the channel does not change the type of system, i.e., input and output algebra are the same. Then
$$\nu _{\mathrm{}}(\mathrm{\Phi }\mathrm{Id})=\nu _{\mathrm{}}(\mathrm{\Phi }).$$
(13)
Proof: Again we use Kraus decomposition (1). For $`\psi `$ we use the Schmidt decomposition $`\psi =_\mu \sqrt{c_\mu }e_\mu e_\mu ^{}`$, where the $`e_\mu `$ and $`e_\mu ^{}`$ are orthonormal systems. Then the expression to maximized on the left-hand side becomes
$`{\displaystyle \underset{\mu \nu \alpha \beta k}{}}(c_\mu c_\nu c_\alpha c_\beta )^{1/2}e_\mu e_\mu ^{},(A_kI)e_\nu e_\nu ^{}e_\alpha e_\alpha ^{},(A_k^{}I)e_\beta e_\beta ^{}`$
$`=`$ $`{\displaystyle \underset{\mu \nu \alpha \beta k}{}}(c_\mu c_\nu c_\alpha c_\beta )^{1/2}e_\mu ,A_ke_\nu e_\alpha ,A_k^{}e_\beta \delta _{\mu \nu }\delta _{\alpha \beta }`$
$`=`$ $`{\displaystyle \underset{\mu \alpha k}{}}c_\mu c_\alpha e_\mu ,A_ke_\mu e_\alpha ,A_k^{}e_\alpha ={\displaystyle \underset{k}{}}\mathrm{tr}(\rho _1A_k)\mathrm{tr}(\rho _1A_k^{}),`$
where $`\rho _1=_\mu c_\mu |e_\mu e_\mu |`$ is the reduced density matrix belonging to $`\psi `$. Since the function $`\rho _1|\mathrm{tr}(\rho _1A_k)|^2`$ is convex, this expression attains its maximum with respect to $`\psi `$ when $`\rho _1`$ is pure, i.e., when $`\psi `$ is a product. $`\mathrm{}`$
## 4 Weak Noise
One testing ground for the multiplicativity/additivity conjecture are channels close to the identity. For such channels the purity parameters can be evaluated in lowest order in the deviation from the identity. Doing this for each subchannel and for their tensor product, one can explicitly check the conjecture. As the following result shows, this test supports the conjecture.
Consider a channel with weak noise, i.e. choose some channel $`\mathrm{\Phi }`$ on $`()`$, and set
$$\mathrm{\Phi }^{(ϵ)}=(1ϵ)\mathrm{Id}+ϵ\mathrm{\Phi }.$$
(14)
For small $`ϵ`$ this is a weak noise channel, which has the property that for any pure input the output will be nearly pure.
Theorem. The multiplicativity hypothesis for the quantities $`\nu _p(\mathrm{\Phi })`$ with $`1p\mathrm{}`$ and the additivity hypothesis for the quantity $`\nu _H(\mathrm{\Phi })`$ hold true approximately in the leading order in $`ϵ`$.
Proof. In order to estimate these quantities for the weak noise channels, we need to estimate entropy, and the $`p`$-norms near a pure state. Let $`\rho `$ be a density operator on a $`d`$-dimensional Hilbert space, and suppose that $`\rho =1ϵ+𝐨(ϵ)`$. Then the leading order of the other norms is determined completely by $`ϵ`$:
$`\rho _p`$ $`=`$ $`1ϵ+𝐨(ϵ)\text{for }p>1`$ (15)
$`H(\rho )`$ $`=`$ $`ϵ\mathrm{log}ϵ+𝐨(ϵ\mathrm{log}ϵ),`$ (16)
where as usual $`𝐨(ϵ)`$ stands for terms going to zero faster than $`ϵ`$ as $`ϵ0`$. In this case we can say more: in first line we have $`0\text{remainder}Cϵ^p`$, for $`ϵ<1/2`$, where $`C`$ is a constant depending only on the dimension. Similarly, the estimates in the second line are independent of the details of $`\rho `$. Hence in leading order all the variational expressions are equivalent: each one amounts to maximizing $`ϵ`$.
Let us go back to the weak noise channel (14). To get high fidelity we need to maximize the leading term, so we can take $`\eta =\xi `$ in the following computation:
$`\nu _{\mathrm{}}(\mathrm{\Phi }^{(ϵ)})`$ $`=`$ $`\underset{\xi ,\eta }{sup}\xi ,\mathrm{\Phi }^{(ϵ)}(|\eta \eta |)\xi `$ (17)
$`=`$ $`\underset{\xi ,\eta }{sup}\left((1ϵ)|\xi ,\eta |^2+ϵ\xi ,\mathrm{\Phi }(|\eta \eta |)\xi \right)`$
$`=`$ $`1ϵ+ϵ\underset{\xi }{sup}\xi ,\mathrm{\Phi }(|\xi \xi |)\xi +𝐨(ϵ)`$
$`=`$ $`1ϵ+ϵ\nu _{\mathrm{}}(\mathrm{\Phi })+𝐨(ϵ).`$
Note that in all these estimates the remainder estimates can be done uniformly for all channels, depending only on dimension.
A tensor product of weak noise channels (14) is again of the same form:
$`\mathrm{\Phi }^{(ϵ)}`$ $`=`$ $`\mathrm{\Phi }_1^{(ϵ)}\mathrm{}\mathrm{\Phi }_n^{(ϵ)}`$
$`=`$ $`(1ϵ)^n\mathrm{Id}+ϵ(1ϵ)^{n1}\left(\mathrm{\Phi }_1\mathrm{Id}_{2\mathrm{}n}+\mathrm{}+\mathrm{Id}_{1\mathrm{}n1}\mathrm{\Phi }_n\right)+𝐨(ϵ)`$
$`=`$ $`(1nϵ)\mathrm{Id}+nϵ\delta \mathrm{\Phi }+𝐨(ϵ),`$
where $`\delta \mathrm{\Phi }`$ is the average of the $`n`$ channels $`\mathrm{Id}_{1\mathrm{}k1}\mathrm{\Phi }_k\mathrm{Id}_{k+1\mathrm{}n}`$. Hence, in order to compute the leading order of $`\nu _{\mathrm{}}(\mathrm{\Phi }^{(ϵ)})`$ by formula (17) we have to determine $`\nu _{\mathrm{}}(\delta \mathrm{\Phi })`$. We have
$`{\displaystyle \frac{1}{n}}{\displaystyle \underset{k=1}{\overset{m}{}}}\nu _{\mathrm{}}(\mathrm{\Phi }_k)`$ $``$ $`\nu _{\mathrm{}}(\delta \mathrm{\Phi })`$
$``$ $`{\displaystyle \frac{1}{n}}{\displaystyle \underset{k=1}{\overset{m}{}}}\nu _{\mathrm{}}(\mathrm{\Phi }_k\mathrm{Id})`$
$`=`$ $`{\displaystyle \frac{1}{n}}{\displaystyle \underset{k=1}{\overset{m}{}}}\nu _{\mathrm{}}(\mathrm{\Phi }_k)`$
where we have used in turn: insertion of product states into the supremum defining $`\nu _{\mathrm{}}(\delta \mathrm{\Phi })`$, convexity of $`\nu _{\mathrm{}}`$ as a supremum of affine functionals, and finally the restricted additivity result (13). Hence equality holds, which means that in the leading order in $`ϵ`$ all the variational expressions for the purity quantities $`\nu _{}()`$, with $`=p,H,\mathrm{}`$ are attained at product states. $`\mathrm{}`$
## 5 Depolarizing Channels
A channel is called bistochastic if $`\mathrm{\Phi }(I)=I`$, where $`I`$ is the unit operator in $`()`$. An important example is the depolarizing channel
$$\mathrm{\Phi }(\rho )=(1p)\rho +\frac{p}{d}(\text{Tr}\rho )I,\rho (),0<p<1,$$
where $`d=\text{dim}`$. A channel is called binary if $`d=2`$.
Lemma. Let $`\mathrm{\Phi }`$ be binary bistochastic channel, then
$$C(\mathrm{\Phi })=\mathrm{log}2\underset{\rho 𝒮()}{\mathrm{min}}H(\mathrm{\Phi }(\rho )).$$
(18)
If $`\mathrm{\Phi }_i`$ are binary bistochastic channels, then (3) implies ( 2).
Proof. The $``$ part of (18) is obvious from the fact that for any channel
$$C(\mathrm{\Phi })\mathrm{log}\text{dim}\underset{\rho 𝒮()}{\mathrm{min}}H(\mathrm{\Phi }(\rho )),$$
(19)
so we need to prove only $``$ part. Since the entropy is convex, the minimum is achieved at the set of extreme points of $`𝒮()`$ which is $`𝒫_1()`$. Let $`\rho `$ be the minimum point, then taking equiprobably $`\rho _0=\rho ,\rho _1=I\rho `$, we obtain
$$C(\mathrm{\Phi })H(\frac{1}{2}\mathrm{\Phi }(I))\frac{1}{2}[H(\mathrm{\Phi }(\rho ))+H(\mathrm{\Phi }(I\rho ))].$$
Since the channel is bistochastic, this is equal to $`H(\frac{1}{2}I)\frac{1}{2}[H(\mathrm{\Phi }(\rho ))+H(I\mathrm{\Phi }(\rho ))]`$, and since it is binary, this is equal to the right-hand side of (18).
To prove the second statement, it is sufficient to prove the $``$ part of ( 2), since $``$ part follows from the definitions. But this follows from (19) and (18). $`\mathrm{}`$
In the paper the relation(2) was proven for the two binary depolarizing channels $`\mathrm{\Phi }_1,\mathrm{\Phi }_2`$. The proof heavily uses Schmidt decomposition and as such does not generalizes to the case $`n>2`$. The main difficulty is evaluating the entropy of the product channel. However, it appears to be possible to check the additivity in the limiting cases of “weak” and “strong” depolarization in the leading order. Let us consider a collection of depolarizing channels $`\{\mathrm{\Phi }_i\}`$ in the Hilbert spaces $`_i`$, with parameters $`p_i,d_i,i=1,2,\mathrm{},n`$, and denote $`\mathrm{\Phi }=_{i=1}^n\mathrm{\Phi }_i`$, $`=_{i=1}^n_i`$, $`d=_{i=1}^nd_i`$. In the following we shall use symbols $`I_i`$ and $`I=_{i=1}^nI_i`$ for the identity operators in $`_i`$ and $``$ respectively. Let $`ϵ_L`$ be the tensor product $`\varphi _1\mathrm{}\varphi _n`$, where $`\varphi _i(\rho )=\frac{1}{d_i}\text{Tr}(\rho )I_i,iL\{1,2,\mathrm{},n\}`$, and $`\varphi _i(\rho )=\rho `$ otherwise, $`\rho 𝒮(_i)`$. Then $`ϵ_L`$ is a conditional expectation onto the subalgebra $`_L`$, generated by operators of the form $`A_1\mathrm{}A_n`$, where $`A_i=I_i`$ for $`iL`$, and is normalized partial trace with respect to its commutant. It has the property $`\underset{P𝒫(_L)}{\mathrm{min}}\mathrm{dim}P=_{i=1}^nd_i^{\theta _L(i)}`$, where $`\theta _L(i)=1`$ if $`iL`$ and $`\theta _L(i)=0`$ otherwise. Here we denoted by $`𝒫(_L)`$ the set of all orthogonal projections in $`_L`$. Notice that the inclusion $`_{L_1}_{L_2}`$ holds if $`L_2L_1`$. So $`ϵ_{L_1}ϵ_{L_2}=ϵ_{L_1L_2}`$.
We shall use the expansion
$$\mathrm{\Phi }=\underset{L}{}\underset{i=1}{\overset{n}{}}p_i^{\theta _L(i)}(1p_i)^{1\theta _L(i)}ϵ_L.$$
(20)
Weak (strong) depolarization corresponds to the case where all $`p_i`$ (respectively, $`1p_i`$) are small parameters, which we assume to be of the same order.
Proposition. The relation (2) holds in the cases of weak and strong depolarization approximately in the leading order.
Proof. In the case of weak depolarization the statement follows from the Theorem in Section 4.
In the case of strong depolarization we have to retain all the terms up to the second order in $`q_i=1p_i`$. Then the leading terms are
$$\mathrm{\Phi }(P)d^1\left[I+\underset{i=1}{\overset{n}{}}q_i(d_iP_iI)+\underset{1i<jn}{}q_iq_j(1d_iP_id_jP_j+d_id_jP_{ij})\right],$$
where we denoted $`d=_{i=1}^nd_i`$, $`P_i`$ is the partial state of $`P`$ in the $`i`$th Hilbert space, multiplied by the unit operator in the tensor product of the remaining Hilbert spaces, and similarly $`P_{ij}`$. Denoting the first (second) sum in the squared brackets $`A_1`$ (respectively $`A_2`$ ) one easily sees that both are traceless operators. Moreover, up to the second order,
$$\mathrm{\Phi }(P)\mathrm{log}\mathrm{\Phi }(P)d^1\left[(1+A_1+A_2)\mathrm{log}d^1+\left(A_1+A_2+\frac{A_1^2}{2}\right)\right]$$
and
$$H(\mathrm{\Phi }(P))\mathrm{log}d\frac{\mathrm{Tr}A_1^2}{2d}.$$
But
$$\mathrm{Tr}A_1^2=d\underset{i=1}{\overset{n}{}}q_i^2(d_i\mathrm{Tr}\rho _i^21),$$
where $`\rho _i`$ is the partial state of $`P`$ in the $`i`$th Hilbert space, which is maximized if and only if $`\rho _i`$ is one-dimensional projection, i.e. $`P=\rho _1\mathrm{}\rho _n.\mathrm{}`$
Partial answers to the multiplicativity hypothesis are given by the following Theorem. In fact, multiplicativity of $`\nu _{\mathrm{}}(\mathrm{\Phi })`$ for binary bistochastic maps follows from a more general result in .
Theorem.
$$\begin{array}{cccc}(i)\hfill & \nu _2(\mathrm{\Phi })\hfill & =\underset{i=1}{\overset{n}{}}\nu _2(\mathrm{\Phi }_i)\hfill & =\underset{i=1}{\overset{n}{}}\left(\frac{d_i1}{d_i}(1p_i)^2+\frac{1}{d_i}\right)^{1/2},\hfill \\ (ii)\hfill & \nu _{\mathrm{}}(\mathrm{\Phi })\hfill & =\underset{i=1}{\overset{n}{}}\nu _{\mathrm{}}(\mathrm{\Phi }_i)\hfill & =\underset{i=1}{\overset{n}{}}\left(1\frac{p_i(d_i1)}{d_i}\right),\hfill \\ (iii)\hfill & \nu _{\mathrm{}}(\mathrm{\Phi })\hfill & =\underset{i=1}{\overset{n}{}}\nu _{\mathrm{}}(\mathrm{\Phi }_i)\hfill & =\underset{i=1}{\overset{n}{}}\frac{p_i}{d_i}.\hfill \end{array}$$
Proof. $`(i)`$ It follows from the relation $`\underset{P𝒫(_L)}{\mathrm{min}}\text{dimP}=_{i=1}^nd_i^{\theta _L(i)}`$ that
$$\text{Tr}(ϵ_{L_1}(P)ϵ_{L_2}(P))=\text{Tr}(Pϵ_{L_1L_2}(P))=\text{Tr}(ϵ_{L_1L_2}(P)^2)\underset{i=1}{\overset{n}{}}d_i^{\mathrm{max}(\theta _{L_1}(i),\theta _{L_2}(i))}$$
for arbitrary $`P𝒫_1()`$ and equality holds only for factorizable projections. Hence
$$\text{Tr}((\mathrm{\Phi }(P))^2)=\text{Tr}((\underset{L}{}\underset{i=1}{\overset{n}{}}p_i^{\theta _L(i)}(1p_i)^{1\theta _L(i)}ϵ_L(P))^2)$$
$$\underset{L_1,L_2}{}\underset{i=1}{\overset{n}{}}p_i^{\theta _{L_1}(i)+\theta _{L_2}(i)}(1p_i)^{2\theta _{L_1}(i)\theta _{L_2}(i)}d_i^{\mathrm{max}(\theta _{L_1}(i),\theta _{L_2}(i))}=$$
$$=\underset{i=1}{\overset{n}{}}\underset{\theta _1,\theta _2=0,1}{}p_i^{\theta _1+\theta _2}(1p_i)^{2\theta _1\theta _2}d_i^{\mathrm{max}(\theta _1,\theta _2)}=\underset{i=1}{\overset{n}{}}\left(\frac{d_i1}{d_i}(1p_i)^2+\frac{1}{d_i}\right).$$
$`(ii)`$ Let us estimate
$$\mathrm{\Phi }(P)\underset{L}{}\underset{i=1}{\overset{n}{}}p_i^{\theta _L(i)}(1p_i)^{1\theta _L(i)}ϵ_L(P)$$
$$\underset{L}{}\underset{i=1}{\overset{n}{}}\left(\frac{p_i}{d_i}\right)^{\theta _L(i)}(1p_i)^{1\theta _L(i)}=\underset{i=1}{\overset{n}{}}\left(1\frac{p_i(d_i1)}{d_i}\right).$$
which proves the second statement.
$`(iii)`$ We have
$$||\mathrm{\Phi }(P))^1||_{k=0}^+\mathrm{}||\mathrm{\Phi }(IP)||^k.$$
(21)
Let us calculate
$$\underset{P𝒫_1(())}{\mathrm{max}}\mathrm{\Phi }(IP)\underset{Q𝒫_{d1}(())}{\mathrm{max}}\mathrm{\Phi }(Q).$$
We have
$$\mathrm{\Phi }(Q)\underset{i=1}{\overset{n}{}}p_iϵ_{\{1,2,\mathrm{},n\}}(Q)$$
$$+\underset{L\{1,2,\mathrm{},n\}}{}\underset{i=1}{\overset{n}{}}p_i^{\theta _L(i)}(1p_i)^{1\theta _L(i)}ϵ_L(Q)$$
$$\left(1d^1\right)\underset{i=1}{\overset{n}{}}p_i+\underset{L\{1,2,\mathrm{},n\}}{}\underset{i=1}{\overset{n}{}}p_i^{\theta _L(i)}(1p_i)^{1\theta _L(i)}=1\underset{i=1}{\overset{n}{}}\frac{p_i}{d_i}.$$
(22)
Here we have used the equality $`\underset{L}{}\underset{i=1}{\overset{n}{}}p_i^{\theta _L(i)}(1p_i)^{1\theta _L(i)}=1`$. Substituting (22) into ( 21), we get the last statement. $`\mathrm{}`$
Acknowledgment. The second author (ASH) acknowledges the support of the A. von Humboldt Foundation.
|
warning/0003/hep-th0003275.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
In nature there exists the mass hierarchy such as the Planck mass ($`10^{19}`$GeV), the GUT scale ($`10^{15}`$Gev), the electro-weak scale ($`10^2`$GeV), the neutrino mass($`10^{11}10^9`$GeV) and the cosmological size($`10^{41}`$GeV). How to naturally explain these different scales ranging over $`10^{60}`$ (so huge !) order has been the long-standing problem (the mass hierarchy problem). One famous approach is the Dirac’s large number theory. He tried to explain some ratios between basic physical quantities ( the electric force/the gravitational force, the age of the universe/the period during the light’s passing through the (classical) electron, the total mass in the universe/the proton mass) using the idea of the variable gravitational constant. Triggered by the development of the string and D-brane theories, some interesting new approaches to the compactification mechanism have recently been proposed and are applied to the hierarchy problem. Here we examine the Randall-Sundrum (RS) model which has some attractive features compared with the Kaluza-Klein compactification. The model is becoming a strong candidate that could solve the mass hierarchy problem. It looks, however, that the domain wall configuration is usually introduced ”by hand” ( not solving the field equation properly ) and is often ”approximated” by some distribution such as $`\delta `$-function or $`\theta `$-function. Such approximate approach sometimes hinders us from treating delicate (but important) procedures such as the boundary condition, the (infrared) regularization and the cosmological term. We present a solution of the field equation, which clarifies the compactification mechanism much more than the usual treatment. Especially the full-fledged treatment of the vacuum in terms of the 5D Higgs potential is an advantage. For the purpose of treating the model starting from the Lagrangian, we consider the model in a simplified case: One-wall model which was considered in . An interesting stable (kink) solution is found for a family of vacua. The solution does not miss the key points of the original one. Some new features are 1) the (5D) cosmological constant has both the upper bound and the lower bound (15); 2) the wall-thickness parameter $`k`$ (defined later) is regarded as the most important quantity to control the whole configuration properly and should be bounded both from below and from above (26); 3) As a plausible numerical choice, we propose $`k10^4`$GeV to explain quark, lepton and neutrino masses (Sec.5 and Sec.6).
The domain wall configuration, which is exploited in the RS model, has been frequently discussed so far in the literature. Especially the relation between some anomalies is examined in . The regularization of the chiral fermion problem on lattice was examined in . The similarity to these works is shown by clarifying the parameters correspondence. The difference between them is only the interpretation of the extra axis; In the chiral fermion case it is regarded as a purely technical axis for the regularization, whereas, in the RS model, it is a physical axis whose size is too small to measure at present. The analysis using the RS model can be regarded as a geometrical approach to the chiral fermion problem.
At present there exists no sign of the extra dimension(s) experimentally. Hence one might wonder about the worth of the present line of research. We remind you, however, of the following important aspects behind.
1. The higher dimensional view to the present world has been taken , since Kaluza-Klein, by many physicists. The latest approaches are unified models based on the supergravity, the string and the D-brane. This is one basic standpoint to understand the nature from geometry. The RS model is one of such approaches and has many distinguished properties compared with the past ones.
2. Use of the extra dimension(s) is one possible ingredient, independent of the supersymmetry, which can make us go beyond the standard model.
3. Many aspects appearing in the RS model overlap with those in the recent 10-15 years development of the theoretical physics: the chiral fermion problem (the previous paragraph), anomaly problem, D-brane physics, AdS/CFT, etc..
4. As reported in ICHEP2000, some interesting phenomena await the experimental tests.
We introduce the present model in Sec.2. A solution is obtained in Sec.3, where it is essentially described by three parameters specifying the Higgs vacuum. In Sec.4, the asymptotic behaviors, in the dimensional reduction from 5D to 4D, are evaluated for the three parameters. In Sec.5, the orders of magnitude for the two quantities, the thickness parameter ($`k`$) and the 5D Planck mass ($`M`$), are examined from the information of the 4D Planck mass, the 4D cosmological constant and the present experimental status of the Newton’s law. In Sec.6, similarity between the present analysis and the the chiral fermion problem is pointed out. Using the 4D fermion (quarks,leptons) masses, we examine the value of the size of the extra dimension ($`r_c`$). The three parameters referred above are precisely determined in Sec.7, where two constraints coming from the boundary condition are solved for the parameters. We conclude and discuss in Sec.8.
## 2 Model set-up
We start with the 5D gravitational theory, where the metric is Lorenzian, with the 5D Higgs potential.
$`S[G_{AB},\mathrm{\Phi }]={\displaystyle d^5X\sqrt{G}(\frac{1}{2}M^3\widehat{R}\frac{1}{2}G^{AB}_A\mathrm{\Phi }_B\mathrm{\Phi }V(\mathrm{\Phi }))},`$
$`V(\mathrm{\Phi })={\displaystyle \frac{\lambda }{4}}(\mathrm{\Phi }^2v_{0}^{}{}_{}{}^{2})^2+\mathrm{\Lambda },`$ (1)
where $`X^A(A=0,1,2,3,4)`$ is the 5D coordinates and we also use the notation $`(X^A)(x^\mu ,y),\mu =0,1,2,3.`$ $`X^4=y`$ is the extra axis which is taken to be a space coordinate. The signature of the 5D metric $`G_{AB}`$ is $`(++++)`$. $`\mathrm{\Phi }`$ is a 5D scalar field, $`G=detG_{AB}`$, $`\widehat{R}`$ is the 5D Riemannian scalar curvature. $`M(>0)`$ is the 5D Planck mass and is regarded as the fundamental scale of this dimensional reduction scenario. $`V(\mathrm{\Phi })`$ is the Higgs potential and serves for preparing the (classical) vacuum in 5D world. See Fig.1.
The three parameters $`\lambda ,v_0`$ and $`\mathrm{\Lambda }`$ in $`V(\mathrm{\Phi })`$ are called here vacuum parameters. $`\lambda (>0)`$ is a coupling, $`v_0(>0)`$ is the Higgs field vacuum expectation value, and $`\mathrm{\Lambda }`$ is the 5D cosmological constant. It is later shown that the sign of $`\mathrm{\Lambda }`$ must be negative for the proposed domain wall vacuum configuration. Following , we take the line element shown below.
$`ds^2=\mathrm{e}^{2\sigma (y)}\eta _{\mu \nu }dx^\mu dx^\nu +dy^2,`$ (2)
where $`\eta _{\mu \nu }=\text{diag}(1,1,1,1)`$. In this choice, the 4D Poincaré invariance is preserved. The ”warp” factor $`\mathrm{e}^{2\sigma (y)}`$ plays an important role throughout this paper. Note that, for the fixed $`y`$ case ($`dy=0`$), the metric is the Weyl transformation of the flat (Minkowski) space $`\eta _{\mu \nu }dx^\mu dx^\nu `$ (See Sec.6).
## 3 A solution
Let us solve the 5D Einstein equation.
$`M^3(\widehat{R}_{MN}{\displaystyle \frac{1}{2}}G_{MN}\widehat{R})=_M\mathrm{\Phi }_N\mathrm{\Phi }+G_{MN}({\displaystyle \frac{1}{2}}G^{KL}_K\mathrm{\Phi }_L\mathrm{\Phi }+V(\mathrm{\Phi })),`$
$`^2\mathrm{\Phi }={\displaystyle \frac{\delta V}{\delta \mathrm{\Phi }}}.`$ (3)
Following Callan and Harvey, we consider the case that $`\mathrm{\Phi }`$ depends only on the extra coordinate $`y`$, $`\mathrm{\Phi }=\mathrm{\Phi }(y)`$. The above equations reduce to
$`6M^3(\sigma ^{})^2={\displaystyle \frac{1}{2}}(\mathrm{\Phi }^{})^2+V,`$ (4)
$`3M^3\sigma ^{\prime \prime }=(\mathrm{\Phi }^{})^2.`$ (5)
We note that the ”matter equation”, the last one of (3), can also be obtained from $`(M,N)=(4,4)`$ component of the ”gravitational equation”, the first one of (3) which is given by (4). As the extra space (the fifth dimension), we take the real number space $`𝐑=(\mathrm{},+\mathrm{})`$. This is a simplified version of the original RS-model where $`S^1/𝐙_2`$ is taken. We impose the following asymptotic behaviour for the (classical) vacuum of $`\mathrm{\Phi }(y)`$.
$`\mathrm{\Phi }(y)\pm v_0,y\pm \mathrm{}`$ (6)
This means $`\mathrm{\Phi }^{}0`$, and from (5), $`\sigma ^{\prime \prime }0`$. Integrating eq.(5), we obtain
$`3M^3\{\sigma ^{}|_{y=+\mathrm{}}\sigma ^{}|_{y=\mathrm{}}\}={\displaystyle _{\mathrm{}}^{\mathrm{}}}(\mathrm{\Phi }^{})^2𝑑y>0.`$ (7)
From this result, we are led to $`\sigma ^{}\pm \omega ,\sigma \omega |y|`$ as $`y\pm \mathrm{}`$, where $`\omega (>0)`$ is some constant to be determined soon. We can scale out $`M`$ in (4) by rescaling all fields ($`\mathrm{\Phi },\sigma `$), all vacuum parameters ($`\lambda ,v_0,\mathrm{\Lambda }`$) and the coordinate $`y`$ with appropriate powers of $`M`$. ($`\mathrm{\Phi }=M^{3/2}\stackrel{~}{\mathrm{\Phi }},\sigma =\stackrel{~}{\sigma },v_0=M^{3/2}\stackrel{~}{v}_0,\lambda =M^1\stackrel{~}{\lambda },\mathrm{\Lambda }=M^5\stackrel{~}{\mathrm{\Lambda }},X^A=M^1\stackrel{~}{X}^A,`$) Therefore we may set $`M=1`$ without ambiguity. (Only when it is necessary, we explicitly write down $`M`$-dependence.)
First we fix the parameter $`\omega `$, by considering $`y\pm \mathrm{}`$ in (4), as
$`\omega =\sqrt{{\displaystyle \frac{\mathrm{\Lambda }}{6}}}M^{\frac{3}{2}},`$ (8)
where we see the sign of $`\mathrm{\Lambda }`$ must be negative, that is, the 5D geometry must be anti de Sitter in the asymptotic regions.
Let us take the following form for $`\sigma ^{}(y)`$ and $`\mathrm{\Phi }(y)`$ as a solution.
$`\sigma ^{}(y)=k{\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{c_{2n+1}}{(2n+1)!}}\{\mathrm{tanh}(ky+l)\}^{2n+1},`$
$`\mathrm{\Phi }(y)=v_0{\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{d_{2n+1}}{(2n+1)!}}\{\mathrm{tanh}(ky+l)\}^{2n+1},`$ (9)
where $`c`$’s and $`d`$’s are coefficient-constants (with respect to $`y`$) to be determined. The free parameter $`l`$ comes from the translation invariance of (4) and (5). A new mass scale $`k(>0)`$ is introduced here to make the quantity $`ky`$ dimensionless. The physical meaning of $`1/k`$ is the ”thickness” of the domain wall. The parameter $`k`$, with $`M`$ and $`r_c`$(defined later), plays a central role in this dimensional reduction scenario. We call $`M,k`$ and $`r_c`$ fundamental parameters. The distortion of 5D space-time by the existence of the domain wall should be small so that the quantum effect of 5D gravity can be ignored and the present classical analysis is valid. This requires the condition
$`kM.`$ (10)
The coefficient-constants $`c`$’s and $`d`$’s have the following constraints
$`\sqrt{{\displaystyle \frac{\mathrm{\Lambda }}{6}}}=k{\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{c_{2n+1}}{(2n+1)!}},1={\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{d_{2n+1}}{(2n+1)!}},`$ (11)
which are obtained by considering the asymptotic behaviours $`y\pm \mathrm{}`$ in (9). We will use these constraints in Sec.7.
We first obtain the recursion relations between the expansion coefficients, from the field equations (4) and (5). For $`n2`$, they are given by
$`{\displaystyle \frac{c_{2n+1}}{(2n)!}}{\displaystyle \frac{c_{2n1}}{(2n2)!}}={\displaystyle \frac{v_{0}^{}{}_{}{}^{2}}{3}}(D_{}^{}{}_{n}{}^{}2D_{}^{}{}_{n1}{}^{}+D_{}^{}{}_{n2}{}^{}),`$
$`6C_{n1}={\displaystyle \frac{v_{0}^{}{}_{}{}^{2}}{2}}(D_{}^{}{}_{n}{}^{}2D_{}^{}{}_{n1}{}^{}+D_{}^{}{}_{n2}{}^{})+{\displaystyle \frac{\lambda }{4}}{\displaystyle \frac{v_{0}^{}{}_{}{}^{4}}{k^2}}(E_{n2}2D_{n1}),`$ (12)
where
$`D_n={\displaystyle \underset{m=0}{\overset{n}{}}}{\displaystyle \frac{d_{2n2m+1}d_{2m+1}}{(2n2m+1)!(2m+1)!}},D_{}^{}{}_{n}{}^{}={\displaystyle \underset{m=0}{\overset{n}{}}}{\displaystyle \frac{d_{2n2m+1}d_{2m+1}}{(2n2m)!(2m)!}},`$
$`C_n={\displaystyle \underset{m=0}{\overset{n}{}}}{\displaystyle \frac{c_{2n2m+1}c_{2m+1}}{(2n2m+1)!(2m+1)!}},E_n={\displaystyle \underset{m=0}{\overset{n}{}}}D_{nm}D_m.`$ (13)
The first few terms, $`(c_1,d_1),(c_3,d_3)`$, are explicitly given as
$`d_1=\pm {\displaystyle \frac{\sqrt{2}}{v_0k}}\sqrt{\mathrm{\Lambda }+{\displaystyle \frac{\lambda v_{0}^{}{}_{}{}^{4}}{4}}},c_1={\displaystyle \frac{2}{3k^2}}(\mathrm{\Lambda }+{\displaystyle \frac{\lambda v_{0}^{}{}_{}{}^{4}}{4}}),`$
$`{\displaystyle \frac{d_3}{d_1}}=2+{\displaystyle \frac{1}{k^2}}\{{\displaystyle \frac{8}{3}}(\mathrm{\Lambda }+{\displaystyle \frac{\lambda v_{0}^{}{}_{}{}^{4}}{4}})\lambda v_{0}^{}{}_{}{}^{2}\},{\displaystyle \frac{c_3}{c_1}}=2+{\displaystyle \frac{1}{k^2}}\{{\displaystyle \frac{16}{3}}(\mathrm{\Lambda }+{\displaystyle \frac{\lambda v_{0}^{}{}_{}{}^{4}}{4}})2\lambda v_{0}^{}{}_{}{}^{2}\},`$ (14)
where $`\pm `$ sign in $`d_1`$ reflects $`\mathrm{\Phi }\mathrm{\Phi }`$ symmetry in (4) and (5). We take the positive one in the following. We can confirm that the above relations, (12) and (14), determine all $`c`$’s and $`d`$’s recursively in the order of increasing $`n`$. This is because eq.(12) is the coupled linear equation with respect to ($`c_{2n+1},d_{2n+1}`$) when lower-order $`c^{}`$s and $`d^{}`$s are regarded as obtained quantities. (Note: $`D_n^{}=\frac{2d_1}{(2n)!}d_{2n+1}+`$lower-order terms, $`D_n=\frac{2d_1}{(2n+1)!}d_{2n+1}+`$lower-order terms, $`C_n=\frac{2c_1}{(2n+1)!}c_{2n+1}+`$lower-order terms, $`E_n=\frac{8d_{1}^{}{}_{}{}^{3}}{(2n+1)!}d_{2n+1}+`$lower-order terms. ) All coefficients are solved and are described by the three dimensionless vacuum parameters:
$`(\mathrm{\Lambda }+\frac{\lambda v_{0}^{}{}_{}{}^{4}}{4})/k^2M^3,\lambda v_{0}^{}{}_{}{}^{2}/k^2,v_{0}^{}{}_{}{}^{2}/M^3`$.
In order for this solution to make sense, as seen from the expression for $`d_1`$, the 5D cosmological term $`\mathrm{\Lambda }`$ should be bounded also from below, in addition to from above.
$`{\displaystyle \frac{\lambda v_{0}^{}{}_{}{}^{4}}{4}}<\mathrm{\Lambda }<0.`$ (15)
At this stage the two constraints (11) are not taken into account. These impose some relations between vacuum parameters which will be explained in Sec.7.
## 4 Vacuum parameters: $`M`$ and $`k`$-dependence in the dimensional reduction
Let us examine the behaviour of the vacuum parameters ($`\mathrm{\Lambda },v_0,\lambda `$) near the 4D world: $`k\mathrm{}`$(the dimensional reduction). This should be taken consistently with (10). We will specify the above limit in the more well-defined way later. We take the following assumption which will be later checked using the final solution,
$`{\displaystyle \frac{c_{2n+1}}{c_1}}O(k^0)\times O(n^0),{\displaystyle \frac{d_{2n+1}}{d_1}}O(k^0)\times O(n^0),`$
$`\text{as}k\mathrm{},n\mathrm{},`$ (16)
where $`O(k^0)`$ and $`O(n^0)`$ are some constants of order $`k^0`$ and $`n^0`$. $`O(n^0)`$ behaviour for $`n\mathrm{}`$ is a sufficient condition for the convergence of the infinite series (9). Then the expressions (9) has the following asymptotic form, as $`k\mathrm{}`$.
$`\sigma ^{}(y)kc_1\times \theta (ky)\times \text{const}={\displaystyle \frac{2}{3k}}(\mathrm{\Lambda }+{\displaystyle \frac{\lambda v_{0}^{}{}_{}{}^{4}}{4}})\theta (ky)\times \text{const},`$
$`\mathrm{\Phi }(y)v_0d_1\times \theta (ky)\times \text{const}={\displaystyle \frac{\sqrt{2}}{k}}\sqrt{\mathrm{\Lambda }+{\displaystyle \frac{\lambda v_{0}^{}{}_{}{}^{4}}{4}}}\theta (ky)\times \text{const},`$ (17)
where $`\theta (y)`$ is the step function: $`\theta (y)=1`$ for $`y>0`$,$`\theta (y)=1`$ for $`y<0`$. (Note: $`(\mathrm{tanh}ky)^{2n+1}\theta (ky),k\mathrm{}`$.) Taking relations (6) and (8) into account, (17) means $`\lambda v_{0}^{}{}_{}{}^{4}\mathrm{\Lambda }k\sqrt{\mathrm{\Lambda }}k^2v_{0}^{}{}_{}{}^{2}`$. These relations say
$`\mathrm{\Lambda }M^3k^2,v_0M^{3/2},\lambda M^3k^2\text{as}k\mathrm{}.`$ (18)
These are leading behaviour of the vacuum parameters in the dimensional reduction. The first one above is given in the original . The more precise forms of (18) will be obtained, in Sec.7, using the constraints (11).
## 5 Parameter fitting
In order to express some physical scales in terms of the fundamental parameters $`M`$, $`k`$ and $`r_c`$( to be introduced soon), we consider the case that the 4D geometry is slightly fluctuating around the Minkowski (flat) space.
$`ds^2=\mathrm{e}^{2\sigma (y)}g_{\mu \nu }(x)dx^\mu dx^\nu +dy^2,g_{\mu \nu }=\eta _{\mu \nu }+h_{\mu \nu },h_{\mu \nu }O({\displaystyle \frac{1}{k}}).`$ (19)
The leading order $`O(k^0)`$ results of the previous section remain valid.
### 5.1 The Planck mass
The gravitational part of 5D action (1) reduces to 4D action as
$`{\displaystyle d^5X\sqrt{G}M^3\widehat{R}}M^3{\displaystyle _{r_c}^{r_c}}𝑑y\mathrm{e}^{2\sigma (y)}{\displaystyle d^4x\sqrt{g}R}+\mathrm{},`$ (20)
where the infrared regularization parameter $`r_c`$ is introduced. $`r_c`$ specifies the length of the extra axis. Using the asymptotic forms, $`\sigma (y)\omega |y|`$ as $`y\pm \mathrm{}`$ and $`\omega =\sqrt{\frac{\mathrm{\Lambda }}{6}}M^{\frac{3}{2}}k`$ as $`k\mathrm{}`$, we can evaluate the order of $`M_{pl}`$ as
$`M_{pl}^{}{}_{}{}^{2}M^3{\displaystyle _{r_c}^{r_c}}𝑑y\mathrm{e}^{2\omega |y|}={\displaystyle \frac{M^3}{\omega }}(1\mathrm{e}^{2\omega r_c}){\displaystyle \frac{M^3}{k}},`$ (21)
where we have used the 4D reduction condition:
$`kr_c1.`$ (22)
The result (21) is again same as in . The above condition should be interpreted as the precise (well regularized) definition of $`k\mathrm{}`$ used so far. We note $`r_c`$ dependence in (21) is negligible for $`kr_c1`$. This behaviour shows the distinguished contrast with the Kaluza-Klein reduction ($`M_{pl}^{}{}_{}{}^{2}M^3r_c`$) as stressed in .
### 5.2 The cosmological term
The cosmological part of (1) reduces to 4D action as
$`{\displaystyle d^5X\sqrt{G}\mathrm{\Lambda }}\mathrm{\Lambda }{\displaystyle _{r_c}^{r_c}}𝑑y\mathrm{e}^{4\sigma (y)}{\displaystyle d^4x\sqrt{g}}\mathrm{\Lambda }_{4d}{\displaystyle d^4x\sqrt{g}},`$
$`\mathrm{\Lambda }_{4d}{\displaystyle \frac{\mathrm{\Lambda }}{2\omega }}(1\mathrm{e}^{4\omega r_c})M^3k<0,kr_c1.`$ (23)
$`\mathrm{\Lambda }_{4d}`$ is the cosmological term in the 4D space-time. It does not, like $`M_{pl}`$, depend on $`r_c`$ strongly. The result says the 4D space-time should also be anti de Sitter.
### 5.3 Numerical fitting
Let us examine what orders of values should we take for the fundamental parameters $`M`$ and $`k`$. ( $`r_c`$ is later fixed by the information of the 4D fermion masses. ) Using the value $`M_{pl}10^{19}`$GeV , the ”rescaled” cosmological parameter $`\stackrel{~}{\mathrm{\Lambda }}_{4d}\mathrm{\Lambda }_{4d}/M_{pl}^{}{}_{}{}^{2}`$ has the relation:
$`\sqrt{\stackrel{~}{\mathrm{\Lambda }}_{4d}}kM^3\times 10^{38}\text{GeV},`$ (24)
where the relations (21) and (23) are used. The unit of $`M`$ is GeV and this mass unit is taken in the following. The observed value of $`\stackrel{~}{\mathrm{\Lambda }}_{4d}`$ is not definite, even for its sign. If we take into account the quantum effect, the value of $`\stackrel{~}{\mathrm{\Lambda }}_{4d}`$ could run along the renormalization. Furthermore if we consider the parameter $`\stackrel{~}{\mathrm{\Lambda }}_{4d}`$ represents some ”effective” value averaging over all matter fields, the value, no doubt, changes during the evolution of the universe. (Note the model (1) has no (ordinary) matter fields.) Therefore, instead of specifying $`\stackrel{~}{\mathrm{\Lambda }}_{4d}`$, it is useful to consider various possible cases of $`\stackrel{~}{\mathrm{\Lambda }}_{4d}k^2`$. Some typical cases are 1) ($`k=10^{41},M=0.1`$), 2) ($`k=10^{13},M=10^8`$) 3) ($`k=10,M=10^{13}`$) 4) ($`k=10^4,M=10^{14}`$) and 5) ($`k=10^{19},M=10^{19}`$). Case 1) gives the most plausible present value of the cosmological constant. The wall thickness $`1/k=10^{41}`$\[GeV<sup>-1</sup>\] , however, is the radius of the present universe. This implies the extra dimensional effect appears at the cosmological scale, which should be abandoned. Case 2) gives $`1/k=10^{13}`$ GeV<sup>-1</sup> $`1`$mm which is the minimum length at which the Newton’s law is checked. Usually $`k`$ should be larger than this value so that we keep the observed Newton’s law. 5) is an extreme case $`M=M_{pl}`$. The fundamental scale $`M`$ is given by the Planck mass. In this case, $`r_c1/k=1/M_{pl}`$ is acceptable, while $`\sqrt{\stackrel{~}{\mathrm{\Lambda }}_{4d}}M_{pl}`$ is completely inconsistent with the experiment and requires explanation. Most crucially the condition (10) breaks down. Cases 3) and 4) are some intermediate cases which are acceptable except for the cosmological constant. They will be used in Sec.8. At present any choice of ($`k,M`$) looks to have some trouble if we take into account the cosmological constant. We consider the observed cosmological constant ($`10^{41}`$GeV) should be explained by some unknown mechanism. (No successful explanation of the small cosmological constant exists . Ordinarily (without fine-tuning) the quantum-loop correction leads to the case 5). Compared with case 5), the cases 3) and 4) should be regarded as ”much improved” cases in this respect.)
## 6 Domain wall in the chiral fermion problem
We point out the mechanism presented here has a strong similarity to that in the chiral fermion determinant. The interpretation of the extra axis only is the difference. The axis is regarded as a real (but hardly measurable) axis here, whereas it is a regularization axis in the chiral problem. The parameter correspondence is
Randal-Sundrum Chiral Fermion
$`k:\text{(thickness of the wall)}^1`$ $``$ $`M_F:\text{1+4 dim fermion mass or}`$
$`\text{(thickness of the wall)}^1`$
$`M:\text{fundamental scale}`$ $``$ $`1/t:\text{temperature or}`$
$`1/a:\text{(lattice spacing)}^1`$
$`r_c:\text{Infrared reg.}`$ $``$ $`1/|k^\mu |:\text{(4D fermion mom.)}^1\text{or}`$
$`=\text{size of the extra axis}`$ $`1/m_q:\text{(quark mass)}^1`$
The condition on $`k`$ in the RS model, from (10) and (22), is given as
$`{\displaystyle \frac{1}{r_c}}kM.`$ (26)
The corresponding one of the chiral fermion is given by
$`|k^\mu |M_F{\displaystyle \frac{1}{t}}.`$ (27)
Both conditions guarantee the mechanism effectively works.
It is known, in the lattice chiral fermion, the choice of the parameter $`M_F`$ is so important to produce a good numerical output, say, the pion mass( where $`M_F`$ is denoted as $`M_5`$ and is called ”domain wall height”). Only for well-chosen value of $`M_F`$, the chiral properties are controlled. In this analogy the thickness parameter $`k`$, in the RS model, is considered to be a key quantity for controlling the whole configuration and for fitting with the real world quatities such as fermion masses.
The line element of (2) or (19) for a fixed $`y`$ is the Weyl scaling $`g_{\mu \nu }(x)\mathrm{e}^{2\sigma (y)}g_{\mu \nu }(x)`$ of the 4D world: $`(ds^2)_{4D}=g_{\mu \nu }(x)dx^\mu dx^\nu `$. $`\sigma (y)`$ is related to the 4D dynamics through the 5D geometrical setting. The extra dimension $`y`$ plays the role of the scaling parameter. On the other hand, in the chiral problem, the extra axis can be regarded as the Schwinger’s proper time (inverse temperature) $`t`$ through the relation :
$`({\displaystyle \frac{}{t}}+\widehat{D})G(x,y;t)=0,G(x,y;t)=<x|\mathrm{e}^{t\widehat{D}}|y>,`$ (28)
where $`\widehat{D}`$ is the general 4D operator and $`G(x,y;t)`$ is the density matrix. Formally it says $`\frac{G}{t}G^1=\frac{}{t}\mathrm{ln}G=\widehat{D}`$. This shows the scaling property of $`\mathrm{ln}G`$ along the coordinate $`t`$. These similar roles of $`y`$ and $`t`$ strongly indicate the both mechanisms are essentially the same.
In the view of , the ”direction” of the system evolvement of the present model is given by the sign change of the 5D Higgs field around the origin $`y=0`$.
As in the Callan and Harvey’s paper, we can have the 4D massless chiral fermion bound to the wall by introducing 5D Dirac fermion $`\psi `$ into (1).
$`S[G_{AB},\mathrm{\Phi }]+{\displaystyle d^5X\sqrt{G}(\overline{\psi }\text{ }\text{/}\text{ }\psi +g\mathrm{\Phi }\overline{\psi }\psi )}.`$ (29)
If we regulate the extra axis by the finite range $`r_cyr_c`$, the 4D fermion is expected to have a small mass $`m_fk\mathrm{e}^{kr_c}`$ (This is known for the two-walls case in ). If we take the case 3) in Subsec.5.3 ($`k=10,M=10^{13}`$) and regard the 4D fermion as a neutrino ($`m_\nu 10^{11}10^9\text{GeV}`$), we obtain $`r_c=2.762.30\text{GeV}^1`$. If we take case 4) ($`k=10^4,M=10^{14}`$), we obtain $`r_c=(3.452.99)\times 10^3\text{GeV}^1`$. When the quarks or other leptons ($`m_q,m_l10^310^2\text{GeV}`$) are taken as the 4D fermion, and take the case 4) in Subsec.5.3, we obtain $`r_c=(1.610.461)\times 10^3`$GeV<sup>-1</sup>. It is quite a fascinating idea to identify the chiral fermion zero mode bound to the wall with the neutrinos, quarks or other leptons.
## 7 Precise form of vacuum parameters
As shown in (18), an interesting aspect of the present solution is that some family of vacua is selected as the consistent (classical) configuration. Let us determine the precise form of (18) using the two constraints (11). In terms of new parameters $`\mathrm{\Omega }\mathrm{\Lambda }+\frac{\lambda }{4}v_{0}^{}{}_{}{}^{4}(0<\mathrm{\Omega }<\frac{\tau }{4}v_{0}^{}{}_{}{}^{2}),\tau \lambda v_{0}^{}{}_{}{}^{2}`$, instead of $`\mathrm{\Lambda }`$ and $`\lambda `$, the precise forms are obtained by the $`\frac{1}{k^2}`$-expansion for the case $`kr_c1`$ as
$`\mathrm{\Omega }=M^3k^2(\alpha _0+{\displaystyle \frac{\alpha _1}{(kr_c)^2}}+\mathrm{})=M^3k^2{\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{\alpha _n}{(kr_c)^{2n}}},`$
$`\tau =k^2(\gamma _0+{\displaystyle \frac{\gamma _1}{(kr_c)^2}}+\mathrm{})=k^2{\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{\gamma _n}{(kr_c)^{2n}}},`$
$`v_0=M^{3/2}(\beta _0+{\displaystyle \frac{\beta _1}{(kr_c)^2}}+\mathrm{})=M^{3/2}{\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{\beta _n}{(kr_c)^{2n}}},`$ (30)
where $`\alpha `$’s,$`\gamma `$’s and $`\beta `$’s are some numerical (real) numbers to be consistently chosen using (11). If we assume the relation (16), the infinite series of (11) can be safely truncated at the first few terms. In order to demonstrate how the vacuum parameters are fixed, we take into account up to $`n=2`$ in (11) and up to $`O(1/(kr_c)^{2\times 2})`$ in (30). For general $`M,k,r_c`$ except the condition $`kr_c1`$, the coefficients are determined as
$`\text{Vacuum 1:}(\alpha _0,\alpha _1,\alpha _2)(1,0,0)\text{input}`$
$`(\beta _0,\beta _1,\beta _2;\gamma _0,\gamma _1,\gamma _2)=(1.6,0,0;4.2,0,0),`$
$`\text{Vacuum 2:}(\alpha _0,\alpha _1,\alpha _2)(1,1,0)\text{input}`$
$`(\beta _0,\beta _1,\beta _2;\gamma _0,\gamma _1,\gamma _2)=(1.6,1.1,0.67;4.2,1.8,1.7),`$
$`\text{Vacuum 3:}(\alpha _0,\alpha _1,\alpha _2)(1,1,1)\text{input}`$
$`(\beta _0,\beta _1,\beta _2;\gamma _0,\gamma _1,\gamma _2)=(1.6,1.1,0.45;4.2,1.8,3.5).`$ (31)
We notice our solution has one free parameter for each $`n`$-th set ($`\alpha _n,\beta _n,\gamma _n`$). This is because the number of constraints for $`c^{}`$s and $`d^{}`$s is two (11), whereas that of quantities to be determined is three (30). Using this freedom we can adjust one of the three vacuum parameters in the way the observed physical values are explained. In (31), we take $`\alpha `$’s as the input. Taking the value $`kr_c=10`$, Vac.3 has the vacuum expectation value $`v_0M^{3/2}=1.6`$, the cosmological constant $`\mathrm{\Lambda }k^2M^3=1.7`$ and the coupling $`\lambda k^2M^3=1.6`$. Other vacua have almost the same values because the first order term dominate in (30) for the thin wall case $`kr_c1`$. We notice the dimensionless vacuum parameters, $`v_0M^{3/2},\mathrm{\Lambda }k^2M^3`$ and $`\lambda k^2M^3`$, are specified only by the value of $`kr_c`$ and the input data, say, $`\alpha ^{}`$s. If we specify $`k`$ and $`M`$, as considered in Subsec 5.3, the values $`v_0,\mathrm{\Lambda }`$ and $`\lambda `$ are obtained. Any higher-order, in principle, can be obtained by the $`\frac{1}{k^2}`$-expansion.
For the Vac.3, we plot $`\mathrm{\Phi }(y)`$ and $`\mathrm{\Phi }^{}(y)`$ for two cases $`kr_c=10`$ and $`kr_c=20`$ in Fig.2.
## 8 Discussion and conclusion
The assumption used in the present explanation is (16) only. The first some values in the series of $`n`$$`(c_3/c_1,d_3/d_1)=(1.05,0.48),(c_5/c_1,d_5/d_1)=(11.9,0.90)`$ for Vac.3 with $`kr_c=10`$ indicate its validity. Another evidence of the (strong) convergence is the fact that the normalization in Fig.2 is quite correctly reproduced.
If we take the boundary condition: $`\mathrm{\Phi }(y)v_0,y\pm \mathrm{}`$, instead of (6), the opposite chirality solution is obtained. Both of the pair, $`+`$ and $``$ chiralities, are indispensable when the ”vector-like” or non-chiral theory, such as QCD, is taken into account.
One of the fundamental parameters, $`r_c`$, is introduced in Sec.5 and 6 as the infrared regularization. This is quite natural in the standpoint of the discretized approach such as lattice. The treatment, however, should be regarded as an ”effective” approach or a ”temporary” stage of the unknown right treatment. The scale $`r_c`$ should be introduced naturally in the continuum approach. If we can generalize the present analysis to the case of the $`S^1/Z_2`$ extra space (two-walls case), $`r_c`$ is interpreted as the distance between the two walls . Another interesting possibility is that the scale $`r_c`$ could be given by some (at present) unknown mechanism in the space-time manifold such as the non-commutative geometry. It also looks that the AdS/CFT view of the present model could give a clue to the problem.
An important task to establish the RS scenario is to introduce the standard electro-weak model (chiral), QCD (non-chiral) and SUSY theories into this scheme. Recently the bulk standard model has been examined by . In a supersymmetric extension is examined. The RS model has given us richer possibilities for the mass hierarchy problem than before. It is hoped that the future experiments can select them.
Acknowledgment
The author thanks G.W.Gibbons for much discussions about the RS-model. Especially eq.(7) was pointed out by him. The author also thanks M.Yamaguchi for some comment.
|
warning/0003/cs0003037.html
|
ar5iv
|
text
|
# QUIP—A Tool for Computing Nonmonotonic Reasoning TasksThis work was partially supported by the Austrian Science Fund Project N Z29-INF.
## General Information
Designing theorem provers for nonmonotonic reasoning formalisms is challenged by the increased computational complexity compared to the inherent complexity of classical reasoning. As is well-known (????), the main reasoning tasks for almost all propositional nonmonotonic logics are either $`\mathrm{\Sigma }_2^p`$-complete or $`\mathrm{\Pi }_2^p`$-complete, i.e., they are located at the second level of the polynomial hierarchy, whereas classical reasoning resides at the first level of that hierarchy.
Important constituents in establishing the $`\mathrm{\Sigma }_2^p`$-completeness results for the different nonmonotonic reasoning tasks are so-called quantified boolean formulae (QBFs), which generalize ordinary propositional formulae by the admission of quantifiers ranging over propositional variables. To wit, demonstrating $`\mathrm{\Sigma }_2^p`$-hardness of a given nonmonotonic decision problem, say nmp, is usually achieved by reducing the problem $`\text{qsat}_2`$ to nmp, where $`\text{qsat}_2`$ is the problem of deciding the truth of QBFs being in prenex normal-form whose leading quantifiers are ordered like $`PQ`$ ($`P,Q`$ are lists of propositional variables), which is complete for $`\mathrm{\Sigma }_2^p`$. (In general, for any class $`\mathrm{\Sigma }_i^p`$, $`i1`$, there is a corresponding decision problem, $`\text{qsat}_i`$, complete for $`\mathrm{\Sigma }_i^p`$, whose task is to check the truth of QBFs of the form $`P_1P_2\mathrm{}P_{n1}P_n\mathrm{\Phi }`$, where $`P_1,\mathrm{},P_n`$ are lists of propositional variables and $`\mathrm{\Phi }`$ is a propositional formula.) More precisely, $`\mathrm{\Sigma }_2^p`$-hardness of nmp is shown by constructing a (polynomial) transformation $`𝒮`$ such that for every instance $`I`$ of $`\text{qsat}_2`$ it holds that $`I`$ is a yes-instance of $`\text{qsat}_2`$ iff $`𝒮(I)`$ is a yes-instance of nmp.
Now, given the membership of a nonmonotonic reasoning task nmp in the class $`\mathrm{\Sigma }_2^p`$, and from the $`\mathrm{\Sigma }_2^p`$-hardness of $`\text{qsat}_2`$, it follows immediately that every such decision problem nmp can be reduced to $`\text{qsat}_2`$, i.e., there is a (polynomial) transformation $`𝒯`$ such that for every instance $`I`$ of nmp it holds that $`I`$ is a yes-instance of nmp iff $`𝒯(I)`$ is a yes-instance of $`\text{qsat}_2`$.
In this paper, we describe the prototype system QUIP (??), an automated reasoning tool utilizing transformations of the latter kind to implement several different reasoning formalisms. The basic idea is to employ *existing sophisticated theorem provers for quantified boolean formulae*, taking care of evaluating the resultant instances of $`\text{qsat}_2`$.
At present, QUIP handles the following propositional nonmonotonic reasoning approaches:
* abduction;
* autoepistemic logic;
* default logic;
* disjunctive stable model semantics;
* circumscription.
The system has been implemented in C using standard tools like LEX and YACC (comprising a total of 2000 lines of code, excluding the used QBF-solver); it runs currently in a Unix environment (Sun/Solaris and Linux), but is easily portable to other operating systems as well.
The use of QBFs expressing advanced reasoning tasks has been advocated in (??) and is a natural generalization of a similar method applied for problems in NP. Such problems are often solved by a reduction to sat, the satisfiability problem of classical propositional logic, which is NP-complete (see, e.g., (??)). Besides the applications discussed in this paper, a reduction of planning problems to QBFs has been given in (?).
In general, since the evaluation of *arbitrary* QBFs is PSPACE-complete (in contrast to the $`\mathrm{\Sigma }_i^p`$-completeness of the restricted QBFs mentioned above), in principle *any* formalism having a decision problem in PSPACE can be handled by QUIP, provided a proper transformation has been found.
The next section outlines the overall architecture of QUIP and gives some background information on the different reasoning tasks implemented in QUIP. Then, the usability of the approach is described. The last section contains a discussion on benchmark problems and a comparison with other implementations.
## Description of the System
### System Architecture
The overall architecture of QUIP is depicted in Figure 1. QUIP consists of three parts, namely the filter program, the QBF-evaluator boole, and the output interpreter int.
The input filter translates the given problem description (e.g., a default theory, a disjunctive logic program, an abductive problem, etc.) into a quantified boolean formula, which is then sent to the QBF-evaluator boole. The result of boole, usually a formula in disjunctive normal form (often called *sum of products*, SOP), is interpreted by int. The latter part associates a meaningful interpretation to the formulae occurring in SOP and provides an explanation in terms of the underlying problem instance (e.g., an extension, a stable model, abductive explanations, etc.). The interpretation relies on a mapping of internal variables of the generated QBF into concepts of the problem description which is provided by filter.
The QBF-evaluator boole is a commonly available propositional theorem prover based on *binary decision diagrams* (BDDs) (?).<sup>1</sup><sup>1</sup>1For a comparison between different BDD packages, cf. (?). We chose to use this particular tool because of several reasons. For one, the program, together with its source code, is in the public domain and can be downloaded from the web (see http://www.cs.cmu.edu/$``$modelcheck/bdd.html). Second, it represents a sophisticated reasoning engine with several years of development. Third, it does not require of the input formula to be in a specific normal form. The latter point distinguishes boole from other QBF-solvers, like those proposed in (??), which operate on formulae being in *prenex conjunctive normal form*. Although these provers can in principle also be employed in QUIP, their inclusion would require an additional normal form translation, since the “natural” reductions of nonmonotonic reasoning tasks to QBFs do in general *not* result in formulae being in a particular normal form.
In order to incorporate new formalisms into QUIP, one has to extend the filter program responsible for the appropriate reductions, the mapping of the variables, and the interpreter int. The deductive engine remains unchanged in this process.
### Implemented Reasoning Tasks
In this subsection, we briefly discuss the different reasoning tasks currently implemented in QUIP. We will not present the concrete transformations into QBFs expressing these tasks; details are given in (?).
All formalisms processed by QUIP are propositional. In what follows, we assume a propositional language $``$ generated by a finite set of propositional variables $`V`$ using the standard sentential connectives $`\neg `$, $``$, $``$, $``$, and $``$. A theory is a finite set $`T`$, which will be identified with the formula $`_{\varphi T}\varphi `$.
Note that, strictly speaking, the different decision problems presented below return either *yes* or *no*, but usually one is more interested in the corresponding *function problem* returning the actual objects responsible for a *yes* answer (like the extensions of a default theory, the stable models of a logic program, etc.). QUIP permits queries of this form, by engaging the so-called *detail mode* (see discussion below).
#### Abduction.
Classical abduction from a theory $`T`$ on $`V`$ may be defined as follows (???). Let $`HV`$ be a set of hypotheses, and let $`pV`$ be a distinguished atom. A subset $`EH`$ is an abductive explanation for $`p`$ from $`T`$ and $`H`$, if
$`TE`$ is consistent, and
$`TEp`$, i.e., $`TE`$ logically implies $`p`$.
An explanation $`E`$ is minimal, if no proper subset $`E^{}E`$ is an abductive explanation.
The following tasks are implemented in QUIP:
* Given a theory $`T`$, a set $`H`$ of variables, and an atom $`p`$. Is there a (minimal) abductive explanation $`EH`$ for $`p`$ from $`T`$ and $`H`$?
* The *relevance problem*: Given a theory $`T`$, a set $`H`$ of variables, an atom $`p`$, and some hypothesis $`hH`$. Is there a (minimal) abductive explanation $`EH`$ for $`p`$ from $`T`$ and $`H`$ containing $`h`$?
* The *necessity problem*: Given a theory $`T`$, a set $`H`$ of variables, an atom $`p`$, and some hypothesis $`hH`$. Does $`h`$ occur in every (minimal) abductive explanation $`EH`$ for $`p`$ from $`T`$ and $`H`$?
#### Autoepistemic logic.
The language of Moore’s autoepistemic logic (?) contains the modal operator $`L`$, where $`L\varphi `$ intuitively means that $`\varphi `$ is believed. By $`_L`$ we denote the language $``$ extended by $`L`$. Formulae $`L\varphi `$ are viewed as propositional variables, which are called modal atoms.
A *stable expansion* of an autoepistemic theory $`T_L`$ is a set $`E_L`$ such that
$$E=Th(T\{L\varphi \varphi E)\{\neg L\varphi \varphi E\}),$$
where $`Th()`$ is the classical consequence operator with respect to the extended language $`_L`$.
QUIP handles the following tasks:
* Given an autoepistemic theory $`T_L`$, is there a stable expansion $`E`$ of $`T`$?
* *Brave reasoning*: Given an autoepistemic theory $`T_L`$ and some formula $`\mathrm{\Phi }`$, is there a stable expansion $`E`$ of $`T`$ containing $`\mathrm{\Phi }`$?
* *Skeptical reasoning*: Given an autoepistemic theory $`T_L`$ and some formula $`\mathrm{\Phi }`$, is $`\mathrm{\Phi }`$ contained in every stable expansion $`E`$ of $`T`$?
For brave and skeptical reasoning, if the detail mode is engaged, QUIP returns *witnesses* corresponding to theses tasks: for brave reasoning, all stable expansions containing $`\mathrm{\Phi }`$ are returned, whereas for skeptical reasoning, all stable expansions *not* containing $`\mathrm{\Phi }`$ are returned.
#### Default logic.
A *default theory* is a pair $`\mathrm{\Delta }=(T,D)`$, where $`T`$ is a set of formulae and $`D`$ is a set of *defaults* of the form $`\frac{\alpha :\beta }{\gamma }`$.<sup>2</sup><sup>2</sup>2For simplicity, we omit multiple justifications here. Intuitively, the default is applied ($`\gamma `$ is concluded) if $`\alpha `$ is provable and the justification $`\beta `$ can be consistently assumed.
The semantics of $`\mathrm{\Delta }=(T,D)`$ is defined in terms of *extensions* (?). Following (?), extensions can be characterized thus. For any $`S`$, let $`D(S)`$ be the monotonic rules $`\{\frac{\alpha }{\gamma }\frac{\alpha :\beta }{\gamma }D,\neg \beta S\}`$. Then, $`E`$ is an extension of $`\mathrm{\Delta }`$ iff $`E=Th^{D(E)}(T)`$, where $`Th^{D(E)}(T)`$ is the set of all formulae derivable from $`T`$ using classical logic together with the rules from $`D(E)`$.
QUIP expresses the following reasoning tasks:
* Given a default theory $`\mathrm{\Delta }`$, is there an extension $`E`$ of $`\mathrm{\Delta }`$?
* *Brave reasoning*: Given a default theory $`\mathrm{\Delta }`$ and some formula $`\mathrm{\Phi }`$, is there an extension $`E`$ of $`\mathrm{\Delta }`$ containing $`\mathrm{\Phi }`$?
* *Skeptical reasoning*: Given a default theory $`\mathrm{\Delta }`$ and some formula $`\mathrm{\Phi }`$, is $`\mathrm{\Phi }`$ contained in every extension $`E`$ of $`\mathrm{\Delta }`$?
As for stable expansions, QUIP returns witnesses if the detail mode is engaged. Moreover, each of these reasoning tasks has been implemented in terms of two independent transformations: The first category of reductions is based on the characterization of extensions discussed above; the second category is based on a characterization of extensions using the notion of a *full set* (?). Interestingly, although the transformations based on the latter method are more succinct than the corresponding reductions of the first kind, they have often an inferior performance compared to the former transformations.
#### Disjunctive Logic Programming.
A *disjunctive logic program*, $`\mathrm{\Pi }`$, is a set of rules
$$r:H(r)P(r),N(r)$$
where $`H(r)`$ is a disjunction of variables, $`P(r)`$ is a conjunction of variables, and $`N(r)`$ is a conjunction of negated variables. A Herbrand interpretation $`I`$ of $`V`$ is a stable model of $`\mathrm{\Pi }`$ (??), if it is a minimal model (with respect to set-inclusion) of the program $`\mathrm{\Pi }^I`$ resulting from $`\mathrm{\Pi }`$ as follows: remove each rule $`r`$ such that $`Ia`$ for some $`\neg a`$ in $`N(r)`$, and remove $`N(r)`$ from all remaining clauses.
Similar to autoepistemic logic and default logic, QUIP handles the problem whether a given logic program has a stable model, as well as brave and skeptical reasoning.
#### Circumscription.
In contrast to the formalisms described above, propositional circumscription is already a quantified boolean formula, hence it does not require a separate reduction. So, QUIP can handle circumscription in a straightforward way.
In the propositional case, the parallel circumscription of a set of atoms $`P=\{p_1,\mathrm{},p_n\}`$ in a theory $`T`$, where the atoms $`Q`$ are fixed and the remaining atoms $`Z=\{z_1,\mathrm{},z_m\}=V(PQ)`$ may vary, is given by the following QBF $`\mathrm{𝐶𝐼𝑅𝐶}(T;P,Z)`$, cf. (?):
$$TP^{}Z^{}\left((T[P/P^{},Z/Z^{}](P^{}P))(PP^{})\right).$$
Here, $`P^{}=\{p_1^{},\mathrm{},p_n^{}\}`$ and $`Z^{}=\{z_1^{},\mathrm{},z_m^{}\}`$ are sets of new propositional variables corresponding to $`P`$ and $`Z`$, respectively, and $`T[P/P^{},Z/Z^{}]`$ results from $`T`$ by substitution of the variables in $`P^{}Z^{}`$ for those in $`PZ`$. QUIP implements circumscriptive inference of a formula $`\varphi `$ from $`T`$, which is expressed by the QBF
$$V(\mathrm{𝐶𝐼𝑅𝐶}(T;P,Z)\varphi ).$$
## Applying the System
### Methodology
One of the basic motivations for the development of QUIP was to make a rapid prototyping tool available, aimed for experimenting with different knowledge-representation formalisms. Accordingly, QUIP is designed in such a way that the important reasoning tasks corresponding to the different formalisms under consideration can be directly encoded as queries. So, the methodology of specifying queries suitable to be processed by QUIP is the same as the methodology of formalizing a particular problem with one of the formalisms implemented in QUIP (subject to the restriction, of course, that QUIP currently accepts only queries specified over a propositional language; but in a future version it is planned to extend the language to allow function-free formulae with variables as well.)
### Specifics
Let us illustrate how the queries of QUIP are structured. Generally, QUIP takes an input file as argument and writes the output to standard-out:
> quip input\_file
The input file comprises nonmonotonic theories and specifications of reasoning tasks. The format of the input uses two major concepts: *definitions* and *commands*. In the definitions, one can specify abductive theories, autoepistemic theories, default theories, logic programs, and propositional theories in general. The commands specify which reasoning tasks (with respect to the chosen formalism) have to be executed. Theories and formulas can be nested by using suitable names, so definitions can be edited rather conveniently. Further, commands can refer to specified theories without the need to describe them more than once. An input file can contain several definitions and commands, even referring to different formalisms.
In the following we describe some of these features using a simple example from default logic. Consider the following default theory $`\mathrm{\Delta }=(T,D)`$, representing the well-known Nixon-diamond:
$`T`$ $`=`$ $`\{\text{Republican},\text{Quaker}\};`$
$`D`$ $`=`$ $`\{{\displaystyle \frac{\text{Republican}:\neg \text{Pacifist}}{\neg \text{Pacifist}}},{\displaystyle \frac{\text{Quaker}:\text{Pacifist}}{\text{Pacifist}}}\}.`$
This default theory has two extensions:
> $`Th(\{\text{Republican},\text{Quaker}\}\{\neg \text{Pacifist}\})`$;
>
> $`Th(\{\text{Republican},\text{Quaker}\}\{\text{Pacifist}\})`$.
The extensions of this example can be computed by QUIP using the following input file, named nixon:
```
@SET DETAIL
T := Republican & Quaker
@D D := { Republican : !Pacifist | !Pacifist,
Quaker : Pacifist | Pacifist
}
@DL ( T ; @D D )
```
Executing the command
> quip nixon
results in the following output:
> Th( {(Republican)&(Quaker)} u {(Pacifist)} )
> Th( {(Republican)&(Quaker)} u {(!Pacifist)} )
The meaning of the different commands and definitions in the input file nixon can be explained as follows. First of all, the command @SET DETAIL invokes the detail mode, i.e., all extensions will be displayed. (Recall that QUIP admits the processing of both *decision problems* and *function problems*.) Choosing the command @UNSET DETAIL would have resulted in a simple yes/no answer.
The next tokens of the input file specify the constituents of the default theory $`\mathrm{\Delta }`$: T represents the background knowledge of $`\mathrm{\Delta }`$, and D contains the defaults. To avoid ambiguities, references to defaults always have to start with a special string “@D”. The logical operators are represented in the obvious way (“&” denotes conjunction, “!” negation, and “|” separates a default’s consequent from its justification).
The command @DL tells QUIP to compute the extensions of the specified default theory. To perform brave or skeptical reasoning, the input file nixon would be changed as follows:
```
@SET DETAIL
T := Republican & Quaker
@D D := { Republican : !Pacifist | !Pacifist,
Quaker : Pacifist | Pacifist
}
@SET BRAVE
@DL ( W ; @D D ) |= Pacifist
@SET SKEPTICAL
@DL ( W ; @D D ) |= Pacifist
```
The @SET-commands switches between the two reasoning modes; “|=” is the symbol standing for the respective default consequence relations. The first command checks whether *Pacifist* is a brave consequence of $`\mathrm{\Delta }`$, the second commands checks skeptical consequence of *Pacifist* from $`\mathrm{\Delta }`$. Observe that it is permitted that an input file contains a sequence of reasoning tasks.
The output corresponding to the first command will be
> Th( {(Republican)&(Quaker)} u {(Pacifist)} )
while the second command displays the extensions from which *Pacifist* does *not* follow:
> Th( {(Republican)&(Quaker)} u {(!Pacifist)} )
### Users and Usability
Obviously, proper usage of QUIP depends on a potential user’s ability to express a given problem in terms of the implemented formalisms. However, a particular advantage of QUIP is that it incorporates several different knowledge representation formalisms. Hence, users have a *choice* selecting among different (yet closely related) approaches, singling out that particular formalism best suited for a specific purpose, or choosing a method based on a user’s personal preference (e.g., because he/she understands that particular approach best). As well, the problem can be represented with different methods *simultaneously*, specifying the instances of the resultant formalizations in a single input file, which can then be processed by QUIP requiring only one initial execution command.
## Evaluating the System
### Benchmarks
There are different approaches how systems handling knowledge representation formalisms can be evaluated. One is to perform comparisons taking into account the representational power of the implemented formalisms. That is to say, under such a comparison, one chooses some “natural” problems, encodes it with respect to the specific methodologies associated with the implemented formalisms, and uses the resultant instances as queries of the respective systems. So, basically, different systems are compared on the basis of (possibly) different representations *of the same problem*. However, to achieve a fair comparison, it is necessary that the experimenter is able to encode the given problem “in the best possible way” with respect to the particular formalisms. Also, incompatibilities of the underlying formalisms render comparisons between different systems often difficult.
Another possibility is to compare systems on problem classes *common* to each of the considered systems. Such a comparison is appropriate if different implementations of *the same formalism* are evaluated.
Since QUIP incorporates a wide array of different knowledge representation methods, comparisons with other implementations can in general be achieved employing the latter procedure.
Accepted benchmarks for nonmonotonic theorem provers have been realized by the well-known TheoryBase system (?). This test-bed provides encodings of various graph problems in terms of default theories or equivalent logic programs. However, the generated problems are at most NP-hard (or co-NP-hard, depending on the reasoning task), and thus do not take full advantage of the expressibility supported by the nonmonotonic formalisms. The only practically applied benchmark test utilizing a $`\mathrm{\Sigma }_2^p`$-hard problem is the strategic companies example carried out for testing the system dlv (?).
Here, we propose a straightforward method how $`\mathrm{\Sigma }_2^p`$-hard benchmark problems for propositional nonmonotonic reasoning formalisms can be generated. The idea is to use the class of problems establishing the $`\mathrm{\Sigma }_2^p`$-hardness of a formalism. Recall from our previous discussion that $`\mathrm{\Sigma }_2^p`$-hardness of a nonmonotonic formalism is usually demonstrated by constructing a (polynomial) transformation $`𝒮`$ mapping instances $`I`$ of $`\text{qsat}_2`$ (the evaluation problem of QBFs having quantifier order $``$) into instances $`𝒮(I)`$ of the considered nonmonotonic reasoning task, nmp, such that $`I`$ is a yes-instance of $`\text{qsat}_2`$ iff $`𝒮(I)`$ is a yes-instance of nmp. Thus, in some sense, the class of problems $`𝒮(I)`$ represents “worst-case” examples for the problem nmp, and therefore is particularly useful estimating the performance of a theorem prover solving the task nmp. Moreover, these problems are easily scalable by parameterizing different instances of $`\text{qsat}_2`$. An added feature of QUIP is that these examples provide at the same time a simple method for *testing* whether the implementation works correct, because QUIP turns instances $`𝒮(I)`$ of nmp back to instances $`𝒯(𝒮(I))`$ of $`\text{qsat}_2`$, satisfying the condition that $`I`$ is a yes-instance of $`\text{qsat}_2`$ iff $`𝒯(𝒮(I))`$ is a yes-instance of $`\text{qsat}_2`$. The next subsection describes comparisons between QUIP and some state-of-the-art provers on the basis of these benchmark problems.
### Comparison
We compare the default-logic module of QUIP with DeReS (?) and the logic-programming module of QUIP with dlv (?), using the class of examples discussed above. Space limits preclude a discussion on the structure of these problems; details can be found in the relevant literature (e.g., (????)). We do not include a comparison with smodels (?) here, because that system is currently not designed to handle $`\mathrm{\Sigma }_2^p`$-problems (a comparison between QUIP, DeReS, dlv, smodels, and Theorist (?), using examples from TheoryBase and some abduction problems, is given in (?)).
Results of the comparison are given Table 1. All tests have been performed on a SUN ULTRA 60 with 256MB RAM; the run-time is measured in seconds with an upper limit of 90sec (i.e., instances requiring a longer period are not displayed). The first group of entries gives the results for the disjunctive logic programming test; the second group gives the results for the default logic test; and the final row contains some measurements using instances of the corresponding abductive problem class. The respective input-QBFs have been randomly generated and are parameterized by the number $`k`$ of existential quantifiers (the number of variables was held fixed and was set to 20). Although the given results represent only a small sample, they do indicate that our *ad hoc* implementation performs sufficiently well.
### Problem Size
It is rather obvious that QUIP cannot compete with state-of-the-art implementations like dlv or smodels in terms of problem size. These tools are highly optimized systems developed with a particular semantics in mind, whereas the purpose of QUIP is to provide a *uniform* method dealing with several knowledge representation tasks at the same time. Under this perspective, and taking into account that QUIP utilizes at present no optimizations whatsoever, our results demonstrate that implementing nonmonotonic reasoning formalisms using reductions to quantified boolean formulae is a feasible approach. Moreover, the modular architecture of QUIP allows an easy scalability and parallelization, by using, e.g., several QBF-provers simultaneously, each of which with its own optimization method.
|
warning/0003/quant-ph0003048.html
|
ar5iv
|
text
|
# On a quantum version of Shannon’s conditional entropy
## 1 Introduction
The concept of entropy plays a major role in thermodynamics and statistical mechanics. It serves to describe the behavior of macroscopic systems. The name “entropy” was introduced by Clausius (1865) and derives from $`ϵ\nu \tau \rho o\pi \iota \eta `$ “transformation”. It was von Neumann (1927 ), who generalized the classical expression of Boltzmann and Gibbs for the entropy to quantum mechanics by using the concept of what is now called a density matrix, also introduced quite generally by him in the same year . In the special context of radiation damping the density matrix was discovered independently by L. Landau and by F. Bloch , again in the same year (see also the citation in ). For a technical overview of the developments up to 1978 and with further historical references see . For recent expositions see . In the theory of dynamical systems entropy and the derived notion of topological entropy also plays an important role, see e.g. the contributions in .
In a seminal article Shannon (1948, ) introduced the concept of entropy into information theory. Roughly speaking a gain in information means a decrease in entropy. Shannon also provided the concept of conditional entropy. It is a measure how entropy is reduced given a preexisting knowledge. To the author’s best knowledge the first construction in quantum mechanics coming close to such a notion is due to E. Lieb (see also ). It involves tensor product structures and it was called a relative entropy in (but a conditional entropy in , p. 259). In view of recent developments in quantum computation and quantum coding (see for a concise account) it is highly desirable to have such a quantity at ones disposal. There is a construction of a non-commutative analogue of Shannon’s conditional entropy by Connes and Størmer and Connes, Narnhofer and Thirring (for an exposition and a discussion of further developments see e.g. ). More recently attempts have been made to construct a mutual information analogous to Shannon’s conditional entropy in the context of quantum error-correction. In two of these attempts , made independently, yielded the same quantity. The first article exhibits necessary and sufficient conditions for quantum error-correction to be possible in terms of the mutual information like the quantity given there, and a conjecture is made on its connection with quantum channel capacity, explored in more detail . The connection with channel capacity was also analyzed in . In its connection with entanglement is discussed. In yet another approach the starting point is one density matrix on a tensor product. The conditioning is then obtained by looking at the two density matrices in the two sub-systems resulting by taking the corresponding partial traces.
In this article we will propose a different candidate for a quantum mechanical conditional entropy $`S(\rho |\sigma )0`$, a function of two density matrices $`\rho `$ and $`\sigma `$ in a same Hilbert space and having the interpretation of the entropy of $`\rho `$ conditioned by the “knowledge” given by $`\sigma `$. For simplicity we will only discuss the finite dimensional case although an extension to the infinite dimensional case seems possible. If we view $`\rho `$ as the analogue of $`X`$ and $`\sigma `$ the analogue of $`Y`$ such that von Neumann’s entropy $`S(\rho )`$ is the analogue of Shannon’s entropy $`H(X)`$, then this conditional entropy shares several but not all properties of Shannon’s conditional entropy $`H(X|Y)`$ (see section 3 for a brief recapitulation of Shannon’s theory). In particular the “knowledge” of $`\sigma `$ reduces the entropy, i.e. the inequality $`S(\rho |\sigma )S(\rho )`$ holds. This corresponds exactly to Shannon’s famous inequality $`H(X|Y)H(X)`$ and was our main motivation for our construction. Also and again in analogy to the classical theory we wanted the conditioning to be given by a quantity on the same footing as the original density matrix, i.e. conditioning should also be given by a density matrix. If as in the classical case $`\sigma `$ contains no information, i.e. if it is a multiple of the identity such that $`S(\sigma )`$ is maximal, then $`S(\rho |\sigma )=S(\rho )`$. In contrast to the classical case $`H(X|X)=0`$, however, the relation $`S(\rho |\rho )=0`$ holds if and only if the non-zero eigenvalues of $`\rho `$ are non-degenerate. In particular $`S(\rho |\rho )=0`$ if $`\rho `$ is pure. We will not elaborate on the question, whether the failure of our $`S(\rho |\sigma )`$ to satisfy all corresponding classical properties, like this last property, is due to a fundamental difference of quantum and classical information theory. In particular we will not provide a more detailed quantum mechanical interpretation of $`S(\rho |\sigma )`$. Also so far we have not analyzed whether it may be used in the context of channel capacity. Rather we will argue that other quantum mechanical versions of conditional entropy, which share more properties with the classical counterpart $`H(X|Y)`$, do not exist.
The article is organized as follows. In section 2 we provide the construction of a quantum version $`S(\rho |\sigma )`$ of the conditional entropy and establish several properties. In section 3 and after a brief review of Shannon’s theory we compare this with Shannon’s conditional entropy. In section 4 we first present a list of desirable properties for a quantum version of conditional entropy given in terms of two density matrices. We then show that even parts of these desiderata can not be fulfilled simultaneously. In particular there is no version involving two density matrices and which reduces to the classical case, when these two density matrices commute. We will provide an alternative in terms of resolutions of the unit matrix in terms of orthogonal projections and which share more properties with the classical case. Briefly we will compare this ansatz with the algebraic constructions given by Connes and Størmer and Connes, Narnhofer and Thirring.
## 2 Construction of a quantum conditional entropy
Let $`\rho `$ be a density matrix on a finite dimensional Hilbert space $``$, i.e $`\rho 0`$ and $`\mathrm{Tr}\rho =1`$, where $`\mathrm{Tr}`$ denotes the canonical trace on $``$. We write $`\rho =_i\rho _iP_i`$ for the spectral representation of $`\rho `$ where the projections $`P_i0`$ are pairwise orthogonal ( i.e. $`P_iP_j=\delta _{ij}P_i,P_i=P_i^{}`$), such that $`\rho _i0`$, $`\rho _i\rho _j`$ for $`ij`$ and $`_iP_i=𝕀`$, where $`𝕀`$ is the identity operator on $``$. Thus $`\mathrm{Tr}\rho =_idimP_i\rho _i=1`$ with $`dimP=\mathrm{Tr}P=dimP`$ for any projection $`P`$. Here and in what follows projection operators are always understood to be orthogonal. With this notational convention the $`P_i`$ are canonically defined in terms of $`\rho `$. Since this fact will be crucial in what follows, let us briefly recall a standard proof. The eigenvalues $`\rho _i`$ (and their degeneracies $`(=dimP_i)`$) are of course uniquely determined by $`\rho `$ as solutions in $`\lambda `$ of the secular equation $`det(\lambda 𝕀\rho )=0`$, a basis independent relation, such that $`det(\lambda 𝕀\rho )=_i(\lambda \rho _i)^{dimP_i}`$. Order the $`\rho _i`$ in such a way that $`1\rho _1>\rho _2>\rho _3>\mathrm{}`$ . Then $`P_1=lim_n\mathrm{}(\rho /\rho _1)^n`$, $`P_2=lim_n\mathrm{}((\rho \rho _1P_1)/\rho _2)^n`$, etc.
The quantum mechanical entropy of $`\rho `$ is given as $`S(\rho )=_idimP_i\rho _i\mathrm{ln}\rho _i`$, which is continuous and concave in $`\rho `$ (for an account of sub-additivity and convexity properties of the entropy and related quantities see e.g. ). Let $`\sigma `$ be another density matrix on the same space $``$ with the spectral representation $`\sigma =_j\sigma _jQ_j`$ again written in a canonical way. We define the conditional entropy by
$`S(\rho |\sigma )`$ $`=`$ $`{\displaystyle \underset{j}{}}dimQ_j\sigma _jF(\rho ,Q_j)`$ (1)
$`=`$ $`{\displaystyle \underset{j}{}}\mathrm{Tr}Q_j\sigma F(\rho ,Q_j)`$
where
$$F(\rho ,Q)=\mathrm{Tr}(Q\rho Q\mathrm{ln}(Q\rho Q))+\mathrm{Tr}(Q\rho Q)\mathrm{ln}\mathrm{Tr}(Q\rho Q)$$
(2)
for any orthogonal projection $`Q`$. Since the $`Q_j`$’s and $`\sigma _j`$’s are well defined in terms of $`\sigma `$ and since trivially $`Q\rho Q0`$, $`S(\rho |\sigma )`$ is well defined. Also as usual in this context $`A\mathrm{ln}A`$ for any non-negative operator $`A`$ is defined in terms of the spectral representation of $`A`$ with the natural convention that $`x\mathrm{ln}x|_{x=0}=0`$. If $`Q\rho Q0`$ then also $`0\mathrm{Tr}Q\rho Q=\mathrm{Tr}Q\rho `$ and then we may write
$$F(\rho ,Q)=\mathrm{Tr}Q\rho S(\rho _Q)$$
(3)
with
$$\rho _Q=\frac{1}{\mathrm{Tr}(Q\rho )}Q\rho Q$$
(4)
being a density matrix. Actually we might use (3) instead of (2) as a definition for $`F(\rho ,Q)`$ with the convention, usually made in similar contexts (see e.g. ), that 0 times something undefined is 0. Relation (3) shows that $`F(\rho ,Q)0`$ for all $`\rho `$ and $`Q`$. Using (3) we may rewrite $`S(\rho |\sigma )`$ as
$$S(\rho |\sigma )=\underset{j:Q_j\rho Q_j0}{}\mathrm{Tr}Q_j\sigma \mathrm{Tr}Q_j\rho S(\rho _{Q_j}).$$
(5)
There is yet another way of writing $`F(\rho ,Q)`$. It uses the relative entropy $`0S_{rel}(A,B)=\mathrm{Tr}A(\mathrm{ln}A\mathrm{ln}B)\mathrm{}`$, which is defined for any $`A0`$ and $`B0`$. The relative entropy is lower semi-continuous in $`A`$ and jointly convex in $`A`$ and $`B`$, see e.g. . Obviously $`S_{rel}(\lambda A,\lambda B)=\lambda S_{rel}(A,B)`$ holds for any $`\lambda >0`$ and we have
$$F(\rho ,Q)=S_{rel}(Q\rho Q,\mathrm{Tr}(Q\rho Q)𝕀)$$
(6)
such that
$$S(\rho |\sigma )=\underset{j}{}\mathrm{Tr}Q_j\sigma S_{rel}(Q_j\rho Q_j,\mathrm{Tr}(Q_j\rho Q_j)𝕀)$$
(7)
It is instructive to compare $`S(\rho |\sigma )`$ with $`S(E_{\underset{¯}{Q}}(\rho ))`$ and which actually motivated our construction of $`S(\rho |\sigma )`$. $`E_{\underset{¯}{Q}}`$ is the linear map on the set of linear operators $`A`$ on $``$ given as $`E_{\underset{¯}{Q}}(A)=_jQ_jAQ_j`$ . The $`Q_j`$’s are as above, i.e a any set $`\underset{¯}{Q}=\{Q_j\}`$ of pairwise orthogonal nonzero projection operators with $`_jQ_j=𝕀`$ and which is called a resolution of the identity. $`E_{\underset{¯}{Q}}`$ is a conditional expectation (see e.g. ) with range being the $``$-algebra consisting of all linear operators which commute with all $`Q_i`$. In particular $`E_{\underset{¯}{Q}}`$ maps density matrices into density matrices. More precisely, let $`=()`$ be the $``$-algebra of all linear operators on $``$, which is (isomorhic to) a full matrix-algebra. Then $`E_{\underset{¯}{Q}}()`$ is a $``$-sub-algebra of $``$ and the direct sum of the $``$-sub-algebras $`Q_jQ_j=(Q_j)`$, which are (isomorphic to) full matrix algebras. Although any finite dimensional $``$-algebra is (isomorphic to) a direct sum of full matrix algebras, not all $``$-sub-algebras of $``$ are of the form $`E_{\underset{¯}{Q}}()`$ for a suitable $`\underset{¯}{Q}`$. As an example consider the algebra generated by $`𝕀`$ alone. It can easily be shown that any $``$-sub-algebra is of this form if and only if it contains a maximal abelian sub-algebra. Also from $`E_{\underset{¯}{Q}}()\underset{¯}{Q}`$ may be recovered. Indeed the $`Q_j`$’s are just the minimal self-adjoint idempotents (i.e. the orthogonal projections) in $`E_{\underset{¯}{Q}}()`$ and which are central. Also on the set of all spectral resolutions of the identity we introduce a partial ordering $``$ by setting $`\underset{¯}{P}\underset{¯}{Q}`$ if to each $`i`$ there is $`j(i)`$ (which is unique) such that $`P_iQ_{j(i)}`$. Note that each $`j`$ is of the form $`j=j(i)`$ for at least one $`i`$. Then in particular all $`P_i`$ commute with all $`Q_j`$. Also $`\underset{¯}{P}\{𝕀\}`$ holds for all $`\underset{¯}{P}`$. It is easy to see that $`\underset{¯}{P}\underset{¯}{Q}`$ if and only if $`E_{\underset{¯}{P}}()E_{\underset{¯}{Q}}()`$. With respect to these orderings $`\underset{¯}{P}`$ or equivalently $`E_{\underset{¯}{P}}()`$ is minimal if and only if each $`P_i`$ is one-dimensional. $`E_{\underset{¯}{P}}()`$ is then commutative with dimension equal to $`dim`$. To sum up, with respect to the partial ordering $``$ there is a unique maximal element but there are many minimal elements in the set of spectral resolutions $`\underset{¯}{P}`$.
Now one has the well known result $`S(E_{\underset{¯}{Q}}(\rho ))S(\rho )`$ (see e.g. for a direct proof and for the special case when $`dimQ_j=1`$ for all $`j`$. It is a special case of Uhlmann’s monotonicity theorem , see also ). It means that projective measurements increase entropy and compares with the inequality $`S(\rho |\sigma )S(\rho )`$ to be proven below. Its interpretation is that of a projective measurement described by the family $`\underset{¯}{Q}`$ of projections on a system given by $`\rho `$, but where we never learn of the result of the measurement. In contrast $`S(\rho |\sigma )`$ is interpreted as a set of projective measurements given by the projections $`Q_j`$, each performed with the probability $`dimQ_j\sigma _j`$, and where we learn of each outcome $`F(\rho ,Q_j)`$ separately. The sum in (1) and (5) then reflects the occurrence of a quantum decoherence. In other words one considers the family of density operators $`\rho _{Q_j},Q_j\rho Q_j0`$, takes their von Neumannn entropy and then forms the linear combination with the non-negative coefficients $`\mathrm{Tr}Q_j\sigma \mathrm{Tr}Q_j\rho `$.
By definition we have
$$F(\rho ,Q=𝕀)=S(\rho |\sigma =(1/dim𝕀)𝕀)=S(\rho ).$$
(8)
We consider this property to be necessary for any other sensible definition of a conditional entropy involving two density matrices. It holds for Shannon’s conditional entropy $`H(X|Y)`$ in the form $`H(X|Y)=H(X)`$ when $`Y`$ is the trivial partition (see section 3), which means that there is no gain in information, if $`Y`$ contains no information. We will return to this point in section 3.
Some additional remarks are in order. Since the quantity $`S(\rho |\sigma )`$ is supposed to be a quantum mechanical mechanical analogue of Shannon’s conditional entropy $`H(X,Y)`$, $`\rho `$ corresponds to $`X`$ and $`\sigma `$ to $`Y`$. In analogy to the classical case, where $`X`$ and $`Y`$ may be considered to be stochastic variables living on the same space, here the density matrices $`\rho `$ and $`\sigma `$ also live on the same space. Unfortunately with this correspondence $`S(\rho |\sigma )`$ does not reduce to the classical case when $`\rho `$ and $`\sigma `$ commute (see (33) and its discussion in section 3). As matter of fact, we shall argue in section 4 that a quantum conditional entropy with this property does not exist.
By construction we have the obvious invariance under unitary automorphisms
$$F(\rho ,Q)=F(U\rho U^1,UQU^1),$$
(9)
for any $`U𝒰()`$, the group of unitary operators in $``$. This relation (9) immediately implies
$$S(U\rho U^1|U\sigma U^1)=S(\rho |\sigma )$$
(10)
for all $`U`$. Relation (10) reflects the fact that $`S(\rho |\sigma )`$ is defined intrinsically and is in particular basis independent. Therefore this invariance property should also hold for any alternative, sensible definition of a quantum mechanical conditional entropy defined in terms of two density matrices. We shall comment on the classical analogue to (10) in section 4.
The next observation is also important. It is easy to see that $`F(\rho ,Q)`$ is continuous in $`\rho `$ and $`Q`$ by the same arguments used to prove continuity of $`S(\rho )`$. Therefore $`S(\rho |\sigma )`$ is also continuous in $`\rho `$ for fixed $`\sigma `$. However, $`S(\rho |\sigma )`$ is not continuous in $`\sigma `$ everywhere for all fixed $`\rho `$. It is continuous on the dense open subset where the eigenvalues of $`\sigma `$ are non-degenerate.In fact, it is zero there(see below). So this lack of continuity occurs where $`\sigma `$ has degenerate eigenvalues and is due to the fact that for $`Q=Q^{}+Q^{\prime \prime }`$ being the sum of two projections both $`0`$ and which are orthogonal to each other, i.e. $`Q^{}Q^{\prime \prime }=0`$, in general one has
$$dimQF(\rho ,Q)dimQ^{}F(\rho ,Q^{})+dimQ^{\prime \prime }F(\rho ,Q^{\prime \prime }).$$
(11)
To understand this consider the case when $`dim=2`$. Then $`S(\rho |\sigma )=S(\rho )`$ if $`\sigma =1/2𝕀`$ and $`S(\rho |\sigma )=0`$ otherwise. At the moment we do not know whether this lack of continuity of $`S(\rho |\sigma )`$ in $`\sigma `$ is a desirable feature or not, i.e whether this can be understood quantum mechanically, when we interpret $`S(\rho |\sigma )`$ as the entropy of $`\rho `$ conditioned by $`\sigma `$. Observe that a degeneracy typically occurs when a non-trivial symmetry is present. In other words there is then a non-trivial non-abelian subgroup $`𝒢=𝒢(\sigma )`$ of $`𝒰()`$ such that $`U\sigma U^1=\sigma `$ for all $`U𝒢`$. Note that $`𝒢`$ always contains a subgroup isomorphic to the abelian group $`U(N=dim)`$. In this picture a removal of degeneracies is related to a breakdown of symmetry, a familiar phenomenon in physics.
To proceed further, $`F(\rho ,Q)=0`$ if $`Q\rho Q=0`$, which can happen for $`Q0`$ only if $`\rho `$ has zero as an eigenvalue, i.e. if $`\rho `$ is not strictly positive. Then also $`(𝕀Q)\rho Q=Q\rho (𝕀Q)=0`$. In fact, by Schwarz inequality for any $`\psi ,\psi ^{}`$ we have
$$|<\psi ,Q\rho (𝕀Q)\psi ^{}>|||\rho ^{1/2}Q\psi ||||\rho ^{1/2}(𝕀Q)\psi ^{}||=0.$$
This also shows that $`Q\rho Q=0`$ is equivalent to $`Q\rho =0`$, which in turn by the self-adjointness of $`\rho `$ and $`Q`$ is equivalent to $`\rho Q=0`$. By the trivial identity
$$\rho =Q\rho Q+(𝕀Q)\rho Q+Q\rho (𝕀Q)+(𝕀Q)\rho (𝕀Q),$$
(12)
valid for all $`\rho ,Q`$, we therefore also have $`\rho =(𝕀Q)\rho (𝕀Q)`$ whenever $`Q\rho Q=0`$. Obviously (12) gives $`\mathrm{Tr}\rho =\mathrm{Tr}Q\rho Q+\mathrm{Tr}(𝕀Q)\rho (𝕀Q)`$ such that in particular the inequalities $`0\mathrm{Tr}Q\rho Q1`$ and $`0\mathrm{Tr}(𝕀Q)\rho (𝕀Q)1`$ hold for any $`\rho `$ and $`Q`$. By relation (3) we also have $`F(\rho ,Q)0`$ and hence $`S(\rho |\sigma )0`$ for all $`\rho ,Q`$ and $`\sigma `$. Now $`S(\rho _Q)=0,Q\rho Q0`$ holds if and only if $`\rho _Q`$ is a pure state, i.e. a one-dimensional projection. Also for $`dimQ=1`$ one always has $`Q\rho Q=(\mathrm{Tr}Q\rho Q)Q`$. We collect this observation in
###### Lemma 2.1.
$`F(\rho ,Q)=0`$ if and only if $`Q\rho Q`$ is a multiple of a one-dimensional projection.
This multiple is allowed to be zero. To characterize such $`Q`$’s fulfilling the conditions of the lemma, let $`P(\rho )0`$ be the projection operator onto the subspace corresponding to the non-zero eigenvalues, such that $`P(\rho )\rho =\rho =\rho P(\rho )`$ and in particular $`P(\rho )=𝕀`$ if $`\rho >0`$. Using the spectral representation of $`\rho `$ it is easy to see that $`Q\rho Q`$ is a multiple (possibly zero) of a one-dimensional projection if and only if $`Q`$ may be written as $`Q=Q^{}+Q^{\prime \prime }`$ with $`dimQ^{}1`$ and $`P(\rho )Q^{\prime \prime }=\rho Q^{\prime \prime }=0`$.
More generally consider the case where $`Q\rho Q=(\mathrm{Tr}(Q\rho Q)/dimQ^{})Q^{},Q0`$ holds for a suitable projection operator $`Q^{}`$ such that in particular $`0Q^{}Q`$ and $`Q^{}`$ is unique whenever $`Q\rho Q0`$. Then $`F(\rho ,Q)=(\mathrm{Tr}Q\rho Q)\mathrm{ln}dimQ^{}`$ and $`\rho _Q=(1/dimQ^{})Q^{}`$. This gives the
###### Lemma 2.2.
If all non-zero eigenvalues of $`\sigma `$ are non-degenerate then $`S(\rho |\sigma )=0`$ for all $`\rho `$. More generally if $`Q_j\rho Q_j`$ is a multiple (possibly zero) of some projection operator $`Q_j^{}(Q_j)`$ for all $`j`$ with $`\sigma _j>0`$, then
$$S(\rho |\sigma )=\underset{j}{}\mathrm{Tr}Q_j\rho \mathrm{Tr}Q_j\sigma \mathrm{ln}dimQ_j^{}.$$
(13)
Observe that $`S(\rho |\sigma )=0`$ for all pure states $`\sigma `$ and all $`\rho `$. If $`\rho `$ is pure then $`Q\rho Q`$ is always a multiple of a pure state for all $`Q`$. Therefore $`S(\rho |\sigma )=0`$ also holds for all $`\sigma `$ whenever $`\rho `$ is pure. Also if $`\rho \sigma =0`$ which is equivalent to $`\mathrm{Tr}\rho \sigma =0`$ and which can happen only if neither $`\rho `$ nor $`\sigma `$ is strictly positive, then again $`S(\rho |\sigma )=0`$. Sufficient (but not necessary) for the condition of Lemma 2.2 to hold is that to each $`j`$ with $`\sigma _j>0`$ there is $`i(j)`$ with $`Q_jP_{i(j)}`$. For these $`j`$’s $`Q_j^{}=Q_j,Q_j\rho Q_j=\rho _{i(j)}Q_j`$ and hence $`\mathrm{Tr}Q_j\rho =\rho _{i(j)}dimQ_j`$. This gives in particular
$$S(\rho |\rho )=\underset{i}{}\rho _i^2(dimP_i)^2\mathrm{ln}dimP_i.$$
(14)
Therefore the relation $`S(\rho |\rho )=0`$ holds if and only if all the non-zero eigenvalues of $`\rho `$ are non-degenerate, the if part being a special case of Lemma 2.2.
If in addition to the property $`Q_jP_{i(j)}`$ the density matrix $`\sigma `$ is such that
$$\underset{j:i(j)=i}{}\sigma _j(dimQ_j)^2\mathrm{ln}dimQ_j\rho _i(dimP_i)^2\mathrm{ln}dimP_i$$
holds for all $`i`$, then by (13) and (14) $`S(\rho |\sigma )S(\rho |\rho )`$. Note that this last condition is satisfied if
$$\underset{j:i(j)=i}{}\sigma _jdimQ_j\rho _idimP_i$$
holds since trivially $`dimQ_jdimP_{i(j)}`$.
We return to a discussion of the general properties of $`F(\rho ,Q)`$ and $`S(\rho |\sigma )`$. The first main result of this article shows that $`S(\rho |\sigma )`$ shares an important property with $`S(\rho )`$ (see e.g. for the classical and the quantum entropy and for Shannon’s conditional entropy and derived quantities).
###### Theorem 2.1.
$`F(\rho ,Q)`$ and $`S(\rho |\sigma )`$ are both concave in $`\rho `$.
Again we consider this property to be necessary for any sensible definition of a quantum conditional entropy. Like for the entropy $`S(\rho )`$ itself it states that mixing (in $`\rho `$) increases (conditional) entropy. On the other hand the case $`dim=2`$ discussed above shows that in general $`S(\rho |\sigma )`$ for fixed $`\rho `$ is neither convex nor concave in $`\sigma `$. Intuitively it would be desirable to have concavity with respect to $`\sigma `$ since mixing the conditioning should increase conditional entropy.
The proof follows easily from the presentation (6) and (7) and the known convexity property of the relative entropy.
The second main result of this article shows in particular that $`S(\rho |\sigma )`$ satisfies Shannon’s inequality.
###### Theorem 2.2.
The following inequalities hold for all density matrices $`\rho `$ and $`\sigma `$ in a fixed finite dimensional Hilbert space
$$0S(\rho |\sigma )S(\rho ).$$
(15)
If $`\rho >0`$ the last inequality is strict unless $`\sigma =(1/dim𝕀)𝕀`$.
The above comparison of $`S(\rho |\sigma )`$ with $`S(E_{\underset{¯}{Q}}(\rho )`$ suggests another definition of conditional entropy with the conditioning not given in terms of a density matrix $`\sigma `$ but rather only in terms of any resolution $`\underset{¯}{Q}`$ of the identity.
$$S(\rho |\underset{¯}{Q})=\underset{j}{}\frac{dimQ_j}{dim𝕀}F(\rho ,Q_j).$$
(16)
By (17) below we have
$$0S(\rho |\underset{¯}{Q})S(\rho ),$$
where the first inequality is an equality if $`dimQ_j=1`$ for all $`j`$ and the second one an equality if the spectral resolution is trivial, i.e. if $`\underset{¯}{Q}=\{𝕀\}`$. We note that in (16) any sequence of numbers $`\sigma _j^{}0`$ ( labeled in the same way as the $`Q_j`$’s) with $`_j\sigma _j^{}=1`$ and replacing $`dimQ_j/dim𝕀`$ would do equally well. But then we may combine and encode these data $`\underset{¯}{Q}`$ and $`\{\sigma \}`$ in the density matrix $`\sigma =_j\sigma _jQ_j`$ with $`\sigma _j=\sigma _j^{}/dimQ_j`$. If in addition all the $`\sigma _j`$’s are pairwise different, then by our discussion above they and the spectral resolution $`\underset{¯}{Q}`$ may be recovered from $`\sigma `$ and we are back to our construction $`S(\rho |\sigma )`$.
Due to the relation $`1=\mathrm{Tr}\sigma =_jdimQ_j\sigma _j`$ this second theorem is an immediate consequence of the following
###### Lemma 2.3.
For all $`\rho `$ and $`Q`$ the inequality
$$F(\rho ,Q)S(\rho )$$
(17)
holds. If $`\rho >0`$ this inequality is strict unless $`Q=𝕀`$.
Before we turn to a proof we make some remarks. We conjecture that in the general case $`\rho 0`$, the inequality (17) is strict unless $`Q\rho =\rho `$. This would imply that the second inequality in (15) is strict unless $`\sigma \rho =(\mathrm{Tr}\sigma \rho )\rho `$, which means the following. Any $`\sigma `$ with $`\sigma \rho =(\mathrm{Tr}\sigma \rho )\rho `$ is of the form $`\sigma =(\mathrm{Tr}\sigma \rho )P(\rho )+\sigma ^{}`$ with $`(𝕀P(\rho ))\sigma ^{}=\sigma ^{}`$.
Instead of $`F(\rho ,Q)`$ one might be tempted to consider instead the quantity (see (2))
$$\stackrel{~}{F}(\rho ,Q)=\mathrm{Tr}(Q\rho Q\mathrm{ln}Q\rho Q)0$$
and try to prove $`\stackrel{~}{F}(\rho ,Q)S(\rho )`$. Obviously we have $`\stackrel{~}{F}(\rho ,Q)F(\rho ,Q)`$. Consider, however, the case where $`dimQ=1`$ and $`\rho =P,dimP=1`$ (i.e. $`\rho `$ is pure) and with $`P`$ chosen such that $`QPQ=(\mathrm{Tr}QP)Q`$ satisfies $`0<\mathrm{Tr}PQ<1`$. Then $`0=S(\rho =P)<\stackrel{~}{F}(\rho =P,Q)`$. Furthermore one has $`F(\rho ,Q)S(\rho _Q)`$ when $`0<\mathrm{Tr}Q\rho Q(1)`$. But it does not make sense to replace $`F(\rho ,Q)`$ by $`S(\rho _Q)`$ as an alternative, since $`S(\rho _Q)`$ is only defined when $`Q\rho Q0`$. Even if $`Q\rho Q0`$, one does not have $`S(\rho _Q)S(\rho )`$ in general. To see this we will consider an example. For any $`0\psi `$ let $`P_\psi `$ be the 1-dim. projection onto the subspace spanned by $`\psi `$.
###### Example 2.1.
Let $`dim=4`$ with $`\psi _1,\psi _2,\psi _3,\psi _4`$ being an orthonormal basis. Let $`Q=P_{\psi _1}+P_{\psi _2}`$ be the 2-dim. projection onto the sub-space spanned by $`\psi _1`$ and $`\psi _2`$. Choose $`\rho (\varphi _1,\varphi _2)=\rho _1P_{\psi _1^{}}+\rho _2P_{\psi _2^{}},\rho _1+\rho _2=1`$ with
$`\psi _1^{}`$ $`=`$ $`\mathrm{cos}\varphi _1\psi _1+\mathrm{sin}\varphi _1\psi _3,`$
$`\psi _2^{}`$ $`=`$ $`\mathrm{cos}\varphi _2\psi _2+\mathrm{sin}\varphi _2\psi _4,\mathrm{cos}\varphi _10\mathrm{cos}\varphi _2.`$
Then
$$\rho (\varphi _1,\varphi _2)_Q=\frac{\mathrm{cos}^2\varphi _1\rho _1}{\mathrm{cos}^2\varphi _1\rho _1+\mathrm{cos}^2\varphi _2\rho _2}P_{\psi _1}+\frac{\mathrm{cos}^2\varphi _2\rho _2}{\mathrm{cos}^2\varphi _1\rho _1+\mathrm{cos}^2\varphi _2\rho _2}P_{\psi _2}.$$
Assume $`0<\rho _1<1`$ such that $`S(\rho (\varphi _1,\varphi _2))0`$ and choose $`\varphi _1`$ and $`\varphi _2`$ such that $`cos^2\varphi _1\rho _1=\mathrm{cos}^2\varphi _2\rho _2`$. This gives $`\rho (\varphi _1,\varphi _2)_Q=1/2Q`$ with $`S(\rho (\varphi _1,\varphi _2)_Q)=\mathrm{ln}2>S(\rho (\varphi _1,\varphi _2))`$ whenever $`\rho _11/2`$. On the other hand, some easy estimates show that indeed $`F(\rho (\varphi _1,\varphi _2),Q)S(\rho (\varphi _1,\varphi _2))`$ holds for all $`\varphi _1`$ and $`\varphi _2`$.
This example also shows that in general neither $`\rho _Q`$ nor $`S(\rho _Q)`$ for $`Q\rho Q=0`$ may be defined by a limiting procedure. In fact, we may let $`\varphi _1`$ and $`\varphi _2`$ tend to $`\pi /2`$ in such a way that $`\mathrm{cos}^2\varphi _2/\mathrm{cos}^2\varphi _1`$ tends to an arbitrary constant $`0`$ showing that in the limit for $`\rho (\varphi _1,\varphi _2)_Q`$ we may obtain an arbitrary convex combination of $`P_{\psi _1}`$ and $`P_{\psi _2}`$ and hence an arbitrary value between 0 and $`\mathrm{ln}2`$ for the entropy. By the convexity of the relative entropy we also have
$$F(\rho ,Q)+F(\rho ,𝕀Q)S(E_{\{Q,𝕀Q\}}(\rho )).$$
On the other hand, in general $`F(\rho ,Q)+F(\rho ,𝕀Q)`$ is in general not bounded above by $`S(\rho )`$. Indeed, consider the following
###### Example 2.2.
Let the set-up be as in Example 2.1. With repsect to this basis let
$$\rho (\kappa )=\frac{1}{4}\left(\begin{array}{cccc}1& 0& 0& \kappa \\ 0& 1& \kappa & 0\\ 0& \kappa & 1& 0\\ \kappa & 0& 0& 1\end{array}\right)$$
with $`0\kappa 1`$. The two two-fold degenerate eigenvalues are $`1/4(1\pm \kappa )`$. This gives $`F(\rho ,Q)+F(\rho ,𝕀Q)=\mathrm{ln}2`$ whereas $`S(\rho (\kappa ))=\mathrm{ln}21/2((1+\kappa )\mathrm{ln}(1+\kappa )+(1\kappa )\mathrm{ln}(1\kappa ))<\mathrm{ln}2`$, whenever $`0<\kappa `$.
The quantity
$$\mathrm{\Delta }S(\rho )=S(\rho )S(\rho |\rho )\underset{i}{}dimP_i\rho _i\mathrm{ln}(dimP_i\rho _i)$$
(18)
is of special interest. The inequality is a consequence of $`dimP_i\rho _i1`$ and again implies that the right hand side is non-negative and equal to zero if and only if $`\rho =(1/dim𝕀)𝕀`$ such that $`\mathrm{\Delta }S(\rho )>0`$ unless $`\rho =(1/dim𝕀)𝕀`$. In more detail the inequality in (18) may also be written as follows. Let $`S_{cl}(\underset{¯}{p})0`$ be the classical entropy for the probability distribution $`\underset{¯}{p}=(p_1,p_2,\mathrm{}p_n),p_k0,_kp_k=1`$
$$S_{cl}(\underset{¯}{p})=\underset{k=1}{\overset{n}{}}p_k\mathrm{ln}p_k.$$
such that in particular
$$S(\rho )=S_{cl}(\underset{¯}{p}(\rho ))$$
(19)
with
$$\underset{¯}{p}(\rho )=(\underset{dimP_1}{\underset{}{\rho _1,..,\rho _1}},\underset{dimP_2}{\underset{}{\rho _2,..,\rho _2}},\mathrm{}.).$$
(20)
(18) may now be rewritten as
$$0S_{cl}(\underset{¯}{\overset{^}{p}}(\rho ))\mathrm{\Delta }S(\rho )$$
(21)
with
$$\underset{¯}{\overset{^}{p}}(\rho )=(dimP_1\rho _1,dimP_2\rho _2,\mathrm{})$$
and where $`S_{cl}(\underset{¯}{\overset{^}{p}})=0`$ if and only if $`\rho `$ is a pure state. $`\mathrm{\Delta }S(\rho )`$ is easily shown to be continuous in $`\rho `$ and is obviously bounded above by $`\mathrm{ln}dim𝕀=\mathrm{ln}dim=S(\rho =(1/dim𝕀)𝕀)`$. It would be interesting to find its maximum in $`\rho `$ for fixed dimension of $`dim`$. Note also that
$$S(\rho )=S_{cl}(\underset{¯}{p}(\rho ))=S_{cl}(\underset{¯}{\overset{^}{p}}(\rho ))+\underset{i}{}dimP_i\sigma _i\mathrm{ln}dimP_iS_{cl}(\underset{¯}{\overset{^}{p}}(\rho ))$$
(22)
with equality if and only if $`dimP_i=1`$ for all $`i`$ with $`\sigma _i>0`$. We will discuss $`\mathrm{\Delta }S(\rho )`$ below when we compare $`S(\rho |\sigma )`$ with Shannon’s conditional entropy.
We turn to the proof of (17). First recall that $`F(\rho ,Q)`$ is continuous in $`\rho `$ (and $`Q`$). Hence it suffices to consider the case $`\rho >0`$ which implies that $`Q\rho Q0`$ for all $`Q0`$. Since $`F(\rho ,Q)=0`$ for $`Q=0`$ and $`dimQ=1`$ and since $`F(\rho ,𝕀)=S(\rho )`$ it suffices to consider the case $`1<dimQ<dim𝕀`$.
Now $`𝒰()`$ operates transitively and continuously on the Grassmannian of all $`n`$-dimensional subspaces of $``$, $`(1ndim)`$. For each $`n`$ this space is therefore compact and homeomorphic to the set of all projections of dimension $`n`$. Obviously on this set $`𝒰()`$ operates, again continuously, via $`U:PUPU^1`$. By (9)
$$F_n(\rho )=\underset{Q:dimQ=n}{sup}F(\rho ,Q)=\underset{U:U𝒰()}{sup}F(\rho ,UQ_0U^1)=\underset{U:U𝒰()}{sup}F(U\rho U^1,Q_0),$$
(23)
which is finite for each $`n`$. Here $`Q_0`$ is any orthogonal projection with $`dimQ_0=n`$. In particular we may choose $`Q_0`$ such that $`F_n(\rho )=F(\rho ,Q_0)`$. Consider the one-parameter unitary subgroup $`U(t)=\mathrm{exp}(itK)`$, where $`K`$ is an arbitrary self-adjoint operator on $``$. Then we must have $`f_K(t)=F(U(t)\rho U(t),Q_0)F(\rho ,Q_0)=f_K(t=0)`$ for all $`t`$ and all s.a. $`K`$. Now it is well known that for any one parameter family of strictly positive operators $`A(t)`$ which is differentiable in $`t`$ one has
$$\frac{d}{dt}\mathrm{Tr}(A(t)\mathrm{ln}A(t))=\mathrm{Tr}((𝕀+\mathrm{ln}A(t))\frac{d}{dt}A(t)).$$
Recalling the assumption $`\rho >0`$ such that $`Q_0\rho Q_0>0`$ when restricted to the subspace $`Q_0`$, it is easy to see that $`f_K(t)`$ is also differentiable in $`t`$ at $`t=0`$ and
$$\begin{array}{ccc}\frac{d}{dt}f_K(t)|_{t=0}& =& i\mathrm{Tr}((𝕀+\mathrm{ln}Q_0\rho Q_0)Q_0[K,\rho ]Q_0)\\ & & +i(1+\mathrm{ln}\mathrm{Tr}(Q_0\rho Q_0))\mathrm{Tr}Q_0[K,\rho ]Q_0\\ & =& i\mathrm{Tr}([\rho ,\mathrm{ln}\mathrm{Tr}(Q_0\rho Q_0)Q_0Q_0(\mathrm{ln}Q_0\rho Q_0)Q_0]K).\end{array}$$
(24)
By definition of $`Q_0`$ we must have $`d/dtf_K(t=0)=0`$ for all $`K`$. But then (24) implies that $`\rho `$ commutes with $`B=Q_0B=BQ_0`$ given as
$$B=\mathrm{ln}\mathrm{Tr}(Q_0\rho Q_0)Q_0Q_0(\mathrm{ln}Q_0\rho Q_0)Q_0.$$
This in turn implies that $`\rho `$ commutes with $`Q_0`$ itself, which is easy to see. Indeed, use the spectral representation $`Q_0\rho Q_0=_k\rho _k^{}Q_k^{}`$ with $`Q_k^{}Q_0,dimQ_k^{}=1`$ and $`_kQ_k^{}=Q_0`$ to write $`B`$ as
$$B=\underset{k}{}(\mathrm{ln}(\underset{l}{}\rho _l^{})\mathrm{ln}\rho _k^{})Q_k^{}.$$
Now write any $`\psi Q_0`$ as $`\psi =_ka_k\psi _k`$, where $`\psi _k`$ is a unit vector in $`Q_k^{}`$. Set
$$\varphi =\underset{k}{}\frac{a_k}{(\mathrm{ln}(_l\rho _l^{})\mathrm{ln}\rho _k^{})}\psi _kQ_0.$$
$`\varphi `$ is well defined since $`_l\rho _l^{}\rho _k^{}`$ for every $`k`$. This follows from our assumption $`n>1`$, the fact that $`\mathrm{ln}x`$ is strictly monotonic in $`x`$ and that $`\rho _k^{}>0`$ for all $`k`$, since $`Q_0\rho Q_0`$ when restricted to $`Q_0`$ is strictly positive. By construction $`\psi =B\varphi `$ such that $`\rho \psi =\rho B\varphi =B\rho \varphi =Q_0B\rho \varphi Q_0`$. Thus $`\rho `$ leaves $`Q_0`$ invariant and hence commutes with $`Q_0`$, as was claimed. But then we have $`\rho =Q_0\rho Q_0+(𝕀Q_0)\rho (𝕀Q_0)`$ which implies
$$S(\rho )=\mathrm{Tr}(Q_0\rho Q_0\mathrm{ln}Q_0\rho Q_0)\mathrm{Tr}((𝕀Q_0)\rho (𝕀Q_0)\mathrm{ln}(𝕀Q_0)\rho (𝕀Q_0)).$$
This gives
$`S(\rho )`$ $`=F(\rho ,Q_0)`$ $`\mathrm{Tr}((𝕀Q_0)\rho (𝕀Q_0)\mathrm{ln}(𝕀Q_0)\rho (𝕀Q_0))`$ (25)
$`\mathrm{Tr}Q_0\rho Q_0\mathrm{ln}\mathrm{Tr}Q_0\rho Q_0.`$
The two last terms in (25), however, are non-negative. This concludes the proof of the claim (17). To prove the second part of Lemma 2.3, we observe that the last two terms in (25) vanish exactly when $`(𝕀Q_0)\rho (𝕀Q_0)=0`$. But this contradicts the assumption $`\rho >0`$ and $`dimQ_0<dim𝕀`$, the case $`Q=𝕀`$ having been discussed previously. This completes the proof of Lemma 2.3.
## 3 Comparison with the classical case
In this section we provide a comparison with the classical theory of Shannon (see and for expositions e.g. ). For the convenience of the reader and in order to establish notation we recall the basic facts. Let $`\{\mathrm{\Omega },\mu \}`$ be a probability space. Furthermore let $`X=\{X_\alpha \}`$ and $`Y=\{Y_\beta \}`$ be any two partitions (up to measure zero) of $`\mathrm{\Omega }`$ into disjoint subsets of non-zero measure. For simplicity we will assume these partitions to be finite, i.e. we choose the indices $`\alpha `$ and $`\beta `$ to be in the range $`1\alpha n,1\beta m`$. Set $`\underset{¯}{p}(X)=\{p_\alpha \}`$ with $`p_\alpha =\mu (X_\alpha )>0`$ and $`\underset{¯}{p}(Y)=\{q_\beta \}`$ with $`q_\beta =\mu (Y_\beta )>0`$ such that $`_\alpha p_\alpha =1`$ and $`_\beta q_\beta =1`$. Here and in what follows $`\alpha `$ is an index referring to $`X`$ and $`\beta `$ to $`Y`$. Then $`H(X)=_\alpha p_\alpha \mathrm{ln}p_\alpha 0`$ and similarly $`H(Y)=_\beta q_\beta \mathrm{ln}q_\beta 0`$ is Shannon’s entropy. Actually Shannon used $`\mathrm{log}_2`$ instead of $`\mathrm{ln}`$ adapting to the situation where information is coded in bits, but this is not relevant for our purpose. Since $`H(X)=S_{cl}(\underset{¯}{p}(X))`$ this concept of information theory relates to the concept of entropy in classical statistical mechanics. Shannon’s conditional entropy is now given as follows. Let
$$p_{\alpha |\beta }=\frac{\mu (X_\alpha Y_\beta )}{\mu (Y_\beta )},q_{\beta |\alpha }=\frac{\mu (Y_\beta X_\alpha )}{\mu (X_\alpha )}$$
be conditional probabilities associated to $`X`$ and $`Y`$ (i.e. $`p_{\alpha |\beta }`$ is the probability that $`X_\alpha `$ will happen, given that $`Y_\beta `$ has happened). Obviously
$$p_{\alpha |\beta }q_\beta =q_{\beta |\alpha }p_\alpha (=\mu (A_\alpha B_\beta ))$$
(26)
for all $`\alpha ,\beta `$, which is called Bayes rule for $`p_{\alpha |\beta }`$ and $`q_{\beta |\alpha }`$. Let $`\underset{¯}{p}_\beta =(p_{1|\beta },p_{2|\beta },..,p_{n|\beta })`$ and $`\underset{¯}{q}_\alpha =(q_{1|\alpha },q_{2|\alpha },\mathrm{}.q_{m|\alpha })`$, such that
$$\underset{¯}{p}=\underset{\beta =1}{\overset{m}{}}q_\beta \underset{¯}{p}_\beta ,\underset{¯}{q}=\underset{\alpha =1}{\overset{n}{}}p_\alpha \underset{¯}{q}_\alpha .$$
(27)
Shannon’s conditional entropy is now defined as
$$H(X|Y)=\underset{\beta =1}{\overset{m}{}}q_\beta S_{cl}(\underset{¯}{p}_\beta )$$
(28)
and it satisfies
$$0H(X|Y)H(X).$$
(29)
We observe that the second inequality, called Shannon’s inequality, is a consequence of the concavity of the function $`\underset{¯}{p}S_{cl}(\underset{¯}{p})`$ and (27). It states that on average information on $`X`$ is gained if $`Y`$ is known. Also $`0H(X,Y)=H(Y)+H(X|Y)`$ is symmetric in $`X`$ and $`Y`$ and satisfies
$$H(Y)H(X,Y)H(X)+H(Y).$$
(30)
Actually $`H(X,Y)=H(XY)`$, where $``$ denotes the join of two partitions. The inequalities in (29) and (30) turn into equalities if the following conditions hold. $`X`$ and $`Y`$ are said to be independent if $`p_{\alpha |\beta }=p_\alpha `$ holds for all $`\alpha `$ and $`\beta `$. This means that $`\underset{¯}{p}_\beta `$ is actually independent of $`\beta `$ and equals $`\underset{¯}{p}`$ and $`\underset{¯}{q}_\alpha `$ is independent of $`\alpha `$ and equals $`\underset{¯}{q}`$. In particular $`S_{cl}(\underset{¯}{p}_\beta )=S_{cl}(\underset{¯}{p})`$ holds for all $`\beta `$ and $`S_{cl}(\underset{¯}{q}_\alpha )=S_{cl}(\underset{¯}{q})`$ for all $`\alpha `$. The second inequality in (29) and the second inequality in (30) (which are equivalent) are now equalities if and only if $`X`$ and $`Y`$ are independent. It follows from the fact that $`S_{cl}(\underset{¯}{p})`$ is strictly concave in $`\underset{¯}{p}`$. Secondly $`X`$ is called a consequence of $`Y`$ if to each $`\alpha `$ there is $`\beta (\alpha )`$ such that $`p_{\alpha |\beta (\alpha )}=1`$. So this means that $`p_{\alpha |\beta }=0`$ for all $`\beta \beta (\alpha )`$ and hence $`S_{cl}(\underset{¯}{p}_\beta )=0`$ for all $`\beta `$. Therefore the first inequality in (29) and equivalently the first inequality in (30) are equalities if and only if $`X`$ is a consequence of $`Y`$. In particular
$$H(X|X)=0,$$
(31)
i.e. $`H(X,X)=H(X)`$.
With this brief review of Shannon’s theory we turn to a comparison with our quantum mechanical construction. Obviously (29) corresponds to (15) when we let $`X`$ correspond to $`\rho `$ and $`Y`$ to $`\sigma `$. Note, however, the difference between (31) and (14). Moreover for the quantity $`S(\rho ,\sigma )=S(\sigma )+S(\rho |\sigma )`$ we have the inequalities
$$S(\sigma )S(\rho ,\sigma )S(\rho )+S(\sigma ),$$
(32)
which correspond to (30). $`S(\rho ,\sigma )`$ is in general not symmetric in $`\rho `$ and $`\sigma `$ . To see this consider commuting $`\rho `$ and $`\sigma `$. Then we have
$$S(\rho |\sigma )=\underset{j,i}{}dimQ_j\sigma _j\rho _idim(P_iQ_j)\mathrm{ln}\frac{\rho _i}{\mathrm{Tr}(\rho Q_j)}.$$
(33)
We remark that if $`\mathrm{Tr}(\rho Q_j)=0`$ for a fixed $`j`$ then $`\mathrm{Tr}(P_iQ_j)=0`$ for all $`i`$. Also (14) is a special case of (33). (33) shows that even in the commutative case $`S(\rho ,\sigma )`$ is not symmetric in $`\rho `$ and $`\sigma `$. So this implies that in the commutative case $`S(\rho |\sigma )`$ does not reduce to $`H(X|Y)`$ for any choice of $`X=X(\rho )`$ and $`Y=Y(\sigma )`$ with $`H(X)=S(\rho )`$ and $`H(Y)=S(\sigma )`$. This lack of symmetry of $`S(\rho ,\sigma )`$ is in contrast to the symmetry of its classical counterpart $`H(X,Y)`$, which has an important interpretation. The relation $`H(X,Y)=H(Y,X)`$ is equivalent to $`H(Y)+H(X|Y)=H(X)+H(Y|X)`$, a consequence of Bayes rule. But this means that on average the information on $`Y`$ plus the information on $`X`$ given $`Y`$ is equal to the information on $`X`$ plus the information on $`Y`$ given $`X`$. It would be interesting to see whether this failure of symmetry for $`S(\rho ,\sigma )`$ has a sensible interpretation in the context of the familiar Alice and Bob set-up in quantum information theory, see e.g. .
Finally consider
$$0S(\rho ||\sigma )=S(\rho )+S(\sigma )S(\rho ,\sigma )=S(\rho )S(\rho |\sigma )S(\rho )$$
(34)
which corresponds to
$$0I(X||Y)=H(X)+H(Y)H(X,Y)=H(X)H(X|Y).$$
On average $`0I(X||Y)H(X)`$ gives the information gain on $`X`$ when knowing $`Y`$. Thus if there is no information content at all in $`Y`$, i.e. if $`Y`$ is the trivial partition $`\{\mathrm{\Omega }\}`$, then there is no information gain in $`X`$
$$I(X||Y=\{\mathrm{\Omega }\})=0.$$
(35)
Thus (35) corresponds to (8) when rewritten as $`S(\rho ||\sigma =(1/dim𝕀)𝕀)=0`$. Therefore we also interpret the quantum mechanical analogue $`S(\rho ||\sigma )`$ as a quantum information gain for $`\rho `$ given $`\sigma `$ and which by (34) can be at most $`S(\rho )`$. In particular the gain is maximal for all $`\rho `$, if all non-zero eigenvalues of $`\sigma `$ are non-degenerate. The gain is also maximal if $`\rho \sigma =0`$, since then $`S(\rho |\sigma )=0`$, see Lemma 2.2 and the remark thereafter.
Finally $`\mathrm{\Delta }S(\rho )`$ (see (18)) corresponds to $`I(X||X)`$ and describes the situation where $`\rho `$ is conditioned on itself, $`\sigma =\rho `$. Then by (21) there is non-zero information gain unless $`\rho `$ is pure (and then a gain is not necessary). In contrast to the classical situation, $`I(X||X)=H(X)`$, which gives complete information gain when $`X`$ is conditioned on itself, there is complete information gain in the quantum case, $`\mathrm{\Delta }S(\rho )=S(\rho )`$, if and only if all non-zero eigenvalues of $`\sigma `$ are non-degenerate.
## 4 Attempts of alternative constructions
We conclude by addressing the natural question whether there is a quantity $`S^\mathrm{?}(\rho |\sigma )`$ which shares more properties with Shannon’s conditional entropy than the $`S(\rho |\sigma )`$ we have given. More precisely and by the arguments given in the preceding sections it would be desirable for $`S^\mathrm{?}(\rho |\sigma )`$ to have (most of) the following properties
1. Invariance under the group $`𝒰()`$: $`S^\mathrm{?}(U\rho U^1|U\sigma U^1)=S^\mathrm{?}(\rho |\sigma )`$ for all $`U𝒰()`$ (compare (10)).
2. Bounds: $`0S^\mathrm{?}(\rho |\sigma )S(\rho )`$ for all $`\rho `$ and $`\sigma `$ with $`S^\mathrm{?}(\rho |\rho )=0`$ and
$`S^\mathrm{?}(\rho |\sigma =(1/dim𝕀)𝕀)=S(\rho )`$.
3. Classical equivalence with Shannon’s conditional entropy.
4. Symmetry: $`S^\mathrm{?}(\rho ,\sigma )=S(\sigma )+S^\mathrm{?}(\rho |\sigma )`$ is symmetric in $`\rho `$ and $`\sigma `$.
5. Continuity of $`S^\mathrm{?}(\rho |\sigma )`$ in $`\rho `$ and in $`\sigma `$.
6. Concavity of $`S^\mathrm{?}(\rho |\sigma )`$ in $`\rho `$ and $`\sigma `$.
Note that $`S(\rho |\sigma )`$ fulfills condition 1, condition 2 apart from the property $`S(\rho |\rho )=0`$, condition 5 up to a set of measure zero and condition 6 only with respect to $`\rho `$.
Both the equality requirements of condition 2 can never be satisfied simultaneously. Indeed, with the choice $`\rho =\sigma =1/dim𝕀`$ we should have both $`S(1/dim𝕀|1/dim𝕀)=0`$ and $`S(1/dim𝕀|1/dim𝕀)=\mathrm{ln}dim`$. Also the condition $`S(\rho |\rho )=0`$ combined with $`S(\rho |\sigma )0`$ is incompatible with concavity of $`S(\rho |\sigma )`$ in $`\rho `$ (condition 6). In fact, let $`\rho =\lambda \rho _1+(1\lambda )\rho _2,\mathrm{\hspace{0.17em}0}<\lambda <1`$. But this gives $`0=S(\rho |\rho )\lambda S(\rho _1|\rho )+(1\lambda )S(\rho _2|\rho )`$. Hence $`S(\rho _1|\rho )=0`$ for all $`\rho _1`$ for which there is $`\lambda >0`$ with $`\lambda \rho _1<\rho `$. This condition is fulfilled for all $`\rho _1`$, whenever $`\rho >0`$ (I owe these observations to H. Narnhofer).
Next let us look at the condition 3, by which we mean the situation where $`\rho `$ and $`\sigma `$ commute such that $`S(\rho )=H(X),S(\sigma )=H(Y)`$ and $`S^\mathrm{?}(\rho |\sigma )=H(X|Y)`$ holds for suitable $`X=X(\rho )`$ and $`Y=Y(\sigma )`$. Also the dependence of $`X(\rho )`$ and $`Y(\sigma )`$ on $`\rho `$ and $`\sigma `$ respectively should be non-trivial w.r.t. their eigenvalues. In particular condition 3 means that the symmetry condition 4 must hold at least when $`\rho `$ and $`\sigma `$ commute. In view of the destruction of quantum coherence when measurements are performed and due to the occurrence of the sum by which $`S^\mathrm{?}(\rho ,\sigma )`$ is defined, it is unclear to the author whether the symmetry condition 4 also should hold for non-commuting $`\rho `$ and $`\sigma `$ (see below , however, a construction of conditional entropy in terms of spectral resolutions of the identity below). It is natural to make the assumption on $`X(\rho )`$, that $`\mu (X_k)=p_k(\rho )`$, see (19). Then it may be shown that the continuity condition and the classical equivalence condition are not compatible. The concavity condition in $`\sigma `$ is at least intuitively desirable since taking convex combinations decreases conditioning, i.e. increases uncertainty, and hence should increase conditional entropy.
We would also like to point out another difference between the classical and the quantum case in the way we have presented it so far. In the classical case the conditioning $`Y`$ is trivial when $`Y=\{\mathrm{\Omega }\}`$, which means no information content and for which we have $`H(Y)=0`$. Within the context of density matrices the only sensible candidate for a trivial conditioning is $`\sigma =1/dim𝕀`$, since this is the density matrix with no information content. Its von Neumann entropy, however, is maximal. Recall that we used this quantum notion of trivial conditioning in our discussion of the inequality $`S(\rho |\sigma )S(\rho )`$ (see also the discussion following (35)). We note that several authors consider von Neumann’s entropy not to be a good generalization of classical entropy (see e.g. , page 141). In fact, in classical theory finer partitions give rise to higher uncertainty and hence to larger classical entropy. This was the reason for the algebraic approach of Connes and Størmer and of Connes, Narnhofer and Thirring, in which a classical finer partitioning corresponds to a larger algebra. In particular the larger the algebra, the larger the entropy and similarly the larger the conditioning algebra the larger the conditional entropy.
We claim, however, that there is a way to reconcile this with von Neumann’s entropy. Indeed, given a quantum system in the state $`\rho `$, the measurements one can perform without disturbing $`\rho `$ are given by the observables (i.e. the self-adjoint operators) in $`𝒜(\rho )`$, which by definition is the $``$-sub-algebra of $``$ consisting of all elements in $``$ which commute with $`\rho `$. In particular $`𝒜(\rho =1/dim𝕀)=`$. In this sense again larger uncertainties correspond to larger algebras. In other words, the larger the entropy the more measurements on can perform without disturbing the system in the given state $`\rho `$. To be more precise, we introduce a partial ordering $``$ on the set of all density matrices (which differs from the one introduced by Uhlmann, see e.g. ). By definition $`\rho \sigma `$ ($`\sigma `$ is more mixed than $`\rho `$), if and only if a) $`\underset{¯}{P}\underset{¯}{Q}`$ and b) $`\mathrm{Tr}\rho Q_j=\mathrm{Tr}\sigma Q_j=\sigma _j\mathrm{Tr}Q_j`$ holds for all $`j`$. It is easy to see that $`\rho \sigma `$ and $`\sigma \tau `$ implies $`\rho \tau `$ and that $`\rho 1/dim𝕀`$ and $`\rho \rho `$ holds for all $`\rho `$. So whenever $`\rho \sigma `$ then condition a) implies $`𝒜(\rho )𝒜(\sigma )`$ and a) and b) combined imply $`S(\rho )S(\sigma )`$ by the concavity of the von Neumann entropy. Note, however, that the correspondence between $`\rho `$ and $`𝒜(\rho )`$ is not one-to-one. In fact, $`𝒜(\rho )`$ only depends on the spectral resolution of the identity $`\underset{¯}{P}=\underset{¯}{P}(\rho )`$ associated to $`\rho `$ and not on the eigenvalues $`\rho _i`$ of $`\rho `$. Indeed, one has $`𝒜(\rho )=E_{\underset{¯}{P}(\rho )}()`$, as one may easily verify.
Returning to our discussion of conditions 1-6, there is a way out, however, if one considers spectral resolutions of the identity $`\underset{¯}{P}`$ instead of density matrices. It works as follows. First observe that the actual choice of the probability space $`\{\mathrm{\Omega },\mu \}`$ for Shannon’s theory is irrelevant. What is relevant are the the sets of non-negative numbers $`\underset{¯}{p}=\{p_\alpha \},\underset{¯}{q}=\{q_\beta \},`$ $`\underset{¯}{p}\underset{¯}{q}=\{p_{\alpha |\beta }\}`$ and $`\underset{¯}{q}\underset{¯}{p}=\{q_{\beta |\alpha }\}`$ subject to the following conditions of which the last one is Bayes rule
$$\underset{\alpha }{}p_\alpha =\underset{\beta }{}q_\beta =1,\underset{\beta }{}p_{\alpha |\beta }q_\beta =p_\alpha ,\underset{\alpha }{}q_{\beta |\alpha }p_\alpha =q_\beta ,p_{\alpha |\beta }q_\beta =q_{\beta |\alpha }p_\alpha .$$
(36)
Note that then
$`{\displaystyle \underset{\alpha }{}}p_{\alpha |\beta }`$ $`=`$ $`{\displaystyle \frac{1}{q_\beta }}{\displaystyle \underset{\alpha }{}}q_{\beta |\alpha }p_\alpha =1`$
$`{\displaystyle \underset{\beta }{}}q_{\beta |\alpha }`$ $`=`$ $`{\displaystyle \frac{1}{p_\alpha }}{\displaystyle \underset{\beta }{}}p_{\alpha |\beta }q_\beta =1.`$
We consider these conditions (36), which mean independence of a particular realization of partitions $`X`$ and $`Y`$ on a probability space, the classical analogue of the relation (10). Setting $`p_{\alpha ,\beta }=p_{\alpha |\beta }q_\beta `$ and $`q_{\beta ,\alpha }=q_{\beta |\alpha }p_\alpha `$, Bayes rule gives $`p_{\alpha ,\beta }=q_{\beta ,\alpha }`$. We will therefore write $`H(X|Y)=H(\underset{¯}{p}|\underset{¯}{q})`$ by a slight abuse of notation since all the data $`\underset{¯}{p},\underset{¯}{q},\underset{¯}{p}\underset{¯}{q}`$ and $`\underset{¯}{q}\underset{¯}{q}`$ in (36) are necessary for a specification of $`H(X|Y)`$. But given these data it makes sense to say that $`\underset{¯}{p}`$ is a consequence of $`\underset{¯}{q}`$ or that $`\underset{¯}{p}`$ and $`\underset{¯}{q}`$ are independent.
Now let $`\tau =1/dim\mathrm{Tr}`$ denote the normalized trace, i.e. $`\tau (𝕀)=1`$. For any two spectral resolutions $`\underset{¯}{P}`$ and $`\underset{¯}{Q}`$ let $`p_i=\tau (P_i),q_j=\tau (Q_j),p_{i|j}=\tau (P_iQ_j)/\tau (Q_j),q_{j|i}=\tau (Q_jP_i)/\tau (P_i)`$. Note that by definition all $`P_i`$ and all $`Q_j`$ are non-zero projections. The conditions (36) are obviously satisfied. We then set $`H(\underset{¯}{P})=S_{cl}(\underset{¯}{p}),H(\underset{¯}{Q})=S_{cl}(\underset{¯}{q})`$ ,such that $`H(\underset{¯}{Q}=\{𝕀\})=\mathrm{ln}dim`$ and finally $`H(\underset{¯}{P}|\underset{¯}{Q})=H(\underset{¯}{p}|\underset{¯}{q})`$, such that $`0H(\underset{¯}{P}|\underset{¯}{Q})H(\underset{¯}{P})`$ as desired. Note that now $`H(\underset{¯}{P}|\underset{¯}{Q})`$ is completely specified by $`\underset{¯}{P}`$ and $`\underset{¯}{Q}`$. Also $`H(\underset{¯}{P},\underset{¯}{Q})=H(\underset{¯}{Q})+H(\underset{¯}{P}|\underset{¯}{Q})`$ is symmetric in $`\underset{¯}{P}`$ and $`\underset{¯}{Q}`$.
It is easy to see that $`\underset{¯}{p}`$ is a consequence of $`\underset{¯}{q}`$ if and only if $`\underset{¯}{Q}\underset{¯}{P}`$ such that $`H(\underset{¯}{P}|\underset{¯}{Q})=0`$ if and only if $`\underset{¯}{Q}\underset{¯}{P}`$. Similarly $`\underset{¯}{p}`$ and $`\underset{¯}{q}`$ are independent if and only if $`\underset{¯}{P}=\{𝕀\}`$ or $`\underset{¯}{Q}=\{𝕀\}`$. Therefore, whenever $`H(\underset{¯}{P})0`$, $`H(\underset{¯}{P}|\underset{¯}{Q})=H(\underset{¯}{P})`$ if and only if $`\underset{¯}{Q}=\{𝕀\}`$, which in this context is the trivial conditioning and for which the entropy is zero in contrast to our construction in terms of density matrices. Finally we set $`AdU\underset{¯}{Q}=\{UQ_iU^1\}`$ for any $`\underset{¯}{Q}`$ and any unitary $`U`$. Then obviously $`H(AdU\underset{¯}{P}|AdU\underset{¯}{Q})=H(\underset{¯}{P}|\underset{¯}{Q})`$ (compare condition 1).
Since the $`P_i`$’s and the $`Q_j`$’s need not commute, this construction is a non-commutative version of Shannon’s conditional entropy in (commutative) classical probability theory. Thus a classical partition $`X`$ is replaced by a spectral resolution of the identity $`\underset{¯}{P}`$, which in turn corresponds to the $``$-algebra $`E_{\underset{¯}{P}}()`$ and which is abelian if and only if each $`P_i`$ is one-dimensional. The choice $`\underset{¯}{Q}=\{𝕀\}`$ giving maximal entropy $`H(\underset{¯}{Q})`$ and maximal conditional entropy $`H(\underset{¯}{P}|\underset{¯}{Q})`$ corresponds to the maximal algebra $`E_{\underset{¯}{Q}=\{𝕀\}}()=`$. Our construction of $`H(\underset{¯}{P}|\underset{¯}{Q})`$ differs from the construction in .
We might have defined the conditional entropy of two density matrices $`\rho `$ and $`\sigma `$ by $`H(\underset{¯}{P}(\rho )|\underset{¯}{Q}(\sigma ))`$. Conditions 1,2 and 4 are then satisfied but not condition 5 and condition 3, since the dependence on the eigenvalues of $`\rho `$ and $`\sigma `$ drops out. We conjecture that condition 6 is also not satisfied.
Acknowledgements: The author would like to thank M. Karowski, H. Narnhofer, M.A. Nielsen, M. Schmidt and E. Størmer for helpful remarks.
|
warning/0003/math0003118.html
|
ar5iv
|
text
|
# Weak reflection at the successor of singular
## 0 Introduction and the statement of the results.
Stationary reflection is a compactness phenomenon in the context of stationary sets. To motivate its investigation, let us consider first the situation of a regular uncountable cardinal $`\kappa `$ and a closed unbounded subset $`C`$ of $`\kappa `$. For every of $`\kappa `$ many limit points $`\alpha `$ of $`C`$, we have that $`C\alpha `$ is closed unbounded in $`\alpha `$. Now let us ask the same question, but starting with a set $`S`$ which is stationary, not necessarily club, in $`\kappa `$. Is there necessarily $`\alpha <\kappa `$ such that $`S\alpha `$ is stationary in $`\alpha `$-in the lingo of set theorists, $`S`$ reflects at $`\alpha `$? The answer to this question turns to be very intricate, and in fact the notion of stationary reflection is one of the most studied notions of combinatorial set theory. This is the case not only because of the historical significance stationary reflection achieved through by now classical work of R. Jensen \[Je\] and later work of J.E. Baumgartner \[Ba\], L. Harrington and S. Shelah \[HaSh 99\], M. Magidor \[Ma\] and many later papers, but also because of the large number of applications it has in set theory and allied areas. In set theory, stationary reflection is known to have deep connections with various guessing principles, the simplest one of which is Jensen’s $`\mathrm{}`$ (\[Je\]), and the notions from pcf theory, such as good scales (for a long list of results in this area, as well as an excellent list of references, we refer the reader to \[CuFoMa\]), and some connections with saturation of normal filters (\[DjSh 545\]). In set-theoretic topology, various kinds of spaces have been constructed from the assumption of the existence of a non-reflecting stationary set (for references see \[KuVa\]), and in model theory versions of stationary reflection have been shown to have a connection with decidability of monadic second-order logic (\[Sh 80\]).
We investigate the notion of weak reflection, which, as the name suggests, is a weakening of the stationary reflection. For a regular cardinal $`\kappa `$, we say that $`\lambda >\kappa `$ weakly reflects at $`\kappa `$ iff for every function $`f:\lambda \kappa `$, there is $`\delta <\lambda `$ of cofinality $`\kappa `$ (we say $`\delta S_\kappa ^\lambda `$) such that $`fe`$ is not strictly increasing for any $`e`$ a club of $`\delta `$. Its negation is a strong form of non-reflection, called strong non-reflection. The notions were introduced by Džamonja and Shelah in \[DjSh 545\] in connection with saturation of normal filters, as well as the guessing principle $`\mathrm{}_\lambda ^{}(\lambda ^+)`$, which a relative of another popular guessing principle, $`\mathrm{}`$. It is proved in \[DjSh 545\] that, in the case when $`\lambda =\mu ^+`$ and $`\mathrm{}_0<\kappa =\mathrm{cf}(\mu )<\mu `$, if weak reflection of $`\lambda `$ at $`\kappa `$ holds relativized to every stationary subset of $`S_\kappa ^\lambda `$, then $`\mathrm{}_\mu ^{}(S_\kappa ^\lambda )`$ holds. The exact statement of the principle is of no consequence to us here, so we omit the definition. We simply note that this statement is stronger than just $`\mathrm{}_\mu ^{}(\lambda )`$, which holds just from the given cardinal assumptions.
Weak reflection was further investigated by Cummings, Džamonja and Shelah in \[CuDjSh 571\], more about which will be mentioned in a moment. Weak reflection has a very interesting aplication given by Cummings and Shelah in \[CuSh 596\], where they use it as a tool to build models where stationary reflection holds for some cofinalities but fails badly for others.
It was proved in \[DjSh 545\] that if there is $`\lambda `$ which weakly reflects at $`\kappa `$, the first such $`\lambda `$ is a regular cardinal. It is also not difficult to see that the first $`\lambda `$ cannot be weakly compact. On the other hand, in \[CuDjSh 571\] Cummings, Džamonja and Shelah proved that, modulo the existence of certain large cardinals, it is consistent to have a cardinal $`\lambda `$ which weakly reflects at unboundedly many regular $`\kappa `$ below it, and strongly non-reflects at unboundedly many others.
A question left open by these investigations, was if it is consistent to have $`\kappa `$ for which the first $`\lambda `$ which weakly reflects at $`\kappa `$, is a successor of singular. We answer this question positively, modulo the existence of a certain large cardinal, whose strength is in the neighborhood of being 2-huge. To state our results more precisely, let us give the exact definition of weak reflection and the statement of our main theorem.
###### Definition 0.1
Given $`\mathrm{}_0<\kappa =\mathrm{cf}(\kappa )`$ and $`\lambda >\kappa `$. We say that $`\lambda `$ weakly reflects at $`\kappa `$ iff for every function $`f:\lambda \kappa `$, there is $`\delta S_\kappa ^\lambda `$ such that $`fe`$ is not strictly increasing for any $`e`$ a club of $`\delta `$.
###### Theorem 0.2
(1) Let $`V`$ be a universe in which $`GCH`$ holds and $`\mu _0`$ is a cardinal such that there is an elementary embedding $`𝐣:VM`$ with the following properties:
(i) $`\mathrm{crit}(𝐣)=\mu _0`$,
(ii) For some $`\kappa ^{}`$ a successor of singular and $`\chi `$, we have
$$\mu _0<\kappa ^{}<\mu _1\stackrel{\mathrm{def}}{=}𝐣(\mu _0)<\lambda ^{}\stackrel{\mathrm{def}}{=}𝐣(\kappa ^{})<\mathrm{cf}(\chi )=\chi <\mu _2\stackrel{\mathrm{def}}{=}𝐣(\mu _1),$$
(iii) $`{}_{}{}^{\chi }MM`$.
Then there is a generic extension of $`V`$ in which cardinals and cofinalities $`\mu _0`$ are preserved, and the first $`\lambda `$ weakly reflecting at $`\kappa ^{}`$ is $`\lambda ^{}`$ (hence, a successor of singular).
(2) In (1), we can replace the requirement that $`\kappa ^{}`$ is a successor of singular by “$`\phi (\kappa ^{})`$ holds” for any of the following meanings of $`\phi (x)`$:
(a) $`x`$ is inaccessible,
(b) $`x`$ is strongly inaccessible,
(c) $`x`$ is Mahlo,
(d) $`x`$ is strongly Mahlo,
(e) $`x`$ is $`\alpha `$-(strongly) inaccessible for $`\alpha <x`$,
(e) $`x`$ is $`\alpha `$-(strongly) Mahlo for $`\alpha <x`$,
and have the same conclusion (hence in place of $`\lambda ^{}`$ is a successor of singular, $`V^P`$ will satisfy $`\phi (\lambda ^{})`$).
(3) With the same assumptions as in (1),
(i) there is a generic extension of $`V`$ in which $`\kappa ^{}=\mathrm{}_{53}`$ and $`\lambda ^{}`$, a successor of singular, is the first cardinal weakly reflecting at $`\kappa ^{}`$,
(ii) there is a generic extension of $`V`$ in which $`\kappa ^{}=\mathrm{}_{\omega +1}`$ and $`\lambda ^{}`$, a successor of singular, is the first cardinal weakly reflecting at $`\kappa ^{}`$,
(iii) there is a generic extension of $`V`$ in which $`\lambda ^{}`$ is the first cardinal weakly reflecting at $`\kappa ^{}`$, and $`\lambda ^{}=(\kappa ^{})^{+\gamma }`$ for some $`\gamma \mathrm{}_7`$.
###### Remark 0.3
Our assumptions follow if $`\mu _0`$ is 2-huge and $`GCH`$ holds. The integers 53 and 7 in the statement of part (3) above, are to a large extent arbitrary.
The proof of (1) uses as a building block a forcing notion introduced by Cummings, Džamonja and Shelah in \[CuDjSh 571\], which introduces a function witnessing strong non-reflection of a given cardinal $`\lambda `$ to a cardinal $`\kappa `$. An important feature of this forcing is that it has a reasonable degree of (strategic) closure, provided that strong non-reflection of $`\theta `$ to $`\kappa `$ already holds for $`\theta [\kappa ,\lambda )`$, and hence it can be iterated. This forcing is a rather homogeneous forcing, so the term forcing associated with it has strong decision properties. The forcing that we actually use is a term forcing associated with a certain product of the strong non-reflection forcings and a Laver-like preparation. Using this, we force the strong non-reflection of $`\theta `$ to $`\kappa ^{}`$ for all $`\theta <\lambda ^{}`$, and the point is to prove that in the extension $`\lambda ^{}`$ weakly reflects on $`\kappa ^{}`$. If we are given a condition and a name forced to be a strogly non-reflecting function, we can use the large cardinal assumptions to pick a certain model $`N`$, for which are able to build a generic condition, whose existence contradicts the choice of the name. To build the generic condition we use the preparation and the fact that we are dealing with a term forcing. Proofs of (2) and (3) are easy modifications of the proof of (1).
We recall some facts and definitions.
###### Notation 0.4
(1) Reg stands for the class of regular cardinals.
(2) If $`p,q`$ are elements of a forcing notion, then $`pq`$ means that $`q`$ is an extension of $`p`$.
(3) For $`p`$ a condition in the limit of an iteration Pα,Q
~
β:αα,β<α\langle P_{\alpha},\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\beta}:\,\alpha\leq\alpha^{\ast},\beta<\alpha^{\ast}\rangle, we let
Dom(p)=def{β<α:¬(pβ``p(β)=Q
~
β")}.superscriptdefDom𝑝conditional-set𝛽superscript𝛼p𝛽``p𝛽subscriptsubscript𝑄
~
𝛽"{\rm Dom}(p)\buildrel\rm def\over{=}\{\beta<\alpha^{\ast}:\,\neg(p\mathchar 13334\relax\beta\mathchar 13325\relax``p(\beta)=\emptyset_{\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\beta}}")\}.
(4) The statement that $`\lambda `$ weakly reflects at $`\kappa `$ is denoted by $`WR(\lambda ,\kappa )`$. Its negation of $`WR(\lambda ,\kappa )`$ is denoted by $`SNR(\lambda ,\kappa )`$.
###### Remark 0.5
It is easily seen that $`\lambda `$ weakly reflects at $`\kappa `$ iff $`|\lambda |`$ does, so we can without loss of generality, when discussing weak reflection of $`\lambda `$ to $`\kappa `$ assume that $`\lambda `$ is a cardinal.
###### Definition 0.6
(1) For a forcing notion and a limit ordinal $`\epsilon `$, we define the game $`G(P,\alpha )`$ as follows. The game is played between I and II, and it lasts $`\epsilon `$ steps, unless a player is forced to stop before that time. For $`\zeta <\epsilon `$, we denote the $`\zeta `$-th move of I by $`p_\zeta `$, and that of II by $`q_\zeta `$. The requirements are that I commences by $`\mathrm{}_P`$ and that for all $`\zeta `$ we have $`p_\zeta q_\zeta `$, while for $`\xi <\zeta `$ we have $`q_\xi p_\zeta `$.
I wins a play $`\mathrm{\Gamma }`$ of $`G(P,\epsilon )`$ iff $`\mathrm{\Gamma }`$ lasts $`\epsilon `$ steps.
(2) For $`P`$ and $`\epsilon `$ as above, we say that $`P`$ is $`\epsilon `$-strategically closed iff I has a winning strategy in $`G(P,\epsilon )`$. We say that $`P`$ is $`(<\epsilon )`$-strategically closed iff it is $`\zeta `$-strategically closed for all $`\zeta <\epsilon `$.
###### Definition 0.7
A set $`A`$ of ordinals is an Easton set iff
$$\sigma \mathrm{Reg}(sup(A)+1)sup(A\sigma )<\sigma .$$
###### Definition 0.8
We shall call a forcing notion $`P`$ mildly homogeneous iff for every formula $`\phi (x_0,\mathrm{},x_{n1})`$ of the forcing language of $`P`$ and $`a_0,\mathrm{},a_{n1}`$ (canonical names of) objects in $`V`$, we have $`\mathrm{}_P\mathrm{`}\mathrm{`}\phi (a_0,\mathrm{},a_{n1})\mathrm{"}.`$
## 1 Proofs.
We give the proof of Theorem 0.2. The main part of the proof is to deal with part (1) of the Theorem, and at the very end of the section we indicate the changes needed for the other parts of the theorem.
###### Definition 1.1
Given $`\mathrm{}_0<\kappa =\mathrm{cf}(\kappa )<\sigma `$.
$`P(\kappa ,\sigma )`$ is the forcing notion whose elements are functions $`p`$ with $`\mathrm{dom}(p)`$ an ordinal $`<\sigma `$, the range $`\mathrm{rge}(p)\kappa `$, and the property
$$\beta S_\kappa ^\sigma (c\text{ a club of }\beta )[pc\text{ is strictly increasing}],$$
while $`P(\kappa ,\sigma )`$ is ordered by extension.
###### Fact 1.2 (Cummings, Džamonja and Shelah)
\[CuDjSh 571\] Let $`\kappa `$ and $`\sigma `$ be such that $`P(\kappa ,\sigma )`$ is defined, then
(1) $`|P(\kappa ,\sigma )||{}_{}{}^{<\sigma }\kappa |=\kappa ^{<\sigma }`$.
(2) Suppose that for all $`\theta [\kappa ,\sigma )`$ we have $`SNR(\theta ,\kappa )`$. Then $`P(\kappa ,\sigma )`$ is $`(<\sigma )`$-strategically closed.
###### Definition 1.3
Given $`\mathrm{}_0<\mathrm{cf}(\kappa )=\kappa <\lambda `$.
$`Q_{(\kappa ,\lambda )}`$ is the result of the reverse Easton support iteration of $`P(\kappa ,\sigma )`$ for $`\sigma =\mathrm{cf}(\sigma )(\kappa ,\lambda )`$. More precisely, let
Q¯=Qα,R
~
β:αλ,β<λ,\bar{Q}=\langle Q_{\alpha},\mathchoice{\oalign{$\displaystyle R$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle R$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle R$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle R$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\beta}:\,\alpha\leq\lambda,\beta<\lambda\rangle,
where
(1) Qα``R
~
α={}"subscript𝑄𝛼``subscript𝑅
~
𝛼"Q_{\alpha}\mathchar 13325\relax``\mathchoice{\oalign{$\displaystyle R$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle R$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle R$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle R$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\alpha}=\{\emptyset\}" unless $`\alpha \mathrm{Reg}(\kappa ,\lambda )`$, in which case
Qα``R
~
α=P
~
(κ,α)".subscript𝑄𝛼``subscript𝑅
~
𝛼𝑃
~
𝜅𝛼"Q_{\alpha}\mathchar 13325\relax``\mathchoice{\oalign{$\displaystyle R$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle R$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle R$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle R$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\alpha}=\mathchoice{\oalign{$\displaystyle{{P}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{P}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{P}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{P}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}(\kappa,\alpha)".
(2) For $`\alpha \lambda `$
$`pQ_\alpha `$ iff for all $`\gamma <\alpha `$ we have Qγ``p(γ)R
~
γ"subscriptsubscript𝑄𝛾absent``𝑝𝛾subscript𝑅
~
𝛾"\mathchar 13325\relax_{Q_{\gamma}}``p(\gamma)\in\mathchoice{\oalign{$\displaystyle R$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle R$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle R$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle R$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\gamma}" and
(i) if $`\alpha `$ is inaccessible, then $`|\mathrm{Dom}(p)|<\alpha `$.
(ii) If $`\alpha `$ is a limit but not inaccessible, then
$$Q_\alpha \stackrel{\mathrm{def}}{=}\{\mathrm{p}:(\beta <\alpha )[\mathrm{p}\beta \mathrm{Q}_\beta ]\}.$$
(3) $`pq`$ iff for all $`\beta <\lambda `$ we have $`q\beta \mathrm{`}\mathrm{`}q(\beta )p(\beta )\mathrm{"}`$.
###### Fact 1.4 (Cummings, Džamonja, Shelah)
\[CuDjSh 571\] Let $`\overline{Q}`$, $`\kappa `$ and $`\lambda `$ be as in Definition 1.3. For all $`\alpha \lambda `$:
(1) Whenever $`\alpha `$ is regular, $`|Q_\alpha |\alpha ^{<\alpha }`$,
(2) Qα``|R
~
α||α|<κ"subscriptsubscript𝑄𝛼absent``subscript𝑅
~
𝛼superscript𝛼absent𝜅"\mathchar 13325\relax_{Q_{\alpha}}``{|\mathchoice{\oalign{$\displaystyle R$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle R$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle R$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle R$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\alpha}|}\leq{|\alpha|}^{<\kappa}",
(3) $`Q_{\alpha +1}`$ has $`(|\alpha |^{|<\alpha |})^+`$-cc. In addition, if $`\alpha `$ is strongly Mahlo, then $`Q_\alpha `$ has $`\alpha `$-cc.
(4) Qα``R
~
αsubscriptsubscript𝑄𝛼absent``subscript𝑅
~
𝛼\mathchar 13325\relax_{Q_{\alpha}}``\mathchoice{\oalign{$\displaystyle R$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle R$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle R$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle R$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\alpha} is $`(<\alpha )`$-strategically closed”.
(5) For all $`\beta <\alpha `$, we have that Q
~
α/Qβsubscript𝑄
~
𝛼subscript𝑄𝛽\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\alpha}/Q_{\beta} is $`(<\beta )`$-strategically closed.
(6) $`Q_\alpha `$ preserves all cardinals and cofinalities $`(|\alpha |^{<|\alpha |})^+`$, and all strongly inaccessible cardinals and cofinalities $`|\alpha |`$.
(7) $`_{Q_\alpha }\mathrm{`}\mathrm{`}SNR(\kappa ,\beta )\mathrm{"}`$ for all $`\beta <\alpha `$.
###### Notation 1.5
(1) For a forcing notion $`Q`$ of the form Q=P1P
~
2𝑄subscript𝑃1subscript𝑃
~
2Q=P_{1}\ast\mathchoice{\oalign{$\displaystyle P$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle P$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle P$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle P$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{2}, we denote by $`Q^{}`$ the term forcing associated with $`Q`$, defined by
Q=def{(P1,q
~
):
q
~
is a canonical P1-name for a condition in P2},superscriptdefsuperscript𝑄tensor-productconditional-setsubscriptsubscriptP1𝑞
~
q
~
is a canonical subscriptP1-name for a condition in subscriptP2Q^{\otimes}\buildrel\rm def\over{=}\{(\emptyset_{P_{1}},\mathchoice{\oalign{$\displaystyle q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}):\,\mathchoice{\oalign{$\displaystyle q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}\mbox{ is a canonical }P_{1}\mbox{-name for a condition in }P_{2}\},
(in particular $`Q^{}Q`$), with the order inherited from $`Q`$.
(2) For a triple $`(R,\kappa ,\sigma )`$ with $`\mathrm{}_0<\mathrm{cf}(\kappa )=\kappa <\sigma `$, and $`R`$ a forcing notion preserving $`\kappa =\mathrm{cf}(\kappa )>\mathrm{}_0`$, with $`\mathrm{}_R`$ the minimal element of $`R`$, we define $`Q_{(R,\kappa ,\sigma )}^{}`$ to be [RQ
~
(κ,σ)]superscriptdelimited-[]𝑅subscript𝑄
~
𝜅𝜎tensor-product[R\ast\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{(\kappa,\sigma)}]^{\otimes}.
###### Observation 1.6
$`Q_{(R,\kappa ,\sigma )}^{}`$, when defined, is $`(<\kappa ^+)`$-strategically closed.
###### Claim 1.7
(1) $`P(\sigma ,\lambda )`$ is mildly homogeneous, for $`\mathrm{}_0<\mathrm{cf}(\sigma )=\sigma <\lambda `$,
(2) If $`P`$ is mildly homogeneous and
P``
Q
~
is mildly homogeneous",subscript𝑃absent``
Q
~
is mildly homogeneous"\mathchar 13325\relax_{P}``\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}\mbox{ is mildly homogeneous}",
then PQ
~
𝑃𝑄
~
P\ast\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}} is mildly homogeneous.
(3) $`Q_{(\kappa ,\lambda )}`$ is mildly homogeneous, for $`\mathrm{}_0<\mathrm{cf}(\kappa )=\kappa <\lambda `$.
Proof of the Claim. (1) Suppose not, and let $`p,qP\stackrel{\mathrm{def}}{=}\mathrm{P}(\sigma ,\lambda )`$ force contradictory statements about $`\phi (a_0,\mathrm{},a_{n1})`$. Let $`\alpha =\mathrm{Dom}(q)`$ and consider the function $`F:PP`$ such that $`F(f)=g`$ iff $`qg`$ and for $`i\mathrm{dom}(f)`$ we have $`g(\alpha +i)=f(i)`$.
This function is an isomorphism between $`P`$ and $`P/q\stackrel{\mathrm{def}}{=}\{\mathrm{g}\mathrm{P}:\mathrm{g}\mathrm{q}\}`$, and it induces an isomorphism between $`P`$-names and $`P/q`$-names. However, in $`P/q`$ we have that $`q`$ and $`F(p)q`$ force contradictory statements about $`\phi (a_0,\mathrm{},a_{n1})`$. Contradiction.
(2)-(3) Similar proofs. $`\mathrm{}_{\text{1.7}}`$
###### Remark 1.8
We remark that $`P(\sigma ,\lambda )`$ in fact has stronger homogeneity properties, a fact which will not be used here.
Proof of the Theorem continued.
Let $`V`$, $`𝐣`$ and the cardinals mentioned in the assumptions be fixed. Note that by elementarity, $`\lambda ^{}`$ is the successor of a singular cardinal.
###### Notation 1.9
In the situation when notation $`Q_{(R,\kappa ,\kappa ^{})}^{}`$ makes sense, we abbreviate it as $`Q_{(R,\kappa )}^{}`$.
###### Definition 1.10
We define $`P^{}`$ to be the forcing whose elements are functions $`h`$, with $`\mathrm{dom}(h)`$ an Easton subset of $`\mu _0`$, with the property
$$\alpha <\beta \mathrm{dom}(h)h(\alpha )(\beta ),$$
ordered by extension.
###### Claim 1.11
Forcing with $`P^{}`$ preserves cardinals and cofinalities $`\mu _0`$, and $`GCH`$ above $`\mu _0`$, while any inaccessible $`\sigma \mu _0`$ which is a limit of inaccessibles, remains regular and $`2^\sigma =\sigma ^+`$ holds in the extension by $`P^{}`$.
Proof of the Claim. First notice that $`|P^{}|=\mu _0`$, so $`P^{}`$ has $`\mu _0^+`$-cc and preserves cardinals and cofinalities $`\mu _0^+`$, as well as $`GCH`$ above $`\mu _0`$.
Now suppose that $`\sigma \mu _0`$ is a limit of inaccessibles, but its cofinality is changed by $`P^{}`$ to be $`\theta `$ for some $`\theta <\sigma `$. Let $`p^{}P^{}`$ force this. Without loss of generality, $`\theta `$ is (strongly) inaccessible and $`\theta \mathrm{dom}(p^{})`$.
Let
$$P_{<\theta }\stackrel{\mathrm{def}}{=}\{\mathrm{q}\theta :\mathrm{q}\mathrm{P}^{}\&\mathrm{q}\mathrm{p}^{}\}$$
and
$$P_\theta \stackrel{\mathrm{def}}{=}\{\mathrm{q}[\theta ,\mu _0):\mathrm{q}\mathrm{P}^{}\&\mathrm{q}\mathrm{p}^{}\},$$
both ordered by extension. The mapping $`q(q[\theta ,\mu _0),q\theta )`$ shows that $`P^{}/p\stackrel{\mathrm{def}}{=}\{\mathrm{q}\mathrm{P}^{}:\mathrm{q}\mathrm{p}^{}\}`$ is isomorphic to $`P_\theta \times P_{<\theta }`$. We have that $`P_\theta `$ is $`(<\theta ^+)`$-closed, so $`P_{<\theta }`$ adds a cofinal function from $`\theta `$ to $`\sigma `$. However, $`|P_{<\theta }|\theta `$ (as $`\theta `$ is strongly inaccessible), and so it preserves cardinals and cofinalities $`\theta ^+`$, a contradiction.
We can similarly decompose $`P^{}`$ into $`P_\sigma \times P_{<\sigma }`$ to observe that
$$_P^{}\mathrm{`}\mathrm{`}2^\sigma =\sigma ^+\mathrm{"}.$$
$`\mathrm{}_{\text{1.11}}`$
###### Definition 1.12
(1) For $`\mu <\mu _0`$ let R
~
μsubscript𝑅
~
𝜇\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu} and κ
~
μsubscript𝜅
~
𝜇\mathchoice{\oalign{$\displaystyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu}be the following $`P^{}`$-names: for a condition $`pP^{}`$, if $`\mu \mathrm{dom}(p)`$ and
p(μ)=(κ,R
~
) with μ<cf(κ)=κ<μ0,𝑝𝜇𝜅𝑅
~
with 𝜇cf𝜅𝜅subscript𝜇0p(\mu)=(\kappa,\mathchoice{\oalign{$\displaystyle R$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle R$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle R$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle R$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})\mbox{ with }\mu<{\rm cf}(\kappa)=\kappa<\mu_{0},
and $`R(\kappa ^+)`$ is a forcing notion which preserves cardinals and cofinalities $`\mu `$, then $`p`$ forces κ
~
μsubscript𝜅
~
𝜇\mathchoice{\oalign{$\displaystyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu} to be $`\kappa `$ and R
~
μsubscript𝑅
~
𝜇\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu} to be Q
~
(R,κ
~
μ)subscriptsuperscript𝑄
~
tensor-product𝑅subscript𝜅
~
𝜇\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{\otimes}_{(R,{\mathchoice{\oalign{$\displaystyle\kappa$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\kappa$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\kappa$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\kappa$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}}_{\mu})}. We say that $`R_\mu =R`$. If, $`\mu \mathrm{Dom}(p)`$ but $`p(\mu )`$ does not satisfy the conditions above, then $`p`$ forces R
~
μsubscript𝑅
~
𝜇\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu} to be the trivial forcing, which will for notational purposes be thought of as $`\{\mathrm{},\mathrm{})\}`$. For the same reason, in these circumstances we think of $`R_\mu =\{\mathrm{}\}`$.
Note: each R
~
μsubscript𝑅
~
𝜇\mathchoice{\oalign{$\displaystyle{R}$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{R}$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{R}$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{R}$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu} is (over a dense subset of $`P^{}`$) a $`P^{}`$-name of a forcing notion from $`V`$, κ
~
μsubscript𝜅
~
𝜇\mathchoice{\oalign{$\displaystyle\kappa$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\kappa$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\kappa$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\kappa$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu} is a $`P^{}`$-name of an ordinal $`<\mu _0`$, and μ<μ0R
~
μsubscriptproduct𝜇subscript𝜇0subscript𝑅
~
𝜇\prod_{\mu<\mu_{0}}\mathchoice{\oalign{$\displaystyle{R}$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{R}$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{R}$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{R}$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu} is a $`P`$-name of a product of forcing. But $`R`$ below is forced not to be from $`V`$.
(2) For a $`P^{}`$-name f
~
μ<μ0R
~
μ𝑓
~
subscriptproduct𝜇subscript𝜇0subscript𝑅
~
𝜇\mathchoice{\oalign{$\displaystyle f$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle f$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle f$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle f$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}\in\prod_{\mu<\mu_{0}}\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu} and $`\alpha \mu _0`$, let
A
~
f
~
,α=def{μ<μ0:f
~
(μ)=(,q
~
) with ¬(R
~
μ``αDom(q
~
)")}.\mathchoice{\oalign{$\displaystyle A$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle A$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle A$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle A$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mathchoice{\oalign{$\displaystyle f$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle f$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle f$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle f$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}},\alpha}\buildrel\rm def\over{=}\{\mu<\mu_{0}:\,\mathchoice{\oalign{$\displaystyle f$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle f$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle f$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle f$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}(\mu)=(\emptyset,\mathchoice{\oalign{$\displaystyle q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})\mbox{ with }\neg(\mathchar 13325\relax_{\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu}}``\alpha\notin{\rm Dom}(\mathchoice{\oalign{$\displaystyle q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})")\}.
(23) Let $`R`$ $`\stackrel{~}{}`$ be a $`P^{}`$-name for:
{f
~
μ<μ0R
~
μ:(αμ0)[A
~
f
~
,α is an Easton set ]},\left\{\mathchoice{\oalign{$\displaystyle f$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle f$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle f$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle f$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}\in\prod_{\mu<\mu_{0}}\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu}:\,(\forall\alpha\leq\mu_{0})\,[\mathchoice{\oalign{$\displaystyle A$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle A$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle A$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle A$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mathchoice{\oalign{$\displaystyle f$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle f$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle f$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle f$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}},\alpha}\mbox{ is an Easton set ]}\right\},
ordered by the order inherited from μ<μ0R
~
μsubscriptproduct𝜇subscript𝜇0subscript𝑅
~
𝜇\prod_{\mu<\mu_{0}}\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu}.
hence, $`R`$ $`\stackrel{~}{}`$ is a $`P^{}`$-name of a forcing notion.
###### Notation 1.13
If we write (p,r
~
¯)PR
~
𝑝¯𝑟
~
superscript𝑃𝑅
~
(p,\bar{\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}})\in{{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}, we mean that
Pr
~
¯=(Rμ,r
~
(μ)):μ<μ0.\mathchar 13325\relax_{{{P}}^{-}}\bar{\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}}=\langle(\emptyset_{R_{\mu}},\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}(\mu)):\,\mu<\mu_{0}\rangle.
###### Definition 1.14
(1) Given (p,r
~
¯)PR
~
𝑝¯𝑟
~
superscript𝑃𝑅
~
(p,\bar{\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}})\in{{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}} and $`\sigma =\mathrm{cf}(\sigma )<\mu _0`$. For (q,s
~
¯)PR
~
𝑞¯𝑠
~
superscript𝑃𝑅
~
(q,\bar{\mathchoice{\oalign{$\displaystyle s$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle s$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle s$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle s$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}})\in{{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}, we define
(i) (q,s
~
¯)pr,σ(p,r
~
¯)subscriptpr𝜎𝑞¯𝑠
~
𝑝¯𝑟
~
(q,\bar{\mathchoice{\oalign{$\displaystyle s$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle s$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle s$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle s$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}})\geq_{{\rm pr},\sigma}(p,\bar{\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}}) iff
$`(\alpha )`$ (q,s
~
¯)(p,r
~
¯)𝑞¯𝑠
~
𝑝¯𝑟
~
(q,\bar{\mathchoice{\oalign{$\displaystyle s$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle s$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle s$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle s$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}})\geq(p,\bar{\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}}),
$`(\beta )`$ $`q(\sigma +1)=p(\sigma +1)`$,
($`\gamma )`$ For $`\mu <\mu _0`$ with ¬(q``R
~
μ\neg(q\mathchar 13325\relax``\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu} is trivial”), we have
(q,R
~
μ)``if
κ
~
μ<σ, then
s
~
(μ)(κ
~
μ,σ]=r
~
(μ)(κ
~
μ,σ]".formulae-sequence𝑞subscriptsubscript𝑅
~
𝜇``subscriptif
κ
~
𝜇𝜎 then
s
~
𝜇subscript𝜅
~
𝜇𝜎𝑟
~
𝜇subscript𝜅
~
𝜇𝜎"(q,\emptyset_{\mathchoice{\oalign{$\displaystyle R$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle R$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle R$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle R$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu}})\mathchar 13325\relax``\mbox{if }\mathchoice{\oalign{$\displaystyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu}<\sigma,\mbox{ then }\mathchoice{\oalign{$\displaystyle s$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle s$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle s$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle s$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}(\mu)\mathchar 13334\relax(\mathchoice{\oalign{$\displaystyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu},\sigma]=\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}(\mu)\mathchar 13334\relax(\mathchoice{\oalign{$\displaystyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu},\sigma]".
(ii) (q,s
~
¯)apr,σ(p,r
~
¯)subscriptapr𝜎𝑞¯𝑠
~
𝑝¯𝑟
~
(q,\bar{\mathchoice{\oalign{$\displaystyle s$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle s$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle s$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle s$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}})\geq_{{\rm apr},\sigma}(p,\bar{\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}}) iff
$`(\alpha )`$ (q,s
~
¯)(p,r
~
¯)𝑞¯𝑠
~
𝑝¯𝑟
~
(q,\bar{\mathchoice{\oalign{$\displaystyle s$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle s$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle s$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle s$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}})\geq(p,\bar{\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}}),
($`\beta `$) $`\mathrm{dom}(q)(\sigma +1,\mu _0)=\mathrm{dom}(p)(\sigma +1,\mu _0)`$.
($`\gamma )`$ For $`\mu <\mu _0`$ with ¬(q``R
~
μ\neg(q\mathchar 13325\relax``\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu} is trivial”), we have
(q,R
~
μ)``s
~
(μ)(σ,κ)=r
~
(μ)(σ,κ)".𝑞subscriptsubscript𝑅
~
𝜇``𝑠
~
𝜇𝜎superscript𝜅𝑟
~
𝜇𝜎superscript𝜅"(q,\emptyset_{\mathchoice{\oalign{$\displaystyle R$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle R$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle R$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle R$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu}})\mathchar 13325\relax``\mathchoice{\oalign{$\displaystyle s$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle s$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle s$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle s$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}(\mu)\mathchar 13334\relax(\sigma,\kappa^{\ast})=\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}(\mu)\mathchar 13334\relax(\sigma,\kappa^{\ast})".
(2) For (p,r
~
¯)PR
~
𝑝¯𝑟
~
superscript𝑃𝑅
~
(p,\bar{\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}})\in{{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}} and $`\sigma =\mathrm{cf}(\sigma )\mu _0`$, we let
Q(p,r
~
¯),σ=def{(q,s¯
~
):(q,s¯
~
)apr,σ(p,r
~
¯) for some (p,r
~
¯)pr,σ(p,r
~
¯)},superscriptdefsubscriptsuperscript𝑄𝑝¯𝑟
~
𝜎conditional-setq¯𝑠
~
subscriptapr𝜎q¯𝑠
~
superscriptpsuperscript¯𝑟
~
for some superscriptpsuperscript¯𝑟
~
subscriptpr𝜎p¯𝑟
~
Q^{-}_{(p,\bar{\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}}),\sigma}\buildrel\rm def\over{=}\{(q,\mathchoice{\oalign{$\displaystyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}):\,(q,\mathchoice{\oalign{$\displaystyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})\geq_{{\rm apr},\sigma}(p^{\prime},\bar{\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}}^{\prime})\mbox{ for some }(p^{\prime},\bar{\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}}^{\prime})\leq_{{\rm pr},\sigma}(p,\bar{\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}})\},
ordered as a suborder of PR
~
superscript𝑃𝑅
~
{{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}.
###### Claim 1.15
Given (p,r
~
¯)(q,s¯
~
)𝑝¯𝑟
~
𝑞¯𝑠
~
(p,\bar{\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}})\leq(q,\mathchoice{\oalign{$\displaystyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}) in PR
~
superscript𝑃𝑅
~
{{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}, and a regular $`\sigma <\mu _0`$.
Then there is a unique (t,z¯
~
)𝑡¯𝑧
~
(t,\mathchoice{\oalign{$\displaystyle\bar{z}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\bar{z}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\bar{z}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\bar{z}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}) such that
(p,r
~
¯)pr,σ(t,z¯
~
)apr,σ(q,s¯
~
).subscriptpr𝜎𝑝¯𝑟
~
𝑡¯𝑧
~
subscriptapr𝜎𝑞¯𝑠
~
(p,\bar{\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}})\leq_{{\rm pr},\sigma}(t,\mathchoice{\oalign{$\displaystyle\bar{z}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\bar{z}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\bar{z}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\bar{z}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})\leq_{{\rm apr},\sigma}(q,\mathchoice{\oalign{$\displaystyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}).
Proof of the Claim. Let $`t\stackrel{\mathrm{def}}{=}\mathrm{p}(\sigma +1)\mathrm{q}(\sigma +1,\mu _0)`$. Hence $`tP^{}`$ and $`ptq`$. We define a PR
~
superscript𝑃𝑅
~
{{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}-name z
~
¯¯𝑧
~
\bar{\mathchoice{\oalign{$\displaystyle z$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle z$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle z$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle z$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}} by letting for $`\mu <\mu _0`$
z
~
(μ)=def{r
~
(μ)(κ
~
μ,σ]q
~
(μ)(σ,κ] if definedr
~
(μ) otherwise.superscriptdef𝑧
~
𝜇cases𝑟
~
𝜇subscript𝜅
~
𝜇𝜎𝑞
~
𝜇𝜎superscript𝜅 if defined𝑟
~
𝜇 otherwise.\mathchoice{\oalign{$\displaystyle z$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle z$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle z$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle z$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}(\mu)\buildrel\rm def\over{=}\left\{\begin{array}[]{ll}\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}(\mu)\mathchar 13334\relax(\mathchoice{\oalign{$\displaystyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu},\sigma]\frown\mathchoice{\oalign{$\displaystyle q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}(\mu)\mathchar 13334\relax(\sigma,\kappa^{\ast}]&\mbox{ if defined}\\
\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}(\mu)&\mbox{ otherwise.}\end{array}\right.
$`\mathrm{}_{\text{1.15}}`$
###### Notation 1.16
(t,z¯
~
)𝑡¯𝑧
~
(t,\mathchoice{\oalign{$\displaystyle\bar{z}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\bar{z}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\bar{z}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\bar{z}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}) as in Claim 1.15 is denoted by intr((p,r
~
¯),(q,s¯
~
))𝑝¯𝑟
~
𝑞¯𝑠
~
\left((p,\bar{\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}}),(q,\mathchoice{\oalign{$\displaystyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})\right).
###### Claim 1.17
For $`\sigma =\mathrm{cf}(\sigma )<\mu _0`$, the forcing (PR
~
,pr,σ)superscript𝑃𝑅
~
subscriptpr𝜎({{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}},\leq_{{\rm pr},\sigma}) is $`(<\sigma +1)`$-strategically closed.
Proof of the Claim. For every $`\mu <\mu _0`$ we have
PR
~
``R
~
μ non-trivialQ
~
(κ
~
μ,κ)/Q
~
(κ
~
μ,σ] is (<σ+1)-strategically closed."\mathchar 13325\relax_{{{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}}``\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu}\mbox{ non-trivial}\Longrightarrow\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{(\mathchoice{\oalign{$\displaystyle\kappa$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\kappa$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\kappa$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\kappa$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu},\kappa^{\ast})}/\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{(\mathchoice{\oalign{$\displaystyle\kappa$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\kappa$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\kappa$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\kappa$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu},\sigma]}\mbox{ is }(<\sigma+1)\mbox{-strategically closed}."
Hence we can find names St
~
μσsubscriptsuperscriptSt
~
𝜎𝜇\mathchoice{\oalign{$\displaystyle{\rm St}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{\rm St}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{\rm St}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{\rm St}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{\sigma}_{\mu} of the corresponding winning strategies which exemplify the above statement.
Suppose that $`\zeta \sigma `$ and pξ=pξ0,p¯
~
ξ1:ξ<ζ\langle p_{\xi}=\langle p^{0}_{\xi},\mathchoice{\oalign{$\displaystyle\bar{p}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\bar{p}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\bar{p}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\bar{p}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{1}_{\xi}\rangle:\,\xi<\zeta\rangle, qξ=qξ0,q¯
~
ξ1:ξ<ζ\langle q_{\xi}=\langle q^{0}_{\xi},\mathchoice{\oalign{$\displaystyle\bar{q}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\bar{q}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\bar{q}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\bar{q}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{1}_{\xi}\rangle:\,\xi<\zeta\rangle are sequences of elements of PR
~
superscript𝑃𝑅
~
{{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}} such that
(1) For all $`\xi <\zeta `$ we have $`p_\xi _{\mathrm{pr},\sigma }q_\xi `$,
(2) For all $`\xi <\zeta `$ and $`\epsilon <\xi `$ we have $`q_\epsilon _{\mathrm{pr},\sigma }p_\xi `$ and for $`\mu <\mu _0`$ with ¬(pξ``R
~
μ\neg(p_{\xi}\mathchar 13325\relax``\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu} is trivial”), we have (pξ,R
~
μ,p
~
01(μ)(κ
~
μ,σ])PR
~
μQ
~
(κμ,σ]subscriptsuperscript𝑃subscript𝑅
~
𝜇subscript𝑄
~
subscript𝜅𝜇𝜎subscript𝑝𝜉subscriptsubscript𝑅
~
𝜇subscriptsuperscript𝑝
~
10𝜇subscript𝜅
~
𝜇𝜎absent(p_{\xi},\emptyset_{\mathchoice{\oalign{$\displaystyle R$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle R$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle R$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle R$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu}},\mathchoice{\oalign{$\displaystyle p$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle p$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle p$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle p$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{1}_{0}(\mu)\mathchar 13334\relax(\mathchoice{\oalign{$\displaystyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu},\sigma])\mathchar 13325\relax_{{{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu}\ast\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{(\kappa_{\mu},\sigma]}}
``p
~
ξ1(μ)(σ,κ)=St
~
μσ(p
~
ε1(μ)(σ,κ):ε<ξ,q
~
ε1(μ)(σ,κ):ε<ξ)"``\mathchoice{\oalign{$\displaystyle p$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle p$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle p$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle p$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{1}_{\xi}(\mu)\mathchar 13334\relax(\sigma,\kappa^{\ast})=\mathchoice{\oalign{$\displaystyle{\rm St}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{\rm St}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{\rm St}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{\rm St}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{\sigma}_{\mu}(\langle\mathchoice{\oalign{$\displaystyle p$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle p$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle p$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle p$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{1}_{\varepsilon}(\mu)\mathchar 13334\relax(\sigma,\kappa^{\ast}):\,\varepsilon<\xi\rangle,\langle\mathchoice{\oalign{$\displaystyle q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{1}_{\varepsilon}(\mu)\mathchar 13334\relax(\sigma,\kappa^{\ast}):\,\varepsilon<\xi\rangle)"
We define $`p_\zeta `$ by letting $`p_\zeta ^0\stackrel{\mathrm{def}}{=}\{\mathrm{q}_\xi ^0:\xi <\zeta \}`$. Notice that $`p_\zeta ^0P^{}`$ and $`p_\zeta ^0(\sigma +1)=p_0^0(\sigma +1)`$.
For $`\mu <\mu _0`$ with¬(pξ``R
~
μ\neg(p_{\xi}\mathchar 13325\relax``\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu} is trivial”), we let p
~
ζ1(μ)subscriptsuperscript𝑝
~
1𝜁𝜇\mathchoice{\oalign{$\displaystyle p$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle p$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle p$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle p$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{1}_{\zeta}(\mu) be the name given by
p
~
ζ1(μ)(κ
~
μ,σ]=defp
~
ζ0(μ)(κ
~
μ,σ]subscriptsuperscript𝑝
~
1𝜁𝜇subscript𝜅
~
𝜇𝜎superscriptdefsubscriptsuperscript𝑝
~
0𝜁𝜇subscript𝜅
~
𝜇𝜎\mathchoice{\oalign{$\displaystyle p$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle p$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle p$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle p$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{1}_{\zeta}(\mu)\mathchar 13334\relax(\mathchoice{\oalign{$\displaystyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu},\sigma]\buildrel\rm def\over{=}\mathchoice{\oalign{$\displaystyle p$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle p$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle p$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle p$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{0}_{\zeta}(\mu)\mathchar 13334\relax(\mathchoice{\oalign{$\displaystyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\kappa$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu},\sigma]
and
p
~
ζ1(μ)(σ,κ)=defSt
~
μσ(p
~
ξ1(μ)(σ,κ):ξ<ζ,q
~
ξ1(μ)(σ,κ):ξ<ζ).\mathchoice{\oalign{$\displaystyle p$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle p$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle p$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle p$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{1}_{\zeta}(\mu)\mathchar 13334\relax(\sigma,\kappa^{\ast})\buildrel\rm def\over{=}\mathchoice{\oalign{$\displaystyle{\rm St}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{\rm St}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{\rm St}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{\rm St}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{\sigma}_{\mu}(\langle\mathchoice{\oalign{$\displaystyle p$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle p$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle p$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle p$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{1}_{\xi}(\mu)\mathchar 13334\relax(\sigma,\kappa^{\ast}):\,\xi<\zeta\rangle,\langle\mathchoice{\oalign{$\displaystyle q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{1}_{\xi}(\mu)\mathchar 13334\relax(\sigma,\kappa^{\ast}):\,\xi<\zeta\rangle).
$`\mathrm{}_{\text{1.17}}`$
###### Claim 1.18
Suppose (p,r
~
¯)``τ
~
:σOrd":𝑝¯𝑟
~
``𝜏
~
𝜎Ord"(p,\bar{\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}})\mathchar 13325\relax``\mathchoice{\oalign{$\displaystyle\tau$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}:\,\sigma\rightarrow{\rm Ord}", where $`\sigma `$ is regular $`<\mu _0`$.
Then there is (q,s
~
¯)pr,σ(p,r¯
~
)subscriptpr𝜎𝑞¯𝑠
~
𝑝¯𝑟
~
(q,\bar{\mathchoice{\oalign{$\displaystyle s$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle s$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle s$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle s$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}})\geq_{{\rm pr},\sigma}(p,\mathchoice{\oalign{$\displaystyle\bar{r}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\bar{r}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\bar{r}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\bar{r}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}) and a Q(q,s
~
¯),σsubscriptsuperscript𝑄𝑞¯𝑠
~
𝜎Q^{-}_{(q,\bar{\mathchoice{\oalign{$\displaystyle s$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle s$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle s$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle s$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}}),\sigma}-name τ
~
superscript𝜏
~
\mathchoice{\oalign{$\displaystyle\tau$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{\prime} such that
(q,s
~
¯)``τ
~
=τ
~
".𝑞¯𝑠
~
``𝜏
~
superscript𝜏
~
"(q,\bar{\mathchoice{\oalign{$\displaystyle s$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle s$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle s$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle s$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}})\mathchar 13325\relax``\mathchoice{\oalign{$\displaystyle\tau$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}=\mathchoice{\oalign{$\displaystyle\tau$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{\prime}".
Proof of the Claim. We define a play of G((PR
~
,pr,σ),σ)𝐺superscript𝑃𝑅
~
subscriptpr𝜎𝜎G(({{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}},\leq_{{\rm pr},\sigma}),\sigma) as follows.
I starts by playing (p,r
~
¯)=defp0superscriptdef𝑝¯𝑟
~
subscriptp0(p,\bar{\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}})\buildrel\rm def\over{=}p_{0}. At the stage $`\zeta \sigma `$, player II chooses $`q_\zeta ^{}p_\zeta `$ such that $`q_\zeta ^{}`$ forces a value to τ
~
ζsubscript𝜏
~
𝜁\mathchoice{\oalign{$\displaystyle\tau$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\zeta}, and we let $`q_\zeta \stackrel{\mathrm{def}}{=}\mathrm{intr}(\mathrm{p}_\zeta ,\mathrm{q}_\zeta ^{})`$. At the stage $`0<\zeta <\sigma `$, we let I play according to the winning strategy for G((PR
~
,pr,σ),σ)𝐺superscript𝑃𝑅
~
subscriptpr𝜎𝜎G(({{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}},\leq_{{\rm pr},\sigma}),\sigma) applied to $`(p_\xi :\xi <\zeta ,q_\xi :\xi <\zeta )`$. At the end, we let (q,s
~
¯)=pσ.𝑞¯𝑠
~
subscript𝑝𝜎(q,\bar{\mathchoice{\oalign{$\displaystyle s$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle s$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle s$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle s$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}})=p_{\sigma}. $`\mathrm{}_{\text{1.18}}`$
###### Claim 1.19
If (p,r
~
¯)PR
~
𝑝¯𝑟
~
superscript𝑃𝑅
~
(p,\bar{\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}})\in{{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}, and $`\sigma =\mathrm{cf}(\sigma )<\mu _0`$ is such that $`\sigma \mathrm{dom}(p)`$, then Q(p,r¯
~
),σsubscriptsuperscript𝑄𝑝¯𝑟
~
𝜎Q^{-}_{(p,\mathchoice{\oalign{$\displaystyle\bar{r}$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\bar{r}$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\bar{r}$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\bar{r}$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}),\sigma} satisfies $`\mu _0`$-cc.
Proof of the Claim. Given q¯=qi=qi0,q
~
i1:i<μ0\bar{q}=\langle q_{i}=\langle q^{0}_{i},\mathchoice{\oalign{$\displaystyle q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{1}_{i}\rangle:\,i<\mu_{0}\rangle, with qiQ(p,r¯
~
),σsubscript𝑞𝑖subscriptsuperscript𝑄𝑝¯𝑟
~
𝜎q_{i}\in Q^{-}_{(p,\mathchoice{\oalign{$\displaystyle\bar{r}$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\bar{r}$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\bar{r}$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\bar{r}$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}),\sigma}. Suppose for contradiction that the range of this sequence is an antichain.
We have that for all $`i<\mu _0`$
$$q_i^0(\sigma +1,\mu _0)p(\sigma +1,\mu _0).$$
As $`\mathrm{dom}(q_i^0)`$ is an Easton set, without loss of generality we have that all $`q_i^0(\sigma +1,\mu _0)`$ are the same $`q^{}`$. Let $`G^{}`$ be $`P^{}`$-generic over $`V`$ with $`q^{}G^{}`$. Hence in $`V[G^{}]`$ the sequence qi1=def(,q¯
~
i):i<μ0\langle q^{1}_{i}\buildrel\rm def\over{=}(\emptyset,\mathchoice{\oalign{$\displaystyle\bar{q}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\bar{q}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\bar{q}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\bar{q}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{i}):\,i<\mu_{0}\rangle to an antichain in μ<μ0[RμQ
~
(κμ,κ)]subscriptproduct𝜇subscript𝜇0superscriptdelimited-[]subscript𝑅𝜇subscript𝑄
~
subscript𝜅𝜇superscript𝜅tensor-product\prod_{\mu<\mu_{0}}[{{R}}_{\mu}\ast\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{(\kappa_{\mu},\kappa^{\ast})}]^{\otimes}, and by the choice of the initial sequence, we have that (,q
~
i(μ)(σ+1)):μ<μ0:i<μ0\langle(\emptyset,\mathchoice{\oalign{$\displaystyle q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{i}(\mu)\mathchar 13334\relax(\sigma+1)):\,\mu<\mu_{0}\rangle:\,i<\mu_{0}\rangle gives rise to an antichain in μ<μ0[RμQ
~
κμ,σ]subscriptproduct𝜇subscript𝜇0superscriptdelimited-[]subscript𝑅𝜇subscript𝑄
~
subscript𝜅𝜇𝜎tensor-product\prod_{\mu<\mu_{0}}[{{R}}_{\mu}\ast\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\kappa_{\mu},\sigma}]^{\otimes}. For every $`i<\mu _0`$,
Ai=def{μ<μ0:q
~
μi(σ+1)}superscriptdefsubscript𝐴𝑖conditional-set𝜇subscript𝜇0subscriptsuperscript𝑞
~
i𝜇𝜎1A_{i}\buildrel\rm def\over{=}\{\mu<\mu_{0}:\,\mathchoice{\oalign{$\displaystyle q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{i}_{\mu}\mathchar 13334\relax(\sigma+1)\neq\emptyset\}
has size $`\sigma `$. Hence, without loss of generality, $`A_i`$’s form a $`\mathrm{\Delta }`$-system with root $`A^{}`$. Hence
,q
~
i(μ)(σ+1):μA:i<μ0\left\langle\langle\langle\emptyset,\mathchoice{\oalign{$\displaystyle q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{i}(\mu)\mathchar 13334\relax(\sigma+1)\rangle:\,\mu\in A^{\ast}:\rangle\,i<\mu_{0}\right\rangle
gives rise to an antichain, a contradiction. $`\mathrm{}_{\text{1.19}}`$
###### Claim 1.20
Forcing with PR
~
superscript𝑃𝑅
~
{{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}} preserves cardinals and cofinalities $`\mu _0`$.
Proof of the Claim. Suppose cofinalities $`\mu _0`$ are not preserved and let $`\theta `$ be the first cofinality $`\mu _0`$ destroyed. Hence $`\theta `$ is regular, and for some $`\tau `$ $`\stackrel{~}{}`$ , condition (p,r¯
~
)𝑝¯𝑟
~
(p,\mathchoice{\oalign{$\displaystyle\bar{r}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\bar{r}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\bar{r}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\bar{r}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}) and regular $`\sigma <\theta `$, we have (p,r¯
~
)``τ
~
:σθ:𝑝¯𝑟
~
``𝜏
~
𝜎𝜃(p,\mathchoice{\oalign{$\displaystyle\bar{r}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\bar{r}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\bar{r}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\bar{r}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})\mathchar 13325\relax``\mathchoice{\oalign{$\displaystyle\tau$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}:\,\sigma\rightarrow\theta is cofinal”.
Case 1. $`\sigma <\mu _0`$. By Claim 1.18, there is (q,s¯
~
)(p,r¯
~
)𝑞¯𝑠
~
𝑝¯𝑟
~
(q,\mathchoice{\oalign{$\displaystyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})\geq(p,\mathchoice{\oalign{$\displaystyle\bar{r}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\bar{r}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\bar{r}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\bar{r}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}) and a Q(p,r¯
~
),σsubscriptsuperscript𝑄𝑝¯𝑟
~
𝜎Q^{-}_{(p,\mathchoice{\oalign{$\displaystyle\bar{r}$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\bar{r}$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\bar{r}$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\bar{r}$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}),\sigma}-name τ
~
superscript𝜏
~
\mathchoice{\oalign{$\displaystyle\tau$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{\prime} such that (q,s¯
~
)``τ
~
=τ
~
"𝑞¯𝑠
~
``𝜏
~
superscript𝜏
~
"(q,\mathchoice{\oalign{$\displaystyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})\mathchar 13325\relax``\mathchoice{\oalign{$\displaystyle\tau$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}=\mathchoice{\oalign{$\displaystyle\tau$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{\prime}". Hence (q,s¯
~
)``τ
~
:σθ:𝑞¯𝑠
~
``superscript𝜏
~
𝜎𝜃(q,\mathchoice{\oalign{$\displaystyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})\mathchar 13325\relax``\mathchoice{\oalign{$\displaystyle\tau$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{\prime}:\,\sigma\rightarrow\theta is cofinal”, contradicting the fact that Q(p,r¯
~
),σsubscriptsuperscript𝑄𝑝¯𝑟
~
𝜎Q^{-}_{(p,\mathchoice{\oalign{$\displaystyle\bar{r}$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\bar{r}$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\bar{r}$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\bar{r}$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}),\sigma} has $`\mu _0`$-cc.
Case 2. $`\sigma \mu _0`$.
As for every $`\mu <\mu _0`$ with R
~
μsubscript𝑅
~
𝜇\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu} non-trivial we have that
PR
~
μQ
~
(κ
~
μ,κ)/Q
~
(κ
~
μ,σ] is (<σ+1)-strategically closed,{{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu}\ast\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{(\mathchoice{\oalign{$\displaystyle\kappa$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\kappa$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\kappa$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\kappa$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu},\kappa^{\ast})}/\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{(\mathchoice{\oalign{$\displaystyle\kappa$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\kappa$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\kappa$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\kappa$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu},\sigma]}\mbox{ is }(<\sigma+1)\mbox{-strategically closed},
there is (q,s¯
~
)(p,r¯
~
)𝑞¯𝑠
~
𝑝¯𝑟
~
(q,\mathchoice{\oalign{$\displaystyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})\geq(p,\mathchoice{\oalign{$\displaystyle\bar{r}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\bar{r}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\bar{r}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\bar{r}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}) and a Pμ<μ0[R
~
μQ
~
(κ
~
μ,σ]]superscript𝑃subscriptproduct𝜇subscript𝜇0superscriptdelimited-[]subscript𝑅
~
𝜇subscript𝑄
~
subscript𝜅
~
𝜇𝜎tensor-product{{P}}^{-}\ast\prod_{\mu<\mu_{0}}[{\mathchoice{\oalign{$\displaystyle{R}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{R}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{R}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{R}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}}_{\mu}\ast\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{(\mathchoice{\oalign{$\displaystyle\kappa$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\kappa$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\kappa$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\kappa$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu},\sigma]}]^{\otimes}-name τ
~
superscript𝜏
~
\mathchoice{\oalign{$\displaystyle\tau$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{\prime} such that (q,s¯
~
)``τ
~
:σθ:𝑞¯𝑠
~
``superscript𝜏
~
𝜎𝜃(q,\mathchoice{\oalign{$\displaystyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\bar{s}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})\mathchar 13325\relax``\mathchoice{\oalign{$\displaystyle\tau$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{\prime}:\,\sigma\rightarrow\theta is cofinal”. But this forcing has $`\sigma ^+`$-cc, a contradiction. $`\mathrm{}_{\text{1.20}}`$
###### Corollary 1.21
Forcing with PR
~
Q
~
(κ,λ)superscript𝑃𝑅
~
subscript𝑄
~
superscript𝜅superscript𝜆{{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}\ast\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{(\kappa^{\ast},\lambda^{\ast})} preserves cardinalities and cofinalities $`\mu _0`$, and forces $`SNR(\theta ,\kappa ^{})`$ for $`\theta (\kappa ^{},\lambda ^{})`$.
###### Claim 1.22
The following is forced by $`P^{}`$:
(1) $`R`$ $`\stackrel{~}{}`$ is mildly homogeneous.
(2) R
~
Q
~
(κ,λ)𝑅
~
subscript𝑄
~
superscript𝜅superscript𝜆\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}\ast\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{(\kappa^{\ast},\lambda^{\ast})} is mildly homogeneous.
Proof of the Claim. (1) First note that each R
~
μsubscript𝑅
~
𝜇\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu} is forced to be mildly homogeneous, by Claim 1.7(2) and the definition of $``$ operation.
(2) Follows from (1) and Claim 1.7(2). $`\mathrm{}_{\text{1.22}}`$
###### Main Claim 1.23
After forcing with P=defPR
~
Q
~
(κ,λ)superscriptdef𝑃superscriptP𝑅
~
subscript𝑄
~
superscript𝜅superscript𝜆{{P}}\buildrel\rm def\over{=}{{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}\ast\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{(\kappa^{\ast},\lambda^{\ast})}, we have the weak reflection of $`\lambda ^{}`$ at $`\kappa ^{}`$.
Proof of the Main Claim. Suppose otherwise, and let p=(p,q
~
,r
~
)superscript𝑝𝑝𝑞
~
𝑟
~
p^{\ast}=(p,\mathchoice{\oalign{$\displaystyle q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}},\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}) force $`\tau `$ $`\stackrel{~}{}`$ to be a function exemplifying the strong non-reflection of $`\lambda ^{}`$ at $`\kappa ^{}`$. As R
~
Q
~
(κ,λ)𝑅
~
subscript𝑄
~
superscript𝜅superscript𝜆\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}\ast\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{(\kappa^{\ast},\lambda^{\ast})} is forced to be mildly homogeneous by Claim 1.22, without loss of generality $`p^{}=(p,\mathrm{},\mathrm{})`$.
By a standard argument, our large cardinal assumptions imply that we can find a model $`N(\chi )`$ such that
(i) $`N\mu _0`$ is an ordinal $`\mu <\mu _0`$,
(ii) $`\text{otp(}N\lambda ^{}\text{)}=\kappa ^{}`$,
(iii) $`\text{otp(}N\mu _1\text{)}=\mu _0`$,
(iv) $`{}_{}{}^{\omega }NN`$ (even $`{}_{}{}^{\mu >}NN`$, although we do not use this),
(v) $`(N,)`$ is isomorphic to $`(\chi ^{})`$ for some regular $`\chi ^{}<\chi `$
(vi) $`|N\kappa ^{}|`$ is a regular cardinal.
(vii) κ,μ0,μ1,μ2,λ,P,p,τ
~
Nsuperscript𝜅subscript𝜇0subscript𝜇1subscript𝜇2superscript𝜆𝑃superscript𝑝𝜏
~
𝑁\kappa^{\ast},\mu_{0},\mu_{1},\mu_{2},\lambda^{\ast},{{P}},p^{\ast},\mathchoice{\oalign{$\displaystyle\tau$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}\in N.
\[Why? Consider $`𝐣\mathrm{`}\mathrm{`}((\chi ))`$ and use elementarity. Note that by $`{}_{}{}^{\chi }MM`$ and $`\chi ^{<\chi }=\chi `$ we have that $`𝐣\mathrm{`}\mathrm{`}((\chi ))M`$.\]
First we consider some consequences of our choice of $`N`$. Let $`\kappa \stackrel{\mathrm{def}}{=}|\mathrm{N}\kappa ^{}|`$ and $`\delta \stackrel{\mathrm{def}}{=}sup(\mathrm{N}\lambda ^{})`$.
Our assumptions on $`N`$ imply that $`\text{otp(}N\kappa ^{}\text{)}<\mu _0`$, hence $`\kappa <\mu _0`$. As for $`\delta `$, we have $`\delta S_\kappa ^{}^\lambda ^{}`$. Now notice that $`N\delta `$ is stationary in $`\delta `$, and it remains so after forcing with $`P`$.
\[Why? The set $`S\stackrel{\mathrm{def}}{=}\mathrm{S}_\mathrm{}_0^\delta \mathrm{N}`$ is a stationary subset of $`\delta `$, as $`E`$ defined as the closure of $`N\delta `$ is a club of $`\delta `$, and $`[\alpha N\&\mathrm{cf}(\alpha )=\mathrm{}_0]\alpha S`$ (this is true even with“$`\mathrm{cf}(\alpha )<\mu `$” in place of “$`\mathrm{cf}(\alpha )=\mathrm{}_0`$”). But $`P`$ is an $`\omega _1`$-closed forcing notion, hence $`S`$ remains stationary after forcing with $`P`$.\]
As $`p^{}N`$, we have that $`\mathrm{dom}(p)\mu `$. Hence
p+=defp{μ,(κ,(PR
~
)N)}superscriptdefsuperscript𝑝p𝜇𝜅superscriptsuperscriptP𝑅
~
Np^{+}\buildrel\rm def\over{=}p\cup\{\langle\mu,(\kappa,({{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})^{N})\rangle\}
is a well defined condition in $`P^{}`$, and it extends $`p`$. In fact, $`p^+`$ is a $`P^{}`$-generic condition over $`N`$, and it forces that R
~
μsubscript𝑅
~
𝜇\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu} is (PR
~
)Nsuperscriptsuperscript𝑃𝑅
~
𝑁({{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{R}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{R}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{R}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{R}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})^{N}, hence that [R
~
μQ
~
(κ,κ)]superscriptdelimited-[]subscript𝑅
~
𝜇subscript𝑄
~
𝜅superscript𝜅tensor-product[\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu}\ast\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{(\kappa,\kappa^{\ast})}]^{\otimes} is [(PR
~
)NQ
~
(κ,κ)]superscriptdelimited-[]superscriptsuperscript𝑃𝑅
~
𝑁subscript𝑄
~
𝜅superscript𝜅tensor-product[({{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{R}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{R}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{R}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{R}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})^{N}\ast\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{(\kappa,\kappa^{\ast})}]^{\otimes}, which is ([(PR
~
)Q
~
(κ,λ)])Nsuperscriptsuperscriptdelimited-[]superscript𝑃𝑅
~
subscript𝑄
~
superscript𝜅superscript𝜆tensor-product𝑁([({{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})\ast\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{(\kappa^{\ast},\lambda^{\ast})}]^{\otimes})^{N}. Let $`H`$ $`\stackrel{~}{}`$ be ([(PR
~
)Q
~
(κ,λ)])Nsuperscriptsuperscriptdelimited-[]superscript𝑃𝑅
~
subscript𝑄
~
superscript𝜅superscript𝜆tensor-product𝑁([({{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})\ast\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{(\kappa^{\ast},\lambda^{\ast})}]^{\otimes})^{N}-generic with pH
~
𝑝𝐻
~
p\in\mathchoice{\oalign{$\displaystyle H$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle H$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle H$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle H$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}. The inverse of the Mostowski collapse $`F`$ of $`N`$ maps $`H`$ $`\stackrel{~}{}`$ into a subset H
~
superscript𝐻
~
\mathchoice{\oalign{$\displaystyle H$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle H$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle H$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle H$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{\ast} of [(PR
~
)Q
~
(κ,λ)]superscriptdelimited-[]superscript𝑃𝑅
~
subscript𝑄
~
superscript𝜅superscript𝜆tensor-product[({{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})\ast\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{(\kappa^{\ast},\lambda^{\ast})}]^{\otimes}. Notice then that
p+``
H
~
is
R
~
μ-generic".superscript𝑝``subscript
H
~
is
R
~
𝜇-generic"p^{+}\mathchar 13325\relax``\mathchoice{\oalign{$\displaystyle H$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle H$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle H$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle H$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}\mbox{ is }\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{\mu}\mbox{-generic}".
We wish to define $`q`$ as follows: q=def(p+,R,r
~
)superscriptdef𝑞superscriptpsubscriptR𝑟
~
q\buildrel\rm def\over{=}(p^{+},\emptyset_{{R}},\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}), where $`r`$ $`\stackrel{~}{}`$ is a PR
~
superscript𝑃𝑅
~
{{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}-name over (p+,R
~
)superscript𝑝subscript𝑅
~
(p^{+},\emptyset_{\mathchoice{\oalign{$\displaystyle{R}$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{R}$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{R}$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{R}$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}}) of a condition in Q
~
(κ,λ)subscript𝑄
~
superscript𝜅superscript𝜆\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{(\kappa^{\ast},\lambda^{\ast})} defined by letting
Dom(r
~
)=def{Dom(h
~
):(p+,R
~
)``F1(h
~
)H
~
"},superscriptdefDom𝑟
~
conditional-setDom
~
superscriptpsubscript𝑅
~
``superscriptF1
~
𝐻
~
"{\rm Dom}(\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})\buildrel\rm def\over{=}\bigcup\{{\rm Dom}(\mathchoice{\oalign{$\displaystyle h$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle h$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle h$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle h$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}):\,(p^{+},\emptyset_{\mathchoice{\oalign{$\displaystyle{R}$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{R}$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{R}$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{R}$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}})\mathchar 13325\relax``F^{-1}(\mathchoice{\oalign{$\displaystyle h$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle h$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle h$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle h$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})\in\mathchoice{\oalign{$\displaystyle H$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle H$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle H$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle H$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}"\},
and for $`\theta `$ with (p+,R
~
)``θDom(r
~
)"superscript𝑝subscript𝑅
~
``𝜃Dom𝑟
~
"(p^{+},\emptyset_{\mathchoice{\oalign{$\displaystyle{R}$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{R}$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{R}$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{R}$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}})\mathchar 13325\relax``\theta\in{\rm Dom}(\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})", we let
r
~
(θ)=def{h
~
(θ):(p+,R
~
)``θDom(h
~
)&h
~
H
~
"}.superscriptdef𝑟
~
𝜃conditional-set
~
𝜃superscriptpsubscript𝑅
~
``𝜃Dom
~
~
superscript𝐻
~
"\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}(\theta)\buildrel\rm def\over{=}\bigcup\{\mathchoice{\oalign{$\displaystyle h$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle h$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle h$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle h$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}(\theta):\,(p^{+},\emptyset_{\mathchoice{\oalign{$\displaystyle{R}$\crcr\vbox to0.60275pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{R}$\crcr\vbox to0.60275pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{R}$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{R}$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}})\mathchar 13325\relax``\theta\in{\rm Dom}(\mathchoice{\oalign{$\displaystyle h$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle h$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle h$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle h$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})\,\,\&\,\,\mathchoice{\oalign{$\displaystyle h$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle h$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle h$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle h$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}\in\mathchoice{\oalign{$\displaystyle H$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle H$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle H$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle H$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{\ast}"\}.
We now claim that $`q`$ is a condition in $`P`$ and $`qp^{}`$. Let us check the relevant items:
(a) If $`(p^+,\mathrm{})\mathrm{`}\mathrm{`}\theta \text{ strongly inaccessible }(\kappa ^{},\lambda ^{})`$”, then
(p+,)``|Dom(r
~
)θ|<θ".superscript𝑝``Dom𝑟
~
𝜃𝜃"(p^{+},\emptyset)\mathchar 13325\relax``{|{\rm Dom}(\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})\cap\theta|}<\theta".
\[Why? We have that for some $`\theta ^{}(\kappa ,\kappa ^{})`$,
(p+,)``Dom(r
~
)θ{Dom(F(f
~
):f
~
Q
~
(κ,θ)N}".(p^{+},\emptyset)\mathchar 13325\relax``{\rm Dom}(\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})\cap\theta\subseteq\bigcup\{{\rm Dom}(F(\mathchoice{\oalign{$\displaystyle f$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle f$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle f$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle f$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}):\,\mathchoice{\oalign{$\displaystyle f$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle f$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle f$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle f$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}\in\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{N}_{(\kappa,\theta^{\prime})}\}".
But then, (p+,)``|Q
~
(κ,θ)N||2θN|<κ"superscript𝑝``subscriptsuperscript𝑄
~
𝑁𝜅superscript𝜃superscript2superscript𝜃𝑁superscript𝜅"(p^{+},\emptyset)\mathchar 13325\relax``{|\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{N}_{(\kappa,\theta^{\prime})}|}\leq{|2^{\theta^{\prime}}\cap N|}<\kappa^{\ast}".\]
(b) If (p+,)``θDom(r
~
)"superscript𝑝``𝜃Dom𝑟
~
"(p^{+},\emptyset)\mathchar 13325\relax``\theta\in{\rm Dom}(\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})", then (p+,)``r
~
(θ)superscript𝑝``𝑟
~
𝜃(p^{+},\emptyset)\mathchar 13325\relax``\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}(\theta) is a function whose domain is an ordinal $`<\theta `$ and range a subset of $`\kappa ^{}`$”.
\[Why? As (p+,)``H
~
superscript𝑝``superscript𝐻
~
(p^{+},\emptyset)\mathchar 13325\relax``\mathchoice{\oalign{$\displaystyle H$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle H$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle H$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle H$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{\ast} is directed”, we have that (p+,)``r
~
(θ)superscript𝑝``𝑟
~
𝜃(p^{+},\emptyset)\mathchar 13325\relax``\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}(\theta) is a function”. If
(p+,)``θDom(h
~
)&F1(h
~
)H
~
",superscript𝑝``𝜃Dom
~
superscript𝐹1
~
𝐻
~
"(p^{+},\emptyset)\mathchar 13325\relax``\theta\in{\rm Dom}(\mathchoice{\oalign{$\displaystyle h$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle h$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle h$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle h$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})\,\,\&\,\,F^{-1}(\mathchoice{\oalign{$\displaystyle h$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle h$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle h$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle h$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})\in\mathchoice{\oalign{$\displaystyle H$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle H$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle H$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle H$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}",
then $`(p^+,\mathrm{})`$ forces
``(σDom(F1(h
~
)))[F1(h
~
)(σ) is a function with domain σ]",``for-all𝜎Domsuperscript𝐹1
~
delimited-[]superscript𝐹1
~
𝜎 is a function with domain 𝜎"``(\forall\sigma\in{\rm Dom}(F^{-1}(\mathchoice{\oalign{$\displaystyle h$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle h$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle h$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle h$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})))\,[F^{-1}(\mathchoice{\oalign{$\displaystyle h$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle h$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle h$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle h$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})(\sigma)\mbox{ is a function with domain }\in\sigma]",
so by elementarity
(p+,)``dom(h
~
(θ)) is an element of θ".](p^{+},\emptyset)\mathchar 13325\relax``{\rm dom}(\mathchoice{\oalign{$\displaystyle h$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle h$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle h$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle h$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}(\theta))\mbox{ is an element of
}\theta".]
(c) (p+,)``r
~
Q
~
(κ,λ)superscript𝑝``𝑟
~
subscript𝑄
~
superscript𝜅superscript𝜆(p^{+},\emptyset)\mathchar 13325\relax``\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}\in\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{(\kappa^{\ast},\lambda^{\ast})}.
\[Why? First of all, we know that (p+,)``dom(r
~
)λsuperscript𝑝``dom𝑟
~
superscript𝜆(p^{+},\emptyset)\mathchar 13325\relax``{\rm dom}(\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})\subseteq\lambda^{\ast} with no last element”. Now, for relevant $`\theta `$, dom(r
~
(θ))dom𝑟
~
𝜃{\rm dom}(\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}(\theta)) is the union of a subset of $`(\theta ^{++})N`$ which has cardinality $`|\theta ^{++}N|<\kappa ^{}`$. Hence the union has cofinality $`<\kappa ^{}`$ (as having cofinality $`\kappa ^{}`$ is preserved by the forcing), hence r
~
(θ)𝑟
~
𝜃\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}(\theta) is forced to be in P
~
(θ,κ)𝑃
~
𝜃superscript𝜅\mathchoice{\oalign{$\displaystyle{{P}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{P}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{P}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{P}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}(\theta,\kappa^{\ast}), and hence $`r`$ $`\stackrel{~}{}`$ is forced to be an element of Q
~
(κ,λ)subscript𝑄
~
superscript𝜅superscript𝜆\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{(\kappa^{\ast},\lambda^{\ast})}.\]
(d) $`(p^+,\mathrm{})`$ forces that for all $`\epsilon S_\kappa ^{}^\theta `$, there is a club $`e`$ of $`\epsilon `$ on which r
~
(θ)𝑟
~
𝜃\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}(\theta) is strictly increasing, for θdom(r
~
)𝜃dom𝑟
~
\theta\in{\rm dom}(\mathchoice{\oalign{$\displaystyle r$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle r$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle r$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}).
\[Why? Because this is forced about each h
~
(θ)
~
𝜃\mathchoice{\oalign{$\displaystyle h$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle h$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle h$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle h$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}(\theta) for h
~
H
~
~
superscript𝐻
~
\mathchoice{\oalign{$\displaystyle h$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle h$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle h$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle h$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}\in\mathchoice{\oalign{$\displaystyle H$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle H$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle H$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle H$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{\ast} and there are $`<\kappa ^{}`$ such $`h`$ $`\stackrel{~}{}`$ .\]
But now we shall see that $`q`$ forces $`\tau `$ $`\stackrel{~}{}`$ to be constant on a stationary subset of $`\delta `$, a contradiction, as $`\delta S_\kappa ^{}^\lambda ^{}`$, and remains there after forcing with $`P`$. We need to consider what $`q`$ forces about τ
~
(α)𝜏
~
𝛼\mathchoice{\oalign{$\displaystyle\tau$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}(\alpha) for $`\alpha N`$. Such τ
~
(α)𝜏
~
𝛼\mathchoice{\oalign{$\displaystyle\tau$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}(\alpha) is a $`P`$-name of an ordinal $`<\kappa ^{}`$. Let
α=def{(,,t
~
)P:(,,t
~
) forces
τ
~
(α) to be equal to a P
R
~
-name}.superscriptdefsubscript𝛼conditional-set𝑡
~
P𝑡
~
forces
τ
~
𝛼 to be equal to a superscriptP
R
~
-name{\cal I}_{\alpha}\buildrel\rm def\over{=}\{(\emptyset,\emptyset,\mathchoice{\oalign{$\displaystyle t$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle t$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle t$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle t$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})\in{{P}}:\,(\emptyset,\emptyset,\mathchoice{\oalign{$\displaystyle t$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle t$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle t$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle t$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})\mbox{ forces }\mathchoice{\oalign{$\displaystyle\tau$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}(\alpha)\mbox{ to be equal to a }{{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}\mbox{-name}\}.
Hence $`_\alpha N`$. As PR
~
superscript𝑃𝑅
~
{{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}} forces that Q
~
(κ,λ)subscript𝑄
~
superscript𝜅superscript𝜆\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{(\kappa^{\ast},\lambda^{\ast})} is $`(\kappa ^{}+1)`$-strategically closed, we have that $`_\alpha `$ is dense in [(PR
~
)Q
~
(κ,λ)]superscriptdelimited-[]superscript𝑃𝑅
~
subscript𝑄
~
superscript𝜅superscript𝜆tensor-product[({{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})\ast\mathchoice{\oalign{$\displaystyle Q$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle Q$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle Q$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}_{(\kappa^{\ast},\lambda^{\ast})}]^{\otimes}. By the definition of H
~
superscript𝐻
~
\mathchoice{\oalign{$\displaystyle H$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle H$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle H$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle H$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{\ast}, there is (,,h
~
)αN
~
subscript𝛼𝑁(\emptyset,\emptyset,\mathchoice{\oalign{$\displaystyle h$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle h$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle h$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle h$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})\in{\cal I}_{\alpha}\cap N such that (,,h
~
)q
~
𝑞(\emptyset,\emptyset,\mathchoice{\oalign{$\displaystyle h$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle h$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle h$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle h$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}})\leq q. Let τ
~
superscript𝜏
~
\mathchoice{\oalign{$\displaystyle\tau$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{\prime} exemplify this, so τ
~
Nsuperscript𝜏
~
𝑁\mathchoice{\oalign{$\displaystyle\tau$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{\prime}\in N.
Hence $`q`$ forces τ
~
(α)𝜏
~
𝛼\mathchoice{\oalign{$\displaystyle\tau$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}(\alpha) to be in the set of all τ
~
Nsuperscript𝜏
~
𝑁\mathchoice{\oalign{$\displaystyle\tau$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{\prime}\in N, where τ
~
superscript𝜏
~
\mathchoice{\oalign{$\displaystyle\tau$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}^{\prime} is a PR
~
superscript𝑃𝑅
~
{{P}}^{-}\ast\mathchoice{\oalign{$\displaystyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle{{R}}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}-name of an ordinal $`<\kappa ^{}`$. The cardinality of this set $`|𝒫(\kappa ^{})N|`$, which is $`<\mu _0`$. Since $`\alpha N`$ was arbitrary, $`q`$ forces the range of τ
~
(Nδ)𝜏
~
𝑁𝛿\mathchoice{\oalign{$\displaystyle\tau$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle\tau$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}\mathchar 13334\relax(N\cap\delta) to be a set of size $`<\mu _0`$, hence $`\tau `$ $`\stackrel{~}{}`$ will be constant on a stationary subset of $`N\delta `$ (as $`N\delta `$ is stationary). $`\mathrm{}_{\text{1.23}}`$
Proof of the Theorem continued.
(2) Same proof.
(3) Follow the forcing from (1) by a Levy collapse. We are making use of the following
###### Claim 1.24
Suppose $`\lambda `$ weakly reflects at $`\kappa `$ and $`P`$ is a $`\kappa `$-cc forcing.
Then $`\lambda `$ weakly reflects at $`\kappa `$ in $`V^P`$.
Proof of the Claim. Suppose that
pP``f
~
:λκ".:subscript𝑃𝑝``𝑓
~
𝜆𝜅"p\mathchar 13325\relax_{P}``\mathchoice{\oalign{$\displaystyle f$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle f$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle f$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle f$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}:\,\lambda\rightarrow\kappa".
We define $`f^{}:\lambda \kappa `$ by letting f(α)=defsup{γ<κ:¬(p``f
~
(α)γ")}superscriptdefsuperscript𝑓𝛼supremumconditional-set𝛾𝜅p``𝑓
~
𝛼𝛾"f^{\prime}(\alpha)\buildrel\rm def\over{=}\sup\{\gamma<\kappa:\,\neg(p\mathchar 13325\relax``\mathchoice{\oalign{$\displaystyle f$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle f$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle f$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle f$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}(\alpha)\neq\gamma")\}. As $`P`$ is $`\kappa `$-cc, the range of $`f^{}`$ is indeed contained in $`\kappa `$. Let $`\delta S_\kappa ^\lambda `$ be such that $`f^{}S`$ is constant on a stationary set $`S\delta `$ (the existence of such $`\delta `$ follows as $`WR(\lambda ,\kappa )`$ holds). Hence p``f
~
S𝑝``𝑓
~
𝑆p\mathchar 13325\relax``\mathchoice{\oalign{$\displaystyle f$\crcr\vbox to0.86108pt{\hbox{$\displaystyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\textstyle f$\crcr\vbox to0.86108pt{\hbox{$\textstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptstyle f$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}{\oalign{$\scriptscriptstyle f$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle{\tilde{\mkern-3.0mu}\mkern 3.0mu}{}$}\vss}}}\mathchar 13334\relax S is bounded”, so $`f`$ $`\stackrel{~}{}`$ does not witness $`SNR(\lambda ,\kappa )`$ in $`V^P`$, as $`S`$ remains stationary in $`V^P`$. $`\mathrm{}\text{1.24}`$
So, for example to get $`\kappa ^{}=\mathrm{}_n`$ for $`n1`$, we could in $`V^P`$ from (1) first make $`GCH`$ hold below $`\mu _0`$ by collapsing various cardinals below $`\mu _0`$, and then collapse $`\kappa ^{}`$ to be $`\mathrm{}_n`$.
If we start in $`V`$ by having $`\kappa ^{}=\mu _0^{\omega +1}`$, and in $`V^P`$ collapse $`\mu _0`$ to $`\mathrm{}_1`$, then we get $`\kappa ^{}=\mathrm{}_{\omega +1}`$.
For the last statement, start by $`\kappa ^{}=\mu _0^{+\delta }`$ for some $`\delta \mathrm{}_7`$, so $`\lambda ^{}=\mu _1^{+\delta }`$. By a minor change in the definition of $`P^{}`$, we can make $`P`$ not add any $`\mathrm{}_7`$-sequences from $`V`$. Now in $`V^P`$, collapse first $`\mu _0`$ to $`\mathrm{}_7`$, and then collapse $`\mu _1`$ to $`(\kappa ^{})^{+3}`$. Hence $`\lambda ^{}=(\kappa ^{})^{+3+\delta }`$. $`\mathrm{}_{\text{0.2}}`$
|
warning/0003/cond-mat0003111.html
|
ar5iv
|
text
|
# Nonequilibrium phase transition in a model for social influence.
## Abstract
We present extensive numerical simulations of the Axelrod’s model for social influence, aimed at understanding the formation of cultural domains. This is a nonequilibrium model with short range interactions and a remarkably rich dynamical behavior. We study the phase diagram of the model and uncover a nonequilibrium phase transition separating an ordered (culturally polarized) phase from a disordered (culturally fragmented) one. The nature of the phase transition can be continuous or discontinuous depending on the model parameters. At the transition, the size of cultural regions is power-law distributed.
Recently, the study of complex systems entered social science in order to understand how self-organization, cooperative effects and adaptation arise in social systems. In this context the use of simple automata or dynamical models often elucidates the mechanisms at the basis of the observed complex behaviors.
In this spirit, R. Axelrod has recently proposed an interesting model to mimic how dissemination of culture works. Culture is used here to indicate the set of individual attributes, such as language, art or technical standards subject to social influence, i.e. that can be changed as effect of mutual interactions. The automaton does not consider the effect of central institutions or mass media and focuses on the self-organization resulting from a simple local dynamics representing the social influence. This dynamics is assumed to satisfy two simple properties: i) individuals are more likely to interact with others who already share many of their cultural attributes; ii) interaction increases the number of features that individuals share. Starting from an initial state with features distributed randomly this leads to the formation and coarsening of regions of shared culture.
In this Letter we carry out an accurate numerical analysis of Axelrod’s model that unravels a remarkably rich behavior, not detected in previous investigations. Depending on the initial degree of disorder, the model undergoes a phase transition separating an ordered from a disordered phase. The ordered phase is characterized by the growth of a dominant cultural region spanning a large fraction of the whole system. On the contrary, in the disordered phase the system freezes in a highly fragmented state with a nontrivial distribution of the sizes of cultural regions. Such a fragmented configuration is reached in a finite time, which diverges at the phase transition. In the whole ordered phase instead, the coarsening process lasts for a time proportional to the system size, before freezing into the culturally polarized state. Interestingly, the nature of the transition turns from continuous to discontinuous when the number of cultural features is increased. Close to the transition, the distribution of region sizes follows a power law in agreement with the results of Ref. on the statistical analysis of the diversity of languages. Some of these features are captured by a mean field approximation that we discuss below.
Axelrod’s model is defined on a square lattice of linear size $`L`$. On each site $`i`$ there is a set of $`F`$ integer variables $`\sigma _{i,f}`$ which define the cultural “features” of the individuals living on that site. In the original model, each feature $`f=1,\mathrm{},F`$ on each site $`i`$ is initially drawn randomly from a uniform distribution on the integers between $`1`$ and $`q`$. The parameter $`q`$ is a measure of the initial cultural variability (i.e. disorder) in the system. Here, we relax the constraint of integer $`q`$ by using a Poisson distribution $`\mathrm{Prob}(\sigma _{i,f}=k)=q^ke^q/k!`$, that allows $`q`$ to be a positive real number. Though the results are qualitatively the same, our choice is more convenient to study the behavior of the model as $`q`$ varies.
At each time step, a pair of nearest neighbor sites $`i`$ and $`j`$ is randomly chosen. A feature $`f`$ is chosen randomly and if $`\sigma _{i,f}\sigma _{j,f}`$ nothing happens. If instead $`\sigma _{i,f}=\sigma _{j,f}`$ then an additional feature $`f^{}`$ is randomly chosen among those taking different values across the bond, $`\sigma _{i,f^{}}\sigma _{j,f^{}}`$. Such a feature is then set equal: $`\sigma _{i,f^{}}\sigma _{i,f^{}}^{}=\sigma _{j,f^{}}`$. A “sweep” of the lattice, i.e. $`L^2`$ such time steps, defines the time unit as usual. Axelrod’s model (at least in the original formulation) can be seen as $`F`$ coupled voter models .
During the dynamical evolution the total diversity, measured as the number of different values of a feature $`f`$ which are present in the system, always decreases. Clearly if all features are equal across a bond ($`\sigma _{i,f}=\sigma _{j,f}`$, $`f`$) or if they are all different ($`\sigma _{i,f}\sigma _{j,f}`$ $`f`$) no change can occur on the bond $`ij`$. A configuration such that, on each bond, either all features are equal or they are all different is an absorbing state: Dynamics will stop if such a state is reached. There are clearly many absorbing states: The dynamics on any finite lattice converges to one of them. The final state can be characterized by the distribution of cultural region sizes, where a region is defined as the connected set of sites sharing exactly the same features.
The dynamical evolution is characterized by the competition between the disorder of the initial configuration and the ordering drive due to the local social interactions. It is intuitively clear that when $`q`$ is small the initial state is almost completely uniform, whereas for large values of $`q`$ almost all sites have features $`\sigma _{i,f}`$ totally different from those of their nearest neighbors. In the two cases we expect the system to converge to a uniform or a highly fragmented state, in which interaction or disorder dominate, respectively. In order to understand how these two limit situations are connected as $`q`$ varies, we have simulated the Axelrod’s model for a number of features ranging from $`F=2`$ to $`10`$ and sizes up to $`L=150`$. We first discuss the dependence of the final state on the parameters $`q`$ and $`F`$ and then the dynamical behavior of the model.
The frozen state. In any finite lattice the dynamics converges to a frozen absorbing state. The existence of a transition in the properties of the final absorbing states is very clear from the plot (Fig. 1) of the average size of the largest region $`s_{\mathrm{max}}`$ as a function of $`q`$ for $`F=10`$: For $`q=q_c300`$ we observe a sharp transition characterized by a sudden drop of $`s_{\mathrm{max}}/L^2`$ which becomes steeper and steeper for increasing sizes $`L`$. This points to the existence of a transition between a “culturally polarized phase” for $`q<q_c`$, where one of the regions has a size of the order of the whole system, and a “culturally fragmented” phase, where all domains are finite. The transition is of the first order, with the size of the largest region having a finite discontinuity at $`q=q_c`$. Note also that, for $`q<q_c`$ the largest domain approaches a unitary density, i.e. it invades the whole system. This scenario holds for all values $`F>2`$ investigated.
The situation is different for $`F=2`$ (inset of Fig. 1): The fraction occupied by the largest cluster $`s_{\mathrm{max}}/L^2`$ vanishes continuously as $`qq_c^{}`$. The difference between $`F=2`$ and $`F>2`$ is confirmed and clarified by the study of the size distribution of cultural regions at the transition. Let $`P_L(s,q)`$ be the probability distribution of regions of $`s`$ sites in a system of size $`L`$. The cumulated distribution is plotted in Fig. 2 for several $`F`$ and values of $`q`$ around the transition. Fig. 2 shows that $`P_L(s,q)`$ decays as a power law $`s^\tau `$ and that the exponent $`\tau `$ is universal ($`\tau 2.6`$) for $`F>2`$ but takes a different value for $`F=2`$ ($`\tau 1.6`$). In particular we find $`\tau _{F>2}>2`$ and $`\tau _{F=2}<2`$.
The different nature of the transition for $`F`$ equal or greater than 2 can be related to the exponent $`\tau `$. Let $`N(q,L)`$ be the total number of regions in the system. Requiring the total area to be $`L^2`$ leads to
$$L^2=N(q,L)s=N(q,L)\underset{s=1}{\overset{\mathrm{}}{}}sP_L(s,q).$$
(1)
For $`q>q_c`$ there are $`N(q,L)L^2`$ domains of finite size and the sum on $`s`$ is finite as $`L\mathrm{}`$. On the other hand, for $`q<q_c`$, there are few small domains and a large one of size $`s_{\mathrm{max}}L^2`$. Hence the probability distribution can be written in the generic scaling form $`P_L(s,q)=s^\tau (s/s_{co})+A(q)\delta _{s,s_{\mathrm{max}}}`$ where $`s_{co}`$ is a cut-off scale, the function $`(x)`$ is constant for $`x1`$ and decays very rapidly for $`x1`$ and $`A(q)=0`$ for $`q>q_c`$.
The divergence $`L^2`$ in the l.h.s. of Eq. (1) is matched, for $`q<q_c`$, by the component $`A\delta _{s,s_{\mathrm{max}}}`$ in $`P_L(s,q)`$, with $`s_{\mathrm{max}}L^2`$. The nature of the transition is identified by the behavior of $`A(q)`$ for $`qq_c^{}`$. For $`\tau <2`$, similarly to what happens in percolation theory , the transition occurs through the divergence of $`s_{co}`$ and hence of a correlation length, as $`qq_c^+`$. This causes the divergence of $`s`$ in Eq. (1) (because $`sP_L(s,q)s^{1\tau }`$ with $`\tau <2`$), which matches the divergence of the l.h.s. of Eq. (1) as $`L\mathrm{}`$, through usual finite size scaling arguments . Indeed, at $`q=q_c`$ the cut-off diverges with the system size as $`s_{co}L^D`$. On the other side of the transition a similar divergence of the cutoff $`s_{co}`$ occurs and the amplitude $`A`$ must vanish as $`qq_c^{}`$. This scenario is typical of second order phase transitions and agrees with the value $`\tau <2`$ found for $`F=2`$ and Fig. 1.
When $`\tau >2`$ the previous scenario cannot hold: For $`q>q_c`$, even if $`s_{co}`$ diverges, the sum on $`s`$ in Eq. (1) remains finite. Hence as $`qq_c^+`$ the number of domains $`N(q,L)`$ must remain of order $`L^2`$. On the other side of the transition, the $`L^2`$ term in the l.h.s. of Eq. (1) can only be matched by the term $`s_{\mathrm{max}}L^2`$. Since no divergence arises from the sum on small components as $`qq_c^{}`$, the amplitude $`A`$ of the large component must remain of order one in this limit. We then conclude that, as $`q`$ crosses $`q_c`$, the nature of the distribution changes abruptly for $`\tau >2`$. In particular the amplitude $`A(q)`$ exhibits a discontinuous jump across the transition .
Dynamics: In order to investigate the model dynamics we study the density $`n_a`$ of active bonds. An active bond is a bond across which at least one feature is different and at least one is equal, so that there can be some dynamics. This quantity is indicative of the dynamical state of the system, being zero in frozen configurations. In Fig. 3 we show the behavior of $`n_a`$ as a function of time for different values of $`q`$ and $`F=10`$. For large values of $`q`$, after a short initial transient, the density of active bonds decays rapidly and the system locks into a frozen configuration in a finite time. For small $`q`$ instead, $`n_a(t)`$ displays a slow decay with a large majority of bonds in the active state until it falls abruptly for long times. As the system size is increased, the slow decay extends to longer and longer times: The cut-off time, at which activity suddenly dies scales as $`t_{co}L^2`$ as shown in the inset of Fig. 3. For a large range of intermediate $`q`$ values, $`n_a`$ first decreases almost to zero but then rises again towards a peak of activity, from which the slow decay begins.
The dynamics is essentially a coarsening process of homogeneous regions. When the process lasts only for a finite time, as for $`q>q_c`$, it gives rise to regions of finite size. On the contrary, for $`q<q_c`$ the coarsening process goes on for a time $`t_{co}L^2`$ and produces a region of size comparable with the whole system. The exponent $`z=2`$ relating $`t_{co}`$ to $`L`$ is the same found in nonconserved phase-ordering and in particular in the two-dimensional voter model , which also exhibits a slow decay of the density of active bonds .
For an infinite system ($`L\mathrm{}`$) and $`q<q_c`$ the system is indefinitely in a coarsening state. Therefore we can also define the transition as separating two different dynamical regimes, with a fast decay of $`n_a`$ lasting for a finite characteristic time for $`q>q_c`$ and a slow, infinitely long decay for $`q<q_c`$. For $`F=2`$ the behavior is qualitatively the same: Noticeably, no evident signature of the different nature of the transition (continuous vs. discontinuous) can be inferred from the dynamical evolution.
The dynamical behavior of the model can be studied within a single bond mean field treatment. Let $`P_m(t)`$ be the probability that a randomly picked bond is of type $`m`$ at time $`t`$, i. e. $`m`$ features across the bond are equal and $`Fm`$ are different. At $`t=0`$, since features are assigned uncorrelated random values, we have $`P_m(0)=\left(\genfrac{}{}{0pt}{}{F}{m}\right)\rho _0^m(1\rho _0)^{Fm}`$ where $`\rho _0=\mathrm{Prob}[\sigma _{i,f}=\sigma _{j,f}]`$ is the probability that two sites have feature $`f`$ with the same value at $`t=0`$ In the mean field approximation, $`P_m`$ satisfies the master equation
$`{\displaystyle \frac{dP_m}{dt}}`$ $`=`$ $`{\displaystyle \underset{k=1}{\overset{F1}{}}}{\displaystyle \frac{k}{F}}P_k[\delta _{m,k+1}\delta _{m,k}+`$ (2)
$`(g1){\displaystyle \underset{n=0}{\overset{F}{}}}(P_nW_{n,m}^{(k)}P_mW_{m,n}^{(k)})],`$ (3)
where $`g`$ is the lattice coordination number and $`W_{n,m}^{(k)}`$ is the transition probability from a $`n`$-type bond to a $`m`$-type bond due to the updating of a $`k`$-type neighbor bond. This equation describes how the number of bonds of type $`m`$ is affected by the dynamics: $`P_kk/F`$ is the probability to select a bond of type $`k`$ and one of the $`k`$ features which are equal across it. If $`k=m1`$ a new bond of type $`m`$ is created whereas if $`k=m`$ one bond of type $`m`$ is deleted. This explains the first two terms on the r. h. s. of Eq. (3). But the change of a feature of the site may also affect the state of the other $`g1`$ bonds connecting such site to its neighbors. The remaining terms in Eq. (3) take into account this possibility. We can compute the probabilities $`W_{n,m}^{(k)}`$ by analyzing in detail each possible process. Let us consider e.g. a type $`k=1`$ bond adjacent to a type $`n=0`$ one. Without loss of generality we can let the features on the extreme sites of the $`1`$ bond be $`(0,0,\mathrm{},0)`$ and $`(0,\sigma _2,\mathrm{},\sigma _F)`$, with $`\sigma _j0`$. The latter site is shared also by the $`0`$ bond whose other site has features $`(\sigma _1^{},\sigma _2^{},\mathrm{},\sigma _F^{})`$ with $`\sigma _1^{}0`$ and $`\sigma _j^{}\sigma _j`$ for $`j>1`$. When the $`1`$ bond becomes a type $`2`$ bond, $`\sigma _20`$ the neighbor $`0`$ bond can either remain a $`0`$ bond, if $`\sigma _2^{}0`$ or it can become a $`1`$ bond if $`\sigma _2^{}=0`$. In the spirit of the mean field approximation we introduce the probability $`\rho =\mathrm{Prob}(\sigma _2^{}=0)`$ which now becomes time dependent, as we shall see. Then $`W_{0,0}^{(1)}=1\rho `$ and $`W_{0,1}^{(1)}=\rho `$. In much the same way we can compute the other transition rates. For example, if $`F=2`$, the only non zero elements are
$$\begin{array}{ccc}W_{0,0}^{(1)}=1\rho ,\hfill & W_{0,1}^{(1)}=\rho \hfill & \\ W_{1,0}^{(1)}=1/2,\hfill & W_{1,1}^{(1)}=(1\rho )/2,\hfill & W_{1,2}^{(1)}=\rho /2\hfill \\ W_{2,1}^{(1)}=1.\hfill & & \end{array}$$
(4)
The system of equations (3) is closed once the dynamics of $`\rho (t)`$ is specified. The simplest such equation, in the mean-field spirit, is $`\rho =_kkP_k/F`$. This amounts to say that between any two sites there is a bond, and hence that the probability that a feature across that bond takes the same value can be expressed in terms of the $`P_m`$. The numerical integration of the mean field equations yields the phase diagram of the model as shown in Fig. 4, which exhibits a phase transition. The order parameter $`n_a=_{k=1}^{F1}P_k`$ undergoes a discontinuity at $`q=q_c`$, as it jumps from a finite value for $`q<q_c`$ to zero for $`q>q_c`$. The mean field fairly reproduces also the dynamical evolution of the model as can be seen from the inset of Fig. 4.
Despite the extreme simplicity of Axelrod’s model, it is tempting to compare the picture derived in this Letter with some recent analysis on the diversity of languages, which to some extent can be considered as an indicator of cultural homogeneity. Gomes and coauthors, by analyzing the statistics of more than 6700 languages, have found that the number $``$ of linguistically homogeneous regions inhabited by $`N`$ individuals scales as $`N^\tau `$, with $`\tau 1.5`$ for populations smaller than $`6\times 10^6`$ individuals. For larger populations the exponent changes suggesting a different dynamical evolution. These findings can be compatible with our continuous phase transition scenario ($`\tau 1.6`$) assuming the equivalence between language and cultural features. However it remains unclear why language spreading is self-organized close to the transition point or why the value $`F=2`$ should be relevant for the process.
In summary, we have presented numerical simulations of the Axelrod’s model for social influence. Social interaction that tends to make culture homogeneous competes with the disorder introduced by the number of different traits each cultural feature initially has. We find a critical value separating two phases in which one of the above elements (interaction or disorder) respectively dominates. The nature of the phase transition and the emergent collective behavior is analyzed by common tools of statistical physics. We find that the transition is continuous or discontinuous depending on the model parameters. Our study shows that the use of concepts and methods developed in physics may be of help in the context of social sciences. For instance, the prevalence, for small values of $`q`$, of one of the initially equivalent cultures can be regarded as a spontaneous symmetry breaking due to stochastic fluctuations. It would be interesting, in the future, to analyze extensions of the simple Axelrod’s model to take into account additional ingredients like the presence of migrations or the effect of geographical barriers.
C. C. acknowledges support from the Alexander Von Humboldt foundation. A. V. acknowledges partial support from the European Network contract ERBFMRXCT980183.
|
warning/0003/hep-ph0003139.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
The observation of the atmospheric neutrino by Super-Kamiokande has shown the existence of the neutrino masses and the neutrino mixings. In particular, the data show that the mixing between $`\nu _\mu `$ and $`\nu _\tau `$ is favored and
$`\mathrm{sin}^22\theta _{atm}1,\mathrm{\Delta }_{atm}^2|m_3^2m_2^2|3.5\times 10^3\mathrm{eV}^2.`$ (1)
The solar neutrino problem is now considered to be due to the $`\nu _e`$ and $`\nu _\mu `$ oscillation, but the information on masses and mixing angles is ambiguous. Now four solutions are possible. The another crucial information is given by CHOOZ group that gives
$`|(V_{MNS})_{13}|<0.16,`$ (2)
provided that $`\mathrm{\Delta }_{atm}^2=3.5\times 10^3\mathrm{eV}^2`$. Here, $`V_{MNS}`$ is the Maki-Nakagawa-Sakata (MNS) neutrino mixing matrix.
If we interpret these information in the three generation mixing, we find
$`\mathrm{sin}^22\theta _{atm}\mathrm{sin}^22\theta _{23}c_{13}^4`$ (3)
and
$`|s_{13}|<0.16,`$ (4)
where $`\theta _{ij}`$ is the mixing angle between the i-th and the j-th mass eigenstate neutrinos, $`s_{ij}=\mathrm{sin}\theta _{ij}`$, $`c_{ij}=\mathrm{cos}\theta _{ij}`$, and we used the standard parameterization of mixing matrix given in Appendix A.
If we obtain the large $`\mathrm{sin}^22\theta _{atm}`$ with $`|s_{13}|<0.16`$, we need $`|s_{23}||c_{23}|1/\sqrt{2}`$. Now we face the following questions:
(1) Why $`|s_{23}||c_{23}|1/\sqrt{2}`$?
(2) Why $`s_{13}`$ is so small?
(3) How large is the CP violation phase $`\delta `$?
In order to answer these questions, various mixing schemes have been proposed. Among them, the bi-maximal mixing matrix, $`(V_{MNS})_{bi}`$, and the democratic one, $`(V_{MNS})_{demo}`$, predict the large mixing for both the solar and the atmospheric neutrino mixings,
$`(V_{MNS})_{bi}=\left(\begin{array}{ccc}\frac{1}{\sqrt{2}}& \frac{1}{\sqrt{2}}& 0\\ \frac{1}{2}& \frac{1}{2}& \frac{1}{\sqrt{2}}\\ \frac{1}{2}& \frac{1}{2}& \frac{1}{\sqrt{2}}\end{array}\right),(V_{MNS})_{demo}=\left(\begin{array}{ccc}\frac{1}{\sqrt{2}}& \frac{1}{\sqrt{2}}& 0\\ \frac{1}{\sqrt{6}}& \frac{1}{\sqrt{6}}& \frac{2}{\sqrt{6}}\\ \frac{1}{\sqrt{3}}& \frac{1}{\sqrt{3}}& \frac{1}{\sqrt{3}}\end{array}\right).`$ (5)
The bi-maximal mixing matrix predicts $`\mathrm{sin}^22\theta _{sol}=1`$ and $`\mathrm{sin}^22\theta _{atm}=1`$, and the democratic mixing matrix predicts $`\mathrm{sin}^22\theta _{sol}=1`$ and $`\mathrm{sin}^22\theta _{atm}=8/9`$.
In our previous paper, we proposed a new type of the neutrino mixing matrix based on the democratic-type mass matrix which is derived from the $`Z_3`$ invariant Lagrangian. This model predicts the most needed relation $`|s_{23}|=|c_{23}|=1/\sqrt{2}`$ and in addition $`\delta =\pi /2`$, the largest CP violation phase in the standard parameterization. In this paper, we examine the democratic-type mass matrix further and explore the possibility of constructing more predictive models.
This paper is organized as follows: In Sec.2, we briefly review the democratic-type mass matrix and its predictions. We explain the origin of the mass matrix based on $`Z_3`$ symmetry, by using the see-saw mechanism. In Sec.3, we consider the one Higgs doublet case and introduce the $`Z_3`$ symmetry breaking terms. The predictions are discussed in detail. The summary is given in Sec.4.
## 2 The democratic-type mass matrix and the $`Z_3`$ symmetry
In this section, we present another view of the democratic-type mass matrix and its origin. Throughout of this paper, we consider the neutrino mass matrix in the charged lepton mass eigenstate basis.
(a) The mixing matrix derived from the deformation from the tri-maximal mixing matrix
The tri-maximal mixing matrix was discussed extensively by many authors and is defined by
$`(V_{MNS})_{tri}V_T={\displaystyle \frac{1}{\sqrt{3}}}\left(\begin{array}{ccc}1& 1& 1\\ \omega & \omega ^2& 1\\ \omega ^2& \omega & 1\end{array}\right),`$ (6)
where $`\omega =e^{i2\pi /3}`$ or $`\omega =e^{i4\pi /3}`$, i.e., $`\omega ^3=1`$. This model predicts $`|(V_{MNS})_{13}|=1/\sqrt{3}`$ which conflicts with the CHOOZ data, but it has an interesting property that it predicts the maximal value of the CP violation phase in the standard form, $`\delta =\pi /2`$, and the maximal CP violation, that is, the maximal value of the rephasing invariant Jarlskog parameter, $`|J_{CP}|_{tri}=1/6\sqrt{3}`$.
In order to remedy the deficit of the model, it may be interesting to consider the deformation from the tri-maximal mixing matrix by an orthogonal matrix $`O`$,
$`V=V_TO.`$ (7)
The orthogonal matrix contains three angles so that naively we expect that the CP violation phase is expressed by three angles in a complicated expression. Contrarily to this expectation, we found
$`|\mathrm{sin}\theta _{23}|=|\mathrm{cos}\theta _{23}|={\displaystyle \frac{1}{\sqrt{2}}},\delta ={\displaystyle \frac{\pi }{2}}`$ (8)
and two Majorana phases are fixed uniquely independent of angle parameters in $`O`$. Explicitly, we found that the neutrino mixing matrix is
$`V_{MNS}=\left(\begin{array}{ccc}c_{12}c_{13}& s_{12}c_{13}& is_{13}\\ \frac{s_{12}ic_{12}s_{13}}{\sqrt{2}}& \frac{c_{12}+is_{12}s_{13}}{\sqrt{2}}& \frac{c_{13}}{\sqrt{2}}\\ \frac{s_{12}+ic_{12}s_{13}}{\sqrt{2}}& \frac{c_{12}is_{12}s_{13}}{\sqrt{2}}& \frac{c_{13}}{\sqrt{2}}\end{array}\right)\left(\begin{array}{ccc}1& 0& 0\\ 0& 1& 0\\ 0& 0& i\end{array}\right).`$ (9)
The diagonal matrix diag$`(1,1,i)`$ represents the Majorana phase matrix which is relevant to the purely lepton number violating processes such as the neutrinoless double beta decay. One parameter out of three parameters in $`O`$ is converted to the phase which is absorbed by the redefinition of charged leptons. The brief derivation is given in Appendix A.
If we use the CHOOZ constraint, $`|(V_{MNS})_{13}|=|s_{13}|<0.16`$, the model predicts
$`\mathrm{sin}^22\theta _{atm}`$ $`=`$ $`4|(V_{MNS})_{23}|^2(1|(V_{MNS})_{23}|^2)`$
$`=`$ $`1s_{13}^4>0.999,`$
$`{\displaystyle \frac{|J_{CP}|_{ourmodel}}{|J_{CP}|_{max}}}`$ $`=`$ $`{\displaystyle \frac{3\sqrt{3}}{2}}|\mathrm{sin}2\theta _{12}s_{13}c_{13}^2|,`$ (10)
where $`\mathrm{sin}^22\theta _{12}\mathrm{sin}^22\theta _{sol}`$ is determined by the solar neutrino data. Therefore, in our model, the size of the CP violation is determined by the solar neutrino mixing angle and $`|s_{13}|`$ which has been bounded by the CHOOZ data. If we take the large mixing solution for the solar neutrino problem, $`\mathrm{sin}^22\theta _{sol}=0.9`$ and the largest allowed value for $`s_{13}`$, $`|s_{13}|=0.16`$, our model predicts $`|J_{CP}|_{ourmodel}=0.38|J_{CP}|_{max}`$.
(b) The possible origin of the mixing matrix in the form of $`V=V_TO`$
Let us consider what kind of mass matrix leads to $`V=V_TO`$. In order to clarify the structure, we change the flavor eigenstate basis to the one which is obtained by transforming by $`V_T`$ (hereafter we call it as the $`\psi `$ basis),
$`\left(\begin{array}{c}\psi _1\\ \psi _2\\ \psi _3\end{array}\right)=V_T^{}\left(\begin{array}{c}\nu _{eL}\\ \nu _{\mu L}\\ \nu _{\tau L}\end{array}\right)={\displaystyle \frac{1}{\sqrt{3}}}\left(\begin{array}{ccc}1& \omega ^2& \omega \\ 1& \omega & \omega ^2\\ 1& 1& 1\end{array}\right)\left(\begin{array}{c}\nu _{eL}\\ \nu _{\mu L}\\ \nu _{\tau L}\end{array}\right).`$ (11)
The relation $`V=V_TO`$ implies that when we look at the mass matrix in the $`\psi `$ basis, the neutrino mass matrix should be a real symmetric matrix. Thus, we may parameterize
$`\stackrel{~}{m}_\nu =V_T^Tm_\nu V_T=\left(\begin{array}{ccc}m_1^0+\stackrel{~}{m}_1& \stackrel{~}{m}_3& \stackrel{~}{m}_2\\ \stackrel{~}{m}_3& m_2^0+\stackrel{~}{m}_2& \stackrel{~}{m}_1\\ \stackrel{~}{m}_2& \stackrel{~}{m}_1& m_3^0+\stackrel{~}{m}_3\end{array}\right),`$ (12)
with the real parameters, $`m_i^0`$ and $`\stackrel{~}{m}_i`$. Here, $`\stackrel{~}{m}_\nu `$ is the mass matrix in the $`\psi `$ basis and $`m_\nu `$ is the one in the flavor eigenstate basis. If $`\stackrel{~}{m}_\nu `$ is a real symmetric matrix, then it is diagonalized by the orthogonal matrix $`O`$ and thus the mixing matrix becomes $`V_TO`$.
By inverting, we obtain the neutrino mass matrix $`m_\nu `$ as
$`m_\nu `$ $`=`$ $`{\displaystyle \frac{m_1^0}{3}}\left(\begin{array}{ccc}1& \omega ^2& \omega \\ \omega ^2& \omega & 1\\ \omega & 1& \omega ^2\end{array}\right)+{\displaystyle \frac{m_2^0}{3}}\left(\begin{array}{ccc}1& \omega & \omega ^2\\ \omega & \omega ^2& 1\\ \omega ^2& 1& \omega \end{array}\right)+{\displaystyle \frac{m_3^0}{3}}\left(\begin{array}{ccc}1& 1& 1\\ 1& 1& 1\\ 1& 1& 1\end{array}\right)`$ (13)
$`+\stackrel{~}{m}_1\left(\begin{array}{ccc}1& 0& 0\\ 0& \omega & 0\\ 0& 0& \omega ^2\end{array}\right)+\stackrel{~}{m}_2\left(\begin{array}{ccc}1& 0& 0\\ 0& \omega ^2& 0\\ 0& 0& \omega \end{array}\right)+\stackrel{~}{m}_3\left(\begin{array}{ccc}1& 0& 0\\ 0& 1& 0\\ 0& 0& 1\end{array}\right),`$
which we called the democratic-type mass matrix.
(c) The $`Z_3`$ symmetric dimension five Lagrangian
We analyze the Lagrangian which gives the democratic-type neutrino mass matrix. From the transformation in Eq.(11), we define
$`\mathrm{\Psi }_1`$ $`=`$ $`{\displaystyle \frac{1}{\sqrt{3}}}(\mathrm{}_e+\omega ^2\mathrm{}_\mu +\omega \mathrm{}_\tau ),`$
$`\mathrm{\Psi }_2`$ $`=`$ $`{\displaystyle \frac{1}{\sqrt{3}}}(\mathrm{}_e+\omega \mathrm{}_\mu +\omega ^2\mathrm{}_\tau ),`$
$`\mathrm{\Psi }_3`$ $`=`$ $`{\displaystyle \frac{1}{\sqrt{3}}}(\mathrm{}_e+\mathrm{}_\mu +\mathrm{}_\tau ),`$ (14)
where $`\mathrm{}_i`$ is the left-handed lepton doublet defined by, say, $`\mathrm{}_e=(\nu _{eL},e_L)^T`$. With the definition, $`\mathrm{\Psi }_i=(\psi _i,e_i)^T`$, the above relation is interpreted as the transformation from the flavor eigenstate basis to the $`\psi `$ basis.
The fields $`\mathrm{\Psi }_i`$ behave as irreducible representations of $`Z_3`$ symmetry under the permutation of $`\mathrm{}_e`$, $`\mathrm{}_\mu `$ and $`\mathrm{}_\tau `$,
$`\mathrm{\Psi }_1\omega \mathrm{\Psi }_1,\mathrm{\Psi }_2\omega ^2\mathrm{\Psi }_2,\mathrm{\Psi }_3\mathrm{\Psi }_3.`$ (15)
If we introduce two Higgs doublets that behave as
$`H_1\omega ^2H_1,H_2\omega H_2,`$ (16)
then we can construct the $`Z_3`$ invariant dimension five effective Lagrangian as
$`_\mathrm{y}`$ $`=`$ $`((m_1^0+\stackrel{~}{m}_1)\overline{(\mathrm{\Psi }_1)^C}\mathrm{\Psi }_1{\displaystyle \frac{H_1H_1}{u_1^2}}+(m_2^0+\stackrel{~}{m}_2)\overline{(\mathrm{\Psi }_2)^C}\mathrm{\Psi }_2{\displaystyle \frac{H_2H_2}{u_2^2}}`$ (17)
$`+(m_3^0+\stackrel{~}{m}_3)\overline{(\mathrm{\Psi }_3)^C}\mathrm{\Psi }_3{\displaystyle \frac{H_1H_2}{u_1u_2}})`$
$`2\left(\stackrel{~}{m}_1\overline{(\mathrm{\Psi }_2)^C}\mathrm{\Psi }_3{\displaystyle \frac{H_1H_1}{u_1^2}}+\stackrel{~}{m}_2\overline{(\mathrm{\Psi }_1)^C}\mathrm{\Psi }_3{\displaystyle \frac{H_2H_2}{u_2^2}}+\stackrel{~}{m}_3\overline{(\mathrm{\Psi }_1)^C}\mathrm{\Psi }_2{\displaystyle \frac{H_1H_2}{u_1u_2}}\right),`$
where $`u_i`$ is the vacuum expectation value of the neutral component of $`H_i`$. After the Higgs fields acquire the vacuum expectation values, the neutrino mass matrix in the $`\psi `$ basis defined by Eq.(12) is obtained.
(d) The $`Z_3`$ symmetric Lagrangian and the see-saw mechanism
In addition to the $`\mathrm{\Psi }_i`$ and two Higgs doublets, $`H_1`$ and $`H_2`$, we introduce the right-handed neutrinos, $`\nu _{eR}`$, $`\nu _{\mu R}`$ and $`\nu _{\tau R}`$ and one more Higgs $`H_3`$ which behave under the $`Z_3`$ as
$`\nu _{eR}\nu _{eR},\nu _{\mu R}\omega \nu _{\mu R},\nu _{\tau R}\omega ^2\nu _{\tau R},`$ (18)
$`H_3H_3.`$ (19)
Now the $`Z_3`$ invariant Yukawa interaction and the Majorana mass term for the right-handed neutrinos are given by
$`_D`$ $`=`$ $`(a\overline{\nu _e}_R{\displaystyle \frac{H_1}{u_1}}+b\overline{\nu _\tau }_R{\displaystyle \frac{H_2}{u_2}}+d^{}\overline{\nu _\mu }_R{\displaystyle \frac{H_3}{u_3}})\mathrm{\Psi }_1`$ (20)
$`(c\overline{\nu _\mu }_R{\displaystyle \frac{H_1}{u_1}}+d\overline{\nu _e}_R{\displaystyle \frac{H_2}{u_2}}+f^{}\overline{\nu _\tau }_R{\displaystyle \frac{H_3}{u_3}})\mathrm{\Psi }_2`$
$`(e\overline{\nu _\tau }_R{\displaystyle \frac{H_1}{u_1}}+f\overline{\nu _\mu }_R{\displaystyle \frac{H_2}{u_2}}+b^{}\overline{\nu _e}_R{\displaystyle \frac{H_3}{u_3}})\mathrm{\Psi }_3+\mathrm{h}.\mathrm{c}.,`$
where $`u_i`$ is the vacuum expectation value of $`H_i`$ and
$`_R=M\overline{(\nu _{eR})^C}\nu _{eR}M^{}\left(\overline{(\nu _{\mu R})^C}\nu _{\tau R}+\overline{(\nu _{\tau R})^C}\nu _{\mu R}\right).`$ (21)
After the Higgs fields acquire the vacuum expectation values, the Dirac mass term, $`m_D`$ in the $`\psi `$ basis and Majorana mass term $`M_R`$ for $`\nu _R`$ are given by
$`m_D=\left(\begin{array}{ccc}a& d& b^{}\\ d^{}& c& f\\ b& f^{}& e\end{array}\right),M_R=\left(\begin{array}{ccc}M& 0& 0\\ 0& 0& M^{}\\ 0& M^{}& 0\end{array}\right).`$ (22)
By the see-saw mechanism, the neutrino mass matrix for the left-handed neutrinos in the $`\psi `$ basis is given by
$`\stackrel{~}{m}_\nu =m_D^TM_R^1m_D=\left(\begin{array}{ccc}\frac{a^2}{M}+\frac{2bd^{}}{M^{}}& \frac{ad}{M}+\frac{bc}{M^{}}+\frac{d^{}f^{}}{M^{}}& \frac{ab^{}}{M}+\frac{bf}{M^{}}+\frac{d^{}e}{M^{}}\\ \frac{ad}{M}+\frac{bc}{M^{}}+\frac{d^{}f^{}}{M^{}}& \frac{d^2}{M}+\frac{2cf^{}}{M^{}}& \frac{b^{}d}{M}+\frac{ce}{M^{}}+\frac{ff^{}}{M^{}}\\ \frac{ab^{}}{M}+\frac{bf}{M^{}}+\frac{d^{}e}{M^{}}& \frac{b^{}d}{M}+\frac{ce}{M^{}}+\frac{ff^{}}{M^{}}& \frac{b^2}{M}+\frac{2fe}{M^{}}\end{array}\right)`$ (23)
If we parameterize
$`m_1^0+\stackrel{~}{m}_1=\left({\displaystyle \frac{a^2}{M}}+2{\displaystyle \frac{bd^{}}{M^{}}}\right)\stackrel{~}{m}_1=\left({\displaystyle \frac{b^{}d}{M}}+{\displaystyle \frac{ce}{M^{}}}+{\displaystyle \frac{ff^{}}{M}}\right)`$
$`m_2^0+\stackrel{~}{m}_2=\left({\displaystyle \frac{d^2}{M}}+2{\displaystyle \frac{cf^{}}{M^{}}}\right)\stackrel{~}{m}_2=\left({\displaystyle \frac{ab^{}}{M}}+{\displaystyle \frac{bf}{M^{}}}+{\displaystyle \frac{d^{}e}{M^{}}}\right)`$
$`m_3^0+\stackrel{~}{m}_3=\left({\displaystyle \frac{b^2}{M}}+2{\displaystyle \frac{fe}{M^{}}}\right)\stackrel{~}{m}_3=\left({\displaystyle \frac{ad}{M}}+{\displaystyle \frac{bc}{M^{}}}+{\displaystyle \frac{d^{}f^{}}{M^{}}}\right),`$ (24)
we obtain the neutrino mass matrix $`\stackrel{~}{m}_\nu `$ in the $`\psi `$ basis given in Eq.(12). The ansatz is that all parameters $`a`$, $`b`$, $`c`$, $`d`$, $`e`$, $`f`$, $`b^{}`$, $`d^{}`$, $`f^{}`$, $`M`$ and $`M^{}`$ are real. The Lagrangian in Eqs.(20) and (21) are the most general one to derive the democratic mass matrix, although it contains redundant parameters.
The minimal model is those which contain only two Higgs doublets, say, $`H_1`$ and $`H_2`$, where $`\stackrel{~}{m}_\nu `$ contains six independent parameters to specify $`m_i^0`$ and $`\stackrel{~}{m}_i`$.
## 3 A restricted model -One Higgs case-
In order to construct the more predictive mass matrix, we try to reduce the number of parameters in the democratic-type mass matrix, $`m_i^0`$ and $`\stackrel{~}{m}_i`$. Here, we consider the possibility of one Higgs model.
If we keep only $`H_1=H`$ with $`H=u`$ in the Lagrangian (20), then
$`_D={\displaystyle \frac{1}{u}}\left(a\overline{\nu _e}_RH\mathrm{\Psi }_1+c\overline{\nu _\mu }_RH\mathrm{\Psi }_2+e\overline{\nu _\tau }_RH\mathrm{\Psi }_3\right)+\mathrm{h}.\mathrm{c}..`$ (25)
Then, the Dirac mass term in the $`\psi `$ basis is $`m_D=\mathrm{diag}(a,c,e)`$. After the see-saw mechanism, the left-handed neutrino mass matrix in the $`\psi `$ basis is
$`\stackrel{~}{m}_\nu =m_D^TM_R^1m_D=\left(\begin{array}{ccc}m_1^0+\stackrel{~}{m}_1& 0& 0\\ 0& 0& \stackrel{~}{m}_1\\ 0& \stackrel{~}{m}_1& 0\end{array}\right).`$ (26)
In the flavor eigenstate basis
$`m_\nu =V_T^{}\stackrel{~}{m}_\nu V_T^{}={\displaystyle \frac{m_1^0}{3}}\left(\begin{array}{ccc}1& \omega ^2& \omega \\ \omega ^2& \omega & 1\\ \omega & 1& \omega ^2\end{array}\right)+\stackrel{~}{m}_1\left(\begin{array}{ccc}1& 0& 0\\ 0& \omega & 0\\ 0& 0& \omega ^2\end{array}\right).`$ (27)
From Eq.(26), we see that the neutrino mass matrix $`m_\nu `$ is diagonalized by
$`V_1=V_T\left(\begin{array}{ccc}1& 0& 0\\ 0& \frac{1}{\sqrt{2}}& \frac{1}{\sqrt{2}}\\ 0& \frac{1}{\sqrt{2}}& \frac{1}{\sqrt{2}}\end{array}\right).`$ (28)
This mixing scheme is interesting, although this model is unrealistic because it predicts the degenerate mass, $`m_2=m_3`$.
(a) The model with the $`Z_3`$ symmetry breaking terms
We introduce the $`Z_3`$ symmetry breaking terms by assuming that they respect the $`Z_2`$ symmetry which is defined by
$`\mathrm{\Psi }_1\mathrm{\Psi }_1,\nu _{eR}\nu _{eR},`$ (29)
and all other fields are unchanged. Then, the $`Z_2`$ invariant Lagrangian is given by
$`_{SB}`$ $`=`$ $`{\displaystyle \frac{1}{u}}(a\overline{\nu _e}_\mathrm{R}H\mathrm{\Psi }_1+c\overline{\nu _\mu }_\mathrm{R}H\mathrm{\Psi }_2+e\overline{\nu _\tau }_\mathrm{R}H\mathrm{\Psi }_3)`$ (30)
$`{\displaystyle \frac{1}{u}}(f\overline{\nu _\mu }_\mathrm{R}H\mathrm{\Psi }_3+f^{}\overline{\nu _\tau }_\mathrm{R}H\mathrm{\Psi }_2)`$
together with the Majorana mass term for the right-handed neutrinos given in Eq.(21). This model gives the mass matrix in the $`\psi `$ basis in Eq.(12) with non-zero $`m_1^0`$, $`\stackrel{~}{m}_1`$, $`m_2^0`$ and $`m_3^0`$. In other words, in the flavor eigenstate basis, it is
$`m_\nu `$ $`=`$ $`{\displaystyle \frac{m_1^0}{3}}\left(\begin{array}{ccc}1& \omega ^2& \omega \\ \omega ^2& \omega & 1\\ \omega & 1& \omega ^2\end{array}\right)+\stackrel{~}{m}_1\left(\begin{array}{ccc}1& 0& 0\\ 0& \omega & 0\\ 0& 0& \omega ^2\end{array}\right)`$ (31)
$`+{\displaystyle \frac{m_2^0}{3}}\left(\begin{array}{ccc}1& \omega & \omega ^2\\ \omega & \omega ^2& 1\\ \omega ^2& 1& \omega \end{array}\right)+{\displaystyle \frac{m_3^0}{3}}\left(\begin{array}{ccc}1& 1& 1\\ 1& 1& 1\\ 1& 1& 1\end{array}\right).`$
In the new basis which is obtained from the transformation by $`V_1`$ in Eq.(28), the mass matrix is
$`V_1^Tm_\nu V_1=\left(\begin{array}{ccc}m_1^0+\stackrel{~}{m}_1& 0& 0\\ 0& \stackrel{~}{m}_1+\frac{1}{2}(m_3^0+m_2^0)& \frac{1}{2}(m_3^0m_2^0)\\ 0& \frac{1}{2}(m_3^0m_2^0)& \stackrel{~}{m}_1+\frac{1}{2}(m_3^0+m_2^0)\end{array}\right).`$ (32)
Thus, the mixing matrix which diagonalize $`m_\nu `$ is given by
$`V=\left(\begin{array}{ccc}1& 0& 0\\ 0& \omega & 0\\ 0& 0& \omega ^2\end{array}\right)\left(\begin{array}{ccc}\sqrt{\frac{1}{3}}& \sqrt{\frac{2}{3}}& 0\\ \frac{1}{\sqrt{3}}& \frac{1}{\sqrt{6}}& \frac{1}{\sqrt{2}}\\ \frac{1}{\sqrt{3}}& \frac{1}{\sqrt{6}}& \frac{1}{\sqrt{2}}\end{array}\right)\left(\begin{array}{ccc}1& 0& 0\\ 0& 1& 0\\ 0& 0& i\end{array}\right)\left(\begin{array}{ccc}1& 0& 0\\ 0& c^{}& s^{}\\ 0& s^{}& c^{}\end{array}\right),`$ (33)
where $`c^{}=\mathrm{cos}\theta ^{}`$ and $`s^{}=\mathrm{sin}\theta ^{}`$ and
$`\mathrm{tan}\theta ^{}={\displaystyle \frac{\frac{m_3^0m_2^0}{2}}{\stackrel{~}{m}_1+\sqrt{\stackrel{~}{m}_1^2+\left(\frac{m_3^0m_2^0}{2}\right)^2}}},`$ (34)
and neutrino masses are given by
$`m_1`$ $`=`$ $`m_1^0+\stackrel{~}{m}_1,`$
$`m_2`$ $`=`$ $`{\displaystyle \frac{m_3^0+m_2^0}{2}}+\sqrt{\stackrel{~}{m}_1^2+\left({\displaystyle \frac{m_3^0m_2^0}{2}}\right)^2},`$
$`m_3`$ $`=`$ $`{\displaystyle \frac{m_3^0+m_2^0}{2}}\sqrt{\stackrel{~}{m}_1^2+\left({\displaystyle \frac{m_3^0m_2^0}{2}}\right)^2}.`$ (35)
Here we take the convention, $`\stackrel{~}{m}_1>0`$, and also consider that $`\stackrel{~}{m}_1>|(m_3^0+m_2^0)/2|`$, $`|(m_3^0m_2^0)/2|`$, because $`m_3^0`$ and $`m_2^0`$ are the $`Z_3`$ symmetry breaking parameters. Then, we find $`m_2>0`$ and $`m_3<0`$. The parameter $`\stackrel{~}{m}_1`$ controls the average size of neutrino masses, and $`m_3^0+m_2^0`$ does the mass splitting between $`m_2`$ and $`m_3`$, and the parameter $`m_3^0m_2^0`$ does the size of $`(V_{MNS})_{13}`$.
The MNS mixing matrix is explicitly given by
$`V_{MNS}=\left(\begin{array}{ccc}\frac{1}{\sqrt{3}}& \sqrt{\frac{2}{3}}c^{}& i\sqrt{\frac{2}{3}}s^{}\\ \frac{1}{\sqrt{3}}& \frac{1}{\sqrt{6}}(c^{}+i\sqrt{3}s^{})& \frac{1}{\sqrt{6}}(\sqrt{3}c^{}+is^{})\\ \frac{1}{\sqrt{3}}& \frac{1}{\sqrt{6}}(c^{}i\sqrt{3}s^{})& \frac{1}{\sqrt{6}}(\sqrt{3}c^{}is^{})\end{array}\right)\left(\begin{array}{ccc}1& 0& 0\\ 0& 1& 0\\ 0& 0& i\end{array}\right).`$ (36)
It should be noted that the model predicts $`\delta =\frac{\pi }{2}`$ and the angle $`\theta _{12}`$ which is relevant to the solar neutrino mixing. We have
$`\mathrm{sin}^22\theta _{sol}={\displaystyle \frac{8}{9}}c^2,\mathrm{sin}^22\theta _{atm}=1{\displaystyle \frac{4}{9}}s^4.`$ (37)
If we impose the CHOOZ bound, $`|\sqrt{2/3}s^{}|<0.16`$, we have
$`{\displaystyle \frac{8}{9}}>\mathrm{sin}^22\theta _{sol}>0.87,\mathrm{sin}^22\theta _{atm}>0.999.`$ (38)
Below, we consider the special cases.
(b) Some limiting cases
(b-1) The model with $`m_3^0=m_2^0`$
This model is realized by imposing the invariance under the permutation in addition to $`Z_2`$ symmetry as
$`\nu _{\mu R}\nu _{\tau R},\mathrm{\Psi }_2\mathrm{\Psi }_3.`$ (39)
Then, the parameters in Eq.(30) are restricted as $`f=f^{}`$ and $`c=e`$, and the condition $`m_3^0=m_2^0`$ is realized. From Eq.(34), we have $`\theta ^{}=0`$ and neutrino masses are
$`m_1=m_1^0+\stackrel{~}{m}_1,m_2=m_3^0+\stackrel{~}{m}_1,m_3=m_3^0\stackrel{~}{m}_1.`$ (40)
Thus, the model predicts the mixing matrix defined in Eq.(28). By parameterizing as $`V=\mathrm{diag}(1,\omega ,\omega ^2)V_{MNS}`$, we find the MNS matrix is
$`V_{MNS}=\left(\begin{array}{ccc}\frac{1}{\sqrt{3}}& \sqrt{\frac{2}{3}}& 0\\ \frac{1}{\sqrt{3}}& \frac{1}{\sqrt{6}}& \frac{1}{\sqrt{2}}\\ \frac{1}{\sqrt{3}}& \frac{1}{\sqrt{6}}& \frac{1}{\sqrt{2}}\end{array}\right)\left(\begin{array}{ccc}1& 0& 0\\ 0& 1& 0\\ 0& 0& i\end{array}\right),`$ (41)
where the matrix diag$`(1,1,i)`$ is the Majorana phase matrix.
This model predicts
$`\mathrm{sin}^22\theta _{atm}=1,\mathrm{sin}^22\theta _{sol}={\displaystyle \frac{8}{9}}.`$ (42)
This mixing matrix is in contrast to the so called democratic mixing matrix which predicts $`\mathrm{sin}^22\theta _{atm}=8/9`$ and $`\mathrm{sin}^22\theta _{sol}=1`$.
(b-2) The model with $`m_2^0=0`$ or $`m_3^0=0`$
One way to realize this model is to consider two Higgs model by keeping $`H_1`$ and $`H_2`$ in Eq.(20). Then, we impose $`Z_3\times Z_2`$ symmetry. By the $`Z_3`$ symmetry, we have $`b^{}=d^{}=f^{}=0`$. Then, we impose the $`Z_2`$ symmetry given in Eq.(29) and we find that $`b=d=0`$. These conditions give $`m_2^0=0`$. Similarly, $`m_3^0=0`$ case is realized. Since both cases give the same mixing matrix so that we consider $`m_2^0=0`$ case. Then, the mixing angle $`\theta ^{}`$ is completely fixed by the ratio of neutrino masses, $`m_2>0`$ and $`m_3<0`$. Explicitly, we find with $`V_{MNS}=V_{SF}\mathrm{diag}(1,1,i)`$
$`V_{SF}=\left(\begin{array}{ccc}\frac{1}{\sqrt{3}}& \sqrt{\frac{1}{3}\frac{\beta +2}{\beta }}& \pm i\sqrt{\frac{1}{3}\frac{\beta 2}{\beta }}\\ \frac{1}{2}\sqrt{\frac{\beta +2}{\beta +1}}\left[1i\sqrt{\frac{1}{3}\frac{\beta 2}{\beta +2}}\right]& \frac{1}{2}\sqrt{\frac{\beta }{\beta +1}}\left[1\pm i\sqrt{\frac{1}{3}\frac{(\beta +2)(\beta 2)}{\beta ^2}}\right]& \sqrt{\frac{1}{3}\frac{\beta +1}{\beta }}\\ \frac{1}{2}\sqrt{\frac{\beta +2}{\beta +1}}\left[1\pm i\sqrt{\frac{1}{3}\frac{\beta 2}{\beta +2}}\right]& \frac{1}{2}\sqrt{\frac{\beta }{\beta +1}}\left[1i\sqrt{\frac{1}{3}\frac{(\beta +2)(\beta 2)}{\beta ^2}}\right]& \sqrt{\frac{1}{3}\frac{\beta +1}{\beta }}\end{array}\right)`$ (43)
where
$`\beta =\sqrt{\left|{\displaystyle \frac{m_2}{m_3}}\right|}+\sqrt{\left|{\displaystyle \frac{m_3}{m_2}}\right|}2.`$ (44)
In the limit $`m_2=m_3`$, i.e., $`\beta =2`$, the mixing matrix reduces to the mixing scheme in Eq.(41). Due to the CHOOZ bound, $`\beta `$ must be close to 2.
This model predicts
$`\mathrm{sin}^22\theta _{sol}`$ $`=`$ $`{\displaystyle \frac{4}{9}}{\displaystyle \frac{\beta +2}{\beta }},`$
$`\mathrm{sin}^22\theta _{atm}`$ $`=`$ $`{\displaystyle \frac{4}{9}}{\displaystyle \frac{(\beta +1)(2\beta 1)}{\beta ^2}}.`$ (45)
If we impose the CHOOZ bound
$`|(V_{MNS})_{13}|=\sqrt{{\displaystyle \frac{1}{3}}{\displaystyle \frac{\beta 2}{\beta }}}<0.16,`$ (46)
$`\beta `$ is restricted by $`2<\beta <2.17`$ which means
$`0.85<\mathrm{sin}^22\theta _{sol}<{\displaystyle \frac{8}{9}},\mathrm{\hspace{0.33em}\hspace{0.33em}0.999}<\mathrm{sin}^22\theta _{atm}<1,`$ (47)
$`0.44<\left|{\displaystyle \frac{m_2}{m_3}}\right|<2.27.`$ (48)
The effective mass for the neutrinoless double beta decay is given by
$`|<m_\nu >|={\displaystyle \frac{1}{3}}\left|m_1+{\displaystyle \frac{\beta +2}{\beta }}m_2+{\displaystyle \frac{\beta 2}{\beta }}m_3\right|{\displaystyle \frac{1}{3}}|m_1+2m_2|.`$ (49)
If we take the constraint $`|<m_\nu >|<0.3`$ eV, then we find $`|<m_\nu >||m_2|<0.3`$ eV for $`m_1m_2>0`$ and $`3|<m_\nu >||m_2|<0.9`$ eV for $`m_1m_2<0`$.
In order to realize the large CP violation, it is required that the value of $`|(V_{MNS})_{13}|=\sqrt{(\beta 2)/3\beta }`$ should be as large as its bound by the CHOOZ data. This leads to the constraint on the ratio $`|m_2/m_3|`$. Explicitly, we have
$`|m_2|=h\sqrt{{\displaystyle \frac{\mathrm{\Delta }_{atm}^2}{|h^21|}}},|m_3|=\sqrt{{\displaystyle \frac{\mathrm{\Delta }_{atm}^2}{|h^21|}}},`$ (50)
where
$`h=\left\{{\displaystyle \frac{1}{2}}\left(\beta \pm \sqrt{\beta ^24}\right)\right\}^2.`$ (51)
If we take $`\mathrm{\Delta }_{atm}^2=3.5\times 10^3\mathrm{eV}^2`$ and the maximal value $`|(V_{MNS})_{13}|=0.16`$, we find $`(|m_2|,|m_3|)=(6.6,2.9)\times 10^2\mathrm{eV}`$ or $`\mathrm{or}(2.9,6.6)\times 10^2\mathrm{eV}`$. When these masses are larger than the above values, the CP violation becomes smaller.
The parameter $`\stackrel{~}{m}_1`$ is used to fix the overall normalization of neutrino masses and $`m_3^0`$ works to adjust the ratio $`|m_2/m_3|`$. The remaining parameter $`m_1^0`$ determines the mass $`m_1`$, which is related to the neutrino squared mass difference for the solar neutrino mixing, $`\mathrm{\Delta }_{sol}^2|m_2^2m_1^2|`$. Since $`\mathrm{\Delta }_{sol}^2`$ is much smaller than $`\mathrm{\Delta }_{atm}^2`$, the fine tuning is required to obtain the approximate degeneracy between $`m_1`$ and $`m_2`$. The parameter $`m_1^0`$ adjusts the splitting between $`m_1`$ and $`m_2`$.
## 4 Summary
In our previous paper, we proposed the democratic-type neutrino mass matrix and showed that this model predicts the quite interesting relations which are crucial to give the large atmospheric neutrino mixing and the large CP violation which are given in Eq.(9).
In this paper, we explored the mass matrices which inherit the attractive features we found, but give more predictions. In order to examine the origin of the democratic-type mass matrix further, we considered the Yukawa interaction and the Majorana mass matrix for the right-handed neutrinos and derive the neutrino mass matrix for the left-handed neutrinos.
We gave the most general $`Z_3`$ invariant Lagrangian with three Higgs doublets and showed that this Lagrangian gives the democratic-type mass matrix. Then, we restricted the number of the Higgs doublet to be one. By introducing the $`Z_3`$ symmetry breaking terms but keeping the $`Z_2`$ symmetry, we found the interesting model that predicts the large solar neutrino mixing also. The most prominent feature of the present model is the prediction that $`\mathrm{sin}^22\theta _{atm}`$ is very close to unity and also the large CP violation (the CP violation phase in the standard form $`\delta =\pi /2`$). These two features will be tested as unambiguous signals of the model.
We expect that our mass matrix is obtained at the right-handed mass $`M_R`$ scale. We did not discuss the effect of the renormalization in this paper. If $`m_1m_2<0`$, our predictions are stable, while the substantial dependence is expected for $`\mathrm{sin}^22\theta _{sol}`$ for $`m_1m_2>0`$ because of the degeneracy. The possibility that the large $`\mathrm{sin}^22\theta _{sol}`$ at $`M_R`$ reduces to the small value which explains the small mixing angle MSW solution is under estimation and will be published elsewhere.
Acknowledgment This work is supported in part by the Japanese Grant-in-Aid for Scientific Research of Ministry of Education, Science, Sports and Culture, No.11127209.
Appendix A: Derivation of Eq.(8)
Let us consider
$`V`$ $``$ $`V_TO`$
$`=`$ $`{\displaystyle \frac{1}{\sqrt{3}}}\left(\begin{array}{ccc}O_{11}+O_{21}+O_{31}& O_{12}+O_{22}+O_{32}& O_{13}+O_{23}+O_{33}\\ \omega O_{11}+\omega ^2O_{21}+O_{31}& \omega O_{12}+\omega ^2O_{22}+O_{32}& \omega O_{13}+\omega ^2O_{23}+O_{33}\\ \omega ^2O_{11}+\omega O_{21}+O_{31}& \omega ^2O_{12}+\omega O_{22}+O_{32}& \omega ^2O_{13}+\omega O_{23}+O_{33}\end{array}\right),`$
where $`V_T`$ is the tri-maximal mixing matrix in Eq.(6) and $`O`$ is the orthogonal matrix. Then, we find the constraints
$`V_{2j}=V_{3j}^{},(j=1,2,3)`$ (A.2)
because $`O_{ij}`$ are real parameters. Now we define the standard form of the mixing matrix as
$`V_{SF}=\left(\begin{array}{ccc}c_{12}c_{13}& s_{12}c_{13}& s_{13}e^{i\delta }\\ s_{12}c_{23}c_{12}s_{23}s_{13}e^{i\delta }& c_{12}c_{23}s_{12}s_{23}s_{13}e^{i\delta }& s_{23}c_{13}\\ s_{12}s_{23}c_{12}c_{23}s_{13}e^{i\delta }& c_{12}s_{23}s_{12}c_{23}s_{13}e^{i\delta }& c_{23}c_{13}\end{array}\right),`$ (A.3)
where $`s_{ij}=\mathrm{sin}\theta _{ij}`$ and $`c_{ij}=\mathrm{cos}\theta _{ij}`$. Since $`V`$ is related to $`V_{SF}`$ by the phase transformation as $`V=PV_{SF}P^{}`$, the constraints are given by $`|(V_{SF})_{2j}|=|(V_{SF})_{3j}|`$. From the constraint with $`j=3`$, we obtain $`|s_{23}|=|c_{23}|`$. From $`j=1`$ and $`j=2`$, together with $`|s_{23}|=|c_{23}|`$, we find $`\mathrm{cos}\delta =0`$.
|
warning/0003/hep-ph0003178.html
|
ar5iv
|
text
|
# 1. Introduction
## 1. Introduction
The discovery of charged Higgs bosons will provide a concrete evidence of the multi-doublet structure of the Higgs sector. Recent efforts have focused on their relevance to Supersymmetry (SUSY), in particular in the MSSM, which incorporates exactly two Higgs doublets, yielding – after spontaneous EW symmetry breaking – five physical Higgs states: the neutral pseudoscalar ($`A`$), the lightest ($`h`$) and heaviest ($`H`$) neutral scalars and two charged ones ($`H^\pm `$).
In much of the parameter space preferred by SUSY, namely $`M_{H^\pm }M_{W^\pm }`$ and $`1<\mathrm{tan}\beta <m_t/m_b`$ , the LHC will provide the greatest opportunity for the discovery of $`H^\pm `$ particles. In fact, over the above $`\mathrm{tan}\beta `$ region, the Tevatron (Run 2) discovery potential is limited to charged Higgs masses smaller than $`m_t`$ .
However, at the LHC, whereas the detection of light charged Higgs bosons (with $`M_{H^\pm }<m_t`$) is rather straightforward in the decay channel $`tbH^+`$ for most $`\mathrm{tan}\beta `$ values, thanks to the huge top-antitop production rate, the search is notoriously difficult for heavy masses (when $`M_{H^\pm }>m_t`$), because of the large reducible and irreducible backgrounds associated with the main decay mode $`H^{}b\overline{t}`$, following the dominant production channel $`bgtH^{}`$ . (Notice that the rate of the latter exceeds by far other possible production modes , this rendering it the only viable channel at the CERN machine in the heavy mass region.)
The analysis of the $`H^{}b\overline{t}`$ signature has been the subject of many debates , whose conclusion is that the LHC discovery potential is satisfactory, but only provided that $`\mathrm{tan}\beta `$ is small ($`\stackrel{<}{}1.5`$) or large ($`\stackrel{>}{}30`$) enough and the charged Higgs boson mass is below 600 GeV or so.
A recent analysis has shown that the $`\tau \nu `$ decay mode, indeed dominant for light charged Higgs states and exploitable below the top threshold for any accessible $`\mathrm{tan}\beta `$ , can be used at the LHC even in the large $`M_{H^\pm }`$ case, in order to discover $`H^\pm `$ scalars in the parameter range $`\mathrm{tan}\beta \stackrel{>}{}3`$ and 200 GeV $`<M_{H^\pm }<1`$ TeV. Besides, if the distinctive $`\tau `$ polarisation is used in this channel, the latter can provide at least as good a heavy $`H^\pm `$ signature as the $`H^{}b\overline{t}`$ decay mode (for the large $`\mathrm{tan}\beta `$ regime ).
At present then, it is the $`\mathrm{tan}\beta \stackrel{<}{}3`$ region of the MSSM which ought to be explored through other decay modes, especially those where direct mass reconstruction is possible. The most obvious of these is the $`H^\pm W^{\pm ()}h`$ channel (see also ), proceeding via the production of a charged gauge boson and the lightest Higgs scalar of the MSSM, with the former on- or off-shell depending on the relative values of $`M_{H^\pm }`$ and $`M_h`$. In fact, its branching ratio (BR) can be rather large, competing with the bottom-top decay mode and overwhelming the tau-neutrino one for $`M_{H^\pm }\stackrel{>}{}m_t`$ at low $`\mathrm{tan}\beta `$: see Figs. 12. Besides, under the assumption that the $`h`$ scalar has previously been discovered (which we embrace here), its kinematics is rather constrained, around two resonant decay modes, $`W^\pm `$ 2 jets (or lepton-neutrino) and $`hb\overline{b}`$, an aspect which allows for a significant reduction of the QCD background. As demonstrated in Ref. , signals of charged Higgs bosons in the $`2\stackrel{<}{}\mathrm{tan}\beta \stackrel{<}{}3`$ range can be seen in this channel, provided that 200 GeV $`\stackrel{<}{}M_{H^\pm }\stackrel{<}{}220`$ GeV (see also for an experimental simulation). The above lower limit on $`\mathrm{tan}\beta `$ corresponds to the border of the exclusion region drawn from LEP2 direct searches for the MSSM $`h`$ scalar, whose mass bound is now set at $`M_h\stackrel{>}{}100`$ GeV or so .
It is the purpose of this letter that of resuming the studies of Ref. , by analysing the contribution to the background due to several irreducible processes, not considered there, whose presence could spoil the feasibility of charged Higgs searches in the $`W^{\pm ()}h`$ mode of the MSSM.
The plan of this paper is as follows. In the next Section we discuss possible signals and backgrounds, their implementation and list the values adopted for the various parameters needed for their computation. Section 3 is devoted to the presentation and discussion of the results. Conclusions are in Section 4.
## 2. Signals and backgrounds
We generate the signal cross sections by using the formulae of Ref. . That is, we implement the $`23`$ matrix element (ME) for the process
$$ggt\overline{b}H^{}+\mathrm{charge}\mathrm{conjugate}(\mathrm{c}.\mathrm{c}.).$$
(1)
This nicely embeds both the $`ggt\overline{t}t\overline{b}H^{}+\mathrm{c}.\mathrm{c}.`$ subprocess of top-antitop production and decay, which is dominant for $`m_t\stackrel{>}{}M_{H^\pm }`$, as well as the $`bgtH^{}`$ \+ c.c. one of $`b\overline{t}`$-fusion and $`H^\pm `$-bremsstrahlung, which is responsible for charged Higgs production in the case $`m_t\stackrel{<}{}M_{H^\pm }`$ . The ME of process (1) has been computed by means of the spinor techniques of Refs. .
In the $`H^{}W^{()}hW^{()}b\overline{b}`$ channel, assuming high efficiency and purity in selecting/rejecting $`b`$-/non-$`b`$-jets, possible irreducible background processes are the following (we consider only the $`gg`$-initiated channels):
1. the $`t\overline{b}W^{}h`$ continuum;
2. $`t\overline{b}W^{}Z`$ production, especially when $`M_ZM_h`$;
3. the QCD induced case $`t\overline{b}W^{}g`$;
4. and, finally, $`t\overline{b}W^{}H`$ and $`t\overline{b}W^{}A`$ intermediate states;
in which $`H,h,A,Z,gb\overline{b}`$, plus their c.c. channels. Once the top quark appearing in the above reactions decays, two $`W^\pm `$ bosons are present in each event. We will eventually assume the $`W^+W^{}`$ pair to decay semi-leptonically to light-quark jets, electrons/muons and corresponding neutrinos. Furthermore, we will require to tag exactly three $`b`$-jets in the final state (e.g., by using $`\mu `$-vertex or high $`p_T`$ lepton techniques). The same ‘signature’ was considered in Ref. , where only the ‘intrinsic’ $`t\overline{b}H^{}t\overline{t}b\overline{b}`$ background and the QCD noise due to ‘$`t\overline{t}`$ \+ jet’ events were studied (with jet signifying here either a $`b`$-, light-quark or gluon jet, the latter two mistagged for the former).
Both signal and background MEs have been integrated numerically by means of VEGAS and, for test purposes, of RAMBO and Metropolis as well. While proceeding to the phase space integration, one also has to fold in the $`(x,Q^2)`$-dependent Parton Distribution Functions (PDFs) for the two incoming gluons. These have been evaluated at leading-order, by means of the package MRS-LO(05A) .
The numerical values of the SM parameters are ($`\mathrm{}=e,\mu `$):
$$m_{\mathrm{}}=m_\nu _{\mathrm{}}=m_u=m_d=m_s=m_c=0,$$
$$m_b=4.25\mathrm{GeV},m_t=175\mathrm{GeV},$$
$$M_Z=91.2\mathrm{GeV},\mathrm{\Gamma }_Z=2.5\mathrm{GeV},$$
$$M_{W^\pm }=80.2\mathrm{GeV},\mathrm{\Gamma }_W=2.2\mathrm{GeV}.$$
(2)
As for the top width $`\mathrm{\Gamma }_t`$, we have used the LO value calculated within the MSSM (i.e., $`\mathrm{\Gamma }_t=1.55`$ GeV if $`M_{H^\pm }m_t`$).
Concerning the MSSM parameters, we proceed as follows. For a start, we assume that the mass of the lightest neutral Higgs particle (but not $`\mathrm{tan}\beta `$) is already known, thanks to its discovery at either LEP2, Tevatron (Run 2) or from early analyses at the LHC itself. Thus, for us, $`M_h`$ is a fixed parameter, assuming for reference the following discrete values: e.g., 90, 100, 110, 120 and 130 GeV. Then we express all other Higgs masses as a function of $`M_{H^\pm }`$ and $`\mathrm{tan}\beta `$. For the pseudoscalar Higgs boson mass, the tree-level relation $`M_{H^\pm }^2=M_{W^\pm }^2+M_A^2`$ is assumed. Radiative corrections then, of arbitrary perturbative order, are in practice embedded in the $`H`$ mass and the mixing angle $`\alpha `$. In general, notice that, at the ‘Renormalisation Group improved’ one-loop level , it is only for very large values of the lightest stop mass and of the squark mixing parameters that $`M_h`$ can escape the LEP2 bound in the low $`\mathrm{tan}\beta `$ region, on which we will focus most of our attention.
Finally, notice that we develop our discussion at the parton level, without considering fragmentation and hadronisation effects. Thus, jets are identified with the partons from which they originate and all cuts are applied directly to the latter. In particular, when selecting $`b`$-jets, a vertex tagging is implied, with a finite efficiency, $`ϵ_b`$, per each tag. Moreover, we assume no correlations among multiple tags, nor do we include misidentification of light-quark (including $`c`$-quark-)jets produced in $`W^\pm `$ decays as $`b`$-jets.
## 3. Results and discussion
As a preliminary exercise, we study the total production and decay cross sections before any cuts, as all our reactions are finite over their entire phase spaces (recall that $`m_b0`$). This is done in Figs. 34 for the signal and the five background processes discussed in the previous Section, for five values of $`\mathrm{tan}\beta `$, over the range 140 GeV $`\stackrel{<}{}M_{H^\pm }\stackrel{<}{}`$ 500 GeV, for $`M_h=90`$ and 100 GeV, in the channel $`Xb\overline{b}`$, where $`X=h,Z,g,H`$ or $`A`$. (Of course, the $`W^\pm Z`$ and $`W^\pm g`$ backgrounds have no dependence on any of the three parameters above<sup>2</sup><sup>2</sup>2Note that the rates in Figs. 34 account for the c.c. production modes as well..) As for the decay rates of the top (anti)quark and the $`W^\pm `$ boson, for sake of simplicity, we take them equal to $`1`$ for the time being. The signal is always dominated by the QCD background and – at large $`\mathrm{tan}\beta `$ – also by the EW ones. Notice the local maxima of the signal rates at $`M_{H^\pm }M_{W^\pm }+M_h`$, as induced by the opening of the $`H^{}W^{}h`$ decay (compare to Figs. 12), and the minima as well, due to the onset of the $`H^{}b\overline{t}`$ channel instead.
In the reminder of our analysis, we assume semi-leptonic decays of $`W^+W^{}`$ pairs, as in Ref. : i.e., $`W^+W^{}2\mathrm{jets}\mathrm{}^\pm \nu _{\mathrm{}}`$ (hereafter, jet refers to a non-$`b`$-jet and $`\mathrm{}=e,\mu `$). However, as compared to that analysis, we make one simplification. Namely, we assume that one top (anti)quark and the $`W^\pm `$ boson generated in its decay have already been reconstructed, e.g., by using the mass selection procedure advocated in Ref. , either leptonically or hadronically. This allows us to greatly reduce the complexity of our numerical calculation while – we believe – substantially un-affecting the relative rates of signal and backgrounds (in fact, all processes described produce the same final state and all involve at least one top quark). Then we apply the following cuts on the remaining particles (here, the label $`\mathrm{j}`$ refers to the decay products of the second $`W^\pm `$ boson present in the event, which can be either light-quarks or leptons):
$$p_T(b,\mathrm{j},\mathrm{missing})>20\mathrm{GeV}$$
(3)
on the transverse momentum (including the missing one),
$$|\eta (b,\mathrm{j})|<2.5$$
(4)
on the pseudorapidity, and
$$\mathrm{\Delta }R(bb,b\mathrm{j},\mathrm{jj})>0.4$$
(5)
on the relative separation of $`b`$\- and light-quark jets/leptons j, where
$$\mathrm{\Delta }R(ij)=\sqrt{\mathrm{\Delta }\eta (ij)^2+\mathrm{\Delta }\varphi (ij)^2},$$
(6)
is defined in terms of relative differences in pseudorapidity $`\eta (ij)`$ and azimuth $`\varphi (ij)`$, with $`ij=b`$, j$`/\mathrm{}`$. Furthermore, we impose (see also Ref. )
$$|M_{bb}M_h|<10\mathrm{GeV}$$
(7)
on exactly one pair of $`b`$-jets,
$$M_{\mathrm{jj}}>50\mathrm{GeV}$$
(8)
on the light-jet (or lepton-neutrino) pair (recall that the $`W^\pm `$ can be off-shell), and, finally,
$$|M_{bbb\mathrm{jj}}M_t|<20\mathrm{GeV}$$
(9)
around the top mass if three $`b`$’s are present in the event (in addition to the one already used to reconstruct the top (anti)quark). In such a case, one may assume that the charged Higgs boson has predominantly been produced in the decay of a top (anti)quark (when $`M_{H^\pm }\stackrel{<}{}m_t`$). If instead only two appear, then one should conclude that the Higgs has mainly been generated in a bremsstrahlung/fusion process (because $`M_{H^\pm }\stackrel{>}{}m_t`$) with a $`b`$-(anti)quark lost along the beam pipe. Our $`23`$ production mechanism naturally allows one to emulate both dynamics in a gauge invariant fashion, including all interference effects. As already mentioned, however, we will assume a triple $`b`$-tagging, this implying an overall efficiency factor of $`ϵ_b^3`$ multiplying our signal and background rates. (Thus, the third $`b`$-jet in eq. (9) is actually non-$`b`$-tagged: it can be interpreted as the jet system satisfying neither eq. (7) nor eq.(8).) We take $`ϵ_b=0.5`$, like in (and assume 100% lepton identification efficiency).
Given the signal production rates before acceptance and selection cuts, it is clear that – for such an $`ϵ_b`$ – even at high collider luminosity (i.e., $`𝑑t=100`$ fb<sup>-1</sup> per annum), hopes of disentangling the charged Higgs boson of the MSSM in the $`W^{\pm ()}h`$ decay channel are only confined to the very low $`\mathrm{tan}\beta `$ region. We will thus restrict ourselves to study in the reminder of the paper $`\mathrm{tan}\beta `$ values which are, e.g., below seven. The total signal rates after the cuts (3)–(9) have been applied can be found in Fig. 5, for the choices $`\mathrm{tan}\beta =1,2,3`$ and $`7`$, as a function of $`M_{H^\pm }`$. For reference, we illustrate the ‘borderline’ case $`M_h=100`$ GeV. (Indeed, a lower $`M_h`$ value at $`\mathrm{tan}\beta =2`$ is in contradiction with LEP2 data, whereas higher masses induce a far too large suppression on BR($`H^\pm W^\pm h`$): see Fig. 2.) The trends in the figure are the consequence of two effects. On the one hand, the production cross section of $`ggt\overline{b}H^{}`$ \+ c.c. is roughly proportional to $`(m_t^2\mathrm{cot}\beta ^2+m_b^2\mathrm{tan}\beta ^2)`$, so that its maxima occur at very low or very high $`\mathrm{tan}\beta `$. On the other hand, we have seen how the largest $`H^{}W^{()}h`$ decay fraction is attained for $`\mathrm{tan}\beta 2`$. In the end, the largest values for $`\sigma (ggt\overline{b}H^{})\times \mathrm{BR}(H^{}W^{()}h)`$ \+ c.c. are obtained for $`\mathrm{tan}\beta =1`$: see Fig. 5. Unfortunately, such a $`\mathrm{tan}\beta `$ value is already excluded in the MSSM from LEP2 data . For the optimal remaining choice, i.e., $`\mathrm{tan}\beta =2`$, the annual rate never exceeds 140 events (before any $`b`$-tagging efficiency but after acceptance and selection cuts). The maximum occurs at $`M_{H^\pm }200`$ GeV, significantly above the real threshold at $`M_h+M_{W^\pm }180`$ GeV.
We now compare such a signal with the irreducible backgrounds 1.–4., for the same choice of $`\mathrm{tan}\beta `$ and $`M_h`$ (where relevant). This is done in the upper half of Fig. 6, at the level of total production rates. After the cuts (3)–(9) are enforced, all background components in 2.–4. are overwhelmed by the signal in the vicinity of $`M_{H^\pm }=200`$ GeV, whereas the $`W^\pm h`$ continuum production is always larger than the $`H^\pm W^\pm h`$ resonant channel. Thus, it is relevant to compare the last two processes in the ‘reconstructed’ invariant mass $`M_{W^\pm h}`$, i.e., that obtained from pairing the two $`b`$-jets fulfilling condition (7) and the two light-quark jets (or, alternatively, the lepton-neutrino pair) satisfying eq. (8) and not already reconstructing $`M_{W^\pm }`$ on their own and $`m_t`$ in association with any of the $`b`$’s (see Ref. ). The spectrum in this variable is presented in the lower half of Fig. 6, for our ideal case $`M_{H^\pm }=200`$ GeV (and, again, $`\mathrm{tan}\beta =2`$ and $`M_h=100`$ GeV). For such MSSM parameter combination, the charged Higgs signal is well above the continuum for values of $`M_{W^\pm h}`$ which are $`\pm 20`$ GeV from $`M_{H^\pm }`$. (To vary $`M_{H^\pm }`$ and/or $`\mathrm{tan}\beta `$ basically corresponds to rescale the solid line in the last plot by a constant factor, according to the rates in Fig. 5.)
For reference, Tab. 1 presents the number of events of resonant and continuum $`W^\pm h`$ production at the LHC, after 300 inverse femtobarns of collected luminosity, for $`ϵ_b^3=0.125`$, in the window $`|M_{H^\pm }M_{W^\pm h}|<`$ 40 GeV, for the three values $`M_{H^\pm }=180,200`$ and 220 GeV. Although very small, a $`H^{}W^{}h`$ signal is generally observable above the $`W^{}h`$ continuum for $`M_{H^\pm }`$ around $`200`$ GeV. Our numbers are roughly consistent with those in Ref. , if one considers that we neglect the finite efficiency of reconstructing one $`W^\pm `$ boson and the associated top (anti)quark and since we have chosen somewhat different cuts. Therefore, in the end, the dominant backgrounds remain (in the $`3b`$-tagged channel) the $`H^{}b\overline{t}`$ \+ c.c. decay and the QCD noise involving misidentified gluons, i.e., those already identified in Ref. .
## 4. Conclusions
In summary, in this paper, we have complemented a previous analysis of the production and decay of charged Higgs bosons of the MSSM at the LHC, in the channels $`ggt\overline{b}H^{}`$ and $`H^{}W^{()}h`$ (and charged conjugated modes), respectively, by considering several irreducible backgrounds in the $`3b`$-tagged channel, i.e., $`t\overline{b}H^{}`$ \+ c.c. $`3b2\mathrm{jets}\mathrm{}`$ \+ ‘missing energy’ (where the initial $`b`$-(anti)quark is usually lost along the beam pipe), which had not yet been considered.
We have found that, after standard acceptance cuts and a kinematic selection along the lines of the one outlined in Ref. , the dominant background among those considered here is the continuum production $`ggt\overline{b}W^{()}h`$ \+ c.c. However, the latter has been found to lie significantly below the signal in the only region where this is detectable: when $`\mathrm{tan}\beta 23`$ and $`M_{H^\pm }200`$ GeV (with $`M_h`$ around 100 GeV, close to the latest LEP2 constraints). Thus, the chances of detecting the $`H^{}W^{}hW^{}b\overline{b}`$ decay in such a (narrow) region of the MSSM parameter space depend mainly on the interplay between this mode, the competing one $`H^{}b\overline{t}W^{}b\overline{b}`$ and the QCD background with mistagged gluons, as are the latter two that clearly overwhelm the former (recall the last figure in ).
We have carried out our analysis at parton level, without showering and hadronisation effects but emulating typical detector smearing. We are confident that its salient features should survive a more sophisticated simulation, such as the one presented in Ref. . Besides, our results concerning the backgrounds can be transposed to the case of non-minimal SUSY models (where the $`H^\pm `$ discovery potential can extend to a much larger portion of parameter space), such as those considered in Ref. , so that also in these scenarios the irreducible backgrounds analysed here can be brought under control.
### Acknowledgements
The author is grateful to the UK-PPARC for financial support. Furthermore, he thanks D.P. Roy for his remarks, which induced him to eventually considering the subject of this research. He also thanks K.A. Assamagan for several useful discussions. Finally, many conversations with K. Odagiri are acknowledged, as well as many numerical comparisons against and the use of some of his programs.
|
warning/0003/hep-ph0003174.html
|
ar5iv
|
text
|
# Are Messages of R-parity Violating Supersymmetry Hidden within Top Quark Signals ?
## Abstract
In an R-parity nonconserving supersymmetric theory, the lighter stop can dominantly decay into $`b\mu `$ and $`b\tau `$ if R-parity breaking has to explain the neutrino mass and mixing pattern suggested by the data on atmospheric muon neutrinos. This should give rise to $`dilepton+dijet`$ and $`singlelepton+jets`$, signals identical with those of the top quark at the Fermilab Tevatron. One can thus constrain the stop parameter space using the current top search data, and similarly look for the first signals of supersymmetry at the upgraded runs of the Tevatron.
preprint: MRI-PHY/P20000307
It has been repeatedly suggested that in a supersymmetric (SUSY) scenario , one of the two spin-zero superpartners of the top quark, popularly known as stop, could be considerably lighter than any other strongly interacting superparticle . This happens due to the mixing between the left-and right chiral stops, which can be quite large in principle, reducing the mass of the lighter eigenstate. Also, negative contribution from Yukawa coupling in the evolution equations for scalar mass parameters can cause the stop to be lighter than the other squarks in a framework where all squark flavours have the same mass at a high energy scale. It has been shown that in the minimal supersymmetric standard model (MSSM), a light stop ($`\stackrel{~}{t}_1`$) will decay dominantly through the channels $`\stackrel{~}{t}_1c\chi _1^0`$ and $`\stackrel{~}{t}_1b\chi _1^+`$, where $`\chi _1^0`$ and $`\chi _1^+`$ are the lightest neutralino and the lighter chargino respectively. Out of these, the former reigns supreme if the light stop is even lighter than $`\chi _1^+`$ . Based on the ensuing signals, experiments at the Fermilab Tevatron have set a lower limit of about $`120135`$ GeV on the lighter stop depending on the mass of the lightest neutralino.
The observed signals can be quite different when R-parity, defined as $`(1)^{3B+L+2S}`$, is not a conserved quantity, something that is both theoretically and phenomenologically consistent in supersymmetric theories so long as only one of baryon number ($`B`$) and lepton number ($`L`$) is violated. In such cases, the lightest supersymmetric particle (LSP) can be unstable, thereby altering the signals of SUSY . If, in addition, the LSP (which is the lightest neutralino in most theories) is massive enough to evade searches at the large electron-positron collider (LEP), then the observation of SUSY at the Fermilab Tevatron may depend squarely on the production and decays of the light stop for whom R-parity violation opens up new and often dominant decay modes. The limits based on $`\stackrel{~}{t}_1c\chi _1^0`$ require re-examination in such a case.
In this letter, we want to point out that the easiest identifiable signatures of a light stop (and therefore perhaps of SUSY itself) in an R-parity nonconserving framework could be in final states that have been, and still are, widely studied to look for the top quark at the Tevatron, namely, the $`dimuon+dijet`$ as well as $`singlemuon+jets`$ signals. We further argue that stop decays leading to such final states are imperative in R-parity violating scenarios that lead to neutrino masses and mixing as required by the recent SuperKamiokande (SK) data on atmospheric muon neutrinos. Thus a careful reanalysis of the already existing Tevatron Run I results on top search may allow us to explore a large region of the light stop parameter space of such theories, where the search channel employed for MSSM is not going to be effective. In Run II one can have an even wider prospect, not only for constraining the SUSY parameter space but also for the possibility of top signals being actually faked by the stop. In addition, some new signals for such models, very similar to those of the top quark, are suggested here.
Expressed in terms of the quark, lepton and Higgs superfields, the MSSM superpotential is
$`W_{MSSM}`$ $`=`$ $`\mu \widehat{H}_1\widehat{H}_2+h_{ij}^l\widehat{L}_i\widehat{H}_1\widehat{E}_j^c+h_{ij}^d\widehat{Q}_i\widehat{H}_1\widehat{D}_j^c+h_{ij}^u\widehat{Q}_i\widehat{H}_2\widehat{U}_j^c`$ (1)
where $`\mu `$ is the Higgsino mass parameter and the last three terms give all the Yukawa interactions.
The possible additions to this superpotential due to R-parity violation (through lepton number violation only) are given by
$$W_{\mathit{}}=ϵ_i\widehat{L}_i\widehat{H}_2+\lambda _{ijk}\widehat{L}_i\widehat{L}_j\widehat{E}_k^c+\lambda _{ijk}^{}\widehat{L}_i\widehat{Q}_j\widehat{D}_k^c$$
(2)
We consider only the effects of the trilinear additional terms in the superpotential; we shall point out at the end of the paper that the effects we are suggesting can arise also from the bilinear terms $`ϵ_i\widehat{L}_i\widehat{H}_2`$.
If the SK data on atmospheric muon neutrinos (and also similar data from the Soudan-II and MACRO experiments) have to be explained in terms of $`\nu _\mu \nu _\tau `$ oscillations, then the mass-squared splitting between the second and third lightest physical neutrino states will have to be $`\mathrm{\Delta }m_{23}^25\times 10^3`$, in addition to near-maximal mixing between the corresponding flavour eigenstates. R-parity violation in the form of the $`\lambda `$-and $`\lambda ^{^{}}`$-type interactions in equation 2 can give rise to neutrino mass terms at one-loop level, the generic expression for them (in the flavour basis) being
$`(m_\nu ^{\mathrm{loop}})_{ij}`$ $``$ $`{\displaystyle \frac{3}{8\pi ^2}}m_k^dm_p^dM_{SUSY}{\displaystyle \frac{1}{m_{\stackrel{~}{q}}^2}}\lambda _{ikp}^{}\lambda _{jpk}^{}`$ (4)
$`+{\displaystyle \frac{1}{8\pi ^2}}m_k^lm_p^lM_{SUSY}{\displaystyle \frac{1}{m_{\stackrel{~}{l}}^2}}\lambda _{ikp}\lambda _{jpk}`$
where $`m^{d(l)}`$ denote the down-type quark (charged lepton) masses. $`m_{\stackrel{~}{l}}^2`$, $`m_{\stackrel{~}{q}}^2`$ are the average slepton and squark mass squared. $`M_{SUSY}(\mu )`$ is the effective scale of supersymmetry breaking. The mass and mixing patterns suggested by the observed $`\nu _\mu `$ data can be accommodated in the above scheme of mass generation if, with $`M_{SUSY}m\stackrel{~}{q}300500`$ GeV, $`\lambda _{233}^{^{}}\lambda _{333}^{^{}}afewtimes10^4`$ . An immediate consequence of such couplings is the decay of the lighter stop in the channels $`\stackrel{~}{t}_1b\tau ^+`$ and $`\stackrel{~}{t}_1b\mu ^+`$ with comparable widths.
There is a considerably large region of the parameter space, not yet constrained by any experimental data, where the stop is the second lightest supersymmetric particle next to the lightest neutralino. In this region, the three lowest-order decay channels available to the stop are $`\stackrel{~}{t}_1c\chi _1^0`$, $`\stackrel{~}{t}_1b\tau ^+`$ and $`\stackrel{~}{t}_1b\mu ^+`$ . The first one, a well-studied process, is a consequence of neutral flavour violation in SUSY and is suppressed by the small mismatch between quark and squark mass matrices. The latter ones are driven by $`\lambda _{233}^{}`$ and $`\lambda _{333}^{}`$, and we find that they dominate for a light stop with mass $`150`$ GeV so long as $`\lambda _{233}^{^{}},\lambda _{333}^{^{}}`$ lie in the range specified above, in conformity with the SK data.
We focus our attention on single-muon and dimuon final states, together with jets and missing energy, arising from (light) stops pair-produced in $`p\overline{p}`$ collisions at the Fermilab Tevatron, with either both decaying into $`b\tau `$ or one of them into $`b\tau `$ while the other goes to $`b\mu `$. The same final states have been extensively analysed for the determination of the mass and production cross-section of the top quark. It should be noted that the $`\tau `$ produced together with a $`b`$ quark in the two-body decay of a stop (with mass $`100150`$ GeV) can have sufficient $`p_T`$ for the resulting jets/muons/neutrinos to often pass the $`p_T`$ or $`\mathit{}_T`$ cuts associated with the top quark signals. Thus, depending upon whether the tau decays hadronically or semileptonically, the stop decays can contribute to top-like signals of the form $`(i)singlemuon+3jets+\mathit{}_T`$ and $`(ii)dimuon+2jets+\mathit{}_T`$, provided that the jets/leptons satisfy the requisite cuts.
We present some numerical estimates where the five degenerate squark flavours are assumed to have masses $`400`$ GeV. The mass of the lighter stop is made to vary beween $`80`$ and $`150`$ GeV, while the $`SU(2)`$ gaugino mass $`M_2`$ is held fixed at $`150`$ GeV. Unification of the gaugino masses has been assumed. In addition, $`\mathrm{tan}\beta `$, the ratio of the two Higgs expectation values, has been fixed at 3. Although we have *not* assumed any definite high scale SUSY breaking mechanism, it can be verified that such combinations of parameters can indeed be realised in a supergravity (SUSY) framework. This is possible by using one’s freedom with the trilinear soft SUSY breaking term $`A`$ (while still preserving charge and colour invariance ) and using the Higgsino parameter $`\mu `$ as a phenomenological input , something that is justfied if the Higgs mass parameters retain the freedom of differing from the ‘universal’ sfermion mass parameter ($`m_0`$) at the grand unification scale.
The above specified set of parameters allows one to calculate the R-conserving stop decay width. For the R-parity violating two-body decays, we have used three sets of values for $`\lambda _{233}^{^{}}=\lambda _{333}^{^{}}`$, consistent with the expectation from the SK results. Using these, it is straightforward to calculate the branching ratios of the various stop decay modes. It is found that for the region of parameter space under investigation here, the branching ratio for R-parity violating decays ranges from 75 to 99 per cent, with a near-equal share between the $`b\tau `$ and $`b\mu `$ channels.
A parton level Monte Carlo calculation has been performed for both $`singlemuon+threejets+\mathit{}_T`$ and $`dimuon+twojets+\mathit{}_T`$ final states arising from stop pair-production at the Tevatron. In the former case, we have demanded that at least one jet be identified as a $`b`$-induced one. Both the top- and stop-production cross-sections have been QCD corrected using a a so-called $`K`$-factor of 1.4 in case of top and 1.3 for the stop. For hadronic decays of the tau, we have considered only modes with one charged track, arising from $`\tau ^\pm \pi ^\pm \nu _\tau ,\rho ^\pm \nu _\tau ,a_1^\pm \nu _\tau `$. In order to calibrate our parton-level results, we have computed the numbers of the same types of expected signal events from top quark pair production. The numbers of both single-muon and dimuon events thus calculated in the latter case agree, within small errors, with the actual observations , so that our results may be treated as reasonable estimates of how many stop-induced events can be contained in the top signal.
Since our main purpose is to see how stop signal can percolate into top signals for which detectors are already designed, both have been subjected to the same set of cuts, as specified by the CDF top search strategy. In both types of final states of our interest, jets are defined using the cone algorithm as discussed in Ref.. Jets are counted in the analysis only if $`|\eta |_{jet}<2`$ where $`\eta `$ is the pseudorapidity . For single muon final state we require jets with $`E_T>15`$ GeV, at least one of which is identified as a $`b`$-jet (with identification efficiency $`40`$ % ) while for dimuon final state two jets with $`E_T>10`$ GeV are required . Two jets are merged if $`\mathrm{\Delta }R0.4`$, where $`\mathrm{\Delta }R=\sqrt{\mathrm{\Delta }\eta ^2+\mathrm{\Delta }\varphi ^2}`$, $`\mathrm{\Delta }\eta `$ and $`\mathrm{\Delta }\varphi `$ being the separations in pseudorapidity and azimuthal angle. In both the cases *isolated* muon(s) with $`p_T>20`$ GeV are necessary in the central region with $`|\eta |<1`$ . The criterion for muon isolation imposed here is that the total $`E_T`$ within a cone of $`\mathrm{\Delta }R=0.4`$ around it should be less than 10 % of its own $`p_T`$. In dimuon final states, backgrounds from real $`Z`$ are rejected by requiring the dimuon invariant mass to be outside the interval $`75105`$ GeV . For single muon events, missing transverse energy, $`\mathit{}_T>20`$ GeV is demanded while for dimuons $`\mathit{}_T>25`$ GeV is generally required . It is observed in the last of Ref. that this combination of cuts effectively eliminates backgrounds from Drell-Yan production and other relevant processes for the dimuon final state.
The net contribution of the stop to the top-like signals depends on two effects: the production cross section of a stop pair vs that of a top pair, and the relatively large branching fraction for the stops cascading into the final states of our interest. In addition, of course, the susceptibility to cuts plays a role. For $`m_{\stackrel{~}{t}_1}100`$ GeV the production rates approximately match, and the stop production rate falls for higher stop masses.
Figures 1 and 2 contain our main results, based on an integrated luminosity of $`109`$ pb<sup>-1</sup>. We have checked by varying the parameter $`\mu `$ between $`700`$ and $`+700`$ GeV that the results are not altered by more than about 25 per cent. The dimuon events enable us to rule out a slightly larger range of the stop mass. This is because of more suppression of the top signal in this channel through leptonic branching fractions of $`W`$ as opposed to what happens for stop-pairs. It is clear from figure 2 that the dimuon data should definitely rule out the entire stop mass interval that is otherwise constrained assuming $`B(\stackrel{~}{t}_1c\chi _1^0)1`$. In fact, the limit can perhaps be pushed somewhat higher up in the R-parity violating case. For $`\lambda _{233}^{^{}}\lambda _{333}^{^{}}=10^4`$, this limit should be around 125 GeV with the currently available data, while for $`\lambda ^{^{}}`$-values of $`5\times 10^4`$ it touches about 140 GeV.
From the single muon signal too, stop masses upto about $`120125`$ GeV seem to be ruled out. We also want to emphasize that both this channel and the dimuon one can be used effectively in the context of Run II of the Tevatron to look for SUSY signals burried within the top data. A careful analysis of the dimuon versus dielectron data there, invigorated by the additional available luminosity, may lead to successful identification of an excess in the former, thereby indicating R-parity violation of a kind that simultaneously explains the atmospheric neutrino puzzle.
For simplicity, we have confined ourselves in the above discussion to cases where the stop is kinematically forbidden to decay into a $`b`$ quark and a chargino. Such a decay can be possible when $`M_2`$ and consequently the mass of lightest neutralino is smaller. However, in view of the fact that the chargino can subsequently decay into a lepton, the above mode can in effect strengthen the stop contribution.
The values of R-parity violating couplings which are as small as the ones under study here are unlikely to contribute to precision electroweak data or to R-parity violating decays of the top quark such as $`tb\stackrel{~}{l}`$ or $`t\stackrel{~}{b}l`$ even if they are kinematically allowed. However, with couplings several times larger in magnitude (with larger average squark mass, for example), some of these decays could lead to additional signals (final states of higher multiplicity) from the decays of the top, particularly at Run II of the Fermilab Tevatron.
Another signal, similar in nature to the ones discussed above, is one where both the stops decay into the $`b\mu `$ channel. In this case one shall see dimuons with two jets. Proper $`b`$-identifaction in the jets, together with effective measures to eliminate the Drell-Yan backgrounds, can establish such final states as rather effective ones for discovering a light stop in an R-parity violating theory.
As has been mentioned at the beginning, the above discussion has assumed only trilinear R-parity violating interactions. The presence of the bilinear terms of the form $`L_iH_2`$ in the superpotential (and the consequent vacuum expectation values for sneutrinos) can give rise to mixing between charged leptons and charginos , which can again lead to top-like signals of the same kinds from stop decays. A detailed analysis of the signals in such a case will soon be presented by us .
In conclusion, the first signals of R-parity violating SUSY observable in current and upcoming experiments can very well mimic those of the top quark. This is particularly true if the theory has to account for the neutrino masses and mixing suggested by experimental results on atmospheric muon neutrinos. We can use the already available $`singlelepton+jets`$ as well as *dilepton* data from the Fermilab Tevatron to constrain a large range of the lighter stop mass in such a scenario. In the upgraded version of the Tevatron, too, top quark signals will continue to be useful probes for such types of SUSY.
We thank S. Raychaudhuri and M. Guchait for technical help.
|
warning/0003/physics0003080.html
|
ar5iv
|
text
|
# 1 What is relativity?
## 1 What is relativity?
Arguments used in the priority debate: ”who was (were) the author(s) of the special relativity theory?” tacitly involve, without really addressing it, a fundamental physics issue. It is implicitly assumed that the basic physics behind special relativity is exactly what standard textbooks have been teaching in the last eight decades: Lorentz symmetry would be an abstract, intrinsic property of space-time that matter cannot escape. In this approach, all particles are compelled to move inside the minkowskian space-time. However, such a non-trivial interpretation of special relativity is not a well-established physical law and there is no proof of its absolute validity. It does not correspond to the initial formulation of the Poincaré relativity principle (1895-1905), and current particle theory suggests that Lorentz symmetry can be violated. A close look reveals that historical arguments are biased by physical prejudices and arbitrary interpretations.
The French mathematician Henri Poincaré was the first author to consistently formulate the relativity principle, stating (Poincaré, 1895): ”Absolute motion of matter, or, to be more precise, the relative motion of weighable matter and ether, cannot be disclosed. All that can be done is to reveal the motion of weighable matter with respect to weighable matter”. Such a revolutionary claim was not easily accepted. Poincaré was fighting for a decade to convince other scientists as well as the public opinion. He further emphasized the deep content of this new and, at the time, unconventional law of Nature when he wrote (Poincaré, 1901): ”This principle will be confirmed with increasing precision, as measurements become more and more accurate”.
Although textbooks and press usually present special relativity as having been formulated in the celebrated Einstein’s 1905 paper, several authors have emphasized the actual role of H. Poincaré in building relativity theory previous to Einstein and the relevance of Poincaré’s thought (Logunov, 1995 and 1997; Feynmann, Leighton, & Sands, 1964). In his June 1905 paper (Poincaré, 1905), published before Einsteins’s article (Einstein, 1905) arrived (on June 30) to the editor, Henri Poincaré explicitly wrote the relativistic transformation law for the charge density and velocity of motion and applied to gravity the ”Lorentz group”, assumed to hold for ”forces of whatever origin”. All the ingredients of special relativity, as well as its basic original concepts, are clearly formulated in this work which, furthermore, emphasizes the need of a new, relativistic, theory of gravitation. But Poincaré’s priority is sometimes denied on the scientific grounds that ”Einstein essentially announced the failure of all ether-drift experiments past and future as a foregone conclusion, contrary to Poincaré’s empirical bias” (Miller, 1996), that Poincaré did never ”disavow the ether” (Miller, 1996) or that ”Poincaré never challenges… the absolute time of newtonian mechanics… the ether is not only the absolute space of mechanics… but a dynamical entity” (Paty, 1996). Is this argumentation correct, is it based on well-established physical evidence? We do not think so. In fact, these authors implicitly assume that A. Einstein was right in 1905 when ”reducing ether to the absolute space of mechanics” (Paty, 1996) and that H. Poincaré was wrong because ”the ether fits quite nicely into Poincaré’s view of physical reality: the ether is real…” (Miller, 1996). But, with the present status of particle physics and cosmology, as well as of condensed-matter physics and of the theory of dynamical systems, there is no scientific evidence for Einstein’s 1905 absolute, ”geometric” view of relativity. The existence of a physical ”ether”, playing an important dynamical role, would not be incompatible with the existing (low-energy) experimental evidence for the relativity principle.
### 1.1 Particle physics point of view:
Modern particle physics has brought back the concept of a non-empty vacuum where free particles propagate: without such an ”ether” where fields can condense, the standard model of electroweak interactions could not be written and quark confinement could not be understood. The mechanism producing the masses of the W and Z bosons is close to the Meissner effect, where the condensed Cooper pairs (equivalent to the Higgs field) prevent the magnetic field (virtual photons) from propagationg beyond a certain distance (the London length) inside a superconductor: in the case of the standard model, the effect is mainly observed through the masses (inverse London lengths) of the intermediate bosons propagating in the (vacuum) Higgs field condensate. Furthermore, modern cosmology is not incompatible with the idea of an ”absolute local frame” close to that suggested by the study of cosmic microwave background radiation (see f.i. Peebbles, 1993).
Therefore, the ”ether” may well turn out to be a real entity in the XXI-th century physics and astrophysics. Then, the relativity principle would become a symmetry of physics, another revolutionary concept whose paternity was attributed to H. Poincaré by R.P. Feynman (as quoted by Logunov, 1995): ”Precisely Poincaré proposed investigating what could be done with the equations without altering their form. It was precisely his idea to pay attention to the symmetry properties of the laws of physics”. Actually, Poincaré did even more: in all his papers since 1895, he emphasized another deep concept: the dynamical origin of special relativity.
Dynamics is, by definition, a scale-dependent property of matter. In a global view of physics, a dynamical property of matter would be the opposite concept to an intrinsic, geometric property of space-time. A basic, unanswered question for particle physics is therefore: was Poincaré right when, in his papers since 1895 and in particular in his note of June 1905 strongly premonitory of nowadays grand-unified theories, he considered the relativity principle as a dynamical phenomenon, related to a common origin of all the existing forces?
As symmetries in particle physics are in general violated, Lorentz symmetry may be broken and an absolute local rest frame may be detectable through experiments performed beyond some critical scale or close to that scale. Poincaré’s special relativity (a symmetry applying to physical processes) could live with this situation, but not Einstein’s approach such as it was formulated in 1905 (an absolute geometry of space-time that matter cannot escape). But, how to check whether Lorentz symmetry is actually broken? We discuss here two issues: a) the scale where we may expect Lorentz symmetry to be violated and the scale(s) at which the effect may be observable; b) the physical phenomena and experiments potentially able to uncover Lorentz symmetry violation (LSV). Previous papers on the subject are (Gonzalez-Mestres, 1998a, 1998b, 1998c and 1999a) and references therein. We have proposed that Lorentz symmetry be a low-energy limit, broken following a $`k^2`$-law ($`k`$ = wave vector) between the low-energy region and some fundamental energy (length) scale.
### 1.2 Condensed-matter point of view:
It seems obviously justified to examine what could be a ”condensed-matter point of view”, as particle physics and cosmology have used many condensed-matter analogues in the last five decades (e.g. the pattern of spontaneous symmetry breaking) and the Klein-Gordon equation is a typical equation for wave propagation. Particle physics uses nowadays concepts such as strings, vortices, monopoles, topological defects… which are closely related to condensed-matter phenomena and make reference to the ”vacuum” as a material medium.
Lorentz symmetry, viewed as a property of dynamics, implies no reference to absolute properties of space and time (Gonzalez-Mestres, 1995). In a two-dimensional galilean space-time, the wave equation:
$$\alpha ^2\varphi /t^2^2\varphi /x^2=F(\varphi )$$
(1)
with $`\alpha `$ = $`1/c_o^2`$ and $`c_o`$ = critical speed, remains unchanged under ”Lorentz” transformations leaving invariant the squared interval $`ds^2=dx^2c_o^2dt^2`$ . Any form of matter made with solutions of equation (1) , built in the laboratory in a set-up at rest, would feel a relativistic space-time even if the real space-time is galilean and if an absolute rest frame exists in the underlying dynamics beyond the wave equation. The solitons of the sine-Gordon equation are obtained taking in (1) :
$$F(\varphi )=(\omega /c_o)^2sin\varphi $$
(2)
$`\omega `$ being a characteristic frequency of the dynamical system. The two-dimensional universe made of such sine-Gordon solitons would indeed behave like a two-dimensional minkowskian world with the laws of special relativity. The actual structure of space and time can only be found by going to deeper levels of resolution where the equation fails, similar to the way high-energy accelerator experiments explore the inner structure of ”elementary” particles (but cosmic rays have the highest attainable energies). As modern particle physics views ”elementary” particles as excitations of vacuum, there would be no inconsistency in assumig that the space-time felt by such particles is similar to the soliton analogy. and does not have and absolute meaning. In such a scenario, that cannot be ruled out by any present experiment, superluminal sectors of matter can exist and even be its ultimate building blocks. This clearly makes sense, as: a) in a perfectly transparent crystal, at least two critical speeds can be identified, those of light and sound; b) the potential approach to lattice dynamics in solid-sate physics is precisely the form of electromagnetism in the limit $`c_sc^10`$ , where $`c_s`$ is the speed of sound and $`c`$ that of light. See, for instance, (Gonzalez-Mestres, 1999a) and references therein.
At the fundamental lenght scale, and taking a simplified two-dimensional illustration, gravitation may even be a composite phenomenon (Gonzalez-Mestres, 1997a), related for instance to fluctuations of the parameters of equations like:
$$Ad^2/dt^2[\varphi (n)]+Hd/dt[\varphi (n+1)\varphi (n1)]\mathrm{\Phi }_n[\varphi ]=0$$
(3)
where we have quantized space to schematically account for the existence of the fundamental length $`a`$, $`\varphi `$ is a wave function, $`n`$ designs by an integer lattice sites spaced by a distance $`a`$, $`A`$ and $`H`$ are coeficients and $`\mathrm{\Phi }_n[\varphi ]`$ is defined by:
$$\mathrm{\Phi }_n[\varphi ]=K_{fl}[2\varphi (n)\varphi (n1)\varphi (n+1)]+\omega _{rest}^2\varphi $$
(4)
$`K_{fl}`$ being a coefficient and $`(2\pi )^1\omega _{rest}`$ a rest frequency.
In the continuum limit, the coefficients $`A=g_{00}`$ , $`H=g_{01}=g_{10}`$ and $`K_{fl}=g_{11}`$ can be regarded as the matrix elements of a space-time bilinear metric with equilibrium values: $`A=1`$ , $`H=0`$ and $`K_{fl}=K`$ . Then, a small local fluctuation:
$$A=1+\gamma $$
(5)
$$K_{fl}=K(1\gamma )$$
(6)
with $`\gamma 1`$ would be equivalent to a small, static gravitational field created by a far away source.
In conclusion, our present knowledge of condensed-matter physics does not plead in favour of Einstein’s 1905 , intrinsically geometric, approach to relativity. It instead suggests that Poincaré was right in not ruling out the Ether and in developing instead the concept of physical symmetry (the ”Lorentz group”).
## 2 Lorentz Symmetry As a Low-Energy Limit
Low-energy tests of special relativity have confirmed its validity to an extremely good accuracy ($`10^{21}`$ from nuclear magnetic resonance experiments), but the situation at very high energy remains unclear. Not only high-energy measurements are less precise, but the violation of the relativity principle can be driven by energy-dependent parameters. To discuss possible Lorentz symmetry violation, the hypothesis of a preferred refrence frame seems necessary. In what follows, all discussions are performed in this frame, that we assume to be close to the natural cosmological one defined by cosmic background radiation (see, f.i. Peebles 1993).
If Lorentz symmetry violation (LSV) follows a $`E^2`$ law ($`E`$ = energy), similar to the effective gravitational coupling, it can be $`1`$ at $`E10^{21}eV`$ (just above the highest observed cosmic-ray energies) and $`10^{26}`$ at $`E100MeV`$ (corresponding to the highest momentum scale involved in nuclear magnetic resonance experiments). Such a pattern of LSV (deformed Lorentz symmetry, DLS) will escape all existing low-energy bounds. If LSV is of order 1 at Planck scale ($`E10^{28}eV`$), and following a similar law, it will be $`10^{40}`$ at $`E100MeV`$ . Our suggestion is not in contradiction with Einstein’s thought such as it became after he had developed general relativity. In 1921 , A. Einstein wrote in ”Geometry and Experiment” (Einstein, 1921): ”The interpretation of geometry advocated here cannot be directly applied to submolecular spaces… it might turn out that such an extrapolation is just as incorrect as an extension of the concept of temperature to particles of a solid of molecular dimensions”. The absoluteness of the minkowskian space-time was clearly abandoned through this statement. It is in itself remarkable that special relativity holds at the attained accelerator energies, but there is no fundamental reason for this to be the case above Planck scale.
### 2.1 Deformed relativistic kinematics
A typical example of patterns violating Lorentz symmetry at very short distance is provided by nonlocal models where an absolute local rest frame exists and non-locality in space is introduced through a fundamental length scale $`a`$ where new physics is expected to occur (Gonzalez-Mestres, 1997a). Such models naturally lead to a deformed relativistic kinematics (DRK) of the form (Gonzalez-Mestres, 1997a and 1997b):
$$E=(2\pi )^1hca^1e(ka)$$
(7)
where $`h`$ is the Planck constant, $`c`$ the speed of light, $`k`$ the wave vector, and $`[e(ka)]^2`$ is a convex function of $`(ka)^2`$ obtained from vacuum dynamics. Such an expression is equivalent to special relativity in the small $`k`$ limit. Expanding equation (1) for $`ka1`$ , we can write (Gonzalez-Mestres, 1997a and 1997c):
$`e(ka)`$ $``$ $`[(ka)^2\alpha (ka)^4+(2\pi a)^2h^2m^2c^2]^{1/2}`$ (8)
$`\alpha `$ being a model-dependent constant, in the range $`0.10.01`$ for full-strength violation of Lorentz symmetry at the fundamental length scale, and m the mass of the particle. For momentum $`pmc`$ , we get:
$`E`$ $``$ $`pc+m^2c^3(2p)^1pc\alpha (ka)^2/2`$ (9)
The ”deformation” approximated by $`\mathrm{\Delta }E=pc\alpha (ka)^2/2`$ in the right-hand side of (9) implies a Lorentz symmetry violation in the ratio $`Ep^1`$ varying like $`\mathrm{\Gamma }(k)\mathrm{\Gamma }_0k^2`$ where $`\mathrm{\Gamma }_0=\alpha a^2/2`$ . If $`c`$ is a universal parameter for all particles, the DRK defined by (7) - (9) preserves Lorentz symmetry in the limit $`k0`$, contrary to the standard $`THϵ\mu `$ model (Will, 1993). If, besides $`c`$ , $`\alpha `$ is also universal, LSV does not lead (Gonzalez-Mestres, 1997a, c and e) to the spontaneous decays predicted in (Coleman, & Glashow, 1997 and subsequent papers) at ultra-high energy using a $`THϵ\mu `$-type approach. On more general grounds, as we also pointed out, the existence of very high-energy cosmic rays can by no means be regarded as an evidence against LSV, as the relevant kinematical balances can be sensitive to many small parameters. In particular, any form of LSV should be considered (Gonzalez-Mestres, 1997d and 1997e) and not only (like in the papers by Coleman and Glashow) models driven by low-energy constant parameters.
As previously emphasized, the above non-locality may actually be an approximation to an underlying dynamics involving superluminal particles (Gonzalez-Mestres, 1996, 1997b, 1997f and 1997g), just as electromagnetism looks nonlocal in the potential approximation to lattice dynamics in solid-state physics: it would then correspond to the limit $`cc_i^10`$ where $`c_i`$ is the superluminal critical speed. Contrary to the $`THϵ\mu `$-type scenario considered by Coleman and Glashow, where LSV occurs explicitly in the hamiltonian already at $`k=0`$ through a non-universality of the critical speed in vacuum (the rest masses of charged particles are no longer given by the relation $`E=mc^2`$ , $`c`$ being the speed of light in the $`k0`$ limit), our DLS approach can preserve standard gravitation and general relativity as low-energy limits. Furthermore, phenomenological estimates by Coleman and Glashow do not consider possible deformations of the relativistic kinematics: they use undeformed kinematics for single particles at ultra-high cosmic-ray energies. More recent (1998) papers by these authors bring no new result as compared to our 1997 papers and present the same fundamental limitation as their 1997 article. Physically, the two approaches are really different: by choosing a $`THϵ\mu `$ scenario, Coleman and Glashow implicitly assume that Lorentz symmetry is broken in an ”external” way, by a small macroscopic effect. On the contrary, as explained above, our pattern attributes LSV to an ”internal” very high-energy, very low-distance phenomenon completely disappearing at macroscopic scale.
A fundamental question is whether $`c`$ and $`\alpha `$ are universal. This may be the case for all ”elementary” particles, i.e. quarks, leptons, gauge bosons…, but the situation is less obvious for hadrons, nuclei and heavier objects. From a naive soliton model (Gonzalez-Mestres, 1997b and 1997f), we inferred that: a) $`c`$ is expected to be universal up to extremely small corrections ($`10^{40}`$ , far below the values considered by Coleman and Glashow) escaping all existing bounds; b) an approximate rule can be to take $`\alpha `$ universal for leptons, gauge bosons and light hadrons (pions, nucleons…) and assume a $`\alpha m^2`$ law for nuclei and heavier objects, the nucleon mass setting the scale. With this rule, DRK introduces no anomaly in the relation between inertial and gravitational masses at large scale (Gonzalez-Mestres, 1998c). Basically, the $`\alpha m^2`$ law makes compatible DRK for a large body with a similar DRK for smaller parts of it which, otherwise, could not travel at the same speed as the whole body if the relation $`v=dE/dp`$ (Gonzalez-Mestres, 1997a , 1997d and 1997f) is used.
The main effect of DRK can be decribed as follows. The deformation term increases with energy roughly like $`E^3`$ , whereas the ”mass term” in (9) decreases like $`E^1`$ . The ratio between the two terms varies like $`E^4`$ and, above some energy depending on the parameters involved, the deformation becomes dominant as compared to the mass term. Very-high energy kinematics in the laboratory rest frame is, basically, dominated by longitudinal momentum: as everything else becomes ”small”, and longitudinal momentum has to be exactly conserved, the real kinematical balances occur entirely between ”small” terms. Therefore, a ”small” violation of the relativity principle can potentially play a crucial role in these balances. If $`c`$ is universal, $`\alpha `$ must be positive to avoid the spontaneous decay of UHE (ultra-high energy) particles (unless the effect is not observable below $`3.10^{20}eV`$). The universality of $`\alpha `$ up to small corrections is imposed by the requirement that elementary particles be able to reach very high energies (again, if the effect is to be obervable below $`3.10^{20}eV`$). Otherwise, particles with smaller positive values of $`\alpha `$ would decay into those with larger $`\alpha `$ (Gonzalez-Mestres, 1997e) and the effect would manifest itself in high-energy cosmic-ray events.
### 2.2 Alternative models
The above model is not the only possible way to deform relativistic kinematics. Alternatives are:
Mixing with superluminal sectors (MSLS). A form of DRK was predicted in our papers since 1995, where we attributed (Gonzalez-Mestres, 1995) Lorentz symmetry violation to a very high-energy, very low-distance phenomenon which would modify the propagators of ”ordinary” particles (those with critical speed in vacuum equal to $`c`$ , the speed of light): the dynamics driving LSV was expected to be generated at Planck scale or at some other fundamental length scale. The energy-dependence of LSV was claimed to be the explanation to the apparent validity of the Poincaré relativity principle, as inferred from low-energy tests. We also pointed out (see, e.g. Gonzalez-Mestres, 1996) that ultra-high-energy cosmic rays would be a natural experimental framework to explore possible Lorentz symmetry violation phenomena. A LSV scenario suggested in all these papers was mixing between ”ordinary” and superluminal particles directly deforming propagators (see also Gonzalez-Mestres, 1997d where energy-dependent mixing parameters were explicitly used, preserving Lorentz symmetry in the $`k0`$ limit). Using such models, counterexamples to the claims made in (Coleman and Glashow, 1997) were presented, based on the energy-dependence of LSV effective parameters. Through the parametrizations thus considered, it was also pointed out that, besides a fundamental length scale, masses of heavy superluminal particles can play a significant role in killing low-energy LSV.
The basic idea of our superluminal particle (superbradyons, see e.g. Gonzalez-Mestres, 1997g) model was that several sectors of matter are generated at the fundamental length scale(s), each sector possibly satisfying a ”sectorial” Lorentz invariance vith a ”sectorial” critical speed in vacuum ($`c_i`$ for the $`i`$-th superluminal sector). Superbradyons would have positive mass and energy, and satisfy sectorial motion equations (e.g. Klein-Gordon) with critical speed $`c_i`$ : thay are not tachyons. Dynamical mixing between two sectors would break both Lorentz invariances, and mixing with superluminal sectors (MSLS) would be enough to produce a consistent DRK for ”ordinary” matter. But, if LSV is generated in this way, we also expect more conventional deformations of particle propagators to occur, other than direct mixing between different sectors of matter.
Linear deformation. When building (1997) the DRK approach given by (7)-(9), where the effective deformation parameter depends quadratically on energy (quadratically deformed relativistic kinematics, QDRK), we were also naturally led to consider models where this dependence is linear (linearly deformed relativistic kinematics, LDRK), i.e. where $`e(ka)`$ is a function of $`ka`$ and, for $`ka1`$ :
$`e(ka)[(ka)^2\beta (ka)^3+(2\pi a)^2h^2m^2c^2]^{1/2}`$ (10)
$`\beta `$ being a model-dependent constant. For momentum $`pmc`$ :
$`Epc+m^2c^3(2p)^1pc\beta (ka)/2`$ (11)
the deformation $`\mathrm{\Delta }E=pc\beta (ka)/2`$ being now driven by an effective parameter proportional to momentum, $`\mathrm{\Gamma }(k)=\mathrm{\Gamma }_0^lk`$ where $`\mathrm{\Gamma }_0^l=\beta a/2`$ .
QDRK naturally emerges when a fundamental length scale is introduced to deform the Klein-Gordon equations. It is typical, for instance, of phonons in condensed-matter physics. As recently pointed out (Ellis et al., 1999a and b) using a class of string models, LDRK can be generated by introducing a background gravitational field in the propagation equations of free particles. In the first case, the Planck scale is an internal parameter of the basic wave equations generating the ”elementary” particles as vacuum excitations. In the second case, it manifests itself only as a parameter of the background gravitational field, similar to a refraction phenomenon. By discriminating between the two parametrizations, or excluding both approaches, feasible experiments at available energies can potentially provide very valuable information on fundamental Planck-scale physics. Our choice in 1997 was to concentrate on the QDRK model and disregard LDRK, for phenomenological reasons which seem to remain valid if the new physics is expected to be generated not too far from Planck scale.
If existing bounds on LSV from nuclear magnetic resonance experiments are to be intepreted as setting a bound of $`10^{21}`$ on relative LSV at the momentum scale $`p100MeV`$ , this implies $`\beta a<10^{34}cm`$ . However, as it will be explained later, it turns out that LDRK can lead to many inconsistencies with cosmic-ray experiments unless $`\beta a`$ is much smaller. Concepts and formulae presented in our previous papers for QDRK (see next section) can be readily extended to LDRK, using similar techniques. In particular:
\- the linear deformation term $`pc\beta (ka)/2`$ and the mass term $`m^2c^3(2p)^1`$ become of the same order at the energy scale $`E_{trans}\pi ^{1/3}h^{1/3}(2\beta )^{1/3}a^{1/3}m^{2/3}c^{5/3}`$ ;
\- the linear deformation term and the target energy $`E_T`$ become of the same order at the energy scale $`E_{lim}(2\pi )^{1/2}(E_Ta^1\beta ^1hc)^{1/2}`$ .
\- if the same philosophy as for QDRK is to be followed, $`c`$ and $`\beta `$ would be universal for all ”elementary” particles including light hadrons, whereas for larger objects $`\beta m^1`$ , the nucleon mass setting the scale.
Modifications of QDRK. Our conjecture that $`\alpha `$ has the same value for light hadrons as for the photon and leptons derives from the result (Gonzalez-Mestres, 1997f) that the elementary soliton solution on a one-dimesional space lattice obeys the same deformed kinematics as plane waves on the same lattice and with the same dalembertian operator (discretized in space), if the soliton size scale is basically the quantum inverse of its mass scale. As the highest-energy observed cosmic-ray events seem to be hadronic and not electromagnetic, the conjecture seems sensible on practical grounds. But it can also be argued that quarks are the real elementary particles and that, to be consistent with quark propagation, the value of $`\alpha `$ should be divided by a factor $`4`$ for mesons and $`9`$ for baryons. Actually, the parton picture seems impossible to implement when the deformation becomes important, as in the conventional parton model the constituents can carry arbitrary fractions of energy and momentum at very high energy and the deformation energy depends crucially on these fractions (Gonzalez-Mestres, 1997f): we therefore expect the failure of any parton model at these energies.
A phenomenological discussion of the latter hypothesis is nevertheless worth attempting: for instance, if alpha is to be divided by a factor of 9 for the proton and 4 for the pion, and the highest-energy cosmic-ray events are due to protons, the requirement that the spontaneous decay $`pp+\gamma `$ does not occur would then imply $`\alpha a^2<2.10^{73}cm^2`$ . This would exclude LSV with strong coupling at the Planck scale. A similar kinematical analysis seems to hold for nuclei unless dynamics prevents the decay (see next section). With $`\alpha a^2=2.10^{73}cm^2`$ , a proton with $`E10^{21}eV`$ could also emit pions and, above $`E10^{20}eV`$ , pion lifetimes would become much shorter than predicted by special relativity and charged pions can emit photons.
## 3 QDRK and Ultra-High Energy Cosmic-Ray (UHECR) Physics
If Lorentz symmetry is broken at Planck scale or at some other fundamental length scale, the effects of LSV may be accessible to experiments well below this energy: in particular, they can produce detectable phenomena at the highest observed cosmic ray energies. DRK (Gonzalez-Mestres 1997a, 1997b, 1997c, 1997h and 1998a) plays a crucial role. Taking the quadratic deformation (QDRK) version of DRK, it is found that, at energies above $`E_{trans}\pi ^{1/2}h^{1/2}(2\alpha )^{1/4}a^{1/2}m^{1/2}c^{3/2}`$, the very small deformation $`\mathrm{\Delta }E`$ dominates over the mass term $`m^2c^3(2p)^1`$ in (3) and modifies all kinematical balances: physics gets thus closer to Planck scale than to electroweak scale (this is actually the case in a logarithmic plot of energy scales) and UHECR become an efficient probe of Planck-scale physics. Because of the negative value of $`\mathrm{\Delta }E`$ , it costs more and more energy, as energy increases above $`E_{trans}`$, to split the incoming logitudinal momentum in the laboratory rest frame. As the ratio $`m^2c^3(2p\mathrm{\Delta }E)^1`$ varies like $`E^4`$ , the transition at $`E_{trans}`$ is very sharp.
With such a LSV pattern, we also inferred (Gonzalez-Mestres, 1997f) from a toy soliton model that the parton picture (in any version), as well as standard relativistic formulae for Lorentz contraction and time dilation, are expected to fail above this energy (Gonzalez-Mestres, 1997b and 1997f) which corresponds to $`E10^{20}eV`$ for $`m`$ = proton mass and $`\alpha a^210^{72}cm^2`$ (f.i. $`\alpha 10^6`$ and $`a`$ = Planck length), and to $`E10^{18}eV`$ for $`m`$ = pion mass and $`\alpha a^210^{67}cm^2`$ (f.i. $`\alpha 0.1`$ and $`a`$ = Planck length). Such effects are in principle detectable. A phenomenological study of the implications of DRK allowed to draw several important conclusions for UHECR experiments (Gonzalez-Mestres, 1997-99).
### 3.1 Our previous predictions with QDRK
Assuming that the earth moves slowly with respect to the absolute rest frame (the ”vacuum rest frame”), so that the approximation (9) remains valid in the laboratory rest frame, QDRK can lead to observable phenomena in future experiments devoted to the highest-energy cosmic rays:
a) For $`\alpha a^2>10^{72}cm^2`$ , assuming universal values of $`\alpha `$ and $`c`$ , there is no Greisen-Zatsepin-Kuzmin (GZK) cutoff (Greisen 1966; Zatsepin & Kuzmin, 1966) for the particles under consideration. Due to the new kinematics, interactions with cosmic microwave background (CMB) photons are strongly inhibited or forbidden, and ultra-high energy cosmic rays (e.g. protons) from anywhere in the presently observable Universe can reach the earth (Gonzalez-Mestres, 1997a and 1997c). In particular, for an incoming UHE nucleon hitting a CMB photon, the $`\mathrm{\Delta }`$ resonance can no longer be formed due to the deformation term. Proton deceleration in astrophysical objects (e.g. gamma-ray bursters) can be inhibited in a similar way.
b) With the same hypothesis, unstable particles with at least two stable particles in the final states of all their decay channels become stable at very high energy. Above $`E_{trans}`$, the lifetimes of all unstable particles (e.g. the $`\pi ^0`$ in cascades) become much longer than predicted by relativistic kinematics (Gonzalez-Mestres, 1997a, 1997b and 1997c). Then, for instance, the neutron or even the $`\mathrm{\Delta }^{++}`$ can be candidates for the primaries of the highest-energy cosmic ray events. If $`c`$ and $`\alpha `$ are not exactly universal, many different scenarios can happen concerning the stability of ultra-high-energy particles (Gonzalez-Mestres, 1997a, 1997b and 1997c).
c) In astrophysical processes at very high energy, similar mechanisms can inhibit radiation under external forces (e.g. synchrotron-like, where the interactions occur with virtual photons), GZK-like cutoffs, decays, photodisintegration of nuclei, momentum loss trough collisions (e.g. with a photon wind in reverse shocks), production of lower-energy secondaries… potentially contributing to solve all basic problems raised by the highest-energy cosmic rays (Gonzalez-Mestres, 1997e), including acceleration mechanisms.
d) With the same hypothesis, the allowed final-state phase space of two-body collisions is strongly reduced at very high energy, leading (Gonzalez-Mestres, 1997e) to a sharp fall of partial and total cross-sections for incoming cosmic ray energies above $`E_{lim}(2\pi )^{2/3}(E_Ta^2\alpha ^1h^2c^2)^{1/3}`$, where $`E_T`$ is the energy of the target. As a consequence, and with the previous figures for Lorentz symmetry violation parameters, above some energy $`E_{lim}`$ between 10<sup>22</sup> and $`10^{24}`$ $`eV`$ a cosmic ray will not deposit most of its energy in the atmosphere and can possibly fake an exotic event with much less energy (Gonzalez-Mestres, 1997e).
e) Actually, requiring simultaneously the absence of GZK cutoff in the region $`E10^{20}eV`$ , and that cosmic rays with energies below $`3.10^{20}eV`$ deposit most of their energy in the atmosphere, leads in the DRK scenario to the constraint: $`10^{72}cm^2<\alpha a^2<10^{61}cm^2`$ , equivalent to $`10^{20}<\alpha <10^9`$ for $`a10^{26}cm`$ ($`10^{21}GeV`$ scale). Remarkably enough, assuming full-strength LSV forces $`a`$ to be in the range $`10^{36}cm<a<10^{30}cm`$ . But a $`10^6`$ LSV at Planck scale can still explain the data. Thus, the simplest version of QDRK naturally fits, on phenomenological grounds, with the expected potential role of Planck scale in generating the standard ”elementary” particles and opening the door to new physics.
f) Effects a) to e) are obtained applying DRK to single particles and collisions. If further dynamical anomalies are added (failure, at very small distance scales, of the parton model and of the standard relativistic formulae for Lorentz contraction and time dilation…), we can expect much stronger effects in the early cascade development profiles of cosmic-ray events (Gonzalez-Mestres, 1997b, 1997f and 1998a). Detailed phenomenology and data analysis in next-generation experiments may uncover spectacular new physics and provide a powerful microscope directly focused on the fundamental length (Planck?) scale.
g) Cosmic superluminal particles would produce atypical events with very small total momentum (due to the high $`E/p`$ ratio), isotropic or involving several jets (Gonzalez-Mestres, 1996, 1997b, 1997d, 1997 and 1998b). In the atmosphere (f.i. AUGER or satellite-based experiments), such events would generate exceptional cascade development profiles and muon spectra, as will be discussed in a forthcoming paper.
\***************************************************************************************
It should be noticed that our description of all these phenomena and, in particular, of points a) and b) , on the grounds of DRK was prior to any similar claim by Coleman and Glashow from $`THϵ\mu `$-like models.
\***************************************************************************************
It follows from b) and f) that early cascade development is a crucial point for possible tests of special relativity in UHCR cosmic-ray events. Unfortunately, this part of the interactions induced by the incoming cosmic ray is not easily detectable: it occurs early in the atmosphere, and it takes a few collisions before energy is degraded into a large enough number of particles to produce an observable fluorescence signal. Data on cascade development start well below the first interaction of the primary with the atmosphere whereas we expect LSV, if present, to manifest itself only in the first few collisions. But the effect will propagate to later multiparticle production and be observable: for instance, if the $`\pi ^0`$ does not decay at very high energy, we expect a smaller electromagetic component developing later and more muons produced. A serious drawback is the present ambiguousness of phenomenological air shower models, but combined data from AUGER and satellite-based experiments should help to clarify this situation. A recent fit to the UHECR spectrum with a model close to QDRK, reproducing the absence of GZK cutoff, can be found in (Chechin & Vavilov, 1999).
### 3.2 On cosmic-ray composition
Cosmic-ray composition above $`10^{17}eV`$ is a crucial question. Protons are often preferred as candidates to UHECR events (Bird et al., 1993), but detailed analysis are not yet conclusive (f.i. AGASA Collaboration, 1999) and there are claims in favor of light nuclei up to 2.10$`{}_{}{}^{19}eV`$ (f.i. Wolfendale & Wibig, 1999). It is commonly agreed that the UHECR composition becomes lighter above $`5.10^{17}eV`$ , as energy increases, and tends to be protons at the highest observed energies. It may be interesting to compare this phenomenon with QDRK predictions, keeping in mind the proposed $`\alpha m^2`$ law. This law naturally allows, kinematically, spontaneous $`NN+\gamma `$ decays ($`N`$ = nucleus) at very high energy. If $`\alpha a^210^{67}cm^2`$ , the threshold for spontaneous gamma emission would be $`3.10^{19}eV`$ for $`Fe`$ and $`10^{19}eV`$ for $`He^4`$ .
It should be noticed, however, that although spontaneous gamma emission will be kinematically allowed for large objects because of the $`\alpha m^2`$ law, it will become more and more unnatural dynamically as their size increases. Such an outgoing photon would carry an energy much larger than that of any incoming nucleon and its wavelength would be smaller, by orders of magnitude, than the size of the composite object.
### 3.3 Neutrinos from gamma-ray bursts
The basic idea (Waxman and Bahcall, 1999) is that, in reverse shocks, protons are accelerated to energies $`10^{20}eV`$ and collide with ambient photons producing, in the kinematical region where the center of mass energy corresponds to the $`\mathrm{\Delta }^+`$ resonance, charged pions with about 20$`\%`$ of the initial proton energy. These pions subsequently decay into muons and muon neutrinos ($`\pi ^+\mu ^++\nu _\mu `$), and the muons decay later into electrons, electron neutrinos and muon antineutrinos ($`\mu ^+e^++\nu _e+\overline{\nu }_\mu `$) . It should be noticed that, beacuse of Lorentz dilation, the lifetime for the first decay at pion energy $`2.10^{19}eV`$ is $`3000s`$ whereas the muon lifetime at $`10^{19}eV`$ is $`10^5s`$ . These time scales are already longer than those of gamma-ray bursts. But the situation, for UHE neutrino production, may be considerably worsened by LSV. For $`\alpha a^2>10^{72}cm^2`$ , the lifetime of a $`10^{20}eV`$ $`\mathrm{\Delta }^+`$ is modified by QDRK and gets much longer. If $`\alpha a^2>10^{70}cm^2`$ , the resonance becomes stable at the same energy with respect to the $`ne^+\nu _\mu \overline{\nu }_\mu \nu _e`$ decay channel. Also, as previoulsy stated, much higher photon energies are required to form a $`\mathrm{\Delta }`$ resonance. In both cases, pion photoproduction is strongly inhibited and the calculation presented in (Waxman and Bahcall, 1999) is to be modified leading to a lower neutrino flux. Similarly, the lifetimes of UHE charged pions and muons become much longer and lower the neutrino flux further. At the same time, QDRK may inhibit UHE proton syncrotron radiation and allow the proton to be accelerated to energies above $`10^{20}eV`$ .
As DRK makes velocity, $`v=dE/dp`$ , energy-dependent (Gonzalez-Mestres, 1997a and 1997d), the arrival time on earth of particles produced in a single burst is expected to dependent on the particle energy. But, contrary to the claim presented in (Ellis et al., 1999a), it follows from the above considerations that observing UHE neutrinos from GRB bursts for QDRK with $`\alpha a^2>10^{72}cm^2`$ will most likely be much more difficult (if at all possible) than naively expected from the Waxman-Bahcall model.
## 4 LDRK, gamma-ray bursts and TeV physics
Considering vacuum as a medium similar to an electromagnetic plasma, it was suggested in (Amelino-Camelia et al., 1998) that quantum-gravitational fluctuations may lead to a correction, linear in energy, to the velocity of light. This is equivalent to a LDRK that, for $`ka1`$ , can be parameterized as:
$`Epcpc\beta (ka)/2=pcp^2M^1`$ (12)
where $`M`$ is an effective mass scale. Possible tests of this model through gamma-ray bursts, measuring the delays in the arrival time of photons of different energies, have been considered in (Norris et al., 1999) having in mind the Gamma-ray Large Area Space Telescope (GLAST) and in (Ellis et al., 1999b) with a more general scope. Biller et al. (1998) claim a lower bound on $`M`$ slightly above $`10^{16}GeV`$ .
However, from the same considerations already developed in our 1997-99 papers taking QDRK as an exemple, stringent bounds on LDRK can be derived. Assume that LDRK applies only to photons, and not to charged particles, so that at high energy we can write for a charged particle, $`ch`$ , the dispersion relation:
$`E_{ch}`$ $``$ $`p_{ch}c+m_{ch}^2c^3(2p_{ch})^1`$ (13)
where the $`ch`$ subscript stands for the charged particle under consideration. Then, it can be readily checked that the decay $`chch+\gamma `$ would be allowed for $`p`$ above $`(2m_{ch}^2Mc^3)^{1/3}`$ , i.e:
\- for an electron, above $`E2TeV`$ if $`M=10^{16}GeV`$ and $`20TeV`$ if $`M=10^{19}GeV`$
\- for a muon or charged pion, above $`E80TeV`$ if $`M=10^{16}GeV`$ and $`800TeV`$ if $`M=10^{19}GeV`$
\- for a proton, above $`E240TeV`$ if $`M=10^{16}GeV`$ and $`2.4PeV`$ if $`M=10^{19}GeV`$
\- for a $`\tau `$ lepton, above $`E400TeV`$ if $`M=10^{16}GeV`$ and $`4PeV`$ if $`M=10^{19}GeV`$
so that none of these particles would be oberved above such energies, apart from very short paths. Such decays seem to be in obvious contradiction with cosmic ray data, but avoiding them forces the charged particles to have the same kind of propagators as the photon, with the same effective value of $`M`$ up to small differences. Similar conditions are readily derived for all ”elementary” particles, leading for all of them, up to small devations, to a LDRK given by the universal dispersion relation:
$`Epc+m^2c^3(2p)^1p^2M^1`$ (14)
For instance, $`\pi ^0`$ production would otherwise be inhibited. But if, as it seems compulsory, the $`\pi ^0`$ kinematics follows a similar law, then the decay time for $`\pi ^0\gamma \gamma `$ will become much longer than predicted by special relativity at energies above $`50TeV`$ if $`M=10^{16}GeV`$ and $`500TeV`$ if $`M=10^{19}GeV`$ . Again, this seems to be in contradiction with cosmic-ray data. Requiring that the $`\pi ^0`$ lifetime agrees with special relativity at $`E10^{17}eV`$ would force $`M`$ to be above $`10^{26}GeV`$ , far away from the values to be tested at GLAST. Another bound is obtained from the condition that there are $`3.10^{20}eV`$ cosmic-ray events. Setting $`E_{lim}`$ to this value, and taking oxygen to be the target, yields $`M3.10^{21}GeV`$ . In view of these bounds, it appears very difficult to make LDRK , with $`M`$ reasonably close to Planck scale, compatible with experimental data. It therefore seems necessary to reconsider the models to be tested at GLAST.
## 5 Conclusions and comments
If, as conjectured by Poincaré, special relativity is a symmetry of dynamical origin, Lorentz symmetry violation at very high energy would not be unnatural. But checking it appears to be a difficult task. If LSV is generated at some fundamental length (e.g. Planck) scale, we expect it to be driven by energy-dependent parameters becoming very small in the low-energy region where impressive tests of the validity of the relativity principle have been performed (Lamoreaux et al., 1986; Hills & Hall, 1990). It therefore seems natural, contrary to the $`THϵ\mu `$ model, to preserve Lorentz symmetry as a low-energy limit. In deformed Lorentz symmetry models, leading in particular to various versions of deformed relativistic kinematics, the deformation disappears in the limit $`k0`$ . DRK is far from being unique, as it can be generated in many different ways, and its predictions are strongly model-dependent. Unlike a previous attempt (Kirzhnits & Chechin, 1972), we have pursued the idea that: a) the fundamental length scale is at the origin of LSV and should be taken seriously in all respects; b) physics can vary smoothly down to this scale where a new dynamics manifests itself; c) LSV is not related to any other fundamental symmetry generalizing special relativity. It then turns out that, at energies well below the fundamental length scale, a very small LSV can generate observable leading-order effects and even allow to discriminate between models of vacuum at Planck scale.
Ambitious prospects, based on LDRK (linearly deformed relativistic kinematics, obtained from Planck-scale ”vacuum recoil” models), to measure a possible systematic energy-dependence in time delays from gamma-ray bursts are confronted to apparent incompatibilities between cosmic-ray data and the orders of magnitude of LDRK parameters considered. The basic reason is that LSV, if realized in this way, would manifest itself in many other phenomena already accessible to experiment (cosmic rays in the TeV - PeV range). However, the fact that the study of LDRK has led to consider such comparatively low energies pleads in favour of precision tests of special relativity at the highest-energy accelerators (LHC, VLHC and beyond). Although the existence of observable effects of LDRK at the gamma-ray burst photon energies seems unlikely, systematic tests of special relativity at energies between 1 TeV and 1 PeV are missing and should be performed.
On phenomenological grounds, QDRK (quadratically deformed relativistic kinematics, obtained when a fundamental length scale is introduced in the dynamics generating ”elementary” particles) seems to be the most performant model and naturally fits with possible new Planck-scale physics. UHE (ultra-high energy) CR (cosmic-ray) physics appears to be the safest laboratory to test LSV, allowing for unconventional phenomena in early cascade development. If Lorentz symmetry is violated, the study of UHECR events may be a powerful tool to get direct information on fundamental physics at Planck scale. At the highest oberved cosmic-ray energies, the effect of LSV at Planck scale can already lead to spectacular signatures: absence of the GZK cutoff; drastic modifications of lifetimes, as well as of total and partial cross-sections; failure of any parton picture as well as of standard formulae for time dilation and Lorentz contraction… But further work is required to clearly translate the physical signatures into measurable data (cascade development profile, muons, electromagnetic yield…). Models of air shower formation should be improved in order to remove many interpretation uncertainties. A basic difficulty is that, although the primary interaction occurs quite early in the atmosphere, a detectable fluorescence yield is emitted only after a few interactions, when many charged particles have been produced and the atmosphere is denser. Combined information from future experiments, such as AUGER and satellite-based measurements, will hopefully make this task easier.
The existence of superluminal particles (superbradyons) is not excluded if LSV occurs at Planck scale: they may even be the ultimate constituents of matter. This subject will be discussed at length elsewhere.
References
AGASA Collaboration, 1999, Proceedings of the 26<sup>th</sup> International Cosmic Ray Conference (ICRC 1999), Salt Lake City August 1999, paper HE 2.3.02.
Amelino-Camelia, G., Ellis, J., Mavromatos, N.E., Nanopoulos, D.V. & Sarkar, S., Nature 393, 319 (1998).
Biller, S.D., Breslin, A.C., Buckley, J., (Biller et al.), 1998, paper gr-qc/9810044 .
Bird, D.J., Corbato, S.C., Dhai, H.Y. et al. (Bird et al.), 1993, Proceedings of the 23<sup>th</sup> International Cosmic Ray Conference (ICRC 1993), Calgary 1993, Vol. 2, p. 38.
Chechin, V.A., & Vavilov, Yu.N., ICRC 1999 Proceedings, paper HE 2.3.07.
Coleman, S., & Glashow, S., 1997, Phys. Lett. B 405, 249 and subsequent (1998) papers in LANL (Los Alamos) electronic archive as well as in the TAUP 97 Proceedings.
Einstein, A., 1905, ”Zur Elecktrodynamik bewegter Körper”, Annalen der Physik 17, 891.
Einstein, A., 1921, ”Geometrie und Erfahrung”, Preus. Akad. der Wissench., Sitzungsberichte, part I, p. 123.
Ellis, J., Mavromatos, N.E., Nanopoulos, D.V. & Volkov, G. (Ellis et al.) 1999a, paper gr-qc/9911055.
Ellis, J., Farakos, K., Mavromatos, N., Mitsou, V.A. & Nanopoulos, D.V. (Ellis et al.), paper astro-ph/9907340.
Feynman, R.P., Leighton, R.B., & Sands, M., 1964 ”The Feynman Lectures on Physics”, Addison-Wesley.
Greisen, K., 1966, Phys. Rev. Lett. 16 , 748.
Gonzalez-Mestres, L., 1995, paper astro-ph/9505117 of LANL electronic archive, Proceedings of the Moriond Workshop on ”Dark Matter, Cosmology, Clocks and Fundamental Laws”, January 1995, Ed. Frontières.
Gonzalez-Mestres, L., 1996, paper hep-ph/9610474 of LANL electronic archive and references therein.
Gonzalez-Mestres, L., 1997a, paper physics/9704017 of LANL archive.
Gonzalez-Mestres, L., 1997b, Proc. of the International Conference on Relativistic Physics and some of its Applications, Athens June 25-28 1997, Apeiron, p. 319.
Gonzalez-Mestres, L., 1997c, Proc. 25th ICRC (Durban, 1997), Vol. 6, p. 113.
Gonzalez-Mestres, L., 1997d, papers physics/9702026 and physics/9703020 of LANL archive.
Gonzalez-Mestres, L., 1997e, papers physics/9706022 and physics/9706032 of LANL archive.
Gonzalez-Mestres, L., 1997f, paper nucl-th/9708028 of LANL archive.
Gonzalez-Mestres, L., 1997g, Proc. 25th ICRC (Durban, 1997), Vol. 6, p. 109.
Gonzalez-Mestres, L., 1997h, talk given at the International Workshop on Topics on Astroparticle and Underground Physics (TAUP 97), paper physics/9712005 of LANL archive.
Gonzalez-Mestres, L., 1998a, talk given at the ”Workshop on Observing Giant Cosmic Ray Air Showers From $`>`$ 10<sup>20</sup> Particles From Space”, College Park, November 1997, AIP Conference Proceedings 433, p. 148.
Gonzalez-Mestres, L., 1998b, same Proceedings, p. 418.
Gonzalez-Mestres, L., 1998c, Proc. of COSMO 97, Ambleside September 1997, World Scientific, p. 568.
Gonzalez-Mestres, L., 1999a, ICRC 1999 Proceedings, paper HE. 1.3.16.
Gonzalez-Mestres, L., 1999b, ICRC 1999 Proceedings, paper OG 3.1.10.
Hills, D. & Hall, J.L., 1990, Phys. Rev. Lett. 64 , 1697.
Kirzhnits, D.A., and Chechin, V.A., 1972, Soviet Journal of Nuclear Physics, 15 , 585.
Lamoreaux, S.K., Jacobs, J.P., Heckel, B.R., Raab, F.J. & Forston, E.N., 1986, Phys. Rev. Lett. 57 , 3125.
Logunov, A.A., 1995, ”On the articles by Henri Poincaré on the dynamics of the electron”, Ed. JINR, Dubna.
Logunov, A.A., 1997, ”Relativistic theory of gravity and the Mach principle”, Ed. JINR, Dubna.
Miller, A.I. 1996, Why did Poincaré not formulate Special Relativity in 1905?”, in ”Henri Poincaré, Science and Philosophy”, Ed. Akademie Verlag, Berlin, and Albert Blanchard, Paris.
Norris, J.P., Bricell, J.T., Marani, G.F. & Scargle, J.D., 1999, ICRC 1999 Proceedings, paper OG 2.3.11.
Paty, M., 1996, ”Poincaré et le principe de relativité”, same reference.
Peebles, P.J.E., 1993, ”Principles of Physical Cosmology”, Princeton University Press. Poincaré, H., 1895, ”A propos de la théorie de M. Larmor”, L’Eclairage électrique, Vol. 5, 5.
Poincaré, H., 1901, ”Electricité et Optique: La lumière et les théories électriques”, Ed. Gauthier-Villars, Paris.
Poincaré, H., 1905, Sur la dynamique de l’électron”, Comptes Rendus Acad. Sci. Vol. 140, p. 1504, June 5.
Waxman, E. & Bahcall, J., 1999, paper hep-ph/9909286.
A.W. Wolfendale & T. Wibig, 1999, ICRC 1999 Proceedings, paper OG 1.3.01.
Will, C., 1993, ”Theory and Experiment in Gravitational Physics”, Cambridge University Press.
Zatsepin, G.T. and Kuzmin, V.A., Pisma Zh. Eksp. Teor. Fiz. 4 , 114.
|
warning/0003/gr-qc0003080.html
|
ar5iv
|
text
|
# Self-similar spherically symmetric cosmological models with a perfect fluid and a scalar field
## 1 Introduction
Scalar field cosmology is of importance in the study of the early Universe and particularly in the investigation of inflation (during which the universe undergoes a period of accelerated expansion ). One particular class of inflationary cosmological models are those with a scalar field and an exponential potential of the form $`V(\varphi )=V_0e^{\kappa \varphi }`$, where $`\kappa `$ and $`V_0`$ are constants. Models with an exponential scalar field potential arise naturally in alternative theories of gravity, such as, for example, scalar-tensor theories, and are currently of particular interest since such theories occur as the low-energy limit in supergravity theories .
A number of authors have studied scalar field cosmological models with an exponential potential within general relativity. Homogeneous and isotropic Friedmann-Robertson-Walker (FRW) models were first studied by Halliwell using phase-plane methods. Homogeneous but anisotropic models of Bianchi types I and III (and Kantowski-Sachs models) were studied by Burd and Barrow in which they found exact solutions and discussed their stability. Bianchi models of types III and VI were studied by Feinstein and Ibáñez , in which exact solutions were found. An analysis of Bianchi models, including standard matter satisfying various energy conditions, was completed by Kitada and Maeda . They found that the well-known power-law inflationary solution is an attractor for all initially expanding Bianchi models (except a subclass of the Bianchi type IX models which will recollapse) when $`\kappa ^2<2`$.
The governing differential equations in spatially homogeneous Bianchi cosmologies containing a scalar field with an exponential potential reduce to a dynamical system when appropriate normalised variables are defined; this dynamical system was studied in detail in . In a follow-up paper the isotropisation of the Bianchi VII<sub>h</sub> cosmological models possessing a scalar field with an exponential potential was further investigated; in the case $`\kappa ^2>2`$, it was shown that there is an open set of initial conditions in the set of anisotropic Bianchi VII<sub>h</sub> initial data such that the corresponding cosmological models isotropise asymptotically. Hence, spatially homogeneous scalar field cosmological models having an exponential potential with $`\kappa ^2>2`$ can isotropise to the future. The Bianchi type IX models have also been studied in more detail .
Recently, cosmological models which contain both a perfect fluid and a scalar field with an exponential potential have come under heavy analysis . One of the exact solutions found for these models has the property that the energy density due to the scalar field is proportional to the energy density of the perfect fluid, hence these models have been labelled scaling cosmologies . These scaling solutions, which are spatially flat isotropic models, are of particular physical interest. For example, in these models a significant fraction of the current energy density of the Universe may be contained in the scalar field whose dynamical effects mimic cold dark matter. In the stability of these cosmological scaling solutions within the class of spatially homogeneous cosmological models with a perfect fluid subject to the equation of state $`p=(\gamma 1)\mu `$ (where $`\gamma `$ is a constant satisfying $`0<\gamma <2`$) and a scalar field with an exponential potential was studied. It was found that when $`\gamma >2/3`$, and particularly for realistic matter with $`\gamma 1`$, the scaling solutions are unstable; essentially they are unstable to curvature perturbations, although they are stable to shear perturbations. Curvature scaling solutions and anisotropic scaling solutions are also possible. In particular, in homogeneous and isotropic spacetimes with non-zero spatial curvature were studied.
Clearly it is of interest to study more general cosmological models, and in this paper we shall comprehensively study the qualitative properties of the class of self-similar spherically symmetric models with a barotropic fluid and a non-interacting scalar field with an exponential potential. Self-similar spherically symmetric perfect fluid models with a linear equation of state have been much studied in general relativity . Carr & Coley have presented a complete asymptotic classification of such solutions and, by reformulating the field equations for these models, Goliath *et al* have obtained a compact three-dimensional state space representation of the solutions which leads to another complete picture of the solution space. Recently, these models have been further studied in a combined approach . The present analysis can be thought of as a natural extension of this recent work .
The Kantowski-Sachs models appear as a limiting case of the class of spherically symmetric models under investigation, and hence this analysis complements recent analyses of spatially homogeneous Bianchi models . Models with positive spatial curvature have attracted less attention than Bianchi models with zero spatial curvature or negative spatial curvature since they are more complicated mathematically. However, the properties of positive-curvature FRW models and Kantowski-Sachs models have been studied previously.
In the next section we shall describe the governing equations of the class of models under investigation. In section 3, compact variables are defined and the resulting dynamical system is derived in the case of spatial self-similarity. A monotone function is obtained. The equilibrium points and their local stability is discussed in section 4. The timelike self-similar case is then considered in sections 5 and 6. The special equilibrium points for values of “extreme tilt” are discussed separately in section 7. The global results and a discussion is given in section 8. Applications in the absence of a barotropic fluid and in the further subcase of a massless scalar field are discussed, respectively, in sections 9 and 10, partially to illustrate the early-time and late-time behaviour of the models.
## 2 Governing equations
We shall consider spherically symmetric similarity solutions in which the source for the gravitational field is a perfect fluid and a non-interacting scalar field with an exponential potential in which the total energy-momentum tensor is given by:
$$T_{ab}=(T_{\mathrm{pf}ab}+T_{\mathrm{sf}ab}),$$
(1)
where
$$T_{\mathrm{pf}ab}=\mu u_au_b+p\left(u_au_b+g_{ab}\right).$$
(2)
The perfect fluid obeys the equation of state
$$p=(\gamma 1)\mu ,$$
(3)
where $`\gamma `$ is a constant satisfying $`1<\gamma <2`$. The scalar-field contribution is given by
$$T_{\mathrm{sf}ab}=\varphi _{,a}\varphi _{,b}\left[\frac{1}{2}\varphi _{,c}\varphi ^{,c}+V(\varphi )\right]g_{ab},$$
(4)
where
$$V(\varphi )=V_0e^{\kappa \varphi }.$$
(5)
Since the fluid and scalar field are non-interacting, we have the following separate conservation laws:
$$_aT_{\mathrm{pf}}^{ab}=0=_aT_{\mathrm{sf}}^{ab}.$$
(6)
The spacetime is self-similar and consequently admits a homothetic vector $`\eta ^a`$. This implies that the matter fields must be of a particular form. Thus, a barotropic fluid must have an equation of state of the form (3) . The energy-momentum tensor of the scalar field must satisfy:
$$_\eta T_{\mathrm{sf}ab}=0.$$
(7)
This implies that
$`\varphi `$ $`=`$ $`\mathrm{\Phi }(\xi )+{\displaystyle \frac{2}{\kappa }}\eta ,`$ (8)
$`V`$ $`=`$ $`e^{2\eta }𝒱(\mathrm{\Phi }),`$ (9)
$`𝒱(\mathrm{\Phi })`$ $`=`$ $`V_0e^{\kappa \mathrm{\Phi }(\xi )},`$ (10)
where $`\eta `$ is the variable defined by the homothetic vector $`\eta ^a`$, and $`\xi `$ is the similarity variable. When the similarity variable is timelike, we will use the notation $`t\xi ,x\eta `$. The homothetic vector then is spacelike, and we denote this as the spatially self-similar case. When the homothetic vector is timelike, we have the timelike self-similar case, for which $`t\eta ,x\xi `$. A dot denotes differentiation with respect to the similarity variable. Finally, we shall define the new variable:
$$X=\frac{1}{\sqrt{2}}\dot{\mathrm{\Phi }},$$
(11)
and for convenience we introduce the new constant
$$k\frac{\sqrt{2}}{\kappa }.$$
(12)
## 3 Spatially self-similar case
In the spatially self-similar case, the line element can be written
$`ds^2`$ $`=`$ $`e^{2x}d\overline{s}^2,`$ (13)
$`d\overline{s}^2`$ $`=`$ $`dt^2+D_1^2dx^2+D_2^2d\mathrm{\Omega }^2,`$ (14)
$`D_1`$ $`=`$ $`e^{\beta ^02\beta ^+},D_2=e^{\beta ^0+\beta ^+}.`$ (15)
The kinematic quantities of the congruence normal to the symmetry surfaces are related to the Misner variables ($`\beta ^0`$, $`\beta ^+`$) as follows:
$$\theta =3\dot{\beta }^0,\sigma _+=3\dot{\beta }^+.$$
(16)
Following , we will work with boosted kinematic quantities ($`\overline{\theta },\overline{\sigma }_+`$)
$$\theta =\frac{1}{\sqrt{3}}\left(2\overline{\theta }+\overline{\sigma }_+\right),\sigma _+=\frac{1}{\sqrt{3}}\left(\overline{\theta }+2\overline{\sigma }_+\right).$$
(17)
The reason for this is that it simplifies the constraint obtained from the non-vanishing off-diagonal component of the field equations. The metric functions $`B_1D_1^1`$ and $`B_2D_2^1`$ then have the following evolution equations:
$`\dot{B_1}`$ $`=`$ $`{\displaystyle \frac{1}{\sqrt{3}}}\overline{\sigma }_+B_1,`$ (18)
$`\dot{B_2}`$ $`=`$ $`{\displaystyle \frac{1}{\sqrt{3}}}(\overline{\theta }+\overline{\sigma }_+)B_2.`$ (19)
The physical quantities associated with the perfect fluid are as follows:
$`u_{\mathrm{pf}}^a`$ $`=`$ $`{\displaystyle \frac{e^x}{\sqrt{1v^2}}}(1,ve^{2\beta ^+\beta ^0},0,0),`$ (20)
$`\mu `$ $`=`$ $`{\displaystyle \frac{1v^2}{1+(\gamma 1)v^2}}e^{2x}\mu _n,`$ (21)
where $`v`$ is the tilt variable, and $`\mu _n`$ is the energy density along the normal congruence. From the conservation equations (6) for the perfect fluid, we have:
$`\dot{\mu }_n`$ $`=`$ $`{\displaystyle \frac{\gamma }{\sqrt{3}\left[1+(\gamma 1)v^2\right]}}\left[2\overline{\theta }+(1v^2)\overline{\sigma }_++2\sqrt{3}vB_1\right]\mu _n,`$ (22)
$`\dot{v}`$ $`=`$ $`{\displaystyle \frac{1v^2}{\sqrt{3}\gamma \left[1(\gamma 1)v^2\right]}}\{\gamma [2(\gamma 1)\overline{\theta }+\gamma \overline{\sigma }_+]v`$ (23)
$`+\sqrt{3}[(\gamma 1)(3\gamma 2)v^2(2\gamma )]B_1\},`$
and the conservation equation for the scalar field yields
$`\dot{X}`$ $`=`$ $`{\displaystyle \frac{1}{\sqrt{3}}}(2\overline{\theta }+\overline{\sigma }_+)X+2kB_1^2+{\displaystyle \frac{𝒱}{k}}.`$ (24)
The Einstein field equations then yield the following:
The Friedmann equation:
$`\mu _n`$ $`=`$ $`{\displaystyle \frac{1}{3}}\left[\overline{\theta }^2\overline{\sigma }_{+}^{}{}_{}{}^{2}3(1+k^2)B_1^2+3B_2^23X^23𝒱\right].`$ (25)
Constraint equation:
$`0`$ $`=`$ $`\gamma v\mu _n{\displaystyle \frac{2}{\sqrt{3}}}\left[1+(\gamma 1)v^2\right]\left(\overline{\sigma }_++\sqrt{3}kX\right)B_1.`$ (26)
Evolution equations for $`\overline{\theta }`$ and $`\overline{\sigma }_+`$:
$`\dot{\overline{\theta }}`$ $`=`$ $`{\displaystyle \frac{1}{\sqrt{3}}}[\overline{\theta }^2+\overline{\sigma }_{+}^{}{}_{}{}^{2}+\overline{\theta }\overline{\sigma }_+3(1+k^2)B_1^2+3X^23𝒱`$ (27)
$`+`$ $`{\displaystyle \frac{3(\gamma 1)(1v^2)}{1+(\gamma 1)v^2}}\mu _n],`$
$`\dot{\overline{\sigma }_+}`$ $`=`$ $`{\displaystyle \frac{1}{\sqrt{3}}}[\overline{\sigma }_{+}^{}{}_{}{}^{2}+2\overline{\theta }\overline{\sigma }_++6k^2B_1^2+3𝒱`$ (28)
$`+{\displaystyle \frac{3}{2}}{\displaystyle \frac{(2\gamma )+(3\gamma 2)v^2}{1+(\gamma 1)v^2}}\mu _n].`$
From equation (25), by demanding $`\mu _n0`$, a dominant quantity
$`\overline{Y}`$ $`=`$ $`\sqrt{\overline{\theta }^2+3B_2^2}`$ (29)
is identified. Thus, we define bounded variables according to
$`\overline{Q}_0={\displaystyle \frac{\overline{\theta }}{\overline{Y}}},\overline{Q}_+={\displaystyle \frac{\overline{\sigma }_+}{\overline{Y}}},\overline{C}_1={\displaystyle \frac{\sqrt{3}B_1}{\overline{Y}}},\overline{U}={\displaystyle \frac{\sqrt{3}X}{\overline{Y}}},\overline{W}={\displaystyle \frac{\sqrt{3𝒱}}{\overline{Y}}}.`$ (30)
Defining an appropriate density parameter with respect to $`\mu _n`$, the Friedmann equation takes the form:
$`\mathrm{\Omega }_n`$ $`=`$ $`{\displaystyle \frac{3\mu _n}{\overline{Y}^2}}`$ (31)
$`=`$ $`1\overline{Q}_{+}^{}{}_{}{}^{2}(1+k^2)\overline{C}_{1}^{}{}_{}{}^{2}\overline{U}^2\overline{W}^2,`$ (32)
while the constraint equation becomes
$`0`$ $`=`$ $`\gamma v\mathrm{\Omega }_n2\left[1+(\gamma 1)v^2\right](\overline{Q}_++k\overline{U})\overline{C}_1.`$ (33)
By defining a new independent variable
$`=`$ $`{\displaystyle \frac{d}{d\tau }}={\displaystyle \frac{\sqrt{3}}{\overline{Y}}}{\displaystyle \frac{d}{dt}},`$ (34)
the evolution equation for $`\overline{Y}`$
$`\overline{Y}^{}`$ $`=`$ $`\left\{\overline{Q}_++\overline{Q}_0\left[2(\overline{Q}_{+}^{}{}_{}{}^{2}+\overline{U}^2)+{\displaystyle \frac{\gamma }{1+(\gamma 1)v^2}}\mathrm{\Omega }_n\right]\right\}\overline{Y}`$ (35)
decouples, and we are left with a reduced set of evolution equations:
$`\overline{Q}_{0}^{}{}_{}{}^{}`$ $`=`$ $`(1\overline{Q}_{0}^{}{}_{}{}^{2})\left[12(1\overline{Q}_{+}^{}{}_{}{}^{2}\overline{U}^2)+{\displaystyle \frac{\gamma }{1+(\gamma 1)v^2}}\mathrm{\Omega }_n\right],`$ (36)
$`\overline{Q}_{+}^{}{}_{}{}^{}`$ $`=`$ $`\overline{Q}_0\overline{Q}_+\left[2(1\overline{Q}_{+}^{}{}_{}{}^{2}\overline{U}^2)+{\displaystyle \frac{\gamma }{1+(\gamma 1)v^2}}\mathrm{\Omega }_n\right]`$ (37)
$`\left[2k^2\overline{C}_{1}^{}{}_{}{}^{2}+\overline{W}^2+{\displaystyle \frac{1}{2}}{\displaystyle \frac{(2\gamma )+(3\gamma 2)v^2}{1+(\gamma 1)v^2}}\mathrm{\Omega }_n\right],`$
$`\overline{C}_{1}^{}{}_{}{}^{}`$ $`=`$ $`2\overline{C}_1\left[\overline{Q}_++\overline{Q}_0(\overline{Q}_{+}^{}{}_{}{}^{2}+\overline{U}^2)+{\displaystyle \frac{1}{2}}{\displaystyle \frac{\gamma }{1+(\gamma 1)v^2}}\overline{Q}_0\mathrm{\Omega }_n\right],`$ (38)
$`v^{}`$ $`=`$ $`{\displaystyle \frac{1v^2}{\gamma \left[1(\gamma 1)v^2\right]}}\{\gamma [2(\gamma 1)\overline{Q}_0+\gamma \overline{Q}_+]v+`$ (39)
$`+[(\gamma 1)(3\gamma 2)v^2(2\gamma )]\overline{C}_1\},`$
$`\overline{U}^{}`$ $`=`$ $`\overline{Q}_0\overline{U}\left[2(1\overline{Q}_{+}^{}{}_{}{}^{2}\overline{U}^2)+{\displaystyle \frac{\gamma }{1+(\gamma 1)v^2}}\mathrm{\Omega }_n\right]`$ (40)
$`+2k\overline{C}_{1}^{}{}_{}{}^{2}+{\displaystyle \frac{\overline{W}^2}{k}},`$
$`\overline{W}^{}`$ $`=`$ $`{\displaystyle \frac{\overline{W}}{k}}\left\{k\overline{Q}_+\overline{U}+k\overline{Q}_0\left[2(\overline{Q}_{+}^{}{}_{}{}^{2}+\overline{U}^2)+{\displaystyle \frac{\gamma }{1+(\gamma 1)v^2}}\mathrm{\Omega }_n\right]\right\}.`$ (41)
This system is invariant under the transformation
$$(\tau ,\overline{Q}_0,\overline{Q}_+,\overline{C}_1,v,\overline{U},\overline{W})(\tau ,\overline{Q}_0,\overline{Q}_+,\overline{C}_1,v,\overline{U},\overline{W}).$$
(42)
Furthermore, $`\overline{W}0`$, and by noting the invariance under the transformation $`(\overline{C}_1,v)(\overline{C}_1,v)`$, we can without loss of generality restrict the analysis to $`\overline{C}_10`$. There is also an auxiliary evolution equation for $`\mathrm{\Omega }_n`$:
$`\mathrm{\Omega }_n^{}`$ $`=`$ $`\mathrm{\Omega }_n\left\{{\displaystyle \frac{\gamma }{1+(\gamma 1)v^2}}\left[2\overline{Q}_0+(1v^2)\overline{Q}_++2v\overline{C}_1\right]+2{\displaystyle \frac{\overline{Y}^{}}{\overline{Y}}}\right\}.`$ (43)
In appendix A, expressions for some important fluid quantities are given.
### 3.1 Invariant submanifolds
A number of invariant submanifolds can be identified:
* Plane symmetric: $`\overline{Q}_0=\pm 1`$, which implies $`B_2=0`$ (see ).
* Non-self-similar Kantowski-Sachs: $`\overline{C}_1=0,v=0`$ (see ).
* Massless scalar field: $`\overline{W}=0`$.
* No perfect fluid: $`\mathrm{\Omega }_n=0`$ (and $`v`$ decouples).
* No scalar field: $`\overline{U}=0`$, $`\overline{W}=0`$ and $`k=0`$.
The case with no scalar field was studied in , where the global dynamics of these models was investigated in detail. The resulting state space in this case is effectively three-dimensional and is illustrated in that reference. We note that the only self-similar FRW models are the zero-curvature models which occur both as an equilibrium point in the plane symmetric invariant set (with $`\overline{Q}_0=\pm 1=2\overline{Q}_+`$), and as an orbit in the interior of the state space.
### 3.2 Monotone function
The evolution equation for $`\mathrm{\Xi }\overline{Q}_++k\overline{U}`$ is given by
$`\mathrm{\Xi }^{}`$ $`=`$ $`\overline{Q}_0\mathrm{\Xi }\left[2(1\overline{Q}_{+}^{}{}_{}{}^{2}\overline{U}^2)+{\displaystyle \frac{\gamma }{1+(\gamma 1)v^2}}\mathrm{\Omega }_n\right]`$ (44)
$`{\displaystyle \frac{1}{2}}{\displaystyle \frac{(2\gamma )+(3\gamma 2)v^2}{1+(\gamma 1)v^2}}\mathrm{\Omega }_n,`$
which we note is of the same form as the evolution equation for $`\overline{Q}_+`$ in the perfect fluid case. Noting the form of the constraint (33), we consider the following function $`\overline{M}`$ (cf. ):
$`\overline{M}`$ $`=`$ $`(1\overline{Q}_{0}^{}{}_{}{}^{2})^{2\gamma }\mathrm{\Xi }^{3\gamma 4}\overline{C}_{1}^{}{}_{}{}^{\gamma }v^2(1v^2)^{(2\gamma )}.`$ (45)
A direct calculation then yields:
$`\overline{M}^{}`$ $`=`$ $`\left[{\displaystyle \frac{(3\gamma 2)(2\gamma )}{\gamma v}}(1v^2)\overline{C}_1\right]\overline{M},`$ (46)
that is, $`\overline{M}`$ is monotonic in both the $`v<0`$ and the $`v>0`$ regions.
When $`v=0`$ the constraint yields $`\mathrm{\Xi }\overline{C}_1=0`$. But since $`\overline{C}_1=0`$, $`v=0`$ is an invariant (boundary) set, if $`\overline{C}_1=0`$ then $`\overline{C}_1=0`$, $`v=0`$ always. Hence on the surface $`v=0`$ and in the interior of the state space $`\overline{C}_1>0`$. Setting $`v=0`$ in the evolution equation for $`v`$ then gives
$`v^{}`$ $`=`$ $`{\displaystyle \frac{2\gamma }{\gamma }}\overline{C}_1,`$ (47)
which is strictly negative. That is, all orbits in the interior region pass from $`v>0`$ to $`v<0`$ across the surface $`v=0`$ (i.e. the surface $`v=0`$ acts as a membrane). Consequently there can be no closed or recurrent orbits in the interior of the state space.
## 4 Equilibrium points for the spatially self-similar case
We shall display all of the equilibrium points below along with their eigenvalues. We will not present the corresponding eigenvectors explicitly. In what follows, $`ϵ=\pm 1`$ is the sign of $`\overline{Q}_0`$, which indicates whether the corresponding solution is expanding $`(+)`$ or contracting $`()`$. The quantity $`\overline{\mathrm{\Omega }_\varphi }=\overline{U}^2+\overline{W}^2`$ is the scalar-field contribution to the density parameter $`\mathrm{\Omega }_n`$. The order of the dependent variables is $`(\overline{Q}_0,\overline{Q}_+,\overline{C}_1,v,\overline{U},\overline{W})`$. The ’$`\pm `$’ suffices on the labels for equilibrium points correspond to the sign of $`\overline{Q}_0`$ (i.e. the value of $`ϵ`$). Equilibrium points that act as attractors are listed in table 1.
### 4.1 No scalar field ($`\overline{U}=0`$, $`\overline{W}=0`$)
#### 4.1.1 K-points <br>
These are special points on the K-rings, defined in subsection 4.2.1. They all have $`\overline{Q}_0=\pm 1`$, $`\overline{Q}_+=\pm 1`$, and all other variables equal to zero.
#### 4.1.2 Flat Friedmann <br>
$`{}_{\pm }{}^{}\mathrm{F}`$: $`(ϵ,\frac{1}{2}ϵ,0,0,0,0)`$.
$`\mathrm{\Omega }_n=\frac{3}{4}`$, $`\overline{\mathrm{\Omega }_\varphi }=0`$, $`q_{\mathrm{pf}}=\frac{3\gamma 2}{2}`$.
Eigenvalues ($`\overline{C}_1`$ eliminated):
$`{\displaystyle \frac{1}{2}}(3\gamma 2)ϵ,{\displaystyle \frac{3}{4}}(2\gamma )ϵ,{\displaystyle \frac{1}{4}}(3\gamma 2)ϵ,{\displaystyle \frac{3}{4}}(2\gamma )ϵ,{\displaystyle \frac{3\gamma }{4}}ϵ.`$
These points are saddles.
#### 4.1.3 Self-similar Kantowski-Sachs <br>
The state space contains the non-self-similar Kantowski-Sachs solutions as a boundary submanifold. In this submanifold, the self-similar Kantowski-Sachs solution appears as an equilibrium point.
$`{}_{\pm }{}^{}\mathrm{KS}`$: $`(\sqrt{\frac{2\gamma }{4(\gamma 1)}}ϵ,\sqrt{\frac{\gamma 1}{2\gamma }}ϵ,0,0,0,0)`$, $`\gamma <2/3`$.
As this solution is physical only when $`\gamma <2/3`$, we will not consider it further.
### 4.2 Massless scalar field ($`\overline{U}0`$, $`\overline{W}=0`$)
#### 4.2.1 K-rings <br>
$`{}_{\pm }{}^{}\mathrm{K}`$: $`(ϵ,\pm \sqrt{1\overline{U}^2},0,0,\overline{U},0)`$.
$`\mathrm{\Omega }_n=0`$, $`\overline{\mathrm{\Omega }_\varphi }=\overline{U}^2`$, $`q_{\mathrm{pf}}=2`$.
Eigenvalues ($`\overline{C}_1`$ eliminated):
$`2ϵ,(2\gamma )\left[\overline{Q}_++2ϵ\right],2(\gamma 1)ϵ+\gamma \overline{Q}_+,0,2ϵ+\overline{Q}_+{\displaystyle \frac{\overline{U}}{k}}.`$
Each K-ring corresponds to a one-parameter family of equilibrium points (and hence gives rise to a zero eigenvalue). They are analogues of the Kasner solutions in the case with no scalar field . For each K-ring, there is a subset of future or past attractors.
$`{}_{+}{}^{}\mathrm{K}`$: sources and saddles.
$`{}_{}{}^{}\mathrm{K}`$: sinks and saddles.
#### 4.2.2 M-points <br>
$`{}_{\pm }{}^{}\mathrm{M}_{}^{\stackrel{~}{v}}`$: $`(ϵ,k^2fϵ,f,\stackrel{~}{v}ϵ,kfϵ,0)`$, $`f=1/(1+k^2)`$, $`\gamma >\frac{6}{5}`$,
$`\stackrel{~}{v}_{1,2}=\frac{\gamma \left[(\gamma 1)\frac{2\gamma }{\gamma }k^2\right]\pm \sqrt{(\gamma 1)(2\gamma )(3\gamma 2)+\gamma ^2\left[(\gamma 1)\frac{2\gamma }{\gamma }k^2\right]^2}}{(\gamma 1)(3\gamma 2)}`$.
$`\mathrm{\Omega }_n=0`$, $`\overline{\mathrm{\Omega }_\varphi }=\left(\frac{k}{1+k^2}\right)^2`$.
Eigenvalues ($`\overline{C}_1`$ eliminated):
$`2fϵ,2(1k^2)fϵ,(1k^2)fϵ,F_1(\gamma ,k)fϵ,`$
$`{\displaystyle \frac{1}{\gamma v}}\left[2\gamma +(3\gamma 2)v^2+2\gamma v\right]fϵ,`$
These equilibrium points are related to the Milne points $`\stackrel{~}{\mathrm{M}}`$ in . They are only physical ($`|\stackrel{~}{v}|<1`$) for certain ranges of $`\gamma `$ and $`k`$. For instance, when $`k<1`$ it follows that $`|\stackrel{~}{v}_2|>1`$. When $`k>1`$, these points are saddles. Furthermore, examining the eigenvalues numerically for $`0<k<1`$, it turns out that the points always are saddles.
#### 4.2.3 Curvature-scaling solutions <br>
$`{}_{\pm }{}^{}\mathrm{X}_{}^{\widehat{v}}`$: $`(2kgϵ,kgϵ,g,\widehat{v}ϵ,gϵ,0)`$, $`g=1/(\sqrt{2}\sqrt{1+k^2})`$, $`k<1`$ ($`\kappa ^2>2`$),
$`\widehat{v}_{1,2}=\frac{(3\gamma 4)\gamma \frac{k}{2}\pm \sqrt{(\gamma 1)(2\gamma )(3\gamma 2)+(3\gamma 4)^2\gamma ^2\frac{k^2}{4}}}{(\gamma 1)(3\gamma 2)}`$.
$`\mathrm{\Omega }_n=0`$, $`\overline{\mathrm{\Omega }_\varphi }=\frac{1}{2(1+k^2)}`$.
Eigenvalues ($`\overline{C}_1`$ eliminated):
$`{\displaystyle \frac{1k^2}{k}}gϵ,F_2(\gamma ,k,\widehat{v})ϵ,F_3(\gamma ,k,\widehat{v})ϵ,`$
$`F_4(\gamma ,k,\widehat{v})ϵ,F_5(\gamma ,k,\widehat{v})ϵ.`$
For $`k=1`$ ($`\kappa ^2=2`$) these points coincide with $`{}_{\pm }{}^{}\mathrm{M}_{}^{\stackrel{~}{v}}`$, and for $`k>1`$ ($`\kappa ^2<2`$) they are unphysical. Note that $`0<\widehat{v}_1<1`$ and $`1<\widehat{v}_2<0`$ only when $`\gamma >4/3`$, assuming $`k<1`$. For $`\gamma <4/3`$, it follows that $`|\widehat{v}|>1`$, and the equilibrium points are unphysical. Numerical evaluation of the eigenvalues shows that these points are saddles.
#### 4.2.4 Equilibrium lines with arbitrary $`v`$ <br>
$`{}_{\pm }{}^{}\mathrm{\Phi }_{}^{v}`$: $`(ϵ,2\frac{\gamma 1}{\gamma }ϵ,0,v,\pm \frac{1}{\gamma }\sqrt{(2\gamma )(3\gamma 2)},0)`$.
$`\mathrm{\Omega }_n=0`$, $`\overline{\mathrm{\Omega }_\varphi }=(2\gamma )(3\gamma 2)/\gamma ^2`$.
Eigenvalues ($`\overline{Q}_+`$ eliminated):
$`0,0,2ϵ,{\displaystyle \frac{2}{\gamma }}(2\gamma )ϵ,{\displaystyle \frac{2}{\gamma }}ϵ{\displaystyle \frac{\overline{U}}{k}}.`$
There are two zero eigenvalues for these points. The first zero eigenvalue corresponds to the fact that we have a line of equilibrium points. The second zero eigenvalue indicates that the equilibria are non-hyperbolic. For $`v=0,\pm 1`$, these equilibrium lines coincide with the various K-rings, see subsections 4.2.1 and 7.2.1, and these exceptional points mark where the K-rings change stability. The higher-order zero eigenvalue of $`{}_{\pm }{}^{}\mathrm{\Phi }_{}^{v}`$ corresponds to the eigenvalue associated with the fact that $`{}_{\pm }{}^{}\mathrm{K}`$ is a line of equilibrium points (and not to the eigenvalue that becomes zero due to the stability change of $`{}_{\pm }{}^{}\mathrm{K}`$), and the corresponding eigenvector is $`\stackrel{}{v}=\frac{\overline{U}}{\overline{Q}_+}\stackrel{}{e}_{\overline{Q}_+}+\stackrel{}{e}_{\overline{U}}`$. Perturbing the equilibrium lines $`{}_{\pm }{}^{}\mathrm{\Phi }_{}^{v}`$ along this eigenvector, we find that
$`\overline{Q}_{+}^{}{}_{}{}^{}`$ $`=`$ $`2(1\overline{Q}_{+}^{}{}_{}{}^{2}\overline{U}^2)ϵ\overline{Q}_+,`$ (48)
$`\overline{U}^{}`$ $`=`$ $`2(1\overline{Q}_{+}^{}{}_{}{}^{2}\overline{U}^2)ϵ\overline{U}.`$ (49)
This is precisely the dynamical system restricted to the invariant set $`\overline{Q}_0=ϵ`$, $`\overline{C}_1=0`$, $`\overline{W}=0`$, $`\mathrm{\Omega }_n=0`$. We can explicitly integrate equations (48) and (49). It follows that $`\overline{Q}_+`$ is proportional to $`\overline{U}`$, and the orbits in the ($`\overline{Q}_+,\overline{U}`$) plane consist of straight lines through the origin with additional equilibrium points at $`{}_{\pm }{}^{}\mathrm{\Phi }_{}^{v}`$ (where $`\overline{Q}_+=2\frac{\gamma 1}{\gamma }\overline{Q}_0`$), which are thus non-linear saddles.
### 4.3 Scalar field with potential ($`\overline{U}0`$, $`\overline{W}0`$)
There are a number of solutions with a non-zero potential listed below. There are also equilibrium points $`{}_{\pm }{}^{}\mathrm{Z}_{}^{v^{}}`$ with variable values
$`(ϵ,0,\frac{1}{\sqrt{1k^2}},v^{},0,\sqrt{\frac{2k^2}{1k^2}})`$, $`v_{1,2}^{}=\frac{(\gamma 1)\gamma \sqrt{1k^2}ϵ\pm \sqrt{(\gamma 1)(2\gamma )(3\gamma 2)+(\gamma 1)^2\gamma ^2(1k^2)}}{(\gamma 1)(3\gamma 2)}`$,
but these points are unphysical, since either $`\overline{C}_1`$ or $`\overline{W}`$ is imaginary.
#### 4.3.1 Scalar-field dominated solutions <br>
$`{}_{\pm }{}^{}\mathrm{\Phi }`$: $`(ϵ,\frac{1}{2}ϵ,0,0,\frac{1}{2k}ϵ,\frac{1}{2k}\sqrt{3k^21})`$, $`k^2>\frac{1}{3}`$ ($`\kappa ^2<6`$).
$`\mathrm{\Omega }_n=0`$, $`\overline{\mathrm{\Omega }_\varphi }=\frac{3}{4}`$, $`q_{\mathrm{pf}}=\frac{1}{k^2}(1k^2)=\frac{1}{2}(2\kappa ^2)`$.
Eigenvalues (constraint degenerate):
$`{\displaystyle \frac{k^21}{k^2}}ϵ,{\displaystyle \frac{3k^21}{2k^2}}ϵ,{\displaystyle \frac{3k^21}{2k^2}}ϵ,{\displaystyle \frac{k^21}{2k^2}}ϵ,`$
$`{\displaystyle \frac{3\gamma 4}{2}}ϵ,{\displaystyle \frac{3\gamma k^22}{2k^2}}ϵ.`$
When $`\mathrm{\Omega }_n=0`$, the constraint defines two hypersurfaces ($`\overline{C}_1=0`$ and $`\overline{Q}_++k\overline{U}=0`$), and these hypersurfaces coincide at $`{}_{\pm }{}^{}\mathrm{\Phi }`$. It turns out that the constraint is degenerate ($`G=0`$) at these equilibrium points. Consequently, all eigenvector directions are physical there, so we need to keep all six eigenvalues. For $`k^2=1/3`$ ($`\kappa ^2=6`$) these points coincide with points in the K-rings, and for $`k^2<1/3`$ ($`\kappa ^2>6`$) they are unphysical.
$`{}_{+}{}^{}\mathrm{\Phi }`$: sink when $`\gamma <\frac{4}{3}`$, $`k^2>1`$ ($`\kappa ^2<2`$); saddle otherwise.
$`{}_{}{}^{}\mathrm{\Phi }`$: source when $`\gamma <\frac{4}{3}`$, $`k^2>1`$ ($`\kappa ^2<2`$); saddle otherwise.
We note that $`{}_{+}{}^{}\mathrm{\Phi }`$ corresponds to the flat FRW power-law inflationary solution .
#### 4.3.2 Curvature-scaling solutions <br>
$`{}_{\pm }{}^{}\mathrm{\Xi }`$: $`(\frac{1}{2kg}ϵ,kgϵ,0,0,gϵ,\frac{1}{\sqrt{2}})`$, $`g=1/(\sqrt{2}\sqrt{1+k^2})`$, $`k>1`$ ($`\kappa ^2<2`$).
$`\mathrm{\Omega }_n=0`$, $`\overline{\mathrm{\Omega }_\varphi }=\frac{2+k^2}{2(1+k^2)}`$, $`q_{\mathrm{pf}}=\frac{1k^2}{2+k^2}=\frac{\kappa ^22}{2(\kappa ^2+1)}<0`$.
Eigenvalues (constraint degenerate):
$`{\displaystyle \frac{1}{k}}\left[2(\gamma 1)+\gamma k^2\right]gϵ,{\displaystyle \frac{1}{2k}}\left[k^2+1\pm \sqrt{(k^2+1)(9k^27)}\right]gϵ,`$
$`{\displaystyle \frac{1}{k}}(k^2+1)gϵ,{\displaystyle \frac{1}{k}}(k^21)gϵ,{\displaystyle \frac{1}{k}}\left[\gamma (2\gamma )(k^2+1)\right]gϵ.`$
For $`k=1`$ ($`\kappa ^2=2`$) these points coincide with $`{}_{\pm }{}^{}\mathrm{\Phi }`$, and for $`k<1`$ ($`\kappa ^2>2`$) they are unphysical. When physical, these points are always saddles.
#### 4.3.3 Friedmann scaling solution <br>
$`{}_{\pm }{}^{}\mathrm{FS}`$: $`(ϵ,\frac{1}{2}ϵ,0,0,\frac{3k}{4}\gamma ϵ,\frac{3k}{4}\sqrt{\gamma (2\gamma )})`$, $`k^2<\frac{2}{3\gamma }`$ ($`\kappa ^2>3\gamma `$).
$`\mathrm{\Omega }_n=\frac{3}{8}(23\gamma k^2)`$, $`\overline{\mathrm{\Omega }_\varphi }=\frac{9}{8}\gamma k^2`$, $`q_{\mathrm{pf}}=\frac{3\gamma 2}{2}`$.
Eigenvalues ($`\overline{C}_1`$ eliminated):
$`{\displaystyle \frac{1}{2}}(3\gamma 2)ϵ,{\displaystyle \frac{3}{4}}(2\gamma )ϵ,1,`$
$`{\displaystyle \frac{3}{8}}\left[(2\gamma )\pm \sqrt{(2\gamma )(12\gamma ^2k^29\gamma +2)}\right]ϵ.`$
For $`k^2=\frac{2}{3\gamma }`$ ($`\kappa ^2=3\gamma `$) these points coincide with $`{}_{\pm }{}^{}\mathrm{\Phi }`$, and for $`k^2>\frac{2}{3\gamma }`$ ($`\kappa ^2<3\gamma `$) they are unphysical. When physical, these points are always saddles.
## 5 Timelike self-similar case
In the timelike self-similar case, the line element can be written
$`ds^2`$ $`=`$ $`e^{2t}d\stackrel{~}{s}^2,`$ (50)
$`d\stackrel{~}{s}^2`$ $`=`$ $`D_1^2dt^2+dx^2+D_2^2d\mathrm{\Omega }^2,`$ (51)
$`D_1`$ $`=`$ $`e^{\beta ^02\beta ^+},D_2=e^{\beta ^0+\beta ^+}.`$ (52)
The kinematic quantities of the congruence normal to the symmetry surfaces are related to ($`\beta ^0`$, $`\beta ^+`$) by:
$$\theta =3\dot{\beta }^0,\sigma _+=3\dot{\beta }^+.$$
(53)
Note that for the timelike self-similar case, the symmetry surfaces $`x=`$ constant are timelike, so the normal congruence is spacelike. As for the spatial case, it is convenient to boost the kinematic quantities in order to simplify the constraint :
$$\theta =\frac{1}{\sqrt{3}}\left(2\overline{\theta }+\overline{\sigma }_+\right),\sigma _+=\frac{1}{\sqrt{3}}\left(\overline{\theta }+2\overline{\sigma }_+\right).$$
(54)
The evolution equations for the metric functions $`B_1D_1^1`$ and $`B_2D_2^1`$ are
$`\dot{B_1}`$ $`=`$ $`{\displaystyle \frac{1}{\sqrt{3}}}\overline{\sigma }_+B_1,`$ (55)
$`\dot{B_2}`$ $`=`$ $`{\displaystyle \frac{1}{\sqrt{3}}}(\overline{\theta }+\overline{\sigma }_+)B_2.`$ (56)
In terms of the coordinates we use, the physical quantities associated with the perfect fluid are given by
$`u_{\mathrm{pf}}^a`$ $`=`$ $`{\displaystyle \frac{e^t}{\sqrt{1u^2}}}(e^{2\beta ^+\beta ^0},u,0,0),`$ (57)
$`\mu `$ $`=`$ $`{\displaystyle \frac{1u^2}{1+(\gamma 1)u^2}}e^{2t}\mu _t,`$ (58)
where $`u`$ is related to the tilt, and $`\mu _t`$ is the energy density with respect a congruence projected onto the surfaces of symmetry. The conservation equations for the perfect fluid yield
$`\dot{u}`$ $`=`$ $`{\displaystyle \frac{1u^2}{\sqrt{3}\gamma \left[u^2(\gamma 1)\right]}}\{\gamma [2(\gamma 1)\overline{\theta }+\gamma \overline{\sigma }_+]u`$ (59)
$`+\sqrt{3}[(\gamma 1)(3\gamma 2)(2\gamma )u^2]B_1\}.`$
There is also an evolution equation for $`\dot{\mu }_t`$, but as it is rather lengthy and not used elsewhere, we will not give it here. The conservation equation for the scalar field is
$`\dot{X}`$ $`=`$ $`{\displaystyle \frac{1}{\sqrt{3}}}(2\overline{\theta }+\overline{\sigma }_+)X+2kB_1^2{\displaystyle \frac{𝒱}{k}},`$ (60)
and the Einstein field equations give:
The Friedmann equation:
$`\mu _t`$ $`=`$ $`{\displaystyle \frac{1}{3}}{\displaystyle \frac{1+(\gamma 1)u^2}{u^2+(\gamma 1)}}\left(\overline{\theta }^2\overline{\sigma }_{+}^{}{}_{}{}^{2}3(1+k^2)B_1^23B_2^23X^2+3𝒱\right).`$ (61)
Constraint equation:
$`0`$ $`=`$ $`\gamma u\mu _t{\displaystyle \frac{2}{\sqrt{3}}}\left[1+(\gamma 1)u^2\right](\overline{\sigma }_++\sqrt{3}kX)B_1.`$ (62)
Evolution equations for $`\overline{\theta }`$ and $`\overline{\sigma }_+`$:
$`\dot{\overline{\theta }}`$ $`=`$ $`{\displaystyle \frac{1}{\sqrt{3}}}(\overline{\theta }^2+\overline{\sigma }_{+}^{}{}_{}{}^{2}+\overline{\theta }\overline{\sigma }_+3(1+k^2)B_1^2+3X^2+3𝒱`$ (63)
$`{\displaystyle \frac{3(\gamma 1)(1u^2)}{1+(\gamma 1)u^2}}\mu _t),`$
$`\dot{\overline{\sigma }_+}`$ $`=`$ $`{\displaystyle \frac{1}{\sqrt{3}}}(\overline{\sigma }_{+}^{}{}_{}{}^{2}+2\overline{\theta }\overline{\sigma }_++6k^2B_1^23𝒱`$ (64)
$`+{\displaystyle \frac{3}{2}}{\displaystyle \frac{(3\gamma 2)+(2\gamma )u^2}{1+(\gamma 1)u^2}}\mu _t).`$
From (61) and by demanding $`\mu _t0`$ it follows that
$`\stackrel{~}{Y}`$ $`=`$ $`\sqrt{\overline{\theta }^2+3𝒱}`$ (65)
is a dominant quantity. Consequently, we define bounded variables as follows:
$`\stackrel{~}{Q}_0={\displaystyle \frac{\overline{\theta }}{\stackrel{~}{Y}}},\stackrel{~}{Q}_+={\displaystyle \frac{\overline{\sigma }_+}{\stackrel{~}{Y}}},\stackrel{~}{C}_1={\displaystyle \frac{\sqrt{3}B_1}{\stackrel{~}{Y}}},\stackrel{~}{C}_2={\displaystyle \frac{\sqrt{3}B_2}{\stackrel{~}{Y}}},\stackrel{~}{U}={\displaystyle \frac{\sqrt{3}X}{\stackrel{~}{Y}}}.`$ (66)
The Friedmann equation becomes an equation for the density parameter $`\mathrm{\Omega }_t`$:
$`\mathrm{\Omega }_t`$ $`=`$ $`{\displaystyle \frac{3\mu _t}{\stackrel{~}{Y}^2}}`$ (67)
$`=`$ $`{\displaystyle \frac{1+(\gamma 1)u^2}{u^2+(\gamma 1)}}\left(1\stackrel{~}{Q}_{+}^{}{}_{}{}^{2}(1+k^2)\stackrel{~}{C}_{1}^{}{}_{}{}^{2}\stackrel{~}{U}^2\stackrel{~}{C}_{2}^{}{}_{}{}^{2}\right),`$ (68)
while the constraint becomes
$`0`$ $`=`$ $`\gamma u\mathrm{\Omega }_t2\left[1+(\gamma 1)u^2\right](\stackrel{~}{Q}_++k\stackrel{~}{U})\stackrel{~}{C}_1.`$ (69)
By defining a new independent variable
$`=`$ $`{\displaystyle \frac{d}{d\xi }}={\displaystyle \frac{\sqrt{3}}{\stackrel{~}{Y}}}{\displaystyle \frac{d}{dx}},`$ (70)
the evolution equation for $`\stackrel{~}{Y}`$
$`\stackrel{~}{Y}^{}`$ $`=`$ $`\{\stackrel{~}{Q}_0[1+\stackrel{~}{Q}_{+}^{}{}_{}{}^{2}+\stackrel{~}{Q}_0\stackrel{~}{Q}_+(1+k^2)\stackrel{~}{C}_{1}^{}{}_{}{}^{2}+\stackrel{~}{U}^2`$ (71)
$`{\displaystyle \frac{(\gamma 1)(1u^2)}{1+(\gamma 1)u^2}}\mathrm{\Omega }_t]+{\displaystyle \frac{1}{k}}\stackrel{~}{U}(1\stackrel{~}{Q}_{0}^{}{}_{}{}^{2})\}\stackrel{~}{Y}`$
decouples, and we obtain the following reduced set of evolution equations:
$`\stackrel{~}{Q}_{0}^{}{}_{}{}^{}`$ $`=`$ $`(1\stackrel{~}{Q}_{0}^{}{}_{}{}^{2})[1+\stackrel{~}{Q}_{+}^{}{}_{}{}^{2}+\stackrel{~}{Q}_0\stackrel{~}{Q}_+(1+k^2)\stackrel{~}{C}_{1}^{}{}_{}{}^{2}+\stackrel{~}{U}^2{\displaystyle \frac{1}{k}}\stackrel{~}{Q}_0\stackrel{~}{U}`$ (72)
$`{\displaystyle \frac{(\gamma 1)(1u^2)}{1+(\gamma 1)u^2}}\mathrm{\Omega }_t],`$
$`\stackrel{~}{Q}_{+}^{}{}_{}{}^{}`$ $`=`$ $`\stackrel{~}{Q}_+\{(\stackrel{~}{Q}_0+\stackrel{~}{Q}_+)(1\stackrel{~}{Q}_0\stackrel{~}{Q}_+){\displaystyle \frac{1}{k}}\stackrel{~}{U}(1\stackrel{~}{Q}_{0}^{}{}_{}{}^{2})`$ (73)
$`+\stackrel{~}{Q}_0[(1+k^2)\stackrel{~}{C}_{1}^{}{}_{}{}^{2}\stackrel{~}{U}^2+{\displaystyle \frac{(\gamma 1)(1u^2)}{1+(\gamma 1)u^2}}\mathrm{\Omega }_t]\}`$
$`\left[2k^2\stackrel{~}{C}_{1}^{}{}_{}{}^{2}(1\stackrel{~}{Q}_{0}^{}{}_{}{}^{2})+{\displaystyle \frac{1}{2}}{\displaystyle \frac{(3\gamma 2)+(2\gamma )u^2}{1+(\gamma 1)u^2}}\mathrm{\Omega }_t\right],`$
$`\stackrel{~}{C}_{1}^{}{}_{}{}^{}`$ $`=`$ $`\stackrel{~}{C}_1\{\stackrel{~}{Q}_0+\stackrel{~}{Q}_++\stackrel{~}{Q}_0[\stackrel{~}{Q}_{+}^{}{}_{}{}^{2}+\stackrel{~}{Q}_0\stackrel{~}{Q}_+(1+k^2)\stackrel{~}{C}_{1}^{}{}_{}{}^{2}+\stackrel{~}{U}^2`$ (74)
$`{\displaystyle \frac{(\gamma 1)(1u^2)}{1+(\gamma 1)u^2}}\mathrm{\Omega }_t]+{\displaystyle \frac{1}{k}}\stackrel{~}{U}(1\stackrel{~}{Q}_{0}^{}{}_{}{}^{2})\},`$
$`u^{}`$ $`=`$ $`{\displaystyle \frac{1u^2}{\gamma \left[u^2(\gamma 1)\right]}}\{\gamma [2(\gamma 1)\stackrel{~}{Q}_0+\gamma \stackrel{~}{Q}_+]u`$ (75)
$`+[(\gamma 1)(3\gamma 2)(2\gamma )u^2]\stackrel{~}{C}_1\},`$
$`\stackrel{~}{U}^{}`$ $`=`$ $`\stackrel{~}{U}\{(\stackrel{~}{Q}_0+\stackrel{~}{Q}_+)+\stackrel{~}{Q}_0[\stackrel{~}{Q}_{+}^{}{}_{}{}^{2}+\stackrel{~}{Q}_0\stackrel{~}{Q}_+(1+k^2)\stackrel{~}{C}_{1}^{}{}_{}{}^{2}+\stackrel{~}{U}^2`$ (76)
$`{\displaystyle \frac{(\gamma 1)(1u^2)}{1+(\gamma 1)u^2}}\mathrm{\Omega }_t]\}+2k\stackrel{~}{C}_1^2{\displaystyle \frac{1}{k}}(1\stackrel{~}{Q}_{0}^{}{}_{}{}^{2})(1\stackrel{~}{U}^2),`$
$`\stackrel{~}{C}_{2}^{}{}_{}{}^{}`$ $`=`$ $`\stackrel{~}{C}_2\{\stackrel{~}{Q}_++\stackrel{~}{Q}_0[\stackrel{~}{Q}_{+}^{}{}_{}{}^{2}+\stackrel{~}{Q}_0\stackrel{~}{Q}_+(1+k^2)\stackrel{~}{C}_{1}^{}{}_{}{}^{2}+\stackrel{~}{U}^2`$ (77)
$`{\displaystyle \frac{(\gamma 1)(1u^2)}{1+(\gamma 1)u^2}}\mathrm{\Omega }_t]+{\displaystyle \frac{1}{k}}\stackrel{~}{U}(1\stackrel{~}{Q}_{0}^{}{}_{}{}^{2})\}.`$
This system is invariant under the transformation
$$(\xi ,\stackrel{~}{Q}_0,\stackrel{~}{Q}_+,\stackrel{~}{C}_1,u,\stackrel{~}{U},\stackrel{~}{C}_2)(\xi ,\stackrel{~}{Q}_0,\stackrel{~}{Q}_+,\stackrel{~}{C}_1,u,\stackrel{~}{U},\stackrel{~}{C}_2).$$
(78)
Furthermore, noting the invariance under $`(\stackrel{~}{C}_1,u)(\stackrel{~}{C}_1,u)`$ and under $`\stackrel{~}{C}_2\stackrel{~}{C}_2`$, we can without loss of generality restrict the analysis to $`\stackrel{~}{C}_10,\stackrel{~}{C}_20`$.
Note that the denominator of the evolution equation for $`u`$ (75) is zero when $`u=\pm \sqrt{\gamma 1}`$. The only way to pass this sonic hypersurface without introducing a shock wave is when the numerator also is zero. This defines a submanifold of the sonic hypersurface, and this submanifold is only physical for the $`u=\sqrt{\gamma 1}`$ case. This severely restricts the global dynamics .
### 5.1 Invariant submanifolds
A number of invariant submanifolds can be identified:
* Massless scalar field: $`V=0`$, which implies $`\stackrel{~}{Q}_0=\pm 1`$.
* Static: $`\stackrel{~}{C}_1=0,u=0`$ (see ).
* Plane-symmetric: $`\stackrel{~}{C}_2=0`$ (see ).
* No perfect fluid: $`\mathrm{\Omega }_t=0`$ (and $`u`$ decouples).
* No scalar field: $`\stackrel{~}{U}=0`$, $`\stackrel{~}{Q}_0=\pm 1`$ and $`k=0`$.
The global dynamics of the submanifold with no scalar field has been studied previously in , where state-space diagrams can be found.
### 5.2 Monotone function
As for the SSS case, it is possible to generalise the monotone function for the perfect fluid TSS case to the case of a perfect fluid with a scalar field. This is done by replacing $`\overline{\mathrm{\Sigma }}_+`$ with $`\stackrel{~}{Q}_++k\stackrel{~}{U}`$ in equation (32) of . Thus,
$`\stackrel{~}{M}`$ $`=`$ $`(\stackrel{~}{Q}_++k\stackrel{~}{U})^{3\gamma 4}\stackrel{~}{C}_{1}^{}{}_{}{}^{\gamma }u^{2(\gamma 1)}(1u^2)^{(2\gamma )}\stackrel{~}{C}_{2}^{}{}_{}{}^{2(2\gamma )},`$ (79)
with
$$\stackrel{~}{M}^{}=\left[\frac{(3\gamma 2)(2\gamma )}{\gamma u}(1u^2)\stackrel{~}{C}_1\right]\stackrel{~}{M},$$
(80)
is monotonic in the regions $`u>0`$ and $`u<0`$. Furthermore,
$$u^{}|_{u=0}=\frac{3\gamma 2}{\gamma }\stackrel{~}{C}_1,$$
(81)
which is strictly negative. Consequently, as in the SSS case, there can be no closed or recurrent orbits in the interior of the state space.
## 6 Equilibrium points for the timelike self-similar case
We shall display all of the equilibrium points below along with their eigenvalues. We will not present the corresponding eigenvectors explicitly. In what follows, $`ϵ=\pm 1`$ is the sign of $`\stackrel{~}{Q}_0`$, which indicates whether the corresponding solution is expanding $`(+)`$ or contracting $`()`$. The order of the dependent variables is $`(\stackrel{~}{Q}_0,\stackrel{~}{Q}_+,\stackrel{~}{C}_1,u,\stackrel{~}{U},\stackrel{~}{C}_2)`$. The ’$`\pm `$’ suffices on the labels for equilibrium points correspond to the sign of $`\stackrel{~}{Q}_0`$ (i.e. the value of $`ϵ`$). The quantity $`\stackrel{~}{\mathrm{\Omega }_\varphi }=\stackrel{~}{U}^2+\stackrel{~}{Q}_{0}^{}{}_{}{}^{2}1`$ indicates the presence of a non-zero scalar field. Equilibrium points that act as attractors are listed in table 2.
### 6.1 No scalar field ($`\stackrel{~}{U}=0`$, $`V=0`$)
The state space contains a number of solutions with no scalar field, as presented below. There is also a solution with variable values $`(ϵ,\frac{3\gamma 2}{4(\gamma 1)}ϵ,0,0,0,0)`$, $`\mathrm{\Omega }_t=\frac{(2\gamma )(7\gamma 6)}{16(\gamma 1)^3}`$. As this solution is physical only when $`6/7<\gamma <1`$, we will not consider it further.
#### 6.1.1 K-points <br>
These are special points on the K-rings, defined in section 6.2.1. They all have $`\stackrel{~}{Q}_0=\pm 1`$, $`\stackrel{~}{Q}_+=\pm 1`$, and all other variables equal to zero.
#### 6.1.2 Static solution <br>
$`{}_{\pm }{}^{}\mathrm{T}`$: $`(ϵ,2\frac{\gamma 1}{3\gamma 2}ϵ,0,0,0,\frac{\sqrt{\gamma ^2+4(\gamma 1)}}{3\gamma 2})`$.
$`\mathrm{\Omega }_t=\frac{4(\gamma 1)}{(3\gamma 2)^2}`$, $`\stackrel{~}{\mathrm{\Omega }_\varphi }=0`$.
Eigenvalues ($`\stackrel{~}{C}_1`$ eliminated):
$`ϵ,{\displaystyle \frac{2\gamma }{3\gamma 2}}ϵ,{\displaystyle \frac{2\gamma }{3\gamma 2}}ϵ,{\displaystyle \frac{1}{2}}ϵ\pm {\displaystyle \frac{\sqrt{\gamma ^244\gamma +36}}{2(3\gamma 2)}}ϵ.`$
These points are always saddles.
#### 6.1.3 Regular centre <br>
$`{}_{\pm }{}^{}\mathrm{C}_{}^{0}`$: $`(ϵ,0,0,0,0,1)`$.
$`\mathrm{\Omega }_t=0`$, $`\stackrel{~}{\mathrm{\Omega }_\varphi }=0`$.
Eigenvalues (constraint degenerate):
$`2ϵ,ϵ,ϵ,ϵ,2ϵ,2ϵ.`$
These points are always saddles.
### 6.2 Massless scalar field ($`\stackrel{~}{U}0`$, $`V=0`$)
#### 6.2.1 K-rings <br>
$`{}_{\pm }{}^{}\mathrm{K}`$: $`(ϵ,\pm \sqrt{1\stackrel{~}{U}^2},0,0,\stackrel{~}{U},0)`$.
$`\mathrm{\Omega }_t=0`$, $`\stackrel{~}{\mathrm{\Omega }_\varphi }=\stackrel{~}{U}^2`$.
Eigenvalues ($`\stackrel{~}{C}_1`$ eliminated):
$`ϵ,2ϵ{\displaystyle \frac{\gamma }{\gamma 1}}\stackrel{~}{Q}_+,4ϵ+2\stackrel{~}{Q}_+{\displaystyle \frac{2}{k}}\stackrel{~}{U},0,4ϵ+{\displaystyle \frac{3\gamma 2}{\gamma 1}}\stackrel{~}{Q}_+.`$
Each K-ring corresponds to a one-parameter family of equilibrium points (and hence gives rise to a zero eigenvalue). They are analogues of the Kasner solutions in the case with no scalar field . For each K-ring, there is a subset of future or past attractors.
$`{}_{+}{}^{}\mathrm{K}`$: sources and saddles.
$`{}_{}{}^{}\mathrm{K}`$: sinks and saddles.
#### 6.2.2 M-points <br>
$`{}_{\pm }{}^{}\mathrm{M}_{}^{\stackrel{~}{u}}`$: $`(ϵ,k^2fϵ,f,\stackrel{~}{u}ϵ,kfϵ,0)`$, $`f=1/(1+k^2)`$, $`\gamma >\frac{6}{5}`$,
$`\stackrel{~}{u}_{1,2}=\frac{\gamma (\gamma 1)\frac{2\gamma }{2}\gamma k^2\pm \sqrt{(\gamma 1)(2\gamma )(3\gamma 2)+\gamma ^2\left[(\gamma 1)\frac{2\gamma }{2}\gamma k^2\right]^2}}{2\gamma }`$.
$`\mathrm{\Omega }_t=0`$, $`\stackrel{~}{\mathrm{\Omega }_\varphi }=f^2k^2`$.
Eigenvalues ($`\stackrel{~}{C}_1`$ eliminated):
$`2fϵ,(1k^2)fϵ,2(1k^2)fϵ,F_6(\gamma ,k,u)ϵ,`$
$`{\displaystyle \frac{1}{\gamma u}}\left[3\gamma 2+(2\gamma )u^2+2\gamma u\right]fϵ.`$
These equilibrium points are related to the Milne points $`\stackrel{~}{\mathrm{M}}`$ in . They are only physical ($`|\stackrel{~}{u}|<1`$) for certain ranges of $`\gamma `$ and $`k`$. For instance, when $`k>1`$ it follows that $`|\stackrel{~}{u}_2|>1`$. When $`k>1`$, these points are saddles. Furthermore, for $`u=\stackrel{~}{u}_2`$ they are saddles even when $`k<1`$. Examining the eigenvalues numerically for $`0<k<1`$, there are values of $`\gamma `$ and $`k`$ for which the $`u=\stackrel{~}{u}_1`$ points act as attractors. However, this only occurs when $`\stackrel{~}{u}_1>\sqrt{\gamma 1}`$, i.e., when the corresponding equilibrium point is beyond the sonic hypersurface located at $`u=\sqrt{\gamma 1}`$. As it is impossible to cross this sonic hypersurface in a regular way, the $`{}_{+}{}^{}\stackrel{~}{M}`$ points will not affect the dynamics of the models we are interested in, even though the points are attractors for some values of $`\gamma `$ and $`k`$. This situation is also present in the case without a scalar field .
#### 6.2.3 Curvature-scaling solutions <br>
$`{}_{\pm }{}^{}\mathrm{X}_{}^{\widehat{u}}`$: $`(ϵ,\frac{1}{2}ϵ,\frac{1}{2k},\widehat{u}ϵ,\frac{1}{2k}ϵ,\frac{1}{\sqrt{2}k}\sqrt{k^21})`$, $`k>1`$ ($`\kappa ^2<2`$),
$`\widehat{u}_{1,2}=\frac{\frac{3\gamma 4}{2}\gamma k\pm \sqrt{(\gamma 1)(2\gamma )(3\gamma 2)+\left(\frac{3\gamma 4}{2}\right)^2\gamma ^2k^2}}{2\gamma }`$.
$`\mathrm{\Omega }_t=0`$, $`\stackrel{~}{\mathrm{\Omega }_\varphi }=\frac{1}{4k^2}`$.
Eigenvalues ($`\stackrel{~}{C}_1`$ eliminated):
$`{\displaystyle \frac{k^21}{k^2}}ϵ,{\displaystyle \frac{1}{2k}}\left[k\pm \sqrt{43k^2}\right]ϵ,F_7(\gamma ,k,u),`$
$`\left[1+{\displaystyle \frac{3\gamma 2+(2\gamma )u^2}{2\gamma ku}}\right]ϵ.`$
For $`k=1`$ ($`\kappa ^2=2`$) these points coincide with $`{}_{\pm }{}^{}\mathrm{M}_{}^{\stackrel{~}{u}}`$, and for $`k<1`$ ($`\kappa ^2>2`$) they are unphysical. Noting that $`k>1`$, these points are always saddles.
#### 6.2.4 Equilibrium lines with arbitrary $`u`$ <br>
$`{}_{\pm }{}^{}\mathrm{\Phi }_{}^{u}`$: $`(ϵ,2\frac{\gamma 1}{\gamma }ϵ,0,u,\pm \frac{1}{\gamma }\sqrt{(2\gamma )(3\gamma 2)},0)`$.
$`\mathrm{\Omega }_t=0`$, $`\stackrel{~}{\mathrm{\Omega }_\varphi }=(2\gamma )(3\gamma 2)/\gamma ^2`$.
Eigenvalues ($`\stackrel{~}{Q}_+`$ eliminated):
$`0,0,ϵ,{\displaystyle \frac{2}{\gamma }}(2\gamma )ϵ,{\displaystyle \frac{4}{\gamma }}ϵ{\displaystyle \frac{2\stackrel{~}{U}}{k}}.`$
There are two zero eigenvalues for these points. The first zero eigenvalue corresponds to the fact that we have a line of equilibrium points. The second zero eigenvalue indicates that the equilibria are non-hyperbolic. For $`u=0,\pm 1`$, these equilibrium lines coincide with the various K-rings (see subsections 6.2.1 and 7.2.1) and these exceptional points mark where the K-rings change stability. The higher-order zero eigenvalue of $`{}_{\pm }{}^{}\mathrm{\Phi }_{}^{u}`$ corresponds to the one that indicates that $`{}_{\pm }{}^{}\mathrm{K}`$ is a line of equilibrium points (and not to the one that becomes zero due to the stability change of $`{}_{\pm }{}^{}\mathrm{K}`$). The corresponding eigenvector is $`\stackrel{}{v}=\frac{\stackrel{~}{U}}{\stackrel{~}{Q}_+}\stackrel{}{e}_{\stackrel{~}{Q}_+}+\stackrel{}{e}_{\stackrel{~}{U}}`$. Perturbing the equilibrium lines $`{}_{\pm }{}^{}\mathrm{\Phi }_{}^{u}`$ along this eigenvector, we find that
$`\stackrel{~}{Q}_{+}^{}{}_{}{}^{}`$ $`=`$ $`2(1\stackrel{~}{Q}_{+}^{}{}_{}{}^{2}\stackrel{~}{U}^2)ϵ\stackrel{~}{Q}_+,`$ (82)
$`\stackrel{~}{U}^{}`$ $`=`$ $`2(1\stackrel{~}{Q}_{+}^{}{}_{}{}^{2}\stackrel{~}{U}^2)ϵ\stackrel{~}{U}.`$ (83)
This is precisely the dynamical system restricted to the invariant set $`\stackrel{~}{Q}_0=ϵ`$, $`\stackrel{~}{C}_1=0`$, $`\stackrel{~}{C}_2=0`$, $`\mathrm{\Omega }_t=0`$. We can explicitly integrate equations (82) and (83). It follows that $`\stackrel{~}{Q}_+`$ is proportional to $`\stackrel{~}{U}`$, and the orbits in the ($`\stackrel{~}{Q}_+,\stackrel{~}{U}`$) plane consist of straight lines through the origin with additional equilibrium points at $`{}_{\pm }{}^{}\mathrm{\Phi }_{}^{u}`$ (where $`\stackrel{~}{Q}_+=2\frac{\gamma 1}{\gamma }\stackrel{~}{Q}_0`$), which are thus non-linear saddles.
### 6.3 Scalar field with potential ($`\stackrel{~}{U}0`$, $`V0`$)
There is a number of solutions with non-zero potential listed below. There are also equilibrium points $`{}_{\pm }{}^{}\mathrm{\Xi }`$ with variable values $`(\frac{1}{h}ϵ,k^2hϵ,0,0,khϵ,\sqrt{1k^2})`$, $`h=1/\sqrt{1+k^2}`$, $`k^2<1`$ ($`\kappa ^2>2`$), but these points are unphysical since $`|\stackrel{~}{Q}_0|>1`$ when $`k>0`$.
#### 6.3.1 Scalar-field dominated solutions <br>
$`{}_{\pm }{}^{}\mathrm{\Phi }`$: $`(2khϵ,khϵ,0,0,hϵ,0)`$, $`k^2<\frac{1}{3}`$ ($`\kappa ^2>6`$), $`h=1/\sqrt{1+k^2}`$.
$`\mathrm{\Omega }_t=0`$, $`\stackrel{~}{\mathrm{\Omega }_\varphi }=(23k^2)h^2`$.
Eigenvalues (constraint degenerate):
$`{\displaystyle \frac{3\gamma 4}{\gamma 1}}khϵ,{\displaystyle \frac{13k^2}{k}}hϵ,{\displaystyle \frac{13k^2}{k}}hϵ,{\displaystyle \frac{1k^2}{k}}hϵ,{\displaystyle \frac{1k^2}{k}}hϵ,`$
$`{\displaystyle \frac{2(\gamma 1)\gamma k^2}{(\gamma 1)k}}hϵ.`$
As in the timelike case, since the constraint is degenerate we must retain all six eigenvalues. For $`k=1/3`$ ($`\kappa ^2=6`$), these points coincide with points in the K-rings, and for $`k^2>\frac{1}{3}`$ ($`\kappa ^2<6`$) they are unphysical.
$`{}_{+}{}^{}\mathrm{\Phi }`$: source when $`\gamma <4/3`$ and $`k^2<2\frac{\gamma 1}{\gamma }`$ ($`\kappa ^2>\frac{\gamma }{\gamma 1}`$); saddle otherwise.
$`{}_{}{}^{}\mathrm{\Phi }`$: sink when $`\gamma <4/3`$ and $`k^2<2\frac{\gamma 1}{\gamma }`$ ($`\kappa ^2>\frac{\gamma }{\gamma 1}`$); saddle otherwise.
#### 6.3.2 Potential-dominated solutions <br>
$`{}_{\pm }{}^{}\mathrm{Z}_{}^{u^{}}`$: $`(\sqrt{\frac{1k^2}{1+k^2}}ϵ,0,\frac{1}{\sqrt{1+k^2}},u^{},0,0)`$, $`k<1`$ $`(\kappa ^2>2)`$,
$`u_{1,2}^{}=\frac{(\gamma 1)\gamma \sqrt{1k^2}ϵ\pm \sqrt{(\gamma 1)(2\gamma )(3\gamma 2)+(\gamma 1)^2\gamma ^2(1k^2)}}{2\gamma }`$.
$`\mathrm{\Omega }_t=0`$, $`\stackrel{~}{\mathrm{\Omega }_\varphi }=\frac{2k^2}{1+k^2}`$.
Eigenvalues ($`\stackrel{~}{C}_1`$ eliminated):
$`\stackrel{~}{Q}_0,(1\pm \sqrt{5})\stackrel{~}{Q}_0,F_8(\gamma ,k),`$
$`{\displaystyle \frac{2(1u^2)+\gamma (u^22\sqrt{1+k^2}\stackrel{~}{Q}_0u3)}{\gamma u\sqrt{1+k^2}}}.`$
For these solutions, the potential is non-zero, since $`|\stackrel{~}{Q}_0|<1`$. In the context of this paper, these solutions may be unphysical. They are on the boundary and correspond to non-self-similar solutions with a cosmological constant. They are always saddles.
## 7 Equilibrium points at extreme tilt
In addition to the above equilibrium points, there is a number of equilibrium points for which the tilt is extreme, i.e., $`v=\pm 1`$ or $`u=\pm 1`$. These are artifacts of the particular approach that we have adopted, and signify that the coordinates break down. These points are still important since orbits that are asymptotic to them may pass between the spacelike and the timelike self-similar regions (at least at non-vacuum equilibrium points). Furthermore, for submanifolds where the tilt variable is not specified (e.g. the fluid vacuum submanifold), some of these solutions are indistinguishable from similar ones with non-extreme tilt. Equilibrium points in this class that act as attractors are listed in table 3. In what follows, the equilibrium points will be given both in the SSS variables $`(\overline{Q}_0,\overline{Q}_+,\overline{C}_1,v,\overline{U},\overline{W})`$ and in the TSS variables $`(\stackrel{~}{Q}_0,\stackrel{~}{Q}_+,\stackrel{~}{C}_1,u,\stackrel{~}{U},\stackrel{~}{C}_2)`$.
### 7.1 No scalar field
#### 7.1.1 K-points <br>
These are special points on the K-rings, defined in section 7.2.1. They all have $`\overline{Q}_0=\stackrel{~}{Q}_0=\pm 1`$, $`\overline{Q}_+=\stackrel{~}{Q}_+=\pm 1`$, $`v=u=\pm 1`$, and all other variables equal to zero.
#### 7.1.2 C points <br>
In the TSS case there are equilibrium points that resemble the regular centre $`{}_{\pm }{}^{}\mathrm{C}_{}^{0}`$, but have $`u=\pm 1`$.
$`{}_{\pm }{}^{}\mathrm{C}_{}^{\pm }`$: $`(ϵ,0,0,\pm 1,0,1)`$ (TSS).
$`\mathrm{\Omega }_t=0`$, $`\stackrel{~}{\mathrm{\Omega }_\varphi }=0`$.
Eigenvalues ($`\stackrel{~}{C}_2`$ eliminated):
$`ϵ,ϵ,4{\displaystyle \frac{\gamma 1}{2\gamma }}ϵ,{\displaystyle \frac{ϵ\pm 3}{2}}.`$
These points are always saddles.
### 7.2 Massless scalar field
#### 7.2.1 K-rings <br>
$`{}_{+}{}^{}\mathrm{K}_{}^{\pm }`$, $`{}_{}{}^{}\mathrm{K}_{}^{\pm }`$: $`(ϵ,\pm \sqrt{1U^2},0,\pm 1,U,0)`$ (SSS and TSS).
$`\mathrm{\Omega }=0`$, $`\mathrm{\Omega }_\varphi =U^2`$.
Eigenvalues (SSS: $`\overline{C}_1`$ eliminated, TSS: $`\stackrel{~}{C}_1`$ eliminated):
$`\mathrm{SSS}`$ $`:`$ $`2ϵ,2(ϵ+\overline{Q}_+),2{\displaystyle \frac{2(\gamma 1)ϵ+\gamma \overline{Q}_+}{2\gamma }},`$
$`0,2ϵ+\overline{Q}_+{\displaystyle \frac{\overline{U}}{k}}.`$
$`\mathrm{TSS}`$ $`:`$ $`ϵ,2(ϵ+\stackrel{~}{Q}_+),2{\displaystyle \frac{2(\gamma 1)ϵ+\gamma \stackrel{~}{Q}_+}{2\gamma }},`$
$`0,2\left(2ϵ+\stackrel{~}{Q}_+{\displaystyle \frac{\stackrel{~}{U}}{k}}\right).`$
For each K-ring, there is a subset of future or past attractors.
$`{}_{+}{}^{}\mathrm{K}_{}^{\pm }`$: sources and saddles.
$`{}_{}{}^{}\mathrm{K}_{}^{\pm }`$: sinks and saddles.
#### 7.2.2 M-points <br>
a) $`{}_{\pm }{}^{}\mathrm{M}_{}^{\pm }`$: $`(ϵ,k^2fϵ,f,ϵ,kfϵ,0)`$, $`f=1/(1+k^2)`$ (SSS and TSS).
$`\mathrm{\Omega }=0`$, $`\mathrm{\Omega }_\varphi =\left(\frac{k}{1+k^2}\right)^2`$.
Eigenvalues (SSS: $`\overline{C}_1`$ eliminated, TSS: $`\stackrel{~}{C}_1`$ eliminated):
$`2fϵ,4fϵ,2{\displaystyle \frac{(5+k^2)\gamma 2(3+k^2)}{2\gamma }}fϵ,`$
$`2(1k^2)fϵ,(1k^2)fϵ.`$
$`{}_{+}{}^{}\mathrm{M}_{}^{+}`$: sink when $`\gamma >\frac{2(3+k^2)}{5+k^2}`$, $`k<1`$ ($`\kappa ^2>2`$); saddle otherwise.
$`{}_{}{}^{}\mathrm{M}_{}^{}`$: source when $`\gamma >\frac{2(3+k^2)}{5+k^2}`$, $`k<1`$ ($`\kappa ^2>2`$); saddle otherwise.
b) $`{}_{\pm }{}^{}\mathrm{M}_{}^{}`$: $`(ϵ,k^2fϵ,f,ϵ,kfϵ,0)`$, $`f=1/(1+k^2)`$ (SSS and TSS).
$`\mathrm{\Omega }=0`$, $`\mathrm{\Omega }_\varphi =\left(\frac{k}{1+k^2}\right)^2`$.
Eigenvalues (SSS: $`\overline{C}_1`$ eliminated, TSS: $`\stackrel{~}{C}_1`$ eliminated):
$`2fϵ,(1k^2)fϵ,2(1k^2)fϵ,2(1k^2)fϵ,0.`$
The zero eigenvalue is due to the fact that these points are the end points of the equilibrium lines $`{}_{\pm }{}^{}\mathrm{H}`$ (see section 7.2.3).
$`{}_{+}{}^{}\mathrm{M}_{}^{}`$: sink when $`k<1`$ ($`\kappa ^2>2`$); saddle otherwise.
$`{}_{}{}^{}\mathrm{M}_{}^{+}`$: source when $`k<1`$ ($`\kappa ^2>2`$); saddle otherwise.
#### 7.2.3 H-lines <br>
$`{}_{\pm }{}^{}\mathrm{H}_{}^{}`$: $`(ϵ,Q_+,1+ϵQ_+,ϵ,k(1+ϵQ_+)ϵ,0)`$, $`1<ϵQ_+<\frac{k^2}{1+k^2}`$ (SSS and TSS).
$`\mathrm{\Omega }=2ϵ(1+ϵQ_+)\left[(1+k^2)Q_++ϵk^2\right]`$, $`\mathrm{\Omega }_\varphi =k^2(1+ϵQ_+)^2`$.
Eigenvalues (SSS: $`\overline{C}_1`$ eliminated, TSS: $`\stackrel{~}{C}_1`$ eliminated):
$`2(1+2ϵQ_+)ϵ,2(1+2ϵQ_+)ϵ,(1+2ϵQ_+)ϵ,`$
$`2(1+ϵQ_+)ϵ,0.`$
These equilibria consist of lines of equilibrium points. The eigenvector direction along the lines is $`\stackrel{}{v}=\stackrel{}{e}_{Q_+}+k\stackrel{}{e}_U`$. The end points of the lines are the M-points $`{}_{\pm }{}^{}\mathrm{M}_{}^{}`$ at one end and points of stability change on the K-rings $`{}_{\pm }{}^{}\mathrm{K}_{}^{}`$ at the other end.
$`{}_{+}{}^{}\mathrm{H}_{}^{}`$: always contains at least a subset of sinks. Solely sinks when $`k>1`$ ($`\kappa ^2<2`$); sources and saddles otherwise.
$`{}_{}{}^{}\mathrm{H}_{}^{+}`$: always contains at least a subset of sources. Solely sources when $`k>1`$ ($`\kappa ^2<2`$); sources and saddles otherwise.
#### 7.2.4 Curvature-scaling solutions <br>
$`{}_{+}{}^{}\mathrm{X}_{}^{\pm }`$, $`{}_{}{}^{}\mathrm{X}_{}^{\pm }`$: $`(2kgϵ,kgϵ,g,\pm 1,gϵ,0)`$,
$`g=1/(\sqrt{2}\sqrt{1+k^2})`$, $`k<1`$ ($`\kappa ^2>2`$) (SSS),
$`(ϵ,\frac{1}{2}ϵ,\frac{1}{2k},\pm 1,\frac{1}{2k}ϵ,\frac{1}{\sqrt{2}k}\sqrt{k^21})`$, $`k>1`$ ($`\kappa ^2<2`$) (TSS).
$`\mathrm{\Omega }=0`$, $`\mathrm{\Omega }_\varphi =\frac{1}{2(1+k^2)}`$.
Eigenvalues (SSS: $`\overline{C}_1`$ eliminated, TSS: $`\stackrel{~}{C}_1`$ eliminated):
$`\mathrm{SSS}`$ $`:`$ $`2{\displaystyle \frac{3\gamma 4}{2\gamma }}\left[\mathrm{sgn}\left(v\right)+k\right]gϵ,{\displaystyle \frac{1k^2}{k}}gϵ,F_9(k)ϵ,`$
$`F_{10}(k)ϵ,F_{11}(k,v),\mathrm{where}F_10(k)>0,\mathrm{and}F_{11}(k)>0.`$
$`\mathrm{TSS}`$ $`:`$ $`{\displaystyle \frac{1}{k^2}}(k^21)ϵ,{\displaystyle \frac{1}{k}}\left[kϵ+\mathrm{sgn}\left(v\right)\right],`$
$`{\displaystyle \frac{1}{k}}[kϵ+\mathrm{sgn}\left(v\right)]{\displaystyle \frac{(3\gamma 4)}{2\gamma }},{\displaystyle \frac{1}{2}}ϵ\pm {\displaystyle \frac{1}{2k}}\sqrt{43k^2}).`$
For $`k=1`$ ($`\kappa ^2=2`$) these points coincide with $`{}_{\pm }{}^{}\mathrm{M}_{}^{\pm }`$. They are always saddles.
### 7.3 Scalar field with potential ($`U0`$, $`W0`$)
#### 7.3.1 Scalar-field dominated solutions <br>
$`{}_{+}{}^{}\mathrm{\Phi }_{}^{\pm }`$, $`{}_{}{}^{}\mathrm{\Phi }_{}^{\pm }`$: $`(ϵ,\frac{1}{2}ϵ,0,\pm 1,\frac{1}{2k}ϵ,\frac{1}{2k}\sqrt{3k^21})`$, $`k^2>\frac{1}{3}`$ ($`\kappa ^2<6`$) (SSS),
$`(2khϵ,khϵ,0,\pm 1,hϵ,0)`$, $`h=1/\sqrt{1+k^2}`$, $`k^2<\frac{1}{3}`$ ($`\kappa ^2>6`$) (TSS).
$`\mathrm{\Omega }=0`$, $`\mathrm{\Omega }_\varphi =3/4`$.
Eigenvalues (SSS: $`\overline{Q}_+`$ eliminated, TSS: $`\stackrel{~}{Q}_+`$ eliminated):
$`\mathrm{SSS}`$ $`:`$ $`{\displaystyle \frac{k^21}{k^2}}ϵ,{\displaystyle \frac{k^21}{2k^2}}ϵ,{\displaystyle \frac{3\gamma 4}{2\gamma }}ϵ,2\overline{W}^2ϵ,2\overline{W}^2ϵ.`$
$`\mathrm{TSS}`$ $`:`$ $`{\displaystyle \frac{h}{k}}(13k^2),{\displaystyle \frac{h}{k}}(13k^2),{\displaystyle \frac{h}{k}}(1k^2),{\displaystyle \frac{h}{k}}(1k^2),`$
$`2{\displaystyle \frac{3\gamma 4}{2\gamma }}hk.`$
For $`k^2=1/3`$ ($`\kappa ^2=6`$) these points coincide with points on the K-rings $`{}_{\pm }{}^{}\mathrm{K}_{}^{\pm }`$.
SSS:
$`{}_{+}{}^{}\mathrm{\Phi }_{}^{\pm }`$: sinks when $`\gamma >\frac{4}{3}`$, $`k>1`$ ($`\kappa ^2<2`$); saddles otherwise.
$`{}_{}{}^{}\mathrm{\Phi }_{}^{\pm }`$: sources when $`\gamma >\frac{4}{3}`$, $`k>1`$ ($`\kappa ^2<2`$); saddles otherwise.
TSS:
$`{}_{+}{}^{}\mathrm{\Phi }_{}^{\pm }`$: sources when $`\gamma <\frac{4}{3}`$, $`k^2<1/3`$ ($`\kappa ^2>6`$); saddles otherwise.
$`{}_{}{}^{}\mathrm{\Phi }_{}^{\pm }`$: sinks when $`\gamma <\frac{4}{3}`$, $`k^2<1/3`$ ($`\kappa ^2>6`$); saddles otherwise.
#### 7.3.2 Curvature-scaling solutions <br>
$`{}_{+}{}^{}\mathrm{\Xi }_{}^{\pm }`$, $`{}_{}{}^{}\mathrm{\Xi }_{}^{\pm }`$: $`(\frac{1}{2kg}ϵ,kgϵ,0,\pm 1,gϵ,\frac{1}{\sqrt{2}})`$, $`g=1/(\sqrt{2}\sqrt{1+k^2})`$, $`k>1`$ ($`\kappa ^2<2`$) (SSS),
$`(\frac{1}{h}ϵ,k^2hϵ,0,\pm 1,khϵ,\sqrt{1k^2})`$, $`h=1/\sqrt{1+k^2}`$, $`k<1`$ ($`\kappa ^2>2`$) (TSS).
$`\mathrm{\Omega }=0`$, $`\mathrm{\Omega }_\varphi =\frac{2+k^2}{2(1+k^2)}`$.
Eigenvalues (SSS: $`\overline{Q}_+`$ eliminated, TSS: $`\stackrel{~}{Q}_+`$ eliminated):
$`\mathrm{SSS}`$ $`:`$ $`{\displaystyle \frac{\sqrt{k^2+1}\pm \sqrt{9k^27}}{2\sqrt{2}k}}ϵ,{\displaystyle \frac{1}{k}}(k^21)gϵ,{\displaystyle \frac{1}{2kg}}ϵ,`$
$`{\displaystyle \frac{2}{(2\gamma )k}}\left[2(\gamma 1)+(3\gamma 2)k^2\right]gϵ.`$
$`\mathrm{TSS}`$ $`:`$ $`{\displaystyle \frac{1}{h}}ϵ,(1k^2)hϵ,{\displaystyle \frac{1}{2h}}ϵ\pm {\displaystyle \frac{1}{2}}\sqrt{79k^2},`$
$`2{\displaystyle \frac{2(\gamma 1)(2\gamma )k^2}{2\gamma }}hϵ.`$
For $`k=1`$ ($`\kappa ^2=2`$) these points coincide with $`{}_{\pm }{}^{}\mathrm{\Phi }_{}^{\pm }`$. They are always saddles.
#### 7.3.3 Potential-dominated solutions <br>
$`{}_{\pm }{}^{}\mathrm{Z}_{}^{\pm }`$: $`(\sqrt{\frac{1k^2}{1+k^2}}ϵ,0,\frac{1}{\sqrt{1+k^2}},\pm 1,0,0)`$, $`k<1`$ $`(\kappa ^2>2)`$.
$`\mathrm{\Omega }_t=0`$, $`\mathrm{\Omega }_\varphi =\frac{2k^2}{1+k^2}`$.
Eigenvalues ($`\stackrel{~}{C}_1`$ eliminated):
$`\stackrel{~}{Q}_0,(1\pm \sqrt{5})\stackrel{~}{Q}_0,2{\displaystyle \frac{\sqrt{1+k^2}\stackrel{~}{Q}_0+\mathrm{sgn}\left(u\right)}{\sqrt{1+k^2}}},`$
$`2{\displaystyle \frac{(3\gamma 4)\mathrm{sgn}\left(u\right)+2(\gamma 1)\sqrt{1+k^2}\stackrel{~}{Q}_0}{(2\gamma )\sqrt{1+k^2}}}.`$
These solutions are only physical in the TSS region. They are always saddles.
## 8 Global results and discussion
Due to the existence of monotone functions and the fact that there are consequently no closed or periodic orbits in the physical state spaces we can obtain global results for the dynamics by studying the local stability of the equilibria.
Indeed, from the monotone functions obtained in the spatially self-similar (SSS) case (45) and the timelike self-similar (TSS) case (79) we can immediately deduce from the monotonicity principle that all orbits have $`\overline{Q}_{0}^{}{}_{}{}^{2}1`$, $`\overline{Q}_++k\overline{U}0`$ or $`\overline{C}_10`$ (or an extreme value for $`v`$) asymptotically in the SSS case (and similarly in the TSS case). Moreover, we can also see immediately that by setting the right-hand-sides of equations (38) and (41) to zero that either $`\overline{C}_1=0`$ or $`\overline{W}=0`$, or if both are non-zero then $`k\overline{Q}_++\overline{U}=0`$; this latter case yields very severe constraints on any possible equilibrium points. In fact, from the local analysis of the equilibria we can determine all of the sinks and sources. In both the SSS and TSS cases a set of massless scalar field models lying on the $`{}_{+}{}^{}\mathrm{K}`$-ring act as sources (i.e., early-time attractors) and a set of massless scalar field models lying on the <sub>-</sub>K-ring act as sinks (i.e., late-time attractors), and for certain ranges of the parameters (e.g., $`\kappa ^2<2`$) the equilibrium point $`{}_{+}{}^{}\mathrm{\Phi }`$ with $`U0`$ and $`V0`$, corresponding to the ever-expanding inflationary flat FRW power-law solution , are sinks (i.e., late-time attractors). \[We note that the equilibrium point $`{}_{}{}^{}\mathrm{\Phi }`$ corresponding to $`ϵ=1`$, which acts as a source, represents an ever-contracting solution and therefore is of less physical importance, although it does serve to classify all of the possible orbits in the state space.\] Hence we have the global results that models that are initially expanding always expand from an initial singularity and always recollapse to a second singularity (when $`\kappa ^2>2`$) or either recollapse or expand forever towards a flat FRW power-law solution (for $`\kappa ^2<2`$). This global behaviour is the same as that for positive curvature FRW models and Kantowski-Sachs models and for Bianchi type IX models . Models that expand from an initial singularity and recollapse to a second singularity are said to satisfy the positive-curvature recollapse property . Models that expand towards the flat FRW power-law solution isotropise and inflate to the future and are said to satisfy the cosmic no-hair theorem . The time-reverse of the above dynamics is also possible (essentially $`ϵϵ`$; although we have included this in the analysis these models are of less interest physically). Solutions in which the shear and the kinetic energy of the scalar field dominate are analogues of the Kasner and Jacobs solutions , and a rigorous study of the structure of the singularity, which is non-oscillatory, for a general class of analytic solutions of the Einstein field equations coupled to a massless scalar field has recently been presented .
However, there are some aspects of the global dynamics of the self-similar, spherically symmetric models that are different. The complete set of attractors for different values of the parameters $`\gamma `$ and $`\kappa `$ are summarised in table 4. Some of these differences are quite subtle. First, we note that the flat FRW power-law inflationary solution corresponds both to a set of non-tilted equilibrium points $`{}_{\pm }{}^{}\mathrm{\Phi }`$ and to points at extreme tilt $`{}_{\pm }{}^{}\mathrm{\Phi }_{}^{\pm }`$. There is a bifurcation of the equation-of-state parameter at $`\gamma =4/3`$ in that the power-law inflationary solution $`\mathrm{\Phi }`$ is a non-tilted attractor $`{}_{\pm }{}^{}\mathrm{\Phi }`$ for $`\gamma <4/3`$ and an extreme-tilt attractor $`{}_{\pm }{}^{}\mathrm{\Phi }_{}^{\pm }`$ when $`\gamma >4/3`$. For $`\gamma =4/3`$ there exist lines of equilibrium points with arbitrary tilt $`v`$ (or $`u`$ in the TSS case). This type of $`\gamma `$-dependent behaviour has also been found in Bianchi type V two-fluid models . Second, there exist additional M-point attractors (for $`\gamma >\frac{2(3\kappa ^2+2)}{5\kappa ^2+2}`$, $`\kappa ^2>2`$) at extreme tilt. The significance of these is less clear, although they are important for the matching of orbits and they are related to critical phenomena . We shall discuss this further in sections 9 and 10. We recall that solutions in the SSS region and the TSS region can be matched across the surface of extreme tilt via the equilibrium points with extreme tilt (see earlier and ). In addition, a comprehensive analysis of the matching of solutions would be necessary in order to obtain a complete knowledge of the intermediate dynamics of the models. Clearly the intermediate behaviour of the models under investigation will be quite different to that of the models previously studied.
We note that all of the equilibrium points with non-negligible matter, namely the non-vacuum Flat Friedmann $`{}_{\pm }{}^{}\mathrm{F}`$ and the Friedmann Scaling $`{}_{\pm }{}^{}\mathrm{FS}`$ equilibrium points, are saddles. This means that the perfect fluid is not dynamically important asymptotically. In order to understand the asymptotic behaviour of the models we consequently need only study the vacuum models (i.e., the invariant boundary with $`\mathrm{\Omega }_n=0`$ in the SSS case and the invariant boundary with $`\mathrm{\Omega }_t=0`$ in the TSS case). We shall discuss the fluid vacuum models further in section 9. The matter will play an important role in describing the dynamics of the models at intermediate times, and hence the physical properties of the models. We shall illustrate some of the intermediate dynamics in the next two sections. However, in the SSS case we know from the behaviour of the monotone function that the subcase $`\overline{C}_1=0`$ is important asymptotically (in the TSS case the analogous case is $`\stackrel{~}{C}_1=0`$, leading to the static models ). Moreover, when $`\mathrm{\Omega }_n=0`$, the constraint (33) leads to $`(\overline{Q}_++k\overline{U})\overline{C}_1=0`$. Clearly the invariant set $`\overline{C}_1=0`$ (and $`v=0`$), corresponding to the (non-self-similar) Kantowski-Sachs models, is of vital importance, and knowledge of the dynamics of the Kantowski-Sachs models is crucial for a complete understanding of the dynamics of the models under consideration here. In addition, all of the interesting transient dynamics with non-negligible matter (e.g., the non-vacuum Flat Friedmann and the Friedmann Scaling saddle equilibrium points, as well as the power-law attractors) occurs in the Kantowski-Sachs invariant submanifold. Consequently, we shall study the Kantowski-Sachs models in more detail elsewhere .
We note that a set of massless scalar field models lying on the K-rings act as sources and sinks (i.e., early- and late-time attractors). It is therefore also of interest to study the self-similar, spherically symmetric massless scalar field models more fully. Indeed, such a study will also be of relevance in the study of critical phenomena (see section 10). We shall return to this in future work.
## 9 Fluid vacuum
When there is no barotropic fluid present ($`\mathrm{\Omega }_n=0`$ and $`\mathrm{\Omega }_t=0`$, respectively), the constraint gives rise to two separate invariant submanifolds: either $`\overline{Q}_+=k\overline{U}`$ ($`\stackrel{~}{Q}_+=k\stackrel{~}{U}`$), or else $`\overline{C}_1=0`$ ($`\stackrel{~}{C}_1=0`$). The $`\overline{Q}_+=k\overline{U}`$ ($`\stackrel{~}{Q}_+=k\stackrel{~}{U}`$) submanifold is particularly interesting, as it contains all the sinks and sources of the more general models under consideration. Furthermore, this submanifold is three-dimensional, and hence lends itself to visual presentation. In addition, the submanifold $`\overline{C}_1=0`$ will be studied in detail in .
### 9.1 Spatially self-similar case
In the SSS case, the reduced dynamical system becomes (eliminating the variable $`\overline{C}_1`$):
$`\overline{Q}_{0}^{}{}_{}{}^{}`$ $`=`$ $`(1\overline{Q}_{0}^{}{}_{}{}^{2})\left[12(1+k^2)\overline{U}^2\right],`$ (84)
$`\overline{U}^{}`$ $`=`$ $`2\left(\overline{Q}_0\overline{U}{\displaystyle \frac{k}{1+k^2}}\right)\left[1(1+k^2)\overline{U}^2\right]`$ (85)
$`+{\displaystyle \frac{1}{k}}\left(1{\displaystyle \frac{2k^2}{1+k^2}}\right)\overline{W}^2,`$
$`\overline{W}^{}`$ $`=`$ $`(1+k^2)\left(2\overline{Q}_0\overline{U}{\displaystyle \frac{1}{k}}\right)\overline{U}\overline{W}.`$ (86)
The equilibrium points of this system are listed in table 5. They constitute a subset of the points listed in sections 4 and 7. In figures 1 and 2, some examples of state-space diagrams for this model are displayed.
### 9.2 Timelike self-similar case
In the TSS case, the reduced dynamical system becomes (eliminating the variable $`\stackrel{~}{C}_1`$):
$`\stackrel{~}{Q}_{0}^{}{}_{}{}^{}`$ $`=`$ $`(1\stackrel{~}{Q}_{0}^{}{}_{}{}^{2})\left[\stackrel{~}{C}_{2}^{}{}_{}{}^{2}{\displaystyle \frac{1+k^2}{k}}(\stackrel{~}{Q}_02k\stackrel{~}{U})\stackrel{~}{U}\right],`$ (87)
$`\stackrel{~}{U}^{}`$ $`=`$ $`{\displaystyle \frac{1}{k}}\left(1\stackrel{~}{Q}_{0}^{}{}_{}{}^{2}+2k\stackrel{~}{Q}_0\stackrel{~}{U}{\displaystyle \frac{2k^2}{1+k^2}}\right)\left[1(1+k^2)\stackrel{~}{U}^2\right]`$ (88)
$`+\left(\stackrel{~}{Q}_0\stackrel{~}{U}{\displaystyle \frac{2k}{1+k^2}}\right)\stackrel{~}{C}_{2}^{}{}_{}{}^{2},`$
$`\stackrel{~}{C}_{2}^{}{}_{}{}^{}`$ $`=`$ $`\stackrel{~}{C}_2\left[(1\stackrel{~}{Q}_{0}^{}{}_{}{}^{2}+2k\stackrel{~}{Q}_0\stackrel{~}{U}){\displaystyle \frac{1+k^2}{k}}\stackrel{~}{Q}_0(1\stackrel{~}{C}_{2}^{}{}_{}{}^{2})\right].`$ (89)
The equilibrium points of this system are listed in table 6. They constitute a subset of the points listed in sections 6 and 7. In figures 3 and 4, some examples of state-space diagrams for this model are displayed.
### 9.3 Discussion
In both the SSS and the TSS cases there are always attractors dominated by the kinetic part of the scalar field, corresponding to the K-equilibrium points. Consequently, there will always be solutions that expand from a K-singularity and recollapse to a K-singularity. When $`\kappa ^2>2`$, the M-equilibrium points also are attractors. These correspond to dispersing solutions. Finally, there are $`\mathrm{\Phi }`$-equilibrium points, which correspond to power-law inflationary solutions when $`\kappa ^2<2`$. In the SSS case they act as attractors when $`\kappa ^2<2`$, while in the TSS case they are attractors when $`\kappa ^2>6`$.
To summarise, when $`\kappa ^2<2`$ the SSS case contains recollapsing K $``$ K solutions and power-law inflationary solutions K $``$ $`\mathrm{\Phi }`$. This is the same situation as for the Kantowski-Sachs models and closed Friedmann models examined in . In contrast, when $`\kappa ^2<2`$ the TSS case contains only recollapsing solutions. When $`\kappa ^2>2`$, both the SSS and the TSS cases contain recollapsing solutions and also dispersing solutions K $``$ M. The asymptotics is thus similar to the fluid-only case , where the generic regular solutions either are recollapsing K $``$ K solutions or ever-expanding solutions K $``$ M. Additionally, when $`\kappa ^2>6`$ the TSS case also contains dispersing K $``$ $`\mathrm{\Phi }`$ solutions that are non-inflationary.
## 10 Massless scalar field
Critical phenomena in gravitational collapse were first found by Choptuik in the study of a massless scalar field, and remain an active field of research (see, e.g., and references therein). The solution at the threshold of black-hole formation in spherically symmetric radiation fluid collapse, corresponding to $`\alpha =1/3`$, was studied by Evans & Coleman . In a new class of ‘asymptotically Minkowski’ self-similar spacetimes were presented, which were shown to be intimately related to the so-called critical phenomena which arise in spherically symmetric gravitational collapse calculations .
Here, we present the governing equations for a self-similar massless scalar field in spherical symmetry. In this case, $`\varphi `$ is still of the form
$$\varphi =\mathrm{\Phi }(\xi )+\sqrt{2}k\eta ,$$
(90)
and the corresponding equations are formally obtained by setting $`V=0`$. In the presence of a barotropic fluid we then essentially have a non-interacting two-fluid model , in which the massless scalar field can be identified with a stiff perfect fluid and the two fluids are separately conserved. We can then deduce that the models evolve from the massless scalar field model to the single-perfect fluid model . Hereafter, we shall assume that there is no barotropic fluid present, and that we are investigating a special case of the fluid vacuum model studied above. The massless scalar field equations without perfect fluid are then obtained by subsequently setting $`\mathrm{\Omega }_n=0`$ and $`\mathrm{\Omega }_t=0`$, respectively. This leads to the decoupling of $`v`$ and $`\overline{C}_1`$ in the SSS case, and $`u`$ and $`\stackrel{~}{C}_1`$ in the TSS case, respectively. Furthermore, in the SSS case $`\overline{W}=0`$, while in the TSS case, it follows that $`\stackrel{~}{Q}_0=ϵ`$ $`(=\pm 1)`$. The remaining dynamical systems are thus three-dimensional systems in $`(\overline{Q}_0,\overline{Q}_+,\overline{U})`$ and $`(\stackrel{~}{Q}_+,\stackrel{~}{U},\stackrel{~}{C}_2)`$, respectively. The constraint leads to two separate regions: either $`C_1=0`$ or else $`Q_+=kU`$. Note that the latter is a special case of the models treated in section 9. Here, we briefly summarise the governing equations for these models and list all of the equilibrium points.
### 10.1 Spatially self-similar case
In the SSS case, the Friedmann equation becomes
$$0=1\overline{Q}_{+}^{}{}_{}{}^{2}(1+k^2)\overline{C}_{1}^{}{}_{}{}^{2}\overline{U}^2,$$
(91)
which implies that
$$\overline{Z}1\overline{Q}_{+}^{}{}_{}{}^{2}\overline{U}^2=(1+k^2)\overline{C}_{1}^{}{}_{}{}^{2}.$$
(92)
The constraint becomes
$$0=(\overline{Q}_++k\overline{U})\overline{Z},$$
(93)
and the reduced dynamical system is
$`\overline{Q}_{0}^{}{}_{}{}^{}`$ $`=`$ $`(1\overline{Q}_{0}^{}{}_{}{}^{2})(12\overline{Z}),`$ (94)
$`\overline{Q}_{+}^{}{}_{}{}^{}`$ $`=`$ $`2\overline{Z}\left(\overline{Q}_0\overline{Q}_++{\displaystyle \frac{k^2}{1+k^2}}\right),`$ (95)
$`\overline{U}^{}`$ $`=`$ $`2\overline{Z}\left(\overline{Q}_0\overline{U}{\displaystyle \frac{k}{1+k^2}}\right).`$ (96)
The constraint implies that either $`\overline{Z}`$ or $`\overline{Q}_++k\overline{U}`$ must be zero. When $`\overline{Z}=0`$, both $`\overline{Q}_+`$ and $`\overline{U}`$ are constants (subject to $`1\overline{Q}_{+}^{}{}_{}{}^{2}\overline{U}^2=0`$), whereby the dynamical system reduces to a single evolution equation for $`\overline{Q}_0`$, and consequently $`\overline{Q}_0`$ is monotonically decreasing. If the constant values of $`\overline{Q}_+`$ and $`\overline{U}`$ do not satisfy $`\overline{Q}_++k\overline{U}=0`$, then the dynamics in the invariant set $`\overline{Z}=0`$ does not intersect with the dynamics in the invariant set $`\overline{Q}_++k\overline{U}=0`$. The latter is contained within the fluid vacuum case studied in the previous section (see figure 1). The equilibrium points of the system are listed in table 7.
### 10.2 Timelike self-similar case
In the TSS case, the Friedmann equation becomes
$$0=1\stackrel{~}{Q}_{+}^{}{}_{}{}^{2}(1+k^2)\stackrel{~}{C}_{1}^{}{}_{}{}^{2}\stackrel{~}{U}^2\stackrel{~}{C}_{2}^{}{}_{}{}^{2},$$
(97)
which implies that
$$\stackrel{~}{Z}1\stackrel{~}{Q}_{+}^{}{}_{}{}^{2}\stackrel{~}{U}^2=(1+k^2)\stackrel{~}{C}_{1}^{}{}_{}{}^{2}+\stackrel{~}{C}_{2}^{}{}_{}{}^{2}.$$
(98)
The constraint becomes
$$0=(\stackrel{~}{Q}_++k\stackrel{~}{U})(\stackrel{~}{C}_{2}^{}{}_{}{}^{2}\stackrel{~}{Z}),$$
(99)
and the reduced dynamical system is
$`\stackrel{~}{Q}_{+}^{}{}_{}{}^{}`$ $`=`$ $`ϵ\stackrel{~}{Q}_+(2\stackrel{~}{Z}\stackrel{~}{C}_{2}^{}{}_{}{}^{2})+{\displaystyle \frac{2k^2}{1+k^2}}(\stackrel{~}{C}_{2}^{}{}_{}{}^{2}\stackrel{~}{Z}),`$ (100)
$`\stackrel{~}{U}^{}`$ $`=`$ $`ϵ\stackrel{~}{U}(2\stackrel{~}{Z}\stackrel{~}{C}_{2}^{}{}_{}{}^{2}){\displaystyle \frac{2k}{1+k^2}}(\stackrel{~}{C}_{2}^{}{}_{}{}^{2}\stackrel{~}{Z}),`$ (101)
$`\stackrel{~}{C}_{2}^{}{}_{}{}^{}`$ $`=`$ $`ϵ\stackrel{~}{C}_2(1+\stackrel{~}{C}_{2}^{}{}_{}{}^{2}2\stackrel{~}{Z}).`$ (102)
The constraint implies that either $`\stackrel{~}{Z}=\stackrel{~}{C}_{2}^{}{}_{}{}^{2}`$ or $`\stackrel{~}{Q}_+=k\stackrel{~}{U}`$. When $`\stackrel{~}{Z}=\stackrel{~}{C}_{2}^{}{}_{}{}^{2}`$, the dynamical system becomes two-dimensional. The evolution equation for $`\stackrel{~}{C}_2`$ decouples and $`\stackrel{~}{C}_2`$ and $`\stackrel{~}{Q}_+`$ (or $`\stackrel{~}{U}`$) are monotonic. If the values of $`\stackrel{~}{Q}_+`$ and $`\stackrel{~}{U}`$ do not satisfy $`\stackrel{~}{Q}_++k\stackrel{~}{U}=0`$, then the dynamics in the invariant set $`\stackrel{~}{Z}=\stackrel{~}{C}_{2}^{}{}_{}{}^{2}`$ does not intersect with the dynamics in the invariant set $`\stackrel{~}{Q}_++k\stackrel{~}{U}=0`$. The latter is contained within the fluid vacuum case studied in the previous section, and corresponds to the semi-disks at $`\stackrel{~}{Q}_0=\pm 1`$ in figure 4. The equilibrium points of the system are listed in table 8.
### 10.3 Discussion
The dynamics is different in the various invariant submanifolds. The $`\overline{Z}=0`$ submanifold of the SSS case only contains recollapsing K $``$ K solutions. The dynamics in the $`\overline{Q}_+=k\overline{U}`$ submanifold of the SSS case is more complicated (see figure 1); when $`\kappa ^2<2`$ there are only recollapsing K $``$ K solutions, when $`2<\kappa ^2<6`$ there are recollapsing solutions and dispersing K $``$ M solutions, and when $`\kappa ^2>6`$ there are also singularity-free bouncing M $``$ M solutions, analogous to the bouncing Friedmann – Lemaître solutions. The TSS case only contains dispersing solutions. In the $`\stackrel{~}{Z}=\stackrel{~}{C}_{2}^{}{}_{}{}^{2}`$ submanifold there are K $``$ C solutions, while in the $`\stackrel{~}{Q}_+=\stackrel{~}{U}`$ submanifold there are K $``$ M solutions.
## Appendix A Fluid quantities
For the non-tilted ($`v=0`$) equilibrium points of the SSS case, we have given the deceleration parameter $`q_{\mathrm{pf}}`$ with respect to the fluid congruence. The necessary expressions are summarised below.
The expansion of the fluid congruence is given by
$`\theta _{\mathrm{pf}}`$ $`=`$ $`_au_{\mathrm{pf}}^a`$ (103)
$`=`$ $`{\displaystyle \frac{\overline{Y}e^x}{\sqrt{3}\sqrt{1v^2}}}\left({\displaystyle \frac{vv^{}}{1v^2}}+2\overline{Q}_0+\overline{Q}_++3v\overline{C}_1\right),`$
and the deceleration parameter, defined by
$`q_{\mathrm{pf}}`$ $`=`$ $`\left(1+3{\displaystyle \frac{u_{\mathrm{pf}}^a_a\theta _{\mathrm{pf}}}{\theta _{\mathrm{pf}}^2}}\right)`$ (104)
is given by
$`q_{\mathrm{pf}}`$ $`=`$ $`1{\displaystyle \frac{3}{\left[\frac{vv^{}}{1v^2}+2\overline{Q}_0+\overline{Q}_++3v\overline{C}_1\right]^2}}\times `$ (105)
$`(2\overline{Q}_{0}^{}{}_{}{}^{}+\overline{Q}_{+}^{}{}_{}{}^{}+{\displaystyle \frac{v(2\overline{Q}_0+\overline{Q}_+)+\overline{C}_1(3v^2)}{1v^2}}v^{}+{\displaystyle \frac{1+2v^2}{(1v^2)^2}}v_{}^{}{}_{}{}^{2}`$
$`+{\displaystyle \frac{vv^{\prime \prime }}{1v^2}}(2\overline{Q}_0+\overline{Q}_++v^{})\{\overline{Q}_++\overline{Q}_0[2(\overline{Q}_{+}^{}{}_{}{}^{2}+\overline{U}^2)`$
$`+{\displaystyle \frac{\gamma }{1+(\gamma 1)v^2}}\mathrm{\Omega }_n]\}\{2(\overline{Q}_0\overline{Q}_+)+3v\overline{C}_1\}v\overline{C}_1).`$
## Appendix B Transformation between the SSS and TSS variables
The two sets of variables basically differ only in the choice of dominant quantities (although one has to be careful, as the change of causality may result in sign change). Defining
$`\zeta `$ $`=`$ $`{\displaystyle \frac{\stackrel{~}{Y}}{\overline{Y}}}`$ (106)
$`=`$ $`\sqrt{\overline{Q}_{0}^{}{}_{}{}^{2}\overline{W}^2}`$ (107)
$`=`$ $`{\displaystyle \frac{1}{\sqrt{\stackrel{~}{Q}_{0}^{}{}_{}{}^{2}\stackrel{~}{C}_{2}^{}{}_{}{}^{2}}}},`$ (108)
the relations between the variable sets become
$`\overline{Q}_0=\zeta \stackrel{~}{Q}_0,\overline{Q}_+=\zeta \stackrel{~}{Q}_+,\overline{C}_1=\zeta \stackrel{~}{C}_1,v=u^1,U=\zeta \stackrel{~}{U},`$ (109)
$`\overline{W}=\zeta \sqrt{\stackrel{~}{Q}_{0}^{}{}_{}{}^{2}1},\stackrel{~}{C}_2=\zeta ^1\sqrt{\overline{Q}_{0}^{}{}_{}{}^{2}1}.`$ (110)
## Acknowledgements
AC would like to acknowledge financial support from NSERC of Canada. MG would like to thank the Department of Mathematics and Statistics at Dalhousie University for hospitality while this work was carried out.
## References
|
warning/0003/gr-qc0003008.html
|
ar5iv
|
text
|
# A New Study about the Two Formulations of Conservation Laws for Matter Plus Gravitational Field and Their Experimental Test
## I The Debate in General Relativity Eighty Years Ago
In 1914, Einstein obtained the conservation laws for matter plus gravitational field in the form
$$\frac{}{x^\mu }(𝒯_{(M)\nu }^\mu +t_{(G)\nu }^\mu )=0,$$
(1)
$`𝒯_{(M)\nu }^\mu `$ is the energy-momentum tensor density for matter field, Einstein called $`t_{(G)\nu }^\mu `$ the energy-momentum pseudotensor density of gravitational field.
Lorentz in 1916 and Levi-Civita in 1917 proposed successively to use
$$𝒯_{(M)\nu }^\mu (x)+𝒯_{(G)\nu }^\mu (x)=0$$
(2)
or
$$\frac{}{x^\mu }(𝒯_{(M)\nu }^\mu +𝒯_{(G)\nu }^\mu )=0$$
(3)
as conservation laws for matter plus gravitational field, their propositions evoked an important debate on the correct formulation of conservation laws in general relativity eighty years ago.
The definition of $`𝒯_{(M)\nu }^\mu `$ is $`𝒯_{(M)\nu }^\mu \stackrel{\mathrm{def}}{=}2\frac{\delta W_M}{\delta g_{\mu \alpha }}g_{\nu \alpha }`$ . This definition has been accepted universally in theoretical physics. It is naturally to define the energy-momentum tensor density for gravitational field by $`𝒯_{(G)\nu }^\mu \stackrel{\mathrm{def}}{=}2\frac{\delta W_G}{\delta g_{\mu \alpha }}g_{\nu \alpha }`$ . So Lorentz and Levi-Civita adopted this definition. In the above definition, $`W_M=_M(x)d^4x`$, $`\delta W_M=\frac{\delta W_M}{\delta g_{\mu \nu }}\delta g_{\mu \nu }d^4x`$; $`W_G=_G(x)d^4x,\delta W_G=\frac{\delta W_G}{\delta g\mu \nu }d^4x`$; $`(x)=_M(x)+_G(x)`$ is the Lagrangian density of the whole system, $`_M(x)`$ and $`_G(x)`$ are the matter field part and the pure gravitational field part of $`(x)`$ respectively. In General relativity, $`𝒯_{(G)\nu }^\mu =\frac{c^4}{8\pi G}\sqrt{g}(R_\nu ^\mu \frac{1}{2}g_\nu ^\mu R)`$. Some people think that $`𝒯_{(G)\nu }^\mu `$ is a pure geometric quantity and can not be used as the definition of energy-momentum tensor for gravitational field. This view is incorrect, because the metric tensor $`g_{\mu \nu }`$ is both geometric quantity and dynamic quantity in the theory of gravitation, so is $`𝒯_{(G)\nu }^\mu `$ .
Eq. (1) can be derived from the local translational symmetry of the gravitational system . There exist the relations
$$t_{(G)\nu }^\mu =2\frac{\delta W_G}{\delta g_{\mu \alpha }}g_{\nu \alpha }\frac{}{x^\sigma }v_{(G)\nu }^{\mu \sigma },\frac{}{x^\sigma }v_{(G)\nu }^{\mu \sigma }=\frac{}{x^\sigma }v_{(G)\nu }^{\sigma \mu }$$
(4)
where $`v_{(G)\nu }^{\mu \sigma }`$ is determined by the Lagrangian density $`_G`$. Eq. (1) can also be derived from Einstein field equations or from the covariant generalized conservation laws $`𝒯_{(M)\nu ;\mu }^\mu =0`$ . $`t_{(G)\nu }^\mu `$ obtained from distinct methods are different, but their difference can always be expressed by the relations: $`{}_{}{}^{\prime \prime }t_{(G)\nu }^{\mu }^{}t_{(G)\nu }^\mu =_\alpha u_\nu ^{\mu \alpha }`$, where $`_\alpha u_\nu ^{\mu \alpha }=_\alpha u_\nu ^{\alpha \mu }`$ or $`u_\nu ^{\mu \alpha }=u_\nu ^{\alpha \mu }`$.
The quantity $`𝒯_{(G)\nu }^\mu `$ is a tensor density, but $`t_{(G)\nu }^\mu `$ is not. The conservation law Eq. (3) is covariant, but the conservation law Eq. (1) is not. The well-known serious difficulties in connection with $`t_{(G)\nu }^\mu `$ do not exist for $`𝒯_{(G)\nu }^\mu `$ and Eq. (3) is more in line with the spirit of general relativity .
The logical rationality for the definition of $`𝒯_{(G)\nu }^\mu `$ had been acknowledged by Einstein , he also acknowledged that one is not entitled to define $`t_{(G)\nu }^\mu `$ as a quantity representing the energy-momentum of gravitational field; but Einstein doubted about the physical meaning indicated by the relation in Eq. (2). He said that “Eq. (2) does not exclude the possibility that a material system disappears completely, leaving no trace of its existence. In fact, the total energy in Eq. (2) is zero from the beginning, and the conservation of this energy value does not guarantee the persistence of the system in any form” ; so he opposed to choose $`𝒯_{(G)\nu }^\mu `$ as the energy-momentum tensor density for gravitational field. Care must be taken to that the reason which Einstein opposed $`𝒯_{(G)\nu }^\mu `$ is not because of its logical trouble or is not dependent on any experimental result; it is only owing to that he thought Eq. (2) being nonsensical. In the following the identities Eq. (2) is reexamined and a new explanation is given, moreover an experimental test is offered to decide which definition of gravitational energy-momentum tensor density and which formulation conservation laws are correct. It can be shown that Eq. (2) have a plentiful physical contents and might be tested by experiments.
## II Reexamination of the Identities $`𝒯_{(M)\nu }^\mu (x)+𝒯_{(G)\nu }^\mu (x)=0`$ and New Interpretations of Gravitational Wave
Should Eq. (2) cause inevitably a material system disappear completely? This is the crux of the problem. We must point out that it is infeasible to determine solely how a material system changes merely using the conservation law of energy alone. Moreover it is impossible to determine whether this material system disappears completely. The change of a material system must yet obey other laws, such as the conservation law of baryon number, the second law of thermodynamic, etc. Therefore, $`𝒯_{(M)\nu }^\mu +𝒯_{(G)\nu }^\mu =0`$ does not necessarily give $`𝒯_{(M)\nu }^\mu =0`$, it only implies $`𝒯_{(G)\nu }^\mu (x)=𝒯_{(M)\nu }^\mu (x)`$.
Eq. (2) shows that the energy-momentum tensor of the gravitational field must coexist with the energy-momentum tensor of material. Their sum total is equal to zero invariably and their distributional region in space-time is the same. These properties make Eq. (2) have a plentiful physical content which we shall show below, the consequences of these properties would be verified in future experiments. So Eq. (2) is not nonsensical from a physical point of view.
It is worth noting that, $`t_{(G)\nu }^\mu `$ and $`𝒯_{(G)\nu }^\mu `$ are not independent, they are interrelated by Eq. (4); via definition of $`𝒯_{(G)\nu }^\mu `$ and Eq. (4), Eq. (3) can be derived from Eq. (1), and on the other hand via Eq. (4), Eq. (1) can also be derived from Eq. (3), so Eq. (1) and Eq. (3) are not independent either. As a better choice, in this paper we shall follow Lorentz and Levi-Civita to treat $`𝒯_{(G)\nu }^\mu `$ , but not $`t_{(G)\nu }^\mu `$ , as the energy-momentum tensor density for gravitational field, and to treat Eq. (3), but not Eq. (1), as the conservation laws of energy-momentum tensor density; and we shall treat $`t_{(G)\nu }^\mu `$ as a subsidiary quantity, then Eq. (1) represents a number of subsidiary relations. According to Lorentz and Levi-Civita’s formulation of conservation laws, the Einstein’s gravitational equations $`\sqrt{g}(R_\nu ^\mu \frac{1}{2}g_\nu ^\mu R)=\frac{8\pi G}{c^4}𝒯_{(M)\nu }^\mu `$ are interpreted both as field equations and as conservation laws , therefore, except using $`𝒯_{(G)\nu }^\mu `$ as the energy-momentum tensor density for gravitational field, all deductions of gravitational field equations remain unchanged. For instance there was still a singularity at the beginning of the present expansion phase of the universe.
The existence of gravitational wave is determined by Einstein’s gravitational field equations, the characteristics of these equations show that the gravitational wave propagate with speed of light .
A great number of people follow Einstein’s viewpoint, they believe that gravitational wave, like electromagnetic wave, is accompanied by radiation of energy. They also believe that this radiation has been verified by PSR 1913 + 16. These views are incorrect. Now let us show that gravitational wave could not transmit energy but could transmit information. The basis for the belief that gravitational wave radiates energy is using $`t_{(G)\nu }^\mu `$ and applying the following equation
$$\frac{}{t}_V(𝒯_{(M)0}^0+t_{(G)0}^0)𝑑V=c_St_{(G)0}^i𝑑S_i,$$
(5)
which can be derived from Eq. (1). The surface $`S`$ enclosing the volume $`V`$ is taken in vacuum where $`𝒯_{(M)\nu }^\mu =0`$. The right integral in Eq. (5) is interpreted as the amount of gravitational energy transferred by the gravitational wave across the surface $`S`$ in unit time.
If we treat $`𝒯_{(G)\nu }^\mu `$ , but not $`t_{(G)\nu }^\mu `$ , as the energy-momentum tensor density for gravitational field, new insights about the gravitational wave can be gained from Eq. (3) (or Eq. (2)). Using these equations, it is easy to obtain the identity
$$\frac{}{t}_V(𝒯_{(M)0}^0+𝒯_{(G)0}^0)𝑑V=0.$$
(6)
This identity can also be derived from Eq. (5) via Eq. (4); and inversely Eq. (5) can be derived from Eq. (6) via Eq. (4). Eq. (6) indicates that there are no energy transferred from $`V`$ to outside via the gravitational wave, $`i.e.`$ gravitational wave does not transmit energy. It must be pointed out that although $`𝒯_{(M)0}^0+𝒯_{(G)0}^0=0`$, the values of $`𝒯_{(M)0}^0`$ and $`𝒯_{(G)0}^0`$ may change, they can be transformed to each other.
It should emphasize that, although the gravitational wave does not transmit energy (and momentum), but it could transmit information. When the gravitational wave passes through a space point, the gravitational field, $`i.e.`$ the metric field of this point will change from $`g_{\mu \nu }(\stackrel{}{r},t)`$ into $`g_{\mu \nu }(\stackrel{}{r},t+\mathrm{\Delta }t)`$ in the time interval $`(t,t+\mathrm{\Delta }t)`$, these changes in $`g_{\mu \nu }`$ convey the information from the source of gravitational wave. Some people guess that information should be closely bound up with energy, they think that the information without energy must not exist. Their guess is not correct. The gravitational wave is determined fully by Einstein field equations , which are irrespective how to define the energy-momentum tensor density for gravitational field. If we adopt the definition $`𝒯_{(G)\nu }^\mu \stackrel{\mathrm{def}}{=}2\frac{\delta W_G}{\delta g_{\mu \alpha }}g_{\nu \alpha }`$ the gravitational wave would transfer information without energy.
When there exists gravitational wave, the space-time metric $`g_{\mu \nu }(x)`$ must change. This change may be considered as that the metric undergoes a perturbation by writing $`g_{\mu \nu }(x)=\underset{\mu \nu }{\overset{}{g}}(x)+h_{\mu \nu }(x)`$ , $`\underset{\mu \nu }{\overset{}{g}}`$ is the background metric and $`h_{\mu \nu }`$ is the perturbation; usually $`|h_{\mu \nu }||\underset{\mu \nu }{\overset{}{g}}|`$, the magnitude of $`h_{\mu \nu }(x)`$ reflects the intensity of gravitational wave.
The solution for weak-field approximations of Einstein field equations has been given as $`\phi _\nu ^\mu (\stackrel{}{r},t)=\frac{4G}{c^4}\frac{(T_\nu ^\mu )_{ret.}d^3x^{}}{|\stackrel{}{r}\stackrel{}{r^{}}|}`$ , where $`\phi _\nu ^\mu =h_\nu ^\mu \frac{1}{2}\delta _\nu ^\mu h`$. From this formula, it is evident that the gravitational wave can be generated only if the energy-momentum tensor of material source is changed in space-time. Let $`R_{\mu \nu }(x)`$ and $`\underset{\mu \nu }{\overset{}{R}}(x)`$ be the Ricci tensors corresponding to $`g_{\mu \nu }(x)`$ and $`\underset{\mu \nu }{\overset{}{g}}(x)`$, $`R_{\mu \nu }(x)`$ and $`\underset{\mu \nu }{\overset{}{R}}(x)`$ must obey with Einstein field equation $`(R_{\mu \nu }\frac{1}{2}g_{\mu \nu }R)=\frac{8\pi G}{c^4}T_{(M)\mu \nu }`$ and $`(\underset{\mu \nu }{\overset{}{R}}\frac{1}{2}\underset{\mu \nu }{\overset{}{g}}\stackrel{}{R})=\frac{8\pi G}{c^4}\underset{(M)\mu \nu }{\overset{}{T}}`$ respectively. It has been proved that $`R_{\mu \nu }=\underset{\mu \nu }{\overset{}{R}}+R_{\mu \nu }^{(1)}(h)+R_{\mu \nu }^{(2)}(h)+\mathrm{}`$, where $`R_{\mu \nu }^{(1)}`$ and $`R_{\mu \nu }^{(2)}`$ are composed of the covariant derivatives of $`h_{\mu \nu }`$, $`R_{\mu \nu }^{(1)}`$ is the linear term and $`R_{\mu \nu }^{(2)}`$ is in the second degree term. Either of these two equations can be rewritten as
$$\underset{\mu \nu }{\overset{}{R}}\frac{1}{2}\underset{\mu \nu }{\overset{}{g}}\stackrel{}{R}=\frac{8\pi G}{c^4}(T_{(M)\mu \nu }+W_{(G)\mu \nu })$$
(7)
Where
$$W_{(G)\mu \nu }\stackrel{\mathrm{def}}{=}\frac{c^4}{8\pi G}\{(R_{\mu \nu }\frac{1}{2}g_{\mu \nu }R)(\underset{\mu \nu }{\overset{}{R}}\frac{1}{2}\underset{\mu \nu }{\overset{}{g}}\stackrel{}{R})\}=\underset{(M)\mu \nu }{\overset{}{T}}T_{(M)\mu \nu }$$
(8)
It is suggested to interpret $`W_{(G)\mu \nu }`$ or its average value as the energy-momentum tensor for gravitational wave. But in our opinion, although $`W_{(G)\mu \nu }`$ might be used to express gravitational field’s energy-momentum tensor relative to the background space-time, but it does not indicate the energy and momentum transmitted by gravitational wave. Since $`W_{(G)\mu \nu }=\underset{(M)\mu \nu }{\overset{}{T}}T_{(M)\mu \nu }=T_{(G)\mu \nu }\underset{(G)\mu \nu }{\overset{}{T}}`$ $`(T_{(G)\mu \nu }=\frac{1}{\sqrt{g}}𝒯_{(G)\nu }^\mu `$ and $`\underset{(G)\mu \nu }{\overset{}{T}}=\frac{1}{\sqrt{\stackrel{}{g}}}\stackrel{}{𝒯_{(G)\nu }^\mu })`$ and $`T_{(M)\mu \nu }(x)+T_{(G)\mu \nu }(x)=\underset{(M)\mu \nu }{\overset{}{T}}(x)+\underset{(G)\mu \nu }{\overset{}{T}}(x)=0`$, therefore at any point of space-time, the total energy-momentum tensor for matter plus gravitational field is identically equal to zero when the gravitational wave generate.
A question then arises as to how gravitational wave can be detected. Although the gravitational wave does not transmit energy, the equation of geodesic deviation remains correct. As a result one can still design the detectors based on the effect of this equation . For instance, Weber’s cylinder can be used to detect the gravitational wave. From the conventional viewpoint, gravitational wave can transmit energy to a detector. But in our viewpoint, when the gravitational wave is received, the oscillation energy of the detector does not come from the gravitational wave, it comes from the local gravitational field where the detector resides. Because the detector is made of matter, let $`𝒯_{(M)\nu }^\mu `$ be its energy-momentum tensor density, $`𝒯_{(M)\nu }^\mu `$ must satisfy Eq. (2). Therefore $`\mathrm{\Delta }𝒯_{(M)\nu }^\mu =\mathrm{\Delta }𝒯_{(G)\nu }^\mu ,\mathrm{\Delta }𝒯_{(M)\nu }^\mu `$ represents the increased energy of the detector and $`\mathrm{\Delta }𝒯_{(G)\nu }^\mu `$ represents the decreased energy of the local gravitational field; $`\mathrm{\Delta }𝒯_{(M)\nu }^\mu `$ or $`\mathrm{\Delta }𝒯_{(G)\nu }^\mu `$ can be calculated from Einstein field equation.
## III An Experimental Test to Decide Which Formulation Is Correct
The microwave background radiation gives us important information of the early epoch of the universe , the discovery and theoretical interpretation of this radiation spurs the cosmology to get ahead. It is believed that there existed also a vast amount of gravitational waves at the earliest epoch of the universe immediately after “big bang”, with the expansion of the universe there should also exist background gravitational waves at present .
The microwave background radiation has the type of spectrum of black body radiation . Do these background gravitational waves also have the type of spectrum of black body radiation? We shall study this question at once. The black body radiation is thermal radiation in equilibrium . Einstein advanced a semi-quantum theory of thermal radiation in equilibrium and derived the equation
$$\rho (\nu ,T)=\frac{8\pi h\nu ^3}{c^3}\frac{1}{e^{h\nu /KT}1}$$
(9)
from his theory. Eq. (9) expresses Plank’s law of radiation.
The Semi-quantum theory of thermal radiation in equilibrium advanced by Einstein is concise and reflect well the essence of this radiation, the key of Einstein’s theory is to think that the radiation wave transmits energy and it is quantized . The prevalent theory assumes that the gravitational wave transmits energy, and the graviton, which is similar to photon, is the quantum of energy for gravitational wave. Therefore Eq. (9) must be also applicable for gravitational radiation. It is possible that this distribution might deform during the evolution of the universe, however, the pattern of spectrum must be yet regular.
On the other hand, according to Lorentz and Levi-Civita’s definition of energy-momentum tensor density $`𝒯_{(G)\nu }^\mu `$ for gravitational field and the conservation law Eq. (2), the gravitational wave does not transmit energy, consequently for gravitational wave the concepts such as thermal radiation in equilibrium, the quantum of energy and distribution of radiation energy are all meaningless. These distinguishing features imply that Eq. (9) does not apply for gravitational waves, $`i.e.`$ the background gravitational waves do not tally with the black body radiation or its variety. The types of spectrum of background gravitational waves should not have any regularity, they are the results of random process. Therefore, through the observations of the spectrum types for background gravitational waves, it might provide an experimental test to decide whether gravitational wave transmits energy, $`i.e.`$ it might judge which is the correct definition of energy-momentum tensor for gravitational field, $`𝒯_{(G)\nu }^\mu `$ or $`t_{(G)\nu }^\mu `$? And which is the correct formulation of conservation laws for matter plus gravitational field, Eq. (1) or Eq. (3) (and Eq. (2))? We shall wait and see.
## IV Has the Gravitational Energy Radiation Been Verified?
The orbital energy loss of the binary pulsar system PSR 1913 + 16 has been confirmed from the observation of the decrease in its orbital period . This observation has been widely interpreted as verification for energy radiation of gravitational wave. However based on our analysis given above, this interpretation is problematic.
The theoretical basis of the so called verification is Eq. (5). When using Eq. (5) to compute the energy change of PSR 1913 + 16, one always assumes: 1 $`\frac{}{t}_V(t_{(G)0}^0)𝑑V=0`$ and 2 except the orbital kinetic energy and gravitational interaction energy (which belongs to $`𝒯_{(M)\nu }^\mu `$ according to definition), the other kinds of matter energy do not change for this binary pulsar. Evidently the result of computation can not be very accurate based on these assumptions. The ratio of the observed to the predicted damping rate $`\dot{P}_b^{obs}/\dot{P}_b^{pred}`$ for PSR 1913 + 16 is $`1.00032\pm 0.0035`$ , the coincidence between $`\dot{P}_b^{obs}`$ and $`\dot{P}_b^{pred}`$ is overprecise. As another example $`\dot{P}_b^{obs}/\dot{P}_b^{pred}`$ for PSR 1534 + 12 is about $`0.83`$ , the coincidence is less evident. Taking the approximation in calculation of $`\dot{P}_b^{pred}`$ into account, one can not rule out the possibility that the coincidence of observation and prediction for the damping rate $`\dot{P}_b`$ of PSR1913 + 16 might be accidental .
Even if the above coincidence for PSR 1913 + 16 is true, yet it can not fully prove that gravitational wave transmits energy. Because Eq. (5) can be derived from Eq. (6), the same value of $`\dot{P}_b^{pred}`$ may also be obtained from Eq. (6) and $`𝒯_{(G)\nu }^\mu `$ . Since Eq. (6) means that gravitational wave does not transmit energy, it is inevitable to conclude that the gravitational energy radiation has not been verified.
|
warning/0003/hep-th0003153.html
|
ar5iv
|
text
|
# Kaluza-Klein Induced Gravity Inflation
## I Introduction
Scale invariant model explains the origin of the scale parameters such as gravitational constant, cosmological constant as well as the masses for the fermion fields. Accordingly, all dimensionful parameters in Einstein Lagrangian are replaced by appropriate scalar measuring field with proper power according to their conformal dimensions.
Scale invariance also appears to be very important in various branches of physics such as QCD and many other inflationary models . Local scale symmetry has also been suggested to be related to the missing Higgs problem in electro-weak theory as well as many other research interests . It is also argued that scale invariant effective theory has to do with physics near the fixed points of renormalization group trajectory .
On the other hand, higher dimensional Kaluza-Klein theory has been a focus of research interest for a long time. In addition, Kaluza-Klein theory should be related to the evolution of our early universe if the compactification process is completed during the early stage of the universe. Hence one is naturally lead to the question whether scale invariant effective action is manifest before dimensional reduction process takes place. Therefore we propose to study the effect of a $`D`$ dimensional induced gravity in the very early universe.
One notes that there have been studies based on a $`D`$-dimensional Friedmann-Robertson-Walker (DFRW) metric
$$ds^2\widehat{g}_{MN}dz^Mdz^N=dt^2+a^2(t)\left[\frac{1}{1kr^2}dr^2+r^2d\mathrm{\Omega }_{D1}\right]$$
(1)
in search of a physically acceptable low energy effective theory . Here $`d\mathrm{\Omega }_n`$ denotes the n-dimensional solid angle with appropriate angular coordinates. Note, however, that the internal and external spaces will inflate all together if $`a(t)`$ undergoes an exponentially expanding process under the DFRW metric. Hence it would be interesting to see if a more general $`4`$+$`d`$ dimensional FRW (4dFRW) space is capable of inducing inflating external-space and contracting internal-space at the same time. Here $`dD4`$ denotes the dimension of the internal space. This kind of generalization is in fact a dimensional reduction from $`M^DR^1\times F^3\times F^d`$ with $`F^n`$ denoting the $`n`$-dimensional maximally symmetric space .
In this paper, we will show that an induced inflationary solution with expanding external-space and contracting internal-space will require a very special symmetry breaking scalar potential. The explicit form of the symmetry breaking potential required by the expanding solution with expanding external-space and constant internal-space will also be solved. In addition, we will also show that the most favorable solution appears to be the same as the result of DFRW space for the induced conventional-$`\varphi ^4`$ model. Our result indicates that the compactification process must have been completed before the inflationary process unless the symmetry breaking potential takes an unordinary form.
This paper will be organized as follows: (i) In section II, we will introduce the $`D`$-dimensional induced gravity theory. The $`D`$-dimensional equations of motion will also be compactified into $`4+d`$ dimensional in this section. (ii) We will also solve the equations of motion for an inflationary solution in the limit of slow-rollover approach in section III. It will be shown first that the existence of a solution with expanding external-space and contracting internal-space will impose a number of constraints on the coupled symmetry breaking potential. In particular, solution with expanding external-space and constant internal-space is also solved in search of the possible candidate for the scalar potential. (iii) the conventional $`\varphi ^4`$ model with a coupled SSB $`\varphi ^4`$ potential is solved and analyzed in section IV. (iv) Finally, conclusions are drawn in section V.
## II Induced Gravity in $`D`$ Dimension
In this paper we will consider the following induced gravity action :
$$S=d^Dz\sqrt{\widehat{g}}\left[\frac{1}{2}ϵ\varphi ^2\widehat{R}+\frac{1}{2}^M\varphi _M\varphi +V(\varphi )\right].$$
(2)
The scalar field $`\varphi `$ in (2) is the measuring field designed to replace the dimensionful Newtonian constant as appeared in the four dimensional induced gravity models. One should replace all dimensionful coupling constants with appropriate scalar fields according to their dimensions. After this replacement, one can show that, apart from a possible symmetry breaking potential $`V`$, the action (2) is invariant under the following scale transformation: $`g_{MN}\mathrm{\Lambda }^2g_{MN}`$ , $`\varphi \mathrm{\Lambda }^{(1D/2)}\varphi `$.
We will denote $`D=4+d`$ from now on in this paper. Here $`d=D4`$ is the dimension of the internal space in the Kaluza-Klein theory we are going to study in this paper. We will also use a hat notation, $`\widehat{R}_{MN}`$, to represent the physical variables in $`D`$-dimension and non-hatted variable, $`R_{ab}`$, will represent the same physical variables evaluated solely in the $`4`$-dimensional physical external-space. Barred notation, $`\overline{R}_{mn}`$, will also be employed to denote the same physical variables evaluated in $`d`$-dimensional internal-space. Note that by the same physical variables we meant that they are defined by the same notation except the metric is replaced by the appropriate metric defined in its own space. Furthermore, we will use capital indices $`M,N,\mathrm{}`$ $`(=0,1,2,\mathrm{},D1)`$ to denote the $`D`$ dimensional space-time indices. Also lower case Latin indices from the beginning ($`a,b,c,\mathrm{}`$) of the alphabet will denote the four dimensional space time indices $`(a,b,c=0,1,2,3)`$. In addition, $`i,j,k,l(=1,2,3)`$ labels the spatial $`3`$-manifold. Finally, we will use lower case Latin indices from the middle ($`m,n,\mathrm{}`$) of the alphabet to label the $`d`$-dimensional compactified internal-space.
Note that the Kaluza-Klein dimensional reduction process we will adopt in this paper is the following $`4+d`$ dimensional Friedmann-Robertson-Walker metric (4dFRW)
$$ds^2\widehat{g}_{MN}dz^Mdz^N=dt^2+a^2(t)h_{ij}(x)dx^idx^j+b^2(t)\overline{h}_{mn}(y)dy^mdy^n.$$
(3)
Here $`h_{ij}dx^idx^j(1k_1r^2)^1dr^2+r^2d\mathrm{\Omega }_3`$ and $`\overline{h}_{mn}dy^mdy^n(1k_2s^2)^1ds^2+s^2d\mathrm{\Omega }_d`$ with $`k_1,k_2=0,\pm 1`$ denoting the signature of the external space and internal space respectively.
Note that if we adopt the compactification ansatz $`\varphi (z)=\varphi (x)\kappa ^{\frac{d}{2}}`$, the compactified $`4`$-dimensional effective Einstein action, except the SSB $`\varphi ^4`$ potential term, will remain scale invariant under the $`4`$-dimensional scale transformation: $`g_{ab}(x)\mathrm{\Lambda }(x)^2g_{ab}(x)`$ , $`\varphi (x)\mathrm{\Lambda }(x)^{(2D)/2}\varphi (x)`$. This shows that this is a consistent and scale-invariant way to carry out the compactification process. Note also that $`\kappa `$ is a dimension one constant parameter such that $`d^dy\kappa ^d`$ is dimensionless and will be set as $`1`$ for latter convenience. Note that we will also use the same $`\varphi `$ notation for $`\varphi (z)`$ and $`\varphi (x)`$ for convenience.
The equations of motion can be obtained from varying the action (2) with respect to $`\varphi `$ and $`\widehat{g}_{MN}`$ respectively. As a result, one has:
$`ϵ\varphi \widehat{R}D_M^M\varphi +{\displaystyle \frac{}{\varphi }}V(\varphi )`$ $`=`$ $`0,`$ (4)
$`ϵ\varphi ^2\widehat{G}_{MN}`$ $`=`$ $`ϵ(D_M_N\widehat{g}_{MN}D^P_P)\varphi ^2+\widehat{T}_{MN}^\varphi .`$ (5)
Here $`\widehat{G}_{MN}\frac{1}{2}\widehat{R}\widehat{g}_{MN}\widehat{R}_{MN}`$ defines the Einstein tensor. Moreover, $`\widehat{T}_\varphi ^{MN}^M\varphi ^N\varphi \widehat{g}^{MN}[\frac{1}{2}^P\varphi _P\varphi +V(\varphi )]`$ is the energy momentum tensor associated with $`\varphi `$. Furthermore, the curvature tensor $`\widehat{R}_{MNOP}`$ is defined by $`[D_M,D_N]\widehat{A}_O=\widehat{R}_{ONM}^P\widehat{A}_P`$. In addition, the Ricci tensor and scalar curvature are defined by $`\widehat{R}_{ON}\widehat{R}_{ONP}^P`$ and $`\widehat{R}\widehat{R}_{ON}\widehat{g}^{ON}`$ respectively.
For latter convenience, we will define $`\varphi ^2e^\phi `$, $`ae^\alpha `$, $`be^\beta `$, $`V(\varphi )U(\phi )`$ throughout this paper. Hence one can bring (4) and (5) into a more comprehensive form in terms of the new variables and parameters defined earlier. Indeed, (4) and (5) can be written as
$$\widehat{G}_{MN}=D_M_N\phi +_M\phi _N\phi \widehat{g}_{MN}(D^P_P\phi +^P\phi _P\phi )\widehat{T}_{MN}^\phi ,$$
(6)
$$\widehat{R}=\frac{1}{4ϵ}(^M\phi _M\phi +2D^P_P\phi )\frac{8}{ϵ}e^\phi \frac{U(\phi )}{\phi }.$$
(7)
Hence one has
$`\widehat{R}`$ $`=`$ $`{\displaystyle \frac{1}{4ϵ}}(_a\phi ^a\phi +D_a^a\phi ){\displaystyle \frac{2}{ϵ}}e^\phi _\phi U(\phi ),`$ (8)
$`\widehat{G}_{ab}`$ $`=`$ $`(_a\phi _b\phi +D_a_b\phi )g_{ab}(^c\phi _c\phi +D^P_P\phi )T_{ab}^\phi ,`$ (9)
$`\widehat{G}_{mn}`$ $`=`$ $`D_m_n\phi \overline{g}_{mn}(^c\phi _c\phi +D^P_P\phi )\overline{T}_{mn}^\phi .`$ (10)
Here we have defined the generalized energy momentum tensor for $`\phi `$ and $`\widehat{T}_{MN}^\phi `$ as:
$$\widehat{T}_{MN}^\phi =\frac{1}{4ϵ}(\frac{1}{2}\widehat{g}_{MN}_P\phi ^P\phi _M\phi _N\phi )+\frac{V}{ϵ}e^\phi \widehat{g}_{MN}.$$
(11)
Therefore, one has
$`T_{ab}^\phi `$ $`=`$ $`{\displaystyle \frac{1}{4ϵ}}({\displaystyle \frac{1}{2}}g_{ab}^c\phi _c\phi _a\phi _b\phi )+{\displaystyle \frac{1}{ϵ}}e^\phi U(\phi )g_{ab},`$ (12)
$`\overline{T}_{mn}^\phi `$ $`=`$ $`{\displaystyle \frac{1}{2}}{\displaystyle \frac{1}{4ϵ}}\overline{g}_{mn}^c\phi _c\phi +{\displaystyle \frac{1}{ϵ}}e^\phi U(\phi )\overline{g}_{mn},`$ (13)
respectively. In addition, with the compactified metric
$$ds^2\widehat{g}_{MN}dz^Mdz^N=g_{ab}(z)dx^adx^b+\overline{g}_{mn}(z)dy^mdy^n,$$
(14)
and by setting $`\varphi (z)=\varphi (x)`$, one can derive the following compactified identities for the curvature terms:
$`\widehat{R}`$ $`=`$ $`R+2dD_a^a\beta +d(d+1)_a\beta ^a\beta d(d1)k_2e^{2\beta },`$ (15)
$`\widehat{G}_{ab}`$ $`=`$ $`G_{ab}+t_{ab},`$ (16)
$`\widehat{G}_{mn}`$ $`=`$ $`{\displaystyle \frac{1}{2}}\overline{g}_{mn}[R+2(d1)D_a^a\beta +d(d1)_a\beta ^a\beta `$ (18)
$`(d1)(d2)k_2e^{2\beta }].`$
Here the generalized energy momentum tensor $`t_{ab}`$ is given by
$`t_{ab}`$ $``$ $`{\displaystyle \frac{1}{2}}g_{ab}[\mathrm{\hspace{0.25em}2}dD_c^c\beta +d(d+1)_c\beta ^c\beta d(d1)k_2e^{2\beta }]`$ (20)
$`d(D_a_b\beta +_a\beta _b\beta ).`$
Note that $`\widehat{g}^{MN}\widehat{G}_{MN}=(\frac{D}{2}1)\widehat{R}`$. Hence one can obtain the following equation of $`\phi `$
$$_a\phi ^a\phi +D_a^a\phi +d_a\beta ^a\phi =\kappa _3\frac{e^\phi }{ϵ}[(D2)_\phi UDU]$$
(21)
from eliminating the $`\widehat{R}`$ term in the trace of $`\widehat{G}_{MN}`$ equation (6) and the $`\phi `$ equation (8). Here one has set $`\kappa _3=4ϵ/[4(D1)ϵ+D2]`$. Finally, one can show that the trace of the $`\widehat{G}_{mn}`$ equation (10), $`\overline{g}^{mn}\widehat{G}_{mn}`$, and the trace of the $`\widehat{G}_{mn}`$ equation (18) gives two constraint-equations related to $`R`$. One can eliminate these $`R`$ terms and obtain the following equation for $`b(t)`$:
$`d_a\beta ^a\beta +D_a^a\beta (d1)k_2e^{2\beta }`$ $`=`$ $`(1+{\displaystyle \frac{1}{4ϵ}})[_a\phi ^a\phi +D_a^a\phi +d_a\beta ^a\phi ]`$ (23)
$`_a\beta ^a\phi +{\displaystyle \frac{e^\phi }{ϵ}}[U_\phi U]`$
Therefore we will take (23) as the independent $`\beta `$-equation. This will soon be shown to be helpful in our analysis below. Finally, one can show that $`G_{tt}`$ component of the equation (9), the $`\phi `$ equation (21) and the $`\beta `$ equation (23) becomes
$`\alpha ^2+{\displaystyle \frac{k_1}{a^2}}+d\alpha ^{}\beta ^{}+{\displaystyle \frac{d(d1)}{6}}(\beta ^2+k_2e^{2\beta })`$ $`+`$ $`\alpha ^{}\phi ^{}+{\displaystyle \frac{d}{3}}\beta ^{}\phi ^{}={\displaystyle \frac{1}{24ϵ}}\phi ^2+{\displaystyle \frac{U}{3ϵ}}e^\phi `$ (24)
$`\phi ^{\prime \prime }+3\alpha ^{}\phi ^{}+d\beta ^{}\phi ^{}+\phi ^2`$ $`=`$ $`{\displaystyle \frac{\kappa _3}{ϵ}}e^\phi [(D2)_\phi UDU],`$ (25)
$`\beta ^{\prime \prime }+3\alpha ^{}\beta ^{}+d\beta ^2+(d1)k_2e^{2\beta }+\beta ^{}\phi ^{}`$ $`=`$ $`{\displaystyle \frac{\kappa _3}{ϵ}}e^\phi [_\phi U+(1+{\displaystyle \frac{1}{2ϵ}})U].`$ (26)
Note also that the $`G_{ij}`$ component of the equation (9) can be deduced from the $`4`$-dimensional Bianchi Identity $`D_aG^{ab}=0`$ associated with the $`4`$-dimensional FRW metric. Hence it is in fact redundant. Therefore, equations (24-26) are in fact a complete set of equations of motion one needs for solving $`\alpha `$, $`\beta `$ and $`\phi `$.
## III Inflationary Universe
If one assumes the slow-rollover approximation, namely, $`\frac{a^{}}{a}|\phi ^{}|`$, one can show that
$`\alpha ^2+d\alpha ^{}\beta ^{}+{\displaystyle \frac{d(d1)}{6}}\beta ^2`$ $`=`$ $`{\displaystyle \frac{U}{3ϵ}}e^\phi ,`$ (27)
$`3\alpha ^{}\beta ^{}+d\beta ^2`$ $`=`$ $`{\displaystyle \frac{\kappa _3}{ϵ}}e^\phi [_\phi U+(1+{\displaystyle \frac{1}{2ϵ}})U],`$ (28)
$`(3\alpha ^{}+d\beta ^{})\phi ^{}`$ $`=`$ $`{\displaystyle \frac{\kappa _3}{ϵ}}e^\phi [(d+4)U(d+2)_\phi U].`$ (29)
Here we have set $`k_1=k_2=0`$ for simplicity. Note that the issue of the non-compact internal space has recently been subject of renewed interest . An exotic class of Kaluza-Klein models in which the internal space is neither compact nor even of finite volume was considered and Gravity is used to trap particles near a four-dimensional subminifold of the higher dimensional spacetime.
Moreover, we have also assumed that $`|\phi ^{\prime \prime }|\alpha ^{}|\phi ^{}|`$ and $`|\beta ^{\prime \prime }|\beta ^2`$. We will show shortly that these assumptions can be met rather easily.
We will assume for the moment during the slow-rollover period that $`\alpha =\alpha _0t`$ and $`\beta =k\alpha _0t`$ for some positive real number $`k`$ and $`\alpha _0`$. This kind of solution represents a brief moment of inflating scale factor $`a`$ accompanied by a contracting internal scale factor $`b`$. This will be helpful in finding possible constraint on the form of symmetry breaking potential one would require for a more realistic model. One can also assume that $`UU_0V(\varphi =\varphi _0)`$ while $`\varphi \varphi _0`$ during the inflationary phase. Therefore, one can show that equations (27-29) can be brought to the following form:
$`d(d1)k^26dk+6`$ $`=`$ $`\stackrel{~}{k},`$ (30)
$`k(dk3)`$ $`=`$ $`{\displaystyle \frac{\stackrel{~}{k}\kappa _3}{4}}(ss_{}),`$ (31)
$`dk3`$ $`=`$ $`{\displaystyle \frac{(d+2)\stackrel{~}{k}\kappa _3\alpha _0}{4\phi _0^{}}}(ss_+).`$ (32)
Here $`\stackrel{~}{k}2U_0/ϵ\alpha _0^2\varphi _0^2`$, $`\phi _0^{}\phi ^{}(t_0)`$, $`s_{}21/ϵ`$ and $`s_+2(d+4)/(d+2)`$. In addition, we have also defined $`s\varphi _0(_\varphi U)_0/U_0`$ as the scaling factor of $`U`$ evaluated at $`\varphi =\varphi _0`$.
We will first study the case where $`d>1`$. Note that equation (30) indicates that $`d(d1)k^26dk+6d(d1)(kk_+)(kk_{})>0`$ under the assumption that $`U`$ is positive everywhere. Here we have defined $`k_\pm 3/(d1)\pm \sqrt{3d(d+2)}/d(d1)`$ as the roots of the $`k`$-equation. Therefore one has either $`k>k_+`$ or $`k<k_{}`$ from the $`k`$-inequality. In addition, equation (31) and equation (32) gives
$$\alpha _0=\frac{ss_{}}{k(d+2)(ss_+)}\phi _0^{}.$$
(33)
This shows that $`\phi _0^{}(ss_+)(ss_{})>0`$ since $`k`$ is assumed to be positive.
Moreover, one also assumes that $`dU(\varphi )/dt<0`$ such that the scalar field is rolling down from some initial value $`\varphi _0`$ to the minimum potential energy state $`\varphi _m`$. This means that $`s\phi _0^{}<0`$. Therefore, one finds that there are only two kinds of combination capable of supporting this process. The first one is (1) $`\phi _0^{}<0`$, $`0<s<s_+`$ and the second one is (2) $`\phi _0^{}>0`$, $`s>s_{}`$. One can further rule out case (2) from the assumption that $`\alpha _0|\phi _0^{}|`$. Indeed, equation (33) indicates that the slow-rollover assumption is equivalent to $`|ss_{}|k(d+2)|ss_+|`$. Hence case (1) can be shown to give a constraint
$$s\frac{k(d+2)s_++s_{}}{k(d+2)+1}.$$
(34)
This can easily be achieved provided that $`s_+s_{}`$. Note that this is true if $`ϵ1`$ . Similarly, case (2) will give a contradictory result $`s_{}s_+`$. Therefore, case (2) is ruled out.
In addition, case (1) and equation (32) shows that $`dk3>0`$. Hence the constraint on $`k`$ obtained earlier is further restricted to the case where $`k>k_+`$. This is because $`3/d<k<k_{}`$ leads to a contradiction $`3>d(d+2)`$.
In short, the induced Kaluza-Klein compactification admits chaotic inflation only if the symmetry breaking potential obeys a number of constraints listed earlier. They are :
$`(a)`$ $`s_+s_{},`$ (35)
$`(b)`$ $`s_+>s[k(d+2)s_++s_{}]/[k(d+2)+1],`$ (36)
$`(c)`$ $`\phi _0^{}<0,`$ (37)
$`(d)`$ $`k>k_+.`$ (38)
For example, one would have (a) $`8/5s_{}`$, (b) $`8/5>s(5k+s_{})/(8k+1)`$ and (c) $`k>1`$ as the constraint on $`k`$ and $`s`$ for the case where $`d=6`$ or equivalently $`D=10`$.
One can easily construct an effective symmetry breaking potential by expanding the potential around the initial point $`\varphi _0`$. Explicitly, it will take the following form $`U=U_0+sU_0(\varphi \varphi _0)+\mathrm{}`$ around the initial point. For example, one can show that the conventional $`\varphi ^4`$ model with $`U=(\lambda /8)(\varphi ^2v^2)^2`$ does not satisfy the constraint obtained earlier. Indeed, one can show that equation (32) gives
$$\mathrm{\Lambda }>\frac{\lambda }{8(d+4)}(\varphi _0^2v^2)[d\varphi _0^2+(d+4)v^2]$$
(39)
for the conventional $`\varphi ^4`$ model with an additional positive definite cosmological constant term $`\mathrm{\Lambda }`$. This clearly shows that the chaotic inflation condition $`\varphi _0^2>v^2`$ is inconsistent with the case where $`\mathrm{\Lambda }=0`$. Note that the no-hair conjecture states that cosmologies with a positive cosmological constant would approach the de Sitter solution asymptotically . Even some counter examples are found, it was shown to hold for a very general conditions . Our result appears to favor above conjecture with the inclusion of the higher dimensional space. Therefore, the conventional $`\varphi ^4`$ model with vanishing cosmological constant can not support an inflationary solution with expanding external-space and contracting internal-space. We will solve the conventional $`\varphi ^4`$ model later in section IV.
For the case where $`d=1`$, the situation is rather different. Equation (30) implies that $`k<1`$ for positive $`U_0`$. In addition, equation (31) gives $`s<s_{}(<0)`$ while equation (32) implies $`\phi _0^{}>0`$ (new inflationary solution). In addition, the slow-rollover assumption indicates that $`s_{}sk(103s)`$. Therefore one obtains $`(3k1)s10ks_{}(>0)`$. This implies that $`k<1/3`$. In summary, one has (a) $`s<s_{}(<0)`$, (b) $`k<1/3`$, and (c) $`\phi _0^{}>0`$. Therefore the five dimensional Kaluza-Klein new inflationary solution with expanding external-space and contracting internal-space can also be arranged if the field parameters are chosen appropriately.
One can also study the case where the internal-space scale factor remains constant, $`i.e.`$ $`b=b_0`$ or equivalently $`k=0`$ in the early universe. In this case, the equations will become
$`\alpha ^2+{\displaystyle \frac{k_1}{a^2}}+{\displaystyle \frac{d(d1)}{6}}{\displaystyle \frac{k_2}{b_0^2}}`$ $`+`$ $`\alpha ^{}\phi ^{}={\displaystyle \frac{1}{24ϵ}}\phi ^2+{\displaystyle \frac{U}{3ϵ}}e^\phi ,`$ (40)
$`\phi ^{\prime \prime }+3\alpha ^{}\phi ^{}+\phi ^2`$ $`=`$ $`{\displaystyle \frac{\kappa _3}{ϵ}}e^\phi [(D2)_\phi UDU],`$ (41)
$`(d1){\displaystyle \frac{k_2}{b_0^2}}`$ $`=`$ $`{\displaystyle \frac{\kappa _3}{ϵ}}e^\phi [_\phi U+({\displaystyle \frac{1}{2ϵ}}+1)U].`$ (42)
Therefore one finds that there is a strong constraint (42) left over for the $`b`$ equation. This equation says that $`_\phi U+(1+\frac{1}{2ϵ})U=0`$ for a flat internal-space (i.e. $`k_2=0`$). One can then show that either (i) the potential $`U`$ has to be a special fractional polynomial functional of $`\varphi `$, namely, $`U=k_0\varphi ^{(2+1/ϵ)}`$ with a proportional constant $`k_0`$, or (ii) the dynamics of the scalar field has to be frozen, namely, the scalar field becomes a constant $`\varphi =\varphi _0`$. One can show that the first case would imply that $`\alpha ^{}\phi ^{}/2`$ under the constraint $`ϵ1`$. This contradicts the slow rollover approximation. On the other hand, the case (ii) implies that $`U(\phi _0)=_\phi U(\phi _0)=0`$. Hence one has $`a^2=k_1`$ due to equation (40). Therefore, one needs $`k_1=1`$ in order to admit a power law inflation. One can hence tune the field parameters to induce enough inflation with expanding external-space and constant internal-space. But this model can not tell us when the inflationary phase should come to an end. One would have to expect that this induced gravity model remains valid only during the inflationary period and leave the problem to other resolutions.
On the other hand, one can show that the constraint (42)
$$(D5)\frac{k_2ϵ}{b_0^2\kappa _3}\varphi ^2=_\phi U+(1+\frac{1}{2ϵ})U$$
(43)
implies $`\varphi =\varphi _0`$ for the case $`k_20`$ unless
$$U=k_0\varphi ^{2\frac{1}{4ϵ}}+\frac{2(D5)k_2ϵ^2}{(1+4ϵ)b_0^2\kappa _3}\varphi ^2.$$
(44)
If $`\varphi =\varphi _0`$, equation (41) implies that
$$\left[(D2)_\phi UDU\right]_{\varphi _0}=0.$$
(45)
Equations (43) and (45) mean that all field parameters and initial conditions are constrained by these equations. In addition, equation (40) tells us that
$$\alpha ^2(D5)\frac{k_2}{3b_0^2}$$
(46)
independent of the form of potential $`U`$. Of course, the initial value of the scalar field $`\varphi _0`$ is determined by the form of potential and the two constraints just derived. This solution is an inflationary solution with expanding external-space and constant internal-space as long as $`k_2(D5)>0`$ and $`b_01`$. One can certainly tune $`b_0`$ to induce enough inflation with expanding external-space and constant internal-space. But this solution can not tell us how to exit the inflationary phase at this point either. One would then have to expect again that this kind of induced gravity model would not remain effective as soon as the inflationary process is completed.
On the other hand equations (40) and (41) imply that
$`\alpha ^2`$ $`=`$ $`(1+6ϵ)A+B,`$ (47)
$`\alpha ^{}\phi ^{}`$ $`=`$ $`4ϵA+2B,`$ (48)
under the slow-rollover approximation if $`U`$ is given by equation (44). Here
$`A`$ $`=`$ $`{\displaystyle \frac{(D5)k_2}{3b_0^2(1+4ϵ)}},`$ (49)
$`B`$ $`=`$ $`{\displaystyle \frac{k_0}{3ϵ\varphi _0^{4+\frac{1}{4ϵ}}}}`$ (50)
for $`\varphi \varphi _0`$ in this inflationary phase. Therefore one can choose $`ϵ1`$ and $`BA`$ in order to be consistent with the assumption that $`\alpha ^{}|\phi ^{}|`$. In addition, one can choose $`A>0`$ since $`\alpha ^2>0`$. This implies that $`k_2=1`$ for $`D5`$. In addition, $`AB`$ implies that $`k_0b_0^2(D5)ϵ\varphi _0^{4+\frac{1}{4ϵ}}`$ which can be achieved by tuning the field parameters appropriately. Moreover, one still needs to make sure that the potential $`U`$ given by equation (44) has at least a local minimum $`\varphi _m`$ far away from the initial data $`\varphi _0`$ such that inflation can exit in due time.
Fortunately, a local minimum always exists for a large class of parameters. Indeed, one can show that
$$k_0b_0^2(D5)(D2)ϵ^2\varphi _m^{4+\frac{1}{4ϵ}}$$
(51)
from $`_\varphi U|_{\varphi =\varphi _m}=0`$. Hence one only needs
$$ϵ\varphi _m^{4+\frac{1}{4ϵ}}\frac{2}{D2}\varphi _0^{4+\frac{1}{4ϵ}}.$$
(52)
In addition, the requirement $`U^{\prime \prime }|_{\varphi _m}>0`$ can be made valid very easily.
Therefore, the inflationary process can properly work with the assumption $`b=b_0`$ for the case where $`F^d=S^{d+1}`$. And this has to come along with the potential of the form given by equation (44).
## IV Conventional $`\varphi ^4`$ Model
One can also works on the model with a spontaneously symmetry breaking (SSB) $`\varphi ^4`$ potential $`U=\frac{\lambda }{8}(e^\phi v^2)^2`$. This sort of potential will be referred to as the conventional $`\varphi ^4`$ model in this paper. It is straightforward to show that $`_\phi U=\frac{\lambda }{4}(e^\phi v^2)e^\phi `$. Hence equation (29) becomes
$$(3\alpha ^{}+d\beta ^{})\phi ^{}=\frac{\kappa _3\lambda }{8ϵ}e^\phi (e^\phi v^2)[de^\phi +(d+4)v^2].$$
(53)
This indicates that $`3\alpha +d\beta `$ is always an increasing function as long as the $`\phi `$ field is rolling down to its true vacuum $`e^\phi =v^2`$. It also indicates that $`\varphi `$ cannot go far away from its local minimum, hence it should oscillate around $`\varphi =v`$ after the inflation is over. We will come back to this point shortly. Moreover, the equation (28) becomes
$$3\alpha ^{}\beta ^{}+d\beta ^2=\frac{2U}{(d+2)ϵ}e^\phi $$
(54)
if $`ϵ1`$ and $`|\varphi ^2v^2|/\varphi ^24ϵ`$. These assumptions can be adjusted rather easily. Together with equation (27), one finds that
$$2\alpha ^2+(d2)\alpha ^{}\beta ^{}d\beta ^2=(\alpha ^{}\beta ^{})(2\alpha ^{}+d\beta ^{})0.$$
(55)
This means that $`\alpha ^{}=\beta ^{}`$ because the equation $`\alpha ^{}=\frac{d}{2}\beta ^{}`$ contradicts the equation (54). Hence $`b(t)`$ increases along with the expanding $`a(t)`$ in the inflationary era under the slow-rollover approximation. Therefore, one has shown that the conventional $`\varphi ^4`$ model supports DFRW space instead of the 4dFRW space. Hence the solution with expanding external-space and contracting internal-space can not be found under the slow-rollover approximation. We will still, however, study the DFRW solution in details in this section for completeness. Note that the presence of a nonvanishing cosmological constant in the conventional $`\varphi ^4`$ model will not affect the equation (55) under the same slow rollover approximation.
Note that equation (54) gives us
$`\alpha ^{}`$ $``$ $`\sqrt{{\displaystyle \frac{\lambda v^4}{8(d+2)(d+3)}}}{\displaystyle \frac{v}{\varphi _0}},`$ (56)
$`a`$ $``$ $`a_0e^{\sqrt{\frac{\lambda v^4}{8(d+2)(d+3)}}\frac{v}{\varphi _0}t}.`$ (57)
Here one has set $`\frac{1}{2}ϵv^2=1`$ such that the gravitational constant measured today is set as $`1`$ in Planck unit. Moreover, $`\varphi _0`$, set to be positive, denotes the initial value of $`\varphi `$ field. Moreover, equation (53) gives
$`\varphi ^{}`$ $``$ $`\sqrt{{\displaystyle \frac{\lambda (d+4)^2}{2(d+2)(d+3)}}}v,`$ (58)
$`\varphi `$ $``$ $`\varphi _0+\sqrt{{\displaystyle \frac{\lambda (d+4)^2}{2(d+2)(d+3)}}}vt.`$ (59)
Here we can see that the assumptions $`|\phi ^{\prime \prime }|\alpha ^{}|\phi ^{}|`$ and $`|\beta ^{\prime \prime }|\beta ^2`$ are both satisfied without imposing any further constraints.
One can further derive a few inequalities from the slow-rollover assumption $`\frac{a^{}}{a}|\phi ^{}|`$. First of all, they give
$$v^24(d+4).$$
(60)
Note that the cosmological constant term $`\frac{1}{8}\lambda v^4`$ at initial time should be less than $`1`$, in Planck unit, in order that quantum effect can be neglected. We will be using Planck unit from now on. In addition, if the scale factor $`a(t)`$ is capable of expanding some $`60`$ e-fold in a time interval of roughly $`\mathrm{\Delta }T10^8`$ Planck unit, one should have the following inequality:
$$\frac{\lambda v^4}{8}(d+2)(d+3)\frac{\varphi _0^2}{v^2}\times 3.6\times 10^{13}.$$
(61)
Inequality (61) can be made valid rather easily. Indeed, these inequalities can be easily satisfied by choosing large $`v^2`$ (hence small $`ϵ`$) and a $`\lambda `$ around the order of $`10^{17}`$ as in Ref. . Hence one shows that the slow-rollover approximation is indeed a good approach to this expanding solution.
Note that we can also extract information about $`\alpha `$, $`\beta `$ and $`\phi `$ when $`\phi v`$ near the end of the expansion. This can be done by analyzing equations (24-26) by assuming $`e^\phi =\xi +v^2`$ with $`\xi v^2`$. Moreover, one can show that equations (25-26) become,
$`_t(e^{3\alpha +d\beta +\phi }\phi ^{})`$ $`=`$ $`{\displaystyle \frac{\kappa _3\lambda }{8ϵ}}e^{3\alpha +d\beta }(e^\phi v^2)[de^\phi +(d+4)v^2],`$ (62)
$`_t(e^{3\alpha +d\beta +\phi }\beta ^{})`$ $`=`$ $`{\displaystyle \frac{\kappa _3\lambda }{8ϵ}}e^{3\alpha +d\beta }(e^\phi v^2)[(3+{\displaystyle \frac{1}{2ϵ}})e^\phi (1+{\displaystyle \frac{1}{2ϵ}})v^2].`$ (63)
Hence equations (62-63) become
$`_t(e^{3\alpha +d\beta }\xi ^{})`$ $`=`$ $`m_v(d+2)e^{3\alpha +d\beta }\xi ,`$ (64)
$`_t[e^{3\alpha +d\beta }\beta ^{}(\xi +v^2)]`$ $`=`$ $`m_ve^{3\alpha +d\beta }\xi ,`$ (65)
as one takes the limit $`e^\phi =\xi +v^2`$. Here $`m_v\kappa _3\frac{\lambda v^4}{8}`$.
Moreover, equations (24-26) can be interpreted as a set of equations that allow one to express $`\alpha (t)`$ and $`\beta (t)`$ as functions of $`\xi (t)`$. Therefore one can expand $`\alpha `$ and $`\beta `$ as polynomials of $`\xi `$, i.e. one can write
$`\alpha (t)`$ $`=`$ $`\alpha _0(t)+\alpha _1(t)\xi (t)+\alpha _2(t)\xi ^2(t)+\mathrm{}`$ (66)
$`\beta (t)`$ $`=`$ $`\beta _0(t)+\beta _1(t)\xi (t)+\beta _2(t)\xi ^2(t)+\mathrm{}.`$ (67)
Therefore, the lowest (first) order in $`\xi `$ of equation (64) is
$$_t(e^{3\alpha _0+d\beta _0}\xi ^{})=m_v(d+2)e^{3\alpha _0+d\beta _0}\xi .$$
(68)
Moreover, the zeroth order in $`\xi `$ of equation (65) can be shown to be
$$_t[e^{3\alpha _0+d\beta _0}\beta _0^{}]=0.$$
(69)
This means that $`_t[e^{d\beta _0}]=\mathrm{const}.\times e^{3\alpha _0}`$. Therefore, one has $`_t[e^{d\beta _0}]0`$ since $`e^{3\alpha _0}`$ is very closed to $`0`$ in the post-expansion era. Therefore, one can assume that $`e^{d\beta _0}`$ is changing very slowly as $`\varphi ^2v^2`$. In addition, the zeroth order in $`\xi `$ of equation (24) is
$$\alpha _0^2+d\alpha _0^{}\beta _0^{}+\frac{d^2d}{6}\beta _0^20.$$
(70)
This gives
$$\alpha _0^{}=\frac{\sqrt{3}d\pm \sqrt{d^2+2d}}{2\sqrt{3}}\beta _0^{}0$$
(71)
to this order of limit. Hence equation (68) becomes an equation for a simple harmonic oscillator
$$\xi ^{\prime \prime }=m_v(d+2)\xi .$$
(72)
Note that the left hand side of equation (72) approaches $`\lambda v^2\xi `$ in the limit $`ϵ1`$. In short, one finds that $`\varphi `$ field indeed oscillates about the local minimum of the symmetry breaking potential $`U`$. Furthermore, equations (71) indicates that $`\alpha _0^{}\beta _0^{}<0`$ as $`\varphi `$ approaches the local minimum of $`U`$. Therefore, $`b(t)`$ in fact starts decreasing if $`a(t)`$ remains increasing at later time. Note that above analysis is only a rough estimate, but it gives us a rough picture of what is going on when $`\xi `$ field approaches zero.
## V Conclusions
In summary, a D-dimensional induced gravity model in 4dFRW space is studied carefully. We present a careful and detailed analysis for the compactification process. This model is then solved for the inflationary solution in the slow-rollover approach. A number of constraints on the symmetry breaking potential are found. These constraints are derived from the search for a inflationary solution with expanding external-space and contracting, compactified internal-space.
The result indicates that the possible form of symmetry breaking potential, prescribed by $`s`$, is constrained by equations (30-32) due to the field equations. Here, $`s\varphi _0(_\varphi U)_0/U_0`$ signifies the scaling factor of $`U`$ evaluated at $`\varphi =\varphi _0`$. The cases where $`d>1`$ and $`d=1`$ are analyzed separately. Explicitly, constraints to the coupled potential are listed in equations (35-38) for the case where $`d>1`$. In particular, one shows that these constraints read (a) $`8/5s_{}`$, (b) $`8/5>s(5k+s_{})/(8k+1)`$ and (c) $`k>1`$ in the limit where $`d=6`$. It was then shown that the conventional $`\varphi ^4`$ model with an additional cosmological constant term fails to satisfy the above constraints. On the other hand, one shows that (a) $`s<s_{}(<0)`$, (b) $`k<1/3`$, and (c) $`\phi _0^{}>0`$ for the case where $`d=1`$. In addition, we also solve the case where the internal scale factor $`b`$ remains constant during the inflationary phase.
An expanding solution is also found and analyzed for the conventional $`\varphi ^4`$ model. In order to generate a solution with expanding external-space inflation in the very early universe, one finds that the internal space is expanding too under the slow rollover approximation. Therefore, this indicates that dimensional reduction has to be completed before expanding external-space starts to expand. With properly chosen free parameters and boundary conditions of the scalar field, one shows that enough expansion can be easily achieved regardless of the negative impact of the expanding internal-space in the conventional $`\varphi ^4`$ model.
Acknowledgments : This work is supported in part by the National Science Council under the contract number NSC88-2112-M009-001.
|
warning/0003/math0003139.html
|
ar5iv
|
text
|
# Consistently there is no non trivial ccc forcing notion with the Sacks or Laver property
## 1. Introduction
At the recent set theory conference, Boban Velickovic asked the following question:
###### 1.1 Question.
Is there a nontrivial forcing notion with the Sacks property which is also ccc?
(See below for a definition of the Sacks property.)
A “definable” variant of this question has been answered in \[Sh 480\]:
> Every nontrivial Souslin forcing notion which has the Sacks property has an uncountable antichain.
(A Souslin forcing notion is a forcing notion for which the set of conditions, the comparability relation and the incompatibility relation are all analytic subsets of the reals. See \[JdSh 292\] and \[Sh 480\] for details).
We show here
###### 1.2 Theorem.
The following statement is equiconsistent with ZFC:
* Every nontrivial forcing notion which has the Sacks property has an uncountable antichain.
Our proof follows the ideas from \[Sh 480\].
Independently, Velickovic has also proved the consistency of $`()`$, following \[Sh 480\] and some of his works. In fact, he shows that the proper forcing axiom (PFA), and even the open coloring axiom implies $`()`$.
Our proof shows that also the following strengthening of $`()`$:
* Every nontrivial forcing notion which has the Laver property has an uncountable antichain.
is equiconsistent with ZFC.
Note that if cov(meagre) $`=`$ continuum (which follows e.g. from PFA) then there is a (non principal) Ramsey ultrafilter on $`\omega `$. The “Mathias” forcing notion for shooting making this ultrafilter principal has the Laver property and is ccc, so $`()`$ does not follow from PFA.
So our result and Velickovic’ result are incomparable.
###### 1.3 Definition.
Let $`g\omega ^\omega `$ be increasing. A $`g`$-slalom is a sequence $`\overline{A}=(A_n:n\omega )`$, $`A_n\omega `$, $`|A_n|g(n)`$. We say that $`\overline{A}`$ covers $`\eta \omega ^\omega `$ iff $`n\eta (n)A_n`$.
Let $`V_1V_2`$ e models of set theory. We say that $`(V_1,V_2)`$ is $`(f,g)`$-bounding iff:
> For all $`\eta _nf(n)V_2`$ there is a $`g`$-slalom in $`V_1`$ covering $`f`$.
We say that a forcing notion $``$ is $`(f,g)`$-bounding if the pair $`(V,V^{})`$ is $`(f,g)`$-bounding
A pair $`(V_1,V_2)`$ has the Laver property iff $`(V_1,V_2)`$ is $`(f,g)`$-bounding for all increasing $`f,gV_1`$, or in other words: For all $`gV_1`$, every function in $`V_2`$ which is bounded by a function in $`V_1`$ is covered by a $`g`$-slalom from $`V_1`$.
Similarly, $``$ has the Laver property iff $`(V,V^{})`$ has the Laver property.
$`(V_1,V_2)`$ has the Sacks property if it has the Laver property, and every function in $`V_2`$ is bounded by a function in $`V_1`$.
## 2. A lemma on Mathias forcing
The following lemma is a theorem of ZFC.
###### 2.1 Lemma.
Let $`𝕄`$ be the Mathias forcing, $`\eta `$ $`\stackrel{~}{}`$ the increasing enumeration of the $`𝕄`$-generic subset of $`\omega `$, and assume that $`T`$ $`\stackrel{~}{}`$ is a name of a tree $`{}_{}{}^{\omega >}2`$ such that
(!)𝕄|T
~
2η
~
(n)|2η
~
(n1)(!)\qquad\qquad\Vdash_{\mathbb{M}}|\mathchoice{\oalign{$\displaystyle T$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\textstyle T$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptscriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{T}}$}\vss}}}\cap{}^{\mathchoice{\oalign{$\displaystyle\eta$\crcr\vbox to0.60275pt{\hbox{$\displaystyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\textstyle\eta$\crcr\vbox to0.60275pt{\hbox{$\textstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptstyle\eta$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptscriptstyle\eta$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}(n)}2|\leq 2^{\mathchoice{\oalign{$\displaystyle\eta$\crcr\vbox to0.60275pt{\hbox{$\displaystyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\textstyle\eta$\crcr\vbox to0.60275pt{\hbox{$\textstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptstyle\eta$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptscriptstyle\eta$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}(n-1)}
Then there is a family $`(T_i,q_i:i<2^\mathrm{}_0)`$ such that for all $`i,j`$:
1. $`q_i𝕄`$
2. $`T_i{}_{}{}^{\omega >}2`$ is a tree
3. qiT
~
Tiforcessubscript𝑞𝑖𝑇
~
subscript𝑇𝑖q_{i}\Vdash\mathchoice{\oalign{$\displaystyle T$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\textstyle T$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptscriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{T}}$}\vss}}}\subseteq T_{i}
4. Whenever $`ij`$, then $`limT_ilimT_j`$ is finite.
(We write $`limT`$ for the set of branches of a tree $`T`$.)
###### Proof.
Conditions in the Mathias forcing are of the form $`(w,A)`$, where $`w`$ is a finite set of natural numbers and $`A`$ is an infinite subset of $`\omega `$ such that $`sup(w)<\mathrm{min}A`$. If $`(w,A)𝕄`$, and $`w=\{w_0<\mathrm{}<w_\mathrm{}1\}`$, then
(w,A)η
~
(i)=wi(i=0,,1),η
~
(i)A(i=,+1,)formulae-sequenceforces𝑤𝐴𝜂
~
𝑖subscript𝑤𝑖𝑖01𝜂
~
𝑖𝐴𝑖1(w,A)\Vdash\mathchoice{\oalign{$\displaystyle\eta$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\textstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptscriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}(i)=w_{i}\ (i=0,\ldots,\ell-1),\qquad\mathchoice{\oalign{$\displaystyle\eta$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\textstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptscriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}(i)\in A\ (i=\ell,\ell+1,\ldots)
Using the fact that truth values in $`V^𝕄`$ can be decided by pure extensions, we can find a condition $`(w^{},A^{})`$ which forces that $`T`$ $`\stackrel{~}{}`$ is “decided continuously” by $`\eta `$ $`\stackrel{~}{}`$ , more specifically:
* there is a function $`t`$ with domain $`[A^{}]^{<\omega }`$, $`w:t(w){}_{}{}^{\omega >}2`$, such that
> For all finite $`uA^{}`$ and all $`n`$ with sup$`(u)<nA^{}`$:
> (wu,A(n+1))T
>
> ~
> δ2n=t(w)2n(w^{*}\cup u,A^{*}\setminus(n+1))\Vdash\mathchoice{\oalign{$\displaystyle T$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\textstyle T$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptscriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{T}}$}\vss}}}_{\delta}\cap{}^{n}2=t(w)\cap{}^{n}2.
For any finite $`uA^{}`$ let
$$T_u=^{df}\{\eta {}_{}{}^{\omega >}2:\text{ for any large enough }k<\omega \text{ we have }(w^{},A^{}k)_{}\text{ “}\eta \begin{array}{c}T\hfill \\ \stackrel{~}{}\hfill \end{array}\text{}\}$$
So $`T_u`$ is a subtree of $`{}_{}{}^{\omega >}2`$.
We have
* If $`w^{}u=\{w_0<\mathrm{}<w_\mathrm{}1\}`$, then (wu,Ak)η
~
(1)=maxuforcessuperscript𝑤𝑢𝐴𝑘𝜂
~
1𝑢(w^{*}\cup u,A\setminus k)\Vdash\mathchoice{\oalign{$\displaystyle\eta$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\textstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptscriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}(\ell-1)=\max u, η
~
()k𝜂
~
𝑘\mathchoice{\oalign{$\displaystyle\eta$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\textstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptscriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}(\ell)\geq k
hence (recall condition (!)): (wu,Ak)|T
~
2k|2η(1)=2maxu(w^{*}\cup u,A\setminus k)\Vdash|\mathchoice{\oalign{$\displaystyle T$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\textstyle T$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptscriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{T}}$}\vss}}}\cap{}^{k}2|\leq 2^{\eta(\ell-1)}=2^{\max u}.
It follows that:
* for every finite $`uA^{}`$, non empty for simplicity, we have
$$k<\omega |T_u{}_{}{}^{k}2|2^{\mathrm{max}(u)},$$
hence lim$`(T_u)`$ is a finite subset of $`{}_{}{}^{\omega }2`$.
We also get:
* if $`u\{m,k\}A^{}`$ and sup$`(u)<m<k`$, then $`T_u{}_{}{}^{m}2=`$ $`T_{u\{k\}}{}_{}{}^{m}2`$.
\[Proof: We know that already $`(w^{}u,Am)`$ decides T
~
2m\mathchoice{\oalign{$\displaystyle T$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\textstyle T$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptscriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{T}}$}\vss}}}\cap{}^{m}2, and the conditions $`(w^{}u\{m\},Ak)`$ and $`(w^{}u,Ak)`$ (for $`k>m`$) are both stronger than $`(w^{}u,Am)`$.\]
In particular, we get:
* for all finite $`uA^{}`$ the sequence $`T_{u\{k\}}:k`$ $`A^{}`$ converges to $`T_u`$.
This means that for every $`m<\omega `$ for every large enough $`kA^{}`$ we have $`T_{u\{k\}}{}_{}{}^{m>}2=`$ $`T_u{}_{}{}^{m>}2.`$
* 1. for $`A`$ an infinite subset of $`A^{}`$ we let
$$T_A=\{T_{An}:n<\omega \}T[A]=\{T_u:uA\text{ finite}\}$$
2. for $`u,v`$ finite subsets of $`A^{}`$ we let $`𝐧(u,v)`$ the smallest $`m`$ such that
+ whenever $`\eta ,\nu `$ are distinct members of $`lim(T_u)lim(T_v)`$ then the length of $`\eta \nu `$ is $`<m`$
+ $`sup(w^{}uv)<m`$
Note that $`𝐧(u,v)`$ is well defined, as both $`lim(T_u)`$ and $`lim(T_v)`$ are finite.
3. for $`m<\omega `$ let
$$𝐧(m)=^{df}\mathrm{max}\{𝐧(u,v):u,vA^{}(m+1)\},$$
so $`𝐧(m)<\omega `$ is well defined being the maximum of a finite set of natural numbers.
* Note that (w,A)T
~
T[A]forcessuperscript𝑤𝐴𝑇
~
𝑇delimited-[]𝐴(w^{*},A)\Vdash\mathchoice{\oalign{$\displaystyle T$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\textstyle T$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptscriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{T}}$}\vss}}}\subseteq T[A].
So without loss of generality (as we can replace $`A^{}`$ by any infinite subset):
* if $`nA^{}`$ then $`𝐧(n)<\mathrm{min}(A^{}(n+1))`$
hence
* if $`nA^{},uA^{}`$ $`(n+1),kA^{}(n+1)`$
then $`T_u{}_{}{}^{𝐧(n)}2=`$ $`T_{u\{k\}}`$ $`{}_{}{}^{𝐧(n)}2.`$
* if $`u,v`$ are finite subsets of $`A^{}`$ and $`sup(uv)<mA^{}`$
then $`T_{u\{m\}}T_v`$ $`T_uT_v`$
Hence
* if $`u,v`$ are finite subsets of $`A^{}`$, not disjoint for notational simplicity, and $`m=sup(uv)`$ then
$$T_uT_vT_{u(m+1)}T_{v(m+1)}$$
\[ why? we can prove this by induction on max$`(uv)`$ using (\*)$`{}_{8}{}^{}]`$
Hence, letting $`Y=\{lim(T_u):`$ $`u`$ a finite subset of $`A^{}\}`$ (=a countable set), we have
* if $`A,B`$ are infinite subsets of $`A^{}`$, with intersection finite non empty
then $`limT[A]limT[B]`$ is included in
$$\{limT_ulimT_v:uA(\mathrm{max}(AB)+1),vB(\mathrm{max}(AB)+1)\}$$
so $`limT[A]limT[B]`$ is is a finite subset of $`Y`$.
\[ why ? just use (\*)<sub>9</sub> and the definition of $`T[A]].`$
Now fix an uncountable family $`(A_i:i<2^\mathrm{}_0)`$ of almost disjoint subsets of $`A^{}`$. Let $`q_i:=(w^{},A_i)`$. Then by $`()_7`$, we have qiT
~
T[Ai]forcessubscript𝑞𝑖𝑇
~
𝑇delimited-[]subscript𝐴𝑖q_{i}\Vdash\mathchoice{\oalign{$\displaystyle T$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\textstyle T$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptscriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{T}}$}\vss}}}\subseteq T[A_{i}]. Hence $`(q_i,A_i:i<2^\mathrm{}_0)`$ satisfies the conditions 1.,2.,3.,4. of lemma 2.1. ∎
## 3. An iteration argument
###### 3.1 Notation.
For any $`f:\omega \omega `$ we define
* $`f^{}:\omega \omega `$ by: $`f^{}(n)=f(n1)`$ for $`n>0`$, $`f^{}(0)=1`$.
* $`\widehat{f}:\omega \omega `$ by: $`\widehat{f}(n)=2^{f(n)}`$.
###### 3.2 Framework.
We will start with a universe where $`2^\mathrm{}_0=\mathrm{}_1`$, $`2^\mathrm{}_1=\mathrm{}_2`$ and use an iteration of length $`\kappa =\mathrm{}_2`$, $`\overline{}=_i,_i:i<\mathrm{}_2`$ satisfying the following:
1. $`\overline{}`$ is a countable support iteration of proper forcing notions.
2. $`_\kappa `$, the union of $`_i`$ for $`i<\kappa `$ , satisfies the $`\kappa `$cc
3. The set
$$S:=\{\delta <\kappa :cf(\delta )>\mathrm{}_0,__i\text{}_\delta \text{=Mathias forcing, with generic real }\begin{array}{c}\eta \hfill \\ \stackrel{~}{}\hfill \end{array}_\delta \text{}\}$$
is stationary.
4. Each forcing notion
~
isubscript
~
𝑖\mathchoice{\oalign{$\displaystyle\mathbb{Q}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{\mathbb{Q}}}$}\vss}}}{\oalign{$\textstyle\mathbb{Q}$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{\mathbb{Q}}}$}\vss}}}{\oalign{$\scriptstyle\mathbb{Q}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{\mathbb{Q}}}$}\vss}}}{\oalign{$\scriptscriptstyle\mathbb{Q}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\mathbb{Q}}}$}\vss}}}_{i} has the Laver property.
5. Whenever $`\alpha <\kappa `$ and $`T`$ $`\stackrel{~}{}`$ is a $`_\alpha `$name of an Aronszajn tree, then for some $`i[\alpha ,\kappa )`$, $`__i`$ “if $`T`$ $`\stackrel{~}{}`$ is a Souslin tree then
~
isubscript
~
𝑖\mathchoice{\oalign{$\displaystyle\mathbb{Q}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{\mathbb{Q}}}$}\vss}}}{\oalign{$\textstyle\mathbb{Q}$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{\mathbb{Q}}}$}\vss}}}{\oalign{$\scriptstyle\mathbb{Q}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{\mathbb{Q}}}$}\vss}}}{\oalign{$\scriptscriptstyle\mathbb{Q}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\mathbb{Q}}}$}\vss}}}_{i} is forcing by $`T`$ $`\stackrel{~}{}`$ (or an isomorphic forcing notion)”
(See the remarks 4.2 and 4.1 for weaker assumptions)
###### 3.3 Fact.
Let $`\overline{}`$ satisfy properties (A)–(E) above, and let $`_\kappa `$ be the CS limit of this iteration. Then:
1. $`_\kappa `$ is proper, making $`\kappa `$ to $`\mathrm{}_2`$.
2. If $`\delta S`$ then the forcing notion $`_\kappa /_{\delta +1}`$ has the Laver property.
###### Proof.
(1) By \[Sh:f\]
(2)By \[Sh:f, ch VI, section 3\]
###### 3.4 Remark.
Assume (say) GCH, then there is a forcing iteration as above. Define $`_i`$ as follows: If $`i`$ is even, then let $`_i`$ be the Mathias forcing, and if $`i`$ is odd, then let $`_i`$ be either trivial or a Souslin tree.
Note that in all intermediate universes we will have GCH, all forcing notions $`_i`$ and $`_i`$ (for $`i<\mathrm{}_2`$) will have a dense subset of size $`\mathrm{}_1`$; this will be sufficient for $`\kappa `$-cc. All the forcing notions $`_i`$ will have the Laver property (recall that a Souslin tree does not add reals), and the usual bookkeeping argument can take care of killing all Souslin trees on $`\omega _1`$.
###### 3.5 Theorem.
Let $`\overline{}`$ satisfy conditions (A)–(E) above.
Then in the universe $`𝕍^_\kappa `$ the following holds:
1. Souslin’s hypothesis
2. Any nontrivial ccc forcing notion adds a real. (See 3.6 below)
3. Whenever $`𝕋`$ is a function satisfying the following conditions $`(\alpha )`$$`(\gamma )`$:
* $`\mathrm{Dom}(𝕋)=\{f:f`$ is a function from $`\omega `$ to $`\omega `$, (strictly) increasing $`\}`$
* $`𝕋(f)`$ is a subtree of $`{}_{}{}^{\omega >}2`$
* for every $`f\mathrm{Dom}(𝕋)`$ and $`n<\omega `$ we have
$$1|𝕋(f){}_{}{}^{f(n)}2|f^{}(n)$$
then $`𝕋`$ also satisfies condition $`(\delta )`$:
* there is a countable subset $`Y{}_{}{}^{\omega }2`$ and an an uncountable subset $`𝐀\mathrm{Dom}(𝕋)`$ with:
> whenever $`fg`$ are from $`A`$ then $`lim(𝕋(f))lim(𝕋(g))`$ in a finite subset of $`Y.`$
###### 3.6 Fact.
If there is a nontrivial ccc forcing which does not add reals, then there is a Souslin tree on $`\omega _1`$.
In other words: If Souslin Hypothesis holds then
> for every ccc forcing $``$ which is not trivial, there are $`p`$ and an $``$-name $`\eta `$ $`\stackrel{~}{}`$ such that:
> $`p_{}`$η
>
> ~
> 2ω{\mathchoice{\oalign{$\displaystyle\eta$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\textstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptscriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}}\in{{}^{\omega}2} is new, that is does not belong to $`𝐕`$
###### Proof.
Let $``$ be a nontrivial ccc forcing. So for some $`q`$ we have
> $`q_{}`$G
>
> ~
> subscript𝐺
>
> ~
> \mathchoice{\oalign{$\displaystyle G$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{G}}$}\vss}}}{\oalign{$\textstyle G$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{G}}$}\vss}}}{\oalign{$\scriptstyle G$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{G}}$}\vss}}}{\oalign{$\scriptscriptstyle G$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{G}}$}\vss}}}_{{\mathbb{R}}} does not belongs to V
Hence for some quadruple (p,α,β,η
~
)𝑝𝛼𝛽𝜂
~
(p,\alpha,\beta,{\mathchoice{\oalign{$\displaystyle\eta$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\textstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptscriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}}) we have:
* $`p,\alpha ,\beta `$ are ordinals, $`\eta `$ $`\stackrel{~}{}`$ is a $``$-name and $`p_{}`$η
~
βα𝜂
~
superscript𝛽𝛼{\mathchoice{\oalign{$\displaystyle\eta$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\textstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptscriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}}\in{}^{\alpha}\beta is not from V
We can choose such quadruple with the ordinal $`\alpha `$ minimal. Necessarily $`\alpha `$ is a limit ordinal and for $`\gamma <\alpha `$, $`\eta `$ $`\stackrel{~}{}`$ $`\gamma `$ is forced by $`p`$ to belong to V.
For $`\gamma <\alpha `$, let
$$T_\gamma =\{\nu :\text{ }\nu \text{ is a function from }\gamma \text{ to }\beta \text{ so }p\text{ does not force that }\nu \begin{array}{c}\eta \hfill \\ \stackrel{~}{}\hfill \end{array}\gamma \}$$
and let $`T=\{T_\gamma :\gamma <\alpha \}`$. Clearly $`T`$ is a tree with $`\alpha `$ levels, and $`p`$ forces that $`\eta `$ $`\stackrel{~}{}`$ is a new $`\alpha `$branch of it.
Now $`T`$ cannot have $`\mathrm{}_1`$ pairwise incomparable elements, as if $`\nu _\zeta `$ for $`\zeta <\omega _1`$ are like that, we can find $`p_\zeta `$ such that: $`pp_\zeta `$ and $`p_\zeta _{}`$$`\nu _\zeta `$ is an initial segment of $`\eta `$ $`\stackrel{~}{}`$ ”; now if $`p_\zeta ,p_\xi `$ are compatible in $``$ then $`\nu _\zeta ,\nu _\xi `$ are comparable in $`T`$ (being, both, the initial segment of some possible $`\eta `$ $`\stackrel{~}{}`$ ). So $`\{p_\zeta :\zeta <\omega _1\}`$ are pairwise incompatible contradiction to $``$ satisfies the ccc”
Also in $`T`$, by its choice, every member has above it elements of every higher level and there is no node above which the tree has no two distinct members of the same level (as then $`\eta `$ $`\stackrel{~}{}`$ will be forced to belongs to V by some condition above $`p)`$.
Also as $``$ satisfies the ccc, every level is countable, and by the minimality of $`\alpha `$ (as we are allowed to change $`\beta )`$ clearly $`\alpha `$ is a regular cardinal. Now $`\alpha >\omega _1`$ is impossible by “$``$ satisfies the $`\kappa `$cc”. As there are no Souslin tree also $`\alpha =\omega _1`$ is impossible. So clearly $`p`$ forces that $``$ add reals so there is $`\eta `$ $`\stackrel{~}{}`$ as required.
###### 3.7 Observation.
Theorem 3.5 suffices to prove 1.2 and its strengthening $`()`$ mentioned in the introduction. That is, conditions (2)&(3) of 3.5 imply:
> Any nontrivial forcing with the Laver property has an uncountable antichain.
###### Proof.
Let $``$ be a forcing notion with the Laver property which adds a real, say pη
~
2ω,η
~
Vp\Vdash\mathchoice{\oalign{$\displaystyle\eta$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\textstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptscriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}\in{}^{\omega}2,\mathchoice{\oalign{$\displaystyle\eta$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\textstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptscriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}\notin V.
Consider any increasing function $`f`$. The function $`n\eta f(n+1)`$ has only $`2^{f(n+1)}`$ many possible values, i.e., is bounded. So, by the Laver property there is a tree $`T_f{}_{}{}^{\omega >}2`$ and a condition $`q_f`$ stronger than $`q`$ with
n|Tf2f(n+1)|f(n),qfη
~
limTfformulae-sequencefor-all𝑛subscript𝑇𝑓superscript2𝑓𝑛1𝑓𝑛forcessubscript𝑞𝑓𝜂
~
subscript𝑇𝑓\forall n\ |T_{f}\cap 2^{f(n+1)}|\leq f(n),\qquad\qquad q_{f}\Vdash\mathchoice{\oalign{$\displaystyle\eta$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\textstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptscriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}\in\lim T_{f}
We have thus defined a family $`𝕋=(T_f:f\text{ increasing })`$. By theorem 3.5, there is a family $`(f_i:i\omega _1)`$ such that
> $`ij:`$ $`limT_{f_i}limT_{f_j}`$ is finite
Clearly, for $`ij`$ the conditions $`q_{f_i}`$ and $`q_{f_j}`$ must be incompatible, since any condition $`r`$ stronger than both would force
rη
~
limTfilimTfisubscriptforces𝑟𝜂
~
subscript𝑇subscript𝑓𝑖subscript𝑇subscript𝑓𝑖r\Vdash_{\mathbb{R}}\mathchoice{\oalign{$\displaystyle\eta$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\textstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptscriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}\in\lim T_{f_{i}}\cap\lim T_{f_{i}}
which implies rη
~
Vforces𝑟𝜂
~
𝑉r\Vdash\mathchoice{\oalign{$\displaystyle\eta$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\textstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptscriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}\in V, a contradiction.
(Remark: While $`limT_{f_i}`$ and $`limT_{f_j}`$ can of course contain branches in $`V^{}`$ which did not exist in $`V`$, the fact that their intersection is a certain finite set is absolute between $`V`$ and $`V^{}`$.) ∎
### Proof of theorem 3.5, part 3
Assume that $`𝕋`$ $`\stackrel{~}{}`$ is a $`_\kappa `$-name such that $`p^{}_\kappa `$ forces “ $`𝕋`$ $`\stackrel{~}{}`$ satisfies $`(\alpha ),(\beta ),(\gamma )`$”.
Without loss of generality (replacing the ground model by an intermediate model $`V[G_\alpha ]`$, $`G_\alpha _\alpha `$, $`p^{}G_\alpha `$, if necessary) we can assume that $`p^{}`$ is really the empty condition.
Let $`S\kappa `$ be unbounded, $`\delta S_\delta `$ is Mathias forcing, with generic real η
~
δsubscript𝜂
~
𝛿\mathchoice{\oalign{$\displaystyle\eta$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\textstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptscriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}_{\delta}.
For every $`\delta S`$ let T
~
δ0=𝕋
~
(η
~
δ0)subscriptsuperscript𝑇
~
0𝛿𝕋
~
subscriptsuperscript𝜂
~
0𝛿\mathchoice{\oalign{$\displaystyle T$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\textstyle T$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptscriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{T}}$}\vss}}}^{0}_{\delta}=\mathchoice{\oalign{$\displaystyle\mathbb{T}$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{\mathbb{T}}}$}\vss}}}{\oalign{$\textstyle\mathbb{T}$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{\mathbb{T}}}$}\vss}}}{\oalign{$\scriptstyle\mathbb{T}$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{\mathbb{T}}}$}\vss}}}{\oalign{$\scriptscriptstyle\mathbb{T}$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\mathbb{T}}}$}\vss}}}(\mathchoice{\oalign{$\displaystyle\eta$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\textstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptscriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}^{0}_{\delta}); clearly :
* $`__\kappa `$T
~
δ0subscriptsuperscript𝑇
~
0𝛿{\mathchoice{\oalign{$\displaystyle T$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\textstyle T$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptscriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{T}}$}\vss}}}}^{0}_{\delta} is a subtree of $`{}_{}{}^{\omega >}2`$ such that (n)|T
~
δ02η
~
δ(n)|η
~
δ(n)(\forall n)|{\mathchoice{\oalign{$\displaystyle T$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\textstyle T$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptscriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{T}}$}\vss}}}}^{0}_{\delta}\cap{}^{{\mathchoice{\oalign{$\displaystyle\eta$\crcr\vbox to0.60275pt{\hbox{$\displaystyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\textstyle\eta$\crcr\vbox to0.60275pt{\hbox{$\textstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptstyle\eta$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptscriptstyle\eta$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}}_{\delta}(n)}2|\leq{\mathchoice{\oalign{$\displaystyle\eta$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\textstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptscriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}}^{-}_{\delta}(n).”
There is only a bounded number of possibilities for T
~
δ02η
~
δ(n){\mathchoice{\oalign{$\displaystyle T$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\textstyle T$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptscriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{T}}$}\vss}}}}^{0}_{\delta}\cap{}^{{\mathchoice{\oalign{$\displaystyle\eta$\crcr\vbox to0.60275pt{\hbox{$\displaystyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\textstyle\eta$\crcr\vbox to0.60275pt{\hbox{$\textstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptstyle\eta$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptscriptstyle\eta$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}}_{\delta}(n)}2, so since $`_\kappa /_{\delta +1}`$ has the Laver property, we can find a pair (pδ,T
~
δ)subscript𝑝𝛿subscript𝑇
~
𝛿(p_{\delta},{\mathchoice{\oalign{$\displaystyle T$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\textstyle T$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptscriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{T}}$}\vss}}}}_{\delta}) satisfying
* 1. $`p_\delta _\kappa `$
2. T
~
δsubscript𝑇
~
𝛿{\mathchoice{\oalign{$\displaystyle T$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\textstyle T$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptscriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{T}}$}\vss}}}}_{\delta} is a $`_{\delta +1}`$name
3. $`_{_{\delta +1}}`$T
~
δsubscript𝑇
~
𝛿{\mathchoice{\oalign{$\displaystyle T$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\textstyle T$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptscriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{T}}$}\vss}}}}_{\delta} is a subtree of $`{}_{}{}^{\omega >}2`$ and $`n<\omega `$ |T
~
δ2η
~
δ(n)||{\mathchoice{\oalign{$\displaystyle T$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\textstyle T$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptscriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{T}}$}\vss}}}}_{\delta}\cap{}^{{\mathchoice{\oalign{$\displaystyle\eta$\crcr\vbox to0.60275pt{\hbox{$\displaystyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\textstyle\eta$\crcr\vbox to0.60275pt{\hbox{$\textstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptstyle\eta$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptscriptstyle\eta$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}}_{\delta}(n)}2| η
~
^δ(n)absentsubscriptsuperscript^𝜂
~
𝛿𝑛\leq\hat{\mathchoice{\oalign{$\displaystyle\eta$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\textstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptscriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}}^{-}_{\delta}(n)
4. $`_{}`$T
~
δ0subscriptsuperscript𝑇
~
0𝛿{\mathchoice{\oalign{$\displaystyle T$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\textstyle T$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptscriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{T}}$}\vss}}}}^{0}_{\delta} T
~
δabsentsubscript𝑇
~
𝛿\subseteq{\mathchoice{\oalign{$\displaystyle T$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\textstyle T$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptscriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{T}}$}\vss}}}}_{\delta}
So there is stationary subset $`S_1S`$ and a condition $`q_1_\kappa `$ such that $`\delta S_1p_\delta \delta =q_1.`$ (Again we may assume that $`q_1`$ is the trivial condition.)
Possibly increasing $`p_\delta (\delta )`$ we can find a $`_\delta `$-name such that $`p_\delta \delta `$ forces:
* Above $`p_\delta (\delta )`$, the $`_\delta `$-name T
~
δsubscript𝑇
~
𝛿{\mathchoice{\oalign{$\displaystyle T$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\textstyle T$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptscriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{T}}$}\vss}}}}_{\delta} can be read continuously from $`\eta `$ $`\stackrel{~}{}`$ <sub>δ</sub> as in $`()_1`$, through the function t
~
δsubscript𝑡
~
𝛿\mathchoice{\oalign{$\displaystyle t$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{t}}$}\vss}}}{\oalign{$\textstyle t$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{t}}$}\vss}}}{\oalign{$\scriptstyle t$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{t}}$}\vss}}}{\oalign{$\scriptscriptstyle t$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{t}}$}\vss}}}_{\delta}).
For $`\delta S_1,p_\delta (\delta )`$, t
~
δsubscript𝑡
~
𝛿\mathchoice{\oalign{$\displaystyle t$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{t}}$}\vss}}}{\oalign{$\textstyle t$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{t}}$}\vss}}}{\oalign{$\scriptstyle t$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{t}}$}\vss}}}{\oalign{$\scriptscriptstyle t$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{t}}$}\vss}}}_{\delta} $`T`$ $`\stackrel{~}{}`$ <sub>δ</sub> are members of $`(\mathrm{}_1)^{𝕍[G_\delta ]}`$. So we can find $`q_\delta p_\delta \delta `$ forcing $`p_\delta (\delta )`$, t
~
δsubscript𝑡
~
𝛿\mathchoice{\oalign{$\displaystyle t$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{t}}$}\vss}}}{\oalign{$\textstyle t$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{t}}$}\vss}}}{\oalign{$\scriptstyle t$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{t}}$}\vss}}}{\oalign{$\scriptscriptstyle t$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{t}}$}\vss}}}_{\delta} and T
~
δsubscript𝑇
~
𝛿\mathchoice{\oalign{$\displaystyle T$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\textstyle T$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptscriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{T}}$}\vss}}}_{\delta} to be equal to hereditarily countable $`_\delta `$-names $`p_\delta ^{}(\delta )`$, t
~
δsubscriptsuperscript𝑡
~
𝛿\mathchoice{\oalign{$\displaystyle t$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{t}}$}\vss}}}{\oalign{$\textstyle t$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{t}}$}\vss}}}{\oalign{$\scriptstyle t$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{t}}$}\vss}}}{\oalign{$\scriptscriptstyle t$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{t}}$}\vss}}}^{\prime}_{\delta} and T
~
δsubscriptsuperscript𝑇
~
𝛿\mathchoice{\oalign{$\displaystyle T$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\textstyle T$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptscriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{T}}$}\vss}}}^{\prime}_{\delta}. \[Here, “hereditarily countable” is taken in the sense of \[Sh:f, III 4.1A\].\]
Since $`cf(\delta )>\mathrm{}_0`$ for $`\delta S_1`$ we can find a stationary subset $`S_2S_1`$ on which $`p_\delta ^{}(\delta )`$, t
~
δsubscriptsuperscript𝑡
~
𝛿\mathchoice{\oalign{$\displaystyle t$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{t}}$}\vss}}}{\oalign{$\textstyle t$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{t}}$}\vss}}}{\oalign{$\scriptstyle t$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{t}}$}\vss}}}{\oalign{$\scriptscriptstyle t$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{t}}$}\vss}}}^{\prime}_{\delta} and T
~
δsubscriptsuperscript𝑇
~
𝛿\mathchoice{\oalign{$\displaystyle T$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\textstyle T$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptscriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{T}}$}\vss}}}^{\prime}_{\delta} are all constant, say with values $`p^{}`$, t
~
superscript𝑡
~
\mathchoice{\oalign{$\displaystyle t$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{t}}$}\vss}}}{\oalign{$\textstyle t$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{t}}$}\vss}}}{\oalign{$\scriptstyle t$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{t}}$}\vss}}}{\oalign{$\scriptscriptstyle t$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{t}}$}\vss}}}^{*} and T
~
superscript𝑇
~
\mathchoice{\oalign{$\displaystyle T$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\textstyle T$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptscriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{T}}$}\vss}}}^{*}. Again we change our base universe to some intermediate universe so that T
~
superscript𝑇
~
\mathchoice{\oalign{$\displaystyle T$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\textstyle T$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptscriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{T}}$}\vss}}}^{*} is now a $`𝕄`$-name, and $`t^{}`$ is an actual function, and $`p^{}=(w^{},A^{})𝕄`$.
We now use our main lemma 2.1 to find an almost disjoint family $`(A_i:i\omega _1)`$ and $`(T_i:i\omega _1)`$ such that $`A_iA^{}`$, (w,Ai)𝕄T
~
T[Ai]subscriptforces𝕄superscript𝑤subscript𝐴𝑖superscript𝑇
~
𝑇delimited-[]subscript𝐴𝑖(w^{*},A_{i})\Vdash_{\mathbb{M}}\mathchoice{\oalign{$\displaystyle T$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\textstyle T$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptscriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{T}}$}\vss}}}^{*}\subseteq T[A_{i}].
Note because of $`()_1`$ this relation can already be computed from $`t^{}`$, so we also have:
* $`\delta S_2`$ $`__\delta `$qiδT
~
δT[Ai]subscriptforcessubscript𝛿subscript𝑞𝑖subscript𝑇
~
𝛿𝑇delimited-[]subscript𝐴𝑖q_{i}\Vdash_{{\mathbb{Q}}_{\delta}}\mathchoice{\oalign{$\displaystyle T$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\textstyle T$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{T}}$}\vss}}}{\oalign{$\scriptscriptstyle T$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{T}}$}\vss}}}_{\delta}\subseteq T[A_{i}]”.
Now consider the model $`V[G_\kappa ]`$. A density argument shows that
* $`i:\{\delta S_1:q_iG__\delta \}\mathrm{}`$
So for all $`i`$ there is $`\delta =\delta (i)`$ with $`𝕋(\eta _{\delta (i)})T[A_i]`$. Letting
$$𝐀:=\{\delta _i:i<\omega _1\}$$
we have found an uncountable family as required.
## 4. Refinements
Theorem 3.5 answers the original question, but essentially the same proof gives a somewhat stronger theorem. The following remarks point a few places where assumptions can be weakened or conclusions strengthened. We leave the details to the reader.
###### 4.1 Remark.
$`2^{2^\mathrm{}_0}=\mathrm{}_2`$ in the ground model is not necessary. The length of our iteration can be any regular cardinal $`\kappa `$ satisfying $`\mu ^\mathrm{}_0<\kappa `$ for all $`\mu <\kappa `$. In the final model we will have $`2^\mathrm{}_0=\mathrm{}_2=\kappa `$.
###### 4.2 Remark.
It is not necessary that all forcing notions have the Laver property. All we need is that $`_\kappa /P_{\delta +1}`$ is (η
~
^δ,η
~
^δ/η
~
δ)subscript^𝜂
~
𝛿subscriptsuperscript^𝜂
~
𝛿superscriptsubscript𝜂
~
𝛿(\hat{\mathchoice{\oalign{$\displaystyle\eta$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\textstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptscriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}}_{\delta},\hat{\mathchoice{\oalign{$\displaystyle\eta$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\textstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptscriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}}^{-}_{\delta}/\mathchoice{\oalign{$\displaystyle\eta$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\textstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptscriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}_{\delta}^{-})-bounding, which can be ensured by a slightly stronger condition on all the $`_i`$, $`\delta <i<\kappa `$.
###### 4.3 Remark.
We showed that in our model every forcing notion with the Laver property which adds reals will have an uncountable antichain. We can strengthen this conclusion by remarking that such forcing notions will actually have an antichain of size $`\kappa =\mathrm{}_2`$.
###### Proof.
Recall the construction of the almost disjoint family after condition $`()_{12}`$, which was used in $`()_{17}`$. Instead of using an almost disjoint family of size $`\mathrm{}_1`$ in the intermediate model we can use a $`_\kappa `$-name of an almost disjoint family of size continuum: Identify the set $`A^{}`$ there with $`{}_{}{}^{\omega >}2`$, then every $`_\alpha `$-name $`\rho `$ $`\stackrel{~}{}`$ of an element of $`{}_{}{}^{\omega }2`$ will induce a set Aρ
~
Asubscript𝐴𝜌
~
superscript𝐴A_{\mathchoice{\oalign{$\displaystyle\rho$\crcr\vbox to0.60275pt{\hbox{$\displaystyle\widetilde{\hphantom{\rho}}$}\vss}}}{\oalign{$\textstyle\rho$\crcr\vbox to0.60275pt{\hbox{$\textstyle\widetilde{\hphantom{\rho}}$}\vss}}}{\oalign{$\scriptstyle\rho$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle\widetilde{\hphantom{\rho}}$}\vss}}}{\oalign{$\scriptscriptstyle\rho$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\rho}}$}\vss}}}}\subseteq A^{*}. Clearly,
κρ
~
1ρ
~
2Aρ
~
1Aρ
~
2 finite.subscriptforcessubscript𝜅absentsubscript𝜌
~
1subscript𝜌
~
2subscript𝐴subscript𝜌
~
1subscript𝐴subscript𝜌
~
2 finite\Vdash_{{\mathbb{P}}_{\kappa}}\mathchoice{\oalign{$\displaystyle\rho$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{\rho}}$}\vss}}}{\oalign{$\textstyle\rho$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{\rho}}$}\vss}}}{\oalign{$\scriptstyle\rho$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{\rho}}$}\vss}}}{\oalign{$\scriptscriptstyle\rho$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\rho}}$}\vss}}}_{1}\not=\mathchoice{\oalign{$\displaystyle\rho$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{\rho}}$}\vss}}}{\oalign{$\textstyle\rho$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{\rho}}$}\vss}}}{\oalign{$\scriptstyle\rho$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{\rho}}$}\vss}}}{\oalign{$\scriptscriptstyle\rho$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\rho}}$}\vss}}}_{2}\Rightarrow A_{\mathchoice{\oalign{$\displaystyle\rho$\crcr\vbox to0.60275pt{\hbox{$\displaystyle\widetilde{\hphantom{\rho}}$}\vss}}}{\oalign{$\textstyle\rho$\crcr\vbox to0.60275pt{\hbox{$\textstyle\widetilde{\hphantom{\rho}}$}\vss}}}{\oalign{$\scriptstyle\rho$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle\widetilde{\hphantom{\rho}}$}\vss}}}{\oalign{$\scriptscriptstyle\rho$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\rho}}$}\vss}}}_{1}}\cap A_{\mathchoice{\oalign{$\displaystyle\rho$\crcr\vbox to0.60275pt{\hbox{$\displaystyle\widetilde{\hphantom{\rho}}$}\vss}}}{\oalign{$\textstyle\rho$\crcr\vbox to0.60275pt{\hbox{$\textstyle\widetilde{\hphantom{\rho}}$}\vss}}}{\oalign{$\scriptstyle\rho$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle\widetilde{\hphantom{\rho}}$}\vss}}}{\oalign{$\scriptscriptstyle\rho$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\rho}}$}\vss}}}_{2}}\mbox{ finite}.
As before, a density argument ensures that there will be $`\kappa `$ many different functions $`\rho `$ $`\stackrel{~}{}`$ such that (w,Aρ
~
)superscript𝑤subscript𝐴𝜌
~
(w^{*},A_{\mathchoice{\oalign{$\displaystyle\rho$\crcr\vbox to0.60275pt{\hbox{$\displaystyle\widetilde{\hphantom{\rho}}$}\vss}}}{\oalign{$\textstyle\rho$\crcr\vbox to0.60275pt{\hbox{$\textstyle\widetilde{\hphantom{\rho}}$}\vss}}}{\oalign{$\scriptstyle\rho$\crcr\vbox to0.60275pt{\hbox{$\scriptstyle\widetilde{\hphantom{\rho}}$}\vss}}}{\oalign{$\scriptscriptstyle\rho$\crcr\vbox to0.60275pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\rho}}$}\vss}}}}) appears in one of the generic Mathias filters for some $`_\delta =𝕄`$, so we can strengthen the conclusion in 3.5, 3$`(\delta )`$ to get a $`\kappa `$-size set $`𝐀`$ rather than just an uncountable one. ∎
###### 4.4 Remark.
We do not need that all forcing notions $`_i`$ have size at most $`\mathrm{}_1`$, there are also weaker conditions (e.g. $`\kappa `$-pic, see \[Sh:f, Ch VIII\]) that will ensure $`\kappa `$-cc of $`_\kappa `$.
For example, instead of forcing only with Souslin trees in the odd stages we can use the forcing from \[Sh:f, Ch V, Section 6\], it specializes the tree (so we can specialize all Aronszajn trees). Here we can prove the $`\kappa `$-cc using the $`\kappa `$-pic condition. ,
###### 4.5 Remark.
Finally, in $`V^_\kappa `$ we can strengthen the conclusion
> Every ccc nontrivial forcing fails the Laver property
as follows:
> For every ccc forcing notion $``$, whenever $`\eta `$ $`\stackrel{~}{}`$ is an $``$-name of a new member of $`{}_{}{}^{\omega }2`$ and $`h`$ is a strictly increasing function from $`\omega `$ to $`\omega ,\mathrm{then}`$ we can find an increasing sequence $`n_i:i<\omega `$ of natural numbers such that $`h(n_i)<n_{i+1}`$ and for no $`p,T`$ do we have:
>
> > $`p,T`$ a subtree of $`{}_{}{}^{\omega >}2`$, $`p_{}`$η
> >
> > ~
> > lim(T)𝜂
> >
> > ~
> > 𝑇\mathchoice{\oalign{$\displaystyle\eta$\crcr\vbox to0.86108pt{\hbox{$\displaystyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\textstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\textstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}{\oalign{$\scriptscriptstyle\eta$\crcr\vbox to0.86108pt{\hbox{$\scriptscriptstyle\widetilde{\hphantom{\eta}}$}\vss}}}\in\lim(T) and for every $`i<\omega `$ we have $`|T{}_{}{}^{n_{2i+1}}2|`$ $`h(n_{2i})`$
The proof is similar to the proof above.
|
warning/0003/astro-ph0003093.html
|
ar5iv
|
text
|
# Electron energy losses near pulsar polar caps: a Monte Carlo approach
## 1 Introduction
Curvature radiation (CR) and magnetic inverse Compton scattering (ICS) are usually considered to be the most natural ways of hard gamma-rays production operating at the expense of pulsar rotational energy (eg. Zhang & Harding zh (2000), and references therein). These two radiation mechanisms dominate within two different ranges of Lorentz factors $`\gamma `$ of beam particles (ie. those leaving the polar cap). When $`\gamma <10^6`$, magnetic inverse Compton scattering plays a dominant role in braking beam particles (Xia et al. xqwh (1985); Chang chang (1995); Sturner sturner (1995)) and is the main source of hard gamma-ray photons (Sturner & Dermer sd (1994); Sturner et al. sdm (1995)).
When $`\gamma >10^6`$, the curvature radiation becomes responsible for cooling beam particles (eg. Daugherty & Harding 1982), yet, the inverse Compton scattering may still be important for secondary $`e^\pm `$-pair plasma. Zhang & Harding (zh (2000)) show that different dependencies of X-ray and gamma-ray luminosities on the pulsar spin-down luminosity as inferred from observations by CGRO, ROSAT and ASCA (Arons arons (1996); Becker & Trümper bt (1997); Saito et al. skks (1998), respectively) can be well reproduced when this effect is included.
So far, the influence of ICS on electron energy have been investigated most thoroughly by Chang (chang (1995)), Sturner (sturner (1995)), Harding & Muslimov (hm (1998)), and Supper & Trümper (supper (2000)). One of several interesting results they found was that energy losses due to resonant ICS can limit the Lorentz factors $`\gamma `$ of electrons to a value which depends on both electric field strength $`E_{}`$, temperature $`T`$ and radius $`R_{\mathrm{th}}`$ of thermal polar cap, and on the magnetic field strength $`B_{\mathrm{pc}}`$ at a polar cap. For example Sturner (sturner (1995)) shows that assuming the acceleration model of Michel (michel (1974)) and a thermal polar cap size comparable to that determined by the open field lines, the Lorentz factors $`\gamma `$ are limited to $`10^3`$ if $`B_{\mathrm{pc}}>10^{13}`$ G, and $`T>3\times 10^6`$ K. This acceleration stopping effect becomes more efficient for stronger magnetic fields, and it led Sturner (sturner (1995)) to propose it as a possible explanation for an apparent cutoff around 10 MeV in gamma radiation of B1509$-$58 (Kuiper et al. kuiper (1999)).
Those results were obtained with a numerical method which assumes smooth changes in electron energy and makes use of expressions for *averaged* energy losses due to the magnetic ICS (Dermer dermer (1990)). The continuous changes of the Lorentz factor with height as determined with this fully deterministic treatment are considered as representative for a behaviour of a large number of electrons. Actually however, electrons lose their energy in discontinuous scattering events. Between some of these events (which *occasionally* can be very distant even for a short mean free path) the energy of an accelerated electron may increase considerably (even a few times) and therefore, the electron can find itself in quite different conditions than if it were losing its energy continuously (the mean free path depends very strongly on $`\gamma `$). Since the energy loss rate due to the resonant ICS is not a monotonic function of $`\gamma `$ this effect may be essential for future behaviour of the particle.
The aim of our paper is to investigate the energetics of electrons on individual basis with a Monte Carlo treatment. In particular, a question of the ICS-limited acceleration is adressed. In Section 2 we calculate electron energy losses near a neutron star using both, the Monte Carlo approach and then the smooth method (after Dermer 1990; Chang 1995; Sturner 1995) for comparison. We consider two distinct cases of electron acceleration: 1) in the constant electric field, proposed by Michel (michel (1974)) (hereafter M74); 2) in the electric field elaborated by Harding & Muslimov (hm (1998)) (hereafter HM98). Section 3 presents the results for electron energetics, and the finding that within some range of pulsars parameters, the averaged Monte Carlo behaviour of electrons does not coincide with a solution found with the smooth method. It also contains the explanation of these differences as well as consequences for predicted efficiencies of gamma-ray emission. Conclusions are given in Sect. 4.
## 2 The model
We consider electrons accelerated in a longitudinal electric field induced rotationally within the region adjacent to the surface of a neutron star. The particles lose their energy due to scattering off soft thermal X-ray photons through magnetic inverse Compton mechanism and (marginally) due to emission of curvature radiation. The thermal photons are assumed to originate from a flat ‘thermal polar cap’ with a temperature $`T\mathrm{a}\mathrm{few}\times 10^6`$ K and with a radius $`R_{\mathrm{th}}\mathrm{a}\mathrm{few}\times r_{\mathrm{pc}}`$ where $`r_{\mathrm{pc}}=(2\pi R_{\mathrm{ns}}^3/cP)^{1/2}`$ is the standard polar cap radius as determined for an aligned rotator by the open lines of purely dipolar magnetic field; ($`R_{\mathrm{ns}}`$ denotes the neutron star radius and $`P`$ is the pulsar period).
Since the ICS losses are to dominate over the curvature losses, one has to ensure that electrons do not attain extremely high Lorentz factors ($`\gamma <10^6`$). Therefore, either the electric field should be weak or at least the size of the accelerating region should be small enough to prevent $`\gamma >10^6`$. For this reason most papers focusing on the resonant ICS made use of a relatively weak electric field after Michel michel (1974) (for example: Sturner sturner (1995) and recently Supper & Trümper supper (2000)). Michel’s model is considered as unrealistic since it ignores magnetic field line curvature (detailed treatment of this problem was first presented by Arons & Scharlemann as (1979)) as well as inertial frame dragging (its importance was acknowledged and the effect was worked out for the first time by Muslimov & Tsygan mt (1992)). Both effects lead to significantly stronger electric fields operating in polar-cap accelerators.
However, for the sake of comparison with previous papers on the resonant ICS we first consider the weak electric field after Michel michel (1974). Accordingly, we assume that electrons are accelerated in a constant $`E_{}`$ extending between stellar surface ($`h=0`$) and altitude $`h_{\mathrm{acc}}`$ :
$$E_{}=\{\begin{array}{c}\left(\frac{8\pi mcB_{\mathrm{pc}}}{eP}\right)^{1/2}\hfill \\ \text{ }3.5810^5\left(\frac{B_{12}}{P}\right)^{1/2}\mathrm{V}\mathrm{cm}^1,\hfill \\ \text{ }\mathrm{for}0hh_{\mathrm{acc}};\hfill \\ \text{ }0,\text{ }\mathrm{for}h>h_{\mathrm{acc}}.\hfill \end{array}$$
(1)
with $`h_{\mathrm{acc}}=r_{\mathrm{pc}}`$, and where $`P`$ is in seconds, $`B_{12}`$ is $`B_{\mathrm{pc}}`$ in Teragauss. Fixed location of $`h_{\mathrm{acc}}`$ means that the model is not self-consistent in the sense that the accelerating field will not be shorted out by a pair formation front (should it occur at a height below $`h_{\mathrm{acc}}=r_{\mathrm{pc}}`$).
As the second case for $`E_{}`$ we take the advanced model of an electric field in the form elaborated by Harding & Muslimov (hm (1998)). HM98 were the first to incorporate electron energy losses due to the ICS in a strong-$`E_{}`$ acceleration model. However, this was done to introduce a high altitude accelerator with negligible resonant ICS losses (an idea of an unstable ICS-induced pair formation fronts). In that work Harding and Muslimov also calculated self-consistent values of the acceleration height $`h_{\mathrm{acc}}`$ using an accelerating electric field which takes into account the upper pair formation front (eqs. (18) and (23) in HM98).
For $`h_{\mathrm{acc}}<0.1r_{\mathrm{pc}}`$ the field is well approximated with
$`E_{}(h)3{\displaystyle \frac{\mathrm{\Omega }R_{\mathrm{ns}}}{c}}{\displaystyle \frac{B_{\mathrm{pc}}}{1ϵ_{\mathrm{GR}}}}{\displaystyle \frac{h}{R_{\mathrm{ns}}}}{\displaystyle \frac{h_{\mathrm{acc}}}{R_{\mathrm{ns}}}}(1{\displaystyle \frac{h}{h_{\mathrm{acc}}}})\times `$ (2)
$`\text{ }\times \left[\kappa \mathrm{cos}\chi +A\mathrm{sin}\chi \right],`$
where $`\mathrm{\Omega }=2\pi /P`$, $`(1ϵ_{\mathrm{GR}})0.6`$, $`\kappa 0.15`$, $`\chi `$ is the angle between the magnetic dipole and rotation axes, $`h`$ is the altitude, and the factor $`A\mathrm{sin}\chi `$ is negligible for nonorthogonal rotators (see HM98 for details). For $`h_{\mathrm{acc}}`$ approaching $`r_{\mathrm{pc}}`$ the linear dependence on $`h_{\mathrm{acc}}`$ in Eq.(2) disappears. We have found that for $`0.5r_{\mathrm{pc}}<h_{\mathrm{acc}}<3r_{\mathrm{pc}}`$ and $`h<r_{\mathrm{pc}}/3`$ the electric field given by eq.(18) in HM98 is well aproximated with a formula
$`E_{}(h)3{\displaystyle \frac{\mathrm{\Omega }R_{\mathrm{ns}}}{c}}{\displaystyle \frac{B_{\mathrm{pc}}}{(1ϵ_{\mathrm{GR}})^{1/2}}}{\displaystyle \frac{h}{R_{\mathrm{ns}}}}{\displaystyle \frac{r_{\mathrm{pc}}}{R_{\mathrm{ns}}}}(1{\displaystyle \frac{h}{h_{\mathrm{acc}}}})\times `$ (3)
$`\text{ }\times \left[\kappa \mathrm{cos}\chi +A\mathrm{sin}\chi \right].`$
For purposes of comparison as well as to simplify our calculations, we inject electrons from the center of polar cap and propagate them along the straight magnetic axis field line. The propagation proceeds in steps of a size $`dh`$, which is one hundred times smaller than any of characteristic length scales involved in the problem, like the acceleration length scale or the mean free path for the magnetic ISC. In each of the distance steps the energy of an electron is increased by a value $`eE_{}dh`$. For the sake of completeness, we also include the energy loss due to the curvature radiation $`|\dot{\gamma }_{\mathrm{cr}}|mc^2dh/v`$ where $`\dot{\gamma }_{\mathrm{cr}}`$ is the CR cooling rate given by
$$\dot{\gamma }_{\mathrm{cr}}=\frac{2e^2\beta ^4\gamma ^4}{3mc\rho _{\mathrm{curv}}^2}$$
(4)
and $`v=c\beta `$ is the electron velocity. Nevertheless, the CR is a very inefficient cooling mechanism in the presence of acceleration models we consider below and it could be neglected equally well. When computing energy losses due to the CR we assume artificially (after Sturner sturner (1995)) that the magnetic axis has a fixed radius of curvature $`\rho _{\mathrm{curv}}=10^7`$ cm, though we keep the electrons all the time directly over the polar cap center.
Magnetic inverse Compton scattering has been treated with a Monte Carlo simulations. The framework of our numerical code is based on an approach proposed by Daugherty & Harding (dh89 (1989)). We improve their method of sampling the parameters of incoming photon and account for the Klein-Nishina regime in an approximate way.
At each step in the electron trajectory, the optical depth for magnetic ICS is calculated as $`d\tau =dh/c`$, where $``$ is a scattering rate, to decide whether a scattering event is to occur. The scattering rate $``$ in the observer frame OF is calculated as
$$=c𝑑\mathrm{\Omega }𝑑\sigma \left(\frac{dn_{\mathrm{ph}}}{dd\mathrm{\Omega }}\right)(1\beta \mu )$$
(5)
(eg. Ho & Epstein he (1989)) where $`\mathrm{\Omega }=d\mu d\varphi `$ is the solid angle subtended by the source of soft photons, ($`\mu =\mathrm{cos}\theta `$), $`\sigma `$ is a total cross section (see below), and $`dn_{\mathrm{ph}}/d/d\mathrm{\Omega }`$ is the density of the soft photons per unit energy and per unit solid angle. The symbol $``$ denotes photon energy in dimensional energy units. Hereafter we will use its dimensionless counterpart $`ϵ=/(mc^2)`$ to denote the photon energy in the observer frame OF and the primed symbol $`ϵ^{}=ϵ\gamma (1\beta \mu )`$ in the electron rest frame ERF.
For temperatures of the thermal polar cap and Lorentz factors of electrons considered below, photon energies $`ϵ^{}`$ may fall well above $`ϵ_B=B/B_{\mathrm{cr}}`$, a local magnetic field strength in units of the critical magnetic field $`B_{\mathrm{cr}}=m^2c^3(e\mathrm{})^1`$. This suggests a full form of the relativistic cross section for Compton scattering in strong magnetic fields to be used (Daugherty & Harding dh86 (1986)). However, incoming photons that propagate in the OF at an angle $`\theta =\mathrm{cos}^1\mu `$ are strongly collimated in the ERF with a cosine of polar angle $`\mu ^{}=(\mu \beta )/(1\beta \mu )`$ close to $`1`$. As was shown by Daugherty & Harding (dh86 (1986)), when $`|\mu ^{}|`$ approaches $`1`$, resonances at higher harmonics become narrower and weaker, and scattering into higher Landau states becomes less important. In such conditions, the polarization-averaged relativistic magnetic cross section in the Thomson regime is reasonably well approximated with a nonrelativistic, classical limit:
$$\sigma =\frac{\sigma _T}{2}\left(1\mu _{}^{}{}_{}{}^{2}+(1+\mu _{}^{}{}_{}{}^{2})\left[g_1+\frac{g_2g_1}{2}\right]\right)$$
(6)
where $`\sigma _T`$ is the Thomson cross section, and $`g_1`$ and $`g_2`$ are given by
$$g_1(u)=\frac{u^2}{(u+1)^2},\text{ }g_2(u)=\frac{u^2}{(u1)^2+a^2}$$
(7)
with $`uϵ^{}/ϵ_B`$, $`a2\alpha _fϵ_B/3`$, where $`\alpha _f`$ is a fine-structure constant (eg. Herold herold (1979); Dermer dermer (1990)).
In the Klein-Nishina regime ($`ϵ^{}>1`$) the relativistic magnetic cross section for the $`|\mu ^{}|1`$ case becomes better approximated with the well known Klein-Nishina relativistic nonmagnetic total cross section $`\sigma _{KN}`$ (Daugherty & Harding dh86 (1986); Dermer dermer (1990)). When $`ϵ^{}ϵ_B`$, (the condition fulfilled in the K-N regime since $`ϵ_B<1`$ holds throughout this paper), the resonant term $`(g_2g_1)/2`$ in Eq. (6) becomes negligible and the nonresonant term $`g_1`$ approaches unity which results in $`\sigma \sigma _T`$. Therefore, to approximate the Klein-Nishina decline we replace the single nonresonant term $`g_1`$ in the square bracket of Eq. (6) with $`\sigma _{KN}/\sigma _T`$ for $`ϵ^{}>2ϵ_B`$.
To make calculations less time-consuming, we have used the delta-function approximation for the resonant part of the cross section (Dermer dermer (1990)), when calculating the optical depth for a scattering and the averaged energy loss of electron (Eq. 11, see below).
As the density of soft photons $`dn_{\mathrm{ph}}/d/d\mathrm{\Omega }`$ we take the spectral density of blackbody radiation given by
$$\left(\frac{dn_{\mathrm{ph}}}{dd\mathrm{\Omega }}\right)dd\mathrm{\Omega }=\frac{2}{\lambda _C^3}\frac{ϵ^2dϵd\mathrm{\Omega }}{[\mathrm{exp}(ϵ/𝒯)1]}$$
(8)
where $`𝒯`$ is a dimensionless temperature ($`𝒯\frac{kT}{mc^2}`$) and $`\lambda _C=h/(mc)`$ is the electron Compton wavelength. Hereby we neglect anisotropy of the thermal emission expected in the strong magnetic field (eg Pavlov et al. psvz (1994)). For $`ϵϵ_B`$ a preferred photon propagation direction is that of the magnetic field because the opacity is reduced for radiation polarized across $`B`$. The probability of collision with an electron is greatly reduced for photons propagating at $`\mu 1`$, thus, the efficiency of ICS as calculated below should be treated as an upper limit.
If a scattering event occurs, we draw the energy $`ϵ`$ and the cosine of polar angle $`\mu `$ for an incoming photon from a distribution determined with the integrand of Eq. (5). We do this with the simple two-dimensional rejection method which is reliable though more time-consuming (cf. an approach by Daugherty & Harding dh89 (1989)). Next, we transform $`ϵ`$ and $`\mu `$ to the ERF values $`ϵ^{}`$ and $`\mu ^{}`$, and sample the cosine of polar angle of an outgoing photon $`\mu _s^{}`$ using the differential form of the cross section (6) in the Thomson limit:
$`{\displaystyle \frac{d\sigma }{dϵ_s^{}d\mathrm{\Omega }_s^{}}}={\displaystyle \frac{3\sigma _T}{16\pi }}\delta (ϵ_s^{}ϵ_{\mathrm{scat}}^{})[(1\mu _{}^{}{}_{}{}^{2})(1\mu _{s}^{}{}_{}{}^{2})+`$ (9)
$`\text{ }+{\displaystyle \frac{1}{4}}(1+\mu _{}^{}{}_{}{}^{2})(1+\mu _{s}^{}{}_{}{}^{2})(g_1+g_2)]`$
(eg. Herold herold (1979)), where $`d\mathrm{\Omega }_s^{}=d\varphi _s^{}d\mu _s^{}`$ is an increment of solid angle into which outgoing photons with energy $`ϵ_s^{}`$ in the ERF are directed. As in the case of total cross section (6), when $`ϵ^{}>2ϵ_B`$ we replace the factor $`(g_1+g_2)`$ in (9) with $`(2\sigma _{KN}/\sigma _T+g_2g_1)`$. The sampled value of $`\mu _s^{}`$ determines the energy $`ϵ_{\mathrm{scat}}^{}`$ for the particular scattered photon with the relativistic formula:
$`ϵ_{\mathrm{scat}}^{}=(1\mu _{s}^{}{}_{}{}^{2})^1\{\text{}1+ϵ^{}(1\mu ^{}\mu _s^{})+`$ (10)
$`\text{ }[1+2ϵ^{}\mu _s^{}(\mu _s^{}\mu ^{})+ϵ_{}^{}{}_{}{}^{2}(\mu _s^{}\mu ^{})^2]^{1/2}\}`$
appropriate for collisions with a recoiled electron remaining at the ground Landau level (Herold herold (1979)). Finally, a value of the electron’s longitudinal momentum in the ERF is changed due to recoil from zero to $`(ϵ^{}\mu ^{}ϵ_{\mathrm{scat}}^{}\mu _s^{})mc`$, and transformed back to the OF.
The Monte Carlo method will be compared with a smooth integration treatment of the magnetic ICS (Chang chang (1995); Sturner sturner (1995)). In the integration method we assume the same procedure to accelerate an electron and to subtract its energy losses due to the curvature radiation as described above (Eq. 4). The only difference is in accounting for electron energy losses due to the magnetic inverse Compton scattering. These are estimated in *each* step (regardless the value of the optical depth for the scattering process) from the following formula for the mean electron energy loss rate:
$`\dot{\gamma }_{\mathrm{ICS}}=c{\displaystyle }dϵ{\displaystyle }d\mathrm{\Omega }\left({\displaystyle \frac{dn_{\mathrm{ph}}}{dϵd\mathrm{\Omega }}}\right)(1\beta \mu )\times `$ (11)
$`\text{ }\times {\displaystyle 𝑑ϵ_s^{}𝑑\mathrm{\Omega }_s^{}\left(\frac{d\sigma }{dϵ_s^{}d\mathrm{\Omega }_s^{}}\right)(ϵ_sϵ)}`$
where $`ϵ_s=ϵ_s^{}\gamma (1+\beta \mu _s^{})`$ is the scattered photon energy in the OF (eg. Dermer dermer (1990)). In other words, the electron Lorentz factor is determined as a solution of the differential equation:
$$\frac{d\gamma }{dh}=v^1(\dot{\gamma }_{\mathrm{acc}}+\dot{\gamma }_{\mathrm{ICS}}+\dot{\gamma }_{\mathrm{cr}}).$$
(12)
At each step, as an electron moves upwards, a decrease in both the dipolar magnetic field strength and the solid angle subtended by the thermal polar cap is taken into account in both methods.
## 3 Results
We compare the Monte Carlo method described in Sect. 2 with the integration method (numerical integration of Eq. (12)), for a pulsar with the polar magnetic field strength and rotation period as for B1509$-$58 ($`B_{\mathrm{pc}}=15.8\times 10^{12}`$ G, $`P=0.15`$ s).
### 3.1 The case of the M74 electric field
In this subsection the results are presented for the electric field of Michel (michel (1974)) (see Eq.1). The assumed values of $`B_{\mathrm{pc}}`$ and $`P`$ give $`E_{}=3.67\times 10^6`$ V cm<sup>-1</sup> and the standard polar cap radius $`r_{\mathrm{pc}}=3.74\times 10^4`$ cm. Following Sturner (sturner (1995)), we assumed the thermal polar cap radius $`R_{\mathrm{th}}=10^5`$ cm ($`2.7\times r_{\mathrm{pc}}`$) and $`h_{\mathrm{acc}}=r_{\mathrm{pc}}`$. Calculations have been performed for four different temperatures of the thermal polar cap: $`T_6=3.0`$, $`3.5`$, $`4.0`$ and $`4.5`$, where $`T_6=T/(10^6\mathrm{K})`$.
Each panel in Fig. 1 presents in grey the curves $`\gamma (h)`$ as calculated in the Monte Carlo way for one hundred individual electrons. The behaviour determined with Eq. (12) is overplotted as the thick solid line. The inclined dotted line presents changes of electron energy if there were no losses. It becomes horizontal at altitude $`h_{\mathrm{acc}}=r_{\mathrm{pc}}=3.74\times 10^4`$ cm above which the longitudinal component of electric field $`E_{}`$ is assumed to be screened by a charge-separated plasma distribution (see Eq. 1). This height is denoted by the vertical long dashed line.
The thick solid trajectories $`\gamma (h)`$ in four panels of Fig. 1 represent the ”averaged treatment” solutions and are in good agreement with the results of previous calculations by Sturner (sturner (1995)) (cf his Fig. 4). Below $`h10`$ cm they overlap with the case with no energy losses (dotted), since the acceleration rate of electron (as given by $`\dot{\gamma }_{\mathrm{acc}}=\frac{eE_{}}{mc^2}\beta c`$) significantly exceeds the energy loss rate due to the ICS (see Fig. 2a). Energy losses due to the CR are negligible for any Lorentz factor accessible for an electron, given the assumed acceleration model of Michel (Eq. 1). At larger altitudes the thick trajectories illustrate the acceleration stopping effect noted by Chang (chang (1995)) and Kardashëv et al. (kmn (1984)): the energy that the electrons would have gained between altitudes $`h10^2`$ and $`10^4`$ cm due to the electric field $`E_{}`$ is transferred to the thermal photons through the resonant inverse Compton scatterings. Depending on which process – either the resonant ICS cooling or the acceleration – ceases first, the electrons end up with energy which is either closer to $`eE_{}h_{\mathrm{acc}}`$ (ie. the energy acquired due to the full voltage drop at no radiative losses, the case $`T_6=3.0`$) or decreases towards a few tens $`\times mc^2`$ (the cases with $`T_63.5`$).
As can be seen in Fig. 1, the Monte Carlo tracks generally behave qualitatively in the same way: the increasing temperature of the thermal cap increases the altitude at which the acceleration takes over, and eventually, the Lorentz factors of most of electrons become limited below $`10^2`$. However, there are strong differences between the average energy of the Monte Carlo electrons and the value obtained with the integration method. They are especially pronounced for $`T_6=3.5`$, which is the case where a maximum altitude at which the acceleration can be counterbalanced by the electron energy losses due to the resonant ICS is equal to $`h_{\mathrm{acc}}`$. The bulk of the Monte Carlo tracks ends up with Lorentz factors $`\gamma >10^5`$, whereas the integration-method solution gives $`\gamma 64`$. To understand these differences it is worth investigating closely the behaviour of one exemplary electron as determined by the two methods.
In the integration method, the electron’s energy increases monotonically until an equilibrium settles between the energy loss process and the acceleration. Fig. 2b shows $`|\dot{\gamma }_{\mathrm{ICS}}|`$ and $`\dot{\gamma }_{\mathrm{acc}}`$ as a function of the Lorentz factor $`\gamma `$ for three different altitudes in the case $`T_6=3.0`$. One can see that $`|\dot{\gamma }_{\mathrm{ICS}}|`$ exceeds $`\dot{\gamma }_{\mathrm{acc}}`$ between the two equilibrium Lorentz factors $`\gamma _{\mathrm{eq1}}`$ and $`\gamma _{\mathrm{eq2}}`$ the values of which are determined by the condition $`|\dot{\gamma }_{\mathrm{ICS}}|=\dot{\gamma }_{\mathrm{acc}}`$. We have calculated them using the approximation $`\dot{\gamma }_{\mathrm{ICS}}\dot{\gamma }_{\mathrm{res}}`$ where $`\dot{\gamma }_{\mathrm{res}}`$ is the energy loss rate due to the resonant part of the cross section alone (cf. Eq. (22) in Sturner sturner (1995)). The two equilibrium Lorentz factors $`\gamma _{\mathrm{eq1}}`$ and $`\gamma _{\mathrm{eq2}}`$ as calculated for different altitudes $`h`$ are presented in Fig. 1 by the thick dashed line. For increasing altitude their values approach each other because a decrease in the density of black-body photons moves down the whole curve $`|\dot{\gamma }_{\mathrm{ICS}}(\gamma )|`$ in Fig. 2b. Thus, the energy-altitude space is divided into two regimes: the region where $`|\dot{\gamma }_{\mathrm{ICS}}|>\dot{\gamma }_{\mathrm{acc}}`$, hereafter called the ”energy loss dominated” region, (it is surrounded by the thick dashed line in Fig. 1), and the ”acceleration dominated” region where $`|\dot{\gamma }_{\mathrm{ICS}}|<\dot{\gamma }_{\mathrm{acc}}`$. As can be seen in Fig. 1, according to the integration method the trajectory of an electron in the energy-altitude space does not penetrate the energy loss dominated region. As the electron moves upwards, its Lorentz factor settles at the lower equilibrium value $`\gamma _{\mathrm{eq1}}`$, and stays there as long as the equilibrium can exist ie. until the curves $`|\dot{\gamma }(\gamma )|`$ and $`\dot{\gamma }_{\mathrm{acc}}(\gamma )`$ in Fig. 2b disconnect. For $`T_6=3.0`$ this happens near the altitude $`h_{\mathrm{eq}}=2\times 10^4`$. Above the height $`h_{\mathrm{eq}}`$ the acceleration cannot be counterballanced by the ICS energy losses at any Lorentz factor accessible for the electron and the increase in the electron energy resumes.
In the Monte Carlo method, the electron’s energy losses due to the scattering occur in a discontinuous way, thus, $`\gamma `$ oscillates around the value $`\gamma _{\mathrm{eq1}}`$ (see Fig. 3). When the energy loss rate $`|\dot{\gamma }_{\mathrm{ICS}}|`$ hardly exceeds the acceleration rate (see Fig. 2) the difference between the equilibrium Lorentz factors $`(\gamma _{\mathrm{eq1}}\gamma _{\mathrm{eq2}})`$ is small and the electron is able to pass *through* the energy-loss dominated region. Eg. for the case $`T_6=3.0`$, $`\gamma _{\mathrm{eq2}}22004\times \gamma _{\mathrm{eq1}}`$ at $`h=10^3`$ cm. To increase its energy from $`\gamma _{\mathrm{eq1}}`$ to $`\gamma _{\mathrm{eq2}}`$ it is enough for the electron to avoid a scattering over a ‘break-through’-distance $`\mathrm{\Delta }_{\mathrm{bt}}=\frac{\gamma _{\mathrm{eq2}}\gamma _{\mathrm{eq1}}}{eE_{}/mc^2}=228\mathrm{cm}`$ which is only ‘a few’ times larger than the local mean free path for scattering $`\lambda _{\mathrm{ICS}}`$. The local value of $`\lambda _{\mathrm{ICS}}`$ depends strongly on the electron Lorentz factor and for $`\gamma `$ between $`\gamma _{\mathrm{eq1}}`$ and $`\gamma _{\mathrm{eq2}}`$ it ranges from $`28`$ to $`108`$ cm (the case $`T_6=3.0`$, $`h=10^3`$ cm). This gives $`\mathrm{\Delta }_{\mathrm{bt}}(82)\times \lambda _{\mathrm{ICS}}`$. Thus, there is a very large probability for the electron to gain $`\gamma >\gamma _{\mathrm{eq2}}`$ at an altitude $`h_{\mathrm{MC}}`$ lower than $`h_{\mathrm{eq}}`$. Once the electron enters the acceleration dominated region at $`h_{\mathrm{MC}}h_{\mathrm{eq}}`$, its energy starts to increase up to a value much larger than obtained in the integration method.
For increasing temperatures $`T`$, the width $`(\gamma _{\mathrm{eq2}}\gamma _{\mathrm{eq1}})`$ of the energy-loss dominated region increases (Fig. 1) whereas the mean free path for $`\gamma =\gamma _{\mathrm{eq1}}`$ decreases. This makes the diffusion of electrons through the energy loss dominated region more difficult: to increase its energy from $`\gamma _{\mathrm{eq1}}`$ to $`\gamma _{\mathrm{eq2}}`$ an electron must avoid a scattering over a distance which is increasing multiplicity of the local mean free path. As a result, the energy distribution for outgoing electrons becomes softer (Fig. 4).
It should be emphasized that the strong disagreement between the final electron energies as determined with the two methods only appears if conditions similar to those for the case $`T_6=3.5`$ in Fig. 1 are fulfilled. These include the equilibrium between the maximum rate of resonant energy losses and the rate of acceleration at $`h=h_{\mathrm{acc}}`$. Making use of Dermer’s approximation for $`|\dot{\gamma }_{\mathrm{res}}|`$ (Dermer dermer (1990)) one can easily find that it has the maximum at $`\gamma _{\mathrm{res}}=ϵ_B/(w(1\beta \mu _{\mathrm{min}})𝒯)`$ where $`\mu _{\mathrm{min}}=h/(h^2+R_{\mathrm{th}}^2)^{1/2}`$ is a cosine of an angle at which the thermal cap radius is seen from the position of electron and $`w=\mathrm{ln}0.5`$. The equilibrium $`|\dot{\gamma }_{\mathrm{res}}|=\dot{\gamma }_{\mathrm{acc}}`$ holds for
$$T_6^{\mathrm{eq}}=4.14\left(\frac{\beta ^2\frac{eE_{}}{mc^2}}{(1\beta \mu _{\mathrm{min}})B_{12}(h)}\right)^{1/2}$$
(13)
where $`B_{12}(h)=B(h)/(10^{12}\mathrm{G})`$ and $`\frac{eE_{}}{mc^2}`$ is in cm<sup>-1</sup> (cf Eq. (29) in Chang chang (1995)). For the considered accelerating field (Eq. 1) this condition becomes
$$T_6^{\mathrm{eq}}=3.46\beta (1\beta \mu _{\mathrm{min}})^{1/2}\left(B_{12}P\right)^{1/4}$$
(14)
and is shown in Fig. 5 for $`h=0`$ and $`P=0.15`$ s as the solid line.
Other conditions required for the diffusion are low thermal cap temperatures and high magnetic field strengths. Both requirements partially stem from the dependence $`\lambda _{\mathrm{ICS}}(\gamma _{\mathrm{res}})BT^3`$ which results in $`\lambda _{\mathrm{ICS}}\mathrm{\Delta }_{\mathrm{bt}}`$ for high $`T`$ and low $`B_{\mathrm{pc}}`$. \[Note that at the resonance $`\lambda _{\mathrm{ICS}}`$ is *proportional to* $`B`$ just as $`|\dot{\gamma }_{\mathrm{res}}|`$\].
Additionally, high temperatures and low $`B`$-field strengths preclude the diffusion because of relatively large energy losses due to scatterings in the Klein-Nishina regime (see Fig 2a). First, for $`\gamma >\gamma _{\mathrm{res}}`$, $`|\dot{\gamma }_{\mathrm{KN}}|\dot{\gamma }_{\mathrm{acc}}`$ must hold, which requires low $`T`$. Second, the resonant bump in the $`|\dot{\gamma }_{\mathrm{ICS}}(\gamma )|`$ curve must be present, which is possible when $`|\dot{\gamma }_{\mathrm{res}}(\gamma _{\mathrm{res}})||\dot{\gamma }_{\mathrm{KN}}|`$. Since $`\dot{\gamma }_{\mathrm{res}}(\gamma _{\mathrm{res}})/\dot{\gamma }_{\mathrm{KN}}B`$, strong magnetic fields are preferred.
For $`P0.15`$ s and $`R_{\mathrm{th}}>h_{\mathrm{acc}}`$ all these constraints limit the diffusion regime to $`T<5\times 10^6`$ K and $`B_{\mathrm{pc}}>3\times 10^{12}`$ G with the $`T`$ and $`B_{\mathrm{pc}}`$ roughly fulfilling Eq. (14) taken for $`h=h_{\mathrm{acc}}`$. The region of strong discrepancies between the two methods is shown in Fig. 5 (hatched).
Among a few pulsars exhibiting a two-component thermal X-ray spectra only the Vela pulsar has an inferred temperature of the thermal polar cap ($`T15.8\times 10^6`$ K, Ögelman ogelman (1995)) which is sufficient for the acceleration to be initially halted by the resonant ICS (ie. it exceeds $`T_6^{\mathrm{eq}}`$ for $`hR_{\mathrm{th}}`$, see Fig. 5). The equilibrium $`|\dot{\gamma }_{\mathrm{res}}|=\dot{\gamma }_{\mathrm{acc}}`$ holds up to the altitude $`h_{\mathrm{eq}}2.2\times 10^4`$ cm which is two times larger than the inferred thermal cap radius ($`10^4`$ cm, Sturner sturner (1995)) and two times smaller than $`r_{\mathrm{pc}}=4.8\times 10^4`$ cm. Nevertheless, because of the high $`T`$, at $`hh_{\mathrm{eq}}`$ the diffusion is precluded by the extremely small $`\lambda _{\mathrm{ICS}}0.1`$ cm. It becomes efficient only at $`h>0.9h_{\mathrm{eq}}`$, thus, most Monte-Carlo electrons follow closely the behaviour predicted by the continuous approach by Chang (chang (1995)) or Sturner (sturner (1995)).
For other pulsars with a likely thermal X-ray emission from heated polar caps (eg. Geminga, B1055$-$52, and B0656$+$14) inferred temperatures are much lower (between 2 and 4 $`\times 10^6`$ K). Therefore, $`|\dot{\gamma }_{\mathrm{ICS}}|\dot{\gamma }_{\mathrm{acc}}`$ for any Lorentz factor accessible for an electron (Fig. 5) and the Compton scattering cannot considerably redistribute an initial electron energy spectrum.
Both, in the case of negligible ICS losses, ie. for $`TT^{\mathrm{eq}}(h=0)`$, and in the cases when $`|\dot{\gamma }_{\mathrm{res}}(\gamma _{\mathrm{res}})|\dot{\gamma }_{\mathrm{acc}}`$ holds up to $`h=h_{\mathrm{acc}}`$, ie. for $`TT^{\mathrm{eq}}(h=h_{\mathrm{acc}})`$, we find the energy distribution of electrons to be quasi-monoenergetic with energy well approximated by Eq.12. In the case of negligible ICS cooling the energy distribution of electrons is a narrow peak at the energy $`\gamma _{\mathrm{max}}mc^2=eE_{}h_{\mathrm{acc}}`$, whereas for very large efficiency of ICS losses (close to 100% of $`\gamma _{\mathrm{max}}mc^2`$) nearly all particles are cooled down to the energy $`\gamma _{\mathrm{min}}`$ for which the ICS loss rate decreases sharply (see Fig. 2a and Fig. 1). Between these limiting cases, ie. when the ICS cooling is comparable to the acceleration, broad energy distributions of electrons (ranging from $`\gamma _{\mathrm{min}}`$ to $`\gamma _{\mathrm{max}}`$) emerge (Fig. 4).
### 3.2 The case of the HM98 electric field
To enable electron energy losses due to the resonant ICS to compete with acceleration (the resonant ICS damping becomes then important and the diffusion effect does occur) the strength of the electric field $`E_{}`$ should not exceed a critical value which may be roughly estimated as $`E_{}^{\mathrm{eq}}310^4B_{12}T_6^2\mathrm{V}\mathrm{cm}^1`$ (cf. Chang chang (1995)). For $`E_{}(h)`$ given by eq. (18) of HM98 this may only occur in the case of small acceleration length $`h_{\mathrm{acc}}r_{\mathrm{pc}}`$ (see Eq. (2)), when the acceleration height is limited by the ICS-induced pair formation front. Self consistent values of $`h_{\mathrm{acc}}`$ are then of the order of a few$`\times 10^3`$ cm, i.e. much smaller than for a CR-induced case (cf. figs. 5 and 6 in HM98). A value of $`h_{\mathrm{acc}}`$ is considered self consistent if the acceleration of primary electrons by the electric field induced with this $`h_{\mathrm{acc}}`$ generates pair formation front at $`h_{\mathrm{acc}}`$.
We calculated self consistent $`h_{\mathrm{acc}}`$ in an approximate way, following HM98, by finding a minimum of a sum of acceleration length and a mean free path for one photon absorption:
$$h_{\mathrm{acc}}=\mathrm{min}\left\{\left(\frac{2\gamma }{N}\right)^{1/2}+\left(\frac{0.2\rho _{\mathrm{curv}}}{ϵ_Bϵ}\right)\right\}$$
(15)
where to get the first term we used $`E_{}`$ in the limit $`hh_{\mathrm{acc}}`$, given by Eq. (2): $`E_{}(h)=Nh`$ with $`N=3\mathrm{\Omega }R_{\mathrm{ns}}B_{\mathrm{pc}}\kappa \mathrm{cos}\chi (c(1ϵ_{\mathrm{GR}})R_{\mathrm{ns}}^2)^1h_{\mathrm{acc}}`$. As an energy of pair producing photons in the second term we assumed that for the resonant scattering: $`ϵ\gamma ϵ_B`$. In Eq. (15) we neglected a contribution from the mean free path for the resonant ICS which is of the order of $`150(1\beta \mu _{\mathrm{min}})^1B_{12}T_6^3`$ cm near the resonance. Since in the regime $`h_{\mathrm{acc}}r_{\mathrm{pc}}`$ the electric field $`E_{}`$ depends itself on $`h_{\mathrm{acc}}`$ we had to repeat the calculation of $`h_{\mathrm{acc}}`$ starting from some guess value until convergence. For $`B_{\mathrm{pc}}=15.8`$ TG, $`P=0.15`$ s, $`\rho _{\mathrm{curv}}=10^7`$ cm and $`\chi =0.1`$ rad we obtained $`h_{\mathrm{acc}}2\times 10^3`$ cm.
Then we calculated the changes of $`\gamma `$ with $`h`$ for $`R_{\mathrm{th}}=10^5`$ cm and for different thermal cap temperatures (as listed in the preceding subsection). We found that the most notable divergence between the Monte Carlo results and the integration method results occurs for $`T_6=3.5`$. As can be seen in Fig. 6a, at $`h10^3`$ cm most Monte Carlo electrons reach Lorentz factor values between $`10^3`$ and $`10^4`$, contrary to $`470`$ as predicted by the integration method. The difference disappears only above the accelerator (i.e. for $`h>h_{\mathrm{acc}}`$) because such altitude is still low enough (i.e. $`hR_{\mathrm{th}}`$) for the resonant ICS to damp $`\gamma `$ to its final value of about $`50`$ (at which it becomes eventually negligible).
An effect similar to that shown in Fig. 1 appears, however, when the resonant ICS losses are no longer effective above $`h=h_{\mathrm{acc}}`$. Fig. 6b shows the case with $`R_{\mathrm{th}}=4.0\times 10^3`$ cm $`=2\times h_{\mathrm{acc}}`$ and $`T_6=3.0`$. Broad electron energy distribution (with average energy much exceeding a value from the integration method) emerges.
Similarly as in the weak electric field case, the position of the diffusion region in $`T`$-$`B_{\mathrm{pc}}`$ plane (for a fixed value of $`R_{\mathrm{th}}`$) is here determined with a condition analogous to Eq. (13) (but $`E_{}`$ of Eq.(2) requires a more accurate determination of $`h_{\mathrm{acc}}`$) along with a requirement for strong $`B_{\mathrm{pc}}`$ and low $`T`$.
For stronger electric fields ($`E_{}>E_{}^{\mathrm{eq}}`$), or for larger acceleration lengths the resonant ICS has negligible influence on final electron energies. In the first case acceleration greatly dominates the ICS damping; in the second case acceleration is stopped only over a negligible, initial part of accelerator height.
## 4 Conclusions
We have calculated the efficiency of energy losses for electrons accelerated over a hot polar cap of a neutron star. As cooling mechanisms we have considered the inverse Compton scattering and the curvature radiation, (though the latter has been negligible for the considered acceleration models).
The electron energy losses due to the ICS have been calculated with the Monte Carlo method and compared with the integration method (after Chang chang (1995); Sturner sturner (1995)) based on the prescription for the averaged ICS cooling (Dermer dermer (1990)). We confirm general predictions of the integration approach (eg. the limitation of electron Lorentz factors below $`10^2`$ for high temperatures), nevertheless, the “stopping acceleration effect” occurs at a slightly higher temperature than that one resulting from integrating the differential equation (12).
For the considered pulsar parameters ($`B_{\mathrm{pc}}=15.8\times 10^{12}`$ G, $`R_{\mathrm{th}}=10^5`$ cm) and the acceleration potential by Michel (michel (1974)), the Monte Carlo-based value is approximately equal to $`T_{\mathrm{MC}}=4.5\times 10^6`$ K (at this $`T`$ only a few percent of electrons reach $`\gamma >10^2`$ at $`h=10^6`$ cm) in comparison with $`T_{\mathrm{int}}=3.3\times 10^6`$ as given by the integration approach. Although the temperature difference is small, the final electron Lorentz factors as determined by the two methods may easily differ by a few orders of magnitude (Fig. 1, the case $`T_6=3.5`$) if the observed temperature matches this range. Moreover, it should be kept in mind that for most objects with hard X-ray emission identified putatively as originating from the hot polar cap, temperatures lie within the similar range $`(24)\times 10^6`$ K (Ögelman ogelman (1995)).
We find that for high-$`B`$ and low-$`T`$ pulsars the integration method gives a poor estimation of electron Lorentz factors if the energy loss rate due to the resonant ICS hardly exceeds the rate of acceleration. In the case of Michel’s model and for $`R_{\mathrm{th}}>h_{\mathrm{acc}}`$, this occurs if $`T4(B_{12}P)^{1/4}`$ which may be the case for PSR B1509$-$58. In the cases when the ICS loss rate dominate firmly over the acceleration or is negligible the energy distribution of outgoing electrons is quasi-monoenergetic around the value which is well approximated with the continuous approach.
We find that preventing electrons from achieving large Lorentz factors occurs also for accelerating potential by Harding & Muslimov (hm (1998)) in its nonsaturated version with $`h_{\mathrm{acc}}r_{\mathrm{pc}}`$. The discrepancy between the Monte Carlo and the integration method also occurs within the high-$`B`$ and low-$`T`$ regime and is especially pronounced for $`R_{\mathrm{th}}h_{\mathrm{acc}}`$.
###### Acknowledgements.
This work was supported by KBN grants 2P03D 01016 and 2P03D 02117. JD thanks T.Bulik for useful advice in programming. We appreciate sugestions by the anonymous referee and comments by Bing Zhang on the manuscript which helped to improve the paper.
|
warning/0003/hep-ph0003008.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
One of the basic open questions of particle physics is the nature of electroweak symmetry breaking. Beyond the discovery of the Higgs boson(s), however, reconstructing the Higgs potential will be necessary, and that requires the experimental study of the self–couplings of the Higgs bosons.
Pair production of neutral Higgs particles in gluon fusion sensitive to the trilinear couplings of Higgs bosons was studied in the SM and the Minimal Supersymmetric Standard Model (MSSM) \[2–5\]. QCD corrections were included in the limiting case of a heavy top mass .
In the SM the trilinear and the quartic couplings of the physical Higgs particle are fixed by the Higgs mass. There is no resonance in the pair production of SM Higgses, and the cross section is only $``$20–50 fb in the intermediate mass range . No resonance effect is present in the production of three Standard Model Higgs bosons via one Higgs in gluon fusion and it is estimated to be under the discovery limit at LHC.
In the MSSM there are five Higgs bosons ($`h,H,A,H^\pm `$) and many trilinear and quartic couplings among them. Two parameters describe the Higgs sector of the MSSM: $`M_A`$, the mass of the CP-odd Higgs boson A, and tan$`\beta `$, the ratio of the two vacuum expectation values. For a wide range of $`M_A`$, in $`pp`$ collision only the cross section of the hh production is large . Possible other processes, such as the gauge boson fusion and Higgs–strahlung off W bosons or heavy quarks (top) provide less event number.
In order to provide further tests of trilinear and quartic Higgs self–couplings, in this paper we calculate the resonance enhanced production of three lightest supersymmetric Higgs bosons (h) in gluon fusion
$$ppgghhh.$$
(1)
Generally four different graphs contribute to the production of three Higgses. Gluons couple to triangle, box or pentagon quark loops emitting 1,2 or 3 Higgs bosons, as is seen in Fig.1. Squark loops are neglected. We have omitted the pentagon graph producing a flat continuum background for the resonance production and we will give the complete cross section valid in a wider range in a subsequent publication . We learn that the quartic coupling is not accessible by LHC experiment. The resonance contribution to (1), however, sizeable at LHC For instance at $`\mathrm{tan}\beta =3`$ it yields a cross section of about 300 fb. This is about $`1/10`$ times the corresponding $`hh`$–case .
The masses, widths and the couplings were calculated using the complete one–loop and the leading two–loop radiative corrections from . The relevant Higgs self–couplings are the following
$`\lambda _{hhh}`$ $`=3\mathrm{cos}(2\alpha )\mathrm{sin}(\beta +\alpha )+{\displaystyle \frac{3ϵ}{M_Z^2}}{\displaystyle \frac{\mathrm{cos}^3\alpha }{\mathrm{sin}\beta }}`$ (2)
$`\lambda _{Hhh}`$ $`=2\mathrm{sin}(2\alpha )\mathrm{sin}(\beta +\alpha )\mathrm{cos}(2\alpha )\mathrm{cos}(\beta +\alpha )+{\displaystyle \frac{3ϵ}{M_Z^2}}{\displaystyle \frac{\mathrm{sin}\alpha \mathrm{cos}^2\alpha }{\mathrm{sin}\beta }}`$
$`\lambda _{HHh}`$ $`=2\mathrm{sin}(2\alpha )\mathrm{cos}(\beta +\alpha )\mathrm{cos}(2\alpha )\mathrm{sin}(\beta +\alpha )+{\displaystyle \frac{3ϵ}{M_Z^2}}{\displaystyle \frac{\mathrm{sin}^2\alpha \mathrm{cos}\alpha }{\mathrm{sin}\beta }}`$
$`\lambda _{Hhhh}`$ $`=3\mathrm{sin}(2\alpha )\mathrm{cos}(2\alpha )+{\displaystyle \frac{ϵ}{8M_Z^2}}{\displaystyle \frac{\mathrm{cos}\alpha \mathrm{sin}^2\alpha }{\mathrm{sin}^2\beta }}`$
$`\lambda _{hhhh}`$ $`=3\mathrm{cos}^2(2\alpha )+{\displaystyle \frac{ϵ}{4M_Z^2}}{\displaystyle \frac{\mathrm{sin}^4\alpha }{\mathrm{sin}^2\beta }}`$
The trilinear and quartic couplings are normalized to $`\lambda _3=[\sqrt{2}G_F]^{1/2}M_Z^2`$ and $`\lambda _4=\sqrt{2}G_FM_Z^2`$, respectively. The couplings depend on $`\beta `$ and the mixing angle $`\alpha `$ of the CP–even Higgs sector
$$\mathrm{tan}2\alpha =\frac{M_A^2+M_Z^2}{M_A^2M_Z^2+ϵ/\mathrm{cos}2\beta }\mathrm{tan}2\beta .$$
(3)
The leading $`m_t^4`$ one–loop corrections are parametrized by
$$ϵ=\frac{3G_F}{\sqrt{2}\pi ^2}\frac{m_t^4}{\mathrm{sin}^2\beta }\mathrm{log}\left[1+\frac{M_S^2}{m_t^2}\right].$$
(4)
The common squark mass is fixed to $`M_S`$= 1 TeV. The quartic couplings depend on $`M_A`$ and are shown for two values of $`\text{tan}\beta =3`$ and 30 in Fig.2.
The paper is organized as follows. In the next section we give the cross section of the resonance production of three h’s in gluon fusion and in section 3 we show the numerical results.
## 2 The cross section
Three different channels contribute to the resonance production of three lightest supersymmetric Higgs particles (h) in gluon fusion. (a) One virtual Higgs boson (h,H) is produced by the heavy quark triangle and it decays into 3 h’s via the quartic coupling $`\lambda _{Hhhh}`$ or $`\lambda _{hhhh}`$, Fig. 1a. (b) One virtual Higgs boson decays into 3 h’s in two steps testing trilinear couplings, $`ggH,h(H,h)hhhh`$, Fig. 2b. (c) Heavy quark box diagram couples to two Higgses, one of them decays subsequently into two h’s, $`gg(H,h)hhhh`$, Fig. 2c. We have omitted the pentagon graphs contributing only to the continuum production. We believe it increases the production at large tan$`\beta `$ due to the large hbb–coupling, but does not change our conclusion about the measureability of the quartic Higgs–couplings.
There are two different helicity amplitudes contributing to the total cross section: the total spin of the two gluons along the collision axis can be $`S_z=0`$ (contribution F) or $`S_z=2`$ (contribution G). The triangle graphs give only contributions $`F_3`$ ($`S_z=0`$) as they involve a single spin–0 Higgs intermediate state. Box graphs give both contributions $`F_4`$ ($`S_z=0`$) and $`G_4`$ ($`S_z=2`$). $`F_4`$ and $`G_4`$ are Lorentz and gauge invariant decompositions of the box amplitude. The tensor basis is the following
$`S_z=0A^{\mu \nu }=g^{\mu \nu }{\displaystyle \frac{p_1^\nu p_2^\mu }{(p_1p_2)}}`$ (5)
$`S_z=2B^{\mu \nu }=g^{\mu \nu }+{\displaystyle \frac{1}{p_T^2(p_1p_2)}}\left(p_3^2p_1^\nu p_2^\mu 2(p_2p_3)p_1^\nu p_3^\mu 2(p_1p_3)p_2^\mu p_3^\nu +2p_3^\mu p_3^\nu (p_1p2)\right)`$
where $`p_1,p_2`$ are the momenta of the incoming gluons and $`p_3`$ is one of the momenta of the outgoing Higgs bosons ($`p_3,p_4,p_5`$). Here $`p_T^2=2\frac{(p_1p_3)(p_2p_3)}{(p_1p_2)}p_3^2`$ is the transverse momentum of the third particle. Tensors $`A^{\mu \nu }`$ and $`B^{\mu \nu }`$ are orthogonal and normalized to $`2`$ .
The M–matrix of the process is
$$=\frac{(\sqrt{2}G_F)^{3/2}\alpha _s\widehat{s}}{4\pi }ϵ_a^\mu ϵ_b^\nu \delta _{ab}(FA_{\mu \nu }+GB_{\mu \nu }).$$
(6)
The spin and color averaged parton level cross section of the process $`gghhh`$ is
$$d\widehat{\sigma }(gghhh)=\frac{\sqrt{2}G_F^3\alpha _s^2}{1024(2\pi )^6}\left[|F|^2+|G|^2\right]\frac{\lambda ^{1/2}(\widehat{s}_{45},p_4^2,p_5^2)}{\widehat{s}_{45}}d\widehat{s}_{45}d\widehat{s}_{13}d\mathrm{\Omega }_{45}^{CM}.$$
(7)
Here we used the Chew-Low parametrization of the three particle phase space , a trivial angle integration was carried out, $`\widehat{s}_{ik}=(p_i+p_k)^2`$, and the usual lambda function is $`\lambda (x,y,z)=(x(\sqrt{y}+\sqrt{z})^2)(x(\sqrt{y}\sqrt{z})^2)`$. $`d\mathrm{\Omega }_{45}^{CM}`$ is the differential solid angle in the center of mass system of particles 4 and 5.
$`F`$ receives contributions from the triangle and box graphs, while only box diagrams give contributions $`S_z=2`$ to $`G`$
$`F`$ $`=`$ $`{\displaystyle \underset{t,b}{}}(C_3F_3+{\displaystyle \underset{i=3,4,5}{}}C_4^{(i)}F_4^{(i)})`$
$`G`$ $`=`$ $`{\displaystyle \underset{t,b}{}}{\displaystyle \underset{i=3,4,5}{}}C_4^{(i)}G_4^{(i)}.`$ (8)
Here we separated the functions $`F_3,F_4,G_4`$ responsible for the heavy quark triangle and boxes, $`C_3,C_4`$ contain the coupling constants and the propagators. The amplitude has to be summed over all quark flavours. However, the four light quarks can be neglected having very small Higgs couplings. Supersymmetric particles are assumed to be too heavy to participate in the triangle or box loop. The second sum in the box contribution corresponds to the outgoing Higgs particle that couples directly to the quark loop.
We calculated $`F_3,F_4,G_4`$ and found them in agreement with the results of , . For instance,
$$F_3=2\frac{m_Q^2}{\widehat{s}}\left(2+(4m_Q^2\widehat{s})C_{12}\right),\widehat{s}=\widehat{s}_{12}.$$
(9)
where
$$C_{ij}=\frac{d^4q}{i\pi ^2}\frac{1}{(q^2m_Q^2)[(q+p_i)^2m_Q^2][(q+p_i+p_j)^2m_Q^2]}.$$
(10)
$`F_4,G_4`$ have long analytic expressions which can be found in ref . The box functions $`F_4^{(i)},G_4^{(i)}`$ in (8) depend on three momenta $`(p_1,p_2,p_i),i=3,4,5`$ and $`F_4^{(i)}=F_4(p_1,p_2,p_i)`$ etc.
In case of a large quark mass ($`m_Q^2\widehat{s}M_{h,H}^2`$) we get as in
$$F_3=\frac{2}{3}+𝒪\left(\widehat{s}/m_Q^2\right),F_4^{(i)}=\frac{2}{3}𝒪\left(\widehat{s}/m_Q^2\right),G_4^{(i)}=𝒪\left(\widehat{s}/m_Q^2\right)$$
(11)
In the limit of light quark masses ($`m_Q^2\widehat{s}M_{h,H}^2`$) all the form factors vanish to $`𝒪\left(m_Q^2/\widehat{s}\right)`$.
The generalized triangle couplings are the following
$`C_3=C_3^h+C_3^H+C_3^{hh}+C_3^{hH}+C_3^{Hh}+C_3^{HH}`$
$`C_3^{H_a}=\lambda _{hhhH_a}{\displaystyle \frac{M_Z^2}{\widehat{s}M_{H_a}^2+iM_{H_a}\mathrm{\Gamma }_{H_a}}}g_Q^{H_a}`$ (12)
$`C^{H_aH_b}=\lambda _{H_aH_bh}{\displaystyle \frac{M_Z^2}{\widehat{s}M_{H_a}^2+iM_{H_a}\mathrm{\Gamma }_{H_a}}}\lambda _{H_bhh}{\displaystyle \frac{M_Z^2}{\widehat{s}M_{H_b}^2+iM_{H_b}\mathrm{\Gamma }_{H_b}}}g_Q^{H_a},H_{a,b}=h/H.`$
The box couplings are
$`C_4=C_4^h+C_4^H`$
$`C_4^{(h,H)(i)}=\lambda _{hh(h/H)}{\displaystyle \frac{M_Z^2}{\widehat{s}_{kl}M_{h/H}^2+iM_{h/H}\mathrm{\Gamma }_{h/H}}}g_Q^hg_Q^{h/H},`$ (13)
$`i,k,l`$ are cyclic. $`g_Q^{h/H}`$ denotes the Higgs–quark couplings normalized to the SM Yukawa coupling $`[\sqrt{2}G_F]^{1/2}m_Q`$,
$$g_t^h=\mathrm{cos}\alpha /\mathrm{sin}\beta ,g_b^h=\mathrm{sin}\alpha /\mathrm{cos}\beta ,g_t^H=\mathrm{sin}\alpha /\mathrm{sin}\beta ,g_b^H=\mathrm{cos}\alpha /\mathrm{cos}\beta .$$
(14)
## 3 Results
Now, we calculate the production cross section of three lightest neutral supersymmetric Higgs bosons $`pphhh+X`$ via gluon fusion at the LHC energy of $`\sqrt{s}=14`$ TeV in the above picture by folding the parton level fusion cross section with the gluon luminosity. We used the GRV structure functions for the gluon luminosity at $`Q^2=\widehat{s}`$. The numerical integration was made by the VEGAS package . The total cross section vs. $`M_A`$ is shown in Fig.3 for two represantative values of tan$`\beta =3`$ and $`\text{tan}\beta =30`$. In the plotted range of $`M_A`$ the heavy CP–even Higgs boson (H) is nearly degenerate in mass with A while the mass of the lightest neutral Higgs h is quickly approaches approximately 104 GeV from below.
For $`\text{tan}\beta =3`$ in the range of $`M_A=`$ 200–350 GeV the cross section is enhanced by resonance effect and large. The cross section goes up to nearly 300 fb and slowly decreases to about 150 fb giving a large number of events at the expected luminosity $`L𝑑t100fb^1/year`$. The main contribution comes from the box diagram where the heavy CP–even Higgs boson H couples to the quark loop and decays into two bosons h. $`M_h`$ slowly increases around 100 GeV when $`M_HM_A`$ reaches 200 GeV the H propagator enters the resonance region. At $`M_H2m_t`$ there is a threshold effect due to the fall–off of the branching ratio $`BR(Hhh)`$ and partly because on–shell top quark pairs can be produced in the quark loop. For large values of $`M_A`$ the cross section reaches a continuum value of a few fb. The dip at small $`M_A`$, similarly to the case of $`\text{tan}\beta =30`$, is the result of the zero in the trilinear couplings $`\lambda _{(H/h)hh}`$.
For $`\text{tan}\beta =30`$ the main contribution comes from the box diagram where an h couples to the quark loop, propagates and decays into two h’s via $`\lambda _{hhh}`$. The Hhh coupling is small compared to the case of $`\text{tan}\beta =3`$ while the hhh coupling is larger giving a sizeable continuum. There is no observable resonance effect and the cross section is nearly constant, 6 fb, versus $`M_A`$ after the coupling constant changed sign. The K factor is expected to increase the cross section as in other cases .
The contributions of the diagrams containing a heavy quark triangle are also shown in Fig.3 at $`\text{tan}\beta =3`$. The lowest curve corresponds to the diagram with quartic couplings, the one in the middle sums up all the contributions of the graphs involving a quark triangle. There is a clear resonance effect in both curves when $`M_H`$ reaches $`3M_h`$ at $``$310 GeV. Here the first propagator of $`ggHHhhhh`$ becomes resonant too. The first rise in the middle curve is the result of the resonance in the second propagator of Fig. 2b, similarly to the box. The triangle loops yield, however, only a small fraction of the cross section even at their peak at 300-350 GeV.
The suppressed contribution of the quartic coupling implies that in $`pphhh+X`$ the measurement of the quartic coupling is not possible. Not only the box contribution is larger by two orders of magnitude but also the other triangle diagrams are 10 times larger. For $`\text{tan}\beta =30`$ the case is even worse.
In conclusion, in this paper we have calculated the resonance contribution to the cross section of 3h–production at LHC via gluon fusion. The main contribution comes from the trilinear Higgs coupling while the quartic ones turn out to be negligible. In the resonance region at small tan$`\beta `$ the ratio of 2h and 3h production is about 10.
This work was partially supported by Hungarian Science Foundation Grants under Contract Nos. OTKA T029803 and F022998.
FIGURE CAPTIONS
Fig.1: Generic diagrams contributing to the production of three CP–even MSSM Higgs bosons in gluon–gluon collisions, $`gghhh`$: a) triangle with quartic coupling, b)triangle with trilinear couplings c) box , d) pentagon contributions.
Fig.2: Quartic Higgs couplings in the MSSM as functions of the pseudoscalar Higgs mass $`M_A`$ for two representative values of $`\text{tan}\beta =3`$ and 30.
Fig.3: Total cross sections for production of three lightest CP–even MSSM Higgs bosons in gluon–gluon collisions at the LHC for $`\text{tan}\beta =3`$ and 30 (upper two curves). The upper axis presents the scalar Higgs masses $`M_h`$ for $`\text{tan}\beta =3`$ corresponding to the pseudoscalar masses $`M_A`$. The lower two dashed curves represent the resonance contributions of the triangle graphs of Fig. 2a and 2b for $`\text{tan}\beta =3`$.
|
warning/0003/physics0003086.html
|
ar5iv
|
text
|
# XAFS spectroscopy. I. Extracting the fine structure from the absorption spectra
## I Introduction
X-ray-absorption fine-structure (XAFS), $`\chi `$, is determined by :
$$\chi (E)=[\mu (E)\mu _0(E)]/[\mu _0(E)\mu _b(E)],$$
(1)
where $`\mu `$ is the measured absorption, $`\mu _0`$ is the “atomic” absorption due to electrons of considered atomic level, $`\mu _b`$ is the absorption of other processes. Since the electronic state of an embedded atom is, in general, different from its state in gaseous phase, $`\mu _0`$ is not the same as for isolated atom and cannot be found experimentally. Therefore a demand arises for an artificial construction of $`\mu _0`$.
Usually, $`\mu _b`$ is approximated by a Victoreen polynomial $`P=aE^3+bE^4`$ or by a more general polynomial $`P`$, coefficients of which are found by the least squares method from $`\mu (E)=P(E)`$ at energies lower than the edge.
Further, energy dependence is transformed to the photoelectron wave number dependence: $`k=\sqrt{2m_e(EE_0)}/\mathrm{}`$, where $`E_0`$ is the energy of the corresponding absorption edge. Usually, to the $`E_0`$ the energy at half the step is assigned or the energy of inflection point of $`\mu (E)`$. In most practical works the deviation of $`E_0`$ from true value, $`\mathrm{\Delta }E_0`$, is one of the fitting parameters.
The most difficult procedure in extracting of XAFS from the measured absorption is the construction of $`\mu _0`$ since one cannot definitely distinguish the environmental-born part of absorption from the atomic-like one. All methods for determination of the post-edge background are based on the assumption of its smoothness, and the only criterion for its validity is the absence of low-frequncy structure in $`\chi (k)k^w`$, i. e. the small absolute value of the Fourier transform (FT) $`\rho (r)`$ at low $`r`$. The review of existing post-edge background methods and the propositions of some new is the main purpose of the article.
Special attention must be paid to the estimation of noise and uncertainties in XAFS data. Experimental noise is shown to be essentially smaller than the errors of the background approximation, and it is the latter that determines the variances of structural parameters in subsequent fitting. The corresponding section of the present article is closely related with the next article devoted to the determining the errors of structural parameters .
All described in the article methods for background removal, its error estimations, and XAFS-function corrections are realized in the freeware program viper which allows one to vary several parameters by hand and watch the results simultaneously.
## II Methods of $`\mu _0`$ construction
### A Smoothing spline
Owing to fast algorithm and easy program realization, the approximation of $`\mu _0`$ by the smoothing spline has become widespread. Let $`N+1`$ experimental values of $`\mu _i`$ are defined on the mesh $`E_i`$. The smoothing spline $`\mu _0`$ minimizes the functional
$$J(\mu _0,\mu )=_{E_{\mathrm{min}}}^{E_{\mathrm{max}}}[\mu _0^{\prime \prime }(E)]^2𝑑E+\frac{1}{\alpha }\underset{i=0}{\overset{N}{}}(\mu _{0i}\mu _i)^2.$$
(2)
The smoothing parameter (or regularizer) $`\alpha `$ is the measure of compromise between smoothness of $`\mu _0`$ and its deviation from $`\mu `$. At $`\alpha =0`$ the smoothing spline exactly coincides with $`\mu `$, at $`\alpha \mathrm{}`$ it degenerates to $`\mu _0=const`$. Optimal regularizer should lead to $`\mu _0`$ containing only low-frequency oscillations and, hence, to $`\chi `$ containing only structural oscillations. The formulation of a new criterion for optimal $`\alpha `$ we shall consider below.
First, we address another problem, the well-known spline instability with respect to the small variations of input parameters: number of nodes, nodal values of the processed function, and limits on integral. In our case the spline is most sensitive to $`E_{\mathrm{min}}`$ due to fast growth of $`\mu `$ in the edge. To raise the stability the method was put forward in viper program which lies in the use of a prior information specifying the shape of $`\mu _0(E)`$ dependence. It is known in advance that the absorption edge without so-called white line constitutes nearly smooth step; the white line, if presents, is added to the step. Denote this prior function as $`p(E)`$. Now we will tend the second derivative of the sought $`\mu _0(E)`$ not to zero (at the specified deviation of $`\mu _0`$ from $`\mu `$) but to the second derivative of $`p(E)`$. The sought $`\mu _0(E)`$ is now minimizes the functional
$$J^{}(\mu _0,\mu )=_{E_{\mathrm{min}}}^{E_{\mathrm{max}}}[\mu _0^{\prime \prime }(E)p^{\prime \prime }(E)]^2𝑑E+\frac{1}{\alpha }\underset{i=0}{\overset{N+1}{}}[\mu _{0i}\mu _i]^2.$$
(3)
As seen, in fact there is no need to know $`p(E)`$ itself, its second derivative is sufficient. The explicit presence of $`p(E)`$ in the following formulas should be taken as a consequence of the technical trick applied: at first $`p(E)`$ is subtracted from the data, then it is added to the found spline.
Represent the second derivatives in finite-difference approximation, introduce $`\stackrel{~}{\mu }_{0i}=\mu _{0i}p_i`$, and denote $`\mathrm{\Delta }_i=E_{i+1}E_i`$:
$$J^{}(\mu _0,\mu )=\underset{i=1}{\overset{N}{}}[\stackrel{~}{\mu }_{0i1}\mathrm{\Delta }_{i1}^1\stackrel{~}{\mu }_{0i}(\mathrm{\Delta }_{i1}^1+\mathrm{\Delta }_i^1)+\stackrel{~}{\mu }_{0i+1}\mathrm{\Delta }_i^1]^2+\frac{1}{\alpha }\underset{i=0}{\overset{N+1}{}}[\stackrel{~}{\mu }_{0i}(\mu _ip_i)]^2=J(\stackrel{~}{\mu }_0,\mu _ip_i).$$
(4)
Thus, the problem is reduced to the preceding one in which instead of initial data $`\mu _i`$ the difference $`\mu _ip_i`$ is appeared. The sought $`\mu _0`$ is found from the smooth $`\stackrel{~}{\mu }_0`$ as $`\mu _{0i}=\stackrel{~}{\mu }_{0i}+p_i`$. In Fig. 1 is shown an example of the atomic-like absorption approximation by the smoothing spline with and without the use of prior function<sup>*</sup><sup>*</sup>*Here and hereafter for examples is used the spectrum at Bi $`L_3`$ absorption edge in Ba<sub>0.6</sub>K<sub>0.4</sub>BiO<sub>3</sub> at 50 K recorded in transmission mode at D-21 line (XAS-13) of DCI (LURE,Orsay, France) at positron beam energy 1.85 GeV and the average current $`250`$ mA. Energy step — 1 eV, counting time — 1 s. Energy resolution of the double-crystal Si monochromator (detuned to reject 50% of the incident signal in order to minimise harmonic contamination) with a 0.4 mm slit was about 2–3 eV at 13 keV.. Energy $`E_0`$ was determined at half the step height. Here, we constructed $`p(E)`$ in the following manner. Found the average value $`\overline{\mu }`$ of $`\mu (E)`$ in region $`20E70`$ eV above the absorptance maximum. Moving from the beginning of spectrum, assign $`p=\mu `$ until $`\mu >\overline{\mu }`$, further $`p=\overline{\mu }`$. Then $`p(E)`$ was smoothed 5 times on 3 points. To perform the Fourier transform, $`\chi (k)k^2`$ was brought into the uniform scale with $`\delta k=0.03`$ Å<sup>-1</sup> and multiplied by a Kaiser-Bessel window with parameter $`A=1.5`$. As seen, the use of $`p(E)`$ has led to disappearance of the spurious peak on the absolute value of FT at $`r0.5`$ Å.
So far we have considered the atomic-like absorption $`\mu _0`$ to be a smooth function with no peculiarities. However, in some spectra $`\mu _0`$ itself has a fine structure originating from resonance scattering within absorbing atom or from multi-electron transitions. If in these cases, based on theoretical calculations, experimental information, or empirical considerations, one can nearly indicate the location of peculiarities, their width and weight relatively to the step height, then one would readily construct the prior function $`p(E)`$ and find the correct $`\mu _0`$. Instead of constant value above absorption edge, the prior function would have corresponding valleys and/or peaks.
Let us now define the criterion for determination of smoothing parameter. An attempt to solve the problem was made in Ref. , where the requirement was proposed: $`H_RH_N0.05H_M`$, where $`H_R`$ is the average value of the weighted Fourier transform magnitude between 0 and 0.25 Å, $`H_M`$ is the maximum value in the transform magnitude between 1 and 5 Å, $`H_N`$ is the average value of the transform magnitude between 9 and 10 Å attributed to the noise. Obviously, that this criterion cannot pretend to the generality since depends on the weighting (op. cit., $`k^3`$) and the relative contribution of noise and the first coordination shell into spectra.
In the program viper we have proposed another approach to the problem based on the consideration of heights of FT peaks as functions of regularizer $`\alpha `$ (see Fig. 2). On increasing $`\alpha `$ from zero, $`\mu _0`$ starts to deviate from the experimental absorption $`\mu `$, $`\rho (r)`$ is growing and then saturates, the peaks at larger $`r`$ being saturated earlier. Clearly, that $`\alpha `$ should be determined by the first peak height since it is the last to saturate. Define the start of saturation on the minimum of second derivative of the first peak squared with respect to $`\mathrm{ln}\alpha `$. Declare the value of $`\alpha `$ in the minimum to be optimal. It is seen that the increase of $`\alpha `$ from the optimal leads to unwanted rapid growth of $`\rho `$ at low $`r`$.
In the example in Fig. 1 the regularizer is optimized following our new criterion.
Unfortunately, the method of smoothing spline does not include any approach to the estimations of uncertainties in the $`\mu _0`$ obtained, in contrast to the following two methods.
### B Interpolation spline drawn through the varied knots
The method was put forward in Ref. . $`N`$ knots are equally spaced in $`k`$ space, through them an interpolation spline is drawn. The ordinates of the knots are varied to minimize $`\rho `$ or $`|\rho \rho _{\mathrm{st}}|`$ in the chosen low-$`r`$ region $`0rr_0`$, where $`\rho _{\mathrm{st}}`$ is the absolute value of the FT of a “standard” $`\chi _{\mathrm{st}}(k)k^w`$, calculated or experimental. The number of knots must not exceed the value $`N_{\mathrm{max}}=2r_0\mathrm{\Delta }k/\pi +1`$, where $`\mathrm{\Delta }k`$ is the $`k`$ range of useful data. In the Ref. was asserted that one need to know the “standard” $`\chi _{\mathrm{st}}(k)k^w`$ merely approximately since it used only to get an estimate of the leakage from the first shell to the region minimized. The strange thing is that having omitted the question on the accuracy of found knots (as we show below, rather poor), the authors of the cited work made a fine comparison between several theoretical models for $`\chi (k)`$ calculations.
In Fig. 1 is shown an example of the method application. Ordinates of the 13 knots ($`N_{\mathrm{max}}=13.2`$) were varied to minimize the difference $`\rho \rho _{\mathrm{st}}`$ at $`0r1.05`$ Å. The function $`\chi _{\mathrm{st}}(k)`$ was calculated using feff6 program (as was pointed above, a crude estimate is sufficient, so details omitted). In the minimized region the $`\rho (r)`$ is somewhat better than that obtained by the previous method. However, at $`k>15`$ Å<sup>-1</sup> one can distinguish the obviously wrong behavior of $`\chi (k)k^2`$, and the first peak on $`\rho (r)`$ becomes quite distorted.
Consider now the problem of the accuracy of knot positions $`y_j`$, $`j=1,\mathrm{},N`$ in fitting $`\rho (r)`$ to $`\rho _{\mathrm{st}}(r)`$. As a figure of merit, the $`\chi ^2`$-statistics appears:
$$\chi ^2=\frac{N_{\mathrm{max}}}{M}\underset{m=1}{\overset{M}{}}\frac{[\rho (r_m)\rho _{\mathrm{st}}(r_m)]^2}{\sigma _m^2},$$
(5)
where $`\sigma _m`$ are the errors of $`\rho (r_m)`$. It can be shown (detailed analysis see in the next article ) that under the assumption of uncorrelated knot positions, the mean-square deviation of $`y_j`$ from the obtained through the fit optimal value $`\widehat{y}_j`$ equals $`\delta (y_j)=(\frac{1}{2}^2\chi ^2/y_j^2)^{1/2}`$, where the partial derivatives are calculated in the fitting procedure at the minimum. $`\sigma _m`$ are assumed to be constant and equal to the root-mean-square average of $`\rho (r)`$ between 15 and 25 Å, where solely the noise is present. The errors $`\epsilon _j=\delta (y_j)/[\mu _0(E_j)\mu _b(E_j)]`$ found under such assumptions are shown in Fig. 4 as open circles with the solid line. Notice, that the ussumption that the knot positions are not correlated gives quite optimistic $`\epsilon _j`$. Actually, several first knot positions appear to be highly correlated; the proper taking into account of the correlations (here we do not present these calculations) raises $`\epsilon _j`$ at the least as twice. But even these underestimated $`\epsilon _j`$ are appreciably larger than those given by the following method.
### C Bayesian smooth curve
Ideologically similar to the smoothing spline method is the method of bayesian smoothing (see Appendix on p. Bayesian smoothing and deconvolution) proposed in the program viper. This method also finds the regularized function $`\mu _0`$, the regularizer $`\alpha `$ is the measure of compromise between smoothness of $`\mu _0`$ and its deviation from $`\mu `$. In comparison with smoothing spline method, this method has some advantages. (i) Various prior information on $`\mu _0`$ can be considered. (ii) In this method the posterior distributions of all $`\mu _{0j}`$ are sought for. From those distributions one can find not only average values but also any desirable momenta, which appears to be an additional difficulty for other methods. (iii) In the framework of the method it is possible also to deconvolute $`\mu `$ with the monochromator rocking curve. The weakness of the method is its low speed (comparing with method II A, not with II B!). On a modern PC the curve drawn through $`N500`$ points is smoothed for a few minutes.
In Fig. 5 the bayesian smoothing was done on the mesh of 536 experimental points above $`E_0`$, without and with the prior function (its construction is described in Sec. II A). Besides, in the last case another information was used: the atomic-like absorption must coincide with the total absorption (minus pre-edge background) at energies $`E<E_0`$. Therefore, we demanded from the bayesian curve to pass through a point nearest (at left) to $`E_0`$. The values $`\overline{\mu }_{0j}`$ and $`\delta ^2(\mu _{0j})`$ were found by formulas (64) and (67). Since the smoothed values do not lie within the limits of $`\pm \delta (\mu _{0j})`$ from $`\mu _j`$, we did not look for the most probable smoothness (see. Appendix), instead we considered the regularizer to be known and equal to the optimal one found in the method II A. The introduction of the prior information has significantly diminished the errors of $`\chi (k)`$ extraction (see dotted curves in Fig. 4) which were defined as $`\epsilon _j=\delta (\mu _{0j})/(\overline{\mu }_{0j}\mu _{bj})`$. This is quite natural: any decrease of our ignorance about $`\mu _0`$ should narrow the posterior distribution of $`\mu _{0j}`$ for all $`j`$. Of course, this concerns the experimental information as well: errors $`\epsilon _j`$ are the less the more measured points $`N`$ the spectrum has. Comparing Fig. 1( ) and Fig. 5(c), it is seen practically perfect coincidence of the results of bayesian smoothing and smoothing spline. From this one can assume the equality of the errors which both methods give.
Could we take into account possible systematic errors in the framework of the method? Yes, if we have the information on their nature and are able to translate it into the mathematics language; such a translation might be rather non-trivial. In any case, now we have the tool to extract from the prior and experimental information not only the sought values but their errors as well.
### D Other methods
Consider briefly the methods for $`\mu _0`$ construction not included into the viper program.
A rich variety of computer programs for XAS spectra processing is collected on the International XAFS Society Web-site . The vast majority of them use as an approximation for the atomic-like absorption a smoothing spline or more general piecewise-polynomial representation. For example, in the method of Ref. , the construction of $`\mu _0`$ is divided into several stages: $`\mu _0`$ is approximated by a low-degree polynomial, obtained $`\chi (k)`$ is multiplied by $`k^w`$, additional $`\mu _0^{}`$ is drawn again as a low-degree polynomial and subtracted, a smoothing spline then approximates one more additional $`\mu _0^{\prime \prime }`$. The sum of all $`\mu _0`$’s gives the total atomic-like background. The necessity of the preliminary stages was not discussed op. cit., however, clearly it was caused by the instability of spline with respect to the small variations of input parameters. And the point is not that the preliminary stages make the process stable, but that for each specific spectrum, auxiliary parameters (degrees of polynomials) could provide an acceptable construction of the atomic-like background. Above (in Sec. II A) we proposed the way to rise the stability of spline making the preliminary stages to be redundant.
In Ref. an iterative approach to “atomic background” removal was developed. First a spline is used to obtain a rough estimate of the background; this alone is enough to have a reliable $`\chi `$ at $`k>56`$ Å<sup>-1</sup>. Over that range the $`\chi `$ obtained is fitted to the theoretical $`\chi _{\mathrm{th}}`$ in $`r`$-space. The resulting fit parameters are used then to generate $`\chi _{\mathrm{th}}(k)`$ that extends down to low $`k`$. This function is transformed back into $`e`$-space and $`\mu _0`$ is obtained as $`\mu _0=\mu /(\chi _{\mathrm{th}}+1)`$ that need be a little smoothed or fitted by an additional spline. Since the logic of reasoning was inversed: not “find $`\mu _0`$ to find $`\chi `$,” but “find $`\chi `$ to find $`\mu _0`$,” the method is suited for the quest of peculiarities on $`\mu _0`$ curve, not for structural XAFS-researches. Besides, the range of accuracy of the model appears to be unknown in principle: all, that is not described by the model, is included in $`\mu _0`$; the errors of the background approximation are also undefined.
In the old work for the determination of the background absorption $`\mu _0`$ was considered the damping of the XAFS amplitude resulting from the measurements with low resolutions (with a large slit width). The superimposition of two spectra measured with different energy resolutions gives the intersection points, the part of which belong to the $`\mu _0`$. Then through the obtained nodal points a smoothing spline is drawn. As the authors of Ref. noted, the measurements of the spectra with worsened resolution are not necessary; the spectra could be damped by the convolution with a “rocking curve,” approximated by a Gaussian function. Of course, the method is correctly works only with a small variation of the Gaussian curve width since for the large width not only the XAFS amplitude is damped but the very edge is washed out. Because of this only the extended part of a spectrum could be reliably determined.
The damping of the XAFS amplitude can be due to other reasons. For instance, as was pointed in Ref. , the nodal points may be obtained from the variable-temperature study. This idea was realized in Ref. and is more sound since the atomic-like background is really independent of temperature and with temperature the XAFS amplitude is changed, not the shape of the edge. But for all that it is important that the phase difference between XAFS of different temperatures was negligible, which is true only for low wave numbers. Unfortunately, the method is suitable only for some particular cases (to say nothing of need for measured temperature series of spectra). Op. cit. it was demonstrated for the x-ray-absorption data for the $`L_3`$ edge of solid Pb. In those spectra the first crossing of $`\mu `$ and $`\mu _0`$ occurs already at $`15`$ eV above edge. In our sample spectra the first crossing occurs only at $`30`$ eV, which allows one to find at most 2–3 points and the first of them being situated at $`k2.5`$ Å<sup>-1</sup>.
An interesting approach to the problem of $`\mu _0`$ determination was reported in Ref. . It is based on the simple identity that relates the FT of some function with the FT of its $`n`$-th derivative:
$$\mathrm{FT}[f^{(n)}(k)]=(2ir)^n\mathrm{FT}[f(k)],$$
(6)
where the conjugate variables are $`k`$ and $`2r`$. Since the atomic-like background is smooth enough, the higher derivatives $`\mu ^{(n)}(k)`$ ($`n2`$) are oscillatory near zero. Performing the FT of $`\mu ^{(n)}(k)k^w`$ and using Eq. (6), one obtains the FT of unnormalized would-be $`\chi (k)k^w`$ (see Fig. 6). Op. cit. the low-$`r`$ part (which in our example is $`0r1.1`$) was cut off, and then the back FT was done. As a result, one has the unnormalized $`\chi (k)k^w`$ and, having subtracted it from the $`\mu (k)`$, the atomic-like background on which some peculiarities due to multi-electron excitations can be distinguished. Like the method of Ref. , this method is suited for the quest of peculiarities on the $`\mu _0`$ curve, not for structural XAFS-researches because of evident distortion of the first peak on the FT by the contribution from the atomic-like background. To illustrate this assertion, in Fig. 6 we show the FT of the second derivative of the $`\mu _0(k)`$ that was found by the present method. As seen, this contribution is not as small.
If the electronic states of an absorbing atom in gaseous phase and in the compound of interest may be considered as equivalent, $`\mu _0`$ can be set equal to the measured absorption in gas, as was done in Ref for solid, liquid, and gaseous Kr. Some differences in energy positions and relative weights of double-electron excitation channels were taken into account by a model using simple empirical functions which were transferred then to the spectra of liquid and solid Kr. Notice that the proposed in the present paper prior function for the methods of smoothing spline and bayesian smoothing can include additional items corresponding to the multi-electron contributions.
## III Errors in $`\mu _0`$ constructing, noise, and choice of limits $`k_{\mathrm{min}}`$ $`k_{\mathrm{max}}`$
For what we need to know the errors of XAFS-function extraction? First, without knowing of these values one cannot in principle aim at their minimization. Second, they are used in the definition of $`\chi ^2`$-statistics in the fitting problems; their underestimation is a source of unjustified optimistic errors of fitting parameters. Third, along with analysis of the noise, the errors of $`\mu _0`$ construction allow us to choose the limits of reliable EXAFS signal, $`k_{\mathrm{min}}`$ and $`k_{\mathrm{max}}`$.
Unfortunately, the issue of quality of XAFS extraction from the measured absorption has not been addressed properly. We see several reasons for that. On the one hand, not having a correctly developed approach to the estimation of the errors of final results (interatomic distances, Debye-Waller factors etc. found via fitting), the errors of EXAFS extraction are useful. On the other hand, only a few methods include approaches to their estimations.
Easily one can compare the errors of different methods (see Fig. 4) and then choose the most reliable one. The problem of plausible limitations on the absolute value of the errors is more difficult. Define “signal” as the envelope of $`\chi (k)`$ (solid line with gray filling in Fig. 4). It is quite reasonable to demand that the errors of $`\mu _0`$ construction were less than XAFS signal. For the method of the interpolation spline drawn through the varied knots to meet this requirement leads to the restriction on the photoelectron wave numbers: $`2k14`$ Å<sup>-1</sup>. For the bayesian curve $`a`$ this range is $`0k14`$ Å<sup>-1</sup>, for the bayesian curves $`b`$ and $`c`$ this range is wider: $`0k16`$ Å<sup>-1</sup>.
Another factor that limitates the spectrum length is the presence of noise. To determine the noise is a straightforward task for $`r`$-space, where XAFS signals at high $`r`$ have clearly noise character. By Parseval’s identity the noise in $`r`$-space is related with the noise in $`k`$-space :
$`{\displaystyle _{k_{\mathrm{min}}}^{k_{\mathrm{max}}}}|n_kk^w|^2𝑑k=2{\displaystyle _0^{\pi /2dk}}|n_r|^2𝑑r.`$ (7)
Substitute the mean value over the range $`15<r<25`$ Å of the FT magnitude squared for $`|n_r|^2`$. Then
$`n_k^2=|n_r^2|{\displaystyle \frac{\pi }{dk}}{\displaystyle \frac{2w+1}{k_{\mathrm{max}}^{2w+1}k_{\mathrm{min}}^{2w+1}}}.`$ (8)
As seen from the formula, $`n_k`$ depends on $`dk`$, the size of evenly-spaced $`k`$-grid. Although above we already have used the Fourier transform, the question of choice of $`dk`$ was not raised yet. The algorithm of fast FT needs the transformed function to be set on a uniform grid. Having chosen a small $`dk`$, we artificially obtain the large number of “experimental” values. Naturally, this trick would not give more information than we have, and the errors $`n_k`$ must be large at the small $`dk`$. In our example the choice of $`dk`$ (0.03 Å<sup>-1</sup>) was based on the equality of numbers of experimental points and the nodes of the grid. The signal-to-noise ratio obtained is greater than unity for all the spectrum (see Fig. 4). There was no doubt in that: the signal is visually distinguished even for the very extended end of the spectrum (see Fig. 1(b) and Fig. 5(b)).
The noise can be estimated based on the bayesian considerations . Let after measurements we have the values of counts from the solid-state or gas-filled detectors and let there is a positive real number $`\lambda `$ such that the probability that a single count occurs in the time interval $`dt`$ is
$`P(1|\lambda )=\lambda dt.`$ (9)
It can be shown that merely from this assumption follows that the counts obey the Poisson distribution law:
$`P(N|\lambda ,T)={\displaystyle \frac{(\lambda T)^N\mathrm{exp}(\lambda T)}{N!}},`$ (10)
where $`T`$ is the sampling time. The problem is to find the intensity $`\lambda `$ and its variance. Using Bayes theorem and introducing prior probabilities $`P(N)=1/N`$ and $`P(\lambda )=1/\lambda `$ , one obtains:
$`P(\lambda |N,T)={\displaystyle \frac{P(N|\lambda ,T)P(\lambda )}{P(N)}}={\displaystyle \frac{T(\lambda T)^{N1}\mathrm{exp}(\lambda T)}{(N1)!}},`$ (11)
that is after measurement the variate $`2T\lambda `$ follows the $`\chi ^2`$-distribution with $`2N`$ degrees of freedom. It is easy to find that $`\overline{\lambda }=N/T`$, $`\overline{\lambda ^2}=N(N+1)/T^2`$, and the variance of intensity is $`\delta \lambda =\sqrt{N}/T`$.
Denote counts from detectors measuring $`i_0`$ and $`i_1`$ as $`I_0`$ and $`I_1`$. By definition the variate $`\xi =\frac{i_0/2I_0}{i_1/2I_1}`$ follows Fisher’s $`F`$-distribution with $`(2I_0,2I_1)`$ degrees of freedom. Its expected value and variance are known: $`\overline{\xi }=I_1/(I_11)`$, $`\delta ^2\xi =I_1^2(I_0+I_11)/((I_11)^2(I_12)I_0)`$, from where we find for the absorption in the fluorescence mode ($`\mu x=i_0/i_1`$):
$`\overline{i_0/i_1}={\displaystyle \frac{I_0}{I_11}},\delta ^2(i_0/i_1)={\displaystyle \frac{I_0(I_0+I_11)}{(I_11)^2(I_12)}}.`$ (12)
Further, the variate $`\eta =\frac{1}{2}\mathrm{ln}\xi `$ follows $`z`$-distribution (Fisher’s distribution of variance ratio) with $`(2I_0,2I_1)`$ degrees of freedom. Its expected value and variance are known: $`\overline{\eta }=0`$, $`\delta ^2\eta =\frac{1}{4}(I_1+I_1)/(I_0I_1)`$, from where we find for the absorption in the transmission mode ($`\mu x=\mathrm{ln}(i_0/i_1)`$):
$`\overline{\mathrm{ln}(i_0/i_1)}=\mathrm{ln}{\displaystyle \frac{I_0}{I_1}},\delta ^2\mathrm{ln}(i_0/i_1)={\displaystyle \frac{1}{I_0}}+{\displaystyle \frac{1}{I_1}}.`$ (13)
The noise of XAFS-function is
$`n_P={\displaystyle \frac{\delta \mu }{\mu _0\mu _b}}=\left({\displaystyle \frac{1}{I_0}}+{\displaystyle \frac{1}{I_1}}\right)^{1/2}{\displaystyle \frac{1}{\mu _0\mu _b}}.`$ (14)
In our example this noise at $`k15`$ Å<sup>-1</sup> becomes greater than signal (see Fig. 4). What is the reason for such significant difference between the really present noise $`n_k`$ and its statistical estimate $`n_P`$? Of course, the reason is in the false premise (9). In practice this condition is realized as: $`P(c|\lambda )=\lambda dt`$. For example, the photocurrent in an ion-chamber depends on gas pressure, potential applied etc.; these dependencies are contained in $`c`$. In other words, the amplification path works in such a way that one photon gives birth to $`c`$ counts. There is no difficulty in writing the posterior distribution for the generalized premise:
$`P(\lambda |N,T)={\displaystyle \frac{T(\lambda T)^{\frac{N}{c}1}\mathrm{exp}(\lambda T)}{\mathrm{\Gamma }(N/c)}},`$ (15)
with $`\overline{\lambda }=N/(cT)`$ and $`\delta \lambda =\sqrt{N/c}/T`$. Thus, having unknown $`c`$ (and implicitly assigning $`c=1`$), we got wrong variances for $`i_0/i_1`$ and $`\mathrm{ln}(i_0/i_1)`$. Unfortunately, in the most of real experiments the association between the probability of a single count event and the radiation intensity (via $`c`$) is unknown. In spite of this, the Poisson counting statistics is traditionally used for a long time. For example, in Ref. signal-to-noise ratios are evaluated (assuming $`c=1`$) for the different detection schemes.
Practically all programs for XAFS spectra processing to estimate the noise use the Fourier analysis. But then it is the noise that they use as uncertainties $`\epsilon _i`$ of $`\chi (k)`$ determination in definition of $`\chi ^2`$-statistics:
$`\chi ^2={\displaystyle \frac{N_{\mathrm{max}}}{M}}{\displaystyle \underset{i=1}{\overset{M}{}}}{\displaystyle \frac{[(\chi _{\mathrm{exp}})_i(\chi _{\mathrm{mod}})_i]^2}{\epsilon _i^2}}.`$ (16)
It would be more correct to consider as $`\epsilon _i`$ the larger from the two: the noise and the errors of the construction of $`\mu _0`$. In our case (and as a rule) the latter are essentially greater (especially in the method II B) than the noise. In the following paper we shall show how the understated $`\epsilon _i`$ lead to optimistic errors of structural parameters.
## IV XAFS-function correction
Because of one reason or another the experimental XAFS might be distorted. Consider some of them.
(i) Let the counts ($`I`$) from detectors are associated with the intensities ($`i`$) as $`i_0=\varkappa _0I_0`$ and $`i_1=\varkappa _1I_1`$. Then the absorption (in the transmission mode) equals:
$`\mu x=\mathrm{ln}(i_0/i_1)=\mathrm{ln}(I_0/I_1)+\mathrm{ln}(\varkappa _0/\varkappa _1).`$ (17)
The second term is a slightly varied function of energy and can be taken into account in independent experiments. Such a distortion appears, for instance, if the absorptance of the gas in ion-chamber detectors depends on energy.
(ii) If some part of incident radiation is not attenuated in the sample as much as expected (due to the pinholes in the sample, harmonics in the incoming beam etc.), that is $`i_0=\varkappa _0I_0+b`$, then the real absorption is connected with the measured $`I_0`$ and $`I_1`$ in a complicated way. In Ref. the possible decrease of XAFS amplitude shown to be essential even at low $`b/(\varkappa _0I_0)`$ but thick samples. At known ratio $`b/(\varkappa _0I_0)`$, the correcting factor can be easily obtained.
(iii) In the fluorescence mode, due to absorptance of the fluorescent signal in the sample itself XAFS spectra strongly depend on the detection geometry. In Ref. the correcting functions are found explicitly.
(iv) The problem of glitches is widespread in the XAFS analysis. The glitches are due to multiple Bragg reflection being satisfied simultaneously and for each given monochromator are manifested in the strictly determined spectral positions. In most cases the glitches seen on curves $`I_0(E)`$ and $`I_1(E)`$, vanish on $`I_0/I_1`$ ratio. If not, one can easily get rid of them. For instance, the glitch area, usually extremely thin, is smoothed or, with fixed ends, replaced by a straight-line segment. The main thing in the correct analysis of glitches is their detection.
To detect a glitch on curves $`\mu `$ or $`\chi k^w`$ is practically impossible. For this, one needs the primary data $`I_0(E)`$ and $`I_1(E)`$, not $`\mathrm{ln}(I_0/I_1)`$ nor $`I_0/I_1`$. Out of glitches the intensity of incident radiation smoothly, ignoring the noise, depends on energy (see Fig. 7). The idea of detection of glitches via critical level for the derivative $`|d\mathrm{ln}I_0/dE|_c`$ is self-evident. For the presented in Fig. 7 $`I_0(E)`$ curve the absolute value of the derivative in the glitch areas is greater than the critical value chosen to be equal to $`1.7710^3`$<sup>-1</sup>. Having extracted the XAFS, one can see that the first (paired) glitch $`a`$ is not manifested on $`\chi (k)k^2`$, the last two ($`c`$ $`d`$) are obscured by the noise, solely glitch $`b`$ is clearly pronounced. Now, being in the firm belief that this is not a part of the XAFS, one can eliminate the glitch with ease. Here, we fixed its ends on $`\mu (E)`$, replaced it by the straight-line segment, and constructed $`\chi (k)k^2`$ again.
## V Conclusion
In this paper we have considered all stages of XAFS function extraction from the measured absorption. We focused our attention on the most important stage, construction of the atomic-like absorption $`\mu _0`$.
For the wide-spread method of approximation of $`\mu _0`$ by a smoothing spline we have proposed the way to raise the stability by including the prior information about absorption edge shape (“nearly step” or “nearly step with a white line”). Besides we have propose a new reliable criterion for determination of the smoothing regularizer.
A new method for approximation of $`\mu _0`$ is proposed, the method of bayesian smoothing. It can include various prior information, which raises the accuracy of XAFS determination. Following this method one finds the distributions of $`\mu _0`$ in each experimental point, from which one can find not only average values but also any desirable momenta, which appears to be an additional difficulty for other methods. This method was shown to give more accurate atomic-like background than that obtained by the method of Ref .
Particular attention has been given to the analysis of noise. We have discussed the difficulties of its estimates on the basis of statistical approach. More reliable is the determination of noise from the Fourier transform. We have shown that the experimental noise is essentially less than the errors of $`\mu _0`$ construction, and the use of values of noise in the $`\chi ^2`$-statistics definition appears to be erroneous since leads to the unjustified optimistic errors of structural parameters inferred in fitting procedures. For detailed consideration of the accuracy of fitting parameters see the following paper .
## Bayesian smoothing and deconvolution
### 1 Posterior distribution for smoothed data
Consider general linear problem of data smoothing with the use of statistical methods (for introduction see review by Turchin et al. and the articles from Web-site bayes.wustl.edu). Let data $`𝐝`$ are defined on the mesh $`x_1,\mathrm{},x_N`$ and consist of the true values $`𝐭`$ and the additive noise $`𝐧`$:
$$d_i=t_i+n_i,i=1,\mathrm{},N.$$
(18)
The problem of smoothing is to find the best estimates of $`𝐭`$. For an arbitrary node $`j`$, find the probability density function for $`t_j`$ given the data $`𝐝`$:
$$P(t_j|𝐝)=\mathrm{}𝑑t_{ij}\mathrm{}P(𝐭|𝐝),$$
(19)
where $`P(𝐭|𝐝)`$ is the joint probability density function for all values $`𝐭`$, and the integration is done over all $`t_{ij}`$. According to Bayes theorem,
$$P(𝐭|𝐝)=\frac{P(𝐝|𝐭)P(𝐭)}{P(𝐝)},$$
(20)
$`P(𝐭)`$ being the joint prior probability for all $`t_i`$, $`P(𝐝)`$ is a normalization constant. Assuming that the values $`n_i`$ are independent in different nodes and normally distributed with zero expected values, the probability $`P(𝐝|𝐭)`$, so-called likelihood function, is given by
$$P(𝐝|𝐭,\sigma )=(2\pi \sigma ^2)^{N/2}\mathrm{exp}\left(\frac{1}{2\sigma ^2}\underset{k=1}{\overset{N}{}}(d_kt_k)^2\right),$$
(21)
where the standard deviation of the noise, $`\sigma `$, appears as a known value. Later, we apply the rules of probability theory to remove $`\sigma `$ from the problem.
Now define prior probability $`P(𝐭)`$. Let we know in advance that the function $`t(x)`$ is smooth enough. To specify this information, introduce the norm of the second derivative and indicate its expected approximate value:
$$\mathrm{\Omega }(t(x))=\left(\frac{d^2t}{dx^2}\right)^2𝑑x\omega .$$
(22)
Denote $`\mathrm{\Delta }_i=x_{i+1}x_i`$, $`i=1,\mathrm{},N1`$ and represent the second derivative in the finit-difference form:
$$\mathrm{\Omega }(t(x))\mathrm{\Omega }(𝐭)=\underset{i=2}{\overset{N1}{}}[t_{i1}\mathrm{\Delta }_{i1}^1t_i(\mathrm{\Delta }_{i1}^1+\mathrm{\Delta }_i^1)+t_{i+1}\mathrm{\Delta }_i^1]^2\underset{k,l=1}{\overset{N}{}}\mathrm{\Omega }_{kl}t_kt_l.$$
(23)
$`\mathrm{\Omega }_{kl}`$ is a five-diagonal symmetric matrix with the following non-zero elements:
$`\mathrm{\Omega }_{11}`$ $`=`$ $`\mathrm{\Delta }_1^1\mathrm{\Delta }_2^2,\mathrm{\Omega }_{22}=\mathrm{\Delta }_2^1(\mathrm{\Delta }_1^1+\mathrm{\Delta }_2^1)^2+\mathrm{\Delta }_2^2\mathrm{\Delta }_3^1,\mathrm{\Omega }_{12}=(\mathrm{\Delta }_1\mathrm{\Delta }_2)^1(\mathrm{\Delta }_1^1+\mathrm{\Delta }_2^1),`$ (25)
$`\mathrm{}\mathrm{}\mathrm{}\mathrm{}`$
$`\mathrm{\Omega }_{ii}`$ $`=`$ $`\mathrm{\Delta }_i^1(\mathrm{\Delta }_i^1+\mathrm{\Delta }_{i1}^1)^2+\mathrm{\Delta }_i^2\mathrm{\Delta }_{i+1}^1+\mathrm{\Delta }_{i1}^3,`$ (26)
$`\mathrm{\Omega }_{i1,i}`$ $`=`$ $`\mathrm{\Delta }_{i1}^2(\mathrm{\Delta }_{i1}^1+\mathrm{\Delta }_{i2}^1)(\mathrm{\Delta }_{i1}\mathrm{\Delta }_i)^1(\mathrm{\Delta }_{i1}^1+\mathrm{\Delta }_i^1),`$ (27)
$`\mathrm{\Omega }_{i2,i}`$ $`=`$ $`\mathrm{\Delta }_{i2}^1\mathrm{\Delta }_{i1}^2,`$ (29)
$`\mathrm{}\mathrm{}\mathrm{}\mathrm{}`$
$`\mathrm{\Omega }_{NN}`$ $`=`$ $`\mathrm{\Delta }_{N1}^3,\mathrm{\Omega }_{N1,N1}=\mathrm{\Delta }_{N1}^1(\mathrm{\Delta }_{N1}^1+\mathrm{\Delta }_{N2}^1)^2+\mathrm{\Delta }_{N2}^3,\mathrm{\Omega }_{N1,N}=\mathrm{\Delta }_{N1}^2(\mathrm{\Delta }_{N1}^1+\mathrm{\Delta }_{N2}^1).`$ (30)
In order to introduce the minimum information in addition to that contained in (23), from all normalized to unity functions $`P(𝐭)`$ which satisfy the condition (23) we choose a single one that contains minimum information about t that is minimizes the functional
$$I[P(𝐭)]=P(𝐭)\mathrm{ln}P(𝐭)𝑑𝐭+\beta \left[1P(𝐭)𝑑𝐭\right]+\gamma \left[\omega \mathrm{\Omega }(𝐭)P(𝐭)𝑑𝐭\right],$$
(31)
where $`\beta `$ and $`\gamma `$ are the Larrange multipliers. In minimizing $`I[P(𝐭)]`$, one obtains the equation set
$`\mathrm{ln}P(𝐭)+1\beta \gamma \mathrm{\Omega }(𝐭)`$ $`=`$ $`0`$ (32)
$`{\displaystyle P(𝐭)𝑑𝐭}`$ $`=`$ $`1`$ (33)
$`{\displaystyle \mathrm{\Omega }(𝐭)P(𝐭)𝑑𝐭}`$ $`=`$ $`\omega ,`$ (34)
that has a solution:
$$P(𝐭)=(\lambda _1\mathrm{}\lambda _N)^{1/2}\left(\frac{2\pi \sigma ^2}{\alpha }\right)^{N/2}\mathrm{exp}\left(\frac{\alpha }{2\sigma ^2}\mathrm{\Omega }(𝐭)\right),$$
(35)
where $`\alpha /2\sigma ^2=\gamma =N/2\omega `$, and $`\lambda _1,\mathrm{},\lambda _N`$ are the eigenvalues of the matrix $`\mathrm{\Omega }_{kl}`$. The regularizer $`\alpha `$ will be used to control the smoothness of $`𝐭`$. The prior distribution obtained is a “soft” one, that is does not demand from the solution to have a strictly prescribed form.
Thus, we have for the probability density function:
$`P(t_j|𝐝,\sigma ,\alpha )`$ $``$ $`{\displaystyle \mathrm{}𝑑t_{ij}\mathrm{}\sigma ^{2N}\alpha ^{N/2}\mathrm{exp}\left(\frac{\alpha }{2\sigma ^2}\underset{k,l=1}{\overset{N}{}}\mathrm{\Omega }_{kl}t_kt_l\right)\mathrm{exp}\left(\frac{1}{2\sigma ^2}\underset{k=1}{\overset{N}{}}(d_kt_k)^2\right)}`$ (36)
$`=`$ $`{\displaystyle \mathrm{}𝑑t_{ij}\mathrm{}\sigma ^{2N}\alpha ^{N/2}\mathrm{exp}\left(\frac{1}{2\sigma ^2}\left[𝐝^22\underset{k=1}{\overset{N}{}}d_kt_k+\underset{k,l=1}{\overset{N}{}}g_{kl}t_kt_l\right]\right)},`$ (37)
where
$`g_{kl}=\alpha \mathrm{\Omega }_{kl}+\delta _{kl},𝐝^2={\displaystyle \underset{k=1}{\overset{N}{}}}d_k^2.`$ (38)
Since there is no integral over $`t_j`$, separate it from the other integration variables:
$`P(t_j|𝐝,\sigma ,\alpha )`$ $``$ $`\sigma ^{2N}\alpha ^{N/2}\mathrm{exp}\left({\displaystyle \frac{1}{2\sigma ^2}}[𝐝^22d_jt_j+g_{jj}t_j^2]\right)`$ (40)
$`\times {\displaystyle }\mathrm{}dt_{ij}\mathrm{}\mathrm{exp}({\displaystyle \frac{1}{2\sigma ^2}}[\underset{k,l=1}{\overset{N}{{\displaystyle }^j}}g_{kl}t_kt_l2\underset{k=1}{\overset{N}{{\displaystyle }^j}}[d_kg_{kj}t_j]t_k]),`$
Here, the symbol $`j`$ near the summation signs denotes the absence of $`j`$-th item. Further, find the eigenvalues $`\lambda _i^{}`$ and corresponding eigenvectors $`𝐞_i`$ of the matrix $`g_{kl}`$ in which the $`j`$-th row and column are deleted, and change the variables:
$$b_i=\sqrt{\lambda _i^{}}\underset{k=1}{\overset{N}{^j}}t_ke_{ik},t_k=\underset{i=1}{\overset{N}{^j}}\frac{b_ie_{ik}}{\sqrt{\lambda _i^{}}}(i,kj).$$
(41)
Using the properties of eigenvectors:
$$\underset{k=1}{\overset{N}{^j}}g_{lk}e_{ik}=\lambda _i^{}e_{il},\underset{k=1}{\overset{N}{^j}}e_{lk}e_{ik}=\delta _{li}(l,ij),$$
(42)
one obtains:
$`P(t_j|𝐝,\sigma ,\alpha )`$ $``$ $`\sigma ^{2N}\alpha ^{N/2}\mathrm{exp}\left({\displaystyle \frac{1}{2\sigma ^2}}[(𝐝^2𝐡^2)2t_j(d_j\mathrm{𝐡𝐮})+t_j^2(g_{jj}𝐮^2)]\right)`$ (44)
$`\times {\displaystyle }\mathrm{}db_{lj}\mathrm{}\mathrm{exp}({\displaystyle \frac{1}{2\sigma ^2}}\underset{i=1}{\overset{N}{{\displaystyle }^j}}[b_ih_i+u_it_j]^2),`$
where new quantities were introduced:
$`h_i={\displaystyle \frac{1}{\sqrt{\lambda _i^{}}}}\underset{k=1}{\overset{N}{{\displaystyle }^j}}d_ke_{ik},u_i={\displaystyle \frac{1}{\sqrt{\lambda _i^{}}}}\underset{k=1}{\overset{N}{{\displaystyle }^j}}g_{kj}e_{ik},`$ (45)
$`𝐡^2=\underset{i=1}{\overset{N}{{\displaystyle }^j}}h_i^2,𝐮^2=\underset{i=1}{\overset{N}{{\displaystyle }^j}}u_i^2,\mathrm{𝐡𝐮}=\underset{i=1}{\overset{N}{{\displaystyle }^j}}h_iu_i.`$ (46)
Evaluating the $`N1`$ integrals in (44), one finally obtains the posterior probability for $`j`$-th node:
$$P(t_j|𝐝,\sigma ,\alpha )\sigma ^{(N+1)}\alpha ^{N/2}\mathrm{exp}\left(\frac{1}{2\sigma ^2}[(𝐝^2𝐡^2)2t_j(d_j\mathrm{𝐡𝐮})+t_j^2(g_{jj}𝐮^2)]\right).$$
(47)
### 2 Eliminating nuisance parameters
In most real problems $`\sigma `$ and $`\alpha `$ are not known. To eliminate $`\sigma `$ is a quite straightforward problem:
$$P(t_j|𝐝,\alpha )=𝑑\sigma P(t_j,\sigma |𝐝,\alpha )=𝑑\sigma P(\sigma )P(t_j|𝐝,\sigma ,\alpha ),$$
(48)
one needs only to know a prior probability $`P(\sigma )`$. Having no specific information about $`\sigma `$, a Jeffreys prior $`P(\sigma )=1/\sigma `$ is assigned . Then
$`P(t_j|𝐝,\alpha )`$ $``$ $`{\displaystyle _0^{\mathrm{}}}𝑑\sigma \sigma ^{(N+2)}\mathrm{exp}\left({\displaystyle \frac{1}{2\sigma ^2}}[(𝐝^2𝐡^2)2t_j(d_j\mathrm{𝐡𝐮})+t_j^2(g_{jj}𝐮^2)]\right)`$ (49)
$``$ $`[(𝐝^2𝐡^2)2t_j(d_j\mathrm{𝐡𝐮})+t_j^2(g_{jj}𝐮^2)]^{(N+1)/2}.`$ (50)
Introducing the substitution
$`w_j^2=N{\displaystyle \frac{(g_{jj}𝐮^2)^2}{(𝐝^2𝐡^2)(g_{jj}𝐮^2)(d_j\mathrm{𝐡𝐮})^2}}\left(t_j{\displaystyle \frac{d_j\mathrm{𝐡𝐮}}{g_{jj}𝐮^2}}\right)^2,`$ (51)
one obtains the Student $`t`$-distribution with $`N`$ degrees of freedom:
$`P(w_j|𝐝,\alpha )\left(1+{\displaystyle \frac{w_j^2}{N}}\right)^{(N+1)/2}`$ (52)
with zero average and the variance $`N/(N2)`$. From where one finds for $`t_j`$:
$`\overline{t}_j={\displaystyle \frac{d_j\mathrm{𝐡𝐮}}{g_{jj}𝐮^2}},\delta ^2(t_j)={\displaystyle \frac{(𝐝^2𝐡^2)(g_{jj}𝐮^2)(d_j\mathrm{𝐡𝐮})^2}{(g_{jj}𝐮^2)^2}}{\displaystyle \frac{1}{N2}}.`$ (53)
Thus, we have got rid of unknown $`\sigma `$ and found the expressions for mean values $`t_j`$ and their dispersions at known regularizer $`\alpha `$. To eliminate the latter is more difficult. The idea is not to find the smoothest solution, but the solution of the most probable smoothness. For that we will find the posterior probability:
$`P(\alpha |𝐝)={\displaystyle 𝑑𝐭𝑑\sigma P(\alpha ,\sigma ,𝐭|𝐝)}={\displaystyle 𝑑𝐭𝑑\sigma P(\alpha ,\sigma )P(𝐭|\alpha ,\sigma ,𝐝)}.`$ (54)
Assuming that $`\alpha `$ and $`\sigma `$ are independent and using Bayes theorem (20), one obtains:
$`P(\alpha |𝐝){\displaystyle 𝑑𝐭𝑑\sigma P(\alpha )P(\sigma )P(𝐭|\alpha ,\sigma )P(𝐝|𝐭,\alpha ,\sigma )}.`$ (55)
Substituting (35) for the prior probability $`P(𝐭|\alpha ,\sigma )`$, (21) for the likelihood, and a Jeffreys prior $`P(\sigma )=1/\sigma `$ and $`P(\alpha )=1/\alpha `$, one obtains the posterior distribution for the regularizer $`\alpha `$:
$`P(\alpha |𝐝)`$ $``$ $`{\displaystyle 𝑑𝐭𝑑\sigma \sigma ^{2N1}\alpha ^{N/21}\mathrm{exp}\left(\frac{\alpha }{2\sigma ^2}\underset{k,l=1}{\overset{N}{}}\mathrm{\Omega }_{kl}t_kt_l\right)\mathrm{exp}\left(\frac{1}{2\sigma ^2}\underset{k=1}{\overset{N}{}}(d_kt_k)^2\right)}`$ (56)
$`=`$ $`{\displaystyle 𝑑𝐭𝑑\sigma \sigma ^{2N1}\alpha ^{N/21}\mathrm{exp}\left(\frac{1}{2\sigma ^2}\left[𝐝^22\underset{k=1}{\overset{N}{}}d_kt_k+\underset{k,l=1}{\overset{N}{}}g_{kl}t_kt_l\right]\right)},`$ (57)
where matrix $`g_{kl}`$ was defined in (38). After its diagonalization, analogously to what was done above, finally one obtains:
$`P(\alpha |𝐝)(\lambda _1^{}\mathrm{}\lambda _N^{})^{1/2}\alpha ^{N/21}[𝐝^2𝐡^2]^{N/2},`$ (58)
where $`𝐡^2`$ is given by
$`𝐡^2={\displaystyle \underset{i=1}{\overset{N}{}}}h_i^2,h_i={\displaystyle \frac{1}{\sqrt{\lambda _i^{}}}}{\displaystyle \underset{k=1}{\overset{N}{}}}d_ke_{ik},`$ (59)
and $`\lambda _i^{}`$ and $`𝐞_i`$ are the eigenvalues and eigenvectors of $`g_{kl}`$. Having found the maximum of the posterior probability (58) or having averaged over it the expression (53), one has the sought $`𝐭`$ with the most probable smoothness. However it is necessary to point out that this procedure narrows the applicability of the bayesian smoothing down to the class of tasks where the smoothed values lie in most within the limits $`\pm \sigma `$ from the most probable. In practice, there possible other tasks where the condition (18) is treated more wider and the smoothed values exceed the bounds of noise.
### 3 Expressions for smoothed values and their variances
The formulas (53) appear useless in practice since require to find the eigenvalues and eigenvectors for the matrix of rank $`N1`$ on each node. Those formulas have merely a methodological value: the explicit expressions for posterior probabilities enable one to find the average of arbitrary function of $`t_j`$. However, $`\overline{t}_j`$ and $`\delta ^2(t_j)`$ could be found significantly easier. Using (48) and (36), represent $`\overline{t}_j`$ as:
$`P(\sigma )P(t_j|𝐝,\sigma ,\alpha )dt_j`$ (60)
$``$ $`{\displaystyle 𝑑𝐭𝑑\sigma \sigma ^{2N1}t_j\mathrm{exp}\left(\frac{1}{2\sigma ^2}\left[𝐝^22\underset{k=1}{\overset{N}{}}d_kt_k+\underset{k,l=1}{\overset{N}{}}g_{kl}t_kt_l\right]\right)}.`$ (61)
Performing the diagonalization, one obtains:
$`\overline{t}_j`$ $``$ $`{\displaystyle 𝑑𝐛𝑑\sigma \sigma ^{2N1}\mathrm{exp}\left(\frac{1}{2\sigma ^2}[𝐝^2𝐡^2]\right)\left(\underset{i=1}{\overset{N}{}}\frac{b_ie_{ij}}{\sqrt{\lambda _i^{}}}\right)\mathrm{exp}\left(\frac{1}{2\sigma ^2}\underset{i=1}{\overset{N}{}}[b_ih_i]^2\right)}`$ (62)
$``$ $`{\displaystyle \underset{i=1}{\overset{N}{}}}{\displaystyle \frac{h_ie_{ij}}{\sqrt{\lambda _i^{}}}}{\displaystyle 𝑑\sigma \sigma ^{N1}\mathrm{exp}\left(\frac{1}{2\sigma ^2}[𝐝^2𝐡^2]\right)},`$ (63)
from where
$`\overline{t}_j={\displaystyle \underset{i=1}{\overset{N}{}}}{\displaystyle \frac{h_ie_{ij}}{\sqrt{\lambda _i^{}}}}.`$ (64)
Analogously, for the variance $`\delta (t_j)`$ one has:
$`\delta ^2(t_j)`$ $``$ $`{\displaystyle 𝑑𝐛𝑑\sigma \sigma ^{2N1}\mathrm{exp}\left(\frac{1}{2\sigma ^2}[𝐝^2𝐡^2]\right)\left(\underset{i=1}{\overset{N}{}}\frac{(b_ih_i)e_{ij}}{\sqrt{\lambda _i^{}}}\right)^2\mathrm{exp}\left(\frac{1}{2\sigma ^2}\underset{i=1}{\overset{N}{}}[b_ih_i]^2\right)}`$ (65)
$``$ $`{\displaystyle \underset{i=1}{\overset{N}{}}}{\displaystyle \frac{e_{ij}^2}{\lambda _i^{}}}{\displaystyle 𝑑\sigma \sigma ^{N1}\mathrm{exp}\left(\frac{1}{2\sigma ^2}[𝐝^2𝐡^2]\right)\sigma ^2}.`$ (66)
Normalizing, one finally obtains:
$`\delta ^2(t_j)`$ $`=`$ $`{\displaystyle \frac{𝑑\sigma \sigma ^{N+1}\mathrm{exp}\left([𝐝^2𝐡^2]/2\sigma ^2\right)}{𝑑\sigma \sigma ^{N1}\mathrm{exp}\left([𝐝^2𝐡^2]/2\sigma ^2\right)}}{\displaystyle \underset{i=1}{\overset{N}{}}}{\displaystyle \frac{e_{ij}^2}{\lambda _i^{}}}`$ (67)
$`=`$ $`{\displaystyle \frac{\mathrm{\Gamma }\left(\frac{N}{2}1\right)}{\left([𝐝^2𝐡^2]/2\right)^{N/21}}}{\displaystyle \frac{\left([𝐝^2𝐡^2]/2\right)^{N/2}}{\mathrm{\Gamma }\left(\frac{N}{2}\right)}}{\displaystyle \underset{i=1}{\overset{N}{}}}{\displaystyle \frac{e_{ij}^2}{\lambda _i^{}}}={\displaystyle \frac{[𝐝^2𝐡^2]}{N2}}{\displaystyle \underset{i=1}{\overset{N}{}}}{\displaystyle \frac{e_{ij}^2}{\lambda _i^{}}}.`$ (68)
Now we got the usable formulas, which require to find the eigenvalues and eigenvectors for the matrix of rank $`N`$ just one time.
### 4 Addenda to the bayesian smoothing
(i) Let the curvature of the function $`t(x)`$ is approximately known in advance. To specify this information, introduce the norm of the difference between $`d^2t/dx^2`$ and approximately known second derivative $`d^2f/dx^2`$:
$$\mathrm{\Omega }(t(x))=\left(\frac{d^2t}{dx^2}\frac{d^2f}{dx^2}\right)^2𝑑x\omega .$$
(69)
Notice, that there is no need to know $`f(x)`$ itself, its second derivative is sufficient. The explicit presence of $`f(x)`$ in the following formulas should be taken as a consequence of the technical trick applied: at first $`f(x)`$ is subtracted from the data, then it is added to the found solution.
Everywhere in formulas (2367) make the substitutions:
$$\stackrel{~}{t}_i=t_if_i,\stackrel{~}{d}_i=d_if_i,i=1,\mathrm{},N.$$
(70)
Performing the described above procedure for smoothing, one finds $`\stackrel{~}{t}_i`$, from which by inverse transformation the sought vector is given by $`𝐭=\stackrel{~}{𝐭}+𝐟`$.
(ii) In some tasks the value on the starting (zero) node is known without measurement. This sort of prior information represents a “hard” one, that is it restricted the class of possible solutions; in the given case the solution must pass through the known zero node. The quadratic form $`\mathrm{\Omega }(𝐭)`$ (or $`\mathrm{\Omega }(\stackrel{~}{𝐭})`$ in the case of approximately known second derivative) in the expression for the prior probability has changed:
$$\mathrm{\Omega }(𝐭)=\underset{i=1}{\overset{N1}{}}[t_{i1}\mathrm{\Delta }_{i1}^1t_i(\mathrm{\Delta }_{i1}^1+\mathrm{\Delta }_i^1)+t_{i+1}\mathrm{\Delta }_i^1]^2\underset{k,l=1}{\overset{N}{}}\mathrm{\Omega }_{kl}t_kt_l+\mathrm{\Omega }_{00}t_0^2+2\mathrm{\Omega }_{01}t_0t_1+2\mathrm{\Omega }_{02}t_0t_2,$$
(71)
the first few matrix elements of $`\mathrm{\Omega }_{kl}`$ now are:
$`\mathrm{\Omega }_{00}`$ $`=`$ $`\mathrm{\Delta }_0^2\mathrm{\Delta }_1^1,\mathrm{\Omega }_{01}=(\mathrm{\Delta }_0\mathrm{\Delta }_1)^1(\mathrm{\Delta }_0^1+\mathrm{\Delta }_1^1),\mathrm{\Omega }_{02}=\mathrm{\Delta }_0^1\mathrm{\Delta }_1^2,`$ (72)
$`\mathrm{\Omega }_{11}`$ $`=`$ $`\mathrm{\Delta }_1^1(\mathrm{\Delta }_0^1+\mathrm{\Delta }_1^1)^2+\mathrm{\Delta }_1^1\mathrm{\Delta }_2^2,\mathrm{\Omega }_{12}=\mathrm{\Delta }_1^2(\mathrm{\Delta }_1^1+\mathrm{\Delta }_0^1)(\mathrm{\Delta }_1\mathrm{\Delta }_2)^1(\mathrm{\Delta }_1^1+\mathrm{\Delta }_2^1).`$ (73)
If $`t_0=0`$ (or $`\stackrel{~}{t}_0=0`$), none further changes to the formulas of smoothing (2367) are needed; at $`t_00`$ the changes are evident: instead of the scalar product $`\mathrm{𝐝𝐭}`$ in (36) will be $`(𝐝\widehat{𝐝})𝐭`$, where $`\widehat{d}_1=\alpha t_0\mathrm{\Omega }_{01}`$, $`\widehat{d}_2=\alpha t_0\mathrm{\Omega }_{02}`$, all remaining $`\widehat{d}_i=0`$; to the $`𝐝^2`$ the term $`\alpha t_0^2\mathrm{\Omega }_{00}`$ will be added.
(iii) Making some changes in the considered above problem of smoothing allows one to solve the problem of deconvolution. If the experimental value $`d_j`$ on some node $`j`$ is determined not only by $`t_j`$ but also by the values of some neighboring nodes, then instead of (18) we have:
$`d_i={\displaystyle \underset{j=1}{\overset{N}{}}}r_{ij}t_j+n_i,i=1,\mathrm{},N,`$ (74)
where $`r_{ij}`$ is the grid representation of the impulse response function. Instead of expression (21), for the likelihood now we have:
$$P(𝐝|𝐭,\sigma )=(2\pi \sigma ^2)^{N/2}\mathrm{exp}\left(\frac{1}{2\sigma ^2}\underset{k=1}{\overset{N}{}}\left[d_k\underset{i=1}{\overset{N}{}}r_{ik}t_i\right]^2\right),$$
(75)
and instead of (36), the posterior probability for $`t_j`$ is now expressed as:
$`P(t_j|𝐝,\sigma ,\alpha ){\displaystyle \mathrm{}𝑑t_{ij}\mathrm{}\sigma ^{2N}\alpha ^{N/2}\mathrm{exp}\left(\frac{1}{2\sigma ^2}\left[𝐝^22\underset{k=1}{\overset{N}{}}D_kt_k+\underset{k,l=1}{\overset{N}{}}G_{kl}t_kt_l\right]\right)},`$ (76)
where
$`G_{kl}=\alpha \mathrm{\Omega }_{kl}+{\displaystyle \underset{i=1}{\overset{N}{}}}r_{ik}r_{il},D_k={\displaystyle \underset{i=1}{\overset{N}{}}}r_{ik}d_i.`$ (77)
Further steps for finding of $`𝐭`$ are analogous to the described above.
|
warning/0003/quant-ph0003096.html
|
ar5iv
|
text
|
# Ground state cooling, quantum state engineering and study of decoherence of ions in Paul traps
## 1 Quantum information processing with trapped ions
Trapped and laser cooled ions in Paul traps are used for an implementation of quantum information processing. Internal electronic states of individual ions serve to hold the quantum information (qubits) and an excitation of common vibrational modes provides the coupling between qubits, which is necessary for quantum logic operations between qubits, more specifically, for the realization of gate operations between two ions. The Cirac-Zoller proposal requires that initially the ions are cooled to the ground state of motion and that the whole system can be coherently manipulated and controlled. The time scale of decoherence and coupling to the environment is required to be much smaller than the time scale of coherent manipulation.
The paper is organized as follows: In section II, we outline the techniques which are used for storing and detecting single ions in Paul traps. Section III decribes high resolution laser spectroscopy on the S$`{}_{1/2}{}^{}`$ D<sub>5/2</sub> transition, which is the basis of coherent manipulation of the internal electronic and external vibrational quantum state of a trapped ion. Ground state cooling for one and two ions is reported in section IV. For a single ion, we reached 99.9 $`\%`$ of motional ground state occupation. Most recently, to scale up our results from one ion towards strings of ions, we have also shown ground state cooling for two ions. After a single ion is has been cooled, it is prepared for coherent manipulation (section V) of the ion’s vibrational state, including the generation of Fock states. We measured the coherence time of this process as well as the time scale of motional heating. Our experimental finding of 1 ms coherence time, and a heating rate of only one phonon in 190 ms, shows good conditions for quantum information processing. Finally, in section VI, we investigate the maximum speed of gate operations and find that with our system approximately 30 - 50 gate operations are possible within the coherence time.
## 2 Paul traps for single Calcium ions
Ions are stored in Paul traps under ultra high vacuum ($`10^{\mathrm{10..11}}`$ mbar) conditions. In one of the experiments we use a spherical quadrupole Paul trap with a 1.4 mm ring diameter (Fig. 1.a) , where we observe motional frequencies ($`\omega _x`$, $`\omega _y`$, $`\omega _z`$)/(2$`\pi `$) of up to (2.16, 2.07, 4.51) MHz along the respective trap axes. The axial direction of the trap is denoted by $`z`$, the degeneracy of radial directions $`x`$ and $`y`$ is lifted by small asymmetries of the setup. Strings of ions are held in a linear trap (Fig. 1.b) , where we typically reach motional frequencies of ($`\omega _x`$, $`\omega _y`$, $`\omega _z`$)/(2$`\pi `$)= (2.0, 2.0, 0.7) MHz, e.g. three ions crystallize at an axial distance of 7$`\mu `$m.
<sup>40</sup>Ca<sup>+</sup> ions have a single valence electron and no hyperfine structure. All relevant transitions are accessible by solid state or diode lasers (see Fig. 1.c) . In our experiment, we apply Doppler cooling to the ion and detect the internal state on the S<sub>1/2</sub> to P<sub>1/2</sub> transition at 397 nm, excited with a frequency-doubled Ti:Sapphire laser. This transition has a natural linewidth of 20 MHz and is not closed since the ion may decay to the metastable D<sub>3/2</sub> level. A diode laser at 866 nm serves to repump the ion via the P<sub>1/2</sub> state thus closing the cooling cycle. As the upper internal level for quantum state engineering and sub-Doppler cooling, we employ the metastable D<sub>5/2</sub> level with a natural lifetime of approx. 1 s. The S$`{}_{1/2}{}^{}`$ D<sub>5/2</sub> quadrupole transition at 729 nm is excited with a Ti:Sapphire laser. We can detect whether a transition to D<sub>5/2</sub> occurred by applying the beams at 397 nm and 866 nm and monitoring the fluorescence of the ion (electron shelving technique). The internal state of the ion is discriminated with an efficiency close to 100$`\%`$ . Another diode laser at 854 nm is used to repump the ion from the D<sub>5/2</sub> level to the electronic ground state via the P<sub>3/2</sub> level. We observe the ions’ fluorescence on a photomultiplier and an intensified CCD camera.
## 3 Spectroscopy
For coherent spectroscopic investigation and state engineering on the narrow S$`{}_{1/2}{}^{}`$ D<sub>5/2</sub> transition at 729 nm we use a pulsed technique which consists of five consecutive steps.
1. Laser light at 397 nm, 866 nm, and 854 nm is used to pump the ion to the S<sub>1/2</sub> ground state. At the Doppler limit, $`E=\mathrm{}\mathrm{\Gamma }/2`$ , the thermal vibrational state corresponds to a mean vibrational quantum number $`n`$ 10 for $`\omega =(2\pi )\mathrm{\hspace{0.25em}1}`$ MHz.
2. A constant magnetic field of a few Gauss splits the 10 Zeeman components of the S$`{}_{1/2}{}^{}`$ D<sub>5/2</sub> transition in frequency space. The S$`{}_{1/2}{}^{}(m=1/2)`$ sub-state is prepared by optical pumping with $`\sigma ^{}`$ radiation at 397 nm.
3. Optional sideband cooling step (see also section 4): The S$`{}_{1/2}{}^{}(m=1/2)`$ D$`{}_{5/2}{}^{}(m=5/2)`$ transition is excited on one of the red sidebands. Therefore, the laser frequency is detuned red by one motional frequency ($`\omega _{laser}=\omega _{SD}\omega _{trap}`$). The laser power is chosen so that approximately 1 mW laser power is focused to a waist size of 30 $`\mu `$m (spherical trap), or 6 $`\mu `$m (linear trap). The laser at 854 nm is switched on to broaden the D<sub>5/2</sub> level at a power level which is set for optimum cooling. Optical pumping to the S$`{}_{1/2}{}^{}(m=+1/2)`$ level is prevented by interspersing short laser pulses of $`\sigma ^{}`$-polarized light at 397 nm. The duration of those pulses is kept at a minimum to prevent unwanted heating.
4. Spectroscopy, or alternatively, state engineering step: We excite the S$`{}_{1/2}{}^{}(m=1/2)`$ D$`{}_{5/2}{}^{}(m=5/2)`$ transition at 729 nm with a single laser pulse, or a number of pulses of well controlled frequency, power, and timing. These parameters are chosen according to the desired state manipulation.
5. Final state analysis: The ion’s fluorescence is collected under excitation with laser light at 397 nm and 866 nm and thus we detect whether a transition to the shelving level D<sub>5/2</sub> has been previously induced.
This sequence is repeated typically 100 times to measure the D<sub>5/2</sub> state population $`P_D`$ after the state engineering step. We study the dependence of $`P_D`$ on the experimental parameters, such as the detuning $`\delta \omega `$ of the light at 729 nm with respect to the ionic transition, or the length of one of the excitation pulses in step four. The duration of a single sequence is typically 20 ms, so we can synchronize the sequence with the ac line frequency at 50 Hz to reduce ac-magnetic field fluctuations.
A detailed study of the Zeeman and vibrational structure of the S$`{}_{1/2}{}^{}`$ D<sub>5/2</sub> transition for a single ion in the linear trap was performed by Nägerl et. al. . For the carrier transition (electronic excitation only, $`nn`$) we measure a 1 kHz linewidth. We attribute this residual broadening to laser and magnetic field fluctuations.
Recently, we excited Ramsey fringes using two consecutive pulses. See Fig. 2 for the Ramsey signal of the S$`{}_{1/2}{}^{}`$ D<sub>5/2</sub> carrier transition for a single ion in the spherical Paul trap. Ramsey spectroscopy allows us to investigate the coherence of superposition states $`\frac{1}{\sqrt{2}}\{|S_{1/2}+|D_{5/2}\}`$ which are excited with the first pulse. The fit to the data of Fig. 2, allows us to estimate the purely transversal decay constant to be 2 kHz. We attribute this value to laser phase fluctuations and mechanical vibrations of the trap electrodes.
In the last part of this section we will focus on spectroscopy of ion crystals: Ions in a string are strongly coupled by the Coulomb interaction. Small displacements from their equilibrium positions are described in terms of normal modes of the entire ion crystal vibrating at distinct frequencies . As an example, consider two ions confined in a Paul trap. The first normal mode corresponds to an oscillation of the entire crystal of ions moving back and forth as if they were rigidly joined. This oscillation is referred to as the center-of-mass mode (COM) . The second normal mode, the so-called breathing mode, corresponds to an oscillation where the ions move in opposite directions. The frequency of this mode is at $`\sqrt{3}\omega _{COM}`$.
In the experiment two ions were trapped in the spherical trap, and the excitation spectrum exhibits additional sideband resonances. Apart from the center-of-mass vibration frequencies, which remain at $`\omega _x`$, $`\omega _y`$, $`\omega _z`$, one observes additional sidebands which are identified in Fig. 3. The understanding of the full motional sideband structure presents the basis of further experiments, such as ground state cooling and driving coherent dynamics.
## 4 Cooling to the vibrational ground state
Preparation of the motional ground state is accomplished by a two-stage cooling process. First, the ion is pre-cooled on the S<sub>1/2</sub> \- P<sub>1/2</sub> dipole transition. In the second stage, a resolved-sideband cooling scheme, similar to the one used in Ref. , is applied on the narrow S<sub>1/2</sub> \- D<sub>5/2</sub> quadrupole transition: The laser frequency is detuned from the line center by the trap frequency, $`\omega _{laser}=\omega _{SD}\omega _{trap}`$ (first ”red sideband” excitation), thus removing one phonon with each electronic excitation process. The cooling cycle is closed by a spontaneous decay to the ground state which conserves the phonon number in the Lamb-Dicke regime. When the vibrational ground state $`|n=0`$ is reached, the ion decouples from the laser excitation. Due to the weak coupling between light and atom on a quadrupole transition one would expect long cooling times. However, the cooling rate is greatly enhanced by (i) strongly saturating the transition and (ii) shortening the lifetime of the excited state via coupling to a dipole-allowed transition.
After sideband cooling, the ground state occupation is determined by probing the sideband absorption immediately after the cooling pulse. Fig. 4 shows the excitation probability $`P_D`$ for frequencies $`\omega `$ centered around the red and blue $`\omega _z`$ sidebands. For the quantitative determination of the vibrational ground state occupation probability $`p(n=0)`$ after sideband cooling, we compare $`P_D`$ for excitation at $`\delta \omega =\omega `$ and $`\delta \omega =+\omega `$, i.e. on the red and on the blue sideband. For a thermal phonon probability distribution $`p(n)`$, the ratio of excitation on the red and blue sidebands is given by $`P_{red}/P_{blue}=n/(1+n)`$. A Lorentzian fit of the sideband heights yields a ground state occupation of $`p_0`$ = 99.9$`\%`$ for the axial mode when $`\omega _z=(2\pi )\mathrm{\hspace{0.25em}4.51}`$ MHz . By cooling the radial mode with $`\omega _y=(2\pi )\mathrm{\hspace{0.25em}2}`$ MHz, we transfer 98$`\%`$ of the population to the motional ground state. Ground state cooling is also possible at lower trap frequencies, however slightly less efficient. At trap frequencies of $`\omega _z=(2\pi )\mathrm{\hspace{0.25em}2}`$ MHz, $`\omega _y=(2\pi )\mathrm{\hspace{0.25em}0.92}`$ MHz we achieve $`n_z=0.95`$ and $`n_y=0.85`$, respectively. The x-direction is left uncooled because it is nearly perpendicular to the cooling beam. We also succeeded in simultaneously cooling all three vibrational modes by using a second cooling beam and alternating the tuning of the cooling beams between the different red sidebands repeatedly . For two ions in the spherical trap we achieved ground state cooling on the center-of-mass mode in the y and z directions, and on the breathing-mode with more than 96$`\%`$ ground state probability.
In the linear trap, our goal is to implement small scale quantum processing, thus, ions have to be kept apart so that they can be individually addressed. Thus, the laser beam at 729 nm is tightly focused into the linear Paul trap. We have demonstrated individual manipulation of single ions in strings . However, the addressing resolution is restricted by the quality of the optical system to 6 $`\mu `$m. If we demand individual addressing, this resolution restricts our ion - ion distances to greater than 6 $`\mu `$m, which corresponds to axial trap frequencies below 700 kHz (with up to three ions). This low trap frequency aggravates the difficulty of ground state cooling, since the corresponding Doppler cooling limit is close to 20 phonons. Still, we succeed in cooling the axial mode at 700 kHz and reach $`p_0`$= 90$`\%`$ ground state. We also achieved 95$`\%`$ in the ground state for cooling on the ”rocking mode” at $`\omega _{rock}=\sqrt{\omega _{radial}^2\omega _{axial}^2}`$, see Fig. 5. More advanced techniques, such as pre-cooling on the second sideband, prior to the usual first sideband cooling, will probably further increase the ground state occupation.
## 5 Coherent manipulation
Starting from the vibrational ground state, arbitrary quantum states can be created. For a demonstration of coherent state engineering, and in order to investigate decoherence, in step 4 we excite Rabi oscillations with the ion initially prepared in Fock states of its motion. Radiation at 729 nm is applied on the blue sideband transition $`|`$S$`,n_z|`$D$`,n_z+1`$ for a given interaction time $`\tau `$ and the excitation probability $`P_D`$ is measured as a function of $`\tau `$ . The Rabi flopping behaviour allows us to analyse the purity of the initial state and its decoherence . Fig. 5a shows $`P_D(\tau )`$ for the $`|n=0`$ state prepared by sideband cooling. Rabi oscillations at $`\mathrm{\Omega }_{01}=(2\pi )\mathrm{\hspace{0.25em}21}`$ kHz are observed with high contrast indicating that coherence is maintained for times above 1 ms. For the preparation of the Fock state $`|n=1`$, we start from $`|`$S$`,n=0`$, apply a $`\pi `$-pulse on the blue sideband and an optical pumping pulse at 854 nm to transfer the population from $`|`$D$`,n=1`$ to $`|`$S$`,n=1`$. As shown in Fig. 5b for the $`|n=1`$ initial state, we also observe Rabi oscillations with a high contrast, now at $`\mathrm{\Omega }_{12}=(2\pi )\mathrm{\hspace{0.25em}30}`$ kHz. The Rabi frequencies have been theoretically investigated and Blockley et. al. found $`\mathrm{\Omega }_{n,n+1}\sqrt{n+1}`$ in the Lamb-Dicke regime , which is fulfilled in good approximation for our trap. For the ratio of Rabi frequencies $`\mathrm{\Omega }_{12}/\mathrm{\Omega }_{01}`$ in the case of Fock states $`|n=0`$ and $`|n=1`$ we thus expect $`\sqrt{2}`$. The experimental finding agrees with $`\sqrt{2}`$ within 1$`\%`$. The Fourier transform of the flopping signals also yields directly the occupation probabilities for the contributing Fock states $`|n=0,1,2,3\mathrm{}`$ and allows us to calculate the purity of the prepared and manipulated states. For the ’vacuum’ state $`|n=0`$, we obtain $`p_0=0.89(1)`$ with impurites of $`p_1=0.09(1)`$ and $`p_{n2}0.02(1)`$. For the Fock state $`|n=1`$ the populations are $`p_0=0.03(1)`$, $`p_1=0.87(1)`$, $`p_2=0.08(2)`$, and $`p_{n3}0.02(1)`$. The measured transfer fidelity is about 0.9. Note that the Rabi flopping data here were taken with less efficient cooling (at a lower trap frequency), and the number state occupation from the Fourier analysis is consistent with the occupation which we determined by sideband measurements.
## 6 Speed limits of gate operations
As a figure of merit for an ion trap quantum processor, we calculate how many gate operations are feasible within the time scale given by decoherence. We have investigated the speed limit of quantum gate operations. The proposed methods employ the use of laser pulses to entangle the electronic and the vibrational degrees of freedom of trapped ions. A theoretical investigation shows that the proposed methods are limited mainly by the recoil frequency of the relevant electronic transition . We have experimentally studied the basic building block of a gate operation, that is a $`\pi `$-pulse excitation on the first blue sideband. If the blue sideband is driven significantly faster than the inverse of the recoil frequency, carrier transitions are excited off-resonance and the contrast of the observed Rabi oscillations decreases. In the case of the quadrupole S$`{}_{1/2}{}^{}`$ D<sub>5/2</sub> transition we find a time of about 16 $`\mu `$s for this limit, if we demand a gate fidelity of 99 $`\%`$. For further details on fast gates, we refer to A. Steane et. al. . We thus estimate that 30 - 50 quantum gate operations will be feasible with the current setup.
## 7 Conclusion
We have engineered the quantum states of motion $`|n=0,1`$ of a single trapped ion that are relevant for quantum computation, using laser excitation on a forbidden optical transition. We have observed more than 30 periods of Rabi oscillations on the motional sidebands of this transition, thus showing that decoherence is negligible on the time scale of a few oscillations, i.e. the time required for a quantum gate operation. We attribute the observed 1 ms decoherence time to laser and magnetic field fluctuations. Heating of the motional degrees of freedom has also been observed and was measured directly to happen at least 1 order of magnitude more slowly. This confirms that in the comparatively large traps which we use, heating seems not to be a limiting process.
Furthermore, we have been able to cool two ions to the ground state, and also to reach the ground state for single and two ions in the linear trap, under conditions which allow individual addressing of ions in a string. Multiple coherent gate operations with trapped ions seem well within experimental reach.
This work is supported by the Austrian ”Fonds zur Förderung der wissenschaftlichen Forschung” within the project SFB15, and in parts by the European Commission within the TMR networks ”Quantum Information” (ERB-FMRX-CT96-0087) and ”Quantum Structures” (ERB-FMRX-CT96-0077), and the ”Institut für Quanteninformation GmbH”.
|
warning/0003/hep-ph0003149.html
|
ar5iv
|
text
|
# DO–TH 00/02 hep-ph/0003149 Neutrino oscillation experiments and limits on lepton–number and lepton–flavor violating processes
## 1 Introduction
In recent years overwhelming evidence for non–vanishing neutrino masses was collected in atmospheric , solar and accelerator experiments, opening a window into a variety of new phenomena . The results are interpreted in terms of neutrino oscillations governed by a mixing angle and a mass–squared difference. Analyzing the data of the respective experiments yields typically values of
$$\begin{array}{ccc}(\mathrm{\Delta }m^2(\mathrm{eV}^2),\mathrm{sin}^22\theta )& \{\begin{array}{cc}(10^3,>0.7)& \text{ atmospheric}\\ \begin{array}{cc}(10^5,10^3)& \text{ SAMSW}\\ (10^5,>0.7)& \text{ LAMSW}\\ (10^{10},>0.7)& \text{ VO}\end{array}\}& \text{ solar}\\ (1,10^3)& \text{ LSND.}\end{array}& \end{array}$$
(1)
Here SAMSW (LAMSW) denotes the small (large) angle MSW solution and VO the vacuum oscillation solution of the solar neutrino problem. One sees that the atmospheric and solar mass scale obey the following relation:
$$\mathrm{\Delta }m_{}^2\mathrm{\Delta }m_\mathrm{A}^2,$$
(2)
regardless of the solar solution chosen. In order to avoid sterile neutrinos one usually leaves out the LSND result (which also gives time to wait until the conflict with KARMEN is resolved). We are thus working in a three neutrino framework, which allows a simple derivation of the leptonic Maki–Nakagawa–Sakata (MNS) matrix . Only five mass schemes can accommodate the results in a three neutrino framework, three hierarchical and two degenerate schemes with the overall scale given by cosmological arguments to be of the order of a few eV. These mass schemes can be divided into two scenarios. In addition, we distinguish the cases in which one element of the mixing element is zero as indicated by CHOOZ data and the general case of all elements being non–zero. Ordering effects of some elements are also considered.
From the MNS matrix one can infer limits on the $`3\times 3`$ matrix of effective Majorana neutrino masses defined as
$$\begin{array}{c}m_{\alpha \beta }=|(U\mathrm{diag}(m_1,m_2,m_3)U^\mathrm{T})_{\alpha \beta }|\\ =\left|m_iU_{\alpha i}U_{\beta i}\right|m_i|U_{\alpha i}U_{\beta i}|\text{ with }\alpha ,\beta =e,\mu ,\tau ,\end{array}$$
(3)
where $`U`$ is the mixing matrix and the $`m_i`$ are mass eigenvalues. With this approximation, $`m_{\alpha \beta }`$ is symmetrical. The mass measured in neutrinoless double beta decay ($`0\nu \beta \beta `$) is the $`(ee)`$ element of this matrix and has been considered by several authors . However, the complete matrix is rarely analyzed in terms of phenomenological consequences, except for the elements $`m_{e\mu }`$ and $`m_{\mu \mu }`$ in Refs. , although without giving concrete numbers. We introduce processes dependent on every element of $`m_{\alpha \beta }`$, like $`\mu ^{}e^+`$ conversion, $`K^+\pi ^{}\mu ^+\mu ^+`$ or recently proposed high–energy scattering processes, finding rather discouraging results for branching ratios or life times. We compare our results (calling them indirect limits) with current experimental (direct) bounds and find that our limits are up to 14 orders of magnitude more stringent than current experimental data and therefore beyond experimental access in the near future. Special attention is paid to 0$`\nu \beta \beta `$ and its sensitivity on mass and mixing. Though it will be much easier to wait for further astrophysical data to distinguish between vacuum and MSW solutions for the solar neutrino problem, it would be a remarkable experiment to decide via terrestrial nuclear physics experiments if matter effects inside the Sun are of importance or not. Unfortunately, it turns out that the only scheme delivering values inside the range of next generation experiments is insensitive on the solar solutions.
The paper is organized as follows: In Section 2 we present the usual assumptions that lead to the derivation of the MNS matrix and $`m_{\alpha \beta }`$. Section 3 gives the results and pays special attention to the $`(ee)`$ element of $`m_{\alpha \beta }`$. In Section 4 processes are introduced which are sensitive on the respective elements of $`m_{\alpha \beta }`$ just as 0$`\nu \beta \beta `$ is on $`m_{ee}`$. Section 5 concludes the paper.
## 2 Oscillation probabilities and mass schemes
The phenomenology of neutrino oscillations is well reviewed in the literature, see e.g. . Flavor eigenstates $`\nu _\alpha `$ ($`\alpha =e,\mu ,\tau `$) are connected to mass eigenstates $`\nu _i`$ ($`i=1,2,3`$) via an unitary matrix, i.e. $`\nu _\alpha =U_{\alpha i}\nu _i`$. For Majorana neutrinos this matrix can be parametrized as
$$U_{\alpha i}=\left(\begin{array}{ccc}c_1c_3& s_1c_3e^{i\lambda _1}& s_3e^{i\delta }\\ (s_1c_2c_1s_2s_3e^{i\delta })e^{i\lambda _1}& c_1c_2s_1s_2s_3e^{i\delta }& s_2c_3e^{i(\lambda _2\lambda _1)}\\ (s_1s_2c_1c_2s_3e^{i\delta })e^{i\lambda _2}& (c_1s_2s_1c_2s_3e^{i\delta })e^{i(\lambda _2\lambda _1)}& c_2c_3\end{array}\right)$$
(4)
with $`c_i=\mathrm{cos}\theta _i`$ and $`s_i=\mathrm{sin}\theta _i`$. Three CP–violating phases are present. For neutrino oscillations, only one phase ($`\delta `$) contributes . Effects of all phases are discussed e. g. in , for our estimations we shall skip them, since we are not interested in CP violation. In addition, the probabilities we use depend only on the absolute values of the mixing matrix elements. The probability of a flavor state $`\alpha `$ to oscillate into a state $`\beta `$ is given by
$$P_{\alpha \beta }=\delta _{\alpha \beta }2\text{ Re }\underset{j>i}{}U_{\alpha i}U_{\alpha j}^{}U_{\beta i}^{}U_{\beta j}(1\mathrm{exp}i\mathrm{\Delta }_{ji}).$$
(5)
Here
$$\mathrm{\Delta }_{ji}=\frac{L}{2E}\mathrm{\Delta }m_{ji}^2=2.54\frac{L/\mathrm{km}}{E/\mathrm{GeV}}\mathrm{\Delta }m_{ji}^2/\mathrm{eV}^2\text{ with }\mathrm{\Delta }m_{ji}^2=m_j^2m_i^2.$$
Without loss of generality we assume
$$m_3>m_2>m_1>0.$$
(6)
With the above relation there are two possibilities to accommodate $`\mathrm{\Delta }m_{}^2\mathrm{\Delta }m_\mathrm{A}^2`$:
$$\begin{array}{cc}\mathrm{𝐒𝐜𝐞𝐧𝐚𝐫𝐢𝐨}𝐀:& \mathrm{\Delta }m_{21}^2=\mathrm{\Delta }m_{}^2\mathrm{\Delta }m_\mathrm{A}^2=\mathrm{\Delta }m_{31}^2\mathrm{\Delta }m_{32}^2\\ \mathrm{𝐒𝐜𝐞𝐧𝐚𝐫𝐢𝐨}𝐁:& \mathrm{\Delta }m_{32}^2=\mathrm{\Delta }m_{}^2\mathrm{\Delta }m_\mathrm{A}^2=\mathrm{\Delta }m_{21}^2\mathrm{\Delta }m_{31}^2\end{array}$$
(7)
Three mass schemes are capable of providing scenario A, qualitatively shown in Fig. 1:
$$\begin{array}{ccc}\text{Scheme A I:}& m_3m_2m_1:& m_3\sqrt{\mathrm{\Delta }m_{31}^2},m_2\sqrt{\mathrm{\Delta }m_{21}^2}\\ \text{Scheme A II:}& m_3m_2m_1:& m_3\sqrt{\mathrm{\Delta }m_{31}^2},m_2m_10\\ \text{Scheme A III:}& m_3m_2m_1m_0:& 3m_05\mathrm{eV},\end{array}$$
(8)
where $`m_0`$ comes from cosmological considerations .
For scenario B, however, there are only two possibilities as presented in Fig. 2:
$$\begin{array}{ccc}\text{Scheme B I:}& m_3m_2m_1:& m_3m_2\sqrt{\mathrm{\Delta }m_{31}^2},m_10\\ \text{Scheme B II:}& m_3m_2m_1m_0:& 3m_05\mathrm{eV}.\end{array}$$
(9)
We stress that there are no other possibilities when Eqs. (2) and (6) are used. Scenario A I can be obtained from the see–saw mechanism ; A II and B I, i.e. two very close masses and one separated by the others can be a result of mechanisms generating neutrino masses radiatively .
Regardless of the concrete scheme, relations (2) and (4) allow to get the absolute values of the elements of the MNS matrix : In scenario A one finds for a short baseline reactor experiment such as CHOOZ:
$$P_{ee}^{\mathrm{CHOOZ}}=14|U_{e3}|^2(1|U_{e3}|^2)\mathrm{sin}^2\mathrm{\Delta }_{31}/2$$
(10)
Due to the negative results CHOOZ presents , one comes to the conclusion that $`|U_{e3}|^2`$ is either very small or close to 1. Taking into account that the probability $`P_{ee}^{}`$ for solar neutrinos is significantly lower than 1, leads to a small value of $`|U_{e3}|^2`$ . In that case one has
$$P_{ee}^{}=14|U_{e1}|^2|U_{e2}|^2\mathrm{sin}^2\mathrm{\Delta }_{21}/2.$$
(11)
For oscillations of atmospheric $`\nu _\mu `$’s into $`\nu _\tau `$’s one finds
$$P_{\mu \tau }^\mathrm{A}=4|U_{\mu 3}|^2|U_{\tau 3}|^2\mathrm{sin}^2\mathrm{\Delta }_{31}/2.$$
(12)
In obtaining this equation we have assumed that the oscillation triggered by $`\mathrm{\Delta }m_{21}^2`$ washes out. This is the case for $`L(\mathrm{km})/E(\mathrm{GeV})10^5`$ so that it is advisable to use the through–going muons data set of SuperKamiokande (SK). Equations (11) and (12) together with the unitarity of the MNS matrix are now used to get the absolute values of all elements . As the mass and mixing parameters we use the best fit points from for solar neutrinos and for the SK through–going muon sample the values from . The numbers used are the following:
$$\begin{array}{cc}(\mathrm{\Delta }m_{21}^2(\mathrm{eV}^2),\mathrm{\hspace{0.33em}4}|U_{e1}|^2|U_{e2}|^2)=& \{\begin{array}{cc}(5.410^6,6.010^3)& \text{ SAMSW}\\ (1.810^5,0.76)& \text{ LAMSW}\\ (8.010^{11},0.75)& \text{ VO}\end{array}\\ (\mathrm{\Delta }m_{31}^2(\mathrm{eV}^2),\mathrm{\hspace{0.33em}4}|U_{\mu 3}|^2|U_{\tau 3}|^2)=& (1.010^2,\mathrm{\hspace{0.33em}0.78})\end{array}$$
(13)
Note the unusual high value for the atmospheric mass scale. We give now the resulting mixing and mass matrices, starting with scenario A and commenting on scenario B later.
### 2.1 The case $`|U_{e3}|0`$
The value $`\mathrm{\Delta }m_{31}^2=1.010^2\mathrm{eV}^2`$ corresponds in CHOOZ’s exclusion plot of to $`|U_{e3}|^2<\mathrm{\hspace{0.33em}0.04}`$. Translating this in Eq. (4) leads to $`|s_3|<\mathrm{\hspace{0.33em}0.19}`$ and $`|c_3|>\mathrm{\hspace{0.33em}0.98}1`$. This leads to the following mixing matrices:
$$|U_{\alpha i}|<\{\begin{array}{cc}\left(\begin{array}{ccc}0.999& 0.039& 0.189\\ 0.131& 0.860& 0.515\\ 0.182& 0.521& 0.857\end{array}\right)& \text{ SAMSW}\\ \left(\begin{array}{ccc}0.863& 0.505& 0.189\\ 0.517& 0.789& 0.515\\ 0.400& 0.527& 0.857\end{array}\right)& \text{ LAMSW}\\ \left(\begin{array}{ccc}0.866& 0.500& 0.189\\ 0.513& 0.791& 0.515\\ 0.398& 0.527& 0.857\end{array}\right)& \text{ VO.}\end{array}$$
(14)
Here we have used that $`|U_{e1}||U_{e2}|`$ as it is necessary for the MSW solutions and — inspired by the CKM matrix — $`|U_{\tau 3}||U_{\mu 3}|`$. The possibility $`|U_{e2}||U_{e1}|`$ is equivalent to an exchange of the first and second column, as $`|U_{\mu 3}||U_{\tau 3}|`$ is to an exchange of the second and third row.
### 2.2 The case $`|U_{e3}|0`$
The observed smallness of $`m_{ee}`$ as measured in 0$`\nu \beta \beta `$ and the absence of $`\nu _e`$ mixing in atmospheric oscillations has lead many authors (see e. g. ) to the assumption $`|U_{e3}|0`$. In addition, it can also be related (together with bi–maximal mixing) with the observed flatness in $`L/E`$ of SK‘s $`e`$–like events . Since also the phases $`\lambda _{1,2}`$ in Eq. (4) do not contribute to oscillations, this means that there is no observable CP violation in any oscillation experiment. The mixing matrix now reads:
$$|U_{\alpha i}|<\{\begin{array}{cc}\left(\begin{array}{ccc}0.999& 0.039& 0\\ 0.033& 0.856& 0.515\\ 0.020& 0.515& 0.857\end{array}\right)& \text{ SAMSW}\\ \left(\begin{array}{ccc}0.863& 0.505& 0\\ 0.433& 0.740& 0.515\\ 0.260& 0.445& 0.857\end{array}\right)& \text{ LAMSW}\\ \left(\begin{array}{ccc}0.866& 0.500& 0\\ 0.429& 0.742& 0.515\\ 0.258& 0.446& 0.857\end{array}\right)& \text{ VO.}\end{array}$$
(15)
The matrices are of course very similar to the ones derived in where the method was first presented.
### 2.3 Scenario B
Scenario B can be easily obtained from scenario A via cyclic permutation of the columns of the mixing matrices. Hence, the cases to distinguish are $`|U_{e1}|^20`$, $`|U_{e1}|^20`$, $`|U_{e2}|^2()|U_{e3}|^2`$ (the “$``$” case only for the vacuum solution) and $`|U_{\mu 1}|^2()|U_{\tau 1}|^2`$. With the two scenarios A and B and the possibilities for ordering the mixing matrix elements we have a total of 80 different mass matrices, 48 for scenario A and 32 for scenario B. From these 80 matrices, 48 are stemming from nondegenerate mass schemes. We will show that the 80 reduces to 57.
## 3 Results for $`m_{\alpha \beta }`$
As shown, the different possibilities are equivalent to exchanges of rows or columns of the mixing matrices. The same holds for the resulting mass matrices. For example, the difference of the $`|U_{\tau 3}||U_{\mu 3}|`$ case and the $`|U_{\tau 3}||U_{\mu 3}|`$ case translates into an exchange of $`m_{\alpha \tau }`$ with $`m_{\alpha \mu }`$ with $`\alpha =e,\mu `$. Replacing $`|U_{e1}||U_{e2}|`$ with $`|U_{e1}||U_{e2}|`$ has no effect on $`m_{\alpha \beta }`$ as long as $`|U_{e3}|0`$ and generally in the degenerate schemes and in A II and B I. This, with the appropriate permutations mentioned above, holds for scenario B as well. Evidently, the degenerate schemes give the same result for $`m_{\alpha \beta }`$ in both scenarios. Therefore from all 80 possible mass matrices only 57 survive, 14 degenerate and 43 nondegenerate ones. We give now for the nondegenerate cases and for all three solar solutions the bounds of our results:
$$m_{\alpha \beta }<\{\begin{array}{cc}\left(\begin{array}{ccc}3.510^6\mathrm{}0.1& 4.610^5\mathrm{}2.010^2& 4.610^5\mathrm{}2.010^2\\ & 2.710^2\mathrm{}7.610^2& 6.610^3\mathrm{}4.710^2\\ & & 2.710^2\mathrm{}7.610^2\end{array}\right)\mathrm{eV}& \text{ SAMSW }\\ \left(\begin{array}{ccc}1.110^3\mathrm{}0.1& 9.510^4\mathrm{}8.510^2& 9.510^4\mathrm{}8.510^2\\ & 2.710^2\mathrm{}8.910^2& 4.410^2\mathrm{}6.210^2\\ & & 2.710^2\mathrm{}8.910^2\end{array}\right)\mathrm{eV}& \text{ LAMSW }\\ \left(\begin{array}{ccc}2.210^6\mathrm{}0.1& 2.010^6\mathrm{}8.410^2& 2.010^6\mathrm{}8.410^2\\ & 2.710^2\mathrm{}8.910^2& 4.410^2\mathrm{}6.210^2\\ & & 2.710^2\mathrm{}8.910^2\end{array}\right)\mathrm{eV}& \text{ VO. }\end{array}$$
(16)
In the scheme (A II, $`|U_{e3}|0`$) there are zeros as solutions for $`m_{e\alpha }`$, which means values much smaller than the atmospheric scheme, i.e. possibly in the range of $`10^3\mathrm{}10^4`$ eV, e.g.
$$m_{ee}|U_{e1}|^2m_1+|U_{e2}|^2m_2+|U_{e3}|^2m_3=m_1\sqrt{\mathrm{\Delta }m_\mathrm{A}^2}.$$
(17)
The same can happen in scheme A I, where always a contribution of a value much smaller than the solar scheme can be present. These cases reflect the fact that the smallest mass eigenvalue is never known. For the degenerate scheme the solutions for VO and LAMSW are almost identical:
$$m_{\alpha \beta }<\{\begin{array}{cc}\left(\begin{array}{ccc}1.67\mathrm{}1.73& 0.74\mathrm{}1.57& 0.74\mathrm{}1.57\\ & 1.67\mathrm{}1.93& 1.47\mathrm{}1.77\\ & & 1.67\mathrm{}1.95\end{array}\right)\mathrm{eV}& \text{LAMSW and VO}\\ \left(\begin{array}{ccc}1.67\mathrm{}1.73& 0.07\mathrm{}0.61& 0.07\mathrm{}0.61\\ & 1.67\mathrm{}1.73& 1.47\mathrm{}1.52\\ & & 1.67\mathrm{}1.73\end{array}\right)\mathrm{eV}& \text{SAMSW.}\end{array}$$
(18)
The values $`>m_0`$ are explained by the violation of unitarity of the mixing matrices.
### 3.1 Properties of the mass matrices
We start with scenario A: In general, for the hierarchical schemes, $`m_{\mu \mu }`$ is the biggest, $`m_{ee}`$ the smallest entry in $`m_{\alpha \beta }`$. The difference can be up to 4 orders of magnitude (VO and SAMSW, scheme A I, $`|U_{e3}|^20`$). Entries in the electron row of $`m_{\alpha \beta }`$ are smaller than the other elements. The degenerate scheme has always $`m_{ee}`$ $``$ $`m_{\mu \mu }`$ $`>`$ $`m_{e\mu }`$. In addition, the $`|U_{e3}|^20`$ case delivers higher values for all elements of $`m_{\alpha \beta }`$. Scheme III gives higher numbers than scheme I which in turn gives higher values than scheme II.
In scenario B all entries are usually in the same order of magnitude yet somewhat higher than in scenario A. The element $`m_{ee}`$, which is the most natural candidate for experimental access, is the only one which is significantly higher, at least one order of magnitude. It is always bounded by $`\sqrt{\mathrm{\Delta }m_\mathrm{A}^2}`$, see the next section. As said before, the degenerate scheme gives the same numbers in both scenarios. Two typical matrices show most of the mentioned points, the first for scheme (A I, LAMSW, $`|U_{e1}||U_{e2}|`$, $`|U_{\mu 3}||U_{\tau 3}|`$ and $`|U_{e3}|=0`$) while the second is for (B II, VO, $`|U_{e2}||U_{e3}|`$, $`|U_{\mu 1}||U_{\tau 1}|`$ and $`|U_{e3}|0`$):
$$\begin{array}{cc}m_{\alpha \beta }<\left(\begin{array}{ccc}3.510^6& 4.610^5& 7.710^5\\ & 7.410^2& 4.510^2\\ & & 2.810^2\end{array}\right)\mathrm{eV}& \\ m_{\alpha \beta }<\left(\begin{array}{ccc}1.73& 1.28& 1.56\\ & 1.95& 1.77\\ & & 1.92\end{array}\right)\mathrm{eV}.& \end{array}$$
(19)
### 3.2 Electron neutrino mass
From inspection of Eq. (16) one sees that the only element accessible to present or near future experiments, $`m_{ee}`$, has the broadest spectrum, a lucky coincidence. On the other hand, due to $`|U_{e3}|^21`$, $`m_{ee}`$ is always the smallest entry for scenario A. From all 57 matrices 33 different possibilities for $`m_{ee}`$ exist (it does not matter if $`|U_{\mu 3}||U_{\tau 3}|`$ or vice versa). From these 33 values $`m_{ee}`$ takes only 10 different values, which are worth taking a closer look at since they spread 6 orders of magnitude. Leaving the ones obtained from degenerated schemes aside, since they lie already above the current experimental limit, we have 8 different values out of 25 matrices, spanning 5 orders of magnitude. It is now tempting to assume that these 8 values have potential to distinguish between the different solar solutions. This is unfortunately not the case:
As a result of relation (2) many schemes have the same value for $`m_{ee}`$, for example (A I, SAMSW, $`|U_{e1}||U_{e2}|`$ and $`|U_{e3}|0`$), (A II, all solar solutions, $`|U_{e1}||U_{e2}|`$ and $`|U_{e3}|0`$) or (A I, VO, $`|U_{e2}||U_{e1}|`$ and $`|U_{e3}|0`$), all yielding $`3.610^3`$ eV.
In addition, for scenario B a peculiarity occurs, namely the bound always takes the same value, regardless of the solar solution:
$$m_{ee}_\mathrm{B}|U_{e1}|^2m_1+|U_{e2}|^2m_2+|U_{e3}|^2m_3m_3\sqrt{\mathrm{\Delta }m_\mathrm{A}^2}.$$
(20)
In the nondegenerate schemes of scenario B, this value of $`m_{ee}`$ is always the largest entry in $`m_{\alpha \beta }`$. Hence a positive signal in a neutrinoless double beta decay experiment will not be able to distinguish between different solar solutions, if nature has chosen scenario B for its neutrinos and an analysis of this kind is used. In addition, when scenario A is realized it will be extremely challenging to distinguish the precise form of the mass and mixing scheme as well as to tell which solar solution is the right one. What also complicates the analysis is that, as mentioned above, the contribution of the neglected $`m_1`$ to $`m_{ee}`$ can very well be in the order of $`10^3`$ eV, making a definite statement somewhat difficult.
Nevertheless, scenario B is only a factor 2 from current experimental limits away, lying well within access in updates of the <sup>76</sup>Ge experiment. Although we used a somewhat high value for $`\mathrm{\Delta }m_\mathrm{A}^2`$ this specific situation seems to be the only realistic candidate for experimental detection. In the next section we will discuss the possibilities of detecting a process sensitive on $`m_{\alpha \beta }`$ in more detail.
## 4 Lepton–number and –flavor violating processes and Majorana neutrinos
For this section it is important to stress again the difference between direct bounds, i.e. considering processes that depend on the respective matrix element and indirect bounds obtained in the present paper, i.e. using oscillation data and unitarity of the MNS matrix.
As not surprising, neutrinoless double beta decay is the best examined process triggered by Majorana neutrinos, resulting in a limit of $`m_{ee}0.2`$ eV . The electron–muon element $`m_{e\mu }`$ can be inferred from muon–positron conversion in sulfur nuclei. Theoretical estimations from together with the PDG limit of the branching ratio give a limit of $`m_{e\mu }0.4(1.9)`$ GeV, when the final state proton pairs are in spin singlet (triplet) state, respectively. The very same diagram as for 0$`\nu \beta \beta `$ can be applied to other processes like $`K^+\pi ^{}\mu ^+\mu ^+`$, which has an experimental branching ratio limit of $`1.510^4`$ . Taking the calculation of , one finds a limit of $`m_{\mu \mu }1.110^5`$ GeV. Another process depending on $`m_{\mu \mu }`$ is $`\mu ^{}\mu ^+`$ conversion in Titanium, discussed in . Instead of nuclear captions or rare decays it was shown in , that in principle one can use high–energy scattering processes — in this case tri–muon production at fixed target neutrino–nucleon experiments — to get a bound on $`m_{\mu \mu }`$. Without worrying too much about experimental cuts a limit of $`m_{\mu \mu }<\mathrm{\hspace{0.17em}10}^4`$ GeV was obtained. In this procedure was applied to existing HERA data and generalized to the process
$$e^+p\overline{\nu _e}\alpha ^+\beta ^+X\text{ with }(\alpha \beta )=(e\tau ),(\mu \tau ),(\mu \mu )\text{ and }(\tau \tau ),$$
(21)
giving for the first time direct limits on the tau–sector of the mass matrix. Another direct way to obtain information about the tau sector of $`m_{\alpha \beta }`$ might be $`B^+X^{}\tau ^+\alpha ^+`$ with $`\alpha =e,\mu \text{ or }\tau `$ and $`X=\pi ,K,D,\mathrm{}`$. In total, the current situation for bounds deduced from processes directly depending on $`m_{\alpha \beta }`$ reads:
$$m_{\alpha \beta }<\left(\begin{array}{ccc}210^{10}& 0.4(1.9)& 4.210^3\\ & 4.010^3& 4.410^3\\ & & 2.010^4\end{array}\right)\mathrm{GeV}.$$
(22)
A spread over 14 orders of magnitude can be seen. An improvement of the values is surely advisable. The somewhat unusual way to use high–energy scattering as done in is highly compatible with decay analyses: For example, assuming that a branching ratio for $`K^+\pi ^{}\mu ^+\mu ^+`$ of about $`9.210^8`$ (BR for $`K^+\pi ^+\mu ^+\mu ^{}`$, ) can be achieved would result in a limit of $`m_{\mu \mu }<\mathrm{\hspace{0.33em}3.5}10^3`$ GeV, almost the same number as from HERA data, which itself will be improved by luminosity and energy updates.
There are other Majorana induced $`\mathrm{\Delta }L0`$ processes, which are at present however not experimentally accessible: Running an $`e^+e^{}`$ collider in $`e^{}e^{}`$ mode could give rise to the “inverse neutrinoless double beta decay” $`e^{}e^{}W^{}W^{}`$ . The same could be done for a $`\mu \mu `$ collider or even a possible $`e\mu `$ machine. For the case we are interested in ($`sm_i^2`$) the cross section reads
$$\sigma (\alpha ^{}\beta ^{}W^{}W^{})\frac{G_F^2}{4\pi }m_{\alpha \beta }{}_{}{}^{2}4.3010^{17}\left(\frac{m_{\alpha \beta }}{\mathrm{eV}}\right)^2\text{ fb},$$
(23)
leaving no prospects for detection, since the cross section is in the order of $`10^{20}`$ ($`10^{16}`$) fb for hierarchical (degenerate) scheme.
For the sake of completeness one has to add a few words on cancellation of terms in $`m_{\alpha \beta }`$. From Eqs. (3) and (6) it is clear that our bounds are insensitive on the phases of the mixing elements. See for more details on what one could learn from the different phase dependence of $`m_{ee}`$, $`m_{e\mu }`$ and $`m_{\mu \mu }`$. The Majorana nature brings additional complications via the intrinsic CP–parities $`\eta _i^{\mathrm{CP}}`$ of the mass eigenvalues. Even for CP invariance one can write
$$m_{\alpha \alpha }=\left|\left|U_{\alpha i}\right|^2m_i\eta _i^{\mathrm{CP}}\right|.$$
(24)
making destructive interference in the respective amplitudes possible. Then one can assume mass matrices in flavor space which prohibite special entries in $`m_{\alpha \beta }`$, e.g.
$$m_{\alpha \beta }=\left(\begin{array}{cc}0& m\\ m& 0\end{array}\right),m0.$$
(25)
with eigenvalues $`\pm m`$, first introduced to conserve $`L_eL_\mu `$. Here, the requirement $`m_i0`$ can be saved by making the mixing matrix elements complex, leading again to cancellation. However, as long as no evidence for a nonvanishing element of $`m_{\alpha \beta }`$ is found, we have no chance to decide which of the above possibilities is realized by nature. Yet, if in direct mass searches for, say, the $`\nu _e`$ a result of a few eV is found, or the degenerate scheme is somehow verified, one could give bounds on the phases and thus restrict different models.
## 5 Conclusions
Tables 1 (hierarchical scheme) and 2 (degenerate) summarize our results for the different elements of the mass matrices, together with their maximal values obtained from our estimations, the process sensitive to the respective element, a ratio with respect to the relevant standard model process (see below) and a number which indicates how far away we are from detecting the process and thus having access to the element. Also given is the ratio of our indirect bound with the previous direct limits from Eq. (22).
As the ratio we use for the $`K`$ decay the branching ratio and for the HERA processes (21) the quotient of the respective cross section with the usual standard model charged current process of $`\sigma (e^+p\overline{\nu _e}X,Q^2>200\mathrm{GeV}^2)30.3`$ pb . For $`m_{ee}`$ we give the half–life obtained with the matrix elements from for $`{}_{}{}^{76}\mathrm{Ge}`$.
Theory/Data is a measure for how close (better: how far away) we are from detection of the respective process. We use for $`m_{ee}`$ the current experimental limit for $`T_{1/2}^{0\nu }(^{76}\mathrm{Ge})`$ divided by our bound, for the $`K`$ decay our result divided by the current measured BR limit and for the HERA processes the cross section times the mean value of the luminosity analyzed by H1 and ZEUS in searches for isolated lepton events, $`_{e^+}=42.1`$ pb<sup>-1</sup>. Note the almost hierarchical structure from $`ee`$ to $`\tau \tau `$ processes. Theory/Data is a number which characterizes how difficult it is to investigate the respective effective mass. A value less than $`10^3`$ to $`10^4`$ for a given $`m_{\alpha \beta }`$ cannot be regarded as accessible in laboratory experiments, even with very positive upgrade assumptions. The numbers show that no element other than $`m_{ee}`$ provides a realistic chance of accession. Regarding scenario A, the highest value obtained for $`m_{ee}`$ is $`4.710^3`$ eV, about the limit achievable in the most positive assumption of the GENIUS sensitivity of $`210^3`$ eV at 68 $`\%`$ C. L. for a 10 year run with 10 tons of enriched germanium .
Note however that we used the best–fit points of typical oscillation experiment analyses, which of course do not need to be the final answer. However, our value $`\mathrm{\Delta }m_\mathrm{A}^2`$ = 0.01 eV<sup>2</sup>, the best–fit point for through–going muons at SK, is just the maximum of typical general analyses (see for details) and therefore the values can be regarded as a realistic indication. In scenario B $`m_{ee}`$ is bounded by the atmospheric mass scale, thus lying well within the range of next generation neutrinoless double beta decay experiments. A detection of a $`m_{ee}`$ in the range of the atmospheric scale would rule out the hierarchical schemes in scenario A.
Direct/indirect shows the ratio of the indirect limits obtained here and the direct limits from Eq. (22). This number can be as high as $`10^{14}`$. The indirect bounds obtained in this paper are more stringent by this number.
The tables show that the clarification of the question whether neutrinos are Dirac or Majorana particles might have to be postponed. Present and foreseeable experimental possibilities are far beyond verifying some of the given branching ratios or cross sections. This is shown by the ratios in the Theory/Data column. In conclusion, we used a three neutrino framework and studied the range of the elements of the effective mass matrix using all possibilities for the mass and mixing schemes. It turned out that $`m_{ee}`$ has the broadest range of all elements but also is in scenario A the smallest entry and in the nondegenerate schemes of scenario B the highest entry. in general is the smallest value of $`m_{\alpha \beta }`$. We may summarize the situation in saying that if scenario A is realized little hope should one have, whereas scenario B provides a realistic chance of being probed, leaving neutrinoless double beta decay, the “gold–plated process” of lepton–number/flavor violation. The nice possibility of deciding which solution to the solar neutrino problem is realized is unfortunately not given, but this question will be clarified anyway in ongoing and forthcoming oscillation experiments. For a degenerate scheme the bounds are higher than current experimental limits so that a cancellation of terms in $`m_{ee}`$ as discussed at the end of Section 4 has to occur, provided nature has chosen this scheme. Regarding the other elements of $`m_{\alpha \beta }`$ we showed that there is no possibility to investigate processes depending on them. Nevertheless, the improvement with respect to the previous direct bounds is up to 14 orders of magnitude.
Note added: When this paper was finished, Ref. appeared which gives new limits on two elements of $`m_{\alpha \beta }`$.
Acknowledgments
I thank S. M. Bilenky, M. Flanz, E. A. Paschos and K. Zuber for helpful discussions. This work has been supported in part by the “Bundesministerium für Bildung, Wissenschaft, Forschung und Technologie”, Bonn under contract number 05HT9PEA5. Financial support from the Graduate College “Erzeugung und Zerf$`\ddot{\mathrm{a}}`$lle von Elementarteilchen” at Dortmund university is gratefully acknowledged.
|
warning/0003/hep-ph0003114.html
|
ar5iv
|
text
|
# Gluelump spectrum in the QCD string model
## 1 Introduction
Gluelumps are not physical objects and their spectrum cannot be measured on experiment. However, as will be discussed below, they play more fundamental role than any other hadrons since gluelump masses define field correlators in the QCD vacuum and in particular string tension. The purpose of the present study is to calculate analytically the gluelamp spectrum in the framework of the so-called QCD string model (QCDSM). This model was developed in the earlier papers for spinless quarks, and was augmented by the spin-dependent terms treated as a perturbation . The resulting Hamiltonian of the rotating string with quarks or gluons at the ends was recently systematically applied to mesons, hybrids and glueballs (see review in and recent developments in ). Three features are characteristic for the model. First, QCDSM is directly derived from QCD with few assumptions supported by the lattice data, i.e. the minimal area law for the Wilson loop, and the dominant role of valence quarks and gluons. Second, it is fully relativistic and the Hamiltonian can be obtained in the c.m. system , or in the light–cone system or else on any other hypersurface. Third, the model contains the minimal number of parameters: fundamental string tension $`\sigma _f=0.18`$ GeV<sup>2</sup> defined by the meson Regge slope, the strong coupling constant $`\alpha _s`$ (taken as constant in first approximation, $`\alpha _s(\mu )`$, where $`\mu `$ is an inverse size of the system, or with the freezing behaviour $`\alpha _s(r)`$ in more accurate calculations), and finally the quark (antiquark) selfenergy $`C_0=0.25`$ GeV, which should be subtracted for each quark from the mass of a meson, hybrid, or baryon. Note, that for gluons this term is absent because of gauge invariance and therefore spectrum of glueballs and gluelumps is defined only by $`\sigma _f`$ and $`\alpha _s`$ .
It is remarkable, that with these fixed universal parameters one obtains spectrum of all hadronic systems in good agreement with lattice data and experiment . In particular, the glueball spectrum depending only on $`\sigma _f`$ (with spin splittings depending also on $`\alpha _s`$) comes out in remarkable agreement with recent lattice data .
At the same time the heavy-light mesons have been calculated in this method in including quark decay constants $`f_M`$, $`M=B,B_s,D,D_s`$ again in good agreement with experiment and other approaches.
This gives a hope that the QCDSM can be used also for the system containing a valence gluon connected by an adjoint string to the adjoint source (”infinitely heavy gluon”) – the so-called gluelump .
Gluelump has a resemblance both to glueballs and to heavy-light mesons. In the framework of the string model, gluelump is a glueball with the adjoint string between gluons when one gluon is made static, and one expects the same structure of the spectrum of the LS-coupling type, which is obtained in and similar to the lattice spectrum calculated in .
On the other hand, gluelump differs (within the model) from heavy-light mesons only in that the fundamental string is replaced by the adjoint one (plus minor differences in the spin splitting terms and the role of color Coulomb forces).
Therefore the calculation of the gluelump spectrum is expected to be as successful as in the case of glueballs and heavy-light mesons. From physical point of view the gluelumps are important at least in two respects. First of all gluelumps play a fundamental role in the nonperturbative structure of the QCD vacuum, since gluelump Green’s functions are connected to the field correlators $`F_{\mu \nu }(x)F_{\lambda \sigma }(0)`$ or its invariant parts $`D(x),D_1(x)`$, introduced in , which are basic elements in the Method of Field Correlators (MFC). In particular, the gluon correlation length $`T_g`$, defining the nonperturbative dynamics of confinement is simply the inverse mass of the lowest gluelump, and was computed in this way before in several systematic studies .
Secondly, gluelump masses define the screening length for the static potential in given representation $`D`$ of $`SU(3)`$, namely the minimal length of the string, which may decay into two gluelumps.
Recent precise measurements of static potentials $`V_D(r)`$ made in revealed an accurate (within 1%) Casimir scaling without any signs of screening (string breaking) at all distances 0.05 fm $`r`$ 1.2 fm. To understand these results one should know screening length for all $`D`$.
The structure of the paper is as follows. In the first part of the paper the gluelump spectrum is computed, with the corresponding Hamiltonian constructed in section 2, spin splittings considered in section 3, and spectrum calculation in section 4. Comparison to the lattice data and the bag model is given in section 5. Section 6 contains conclusions and outlook, while Appendix is devoted to the details of hyperspherical formalism used for calculation of two-gluon gluelumps.
## 2 The gluelump Hamiltonian
The starting point is the gluelump Green’s function $`G^{glump}(x,y)`$ which is obtained from the initial and final gluelump operators $`\mathrm{\Psi }^{(in,out)}`$, expressed in terms of the valence gluon field $`a_\mu `$ and background gluon field $`B_\mu `$ so that the total field is $`A_\mu =B_\mu +a_\mu `$, with the gauge transformation properties
$$B_\mu U^+(B_\mu +\frac{i}{g}_\mu )U,a_\mu U^+a_\mu U.$$
(1)
Similarly to the glueball case one can write
$$G^{glump}(x,y)=tr_{adj}(\mathrm{\Gamma }^{(out)}(x)G_{\mu \nu }(x,y)\mathrm{\Gamma }^{in}(y)\mathrm{\Phi }(y,x))_B,$$
(2)
where $`\mathrm{\Phi }(x,y)`$ is the parallel transporter representating the adjoint source Green’s function:
$$\mathrm{\Phi }(x,y)=P\mathrm{exp}(ig_y^x\widehat{B}_\mu (z)𝑑z_\mu ),\widehat{B}_\mu =B_\mu ^aT^a$$
(3)
and $`G_{\mu \nu }(x,y)`$ is the valence gluon Green’s function
$$G_{\mu \nu }(x,y)=(D^2\delta _{\mu \nu }+2igF_{\mu \nu }(B))_{x,y}^1=$$
$$=_0^{\mathrm{}}𝑑sDze^KP\mathrm{exp}[ig_y^xB_\mu 𝑑z_\mu +2g_0^sF_{\mu \nu }(z(\tau ))𝑑\tau ]$$
(4)
In (2), $`\mathrm{\Gamma }^{(out,in)}(x)`$ are operators defining quantum numbers of final and initial gluelump states, below in Table 1 listed are examples for the lowest mass states.
The last term in the exponent in (4) describes the interaction of the gluon spin with the background $`B_\mu `$ (and with the adjoint string, which will be created out of it). Here as well as in the glueball case we shall treat this interaction as perturbation to be discussed in section 3, and disregard it in the first approximation. Then Eq. (2) reduces to the average of the adjoint Wilson loop with the contour $`C`$ consisting of the straight line of the source and the path $`z(\tau )`$ of the valence gluon, integrated upon in (4).
Assuming the area law for the adjoint Wilson loop in agreement with lattice measurements , one can go over in a standard way from the Green’s function $`G^{(glump)}`$ to the corresponding Hamiltonian $`H^{(glump)}`$, connected by
$$G^{glump}(x,y)=out|\mathrm{exp}(H^{(glump)}|xy|)|in.$$
(5)
The resulting Hamiltonian can be obtained from that of the $`q\overline{q}`$ system when mass of one quark is zero and mass of another going to infinity, while the quark string tension $`\sigma _f`$ is replaced by $`\sigma _{adj}=\frac{9}{4}\sigma _f`$. One obtains
$$H^{(glump)}=\frac{\mu }{2}+\frac{p_r^2}{2\mu }+\frac{L(L+1)/r^2}{2(\mu +_0^1𝑑\beta \beta ^2\nu (\beta ))}$$
$$+\frac{\sigma _{adj}^2r^2}{2}_0^1\frac{d\beta }{\nu (\beta )}+_0^1\frac{\nu (\beta )}{2}𝑑\beta .$$
(6)
Here $`\mu `$ and $`\nu (\beta )`$ are respectively the constituent gluon mass and energy density along the string coordinate $`\beta `$, to be found from minimization of the Hamiltonian.
For the angular momentum $`L=0`$ the stationary point in $`\nu (\beta )`$ yields a simple answer
$$H^{(glump)}(L=0)=\frac{\mu }{2}+\frac{p_r^2}{2\mu }+\sigma _{adj}r.$$
(7)
For $`L>0`$ instead of solving for the complicated operator (6) we shall treat effect of $`L`$ perturbatively as in , yielding accuracy around 5% for $`L3`$, namely
$$H^{(glump)}(L>0)=\frac{\mu }{2}+\frac{𝐩_r^2}{2\mu }+\sigma _{adj}r+\mathrm{\Delta }M_L$$
(8)
where we have defined as in
$$\mathrm{\Delta }M_L=\frac{4L(L+1)\sigma _{adj}^2}{3M_0^3}$$
(9)
and $`\mu `$ in (8) is to be found from the minimum of the mass eigenvalues of $`H^{(glump)}`$, (accuracy of this procedure was found in to be better than 5%).
## 3 Spin-dependent and color Coulomb interaction
We now take into account the term $`F_{\mu \nu }`$ in (4), which yields the gluon spin $`𝐒`$ dependence, since
$$2iF_{ik}=2𝐒𝐁,ik=1,2,3.$$
(10)
The analysis of the spin-dependent terms can be done as in the case of glueballs and heavy quarkonia (note however, that nowhere we use the nonrelativistic inverse mass expansion, instead the lowest (Gaussian) field correlator $`FF`$ is retained in the average of (4), this procedure was tested in to yield accuracy around 1%).
As a result one obtains
$$\mathrm{\Delta }H_{LS}=\mathrm{\Delta }H_{LS}(Thomas)+\mathrm{\Delta }H_{LS}(pert)$$
(11)
where
$$\mathrm{\Delta }H_{LS}(Thomas)=\sigma _{adj}\frac{𝐋𝐒}{2\mu ^2}\frac{f(r)}{r}b^{(Thomas)}𝐋𝐒$$
(12)
and $`f(r=\mathrm{})=1`$ takes into account that confining interaction starts at small $`r<T_g`$, as $`r^2`$, which significantly decreases the matrix element $`\frac{f(r)}{r}`$, see for details, and for numerical results for the glueball case.
With $`\mathrm{\Delta }H(pert)`$ the situation is more subtle. Indeed the one-gluon exchange does not contribute to the potential $`V_1(r)`$ (we stick to the standard notations of Eichten-Feinberg-Gromes approach), and the latter is the only $`LS`$ piece surviving for the gluelump case. Hence one should consider the one-loop contribution, similar to the one considered in for heavy quarkonia, in the limit of one mass infinite and another equal to $`\mu `$. In absence of actual calculations for the gluelump case, and to have an orientation, we shall translate the result of replacing $`C_FC_A`$, and $`m_1\mu `$, $`m_2\mathrm{}`$. The result is
$$\mathrm{\Delta }H_{LS}(pert)=\frac{𝐋𝐒}{\mu ^2r^3}\frac{\alpha _s^2C_A^2}{2\pi }(2\mathrm{ln}\mu r\gamma _E)b^{(pert)}𝐋𝐒.$$
(13)
As we shall see the overall magnitude of $`\mathrm{\Delta }H_{LS}`$ and of its parts is small, and finally can be neglected. For an estimate we shall take the brackets in (13) to be equal to one, and the matrix element $`\frac{1}{r^3}`$ can be found from the equality
$$L(L+1)\frac{1}{r^3}=\mu \sigma _{adj}$$
(14)
Consider now the contribution of perturbative gluon exchanges to the interaction between the valence gluon and static source.
Here we adopt the same attitude that was suggested and discussed in detail in . Namely, the perturbative ladder for the gluelump is equivalent to the BFKL ladder of the pomeron (more exactly to one-half of that ladder cut along the valence gluon line). For the BFKL ladder it was shown that the higher-order correction practically cancels the lowest order result. This fact was used in the glueball case in to neglect the Coulomb ladder completely, and the resulting glueball masses agree surprisingly well with the lattice data (whereas retaining color Coulomb interaction would reduce the lowest masses by 0.5 GeV, and increase spin splittings 3 times in strong disagreement with lattice data).
Accordingly for the gluelump case we at first also disregard color Coulomb exchanges both in mass eigenvalues and in wave functions, so that relation (14) holds true and the matrix element $`\frac{f(r)}{r}`$ is to be calculated with eigenfunctions of (7) and (8). One finds that the spacial size of gluelump (in absence of color Coulomb force) is $`\sqrt{2}`$ times less than that of two-gluon glueballs, and is around 0.5 fm for lowest states. Hence the influence of (however reduced) Coulomb force in this system can be important.
Therefore also calculations have been done for two other values of effective $`\alpha _s`$, which when taking into account higher loop effects reduce to a smaller value, denoted by $`\overline{\alpha }_s`$. Two values of $`\overline{\alpha }_s`$, $`\overline{\alpha }_s=0.15`$ and $`\overline{\alpha }_s=0.195`$ have been used and the results for the spin-overaged mass are shown in Table 2, while the values of spin-orbit matrix elements $`b^{(Thomas)}`$ and $`b^{(pert)}`$ change for $`\overline{\alpha }_s=0.15`$ at most 1.5 times and not listed in the Table.
The resulting values of spin-splitted masses are given in Table 4 for $`\overline{\alpha }_s=0`$.
## 4 The gluelump spectrum
We are now in position to calculate the spectrum of (8) and the spin splittings (11), (12).
The total mass eigenvalue is written as
$$M(J,L,n_r)=M_0(n,L)+\mathrm{\Delta }M_L+\mathrm{\Delta }M_{LS}$$
(15)
where $`M_0(n,L)`$ is the eigenvalue of the operator $`h(\mu )\mu /2+𝐩^2/2\mu +\sigma _{adj}r`$, minimized over the values of $`\mu `$.
$$h(\mu )\psi (x)=\epsilon (\mu )\psi (x).$$
(16)
Using the standard technic of the QCDSM , one finds
$$\epsilon (\mu )=\frac{\mu }{2}+\frac{\sigma _{adj}^{2/3}a(n,L)}{(2\mu )^{1/3}}$$
(17)
where $`a(n,L)`$ are the eigenvalues of the reduced equation found numerically and given in Table 1 (lower entries). Minimization of (17) over $`\mu `$ yields for $`\mu =\mu _0`$ the values given in Table 1 (upper entries). One has from (17)
$$\mu _0(n,L)=(\frac{a(n,L)}{3})^{3/4}(2\sigma _{adj})^{1/2}.$$
(18)
From $`\epsilon (\mu )`$ (17) one immediately finds that
$$M_0(n,L)=2\mu _0(n,L)$$
(19)
The values of $`M_0(n,L),\mathrm{\Delta }M_L`$ and the resulting spin-averaged masses $`\overline{M}(n,L)M_0(n,L)+\mathrm{\Delta }M_L`$ are given in Table 2.
Now we turn to the spin splittings of the obtained levels. The values of $`\frac{1}{r^3}`$ are taken from (14) and those of $`\frac{f(r)}{r}`$ from the corresponding glueball matrix elements in Table 7 of , with the proper rescaling from the two-gluon Hamiltonian to $`h(\mu )`$ . The resulting splittings $`\mathrm{\Delta }M_{LS}`$ can be written as
$$\mathrm{\Delta }M_{LS}=\frac{J(J+1)L(L+1)2}{2}(b^{(Thomas)}(n,L)+b^{(pert)}(n,L))$$
(20)
The values of $`b^{(Thomas)}`$ and $`b^{(pert)}`$ are given in Table 3. Finally summing up all corrections one obtains the resulting gluelump masses listed in Table 4 for $`\overline{\alpha }_s=0`$.
Now we turn to the two-gluon gluelumps, which correspond to $`C=+1`$. This is a special case of $`3g`$ glueballs when the mass of one gluon tends to infinity. Therefore one can use the same Hamiltonian technic as was exploited for $`3g`$ glueballs in , i.e. the Hamiltonian $`h(\mu )`$ in (16) is replaced by $`h(\mu _1,\mu _2)`$ (with $`\mu _i`$ to be found again from minimization of eigenvalues).
$$h(\mu _1,\mu _2)=\frac{\mu _1}{2}+\frac{\mu _2}{2}+\frac{𝐩_1^2}{2\mu _1}+\frac{𝐩_2^2}{2\mu _2}+\sigma _f\left\{|𝐫_1|+|𝐫_2|+|𝐫_1𝐫_2|\right\}.$$
(21)
Note the appearence of $`\sigma _f`$ in (21), since the string is now fundamental and forms a triangle, passing through two gluons and the adjoin source. The Hamiltonian (21) can be treated using hyperspherical formalism, for details see Appendix. Here we only quote the result for spin-averaged masses (obtained with accuracy better than 5%)
$$M(n,K)=4\mu _0+\omega _0(n+\frac{1}{2})+\frac{K(K+4)}{15}4\mu _0$$
(22)
where $`n=0,1,2,\mathrm{};K=0,1,2,\mathrm{}`$ and
$$\mu _0=2\sqrt{\sigma _f}\left(\frac{2}{15}\right)^{1/4}(\sqrt{2}+2)^{1/2},\omega _0=\frac{4}{\sqrt{5}}\mu _0.$$
(23)
For the lowest state, $`K=0,n=0`$ one obtains
$$M^{(2g)}(0,0)=2.61GeV(\sigma _f=0.18GeV^2).$$
(24)
The hyperfine splitting of this level into $`J^{PC}=0^{++}`$ and $`2^{++}`$ can be calculated in the same way as in with the result that $`0^{++}`$ lies below $`2^{++}`$ with the interval proportional to $`\alpha _s`$ (see Appendix for explicit expression). We place the spin-averaged value (24) in Table 4 and hyperfine-splitted level $`0^{++}`$ in Table 5. It is interesting to note that in two-gluon gluelumps both hyperfine and tensor forces are operating in contrast to one-gluon gluelumps, and therefore the mixing of the state $`𝐁^{(1)}𝐁^{(2)}`$ (which is orbitally excited with respect to $`a_i^{(1)}a_i^{(2)}`$) with the latter is significant, which may explain the appearence of this state in lattice calculations.
## 5 Discussion of the spectrum. Comparison with lattice data
The spectrum given in Tables 2 and 4 demonstrates the following features. First of all one can see that the spin splittings of levels are small and can be neglected in first approximation. The ordering of levels has the same character as for two-gluon glueball masses and can be roughly described by the equation
$$\overline{M}(n,L)(2n+L)\omega +\overline{M}_0$$
(25)
where $`\overline{M}_0(\overline{\alpha }_s)(1.52.5\overline{\alpha }_s)`$ GeV, and $`\omega (\overline{\alpha }_s)(0.35+\overline{\alpha }_s1.2)`$ GeV. This behaviour is typical for other hadrons, see for the discussion of glueballs and for a review of hadronic spectra.
¿From dimensional point of view the spectrum corresponds to the increasing dimension of valence gluon operators $`a_i,_ia_k,_i_ka_l`$ etc. (and not to the dimension of operators $`E_i,B_i,D_iB_k`$ etc.).
Note that the lowest level, $`1^{}`$, is of the electric type and there is a gap of $`0.5÷0.6`$ GeV between the ground and excited level, while the distance between $`L=0`$ and $`L=2`$ levels is equal to 0.73 GeV $`(\overline{\alpha }_s=0)`$ or 1 GeV ($`\overline{\alpha }_s=0.15).`$
It is instructive to compare our spectrum to the gluelump spectrum of the MIT bag model .
For $`\mathrm{\Lambda }^{1/4}=0.315`$ GeV and $`\alpha _s=0.23`$ the masses obtained in are
$$M(1^+)=1.43\mathrm{GeV},M(2^{})=1.97\mathrm{GeV},M(1^{})=1.98\mathrm{GeV}$$
$$M(3^+)=2.44\mathrm{GeV},M(2^+)=2.64\mathrm{GeV}.$$
(26)
One can see the inverse ordering of the first levels in (26), namely magnetic level $`1^+`$ in MIT bag model is below the electric one $`1^{}`$ and the gap between them is approximately the same as in QCD string model, but the ordering is reversed.
Here the difference between two models is most pronounced (in contrast the glueball spectrum, where the lowest two-gluon states have the same ordering in both models, see and ). Hence the independent check of lattice calculations can in principle decide which of the models is closer to reality.
Now the recent lattice calculations (see also for earlier lattice studies) yield the mass levels shown in Table 5. One should note at this point, that the (divergent) mass renormalization terms are not subtracted from the data and therefore absolute scale is missing. We therefore have placed for the sake of comparison in Table 5 the lowest $`1^+`$ level at the same mass as given by our calculations, see Table 2 and 4 for $`\overline{\alpha }_s=0.15`$.
Another feature of data from is that only spacial links have been used for gluelump operators at initial and final states, and hence the states in Table 4 containing $`E_j`$ are not directly excited in measurements in (however these states could in principle be excited indirectly due to mixing, i.e. with smaller overlap). The states in Table 4 which are excited directly in are marked by asterix and compared with lattice results in Table 5.
One can see in Table 5 a good correspondence between the levels, and an approximate mass degeneration of $`2^+,3^+`$ levels is present in both results.
However the $`1^{},2^{}`$ levels, almost degenerate in the QCDSM have a gap of 0.22 GeV on the lattice. This might be explained by the mixing of the $`1^{}`$ level with the ground state, having the same quantum numbers, whereas $`2^{}`$ has no low-lying counterpart.
As was mentioned above, the crucial test of the QCDSM is the search for the ground state level $`1^{}`$, which should lie $`0.5÷0.6`$ GeV below the lowest $`1^+`$ level found on the lattice. If this level is not found, it would possibly mean that gluelumps are more like bags than strings. However no derivation of the bag model from QCD was ever made and theoretical grounds for this model are of intuitive character, whereas the QCDSM follows directly (with few assumptions supported by lattice data) from the QCD lagrangian. Therefore the existence of the $`1^{}`$ ground state for the gluelump seems very plausible. In conclusion of this section we discuss shortly an important connection between gluelumps and field strength correlators $`D(x),D_1(x)`$ introduced in and measured on the lattice in using cooling procedure and in using RG smoothing. It is clear from Table 4 that two lowest states, $`1^{}`$ and $`1^+`$, correspond to the electric and magnetic correlators respectively, and therefore the masses $`M(1^{})1.12`$ GeV and $`M(1^+)1.7`$ GeV $`(\overline{\alpha }_s=0.15)`$ should give the exponential slopes of those correlators (the notation $`D_{}`$ and $`D_{}`$ was also used in coiciding with electric and magnetic respectively). The lowest mass of $`M1`$ GeV is in agreement with measurements in , however for $`D_{}`$ the same slope was found in contrast with present calculations. There are no general symmetry arguments, why masses (slopes) in both functions should be the same, (in contrast to electric and magnetic condensates at zero temperature) and this disagreement between gluelump masses and slopes in $`D_{},D_{}`$ calls for further investigation.
One should note in addition, that the independent calculation of the slope from the selfcoupled equations for correlators made in , also gives the lowest (electric) mass around 1 GeV, in agreement with the present calculations for $`(\overline{\alpha }_s=0.15)`$.An analysis for the same quantity in the framework of the QCD sum rules gives a larger value in the quenched case, while for the magnetic mass the result was unstable. Finally,in the lattice study of electric and magnetic field correlators made without cooling or smoothing procedures,electric and magnetic masses are obtained in good agreement with present calculations, specifically magnetic mass turned out to be around 0.5 GeV heavier.
## 6 Conclusions
Results of the present calculations in the QCDSM framework are in general agreement with recent lattice data , using the restricted set of gluonic operators, and with previous lattice and analytic calculations of the slope of the $`D_{}`$ function. It is not clear from the gluelump spectrum point of view why the correlator $`D_{}`$ (magnetic field correlator) should have the same slope as that of $`D_{}`$, as it was found on the lattice . With the lowest gluelump mass equal to $`1.2÷1.5`$ GeV in the present calculations, the onset of the adjoint string breaking appears at the string length $`r_0=1.2÷1.5`$ fm, in agreement with recent lattice data .
The fundamental role of gluelump states in the theory of the QCD vacuum makes it important to resolve the existing problem of the spectrum and of the relation with vacuum correlators $`D_{},D_{}`$.
The author is grateful to C.Michael for a useful correspondence, and to Yu.S.Kalashnikova and V.I.Shevchenko for numerous discussions. The present work was partially supported in the framework of the RFFI projects 00-02-17836 and 96-15-96740. This work was completed while the author was visiting the Humboldt University Of Berlin in the framework of the joint DFG-RFFI project 96-02-00088G.It is a pleasure for the author to thank M.Mueller- Preussker for a kind hospitality,collaboration and discussions and G.Bali for a series of discussions which stimulated the final version of the paper.
Table 1
Valence gluon masses $`\mu _0(n,L)`$ (upper entries) and reduced eigenvalues $`a(n,l)`$ (lower entries). $`\sigma _f=0.18`$ GeV$`{}_{}{}^{2},\sigma _{adj}=\frac{9}{4}\sigma _f=0.405`$ GeV
| $`Ln`$ | 0 | 1 | 2 |
| --- | --- | --- | --- |
| 0 | 0.746 | 1.135 | 1.422 |
| | 2.3381 | 4.0879 | 5.520 |
| 1 | 0.98 | 1.297 | 1.554 |
| | 3.3613 | 4.8845 | 6.216 |
| 2 | 1.168 | 1.443 | |
| | 4.248 | 5.63 | |
| 3 | 1.33 | | |
| | 5.053 | | |
Table 2
Unperturbed masses $`M_0(n,L)`$, string correction $`\mathrm{\Delta }M_L`$ and spin averaged masses $`\overline{M}(n,L)`$ for $`L=0,1,2,3`$ and $`n=0`$ (upper entries and two last lines) and $`n=1`$ (lower entries).
| $`L`$ | 0 | 1 | 2 | 3 |
| --- | --- | --- | --- | --- |
| $`M_0`$ | 1.492 | 1.96 | 2.336 | 2.66 |
| (GeV) | 2.27 | 2.6 | 2.886 | |
| $`\mathrm{\Delta }M_L`$ | 0 | -0.0581 | -0.103 | -0.139 |
| (GeV) | 0 | -0.025 | -0.0546 | |
| $`\overline{M}`$ | 1.492 | 1.9 | 2.233 | 2.52 |
| (GeV) | 2.27 | 2.575 | 2.83 | |
| $`\overline{M}(\overline{\alpha }_s=0.15)`$ | 1.12 | 1.73 | 2.15 | 2.46 |
| $`\overline{M}(\overline{\alpha }_s=0.195)`$ | 0.982 | 1.67 | 2.10 | |
Table 3
Perturbative and Thomas spin-orbit martix elements defined as in (11-13)
| $`L`$ | 1 | 2 | 3 |
| --- | --- | --- | --- |
| $`b^{(pert)}`$ | 0.033 | 0.0096 | 0.00425 |
| $`b^{(Thomas)}`$ | -0.0025 | -0.0024 | -0.0021 |
Table 4
Gluelump masses for lowest states $`J^{PC}`$ are listed together with gluonic state operators in the background field formalism and in the general notations. Asterix marks the states which are directly generated on the lattice
| State | $`L`$ | $`n`$ | operator | | Mass |
| --- | --- | --- | --- | --- | --- |
| | | | backgr. | general | (GeV) |
| $`1^{}`$ | 0 | 0 | $`a_i`$ | $`E_i`$ | 1.492 |
| $`1^+`$ | 1 | 0 | $`e_{ikl}_ka_l`$ | $`B_i`$ | 1.87\* |
| $`2^+`$ | 1 | 0 | $`(_ia_k)_{symm}`$ | $`(D_iE_k)_{symm}`$ | 1.93 |
| $`1^{}`$ | 0 | 1 | $`a_i`$ | $`E_i`$ | 2.27 |
| $`1^{}`$ | 2 | 0 | $`(_i_ka_l)_J`$ | $`D_iB_ke_{ikl}`$ | 2.21\* |
| $`2^{}`$ | 2 | 0 | | $`(D_iB_k)_{symm}`$ | 2.226\* |
| $`3^{}`$ | 2 | 0 | | $`D_iD_kE_l`$ | 2.24 |
| $`2^+`$ | 3 | 0 | $`(_i_k_la_m)_J`$ | $`(D_iD_kB_l)_J`$ | 2.51\* |
| $`3^+`$ | 3 | 0 | | | 2.52\* |
| $`4^+`$ | 3 | 0 | | $`(D_iD_kD_lE_m)_4`$ | 2.53 |
| $`1^+`$ | 1 | 1 | ($`_ia_k)_1`$ | $`B_i`$ | 2.57\* |
| $`2^+`$ | 1 | 1 | ($`_ia_k)_2`$ | $`D_iE_k`$ | 2.57 |
| $`0^{++}2^{++}`$ | 0 | 0 | $`(a_i^{(1)}a_k^{(2)})_J`$ | $`(𝐁_i^{(1)}𝐁_k^{(2)}+𝐄_i^{(1)}𝐄_k^{(2)})_J`$ | 2.61\* |
Table 5
The gluelump states generated by spacial operators, with masses computed analytically in the present paper $`(\overline{\alpha }_s=0.15)`$ and masses computed on the lattice . For better comparison results of Table 2 corrected for larger $`\sigma _f=0.22`$ used on lattice.
| $`J^{PC}`$ | $`1^+`$ | $`1^{}`$ | $`2^{}`$ | $`2^+`$ | $`3^+`$ | $`0^{++}`$ |
| --- | --- | --- | --- | --- | --- | --- |
| $`M`$ (GeV) | 1.87 | 2.23 | 2.45 | 2.84 | 2.84 | 2.96 |
| lattice | | | | | | |
| $`M`$ (GeV) | 1.87 | 2.34 | 2.36 | 2.70 | 2.71 | 2.78 |
| QCD string | | | | | | |
Appendix
Two-gluon gluelumps
Consider a static source at point $`𝐫_3=0`$ and two gluons at $`𝐫=𝐫_1`$ and $`𝐫=𝐫_2`$, connected by a fundamental string, passing through these three points. The Hamiltonian for the case, when both orbital momenta are zero (no string rotation correction) can be written in analogy to (7) as
$$h(\mu _1,\mu _2)=\frac{\mu _1+\mu _2}{2}+\frac{𝐩_1^2}{2\mu _1}+\frac{𝐩_2^2}{2\mu _2}+\sigma _f\{|𝐫_1|+|𝐫_2|+|𝐫_1𝐫_2|\}$$
(A.1)
This 3 body problem can be treated by the hyperspherical formalism . Introducing the hyperradius $`\rho `$ and denoting $`\mu _1=\mu _2=\mu `$, one has
$$\rho ^2=𝜼^2+𝝃^2,𝜼=𝐫_{12}/\sqrt{2},𝝃=(𝐫_1+𝐫_2)\frac{1}{\sqrt{2}}$$
(A.2)
one obtains the Hamiltonian for the given grand orbital momentum $`K=0,1,2,\mathrm{}`$
$$h=+\mu \frac{1}{2\mu \rho ^5}\frac{}{\rho }(\rho ^5\frac{}{\rho })+\frac{(K+\frac{3}{2})(K+\frac{5}{2})}{2\mu \rho ^2}+U_K(\rho )$$
(A.3)
where $`U_K(\rho )`$ for $`K=0`$ is the potential term in (A.1) averaged over all hyperangles. Using standard expressions from one obtains
$$U_0(\rho )=C_0\sigma \rho ,C_0=\frac{32\sqrt{2}(1+\sqrt{2})}{15\pi }$$
(A.4)
Eq. (A.3) can be solved numerically and results are tabulated, but it appears that the accuracy of about 1% for the eigenvalue can be obtained by the stationary point method applied to the effective potential $`W(\rho )`$
$$W(\rho )=\frac{(K+\frac{3}{2})(K+\frac{5}{2})}{2\mu \rho ^2}+U_K(\rho );W^{}(\rho =\rho _0)=0;$$
(A.5)
while the second derivative at $`\rho =\rho _0`$ defines the radial excitation energy $`\omega `$:
$$W^{\prime \prime }(\rho _0)\frac{(\rho \rho _0)^2}{2}\frac{\mu \omega ^2(\rho \rho _0)^2}{2}$$
(A.6)
In this way one obtains the mass eigenvalue $`M(\mu )`$, and finally the stationary point condition for $`\mu `$, $`\frac{dM}{d\mu }(\mu =\mu _0)=0`$ defines gluon constituent mass $`\mu _0`$ and the resulting mass eigenvalue $`M(\mu _0)`$, given in (22)-(24). One can note that radial excitations are given by $`\omega _0=0.955`$ GeV, and ”grand-orbital excitation” from $`K=0`$ to $`K=1`$ is given by the interval $`\mathrm{\Delta }M=0.712`$ GeV.
The hyperfine splitting is given by the Hamiltonian (cf the discussion in for 3$`g`$ glueballs)
$$H_{SS}=𝐒^{(1)}𝐒^{(2)}\frac{5\pi C_2(fund)\alpha _s}{3\mu _0^2}\delta ^{(3)}(𝐫_{12})_\rho $$
(A.7)
The last factor in (A.7) is easily computed
$$\delta ^{(3)}(𝐫_{12})_\rho =\frac{\sqrt{2}}{\pi ^2\rho _0^3},\rho _0=1.15/\mu _0.$$
(A.8)
This yields the spin splitting of the levels
$$\mathrm{\Delta }M_{SS}=𝐒^{(1)}𝐒^{(2)}0.49\mu _0\frac{4}{3}\alpha _s.$$
(A.9)
For $`\alpha _s=\overline{\alpha }_s=0.15`$ one obtains
$$\mathrm{\Delta }M_{SS}=0.0525\mathrm{GeV}\left(\begin{array}{cc}2,\hfill & J=0\hfill \\ +1,\hfill & J=2\hfill \end{array}\right).$$
(A.10)
In this way one gets the mass $`M(0^{++})`$ in Table 5.
|
warning/0003/hep-ph0003217.html
|
ar5iv
|
text
|
# Reconciling 𝐺_𝐴/𝐺_𝑉, and 𝜇_{𝑝,𝑛} in 𝜒QM with One Gluon Generated Configuration Mixing.
## Abstract
The spin polarization functions $`(\mathrm{\Delta }u,\mathrm{\Delta }d,\mathrm{\Delta }s)`$ for proton are calculated in the chiral quark model ($`\chi `$QM) with SU(3) symmetry breaking as well as configuration mixing generated by one gluon exchange forces for the NMC and the most recent E866 data. Besides reproducing the spin polarization functions $`\mathrm{\Delta }u,\mathrm{\Delta }d,\mathrm{\Delta }s`$ as well as $`G_A/G_V`$, it can accomodate nucleon magnetic moments and neutron charge radius as well, thus resolving the compatibility problem of these parameters which could not be achieved in constituent quark models.
The Constituent Quark Model (CQM), despite its impressive performance in explaining low energy hadronic matrix elements, is unable to account for the EMC effect or the “proton spin crisis” which indicated that only a small portion of the proton spin is carried by the valence quarks, as well as the presence of significant negative strange quark polarization in the proton quark sea . CQM is also unable to explain the results of the NMC and E866 experiments which have shown that the Gottfried sum rule is violated, indicating that the $`\overline{d}`$ density is larger than $`\overline{u}`$ density in the nucleon sea. Apart from the problems faced by CQM in explaining the spin content of nucleons, it is also saddled with another problem, $`viz.`$, it has been shown that it is not possible to have a simultaneous reconciliation of $`G_A/G_V`$, charge form factors of nucleon and magnetic moments of nucleons ($`\mu _{p,n}`$) in the CQM. This problem becomes more acute if one considers neutron charge radius $`(<r_n^2>)`$.
The EMC effect and related issues have been successfully addressed in the chiral quark model ($`\chi `$QM) , originally conceived by Weinberg , subsequently developed by Manohar and Georgi . In particular, the $`\chi `$QM is able to explain not only the spin content of the nucleon but is also able to give a fair description of flavor content of the nucleons, violation of Gottfried sum rule, strange quark content in the nucleon, $`G_A/G_V`$ as well as the magnetic moments of the baryons etc. .
Recently it has been claimed that, within the $`\chi `$QM, neutron charge radius can also be reproduced through the Foldy term , however very recently it has been shown by Isgur that the Foldy term gets cancelled against a contribution to the Dirac form factor $`F_1`$ to leave intact the interpretation of neutron electric charge form factor $`G_E^n`$ as arising from the neutron’s rest frame charge distribution. Therefore, in the light of the work of Isgur, it becomes essential that $`<r_n^2>`$ should be reproduced by the form factor $`F_1`$. Thus, even in the $`\chi `$QM, the question of compatibility of $`G_A/G_V`$, $`<r_n^2>`$ and $`\mu _{p,n}`$ remains.
In the context of CQM, it is well known that there are several low energy parameters which can be explained by including one gluon mediated configuration mixing . It has also been shown that one gluon mediated configuration mixing is able to generate $`<r_n^2>`$ as well as it improves the fit of magnetic moment of baryons . In view of the fact that the $`\chi `$QM incorporates the basic features of constituent quark model, therefore it becomes interesting to examine the implications of configuration mixing in $`\chi `$QM ($`\chi `$QM<sub>gcm</sub>). The purpose of the present communication on the one hand is to examine, within the $`\chi `$QM, the implications of one gluon mediated configuration mixing for spin polarizations and related issues while on the other hand it is to investigate the issue of compatibility of $`G_A/G_V`$, neutron charge radius and $`\mu _{p,n}`$.
To understand the implications of one gluon mediated configuration mixing for $`G_A/G_V`$, we first calculate the quark spin polarizations in the $`\chi `$QM. The basic process, in the $`\chi `$QM, is the emission of a Goldstone Boson which further splits into $`q\overline{q}`$ pair, for example,
$$q_\pm GB^0+q_{}^{^{}}(q\overline{q}^{^{}})+q_{}^{^{}}.$$
(1)
The above process can be expressed through the Lagrangian
$$=g_8\overline{q}\mathrm{\Phi }q,$$
(2)
where $`g_8`$ is the coupling constant,
$$q=\left(\begin{array}{c}u\\ d\\ s\end{array}\right),$$
(3)
and
$$\mathrm{\Phi }=\left(\begin{array}{ccc}\frac{\pi ^0}{\sqrt{2}}+\beta \frac{\eta }{\sqrt{6}}+\zeta \frac{\eta ^{^{}}}{\sqrt{3}}& \pi ^+& \alpha K^+\\ \pi ^{}& \frac{\pi ^0}{\sqrt{2}}+\beta \frac{\eta }{\sqrt{6}}+\zeta \frac{\eta ^{^{}}}{\sqrt{3}}& \alpha K^0\\ \alpha K^{}& \alpha \overline{K}^0& \beta \frac{2\eta }{\sqrt{6}}+\zeta \frac{\eta ^{^{}}}{\sqrt{3}}\end{array}\right).$$
(4)
SU(3) symmetry breaking is introduced by considering different quark masses $`m_s>m_{u,d}`$ as well as by considering the masses of non-degenerate Goldstone Bosons $`M_{K,\eta }>M_\pi `$ , whereas the axial U(1) breaking is introduced by $`M_\eta ^{^{}}>M_{K,\eta }`$ . The parameter a(=$`|g_8|^2`$) denotes the transition probability of chiral fluctuation or the splittings $`u(d)d(u)+\pi ^{+()}`$, whereas $`\alpha ^2a`$ denotes the probability of transition $`u(d)s+K^{(0)}`$. Similarly $`\beta ^2a`$ and $`\zeta ^2a`$ denote the probability of $`u(d,s)u(d,s)+\eta `$ and $`u(d,s)u(d,s)+\eta ^{^{}}`$ respectively.
The one gluon exchange forces generate the mixing of the octet in $`(56,0^+)_{N=0}`$ with the corresponding octets in $`(56,0^+)_{N=2}`$, $`(70,0^+)_{N=2}`$ and $`(70,2^+)_{N=2}`$ harmonic oscillator bands . The corresponding wave function of the nucleon is given by
$$|B>=(|56,0^+>_{N=0}cos\theta +|56,0^+>_{N=2}sin\theta )cos\varphi $$
$$+(|70,0^+>_{N=2}cos\theta +|70,2^+>_{N=2}sin\theta )sin\varphi .$$
(5)
In the above equation it should be noted that $`(56,0^+)_{N=2}`$ does not affect the spin-isospin structure of $`(56,0^+)_{N=0}`$, therefore the mixed nucleon wave function can be expressed in terms of $`(56,0^+)_{N=0}`$ and $`(70,0^+)_{N=2}`$, which we term as non trivial mixing and is given as
$$|8,\frac{1}{2}^+=cos\varphi |56,0^+>_{N=0}+sin\varphi |70,0^+>_{N=2},$$
(6)
where
$$|56,0^+>_{N=0,2}=\frac{1}{\sqrt{2}}(\chi ^{^{}}\varphi ^{^{}}+\chi ^{^{\prime \prime }}\varphi ^{^{\prime \prime }})\psi ^s,$$
(7)
$$|70,0^+>_{N=2}=\frac{1}{2}[(\psi ^{^{\prime \prime }}\chi ^{^{}}+\psi ^{^{}}\chi ^{^{\prime \prime }})\varphi ^{^{}}+(\psi ^{^{}}\chi ^{^{}}\psi ^{^{\prime \prime }}\chi ^{^{\prime \prime }})\varphi ^{^{\prime \prime }}].$$
(8)
The spin and isospin wave functions, $`\chi `$ and $`\varphi `$, are given below
$$\chi ^{^{}}=\frac{1}{\sqrt{2}}(),\chi ^{^{\prime \prime }}=\frac{1}{\sqrt{6}}(2),$$
$$\varphi _p^{^{}}=\frac{1}{\sqrt{2}}(ududuu),\varphi _p^{^{\prime \prime }}=\frac{1}{\sqrt{6}}(2uudududuu),$$
$$\varphi _n^{^{}}=\frac{1}{\sqrt{2}}(udddud),\varphi _n^{^{\prime \prime }}=\frac{1}{\sqrt{6}}(udd+dud2ddu).$$
For the definition of the spatial part of the wave function, ($`\psi ^s,\psi ^{^{}},\psi ^{^{\prime \prime }})`$ as well as the definitions of the spatial overlap integrals we refer the reader to references and .
The contribution to the proton spin, defined through the equation
$$\mathrm{\Delta }q=q_{}q_{}+\overline{q}_{}\overline{q}_{},$$
(9)
using Equation(4) and following Linde $`etal.`$ , can be expressed as
$$\mathrm{\Delta }u=cos^2\varphi \left[\frac{4}{3}\frac{a}{3}(7+4\alpha ^2+\frac{4}{3}\beta ^2+\frac{8}{3}\zeta ^2)\right]+sin^2\varphi \left[\frac{2}{3}\frac{a}{3}(5+2\alpha ^2+\frac{2}{3}\beta ^2+\frac{4}{3}\zeta ^2)\right],$$
(10)
$$\mathrm{\Delta }d=cos^2\varphi \left[\frac{1}{3}\frac{a}{3}(2\alpha ^2\frac{1}{3}\beta ^2\frac{2}{3}\zeta ^2)\right]+sin^2\varphi \left[\frac{1}{3}\frac{a}{3}(4+\alpha ^2+\frac{1}{3}\beta ^2+\frac{2}{3}\zeta ^2)\right],$$
(11)
and
$$\mathrm{\Delta }s=a\alpha ^2.$$
(12)
Before one presents and discusses the results pertaining to Equations (10), (11) and (12), for a better appreciation of the role of configuration mixing and symmetry breaking we have considered the case of SU(3) symmetry as well. The SU(3) symmetric calculations can easily be obtained from the equations (10), (11) and (12) by considering $`\alpha ,\beta =1`$. The corresponding equations can be expressed as
$$\mathrm{\Delta }u=cos^2\varphi \left[\frac{4}{3}\frac{a}{9}(37+8\zeta ^2)\right]+sin^2\varphi \left[\frac{2}{3}\frac{a}{9}(23+4\zeta ^2)\right],$$
(13)
$$\mathrm{\Delta }d=cos^2\varphi \left[\frac{1}{3}\frac{2a}{9}(\zeta ^21)\right]+sin^2\varphi \left[\frac{1}{3}\frac{a}{9}(16+2\zeta ^2)\right],$$
(14)
and
$$\mathrm{\Delta }s=a.$$
(15)
In Table 1, we have presented the results of our calculations. First of all, for $`\chi `$QM with SU(3) symmetry breaking as well as configuration mixing, we have carried out a $`\chi ^2`$ fit to $`\mathrm{\Delta }u,\mathrm{\Delta }d,\mathrm{\Delta }s,G_A/G_V`$ and other related parameters (details of which will be published elsewhere), however, in the fit we have taken $`\varphi `$ 20<sup>0</sup>, a value dictated by consideration of neutron charge radius . In the table we have also considered a few more values of the mixing parameter $`\varphi `$ in order to study the variation of spin distribution functions with $`\varphi `$. The parameter $`a`$ is taken to be 0.1, as considered by other authors . The symmetry breaking parameters obtained from $`\chi ^2`$ fit are $`\alpha =.4`$ and $`\beta =.7`$ for the data corresponding to recent E866 as well as NMC data . The parameter $`\zeta `$ is constrained by the expressions $`\zeta =0.7\beta /2`$ and $`\zeta =\beta /2`$ for the NMC and E866 experiments respectively, which essentially represent the fitting of deviation from Gottfried sum rule . Further, while presenting the results of SU(3) symmetry breaking case without configuration mixing $`(\varphi =0^0)`$, we have used the same values of parameters $`\alpha `$ and $`\beta `$, primarily to understand the role of configuration mixing for this case. The SU(3) symmetry calculations based on Equations (13), (14) and (15) are obtained by taking $`\alpha =\beta =1,\varphi =20^0`$ and $`\alpha =\beta =1,\varphi =0^0`$ respectively for with and without configuration mixing. For the sake of completion, we have also presented the results of CQM with and without configuration mixing. In this case the spin polarization functions can easily be found from equations (10), (11) and (12), for example,
$$\mathrm{\Delta }u=cos^2\varphi [\frac{4}{3}]+sin^2\varphi [\frac{2}{3}],$$
(16)
$$\mathrm{\Delta }d=cos^2\varphi [\frac{1}{3}]+sin^2\varphi [\frac{1}{3}],$$
(17)
and
$$\mathrm{\Delta }s=0.$$
(18)
From Table 1, it is clear that $`\chi `$QM with SU(3) symmetry breaking along with configuration mixing generated by one gluon exchange forces is able to give an excellent fit to the spin polarization data for symmetry breaking parameters $`\alpha =.4`$ and $`\beta =.7`$. In order to appreciate the role of configuration mixing in affecting the fit, we first compare the results of CQM with those of CQM<sub>gcm</sub>. One observes that configuration mixing corrects the result of the quantities in the right direction but this is not to the desirable level. Further, in order to understand the role of configuration mixing and SU(3) symmetry with and without breaking in $`\chi `$QM, we can compare the results of $`\chi `$QM with SU(3) symmetry to those of $`\chi `$QM<sub>gcm</sub> with SU(3) symmetry. Curiously $`\chi `$QM<sub>gcm</sub> compares unfavourably with $`\chi `$QM in case of the calculated quantities. This indicates that configuration mixing alone is not enough to generate an appropriate fit in $`\chi `$QM. However when $`\chi `$QM<sub>gcm</sub> is used with SU(3) symmetry breaking then the results show uniform improvement over the corresponding results of $`\chi `$QM with SU(3) symmetry breaking. It is interesting to note that the values of $`\alpha `$ and $`\beta `$ are in agreement with other recent calculations .
After having seen that $`\chi `$QM<sub>gcm</sub> is able to accomodate $`G_A/G_V`$, one turns to the question of compatibility of $`G_A/G_V`$, $`\mu _{p,n}`$ and $`<r_n^2>`$. In this regard, we first evaluate magnetic moments in the $`\chi `$QM<sub>gcm</sub>. In this context, it has been shown recently by Cheng and Li that $`\chi `$QM, incorporating the symmetry breaking effects, leads to formula which is similar to CQM, for example,
$$\mu (B)=(1+\kappa _{spin}+\kappa _{orbit})\mu (B)_v,$$
(19)
where $`(1+\kappa _{spin}+\kappa _{orbit})`$ is the Cheng and Li scale factor and $`\mu (B)_v`$ is the magnetic moment in the CQM. The Cheng and Li scale factor can be absorbed in the quark masses, thus magnetic moments calculated in $`\chi `$QM essentially reduce to using appropriate ‘quark mass scales’ for fitting the magnetic moments.
In Table 2, we have presented the magnetic moments with the wavefunctions expressed through equation (6), in terms of $`\mu _{56}`$ and $`\mu _{70}`$,
$$\mu (N)=cos^2\varphi <56|M|56>+sin^2\varphi <70|M|70>$$
$$=cos^2\varphi (\mu _{56})+sin^2\varphi (\mu _{70}),$$
where $`M`$ is the magnetic moment operator. It is evident from the table that for a good range of $`\varphi `$ we can fit the magnetic moments of the nucleons. As has been shown , this not only reproduces nucleon magnetic moments but is also able to give a good fit to other magnetic moments . Further, the magnetic moments are independent of the sign of mixing angle $`\varphi `$.
After having realized that the $`\chi `$QM<sub>gcm</sub> can explain $`G_A/G_V`$ and the nuclear magnetic moments, we have tried to investigate whether it is able to reproduce the neutron charge radius $`<r_n^2>`$. As it has been emphasized earlier that the Foldy term, reproducible within $`\chi `$QM, gets cancelled against a contribution to the Dirac form factor , therefore to calculate the neutron charge radius we have effectively replaced $`G_E^n(q^2)_{\chi QM}G_E^n(q^2)_{QM}.`$ In the CQM similar calculations have been done . In order to emphasize its dependence on mixing angle $`\varphi `$ we reproduce here some of the essential details .
The neutron charge radius is usually expressed in terms of the slope of the electric form factor $`G_E^n(q^2)`$, the experimental value of which is given as
$$(\frac{dG_E^n(q^2)}{d|q^2|})_{q^2=0}=0.47\pm 0.01GeV^2.$$
(20)
If we assign the nucleon to a pure 56 (with the spin expressed in terms of Pauli spinors ), the neutron electric form factor vanishes for all $`q^2`$. Considering our complete wavefunction with the $`5670`$ mixing the neutron charge radius can be expressed as
$$<r_n^2>=6(\frac{dG_E^n(q^2)}{d|q^2|})_{q^2=0}=\sqrt{\frac{2}{3}}R^2(tan\varphi ),$$
(21)
where $`\varphi `$ is the mixing angle which is negative and $`R^2`$ is the shape factor for the harmonic oscillator wave function. We have given the calculated values of neutron charge radius $`<r_n^2>(=6b)`$ as function of $`\varphi `$ and $`R^2`$ in Table 2. From the table one finds that one is able to reproduce fairly well the experimental value of $`<r_n^2>`$, viz., 2.82 GeV<sup>-2</sup> in the $`\chi `$QM<sub>gcm</sub> by considering the same values of parameter $`\varphi `$ which reproduced the values of $`G_A/G_V`$ and nuclear magnetic moments.
In conclusion, we would like to mention that we have calculated the spin polarization functions $`(\mathrm{\Delta }u,\mathrm{\Delta }d`$ and $`\mathrm{\Delta }s)`$ in the $`\chi `$QM with SU(3) symmetry breaking as well as one gluon generated configuration mixing for the NMC and the most recent E866 data. Besides reproducing $`G_A/G_V`$ and magnetic moments, it can also accomodate the neutron charge radius, thus resolving the compatibility problem of these parameters which could neither be achieved in the CQM with configuration mixing nor in the $`\chi `$QM with and without SU(3) symmetry breaking.
ACKNOWLEDGMENTS
M.G. would like to thank L.M. Sehgal for some useful discussions as well as initiating his interest in the problem. H.D. would like to thank CSIR, Govt. of India, for financial support and the Chairman, Department of Physics, for providing facilities to work in the department.
|
warning/0003/hep-ph0003252.html
|
ar5iv
|
text
|
# Particle production in the oscillating inflation model
## I Introduction
Inflationary cosmology is one of the most reliable concepts to describe the early stage of the universe. This paradigm not only gives the solution to a number of shortcomings of the standard big bang cosmology, but also provides the density perturbation that may be responsible for the structure formation. During the inflationary stage, a scalar field $`\varphi `$ known as inflaton is slowly rolling down toward a minimum of its potential. Inflation ends when the kinetic energy of the inflaton becomes comparable to the potential energy. After that, the inflaton field begins to oscillate around the minimum of its potential and produces elementary particles. This process by which the energy of the inflaton field is transferred to other particles is called reheating. The original scenario of reheating was considered in Ref. based on the perturbation theory, adding the phenomenological decay term to the equation of the inflaton field. However, since the production of a grand unified theory (GUT) scale boson is kinematically forbidden in this model, the GUT scale baryogenesis does not work well in this scenario.
It was recently recognized that the reheating process begins by an explosive particle production called preheating . During this stage, the fluctuation of produced particles grows quasi-exponentially by parametric resonance. The efficiency of resonance depends on the model of the inflation. The chaotic inflation is one of the most efficient model for the development of the fluctuation. In the case of the massive inflaton potential $`V(\varphi )=m^2\varphi ^2/2`$, another scalar field $`\chi `$ coupled to inflaton with a coupling $`g^2\varphi ^2\chi ^2/2`$ can be enhanced in a certain range of the coupling constant. With the coupling of $`g\stackrel{>}{}10^4`$, resonance turns on from a broad resonance regime and the fluctuation of the $`\chi `$ particle increases overcoming the diluting effect by the cosmic expansion. In the massless inflaton potential $`\lambda \varphi ^4/4`$, the growth of the inflaton fluctuation occurs even if we do not introduce another field $`\chi `$ coupled to inflaton . In this case, however, since the resonance band is restricted to be narrow, the transfer of energy from the homogeneous inflaton to the fluctuation does not occur sufficiently. In the two-field model of $`V(\varphi ,\chi )=\lambda \varphi ^4/4+g^2\varphi ^2\chi ^2/2`$, the $`\chi `$ particle production is typically more efficient than the production of the $`\varphi `$ particle. As for other models of inflation, several authors considered preheating such as the hybrid inflation, and the higher curvature inflation.
Recently it has been recognized by Damour and Mukhanov that inflation occurs even in the oscillating stage for the non-convex type potential $`V(\varphi )`$ where $`d^2V/d\varphi ^2`$ is negative in the regime not too far from the core of the potential. In spite of the rapid oscillation of the inflaton field, inflation takes place during which the field moves in flat regions outside the core part. Liddle and Mazumdar called it oscillating inflation and numerically calculated the revised number of $`e`$-foldings. The amount of inflation becomes larger with the decrease of the critical value $`\varphi _c`$ which determines the scale of the convex core of the potential. However, it is difficult to induce a sufficient inflation only by the oscillating inflation even when $`\varphi _c`$ is lowered to the electro-weak scale $`\varphi _c=10^{17}M_{\mathrm{pl}}`$. Since the ordinary inflation takes place in the slow-roll regime before the oscillating inflation, the total amount of $`e`$-folding is mostly due to this preceding inflation and the contribution of the oscillating inflation is added to some extent. The energy scale of the potential can be determined in the usual manner by fitting the density perturbation observed by the Cosmic Background Explorer (COBE) satellite.
As was pointed out by Damour and Mukhanov, since the square of the effective mass of the inflaton field $`m_\varphi ^2d^2V/d\varphi ^2`$ is mostly negative during the oscillating inflation, the fluctuation of inflaton can be strongly amplified. Taruya considered the evolution of the fluctuation including the metric perturbation in the single field model during the oscillating inflation. Although the super-horizon modes ($`k0`$) are not relevantly amplified, the growth of the modes inside the Hubble horizon occurs significantly. In the case of the single field, since the curvature perturbation on the comoving slice remains constant in the long wavelength limit, we can not expect the significant growth of the long-wave perturbation. It was also suggested by Damour and Mukhanov that another field $`\chi `$ coupled to inflaton would be resonantly amplified during the oscillating inflation and the superheavy GUT scale bosons are expected to be generated. The structure of resonance for the $`\chi `$ particle is different from that of the $`\varphi `$ particle, and we should make clear the parameter range of the coupling constant $`g`$ where the $`\chi `$ particle production occurs sufficiently. Since the effective mass of the inflaton around the minimum of the potential depends on the critical value $`\varphi _c`$, it is also important to investigate how the difference of the shape of the potential would affect the $`\chi `$ particle production. If the $`\chi `$ particle is efficiently created in the oscillating inflation model, this would affect a nonthermal phase transition, and topological defect formation. Also, the possibility of the production of the superheavy particle would make the baryogenesis at the GUT scale possible. In this paper, we consider the production of the $`\chi `$ particle as well as the $`\varphi `$ particle during the oscillating inflation and the subsequent oscillating stage. We make use of the Hartree approximation in order to include the back reaction effect of created particles which is absent in Ref. . Even in the case where the resonance band is very broad, the back reaction effect plays a crucial role to terminate the growth of the fluctuation. We will discuss how the maximum fluctuations of $`\varphi `$ and $`\chi `$ particles depend on the critical value of $`\varphi _c`$ and the coupling constant $`g`$.
This paper is organized as follows. In the next section, we introduce basic equations based on the Hartree approximation in the oscillating inflation model. In Sec. III, the particle production in this model is investigated by making use of the numerical results. We study how the fluctuations of the $`\varphi `$ and $`\chi `$ fields grow during the oscillating inflation and the subsequent oscillating stage. We give our discussions and conclusions in the final section.
## II The basic equations
We investigate a model where an inflaton field $`\varphi `$ is coupled to a scalar field $`\chi `$,
$`=\sqrt{g}\left[{\displaystyle \frac{1}{2\kappa ^2}}R{\displaystyle \frac{1}{2}}(\varphi )^2V(\varphi ){\displaystyle \frac{1}{2}}(\chi )^2{\displaystyle \frac{1}{2}}m_\chi ^2\chi ^2{\displaystyle \frac{1}{2}}g^2\varphi ^2\chi ^2\right],`$ (1)
where $`\kappa ^2/8\pi G=M_{\mathrm{pl}}^2`$ is Newton’s gravitational constant, $`R`$ is a scalar curvature, $`m_\chi `$ is a mass of the $`\chi `$ field, and $`g`$ is a coupling constant. We consider an inflaton potential $`V(\varphi )`$ proposed by Damour and Mukhanov, which is described by
$`V(\varphi )={\displaystyle \frac{M^4}{q}}\left[\left({\displaystyle \frac{\varphi ^2}{\varphi _c^2}}+1\right)^{q/2}1\right],`$ (2)
where $`M`$ is a mass which is constrained by the primordial density perturbation observed by the COBE satellite, $`q`$ is a dimensionless parameter greater than zero, and $`\varphi _c`$ is a critical value of the inflaton field which determines the scale of the core part of the potential (See Fig. 1). Note that the potential $`(\text{2})`$ is a toy model which is not directly related with a real theory of physics, although supergravity and superstring models may give rise to some non-convex type potentials. In the case of $`\varphi ^2\varphi _c^2`$, the potential is approximately written by $`V(\varphi )Aq^1(\varphi /\varphi _c)^q`$, and $`d^2V/d\varphi ^2`$ becomes negative for $`q<1`$. Inflation takes place in the usual manner while inflaton slowly moves in this flat region. This conventional slow-roll inflation is followed by the oscillating inflation, which means that inflation continues to occur during the oscillating stage of inflaton while it evolves in the region of $`\varphi ^2\stackrel{>}{}\varphi _c^2`$. Since the amplitude of the $`\varphi `$ field gradually decreases due to the adiabatic expansion of the universe, the $`\varphi `$ field is finally trapped in the core region of the potential (i.e. $`\varphi ^2\stackrel{<}{}\varphi _c^2`$) and the oscillating inflation ceases. After that, the universe enters the ordinary reheating stage, in which the potential $`(\text{2})`$ is described by the massive inflaton potential.
Hereafter, we study the dynamics of the system in the flat Friedmann-Robertson-Walker metric
$`ds^2=dt^2+a^2(t)d𝐱^2,`$ (3)
where $`a(t)`$ is the scale factor, and $`t`$ is the cosmic time coordinate.
Let us consider the equations of motion based on the Hartree factorization in the oscillating inflation model. We decompose the inflaton field into the homogeneous and fluctuational parts as
$`\varphi (t,𝐱)=\varphi _0(t)+\delta \varphi (t,𝐱),`$ (4)
where the fluctuational part satisfies the tadpole condition
$`\delta \varphi (t,𝐱)=0.`$ (5)
In order to study the quantum particle creation, we expand $`\delta \varphi `$ and $`\chi `$ fields as
$`\delta \varphi ={\displaystyle \frac{1}{(2\pi )^{3/2}}}{\displaystyle \left(a_k\delta \varphi _k(t)e^{i𝐤𝐱}+a_k^{}\delta \varphi _k^{}(t)e^{i𝐤𝐱}\right)d^3𝐤},`$ (6)
$`\chi ={\displaystyle \frac{1}{(2\pi )^{3/2}}}{\displaystyle \left(a_k\chi _k(t)e^{i𝐤𝐱}+a_k^{}\chi _k^{}(t)e^{i𝐤𝐱}\right)d^3𝐤},`$ (7)
where $`a_k`$ and $`a_k^{}`$ are the annihilation and creation operators respectively.
Then, the equations of motion for inflaton are expressed by imposing the Hartree factorization as
$`\ddot{\varphi }_0+3H\dot{\varphi }_0+{\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{1}{2^nn!}}\delta \varphi ^2^nV^{(2n+1)}(\varphi _0)+g^2\chi ^2\varphi _0=0,`$ (8)
$`\delta \ddot{\varphi }_k+3H\delta \dot{\varphi }_k+\left[{\displaystyle \frac{k^2}{a^2}}+{\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{1}{2^nn!}}\delta \varphi ^2^nV^{(2n+2)}(\varphi _0)+g^2\chi ^2\right]\delta \varphi _k=0,`$ (9)
where $`H\dot{a}/a`$, $`V^n(\varphi _0)\delta ^nV(\varphi _0)/\delta \varphi _0^n`$, and $`\delta \varphi ^2`$, $`\chi ^2`$ are defined by
$`\delta \varphi ^2={\displaystyle \frac{1}{2\pi ^2}}{\displaystyle k^2|\delta \varphi _k|^2𝑑k},`$ (10)
$`\chi ^2={\displaystyle \frac{1}{2\pi ^2}}{\displaystyle k^2|\chi _k|^2𝑑k}.`$ (11)
Note that the $`V^{(2)}(\varphi _0)`$ term in Eq. $`(\text{9})`$ is the leading term for the development of the fluctuation $`\delta \varphi ^2`$. In the present model, this term is negative in most stages of the oscillating inflation, and the growth of the fluctuation can be expected during this stage. As $`\varphi `$ particles are produced, however, the back reaction effect by the growth of $`\delta \varphi ^2`$ plays an important role. The development of the fluctuation is suppressed by the $`\varphi `$ particle production itself.
With regard to the $`\chi `$ field, this satisfies the following equation
$`\ddot{\chi }_k+3H\dot{\chi }_k+\left[{\displaystyle \frac{k^2}{a^2}}+m_\chi ^2+g^2(\varphi _0^2+\delta \varphi ^2)\right]\chi _k=0.`$ (12)
The $`g^2\varphi _0^2`$ term leads to the parametric amplification of $`\chi `$ particles when the $`\varphi _0`$ field oscillates around the minimum of its potential. The strength of resonance depends on the coupling constant $`g`$ and the amplitude of the $`\varphi _0`$ field. Moreover, the effective mass $`m_\varphi `$ of the $`\varphi _0`$ field is also important as we will show later. Since $`m_\varphi `$ is closely related to the shape around the core region, the development of the fluctuation $`\chi ^2`$ depends on the critical value $`\varphi _c`$. In the case where the growth rate of the $`\chi `$ fluctuation is large, back reaction effects of $`\chi `$ particles as well as $`\varphi `$ particles finally shut off the parametric resonance.
Next, the evolution of the scale factor is described by
$`\left({\displaystyle \frac{\dot{a}}{a}}\right)^2`$ $`=`$ $`{\displaystyle \frac{\kappa ^2}{3}}[{\displaystyle \frac{1}{2}}\dot{\varphi _0}^2+{\displaystyle \frac{1}{2}}\delta \dot{\varphi }^2+{\displaystyle \frac{1}{2a^2}}(\varphi )^2+{\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{1}{2^nn!}}\delta \varphi ^2^nV^{2n}(\varphi _0)`$ (13)
$`+`$ $`{\displaystyle \frac{1}{2}}\dot{\chi }^2+{\displaystyle \frac{1}{2a^2}}(\chi )^2+{\displaystyle \frac{1}{2}}\{m_\chi ^2+g^2(\varphi _0^2+\delta \varphi ^2)\}\chi ^2],`$ (14)
where
$`\delta \dot{\varphi }^2={\displaystyle \frac{1}{2\pi ^2}}{\displaystyle k^2|\delta \dot{\varphi }_k|^2𝑑k},`$ (15)
$`(\varphi )^2={\displaystyle \frac{1}{2\pi ^2}}{\displaystyle k^4|\delta \varphi _k|^2𝑑k},`$ (16)
$`(\chi )^2={\displaystyle \frac{1}{2\pi ^2}}{\displaystyle k^4|\chi _k|^2𝑑k}.`$ (17)
Before analyzing the above equation of motion, we first discuss the evolution of the scale factor and the homogeneous inflaton field neglecting the fluctuational terms. Then, Eqs. $`(\text{8})`$ and $`(\text{14})`$ are approximately written as
$`\ddot{\varphi }_0+3H\dot{\varphi }_0+V,_{\varphi _0}(\varphi _0)0,`$ (18)
$`\left({\displaystyle \frac{\dot{a}}{a}}\right)^2{\displaystyle \frac{\kappa ^2}{3}}\left[{\displaystyle \frac{1}{2}}\dot{\varphi _0}^2+V(\varphi _0)\right].`$ (19)
We can understand the mean behavior of the scale factor and the inflaton field by ignoring the core region of the potential. For the case of $`\varphi _0^2\varphi _c^2`$, making use of the time averaged relation $`\dot{\varphi }^2_T=qV(\varphi )_T`$ in Eqs. $`(\text{18})`$ and $`(\text{19})`$, we find the following approximate relation
$`at^{(q+2)/3q},`$ (20)
$`\stackrel{~}{\varphi }_0t^{2/q},`$ (21)
where $`\stackrel{~}{\varphi }_0`$ is the amplitude of the $`\varphi _0`$ field. Note that the universe expands acceleratedly for $`0<q<1`$. Hereafter, we consider the case of $`0<q<1`$. Since inflation takes place for $`\varphi _0^2\stackrel{>}{}\varphi _c^2`$, the amount of inflation becomes larger with the decrease of $`\varphi _c`$. Damour and Mukhanov estimated the number of $`e`$-foldings as
$`N\mathrm{ln}{\displaystyle \frac{a_f}{a_s}}{\displaystyle \frac{2+q}{6}}\mathrm{ln}{\displaystyle \frac{\varphi _0(t_s)}{\varphi _c}},`$ (22)
where the subscripts $`s`$ and $`f`$ denote the end of the slow-roll and the end of the oscillating inflation respectively. The slow-roll inflation ends when the slow-roll parameter $`ϵ(V,_{\varphi _0}/V)^2/2\kappa ^2`$ becomes of order unity. For the case of $`\varphi _0^2\varphi _c^2`$, since $`ϵ`$ can be written as
$`ϵ{\displaystyle \frac{q^2M_{\mathrm{pl}}^2}{16\pi \varphi _0^2}},`$ (23)
$`\varphi _0(t_s)`$ is estimated by setting $`ϵ=1`$ as
$`\varphi _0(t_s){\displaystyle \frac{q}{\sqrt{16\pi }}}M_{\mathrm{pl}}.`$ (24)
We find from Eqs. $`(\text{22})`$ and $`(\text{24})`$ that the number of $`e`$-foldings during the oscillating inflation is small compared with the total needed amount of inflation $`N_t\stackrel{>}{}60`$. For example, when $`\varphi _c=10^6M_{\mathrm{pl}}`$ and $`q=0.1`$, $`N=3.3`$. Even if $`\varphi _c`$ is the electro-weak scale $`\varphi _c=10^{17}M_{\mathrm{pl}}`$, $`N=12.2`$ for $`q=0.1`$. As was confirmed by numerical calculations in Ref. , the number of $`e`$-foldings becomes smaller as $`q`$ approaches zero or unity for the fixed value of $`\varphi _c`$. Hence even if we change the values of $`q`$ in the range of $`0<q<1`$, the contribution to the amount of inflation due to the oscillating inflation is still small. This means that the preceding inflation by the ordinary slow-roll is expected to contribute to most of the number of $`e`$-foldings needed to solve cosmological puzzles.
Let us estimate the energy scale of the potential $`(\text{2})`$ which is constrained by the density perturbation observed by the COBE satellite. First, we consider the value of $`\varphi _0`$ ($`=\varphi _0(t_i)`$) at the epoch of horizon exit when physical scales crossed outside the Hubble radius 50 $`e`$-foldings before the start of the oscillating inflation. Calculating the number of $`e`$-foldings
$`N=\kappa ^2{\displaystyle _{\varphi _0(t_i)}^{\varphi _0(t_s)}}{\displaystyle \frac{V}{V,_{\varphi _0}}}𝑑\varphi _0,`$ (25)
in the present model with the condition of $`\varphi _0^2\varphi _c^2`$, we obtain
$`N{\displaystyle \frac{4\pi }{qM_{\mathrm{pl}}^2}}\left[\varphi _0^2(t_i)\varphi _0^2(t_s)\right].`$ (26)
Then the value of $`\varphi _0(t_i)`$ is approximately estimated as
$`\varphi _0(t_i)\sqrt{{\displaystyle \frac{qN}{4\pi }}}M_{\mathrm{pl}},`$ (27)
where we neglected the contribution of the $`\varphi _0(t_s)`$ term. The square of the amplitude of the density perturbation can be calculated for $`\varphi _0^2\varphi _c^2`$ as
$`\delta _H^2={\displaystyle \frac{32}{75}}{\displaystyle \frac{V(\varphi _0(t_i))}{M_{\mathrm{pl}}^4}}ϵ^1(\varphi _0(t_i)){\displaystyle \frac{512\pi }{75}}{\displaystyle \frac{M^4}{q^3M_{\mathrm{pl}}^6}}{\displaystyle \frac{\varphi _0^{q+2}(t_i)}{\varphi _c^q}}.`$ (28)
Then the mass $`M`$ is constrained to be
$`M\left[{\displaystyle \frac{75}{128}}{\displaystyle \frac{q^2}{N}}\left({\displaystyle \frac{\varphi _c}{M_{\mathrm{pl}}}}\sqrt{{\displaystyle \frac{4\pi }{qN}}}\right)^q\delta _H^2\right]^{1/4}M_{\mathrm{pl}}.`$ (29)
The COBE data requires $`\delta _H2\times 10^5`$. Setting $`N=50`$, we can estimate the mass $`M`$ as two functions of $`q`$ and $`\varphi _c`$. The mass $`M`$ is weakly dependent on the value of $`\varphi _c`$ for the fixed value of $`q`$. For example, for $`\varphi _c=10^4M_{\mathrm{pl}}`$ and $`q=0.1`$, $`M=3.74\times 10^4M_{\mathrm{pl}}`$; and for $`\varphi _c=10^6M_{\mathrm{pl}}`$ and $`q=0.1`$, $`M=3.33\times 10^4M_{\mathrm{pl}}`$. In the next section, we use the value of $`M`$ which is obtained by Eq. $`(\text{29})`$, and consider the growth of the fluctuation during and after the oscillating inflation.
## III Particle production in the oscillating inflation model
In this section, we investigate the particle production of $`\varphi `$ and $`\chi `$ fields with the potential $`(\text{2})`$. Defining new scalar fields $`\phi _0a^{3/2}\varphi _0`$, $`\delta \phi _ka^{3/2}\delta \varphi _k`$, and $`X_ka^{3/2}\chi _k`$, Eqs. $`(\text{8})`$, $`(\text{9})`$, and $`(\text{12})`$ can be rewritten as
$`\ddot{\phi }_0+\omega _{\phi _0}^2\phi _0=0,`$ (30)
$`\delta \ddot{\phi }_k+\omega _{\delta \phi _k}^2\delta \phi _k=0,`$ (31)
$`\ddot{X}_k+\omega _{X_k}^2X_k=0,`$ (32)
with
$`\omega _{\phi _0}^2`$ $``$ $`{\displaystyle \frac{M^4}{\varphi _c^2}}\left({\displaystyle \frac{\varphi _0^2}{\varphi _c^2}}+1\right)^{\frac{1}{2}q1}+{\displaystyle \frac{1}{2}}{\displaystyle \frac{M^4}{\varphi _c^4}}\delta \varphi ^2\left({\displaystyle \frac{\varphi _0^2}{\varphi _c^2}}+1\right)^{\frac{1}{2}q3}(q2)\left\{(q1){\displaystyle \frac{\varphi _0^2}{\varphi _c^2}}+3\right\}`$ (33)
$`+`$ $`g^2\chi ^2{\displaystyle \frac{3}{4}}\left({\displaystyle \frac{2\ddot{a}}{a}}+{\displaystyle \frac{\dot{a}^2}{a^2}}\right),`$ (34)
$`\omega _{\delta \phi _k}^2`$ $``$ $`{\displaystyle \frac{k^2}{a^2}}+{\displaystyle \frac{M^4}{\varphi _c^2}}\left({\displaystyle \frac{\varphi _0^2}{\varphi _c^2}}+1\right)^{\frac{1}{2}q2}\left\{(q1){\displaystyle \frac{\varphi _0^2}{\varphi _c^2}}+1\right\}+{\displaystyle \frac{1}{2}}\delta \varphi ^2{\displaystyle \frac{M^4}{\varphi _c^4}}\left({\displaystyle \frac{\varphi _0^2}{\varphi _c^2}}+1\right)^{\frac{1}{2}q4}`$ (35)
$`\times `$ $`(q2)\left\{(q1)(q3){\displaystyle \frac{\varphi _0^4}{\varphi _c^4}}+6(q3){\displaystyle \frac{\varphi _0^2}{\varphi _c^2}}+3\right\}+g^2\chi ^2{\displaystyle \frac{3}{4}}\left({\displaystyle \frac{2\ddot{a}}{a}}+{\displaystyle \frac{\dot{a}^2}{a^2}}\right),`$ (36)
$`\omega _{X_k}^2{\displaystyle \frac{k^2}{a^2}}+m_\chi ^2+g^2(\varphi _0^2+\delta \varphi ^2){\displaystyle \frac{3}{4}}\left({\displaystyle \frac{2\ddot{a}}{a}}+{\displaystyle \frac{\dot{a}^2}{a^2}}\right),`$ (37)
where we have considered the $`\delta \varphi ^2^n`$ terms up to $`n=1`$. The higher order terms do not significantly contribute to the evolution of the system. We find from Eq. $`(\text{34})`$ that the frequency of the $`\phi _0`$ field changes even during one oscillation of the $`\phi _0`$ field. At the stage when the particles are not sufficiently produced ($`\delta \varphi ^2,\chi ^2\stackrel{~}{\varphi }_0^2`$), the second and third terms in Eq. $`(\text{34})`$ are negligible. Moreover, since the last term in Eq. $`(\text{34})`$ is estimated by using Eqs. $`(\text{18})`$ and $`(\text{19})`$ as
$`{\displaystyle \frac{3}{4}}\left({\displaystyle \frac{2\ddot{a}}{a}}+{\displaystyle \frac{\dot{a}^2}{a^2}}\right){\displaystyle \frac{3}{4}}\kappa ^2\left[{\displaystyle \frac{1}{2}}\dot{\varphi _0}^2V(\varphi _0)\right],`$ (38)
this term is also negligible compared with the first term in Eq. $`(\text{34})`$ for the typical values of $`\varphi _c`$ ($`\stackrel{<}{}10^3M_{\mathrm{pl}}`$). Then the effective mass of the inflaton in the first stage of the oscillating inflation for two limiting cases $`\varphi _0^2\varphi _c^2`$ and $`\varphi _0^2\varphi _c^2`$ is given by
$`m_\varphi =\{\begin{array}{cc}\frac{M^2}{\varphi _c}\left(\frac{\varphi _c}{\varphi _0}\right)^{1\frac{q}{2}}\hfill & \text{(}\varphi _0^2\varphi _c^2\text{)}\hfill \\ \frac{M^2}{\varphi _c}\hfill & \text{(}\varphi _0^2\varphi _c^2\text{)}\hfill \end{array}`$ (39)
Note that $`m_\varphi `$ for $`\varphi _0^2\varphi _c^2`$ is smaller than $`m_\varphi `$ for $`\varphi _0^2\varphi _c^2`$ because the potential is flat for $`\varphi _0^2\varphi _c^2`$. As time passes and particles are produced, back reaction effects due to the increase of $`\delta \varphi ^2`$ and $`\chi ^2`$ become relevant and the coherent oscillation of the $`\phi _0`$ field is prevented by the change of the effective mass $`m_\varphi `$.
In the frequency $`(\text{36})`$ of the $`\delta \phi _k`$ field, since we consider the situation of $`0<q<1`$, the second term becomes negative in the region of $`\varphi _0^2\stackrel{>}{}\varphi _c^2`$. Moreover, in the small vicinity of $`\varphi _0^2\stackrel{<}{}\varphi _c^2`$, this term rapidly grows and takes large positive value $`M^4/\varphi _c^2`$. These peculiar behaviors of the frequency lead to the enhancement of the $`\varphi `$ particle during the stage of the oscillating inflation.
As for the $`\chi `$ field, since the oscillation of the $`\varphi _0`$ field is different from that of the ordinary inflaton potential, the increase of $`\chi ^2`$ occurs in a different way. In this case, although analytic investigations of the evolution of the $`\chi `$ particle are rather difficult, we shall examine the parameter range of the critical value $`\varphi _c`$ and the coupling constant $`g`$ where the $`\chi `$ particle production takes place efficiently.
With regard to the initial conditions of the fluctuation, we choose the states which correspond to the conformal vacuum as
$`\delta \phi _k(0)={\displaystyle \frac{1}{\sqrt{2\omega _{\delta \phi _k(0)}}}},\delta \dot{\phi }_k(0)=i\omega _{\delta \phi _k(0)}\delta \phi _k(0),`$ (40)
$`X_k(0)={\displaystyle \frac{1}{\sqrt{2\omega _{X_k(0)}}}},\dot{X}_k(0)=i\omega _{X_k(0)}X_k(0),`$ (41)
We investigate the evolution of the fluctuations of $`\varphi `$ and $`\chi `$ fields with those initial conditions as the semiclassical problem. In order to clarify the situation we consider, we first investigate the case when the coupling $`g`$ is zero (i.e. one-field case) and next investigate the case of $`g0`$ (two-field case).
### A The case of $`g=0`$
In this subsection, we study the evolution of the inflaton quanta in the case of $`g=0`$. This case was originally considered by Taruya including the metric perturbation. However, since his results are obtained neglecting the back reaction effect of created particles, we investigate how this effect would modify the growth of the fluctuation. The first thing we should notice is that the strength of the amplification of the fluctuation $`\delta \varphi ^2`$ strongly depends on the critical value $`\varphi _c`$. With the decrease of $`\varphi _c`$, since the duration of the oscillating inflation becomes longer, parametric amplification of the $`\varphi `$ particle occurs more efficiently before the back reaction effect becomes significant.
We investigate two concrete cases of $`\varphi _c=10^3M_{\mathrm{pl}}`$ and $`\varphi _c=10^4M_{\mathrm{pl}}`$ with the value of $`q=0.1`$ where the growth of the fluctuation can be expected. When $`q=0.1`$, the initial value of $`\varphi _0`$ is estimated as $`\varphi _0=1.4\times 10^2M_{\mathrm{pl}}`$ by Eq. $`(\text{24})`$.
Let us first consider the evolution of the $`\varphi _0`$ field. In the case of $`\varphi _c=10^3M_{\mathrm{pl}}`$, since the initial value of $`\varphi _0`$ is larger than $`\varphi _c`$ only by one order of magnitude, the number of $`e`$-foldings during the oscillating inflation is small \[$`N=0.93`$ by Eq. $`(\text{22})`$\]. In Fig. 2, we depict the evolution of the $`\varphi _0`$ field. The amplitude $`\stackrel{~}{\varphi }_0`$ gradually decreases due to the expansion of the universe, and the oscillating inflation ends when $`\stackrel{~}{\varphi }_0`$ is lowered to $`\varphi _c`$. Numerical calculations indicate that this occurs at $`\overline{t}M^2t/M_{\mathrm{pl}}0.37`$. After this, the $`\varphi _0`$ field oscillates around the core region $`\varphi _0^2\stackrel{<}{}\varphi _c^2`$, whose bare mass is given by $`m_\varphi =M^2/\varphi _c`$. In the case of $`\varphi _c=10^4M_{\mathrm{pl}}`$, since the oscillating inflation continues until $`\stackrel{~}{\varphi }_010^4M_{\mathrm{pl}}`$, we obtain the larger value of $`e`$-foldings $`N=1.73`$ compared with the case of $`\varphi _c=10^3M_{\mathrm{pl}}`$. Numerically, the oscillating inflation ceases at $`\overline{t}0.45`$ in this case.
With regard to the fluctuation of inflaton, the resonance band is very broad and the growth rate of the fluctuation becomes of order unity. We can expect the enormous amplification of the wave modes especially within the Hubble radius. We show in Fig. 3 the evolution of the $`\delta \varphi _k`$ field for two cases of $`\overline{k}k/(M^2/M_{\mathrm{pl}})=0.1`$ and $`\overline{k}=100`$. We find that long-wave modes are not significantly enhanced compared with the modes inside the horizon scale. This result coincides with numerical calculations performed in Ref. which include the metric perturbation. As was presented in Ref. , the effect of the metric perturbation appears in the term $`2\kappa ^2(V/H)^\dot{}`$ of the equation of the Mukhanov variable $`Q\delta \varphi \dot{\varphi }/H`$, where $``$ is the spatial curvature \[See Eq. (8) in Ref. \]. However, since this term decreases faster than the term $`V^{(2)}(\varphi _0)`$ in Eq. $`(\text{9})`$, we expect that adding this term to the equation of the fluctuation will not alter the evolution of the system. We have numerically checked that the growth of the fluctuation in Fig. 3 is almost the same as in the case where the $`2\kappa ^2(V/H)^\dot{}`$ term is included. In the single-field case, there exists an exact solution for the Mukhanov variable in the long wavelength limit \[See Eq. (9) in Ref. \], and we can confirm that the amplitude of the fluctuation remains nearly constant. On the other hand, since the instability band is broad in the present model even for the large $`k`$, the momentum modes up to $`k^2/a^2\stackrel{<}{}M^4/\varphi _{c}^{}{}_{}{}^{2}`$ inside the Hubble horizon are effectively enhanced. In Fig. 3, we find that the fluctuation $`\delta \varphi _k`$ rapidly grows at the stage of the oscillating inflation for the case of $`\overline{k}=100`$. However, this increase is suppressed at $`\overline{t}0.12`$, before the oscillating inflation terminates at $`\overline{t}0.37`$. This is due to the fact that the back reaction effect of created particles becomes significant. We have numerically confirmed that the fluctuation continues to grow during the oscillating inflation if we neglect the back reaction effect as in Ref. . The increase of $`\delta \varphi ^2`$ effectively changes both frequencies of $`\phi _0`$ and $`\delta \phi _k`$ fields as is found by Eqs. $`(\text{34})`$ and $`(\text{36})`$, and the growth of the fluctuation finally stops when the $`\delta \varphi ^2`$ terms become comparable to the preceding terms in Eqs. $`(\text{34})`$ and $`(\text{36})`$. It is expected that the back reaction effect becomes significant when $`\delta \varphi ^2`$ increases of order $`\varphi _c^2`$. In the case of $`\varphi _c=10^3M_{\mathrm{pl}}`$, numerical calculations show that the maximum value of the fluctuation is $`\delta \varphi ^2_f10^7M_{\mathrm{pl}}^2`$ at $`\overline{t}0.12`$ (See Fig. 4). In this case, parametric amplification of the $`\varphi `$ particle terminates when $`\delta \varphi ^2`$ increases up to $`\delta \varphi ^2_f0.1\varphi _c^2`$. As is seen in Fig. 2, the coherent oscillation of the $`\varphi _0`$ field is hardly broken after $`\overline{t}0.12`$ where the variance of the $`\varphi `$ particle reaches the maximum value. This means that the growth of $`\delta \varphi ^2`$ affects the evolution of the $`\delta \varphi _k`$ field more strongly than that of the $`\varphi _0`$ field.
With the decrease of $`\varphi _c`$, since the oscillating inflation continues longer, the growth rate of the fluctuation in the first stage becomes larger. However, the final variance $`\delta \varphi ^2_f`$ is suppressed as $`\varphi _c`$ becomes smaller. We depict in Fig. 5 the evolution of the fluctuation $`\delta \varphi ^2`$ in the case of $`\varphi _c=10^4M_{\mathrm{pl}}`$. In this case, although the oscillating inflation continues until $`\overline{t}0.4`$, the growth of the fluctuation stops much earlier: $`\overline{t}0.05`$. Although the initial growth rate is larger than in the case of $`\varphi _c=10^3M_{\mathrm{pl}}`$, the production of the $`\varphi `$ particle itself terminates the growth of the fluctuation. This tendency is stronger with the decrease of $`\varphi _c`$, and the maximum fluctuation becomes smaller. We have found the following relation for the typical value of $`\varphi _c\stackrel{<}{}10^3M_{\mathrm{pl}}`$ as
$`\delta \varphi ^2_f0.1\varphi _c^2.`$ (42)
Although the $`\varphi `$ particle production is possible at the initial stage of the oscillating inflation, the final variance is strongly suppressed with the decrease of $`\varphi _c`$ as Eq. $`(\text{42})`$. After the amplitude of the inflaton field drops under $`\stackrel{~}{\varphi }_0\stackrel{<}{}\varphi _c`$, the system enters the ordinary reheating stage where the universe expands deceleratedly. In this stage, the growth of the fluctuation $`\delta \varphi ^2`$ can be no longer expected because the inflaton field behaves as the massive inflaton whose mass is $`m_\varphi =M^2/\varphi _c^2`$.
In the next subsection, we consider another field $`\chi `$ coupled to inflaton and analyze how $`\chi `$ particles are produced during the oscillating inflation and subsequent oscillating phase by parametric resonance.
### B The case of $`g0`$
Next, we consider the production of the $`\chi `$ particle coupled to the inflaton field. In the ordinary picture of the slow-roll inflation, the $`\chi `$ particle production is inefficient during the inflationary phase. In the present model, however, since the inflaton field oscillates rapidly during the oscillating inflation, $`\chi `$ particles can be produced in this stage by parametric resonance.
The strength of resonance depends on the coupling $`g`$, the amplitude of $`\varphi _0`$ $`(=\stackrel{~}{\varphi }_0)`$, and the mass of the inflaton field $`(=m_\varphi )`$. In the model of the scalar field $`\chi `$ coupled to the massive inflaton $`\varphi `$: $`V(\varphi ,\chi )=m^2\varphi ^2/2+g^2\varphi ^2\chi ^2/2`$, the equation of the $`\chi `$ field is reduced to the Mathieu equation at the linear stage. In this model, the resonance parameter $`qg^2\stackrel{~}{\varphi }^2/(4m^2)`$ is an important factor in determining whether the $`\chi `$ particle production is efficient or not. When $`q`$ is sufficiently large initially as $`q_i1`$, parametric resonance turns on from broad resonance regimes. Although $`q`$ decreases with the decrease of $`\stackrel{~}{\varphi }_0`$ due to the cosmic expansion, the fluctuation $`\chi ^2`$ grows quasiexponentially until $`q`$ drops under unity or the back reaction effect of created particles becomes significant.
In the oscillating inflation model, since the mass $`m_\varphi ^2=\frac{M^4}{\varphi _c^2}\left(\frac{\varphi _0^2}{\varphi _c^2}+1\right)^{q/21}`$ continually changes during one oscillation of inflaton, the $`\phi _0`$ field does not oscillate sinusoidally and the method based on the Mathieu equation is not valid except when $`\stackrel{~}{\varphi }_0`$ drops under $`\varphi _c`$. In the frequency $`(\text{37})`$ of the $`X_k`$ field, since the last term is negligible when the $`\chi `$ particle production occurs sufficiently, the equation of the $`\chi `$ field for the case of $`\delta \varphi ^2\stackrel{~}{\varphi }_0^2`$ is approximately written by
$`{\displaystyle \frac{d^2}{dz^2}}X_k+\left[{\displaystyle \frac{\overline{k}^2}{a^2}}+\overline{m}_\chi ^2+{\displaystyle \frac{g^2\varphi _0^2}{m_\varphi ^2}}\right]X_k0,`$ (43)
where $`zm_\varphi t`$, $`\overline{k}k/m_\varphi `$, and $`\overline{m}_\chi m_\chi /m_\varphi `$. When the initial value of $`g^2\varphi _0^2/m_\varphi ^2`$ is large, parametric resonance works out more efficiently. The initial value of $`\varphi _0`$ estimated by Eq. $`(\text{24})`$ for the typical value of $`q`$ is smaller than the initial value $`\varphi _00.2`$-$`0.3M_{\mathrm{pl}}`$ in the massive inflaton model. Moreover, as is found by Eq. $`(\text{21})`$, the amplitude of the $`\varphi _0`$ field decays faster than in the model of the massive inflaton with $`\stackrel{~}{\varphi }_0t^1`$.
The effective mass of the $`\varphi _0`$ field is approximately written in two asymptotic regions as Eq. $`(\text{39})`$, and depends on the critical value $`\varphi _c`$. The mass $`m_\varphi M^2/\varphi _c^2`$ around $`\varphi _0^2\stackrel{<}{}\varphi _c^2`$ plays the more important role than that of $`\varphi _0^2\stackrel{>}{}\varphi _c^2`$ for the development of the fluctuation, because $`\chi `$ particles are mainly produced in the vicinity of $`\varphi _0=0`$ where the $`\varphi _0`$ field changes nonadiabatically. During one oscillation of the $`\varphi _0`$ field, $`m_\varphi `$ gradually gets larger as $`\varphi _0`$ approaches the minimum of the potential for the fixed value of $`\varphi _c`$. Moreover, for the typical value of $`\varphi _c\stackrel{<}{}10^3M_{\mathrm{pl}}`$ where the oscillating inflation occurs, $`m_\varphi =M^2/\varphi _c`$ is greater than the mass $`10^6M_{\mathrm{pl}}`$ in the model of the massive inflaton. For example, in the case of $`\varphi _c=10^3M_{\mathrm{pl}}`$ and $`q=0.1`$, $`M^2/\varphi _c=1.57\times 10^4M_{\mathrm{pl}}`$; for $`\varphi _c=10^4M_{\mathrm{pl}}`$ and $`q=0.1`$, $`M^2/\varphi _c=1.40\times 10^3M_{\mathrm{pl}}`$.
In addition to the fact that $`\stackrel{~}{\varphi }_0`$ decreases faster than in the model of the massive inflaton, the larger mass $`m_\varphi `$ results in the restriction of the coupling constant $`g`$ for an efficient particle production. Since the dependence of $`M^2`$ for the value $`\varphi _c`$ is weak, the mass $`M^2/\varphi _c`$ monotonically increases with the decrease of $`\varphi _c`$. This means that parametric resonance does not take place unless we choose a large coupling constant $`g`$ as $`\varphi _c`$ decreases. When $`\varphi _c`$ is fairly large as $`\varphi _c\stackrel{>}{}0.1M_{\mathrm{pl}}`$, $`M^2/\varphi _c`$ becomes comparable to the mass of the massive inflaton model, but in this case the creation of $`\varphi `$ particles can not be expected because of the absence of the oscillating inflationary phase. What we are interested in is the case of $`\varphi _c\stackrel{<}{}10^3M_{\mathrm{pl}}`$ where the oscillating inflation occurs and both of $`\varphi `$ and $`\chi `$ particles are generated. In what follows, we examine the $`\chi `$ particle production in two cases of $`\varphi _c=10^3M_{\mathrm{pl}}`$ and $`\varphi _c=10^4M_{\mathrm{pl}}`$ for the massless $`\chi `$ particle, and add the discussion of the case where the mass of the $`\chi `$ particle is included at the final of this section.
When $`\varphi _c=10^3M_{\mathrm{pl}}`$, numerical calculations indicate that the increase of $`\chi ^2`$ can take place for $`g\stackrel{>}{}5\times 10^3`$. However, parametric resonance is weak for the case of $`g\stackrel{<}{}0.03`$. As compared with the model of $`V(\varphi ,\chi )=m^2\varphi ^2/2+g^2\varphi ^2\chi ^2/2`$ in which the $`\chi `$ particle production takes place for $`g\stackrel{>}{}10^4`$, larger values of $`g`$ are required even in the case of $`\varphi _c=10^3M_{\mathrm{pl}}`$ where the oscillating inflation marginally occurs. Since the fluctuation $`\chi ^2`$ does not grow well for $`g\stackrel{<}{}0.03`$, the evolution of the $`\varphi _0`$ field is hardly affected by the $`\chi `$ particle production. For the case of $`g\stackrel{>}{}0.05`$, the back reaction effect of the created $`\chi `$ particle begins to be significant. We depict in Fig. 6 the evolution of $`\chi ^2`$ for two cases of $`g=0.03`$ and $`g=0.07`$. Although parametric resonance evidently occurs for $`g=0.03`$, this is rather inefficient and $`\chi ^2`$ takes the maximum value $`\chi ^2_f=2.3\times 10^{11}M_{\mathrm{pl}}^2`$ at $`\overline{t}=0.177`$. In this case, the $`\chi `$ field deviates from the resonance bands with the decrease of $`\stackrel{~}{\varphi }_0`$ due to the cosmic expansion before the oscillating inflation ceases at $`\overline{t}0.37`$. Since the maximum fluctuation of $`\delta \varphi ^2`$ is almost the same as the $`g=0`$ case, the back reaction effect of $`\varphi `$ particles onto the $`\varphi _0`$ field can be marginally negligible. Hence in the case of $`g\stackrel{<}{}0.03`$, the coherent oscillation of the $`\varphi _0`$ field is hardly prevented by the back reaction effects of both $`\varphi `$ and $`\chi `$ particles. When $`g=0.07`$, the evolution of the system shows different characteristics compared with the case of $`g\stackrel{<}{}0.03`$. In Fig. 7, the evolution of the $`\varphi _0`$ field for the $`g=0.07`$ case is depicted. We find that the coherent oscillation of the $`\varphi _0`$ field is broken for $`\overline{t}\stackrel{>}{}0.12`$, which is different from the $`g=0`$ case in Fig. 2. As for the $`\varphi `$ particle, the maximum fluctuation is $`\delta \varphi ^2_f=2.7\times 10^6M_{\mathrm{pl}}^2`$ at $`\overline{t}=0.12`$. This maximum variance is greater than in the case of $`g=0`$ by one order of magnitude (See Fig. 8). As was noticed in the previous subsection, the growth of $`\delta \varphi ^2`$ stops when the contribution of the third term in the frequency (3.5) becomes comparable to the second term. Since the third term mostly takes negative values, the increase of the $`g^2\chi ^2`$ term due to the $`\chi `$ particle production generally assists the $`\varphi `$ particle production to continue longer. This results in the larger value of the final fluctuation $`\delta \varphi ^2`$, and the back reaction effect due to the $`\varphi `$ particle production becomes relevant at $`\overline{t}0.12`$. This behavior is found in Fig. 7. However, since the $`\varphi _0`$ field still oscillates after the back reaction effect of the $`\varphi `$ particle becomes important, the $`\chi `$ particle continues to be enhanced after $`\overline{t}0.12`$. The fluctuation of $`\chi ^2`$ does not increase significantly in the initial stage of the oscillating inflation, because the expansion rate of the universe is large. $`\chi ^2`$ begins to increase rapidly after $`\overline{t}0.15`$, and reaches the maximum value $`\chi ^2_f=2.4\times 10^7M_{\mathrm{pl}}^2`$ at $`\overline{t}=0.28`$. The back reaction effect of the $`\chi `$ particles as well as the decrease of $`\stackrel{~}{\varphi }_0`$ terminates the parametric amplification of $`\chi `$ particles. When $`g=0.1`$, the increase of $`\chi ^2`$ is more significant and the final variance is $`\chi ^2_f=3.3\times 10^6M_{\mathrm{pl}}^2`$ . In this case, since the maximum variance is larger than in the case of $`g=0.07`$ by one order of magnitude, the coherent oscillation of the $`\varphi _0`$ field is more strongly prevented by the growth of $`\chi ^2`$. When $`g\stackrel{>}{}0.3`$, the final variance is suppressed because the back reaction effect becomes quite significant. We find that $`\chi ^2_f`$ takes the maximum value $`\chi ^2_{\mathrm{max}}10^5M_{\mathrm{pl}}^2`$ for $`g0.3`$ (See Fig. 9).
As for the momentum modes of the produced $`\chi `$ particle, we show the evolution of the $`\chi _k`$ field in two cases of $`\overline{k}=0`$ and $`\overline{k}=100`$ for $`g=0.1`$ in Fig. 10. Since the larger $`k`$ makes the $`\chi _k`$ field deviate from the resonance band in Eq. $`(\text{43})`$, higher momentum modes of the $`\chi `$ particle are not sufficiently produced. The low momentum modes mainly contribute to the growth of the fluctuation $`\chi ^2`$. This property is the same as in the model of $`V(\varphi ,\chi )=m^2\varphi ^2/2+g^2\varphi ^2\chi ^2/2`$. As was found in Ref. , there is a possibility that long wavelength modes of the metric perturbation are amplified considering the perturbed metric. This issue will be discussed elsewhere.
In the case of $`\varphi _c=10^4M_{\mathrm{pl}}`$, we need further large values of $`g`$ to yield an efficient resonance. Numerically, we find that the coupling is required $`g\stackrel{>}{}0.01`$ for the development of the $`\chi `$ fluctuation. Even when $`g=0.1`$, parametric resonance is not so efficient compared with the case of $`\varphi _c=10^3M_{\mathrm{pl}}`$. We depict in Fig. 11 the evolution of $`\chi ^2`$ for $`g=0.07,0.1,0.5`$ cases respectively. When $`g=0.07`$, the maximum value of the fluctuation is $`\chi ^2_f=1.2\times 10^{10}M_{\mathrm{pl}}^2`$, which is much smaller than in the case of $`\varphi _c=10^3M_{\mathrm{pl}}`$ with the same value of $`g`$. As the coupling increases further, we obtain larger values of the maximum fluctuation $`\chi ^2_f`$. However, for $`g\stackrel{>}{}0.5`$, the back reaction effect of created $`\chi `$ particles becomes significant and the final variance is suppressed. In the case of $`\varphi _c=10^4M_{\mathrm{pl}}`$ and $`g=0`$, the fluctuation of the $`\varphi `$ field rapidly grows in the initial stage and reaches the maximum value $`\delta \varphi ^2_f10^9M_{\mathrm{pl}}^2`$ at $`\overline{t}=0.05`$. Even if we take into account the interaction with the $`\chi `$ field, this maximum variance is hardly altered because the increase of $`\chi ^2`$ is weak for $`\overline{t}\stackrel{<}{}0.05`$ as is found in Fig. 11. The back reaction effect of $`\varphi `$ particles onto the $`\varphi _0`$ field is marginally negligible as the case of $`g=0`$. When $`g=0.1`$, the final variance is $`\chi ^2_f=2.5\times 10^9M_{\mathrm{pl}}^2`$, and the growth of the fluctuation hardly affects the evolution of the $`\varphi _0`$ field. However, in the case of $`g=0.5`$, $`\chi ^2_f`$ increases up to $`\chi ^2_f=1.3\times 10^6M_{\mathrm{pl}}^2`$ at $`\overline{t}0.17`$. In Fig. 12, we find that the increase of $`\chi `$ particles prevents the coherent oscillation of the $`\varphi _0`$ field for $`\overline{t}\stackrel{>}{}0.17`$. When $`g\stackrel{>}{}0.5`$, since the $`\chi `$ particle production stops by the back reaction effect, the final variance gradually decreases with the increase of $`g`$. As a result, the maximum variance for the case of $`\varphi _c=10^4M_{\mathrm{pl}}`$ is $`\chi ^2_{\mathrm{max}}10^6M_{\mathrm{pl}}^2`$ for $`g0.5`$.
With the decrease of $`\varphi _c`$ ($`\stackrel{<}{}10^5M_{\mathrm{pl}})`$, the development of the $`\chi `$ fluctuation is not expected unless $`g`$ is unnaturally large. Moreover, in this case, although the $`\varphi `$ particle production is possible in the initial stage of the oscillating inflation, the final fluctuation $`\delta \varphi ^2_f`$ is strongly suppressed. This means that it is difficult to obtain the sufficient amount of $`\varphi `$ and $`\chi `$ particles with the natural coupling $`g\stackrel{<}{}1`$ in the case of $`\varphi _c\stackrel{<}{}10^5M_{\mathrm{pl}}`$. For large critical values $`\varphi _c>10^3M_{\mathrm{pl}}`$, the $`\chi `$ particle production occurs effectively for the natural coupling. With the increase of $`\varphi _c`$, the structure of resonance for the $`\chi `$ field approaches that of the model $`V(\varphi ,\chi )=m^2\varphi ^2/2+g^2\varphi ^2\chi ^2/2`$, and the investigation based on the Mathieu equation performed in Ref. can be applied. Since the mass $`m_\varphi `$ around the minimum of the inflaton potential becomes smaller, the $`\chi `$ particle production occurs efficiently. In this case, however, since the duration of the oscillating is too short, the $`\varphi `$ particle production is hardly expected. In the case of the massless $`\chi `$ particle, we conclude that both $`\varphi `$ and $`\chi `$ particles are most efficiently created in the parameter range $`\varphi _c=10^3`$-$`10^4M_{\mathrm{pl}}`$ for the natural coupling $`g`$.
Finally, we discuss the case where the mass of the $`\chi `$ particle is included. We expect that the large mass makes the $`\chi `$ field deviate from the resonance band by the relation $`(\text{43})`$. Let us consider the case of $`\varphi _c=10^3M_{\mathrm{pl}}`$ and $`q=0.1`$. In the massless case, the $`\chi `$ particle production occurs relevantly for $`g\stackrel{>}{}0.05`$. If the mass of the $`\chi `$ particle is taken into account, numerical calculations show that heavy particles up to the order of $`m_\chi =10^{14}`$ GeV can be produced enough for $`g\stackrel{>}{}0.05`$. In Fig. 13, the evolution of the fluctuation $`\chi ^2`$ is depicted in two cases of $`m_\chi =3\times 10^{14}`$ GeV and $`m_\chi =1\times 10^{15}`$ GeV for $`g=0.05`$. The final variance $`\chi ^2_f10^8M_{\mathrm{pl}}^2`$ with mass $`m_\chi =3\times 10^{14}`$ GeV is almost the same as in the case of the massless $`\chi `$ particle with the same coupling $`g`$. This means that the massive $`\chi `$ particle whose mass is $`m_\chi \stackrel{<}{}3\times 10^{14}`$ GeV is produced with almost the same amount as the massless $`\chi `$ particle for $`g=0.05`$. However, the achieved amount of the produced $`\chi `$ particle decreases with the increase of $`m_\chi `$ for $`m_\chi \stackrel{>}{}3\times 10^{14}`$ GeV. In the case of $`g=0.05`$, although the $`\chi `$ particle with mass $`m_\chi \stackrel{<}{}7\times 10^{14}`$ GeV can be created, it is difficult to generate the massive particle whose mass is $`m_\chi \stackrel{>}{}1\times 10^{15}`$ GeV (See Fig. 13). In the model of $`V(\varphi ,\chi )=m^2\varphi ^2/2+g^2\varphi ^2\chi ^2/2`$, analytic investigations show that the $`\chi `$ particle which satisfies the condition $`m_\chi \stackrel{<}{}\sqrt{g}10^{15}`$ GeV can be generated. When $`m_\chi =10^{14}`$ GeV, numerical calculations performed in Ref. reveal that the $`\chi `$ particle production occurs for $`g\stackrel{>}{}0.06`$. This condition is almost the same as in the oscillating inflation model in spite of the difficulty in producing the massless $`\chi `$ particle without the larger $`g`$ compared with the $`V(\varphi ,\chi )=m^2\varphi ^2/2+g^2\varphi ^2\chi ^2/2`$ model. The reason is simple. In Eq. $`(\text{43})`$, the massive $`\chi `$ particle production occurs when the initial value of the resonance term $`g^2\varphi _0^2/m_\varphi ^2`$ is much greater than unity \[$`(g^2\varphi _0^2/m_\varphi ^2)_i1`$\] and $`\overline{m}_\chi ^2`$ is much smaller than this term $`[\overline{m}_\chi ^2(g^2\varphi _0^2/m_\varphi ^2)_i]`$. In the case of $`\varphi _c=10^3M_{\mathrm{pl}}`$, $`q=0.1`$, and $`g=0.05`$, since $`M=3.96\times 10^4M_{\mathrm{pl}}`$ by Eq. $`(\text{29})`$, we obtain $`(g^2\varphi _0^2/m_\varphi ^2)_i4\times 10^3`$. Although $`\stackrel{~}{\varphi }_0`$ decreases by the cosmic expansion and $`m_\varphi `$ becomes large in the vicinity of $`\varphi _0=0`$, the massive $`\chi `$ particle production is possible as long as $`g^2\stackrel{~}{\varphi }_0^2/m_\varphi ^21`$ and $`\overline{m}_\chi ^2g^2\stackrel{~}{\varphi }_0^2/m_\varphi ^2`$. Since the mass $`m_\varphi `$ in this case is estimated as $`m_\varphi 10^5M_{\mathrm{pl}}`$ for $`\varphi _0^2\varphi _c^2`$ and $`m_\varphi 10^4M_{\mathrm{pl}}`$ for $`\varphi _0^2\varphi _c^2`$ by Eq. $`(\text{39})`$, this is heavier than the mass $`m10^6M_{\mathrm{pl}}`$ in the $`V(\varphi ,\chi )=m^2\varphi ^2/2+g^2\varphi ^2\chi ^2/2`$ model. Even in the case where the normalized mass is $`\overline{m}_\chi =1`$, this corresponds to the rather heavy particle whose mass is $`m_\chi \stackrel{>}{}10^{14}`$ GeV. For the typical value $`\varphi _c\stackrel{<}{}10^3M_{\mathrm{pl}}`$ in the oscillating inflation model, since $`m_\varphi `$ is at least by one order larger than $`m=10^6M_{\mathrm{pl}}`$, the massive $`\chi `$ particle whose mass is of order $`m_\chi =10^{14}`$ GeV can be generated for the coupling $`g`$ where the massless $`\chi `$ production occurs. As the coupling $`g`$ increases, more massive particles are generated. In the case of $`\varphi _c=10^3M_{\mathrm{pl}}`$, we find that the production of the massive particle whose mass is of order $`m_\chi =10^{15}`$ GeV relevantly occurs for the coupling $`g\stackrel{>}{}0.2`$. Even the GUT scale bosons $`m_\chi =10^{16}`$ GeV are produced in the initial stage of the oscillating inflation for the case of $`g=1`$, although the achieved amount is small (See Fig. 14).
With the decrease of $`\varphi _c`$, the larger values of $`g`$ are required in order to produce the massive $`\chi `$ particle as is the same with the massless $`\chi `$ particle. However, in the parameter range of $`g`$ where the massless $`\chi `$ particles are sufficiently produced, we can also expect the generation of the superheavy particles needed for the success of the GUT baryogenesis. For example, in the case of $`\varphi _c=10^4M_{\mathrm{pl}}`$, numerical calculations indicate that $`\chi `$ particles up to the mass $`m_\chi \stackrel{<}{}10^{14}`$ GeV and $`m_\chi \stackrel{<}{}10^{15}`$ GeV are created for the coupling $`g\stackrel{>}{}0.07`$ and $`g\stackrel{>}{}0.3`$ respectively. It will be interesting to investigate how these superheavy bosons produced in the oscillating inflation model would affect the scenario of the GUT baryogenesis.
## IV Concluding remarks and Discussions
In this paper we have considered parametric amplification of a scalar field $`\chi `$ coupled to an inflaton field $`\varphi `$ as well as the enhancement of the fluctuation of the $`\varphi `$ field in the oscillating inflation model. Since the inflaton potential $`V(\varphi )`$ is the non-convex type where $`d^2V/d\varphi ^2`$ takes negative values in the regions not too far from the minimum of the potential, inflation is realized even during the oscillation of the inflaton field. The oscillating inflation model is characterized by the critical value of $`\varphi _c`$ which indicates the scale of the core part of the potential. Inflation takes place while the $`\varphi `$ field moves in the regions of $`|\varphi |\stackrel{>}{}\varphi _c`$. With the decrease of $`\varphi _c`$, the total amount of inflation becomes larger, because the duration during which the oscillating inflation takes place becomes longer. However, it is rather difficult to obtain the sufficient number of $`e`$-foldings such as $`N\stackrel{>}{}60`$ only by the stage of the oscillating inflation, and the main contribution to the number of $`e`$-foldings is achieved by the ordinary slow-roll inflation which precedes the oscillating inflation. We have estimated the energy scale of the oscillating inflation $`M^4`$ by fitting the primordial density perturbation observed by COBE. The oscillating inflation terminates when the amplitude of the $`\varphi _0`$ field decreases to of the order $`\varphi _c`$ by the expansion of the universe.
In this model, the fluctuation of the $`\varphi `$ field grows during the oscillating inflation due to the existence of the non-convex part of the potential. We examined the growth of the fluctuation for two typical cases $`\varphi _c=10^3M_{\mathrm{pl}}`$ and $`\varphi _c=10^4M_{\mathrm{pl}}`$ including the back reaction effect of created particles based on the Hartree approximation. We have found that the wave modes inside the Hubble horizon are effectively enhanced, while the growth of long-wave modes is small as was shown in Ref. . Although the growth rate of the fluctuation $`\delta \varphi ^2`$ in the initial stage of the oscillating inflation becomes larger with the decrease of $`\varphi _c`$, the final fluctuation is significantly suppressed. This indicates that the generation of the $`\varphi `$ particle itself plays a crucial role for the termination of parametric resonance. As for the final variance of the $`\varphi `$ particle, we find the relation $`\delta \varphi ^2_f0.1\varphi _c^2`$ for the case of $`\varphi _c\stackrel{<}{}10^3M_{\mathrm{pl}}`$.
With regard to the $`\chi `$ field coupled to inflaton, the enhancement of the fluctuation mainly occurs when the $`\varphi _0`$ field moves nonadiabatically around the minimum of the potential. Since the mass $`m_\varphi `$ for $`|\varphi _0|\stackrel{<}{}\varphi _c`$ is approximately written as $`m_\varphi M^2/\varphi _c`$, it is larger than that of the massive inflaton of the chaotic inflation model for the value of $`\varphi _c\stackrel{<}{}10^3M_{\mathrm{pl}}`$. As $`m_\varphi `$ becomes larger, this restricts the effective $`\chi `$ particle production. Moreover, the amplitude of the homogeneous inflaton field decreases faster than in the model of the massive inflaton. This means that the $`\chi `$ particle production is not possible unless we select rather large values of the coupling constant $`g`$. This tendency becomes stronger with the decrease of $`\varphi _c`$. For example, in the case of $`\varphi _c=10^3M_{\mathrm{pl}}`$ and $`q=0.1`$, the coupling $`g\stackrel{>}{}0.05`$ is required to yield the sufficient $`\chi `$ particle production, the value of which is two orders of magnitude larger than in the model of $`V(\varphi ,\chi )=m^2\varphi ^2/2+g^2\varphi ^2\chi ^2/2`$. For $`g\stackrel{>}{}0.05`$, the back reaction effect onto the $`\varphi _0`$ field due to the $`\chi `$ particle production is significant, and finally terminates parametric resonance. The final variance is suppressed by the back reaction effect with the increase of $`g`$, and the maximal variance for the $`\varphi _c=10^3M_{\mathrm{pl}}`$ case is $`\chi ^2_f10^5M_{\mathrm{pl}}^2`$ when $`g0.3`$. In the case of $`\varphi _c=10^4M_{\mathrm{pl}}`$, we need $`g\stackrel{>}{}0.1`$ for an effective resonance, although the $`\chi `$ particle production occurs for $`g\stackrel{>}{}0.01`$. The maximal variance for the $`\varphi _c=10^4M_{\mathrm{pl}}`$ case is $`\chi ^2_f10^6M_{\mathrm{pl}}^2`$ when $`g0.5`$. As $`\varphi _c`$ decreases further, we no longer expect the growth of the $`\chi `$ fluctuation with the natural coupling $`g\stackrel{<}{}1`$. Moreover, since the final fluctuation of the $`\varphi `$ particle is also suppressed, the small $`\varphi _c`$ such as $`\varphi _c\stackrel{<}{}10^5M_{\mathrm{pl}}`$ is not favorable for the production of a sufficient amount of $`\varphi `$ and $`\chi `$ particles.
In the oscillating inflation model, the superheavy $`\chi `$ particle with mass greater than $`m_\chi =10^{14}`$ GeV can be copiously produced. For example, in the case of $`\varphi _c=10^3M_{\mathrm{pl}}`$, we find that the $`\chi `$ particle whose mass is of order $`10^{14}`$GeV and $`10^{15}`$GeV is effectively enhanced for $`g\stackrel{>}{}0.05`$ and $`g\stackrel{>}{}0.2`$ respectively. The GUT scale boson $`m_\chi 10^{16}`$ GeV can be also generated in the initial stage if the coupling $`g`$ is of order unity, although the final amount is small. In Ref. , the authors considered the GUT scale baryogenesis with the massive inflaton assuming that some of the initial inflaton energy is efficiently transferred to the bosons with the mass $`10^{14}`$ GeV. These massive bosons decay into lighter particles, and produce the net baryon number. Although we do not consider such a decaying process in this paper, it is of interest how the production of the superheavy particle would affect the baryon asymmetry in the universe.
In this paper we make use of the Hartree approximation, which is essentially the mean field approximation. This does not include the rescattering effect which becomes important as $`\chi `$ particles are sufficiently produced. The scattering between $`\varphi `$ and $`\chi `$ particles may reduce the final amount of fluctuations. For a complete study including the nonlinear effect of the particle production, we should perform the lattice simulations and compare them with the mean field approximation performed in this paper.
Finally, we comment on the case where the metric perturbation is included. In the single field case, even if we consider the effect of the metric perturbation, the $`V^{(2)}(\varphi _0)`$ term in Eq. $`(\text{9})`$ is much more important than the $`2\kappa ^2(V/H)^\dot{}`$ term which is the gravitational origin. Hence the situation is almost the same as the case when the metric perturbation is neglected. As we have showed, the long-wave modes are not significantly enhanced compared with the modes inside the Hubble Horizon in the single-field case. This result is consistent with other inflation models in the single-field case. In the two-field case, it was suggested that super-Hubble metric perturbations can be amplified in broad classes of models. In the present model, since the $`\chi `$ fluctuation of super-Hubble modes will be exponentially suppressed for large values of $`q1`$, we may not expect the enhancement of super-Hubble metric perturbations significantly. However, it is worth investigating to investigate these issues including mode-mode coupling between field and metric perturbations, because this may lead to some imprints on the spectrum of density perturbations. These issues are under consideration.
## ACKOWLEDGEMENTS
The author would like to thank Bruce A. Bassett, Kei-ichi Maeda, Atsushi Taruya, Takashi Torii, and Hiroki Yajima for useful discussions. This work was supported partially by a Grant-in-Aid for Scientific Research Fund of the Ministry of Education, Science and Culture (No. 09410217 and Specially Promoted Research No. 08102010), and by the Waseda University Grant for Special Research Projects.
Figure Captions
FIG. 1:
The potential for $`q=0.1`$ in the oscillating inflation model, which is composed of the non-convex region $`|\varphi |\stackrel{>}{}\varphi _c`$ and the core region $`|\varphi |\stackrel{<}{}\varphi _c`$.
FIG. 2:
The evolution of the $`\varphi _0`$ field in the case of $`\varphi _c=10^3M_{\mathrm{pl}}`$, $`q=0.1`$, and $`g=0`$. The stage of the oscillating inflation starts from $`\varphi _01.4\times 10^2M_{\mathrm{pl}}`$, and ends at $`\overline{t}0.37`$ when the amplitude of the $`\varphi _0`$ field becomes of order $`\varphi _c`$.
FIG. 3:
The evolution of the real part of the fluctuation $`\delta \varphi _k`$ for two momentum modes of $`\overline{k}=0.1`$ and $`\overline{k}=100`$ in the case of $`\varphi _c=10^3M_{\mathrm{pl}}`$, $`q=0.1`$, and $`g=0`$. For the long wave mode $`\overline{k}=0.1`$, the growth of the fluctuation is weak, but for the mode $`\overline{k}=100`$, the fluctuation grows rapidly until the back reaction effect becomes significant at $`\overline{t}0.12`$.
FIG. 4:
The evolution of $`\delta \varphi ^2`$ in the case of $`\varphi _c=10^3M_{\mathrm{pl}}`$, $`q=0.1`$, and $`g=0`$. The fluctuation reaches the maximum value $`\delta \varphi ^2_f10^7M_{\mathrm{pl}}^2`$ at $`\overline{t}0.12`$.
FIG. 5:
The evolution of $`\delta \varphi ^2`$ in the case of $`\varphi _c=10^4M_{\mathrm{pl}}`$, $`q=0.1`$, and $`g=0`$. Although the initial growth rate is larger than in the case of $`\varphi _c=10^3M_{\mathrm{pl}}`$, the fluctuation soon reaches the maximum value $`\delta \varphi ^2_f8.0\times 10^{10}M_{\mathrm{pl}}^2`$, which is by two orders smaller than in the case of $`\varphi _c=10^3M_{\mathrm{pl}}`$.
FIG. 6:
The evolution of the fluctuation $`\chi ^2`$ in the case of $`\varphi _c=10^3M_{\mathrm{pl}}`$ and $`q=0.1`$ for $`g=0.03`$ (bottom) and $`g=0.07`$ (top). For $`g=0.03`$, the maximum value of the fluctuation is $`\chi ^2_f=2.3\times 10^{11}M_{\mathrm{pl}}^2`$; and for $`g=0.07`$, $`\chi ^2_f=2.4\times 10^7M_{\mathrm{pl}}^2`$.
FIG. 7:
The evolution of the $`\varphi _0`$ field in the case of $`\varphi _c=10^3M_{\mathrm{pl}}`$, $`q=0.1`$, and $`g=0.07`$. The coherent oscillation is a bit broken at $`\overline{t}0.12`$ due to the increase of $`\delta \varphi ^2`$. With the increase of $`\chi ^2`$, this also changes the frequency of the $`\varphi _0`$ field at $`\overline{t}0.28`$.
FIG. 8:
The evolution of $`\delta \varphi ^2`$ in the case of $`\varphi _c=10^3M_{\mathrm{pl}}`$, $`q=0.1`$, and $`g=0.07`$. With the growth of $`\chi ^2`$, this assists the $`\varphi `$ particle production, and the final variance $`\delta \varphi ^2_f=2.7\times 10^6M_{\mathrm{pl}}^2`$ becomes larger than in the case of $`\varphi _c=10^3M_{\mathrm{pl}}`$ and $`g=0`$ by one order of magnitude.
FIG. 9:
The final variance $`\chi ^2_f`$ as the function of $`g`$ in the case of $`\varphi _c=10^3M_{\mathrm{pl}}`$ and $`q=0.1`$. We find that $`\chi ^2_f`$ takes the maximum value $`\chi ^2_{\mathrm{max}}10^5M_{\mathrm{pl}}^2`$ when $`g0.3`$.
FIG. 10:
The evolution of the real part of the fluctuation $`\chi _k`$ for two momentum modes of $`\overline{k}=0`$ and $`\overline{k}=100`$ in the case of $`\varphi _c=10^3M_{\mathrm{pl}}`$, $`q=0.1`$, and $`g=0.1`$. The long wave mode $`\overline{k}=0`$ is strongly enhanced compared with the wave mode $`\overline{k}=100`$.
FIG. 11:
The evolution of the fluctuation $`\chi ^2`$ in the case of $`\varphi _c=10^4M_{\mathrm{pl}}`$, $`q=0.1`$, and for $`g=0.07`$ (bottom), $`g=0.1`$ (middle), and $`g=0.5`$ (top) respectively. The final variances are $`\chi ^2_f=1.2\times 10^{10}M_{\mathrm{pl}}^2`$, $`2.5\times 10^9M_{\mathrm{pl}}^2`$, and $`1.3\times 10^6M_{\mathrm{pl}}^2`$ respectively.
FIG. 12:
The evolution of the $`\varphi _0`$ field in the case of $`\varphi _c=10^4M_{\mathrm{pl}}`$, $`q=0.1`$, and $`g=0.5`$. While the back reaction effect of $`\varphi `$ particles can be negligible, the increase of $`\chi ^2`$ strongly affects the evolution of the $`\varphi _0`$ field from $`\overline{t}0.17`$.
FIG. 13:
The evolution of $`\chi ^2`$ for $`\varphi _c=10^3M_{\mathrm{pl}}`$, $`q=0.1`$, and $`g=0.05`$ in two cases of $`m_\chi =3\times 10^{14}`$ GeV (top) and $`m_\chi =1\times 10^{15}`$ GeV (bottom). The $`\chi `$ particle whose mass is of order $`m_\chi =10^{14}`$ GeV is generated, but the production of the massive particle with $`m_\chi \stackrel{>}{}10^{15}`$ GeV is hardly expected for $`g=0.05`$.
FIG. 14:
The evolution of $`\chi ^2`$ for $`\varphi _c=10^3M_{\mathrm{pl}}`$ and $`q=0.1`$, in two cases of $`g=0.5`$, $`m_\chi =1\times 10^{15}`$ GeV (top); and $`g=1`$, $`m_\chi =1\times 10^{16}`$ GeV (bottom). In the case of $`g=0.5`$, the $`\chi `$ particle whose mass is of order $`m_\chi =10^{15}`$ GeV is effectively enhanced. In the case of $`g=1`$, the GUT scale boson $`m_\chi =10^{16}`$ GeV is a little produced in the initial stage of the oscillating inflation.
|
warning/0003/hep-ph0003192.html
|
ar5iv
|
text
|
# IISc-CTS-6/00 hep-ph/0003192 Constraints on 𝑚_𝑠 and ϵ'/ϵ from Lattice QCD
## 1 Determination of Light Quark Masses
Quark masses are not physical observables in QCD, rather they enter the theory as parameters in the Lagrangian. Their values depend on the QCD renormalisation scale, and three quantitative approaches have been used to determine them.
### 1.1 Chiral perturbation theory
This is a low energy ($`E1\mathrm{GeV}`$) effective field theory of QCD in presence of spontaneous chiral symmetry breaking. With $`m_q\mathrm{\Lambda }_{QCD}`$, the pseudo-Nambu-Goldstone bosons are the dominant fields at low energy. $`O(p^2,m_q)`$ terms in the effective Lagrangian are fixed by the spectrum. $`O(p^4,p^2m_q,m_q^2)`$ terms are estimated using resonance saturation and large $`N_c`$ power counting, as well as from phenomenological fits to various form factors. Small electromagnetic and isospin breaking effects are systematically included, and bounds are obtained on quark mass ratios.
The range of validity of this effective field theory expansion cannot be convincingly established. Also the absolute mass scale has to be fixed from experimental data. Yet renormalisation group invariant dimensionless mass ratios can be tightly constrained .
$$\frac{m_u}{m_d}=0.553\pm 0.043,\frac{m_s}{m_d}=18.9\pm 0.8,\frac{m_s}{m_u}=34.4\pm 3.7,$$
(1)
$$\frac{2m_s}{m_u+m_d}\frac{m_s}{m_{ud}}=24.4\pm 1.5,\frac{m_sm_{ud}}{m_dm_u}=40.8\pm 3.2.$$
(2)
These mass ratios are typically combined with $`m_s(1\mathrm{GeV})`$ extracted from QCD sum rules to give individual quark masses.
### 1.2 QCD sum rules
In this approach, two-point hadronic current correlators are evaluated using the Operator Product Expansion (OPE) and perturbative QCD in the Euclidean region $`q^2<0`$. The expansion in inverse powers of $`q^2`$ is then analytically continued to the physical region $`q^2>0`$, and matched to the experimental data using dispersion relations. The experimental data is organised in terms of contributions from well-known leading poles and subleading branch-cuts from the QCD continuum. Subtractions are used in the dispersion relations to suppress large continuum contributions, while threshold factors and moments help in suppressing contribution of regions near the poles where perturbative QCD is inapplicable. The results critically depend on proper choice of the boundary conditions.
Pseudoscalar, scalar, $`e^+e^{}`$annihilation and $`\tau `$decay sum rules have been used by various groups to extract $`m_s`$. In the $`\overline{MS}`$ scheme, the required $`\beta `$ and $`\gamma `$ functions are known to $`4`$loop precision, giving the most accurate results :
$$\overline{m}_s(1\mathrm{GeV})=162.5\pm 15.5\mathrm{MeV},\overline{m}_s(2\mathrm{GeV})=117.8\pm 12.3\mathrm{MeV}.$$
(3)
Analysis of Cabbibo-suppressed $`\tau `$decay gives $`\overline{m}_s(M_\tau )=119\pm 24`$ MeV .
The major uncertainty in the sum rule approach is the asymptotic nature of the perturbative QCD expansion—in the longitudinal channel it cannot be pushed beyond $`O(\alpha _s^3)`$.
### 1.3 Lattice QCD simulations
These provide a first principle non-perturbative determination of quark masses without additional assumptions. It is convenient to convert the running quark masses to renormalisation group invariant quark masses, completely within the lattice regularisation framework.
$$m_q^{RGI}=\underset{\mu \mathrm{}}{lim}\{(2\beta _0g^2(\mu ))^{\gamma _0/2\beta _0}m_q(\mu )\}.$$
(4)
With improvements in simulation algorithms and computers, the systematic errors from finite lattice spacing, finite lattice size and the quenched approximation are gradually coming under control. Most simulations are in the range $`m_s/3m_q3m_s`$, and $`m_u=m_d`$ is assumed. Results from several simulations are combined to extrapolate to $`\mu \mathrm{}`$.
Perturbative relations connecting lattice and $`\overline{MS}`$ values show that including higher order corrections decreases $`\overline{m}_q`$. Moreover, at $`\mu =2\mathrm{GeV}`$, the $`N^3LO`$ correction is comparable to the $`N^2LO`$ one, and decreases $`\overline{m}_q`$ by about $`10\%`$ . A rigorous comparison of lattice and $`\overline{MS}`$ values should therefore be made at $`\mu >2\mathrm{GeV}`$, contrary to common practice.
Here is a summary of results presented by various groups at LAT99.
$``$ CP-PACS: RG-improved gauge field action and tadpole-improved clover fermion action are used. Ward identities are used to extract $`m_q`$. $`N_f=0`$ and $`N_f=2`$ results are compared, demonstrating that going to $`N_f=2`$ reduces the systematic error in the lattice scale. The results converted to the $`\overline{MS}`$ scheme are :
$$\overline{m}_{ud}(2\mathrm{GeV})=3.3(4)\mathrm{MeV},\overline{m}_s(2\mathrm{GeV})=84(7)\mathrm{MeV}.$$
(5)
$``$ ALPHA/UKQCD: Schrödinger functional method is used with $`O(a)`$ improved action and $`N_f=0`$. PCAC relation is used to extract $`m_q`$. With $`(M_Kr_0)^2=1.5736`$ fixing the reference scale, the results are :
$$2M_{ref}=m_s^{RGI}+m_{ud}^{RGI}=143(5)\mathrm{MeV},\overline{m}_s(2\mathrm{GeV})=94(4)\mathrm{MeV}.$$
(6)
$``$ QCDSF: Schrödinger functional method is used with $`O(a)`$ improved action and $`N_f=0`$. PCAC relation is used to extract $`m_q`$. Using $`r_0=0.5\mathrm{fm}`$ as the reference scale yields :
$$\overline{m}_{ud}(2\mathrm{GeV})=4.4(2)\mathrm{MeV},\overline{m}_s(2\mathrm{GeV})=105(4)\mathrm{MeV}.$$
(7)
$``$ Rome group: $`O(a)`$ improved action is used with $`N_f=0`$. Ward identity for the renormalised quark propagator is used to extract $`m_q`$. With pseudoscalar and vector meson masses fixing the lattice scale, the results are :
$$\overline{m}_{ud}(2\mathrm{GeV})=4.8(5)\mathrm{MeV},\overline{m}_s(2\mathrm{GeV})=111(9)\mathrm{MeV}.$$
(8)
Numerical results show that, for a fixed value of the lattice scale, increasing $`N_f`$ increases the coupling $`g^2`$ through vacuum polarisation. This translates into $`m_q(\mu )`$ decreasing with increasing $`N_f`$. My educated guess, including the effect of a relatively light $`s`$quark, is
$$\overline{m}_s(2\mathrm{GeV},N_f=2.5)=90\pm 10\mathrm{MeV}.$$
(9)
## 2 Determination of $`ϵ^{}/ϵ`$
$`ϵ`$ and $`ϵ^{}`$ parametrise indirect and direct $`CP`$violation effects in neutral kaon decays into two pions. $`ϵ`$ arises from the mixing between $`CP`$eigenstates $`K_L`$ and $`K_S`$, and is experimentally measured to be
$$ϵ=2.280(13)10^3e^{i\varphi _ϵ},\varphi _ϵ\pi /4.$$
(10)
$`ϵ^{}`$ parametrises $`CP`$violation in the decay amplitudes without $`K_LK_S`$ mixing, and in the standard model it arises from the complex phase in the CKM quark mixing matrix. $`\pi \pi `$ scattering data establish that the phases of $`ϵ`$ and $`ϵ^{}`$ are almost identical. Recent results from NA31, E731, KTeV and NA48 experiments have determined (see Ref. for details)
$$Re(ϵ^{}/ϵ)=(21.2\pm 4.6)10^4.$$
(11)
### 2.1 Parametrisation using OPE
All phenomenological explanations of $`ϵ^{}/ϵ`$ take the $`CP`$conserving data from experiments and estimate the $`CP`$violating part. In the standard model, $`ϵ`$ is found by calculating the box diagram, while $`ϵ^{}`$ is found by calculating the $`g,\gamma ,Z^0`$ penguin diagrams. Operator mixing and RG-evolution are crucial in the analysis, and electroweak contributions have become important due to large $`m_t`$. The dominant components are the $`K\pi \pi `$ matrix elements of the $`4`$fermion operators $`Q_6`$ and $`Q_8`$, in the $`\mathrm{\Delta }I=1/2`$ and $`\mathrm{\Delta }I=3/2`$ channels. Let $`B_{i,\mathrm{\Delta }I}`$ denote the ratios of the actual matrix elements to their values in the vacuum saturation approximation (VSA). Using commonly accepted scale parameters, and including isospin breaking effects with $`\mathrm{\Omega }_{\eta +\eta ^{}}=0.25(8)`$, an approximate formula is
$`{\displaystyle \frac{ϵ^{}}{ϵ}}`$ $`13Im(V_{td}V_{ts}^{})\left({\displaystyle \frac{110\mathrm{MeV}}{\overline{m}_s(2\mathrm{GeV})}}\right)^2\left({\displaystyle \frac{\mathrm{\Lambda }_{\overline{MS}}^{(4)}}{340\mathrm{MeV}}}\right)`$ (12)
$`\left[B_{6,(1/2)}(1\mathrm{\Omega }_{\eta +\eta ^{}})0.4B_{8,(3/2)}\right].`$
I want to emphasise that the appearance of $`m_s`$ in the above formula is spurious; it arises because the VSA matrix element values are accompanied by quark mass factors in the chiral limit. It is silly to express one unknown matrix element in terms of two other unknowns, $`m_s`$ and $`B_{i,\mathrm{\Delta }I}`$, but that has become commonplace. To avoid unwanted systematic errors, one should therefore either (a) calculate the matrix elements directly, or (b) use the same calculational framework to evaluate both $`m_s`$ and the $`B_{i,\mathrm{\Delta }I}`$. Much of the confusion in the literature has arisen from not following this simple guideline consistently.
$`ϵ^{}/ϵ`$ has been evaluated in three different frameworks: lattice QCD, large$`N_c`$ approximation, and chiral quark model. Lattice QCD simulations show that $`B_{8,(3/2)}`$ is suppressed below $`1`$, but provide no clean result for $`B_{6,(1/2)}`$. With $`B_{6,(1/2)}=1.0(3)`$ and $`B_{8,(3/2)}=0.8(2)`$, the estimates for $`ϵ^{}/ϵ`$ are about a factor $`2`$ below its experimental value.
### 2.2 Final state interactions
Final state interactions (FSI) strongly influence $`K\pi \pi `$ decays, as shown by the experimental phase shifts. They are absent in VSA, lattice QCD, large$`N_c`$ approximation, and appear at subleading order in chiral perturbation theory. Their contribution to the decay amplitudes can be estimated using dispersion relation analysis of the experimentally measured phase shifts in the elastic $`\pi \pi `$ channel. There is no doubt that incorporating the FSI boosts the $`\mathrm{\Delta }I=1/2`$ amplitude and suppresses the $`\mathrm{\Delta }I=3/2`$ one, and hence increases $`ϵ^{}/ϵ`$. The exact value of the enhancement is debatable, since it depends on the boundary conditions used to analyse the experimental data. In a specific analysis , the enhancement is the required factor of $`2`$. Higher order chiral quark model calculations, which automatically include the FSI, also enhance $`ϵ^{}/ϵ`$ close to its experimental value .
In conclusion, the standard model, with a proper framework to handle non-perturbative QCD effects, is fully capable of explaining the observed value of $`ϵ^{}/ϵ`$; contributions to $`ϵ^{}/ϵ`$ from new effects beyond the standard model must be kept $`510^4`$.
|
warning/0003/quant-ph0003137.html
|
ar5iv
|
text
|
# Fermionic quantum computation
## Introduction
The notion of locality plays the key role in the definition of computation process. The same basic principles apply to classical computers and the circuit model of quantum computation :
1. The computer consists of small pieces, or cells (bits, qubits, qutrits or something else).
2. It is allowed to operate on few cells at a time.
3. All cells are identical, so each operation has a model which can be applied to different sets of cells. We call such a model a gate (like the CNOT gate) while the operation itself is called a gate application (like CNOT applied to qubits $`5`$ and $`8`$).
In fact, the main difference between classical computation and quantum computation is the concrete interpretation of these postulates. The standard quantum interpretation is as follows.
* Each cell is described by a Hilbert space of small dimensionality. (Without loss of generality, these spaces are two-dimensional, in which case the cells are called qubits). The Hilbert space of the entire computer is the tensor product of the spaces associated to the individual cells.
* Each operation is described by a unitary operator which is the tensor product of some operator $`U`$, acting on the selected qubits, and the identity operator acting on the rest of the system.
* A $`p`$-qubit gate can be defined as $`U`$ acting on some standard $`2^p`$-dimensional space (which is the same as a standard set of $`p`$ qubits).
This interpretation might be perfect from the complexity-theoretic point of view, but it does not necessarily correspond to physics. At the fundamental level, Fermi systems do not satisfy the condition (2i). Hence using fermions as carriers of quantum information should be considered as a different computation model, although it is equivalent to the standard one in a certain sense. At the macroscopic level, collective quantum degrees of freedom (or excitations, such as anyons ) do not even satisfy the condition (1i). In all such cases, it makes sense to abstract the nontrivial locality properties from physical details. This will lead us to a definition of “fermionic gates”, “anyonic gates” and other quantum computation models which deserve careful study.
What can one expect from alternative models of quantum computation? It is very unlikely that any physical system would provide more computational power than the standard quantum computation model has. (This might be wrong for quantum gravity but here we can only guess). So, the alternative models should be polynomially equivalent to the standard one. There are indeed several results which support this statement. Firstly, D. Abrams and S. Lloyd have shown that a system of $`m`$ local fermionic modes (i.e. sites which can be empty or occupied) can be simulated on a quantum computer in such a way that one fermionic gate takes $`O(m)`$ qubit operations. In some cases (one of the assumptions is that the number of particles is conserved) faster simulation is possible. We will extend this result by showing that in the general case (when the number of fermions is conserved only modulo $`2`$) each fermionic gate can be simulated by $`O(\mathrm{log}m)`$ qubit gates. Secondly, TQFT computation (which is more general than quantum computation with anyons) can be simulated in polynomial time on an ordinary quantum computer .
Thus alternative quantum computation models do not generate new computational classes. Rather, they provide new descriptions of the standard class $`\mathrm{B}QP`$ (the class of problems that are solvable on a quantum computer in polynomial time). These descriptions may be useful to find new quantum algorithms, error-correcting codes, fault-tolerant procedures, or to prove that $`\mathrm{B}QP`$ is contained in some other computational classes. They may also open the door to new physical implementations of a quantum computer.
## 1 A more general notion of locality
Why isn’t the standard interpretation of locality good in all cases? The answer is in the way we describe quantum evolution. The Hilbert space and state vectors are very convenient tools but they are not directly related to physically observable things. Operators are much more “real” as they allow one to describe interaction between the system and the rest of the world. In fact, the causality principle is usually stated in terms of operators: “operators in spatially separated points commute”. Our definition of locality will be in the same spirit.
A quantum system can be characterized by a finite-dimensional $`C^{}`$-algebra<sup>1</sup><sup>1</sup>1$`C^{}`$-algebra is a generalization of the algebra of linear operators on a Hilbert space. The properties of operator addition, multiplication, Hermitian conjugation and the operator norm are axiomatized in a certain way. However, instead of using the axioms, we will rely on a characterization of finite-dimensional $`C^{}`$-algebras given below. $`𝒢`$ whose elements are called “physical operators”. As a matter of fact, they are operators on a suitable Hilbert space. Indeed, it is a theorem that $`𝒢`$ can be represented as $`_j𝐋(_j)`$, where $`𝐋()`$ stands for the algebra of operators on the space $``$. Hence $`𝒢`$ acts on $`=_j_j`$. (In the case of fermions, $`=_0_1`$ is the Foch space split into the subspaces corresponding to an even and odd number of particles; $`𝒢`$ is the algebra of operators which preserve the parity).
To define locality, we will assume that the system is associated with some set of sites $`M`$. The following properties are postulated:
1. For each subset of sites $`SM`$, there is a $`C^{}`$-subalgebra $`𝒢(S)𝒢`$. Elements of $`𝒢(S)`$ are called physical operators acting on $`S`$. We require that $`𝒢(M)=𝒢`$, $`𝒢(\mathrm{})=𝐂I`$ (where $`𝐂`$ is the algebra of complex numbers, $`I`$ is the unit element of $`𝒢`$), and $`𝒢(S)𝒢(S^{})`$ if $`SS^{}`$.
2. If $`S_1S_2=\mathrm{}`$ then any two operators $`X_1𝒢(S_1)`$ and $`X_2𝒢(S_2)`$ commute.
The concept of unitarity is well defined in this setting: an element $`U𝒢(M)`$ is called unitary if $`UU^{}=U^{}U=I`$ (the operation $``$ is a part of the $`C^{}`$-algebra structure). Note that nonunitary elements of $`𝒢(M)`$ also have physical meaning because they can be used to construct a unitary operator on a larger space $``$, where $``$ represents some external system (e. g. a measurement device). Such an operator generally has the form $`U=_kA_kB_k`$ ($`A_k𝒢(M)`$), i. e. $`U𝒢(M)𝐋()`$.
Thus we have given a more general interpretation of the locality postulates 1 and 2 which were discussed in the introduction. (We put aside the postulate 3 here).
## 2 Local fermionic modes
Consider $`m`$ sites (numbered $`0`$ through $`m1`$) each of which can be either empty or occupied by a spinless fermionic particle. Such sites will be called local fermionic modes (LFMs). The Hilbert space $``$ of this system, known as Foch space, is spanned by $`2^m`$ basis vectors $`|n_0,\mathrm{},n_{m1}`$, where $`n_j=0,1`$ is the occupation number of the $`j`$-th site. Everything related to fermions can be expressed in terms of annihilation and creation operators $`a_j,a_j^{}`$, ($`j=0,\mathrm{},m1`$). The operator $`a_j`$ acts on basis vectors as follows:
$$\begin{array}{c}a_j|n_0,\mathrm{},n_{j1},1,n_{j+1},\mathrm{},n_{m1}=\left(1\right)^{_{s=0}^{j1}n_s}|n_0,\mathrm{},n_{j1},0,n_{j+1},\mathrm{},n_{m1},\hfill \\ a_j|n_0,\mathrm{},n_{j1},0,n_{j+1},\mathrm{},n_{m1}=\mathrm{\hspace{0.17em}0};\hfill \end{array}$$
(1)
$`a_j^{}`$ is the Hermitian conjugate. Note that the definition depends on the order of LFMs! (One may rather say that the basis depends on the order while $`a_j,a_j^{}`$ do not, since all relations between them are permutation-invariant). The annihilation and creation operators generate the algebra $`\overline{}=𝐋()`$ and have the following commutation rules:
$$\begin{array}{ccc}\hfill a_ja_k+a_ka_j& =& 0,\hfill \\ \hfill a_j^{}a_k^{}+a_k^{}a_j^{}& =& 0,\hfill \\ \hfill a_ja_k^{}+a_k^{}a_j& =& \delta _{jk}.\hfill \end{array}$$
(2)
The Hilbert space of $`m`$ LFMs splits into two parts: $`=_0_1`$, where “0” and “1” refers to the total fermionic parity $`n=_{j=0}^{m1}n_j(mod2)`$. Physical operators are those which preserve the parity. Note that the Hamiltonian of a real Fermi system always satisfies this condition,<sup>2</sup><sup>2</sup>2In electon systems, the Hamiltonian also preserves the electric charge, so terms with different numbers of $`a_j`$ and $`a_j^{}`$ are usually forbidden. Our model is mostly relevant to superconductors where the total charge of excitations is not conserved, so terms like $`a_ja_k`$ appear in the effective Hamiltonian. unlike the operators $`a_j,a_j^{}`$ alone. The algebra of physical operators $`=𝐋(_0)𝐋(_1)`$ is spanned by products of even number of $`a_j,a_j^{}`$. (The notation $`𝒢`$ in the previous section referred to the general case whereas $``$ is reserved for LFMs).
Let $`S\{0,\mathrm{},m1\}`$ be a set of LFMs. Physical operators on S are linear combinations of even products of $`a_j,a_j^{}`$, $`jS`$. The algebra of such operators is $`(S)=\overline{}(S)`$, where $`\overline{}(S)\overline{}`$ is generated by $`a_j,a_j^{}`$, $`jS`$. The conditions (1) and (2) from Sec. 1 are obviously satisfied. Moreover, $`(S_1)`$ commutes with $`\overline{}(S_2)`$ if $`S_1S_2=\mathrm{}`$.
The occupation number $`n_j`$ is an eigenvalue of the operator $`\widehat{n}_j=a_j^{}a_j(\{j\})`$, which means it can be measured locally (by acting on the $`j`$-th LFM and some external device). The occupation number can not be changed locally, though.
Here are some examples of unitary operators acting on one or two LFMs: $`\mathrm{exp}(i\beta a_j^{}a_j)`$ (action by an external potential), $`\mathrm{exp}(i\beta a_j^{}a_ja_k^{}a_k)`$ (two-particle’s interaction), $`\mathrm{exp}(i(\gamma a_j^{}a_k+\gamma ^{}a_k^{}a_j))`$ (tunneling) and $`\mathrm{exp}(i(\gamma a_ka_j+\gamma ^{}a_j^{}a_k^{}))`$ (interaction with a superconductor). We will show that these operators (for all or for some particular values of $`\beta 𝐑`$ and $`\gamma 𝐂`$) form a universal set, i.e. any unitary operator can be represented as a composition of these ones to any given precision, using ancillas.
In our computation model we allow to use ancillas in the state $`|0`$. (To be more accurate, we should say that the input state is padded by some number of zeros to the right. Actually, the order does not matter in this case). If we speak about implementing a unitary operator, the ancillas must return to the state $`|0`$ by the end of the procedure. As is usual, this restriction does not apply to computing a Boolean function which proceeds as follows. One starts from a basis vector $`|n_0,\mathrm{},n_{m1}`$ representing the input data, adds some ancillas, applies some sequence of local unitary operators and reads the result by measuring some of the occupation numbers.
## 3 Relation between LFMs and qubits
The Hilbert space of $`m`$ LFMs can be identified with the Hilbert space of $`m`$ qubits $`^m`$ (where $``$ stands for the two-dimensional space $`𝐂^2`$ endowed with the standard basis $`\{|0,|1\}`$):
$$|n_0,n_1,\mathrm{},n_{m1}|n_0|n_1\mathrm{}|n_{m1},n_j=0,1.$$
(3)
Measurement of $`n_j`$ is the same as eigenvalue measurement of $`\sigma _j^z`$. A physical fermionic operator corresponds to a qubit operator which preserves the parity, i.e. commutes with $`_{j=0}^{m1}\sigma _j^z`$. However, “applying a gate to a set of LFMs” is very different from “applying a gate to a set of qubits”.
Let $`X`$ be a parity-preserving $`p`$-qubit operator acting on qubits numbered $`0`$ through $`p1`$. Applying it to the qubits $`j_0,\mathrm{},j_{p1}`$ is a straightforward procedure. The Hilbert space of $`m`$ qubits $`^m`$ can be identified with $`^p^{(mp)}`$ by the qubit permutation $`P:|n_0,\mathrm{}n_{m1}|n_{j_0},\mathrm{}n_{j_{p1}}|\text{the other}n_j`$. Then the action of $`X`$ is defined as follows:
$$X[j_0,\mathrm{},j_{p1}]=P^1(XI_{^{(mp)}})P.$$
If we want to apply $`X`$ to the LFMs $`j_0,\mathrm{},j_{p1}`$, the procedure is different. First, we should expand $`X`$ into products of $`a_0,\mathrm{},a_{p1},a_0^{},\mathrm{},a_{p1}^{}`$. Then we replace each $`a_r`$ by $`a_{j_r}`$ and each $`a_r^{}`$ by $`a_{j_r}^{}`$. The resulting operator will be denoted by $`X\{j_0,\mathrm{},j_{p1}\}`$. For example, if $`X=|1,00,1|=a_0^{}a_1`$ then $`X\{j,k\}=a_j^{}a_k`$. This operator acts as follows:
$$\begin{array}{ccc}\hfill a_j^{}a_k|\mathrm{},0,n_{j+1},\mathrm{},n_{k1},0,\mathrm{}& =& 0,\hfill \\ \hfill a_j^{}a_k|\mathrm{},0,n_{j+1},\mathrm{},n_{k1},1,\mathrm{}& =& \left(1\right)^{_{s=j+1}^{k1}n_s}|\mathrm{},1,n_{j+1},\mathrm{},n_{k1},0,\mathrm{},\hfill \\ \hfill a_j^{}a_k|\mathrm{},1,n_{j+1},\mathrm{},n_{k1},0,\mathrm{}& =& 0,\hfill \\ \hfill a_j^{}a_k|\mathrm{},1,n_{j+1},\mathrm{},n_{k1},1,\mathrm{}& =& 0.\hfill \end{array}$$
Not only $`X\{j,k\}X[j,k]`$ but also $`X\{j,k\}`$ is non-local in terms of qubits: it involves all the qubits with numbers from $`j`$ to $`k`$.
A unitary qubit gate (LFM, or fermionic gate) is a unitary operator $`U`$ meant to be applied to a number of qubits (LFMs); a $`p`$-qubit or a $`p`$-LFM gate acts on the standard Hilbert space $`^p`$. Operators of the form $`U[j,k]`$ or $`U\{j,k\}`$ are called gate applications. We will usually consider unitary gates up to overall phase factors. A set of gates is also called a basis. A circuit of size $`s`$ in a basis $`𝒜`$ is a composition of $`s`$ applications of gates from $`𝒜`$, i.e. an expression of the form $`U_s\{j_{s,0},\mathrm{},j_{s,p_s1}\}\mathrm{}U_1\{j_{1,0},\mathrm{},j_{1,p_11}\}`$, where $`U_k𝒜`$. Such an expression is evaluated by a unitary operator which is said to be represented by the circuit.
Note that $`X\{j\}=X[j]`$, so one-LFM gates are simply parity-preserving one-qubit gates. Up to an overall phase, such gates have the form $`\mathrm{\Lambda }(e^{i\varphi })`$, where $`\mathrm{\Lambda }(U)`$ denotes the controlled $`U`$. (If $`U`$ acts on $`p`$ qubits then $`\mathrm{\Lambda }(U)`$ acts on $`p+1`$ qubits; in our case $`p=0`$).
More generally, $`X\{j,j+1,\mathrm{},j+p1\}=X[j,j+1,\mathrm{},j+p1]`$. This allows one to represent fermionic gates in terms of qubit gates and vice versa. We will now show how to do that in the case $`p=2`$.
Suppose we want to execute a two-LFM operator $`X\{j,k\}`$ (w. l. .o. g. $`j<k`$). First, we move the $`k`$-th qubit next to the $`j`$-th one by swapping it with its nearest neighbors. Then we apply $`X\{j,j+1\}=X[j,j+1]`$ and move the $`k`$-th qubit back to its original position. However, what we actually need here is to interchange LFMs, not qubits. This is different even if the LFMs (qubits) are next to each other!
A swap between two qubits (with numbers $`0`$ and $`1`$) is defined in the obvious way: $`():|n_0,n_1|n_1,n_0`$. A swap between two LFMs is a unitary operator $`()`$ such that
$$()a_0()^{}=a_1,()a_1()^{}=a_0.$$
(4)
These equations have a unique solution (up to an overall phase factor):
$$()=Ia_0^{}a_0a_1^{}a_1+a_1^{}a_0+a_0^{}a_1=\mathrm{exp}\left(i\frac{\pi }{2}(a_0^{}a_1^{})(a_0a_1)\right):\{\begin{array}{c}|0,0|0,0\hfill \\ |0,1|1,0\hfill \\ |1,0|0,1\hfill \\ |1,1|1,1.\hfill \end{array}$$
(5)
This differs from the qubit swap by the “-” sign. So,
$$()=()D,$$
(6)
where $`D=\mathrm{\Lambda }(\sigma ^z):|a,b(1)^{ab}|a,b`$ is a “swap defect” operator.
To perform the procedure described above, we do not have to actually interchange LFMs or qubits. Instead, we can simply apply operators $`D\{l,r\}=D[l,r]`$ to all pairs of qubits that otherwise would be interchanged. So, any two-LFM operator $`X\{j,k\}`$ ($`j<k`$) can be represented by a qubit circuit as follows:
$$X\{j,k\}=D[k1,k]\mathrm{}D[j+1,k]X[j,k]D[j+1,k]\mathrm{}D[k1,k].$$
(7)
Conversely, any parity-preserving two-qubit operator is represented by a fermionic circuit:
$$X[j,k]=D\{k1,k\}\mathrm{}D\{j+1,k\}X\{j,k\}D\{j+1,k\}\mathrm{}D\{k1,k\}.$$
(8)
This method also works for operators which act on more than two LFMs (qubits).
## 4 A universal set of LFM gates.
We have shown that fermionic gates are equivalent to parity-preserving qubit gates modulo the swap defect operator $`D`$. So, to obtain a universal set of fermionic gates, we only need to find a universal set of parity-preserving qubit gates. It will be possible to represent the operator $`D`$ by these gates too, exactly or approximately. (If only approximate representation is possible, one may have to pay extra cost when simulating a single fermionic gate because the more qubit gates are used, the more accurate they should be. We will avoid this problem, though).
We claim that the following gates are sufficient to represent any parity-preserving operator to any given precision (using ancillas):
$$\mathrm{\Lambda }(e^{i\pi /4}),\mathrm{\Lambda }(\sigma ^z),\stackrel{~}{H}:|a,b\frac{1}{\sqrt{2}}\underset{c}{}(1)^{bc}|a+b+c,c.$$
(9)
(Here $`a,b,c𝐅_2=\{0,1\}`$, so the expression $`a+b+c`$ is considered modulo $`2`$). Note that $`D=\mathrm{\Lambda }(\sigma ^z)`$ belongs to this set of gates.
Any parity-preserving operator $`U`$ can be considered as a pair of operators $`(U_0,U_1)`$, where $`U_0`$ acts on the even sector $`_0`$ whereas $`U_1`$ acts on the odd sector $`_1`$. We will first show how to get a given $`U_0`$ while not caring about $`U_1`$ — we will actually get $`U_1=U_0`$. (The comparison between $`U_0`$ and $`U_1`$ is made through identifying $`_0`$ and $`_1`$ by the map $`\sigma ^x[0]`$). Then we will implement operators of the form $`(I,Y)`$ which can be used to correct the first step.
Any operator $`X`$ on $`m1`$ qubits can be transformed to a parity-preserving operator $`\stackrel{~}{X}`$ on $`m`$ qubits by using the extra qubit to maintain the parity:
$$\stackrel{~}{X}=V^1(I_{}X)V,V:|n_0,n_1,\mathrm{},n_{m1}|n_0+\mathrm{}+n_{m1},n_1,\mathrm{},n_{m1}.$$
(10)
Note that the gate $`\stackrel{~}{H}`$ (see eq. (9)) corresponds to the Hadamard gate $`H=\frac{1}{\sqrt{2}}\left(\begin{array}{cc}\hfill 1& \hfill 1\\ \hfill 1& \hfill 1\end{array}\right)`$. If $`X`$ already preserves the parity, as it is the case with the operators $`\mathrm{\Lambda }(e^{i\pi /4})`$ and $`\mathrm{\Lambda }(\sigma ^z)`$, then $`\stackrel{~}{X}=I_{}X`$. (This equality is actually a characteristic property of parity-preserving operators).
The operator $`V`$ is unitary. It maps the even sector $`_0`$ onto the subspace $`_0`$ which consists of vectors $`|0|\xi `$ ($`|\xi ^{(m1)}`$). Any operator on this subspace extends to an operator of the form $`I_{}X`$. Hence any $`U_0𝐋(_0)`$ extends to an operator of the form $`\stackrel{~}{X}`$. Note that $`V`$ and $`I_{}X`$ commute with $`\sigma ^x[0]`$, so $`\stackrel{~}{X}`$ also commutes with $`\sigma ^x[0]`$. Therefore $`\stackrel{~}{X}=(U_0,U_1)`$, where $`U_1=\sigma ^x[0]U_0\sigma ^x[0]`$, or simply $`U_1=U_0`$ if the identification between $`_0`$ and $`_1`$ is used.
If an operator $`X`$ is represented by a quantum circuit $`A_L\mathrm{}A_1`$, one can replace each gate application $`A_k`$ by $`\stackrel{~}{A_k}`$ to get a quantum circuit for $`\stackrel{~}{X}`$. Indeed, eq. (10) defines a $``$-algebra homomorphism, i.e.
$$\stackrel{~}{X_1+X_2}=\stackrel{~}{X_1}+\stackrel{~}{X_2},\stackrel{~}{cX}=c\stackrel{~}{X},\stackrel{~}{X_1X_2}=\stackrel{~}{X_1}\stackrel{~}{X_2},\stackrel{~}{X^{}}=\stackrel{~}{X}^{},\stackrel{~}{I}=I.$$
It follows that any universal gate set $`𝒜`$ transforms to a set of parity-preserving gates $`\stackrel{~}{𝒜}`$ which are universal on the even sector. The following gate set is known to be universal : $`𝒜=\{\mathrm{\Lambda }(e^{i\pi /4}),\mathrm{\Lambda }(\sigma ^z),H\}`$. The corresponding gate set $`\stackrel{~}{𝒜}`$ is given by eq. (9). (The parity-preserving gates $`\mathrm{\Lambda }(e^{i\pi /4})`$ and $`\mathrm{\Lambda }(\sigma ^z)`$ are copied from $`𝒜`$ to $`\stackrel{~}{𝒜}`$ unchanged). Thus the basis (9) allows one to obtain at least unitary gates of the form $`(U_0,U_0)`$.
With parity-preserving operators, we can still use one of the standard techniques in quantum circuit design — gates with quantum control. Indeed, if $`U`$ preserves the parity then $`\mathrm{\Lambda }(U)`$ also does. More generally, $`\stackrel{~}{\mathrm{\Lambda }(X)}=\mathrm{\Lambda }(\stackrel{~}{X})`$ (up to a permutation of the control qubit and the parity qubit). So, if we can represent $`\mathrm{\Lambda }(X)`$ by a circuit,<sup>3</sup><sup>3</sup>3Implementing $`\mathrm{\Lambda }(X)`$ in the basis $`𝒜`$ is only slightly harder than implementing $`X`$. Indeed, for each gate $`X`$ from the basis $`𝒜`$, the operator $`\mathrm{\Lambda }(X)`$ can be represented by a circuit in the same basis exactly. Therefore, the circuit for $`\mathrm{\Lambda }(X)`$ will be larger than the circuit for $`X`$ only by a constant factor. Note, however, that $`X`$ may have been implemented up to a phase factor; this phase factor becomes important when we consider $`\mathrm{\Lambda }(X)`$. The necessary correction can be achieved by an operator $`\mathrm{\Lambda }(e^{i\varphi })`$ which should be implemented separately. we can also obtain a circuit for $`\mathrm{\Lambda }(\stackrel{~}{X})`$ by the procedure described above. For example, the operator $`\mathrm{\Lambda }(\sigma ^x)[1,2]`$ can be represented as $`H[2]\mathrm{\Lambda }(\sigma ^z)[1,2]H[2]`$, hence
$$\mathrm{\Lambda }(\stackrel{~}{\sigma ^x})[1,0,2]=\stackrel{~}{H}[0,2]\mathrm{\Lambda }(\sigma ^z)[1,2]\stackrel{~}{H}[0,2].$$
Here we use qubit $`0`$ to maintain the parity whereas $`1`$ is the control qubit. (The notation $`\mathrm{\Lambda }(\stackrel{~}{X})[\mathrm{}]`$ suggests that the control qubit goes first).
Now we are in a position to implement operators of the form $`(I,Y)`$ using the gates (9). We will also use one ancilla which will be assigned the number $`m`$. First, we execute the operator
$$\begin{array}{c}W:|n_0,n_1,\mathrm{},n_{m1},n_m|n_0+\mathrm{}+n_{m1},n_1,\mathrm{},n_{m1},n_1+\mathrm{}+n_m,\hfill \\ W=\mathrm{\Lambda }(\stackrel{~}{\sigma ^x})[m1,0,m]\mathrm{}\mathrm{\Lambda }(\stackrel{~}{\sigma ^x})[1,0,m],\hfill \end{array}$$
(11)
where $`\mathrm{\Lambda }(\stackrel{~}{\sigma ^x}):|a,b,c|a,b+a,c+a`$. Now qubit $`0`$ indicates the total parity.
Let $`\stackrel{~}{X}=(Y,Y)`$. Since $`\stackrel{~}{X}`$ can be represented according to (10), and because
$$W=V^1[m,1,\mathrm{},m1]V,\text{where}V=V[0,\mathrm{},m1],$$
we conclude that
$$W^1\mathrm{\Lambda }(\stackrel{~}{X})[0,m,1,\mathrm{},m1]W=V^1\mathrm{\Lambda }(X)[0,1,\mathrm{},m1]V=UI_{},U=(I,Y).$$
(12)
Note that in this case, the ancilla is not affected no matter what its initial state was; one can even use a data qubit as the ancilla.
We have proved that the gate set (9) is universal in the class of parity-preserving operators. It remains to represent these gates in terms of creation and annihilation operators. The first two are simple:
$$\mathrm{\Lambda }(e^{i\pi /4})=\mathrm{exp}\left(i\frac{\pi }{4}a_0^{}a_0\right),\mathrm{\Lambda }(\sigma ^z)=\mathrm{exp}(i\pi a_0^{}a_0a_1^{}a_1).$$
(13)
Unfortunately, the gate $`\stackrel{~}{H}`$ in the fermionic representation looks ugly, so we first represent it as follows:
$$\stackrel{~}{H}[0,1]=\mathrm{\Lambda }(i)[1]\stackrel{~}{G}[0,1]\mathrm{\Lambda }(i)[1],G=\left(\begin{array}{cc}1& i\\ i& 1\end{array}\right).$$
So, the fermionic gates (13), together with
$$\stackrel{~}{G}=\mathrm{exp}\left(i\frac{\pi }{4}(a_0a_0^{})(a_1+a_1^{})\right)=\mathrm{exp}\left(i\frac{\pi }{4}(a_0^{}a_1+a_1^{}a_0)\right)\mathrm{exp}\left(i\frac{\pi }{4}(a_1a_0+a_0^{}a_1^{})\right),$$
(14)
form a universal set. Obviously, this gate set is also universal:
$$\left\{\begin{array}{c}\mathrm{exp}\left(i\frac{\pi }{4}a_0^{}a_0\right),\hfill \\ \mathrm{exp}\left(i\frac{\pi }{4}(a_0^{}a_1+a_1^{}a_0)\right),\hfill \\ \mathrm{exp}\left(i\frac{\pi }{4}(a_1a_0+a_0^{}a_1^{})\right),\hfill \\ \mathrm{exp}(i\pi a_0^{}a_0a_1^{}a_1)\hfill \end{array}\right\}.$$
(15)
## 5 Fast simulation procedures
So far we have been using the standard identification (3) between the Foch space $``$ and the Hilbert space of $`m`$ qubits $`^m`$. This identification has allowed us to consider qubit gate applications $`X[j,k]𝐋(^m)`$ and LFM gate applications $`X\{j,k\}𝐋()`$ as operators acting on the same space. Now we are going to discuss a more general way of simulating LFMs by qubits (or vice versa). Let $`J`$ be an encoding of $`m`$ LFMs by $`m^{}`$ qubits, i.e. $`J:^m^{}`$ is a unitary embedding. ($`J`$ being a unitary embedding means that $`J^{}J=I_{}`$. Note that $`JJ^{}`$ is the projector onto $`=ImJ^m^{}`$). We say that an operator $`U^{}𝐋(^m^{})`$ represents an operator $`U𝐋()`$ if
$$\begin{array}{ccc}& \stackrel{U}{}& \\ J& & J\\ ^m^{}& \stackrel{U^{}}{}& ^m^{}\end{array}\text{commutes},\text{i.e.}JU=U^{}J.$$
(16)
The operator $`U`$, as well as the LFMs or qubits equivalent to them, are called logical. The $`m^{}`$ qubits the operator $`U^{}`$ acts on are called code qubits.
In this section we will show that each LFM gate can be simulated by $`O(\mathrm{log}m)`$ qubit gates. Surprisingly, this does not require quantum codes in the proper sense, i.e. $`J:^m`$ is a map onto. Moreover, $`J`$ takes basis vectors to basis vectors. We will also show how to simulate qubit gates by LFM gates at a constant cost using a subspace of the Foch space.
Gate action on a set of qubits (LFMs) can be described as follows. First, we extract two (or some other number) qubits or LFMs from the quantum memory which is now considered as a “black box”. We place these qubits at positions $`2`$ and $`1`$. Then we apply the gate and put the qubits (LFMs) back into the memory. In this description, it does not matter what gate we apply, a qubit one or a fermionic one (because $`X\{2,1\}=X[2,1]`$). Once extracted, LFMs can be regarded as qubits — the difference is in the way we extract them. With qubits, we apply the operator
$$[j]:^m^m:|n_0\mathrm{}|n_j\mathrm{}|n_{m1}|n_j|n_0\mathrm{}|0\mathrm{}|n_{m1}.$$
This is a unitary embedding into a larger Hilbert space; it can be represented as adding an ancilla in the state $`|0`$ followed by a unitary operator. With fermions, we move the $`j`$-th LFM through all the LFMs left to it, so we should take swap defects into account. Thus we get another unitary embedding:
$$\begin{array}{c}\{j\}:,\hfill \\ \{j\}|n_0,\mathrm{},n_j,\mathrm{},n_{m1}=(1)^{n_j_{s=0}^{j1}n_s}|n_j|n_0,\mathrm{},0,\mathrm{},n_{m1}.\hfill \end{array}$$
(17)
It is easy to verify that our recipe is correct, i.e. applying the operator $`X[2,1]`$ after extracting two LFMs is equivalent to applying $`X\{j,k\}`$ before the extraction:
$$X[2,1](I_{}\{k\})\{j\}=(I_{}\{k\})\{j\}X\{j,k\}.$$
(18)
(Our notations are somewhat confusing so this comment could be helpful. When we extract the $`j`$-th LFM by applying $`\{j\}`$, we add one qubit at position $`1`$. When we then extract the $`k`$-th LFM, this qubit moves to position $`2`$ while another one is being inserted; this can be described by the operator $`I_{}\{k\}`$.)
Suppose we want to simulate LFMs by qubits at low cost. The problem with the standard encoding (3) is that multiplication by the factor $`\left(1\right)^{n_j_{s=0}^{j1}n_s}`$ requires too many operations. The simplest solution would be to store $`y_j=_{s=0}^{j1}n_s`$ instead of $`n_j`$. (Remember that we consider $`n_s`$ as residues$`(mod2)`$, so the sum is also taken$`(mod2)`$). This does solve the problem but also creates a new one: when $`n_j`$ becomes $`0`$ as a result of extraction, we have to modify all $`y_k`$, $`k>j`$. So, we need to balance the complexity of computing $`_{s=0}^{j1}n_s`$ and that of updating the encoded quantum memory when $`n_j`$ changes. This kind of trade-off can be achieved by storing some partial sums $`_{s=a}^bn_s`$.
In general, we will use encodings of the form
$$\begin{array}{c}J:^m:|n_0,\mathrm{},n_{m1}|x_0\mathrm{}|x_{m1},\hfill \\ x_j=\underset{sS(j)}{}n_s,S(j)\{0,\mathrm{},m1\}.\hfill \end{array}$$
We start with an example of such an encoding for $`m=8`$ (the diagram next to the equation illustrates grouping of $`n_s`$ into $`x_j`$):
$$\begin{array}{cccc}x_0=n_0\hfill & x_2=n_2\hfill & x_4=n_4\hfill & x_6=n_6\hfill \\ \multicolumn{2}{c}{x_1=n_0+n_1}& \multicolumn{2}{c}{x_5=n_4+n_5}\\ \multicolumn{4}{c}{x_3=n_0+n_1+n_2+n_3}\\ \multicolumn{4}{c}{x_7=n_0+n_1+n_2+n_3+n_4+n_5+n_6+n_7}\end{array}\overline{)\overline{)\overline{)\overline{)n_0}n_1}\overline{)n_2}n_3}\overline{)\overline{)n_4}n_5}\overline{)n_6}n_7}$$
A binary tree structure is apparent here. To proceed, we will represent the qubit indices $`0,\mathrm{},m1`$ by binary strings. The length of these strings is not fixed; we may add an arbitrary number of zeros to the beginning of a string, e.g. $`3=\overline{11}=\overline{011}=\overline{0011}`$.
Let us define a partial order $``$ on the set of binary strings. We write $`\overline{\alpha _{t1}\mathrm{}\alpha _0}\overline{\beta _{t1}\mathrm{}\beta _0}`$ if $`\alpha _l=\beta _l`$ for $`ll_0`$ while $`\beta _{l_01}=\mathrm{}=\beta _0=1`$ (for some $`l_0`$). For example,
$$\begin{array}{c}\hfill \begin{array}{c}\hfill \overline{000}\overline{001}\\ \hfill \overline{010}\end{array}\}\overline{011}\\ \hfill \overline{100}\overline{101}\\ \hfill \overline{110}\end{array}\}\overline{111},$$
where $`jk`$ means that $`jk`$ but $`jk`$. Note that if $`jk`$ then $`j<k`$.
Now we can specify our encoding for arbitrary $`m`$:
$$|n_0,\mathrm{},n_{m1}|x_0\mathrm{}|x_{m1},x_j=\underset{sj}{}n_s.$$
(19)
It is important that each $`n_s`$ enters only $`O(\mathrm{log}m)`$ of $`x_j`$. Hence applying $`\sigma ^x[s]`$ to one logical qubit amounts to logarithmically many $`\sigma ^x`$ gates being applied to the code qubits.
The inverse transformation (from $`x_j`$ to $`n_s`$) is also simple:
$$\begin{array}{c}n_j=x_j\underset{sK(j)}{}x_s,\\ \overline{\alpha _{t1}\mathrm{}\alpha _0}K\left(\overline{\beta _{t1}\mathrm{}\beta _0}\right)\text{if and only if}\hfill \\ \beta _{l_0}=\mathrm{}=\beta _0=1,\alpha _l=\beta _l\text{for}ll_0,\alpha _{l_0}=0\text{(for some }l_0\text{)}.\hfill \end{array}$$
(20)
The sum in (20) contains only $`O(\mathrm{log}m)`$ terms. Moreover, $`y_j=_{s=0}^{j1}n_s`$ can be also expressed as a sum of $`O(\mathrm{log}m)`$ numbers $`x_s`$:
$$\begin{array}{c}y_j=\underset{sL(j)}{}x_s,\\ \overline{\alpha _{t1}\mathrm{}\alpha _0}L\left(\overline{\beta _{t1}\mathrm{}\beta _0}\right)\text{if and only if}\hfill \\ \beta _{l_0}=1,\alpha _l=\beta _l\text{for}l>l_0,\alpha _{l_0}=0,\alpha _{l_01}=\mathrm{}=\alpha _0=1\text{(for some }l_0\text{)}.\hfill \end{array}$$
(21)
It remains to actually represent the LFM extraction operator $`\{j\}`$ by acting on the code qubits. Since $`\{j\}`$ increases the number of qubits by $`1`$, we should add this extra qubit first. We place it at position $`1`$ and initialize by $`|0`$. Then we make its value equal to $`n_j`$ by applying the operators $`\mathrm{\Lambda }(\sigma ^x)[s,1]`$, $`sK(j)\{j\}`$ (see eq. (20)). After that, $`n_j`$ can be turned into $`0`$ by executing the operators $`\mathrm{\Lambda }(\sigma ^x)[1,k]`$, $`kj`$ (see eq. (19)). Finally, we multiply by $`(1)^{n_jy_j}`$ by applying $`\mathrm{\Lambda }(\sigma ^z)[1,s]`$, $`sL(j)`$ (see eq. (21)). To summarize, we execute the operator
$$U^{}=\underset{sL(j)}{}\mathrm{\Lambda }(\sigma ^z)[1,s]\underset{kj}{}\mathrm{\Lambda }(\sigma ^x)[1,k]\underset{sK(j)\{j\}}{}\mathrm{\Lambda }(\sigma ^x)[s,1]$$
(22)
which has the property $`(I_{}J)\{j\}|\xi =U^{}(|0J|\xi )`$ for any $`|\xi `$. This requires $`O(\mathrm{log}m)`$ operations.
Simulating qubits by LFMs is easier and faster. One can use this simple encoding:
$$|n_0|n_1\mathrm{}|n_{m1}|n_0,n_0,n_1,n_1,\mathrm{},n_{m1},n_{m1},$$
(23)
i.e. each qubit is represented by a pair of LFMs with an even number of fermions, $`|0|00`$, $`|1|11`$. Each two-qubit operator $`X`$ is represented by a four-qubit operator $`X^{}`$. The operators $`X^{}[2j,2j+1,2k,2k+1]`$ and $`X^{}\{2j,2j+1,2k,2k+1\}`$ act the same way on the code subspace. So, when we use the encoding (23), it does not matter whether the code elements are qubits or LFMs. The simulation cost is just a constant.
## 6 Majorana fermions
It is possible, at least mathematically, to split each local fermionic mode into two objects. These halves of LFMs are called Majorana fermions.<sup>4</sup><sup>4</sup>4 In field theory, this term usually means something more specific, but it is sometimes used in this sense too.
Let us introduce a set of Hermitian operators $`c_j`$ ($`j=0,\mathrm{},2m1`$):
$$c_{2k}=a_k+a_k^{}=\sigma ^x[k]\underset{j=0}{\overset{k1}{}}\sigma ^z[j],c_{2k+1}=\frac{a_ka_k^{}}{i}=\sigma ^y[k]\underset{j=0}{\overset{k1}{}}\sigma ^z[j].$$
(24)
These operators satisfy the commutation relations
$$c_jc_k+c_kc_j=2\delta _{jk}.$$
(25)
We can define new locality rules on the algebra $`=𝐋(_0)𝐋(_1)`$ which will be now denoted by $``$. (Also $`\overline{}=𝐋()`$ will be denoted by $`\overline{}`$). We say that there are $`2m`$ sites called Majorana fermions. For each set of sites $`S\{0,\mathrm{},2m1\}`$, let $`\overline{}(S)\overline{}`$ be the subalgebra generated by $`c_j`$, $`jS`$. (Such an algebra is known as complex Clifford algebra). Then $`(S)=\overline{}(S)`$, i. e. physical operators on $`S`$ are linear combinations of even products of $`c_j`$, $`jS`$.
According to this definition, Majorana fermions “exist” in any Fermi system. A nontrivial thing is that it is possible (at least theoretically) to pair them up by interaction, so that few Majorana fermions remain unpaired and separated from each other . Such systems could be used as decoherence-free quantum memory. Indeed, a single Majorana fermion can not interact with the environment by itself (because the operator $`c_j`$ is not physical), so the decoherence can only arise from environment-mediated interaction of two Majorana fermions. But if they are well separated in space, such interaction should be exponentially small (a finite correlation length in the environment is assumed). Roughly speaking, we keep two halves of a qubit apart, so the qubit is decoherence-free!
An application of a Majorana gate is defined as in the case of LFMs. Let $`X`$ be a physical operator acting on $`p`$ Majorana fermions numbered $`0`$ through $`p1`$. Then $`X`$ can be expanded into (even) products of $`c_j`$, $`j=0,\mathrm{},p1`$. If we substitute $`c_{j_r}`$ for $`c_r`$, we will get an operator $`X\{\{j_0,\mathrm{},j_{p1}\}\}(\{j_0,\mathrm{},j_{p1}\})`$.
To find a universal set of Majorana gates, it suffice to rewrite the operators (13), (14) in terms of $`c_0=a_0+a_0^{}`$, $`c_1=i(a_0a_0^{})`$, $`c_2=a_1+a_1^{}`$, $`c_3=i(a_1a_1^{})`$.
$$\begin{array}{c}a_0^{}a_0=\frac{1}{2}(1+ic_0c_1),a_1^{}a_1=\frac{1}{2}(1+ic_2c_3),\\ \mathrm{exp}\left(i\frac{\pi }{4}a_0^{}a_0\right)=e^{i\pi /8}\mathrm{exp}\left(\frac{\pi }{8}c_0c_1\right),\mathrm{exp}\left(i\frac{\pi }{4}(a_0a_0^{})(a_1+a_1^{})\right)=\mathrm{exp}(\frac{\pi }{4}c_1c_2),\\ \mathrm{exp}\left(i\pi a_0^{}a_0a_1^{}a_1\right)=e^{i\pi /4}\mathrm{exp}\left(\frac{\pi }{4}c_0c_1\right)\mathrm{exp}\left(\frac{\pi }{4}c_2c_3\right)\mathrm{exp}\left(i\frac{\pi }{4}c_0c_1c_2c_3\right).\end{array}$$
Hence the following gate set is universal (up to phase factors):
$$\{\mathrm{exp}\left(\frac{\pi }{8}c_0c_1\right),\mathrm{exp}\left(i\frac{\pi }{4}c_0c_1c_2c_3\right)\}.$$
(26)
The second gate in the set (26) describes four-particle interaction. When it comes to physical implementation, this gate will be particularly difficult to realize. Unfortunately, it is not possible to do universal quantum computation by acting on three or fewer Majorana fermions at a time. This way one can only generate the group of unitary operators which act by conjugation as follows:
$$Uc_jU^{}=\underset{k}{}\beta _{jk}c_k,$$
(27)
where $`\beta SO(m)`$, i.e. $`(\beta _{jk})`$ is a real orthogonal matrix with determinant $`+1`$.
## 7 An alternative to the four-particle Majorana gate
In this section we show that, for the purpose of universal computation with Majorana fermions, the gate $`\mathrm{exp}(i\frac{\pi }{4}c_0c_1c_2c_3)`$ can be replaced by a nondestructive eigenvalue measurement of the operator $`c_0c_1c_2c_3`$. (A measurement being nondestructive means that vectors in each of the eigenspaces remain intact; in other words, no extra information is learned or leaks to the environment). Such measurements might be easier to implement, they are also useful in some theoretical application of fermionic computation .
Let us assume that the following operations are possible:
1. Applying the unitary gate $`R=\mathrm{exp}\left(\frac{\pi }{4}c_0c_1\right)`$. (Note that $`Rc_0R^{}=c_1`$, $`Rc_1R^{}=c_0`$).
2. Creation of an ancilla pair in a state which is the eigenstate of $`c_{2k}c_{2k+1}=i(12a_k^{}a_k)`$ corresponding to the eigenvalue $`i`$.
3. Eigenvalue measurement of $`c_jc_k`$.
4. Nondestructive eigenvalue measurement of $`c_jc_kc_rc_s`$.
Moreover, we can base the choice of the next operation on the previous measurement outcomes. (One may call such quantum computation adaptive). Of course, the amount of classical computation involved in this choice should not be too large, so we should better include it into the overall size of the quantum circuit.
Remark. In the standard scheme of quantum computation, measurements in the middle of computation are redundant and can be simulated by unitary gates (e. g. see ). However, this is only true if one uses a universal set of unitary gates. Otherwise measurements and adaptiveness can indeed add extra power to unitary operators.
Suppose we want to apply the operator $`\mathrm{exp}(i\frac{\pi }{4}c_0c_1c_2c_3)`$. Let Majorana fermions 4 and 5 form an ancilla pair, so the input state of the system satisfies
$$(c_4+ic_5)|\mathrm{\Psi }_{\mathrm{in}}=0.$$
(28)
We measure the eigenvalue of $`c_0c_1c_3c_4`$. The outcome is either $`+1`$ or $`1`$, which means that $`|\mathrm{\Psi }_{\mathrm{in}}`$ gets multiplied by the projector $`\mathrm{\Pi }_{+1}^{(4)}=\frac{1}{2}(1+c_0c_1c_3c_4)`$ or $`\mathrm{\Pi }_1^{(4)}=\frac{1}{2}(1c_0c_1c_3c_4)`$, respectively. (More exactly, $`|\mathrm{\Psi }_{\mathrm{in}}p_{\pm 1}^{1/2}\mathrm{\Pi }_{\pm 1}^{(4)}|\mathrm{\Psi }_{\mathrm{in}}`$, where $`p_z`$ is the probability to get outcome $`z`$). Then we measure the eigenvalue of $`c_2c_4`$. The possible eigenvalues $`+i`$ and $`i`$ correspond to the projectors $`\mathrm{\Pi }_{+i}^{(2)}=\frac{1}{2}(1ic_2c_4)`$ and $`\mathrm{\Pi }_i^{(2)}=\frac{1}{2}(1+ic_2c_4)`$. We claim that after some correction depending on the measurements outcomes, we will effectively execute the operator $`\mathrm{exp}(i\frac{\pi }{4}c_0c_1c_2c_3)`$ while leaving the ancilla pair intact. Indeed,
$$\begin{array}{cc}\multicolumn{2}{c}{\mathrm{exp}\left(i\frac{\pi }{4}c_0c_1c_2c_3\right)|\mathrm{\Psi }_{\mathrm{in}}=}\\ \hfill =2\mathrm{exp}\left(\frac{\pi }{4}c_2c_5\right)\frac{1}{2}\left(1ic_2c_4\right)\frac{1}{2}\left(1+c_0c_1c_3c_4\right)|\mathrm{\Psi }_{\mathrm{in}}& =\\ \hfill =2i\mathrm{exp}\left(\frac{\pi }{2}c_0c_1\right)\mathrm{exp}\left(\frac{\pi }{2}c_2c_3\right)\mathrm{exp}\left(\frac{\pi }{4}c_2c_5\right)\frac{1}{2}\left(1ic_2c_4\right)\frac{1}{2}\left(1c_0c_1c_3c_4\right)|\mathrm{\Psi }_{\mathrm{in}}& =\\ \hfill =2i\mathrm{exp}\left(\frac{\pi }{2}c_0c_1\right)\mathrm{exp}\left(\frac{\pi }{2}c_2c_3\right)\mathrm{exp}\left(\frac{\pi }{4}c_2c_5\right)\frac{1}{2}\left(1+ic_2c_4\right)\frac{1}{2}\left(1+c_0c_1c_3c_4\right)|\mathrm{\Psi }_{\mathrm{in}}& =\\ \hfill =2\mathrm{exp}\left(\frac{\pi }{4}c_2c_5\right)\frac{1}{2}\left(1+ic_2c_4\right)\frac{1}{2}\left(1c_0c_1c_3c_4\right)|\mathrm{\Psi }_{\mathrm{in}}& \end{array}$$
(29)
(we have used eq. (28)). Thus we can apply a suitable correction operator $`U_{yz}`$ in each of the four cases ($`U_{+i,+1}=\mathrm{exp}(\frac{\pi }{4}c_2c_5)`$ if the outcomes were $`+1`$ and $`+i`$, etc.) so that
$$\mathrm{exp}\left(i\frac{\pi }{4}c_0c_1c_2c_3\right)|\mathrm{\Psi }_{\mathrm{in}}=\mathrm{\hspace{0.17em}2}U_{yz}\mathrm{\Pi }_y^{(2)}\mathrm{\Pi }_z^{(4)}|\mathrm{\Psi }_{\mathrm{in}}.$$
Each of the four outcome combinations occurs with probability $`2^2=\frac{1}{4}`$. The final state is always the desired one, $`|\mathrm{\Psi }_{\mathrm{fin}}=\mathrm{exp}(i\frac{\pi }{4}c_0c_1c_2c_3)|\mathrm{\Psi }_{\mathrm{in}}`$.
## 8 Superfast simulation of fermions on a graph
The results of Sec. 5 suggest that fermions have slightly more computational power than qubits. The logarithmic slowdown in simulation of fermions seems to be inevitable in the general case. However, in the physical world fermions (e. g. electrons) interact locally not only in the sense that the interaction is pairwise, but also in the geometric sense: a particle can not instantly jump to another position far away. Such physical interactions might be easier to simulate. In this section we study an abstract model of geometrically local interactions. The result is that geometrically local gates can indeed be simulated without any substantial slowdown, i. e. the simulation cost is constant. Therefore one can speculate that, in principle, electrons might not be fundamental particles but, rather, excitations in a (nonperturbative) system bosons. Of course, this is only a logical possibility which may or may not be true.
What follows is a definition of the model. Consider a connected unoriented graph $`\mathrm{\Gamma }=(M,E)`$, where $`M=\{0,\mathrm{},m1\}`$ is the set of vertices, and $`EM\times M`$ is the set of edges. (As the graph is unoriented, $`(j,k)`$ and $`(k,j)`$ either both belong to $`E`$ or both do not belong to $`E`$). We will assume that the degree of each vertex is bounded by some constant $`d`$. Let us put a local fermionic mode on each vertex. We will consider only the even sector of the system, $`_0`$, i. e. the total number of fermions is required to be even. The allowed unitary operations are one-LFM gates and two-LFM gates applied to any pair of vertices connected by an edge.
We are going to identify the Hilbert space $`_0`$ with a codespace of a certain symplectic (stabilizer) code <sup>5</sup><sup>5</sup>5The term “stabilizer code” has become traditional but it is somewhat confusing because any code can be defined in terms of stabilizer operators. (It is actually practical to do so for nonbinary codes related to anyons ). We prefer to use the more specific terms “symplectic code” . so that each elementary operator of the form $`a_k^{}a_k`$, as well as $`a_j^{}a_k`$, $`a_k^{}a_j`$, $`a_ja_k`$ or $`a_k^{}a_j^{}`$, where $`(j,k)E`$, be represented by operators acting on $`O(d)`$ qubits. Then each one-LFM gate and each two-LFM gate applied to neighboring vertices will be also represented by an operator acting on $`O(d)`$ qubits. As $`d=\mathrm{const}`$, this means one can simulate each of the allowed fermionic operations by a constant number of one-qubit and two-qubit gates.
It will be convenient to use the Majorana fermions operators $`c_{2k}`$, $`c_{2k+1}`$ (see eq. (24)) instead of $`a_k`$, $`a_k^{}`$. The list of elementary operators to be represented can be reduced to these ones:
$$\begin{array}{cccc}\hfill B_k& =& ic_{2k}c_{2k+1}\hfill & \text{for each vertex}k,\hfill \\ \hfill A_{jk}& =& ic_{2j}c_{2k}\hfill & \text{for each edge}(j,k)E.\hfill \end{array}$$
(30)
These operators satisfy the following relations:
$$B_k^{}=B_k,A_{jk}^{}=A_{jk},B_k^2=1,A_{jk}^2=1,A_{kj}=A_{jk},$$
(31)
$$B_kB_l=B_lB_k,A_{jk}B_l=(1)^{\delta _{jl}+\delta _{kl}}B_lA_{jk},A_{jk}A_{ls}=(1)^{\delta _{jl}+\delta _{js}+\delta _{kl}+\delta _{ks}}A_{ls}A_{jk},$$
(32)
$$i^pA_{j_0,j_1}A_{j_1,j_2}\mathrm{}A_{j_{p2},j_{p1}}A_{j_{p1},j_0}=\mathrm{\hspace{0.17em}1}\text{for any closed path on the graph}.$$
(33)
It is easy to prove that $`B_k`$, $`A_{jk}`$ modulo these relations generate the algebra of physical operators $`=𝐋(_0)𝐋(_1)`$. However, we are considering only the even sector now. Having been restricted to $`_0`$, the operators $`B_k`$ satisfy an additional relation (which was false in $``$):
$$\underset{k}{}B_k=1.$$
(34)
Hence the algebra $`𝐋(_0)`$ is generated by $`B_k`$, $`A_{jk}`$ modulo the relations (31)–(34).
To construct the code, we put a qubit on each edge of the graph. Thus $`\sigma _{jk}^\alpha =\sigma _{kj}^\alpha `$ denotes the Pauli operator $`\sigma ^\alpha `$$`(\alpha =x,y,z)`$ acting on the edge $`(j,k)`$. The operators $`B_k`$, $`A_{jk}`$ defined above will be identified with some operators $`\stackrel{~}{B}_k`$, $`\stackrel{~}{A}_{jk}`$ acting on the code subspace $``$ (which will be defined later). We start with defining the action of $`\stackrel{~}{B}_k`$ and $`\stackrel{~}{A}_{jk}`$ on the entire Hilbert space of the qubits. Our construction depends on two arbitrary choices. Firstly, we choose orientation for each edge of the graph. This can be described by a matrix $`(ϵ_{jk})`$ such that $`ϵ_{kj}=ϵ_{jk}`$, $`ϵ_{jk}=\pm 1`$ for each edge $`(j,k)E`$. Secondly, for each vertex $`k`$, we order all incident edges $`(j,k)`$. This order will be denoted by $`\underset{k}{<}`$. Now we put
$$\begin{array}{ccc}\hfill \stackrel{~}{B}_k& =& \underset{j:(j,k)E}{}\sigma _{jk}^z,\hfill \\ \hfill \stackrel{~}{A}_{jk}& =& ϵ_{jk}\sigma _{jk}^x\underset{l:(l,j)\underset{j}{<}(k,j)}{}\sigma _{lj}^z\underset{s:(s,k)\underset{k}{<}(j,k)}{}\sigma _{sk}^z.\hfill \end{array}$$
(35)
These operators satisfy the relations analogous to (31), (32) and (34), but not (33).
Finally, we define the code subspace $`^u`$ (where $`u`$ is the number of qubits) by imposing stabilizer conditions:
$$\begin{array}{c}|\psi \text{if and only if}\stackrel{~}{C}_{j_0,\mathrm{},j_{p1}}|\psi =|\psi \text{for any closed path}(j_0,\mathrm{},j_{p1},j_0),\\ \stackrel{~}{C}_{j_0,\mathrm{},j_{p1}}=i^p\stackrel{~}{A}_{j_0,j_1}\stackrel{~}{A}_{j_1,j_2}\mathrm{}\stackrel{~}{A}_{j_{p2},j_{p1}}\stackrel{~}{A}_{j_{p1},j_0}.\end{array}$$
(36)
The stabilizer operators $`\stackrel{~}{C}_{j_0,\mathrm{},j_{p1}}`$ are Hermitian and can be represented in the form $`\pm _{(j,k)}\sigma _{jk}^{\alpha _{jk}}`$, where $`(j,k)`$ runs over a set of different qubits. The set of stabilizer operators is obviously redundant but it is consistent, meaning that (i) they commute with each other, and (ii) whenever the product of several stabilizer operators is a constant, this constant is $`1`$. The number of independent stabilizer operators equals the number of linearly$`(mod2)`$ independent cycles which in turn equals $`um+1`$. Hence
$$dim=2^{u(um+1)}=2^{m1}=dim_0.$$
(37)
The operators $`\stackrel{~}{B}_k`$, $`\stackrel{~}{A}_{jk}`$ commute with $`\stackrel{~}{C}_{j_0,\mathrm{},j_{p1}}`$, so they leave the code subspace invariant. The restrictions of these operators, $`\stackrel{~}{B}_k|_{}`$ and $`\stackrel{~}{A}_{jk}|_{}`$, satisfy the relations analogous to (31)–(34). Thus the correspondence $`B_k\stackrel{~}{B}_k|_{}`$, $`A_{jk}\stackrel{~}{A}_{jk}|_{}`$ extends to a $``$-algebra homomorphism $`\mu :𝐋(_0)𝐋()`$. But $`dim=dim_0`$, hence $`\mu `$ is an isomorphism. It can be represented as $`\mu (X)=JXJ^{}`$, where $`J:_0`$ is a unitary map which is unique up to an overall phase.
Now that the main construction is complete, we can give exact rules for converting the allowed (geometrically local) operations on LFMs into qubit gates. These rules are almost obvious but still worth putting them into a rigorous form. For most generality, consider a two-LFM gate application $`U\{j,k\}`$, where $`(j,k)E`$. Here $`U`$ is a physical operator acting on 2 LFMs ($`=`$ 4 Majorana fermions $`=`$ 2 qubits), so it can be expressed in terms of $`A=ic_0c_2=\sigma ^y[0]\sigma ^x[1]`$, $`B^{}=ic_0c_1=\sigma ^z[0]`$ and $`B^{\prime \prime }=ic_2c_3=\sigma ^z[1]`$. Applying $`U`$ to LFMs $`j`$ and $`k`$ means substituting $`A_{jk}`$, $`B_j`$, $`B_k`$ for $`A`$, $`B^{}`$, $`B^{\prime \prime }`$. Instead of that, we actually do another substitution:
$$\nu _{jk}:A\stackrel{~}{A}_{jk},B^{}\stackrel{~}{B}_j,B^{\prime \prime }\stackrel{~}{B}_k.$$
(38)
It extends to a $``$-algebra homomorphism $`𝒢𝐋(^{S_{jk}})`$, where $`𝒢`$ is the algebra generated by $`A`$, $`B^{}`$ and $`B^{\prime \prime }`$ ($`=`$ the algebra of parity-preserving operators on two qubits), and $`S_{jk}`$ is the set of qubits consisting of all edges incident to $`j`$ and $`k`$. (Note that we do not have to restrict $`\stackrel{~}{A}_{jk}`$, $`\stackrel{~}{B}_j`$, $`\stackrel{~}{B}_k`$ to the subspace $``$ because the cycle relation (33) is irrelevant in this context). So, $`U\{j,k\}`$ is simulated by $`\nu _{jk}(U)`$.
It turns out that this simulation is pretty efficient even if $`d`$ ($`=`$ the largest degree of a vertex in the graph) is not a constant. W. l. o. g. $`m>2`$, so $`S_{jk}`$ contains at least one edge besides $`(j,k)`$, say, $`(j,l)`$. Then $`\stackrel{~}{B}_j`$, $`\stackrel{~}{B}_k`$ and $`\stackrel{~}{A}_{jk}`$ have no nontrivial relations (like $`\stackrel{~}{B}_j\stackrel{~}{B}_k=1`$). More exactly, the map $`\nu _{jk}`$ is injective, so $`\stackrel{~}{B}_j`$, $`\stackrel{~}{B}_k`$ and $`\stackrel{~}{A}_{jk}`$ satisfy exactly the same relations as $`\sigma ^z[0]`$, $`\sigma ^z[1]`$ and $`\sigma ^y[0]\sigma ^x[1]`$. It follows that there is a “symplectic transformation” of the form $`XWXW^{}`$ which takes $`\stackrel{~}{B}_j`$, $`\stackrel{~}{B}_k`$ and $`\stackrel{~}{A}_{jk}`$ to $`\sigma _{jl}^z`$, $`\sigma _{jk}^z`$ and $`\sigma _{jl}^y\sigma _{jk}^x`$. The unitary operator $`W`$ acts on the qubits from $`S_{jk}`$ and can be represented as a product of elementary “symplectic gates”: $`H`$, $`\mathrm{\Lambda }(\sigma ^x)`$ and $`\mathrm{\Lambda }(i)`$. It is easy to show that $`O(d)`$ applications of these gates are sufficient. Hence executing the operator
$$\nu _{jk}(U)=W^{}U[(j,l),(j,k)]W$$
(39)
takes $`O(d)`$ one-qubit and two-qubit gate applications.
In the above analysis, we did not take into account the number of operations required to prepare an initial state $`|\psi `$, which is necessary to begin simulation. For definiteness, we will consider the unique quantum state $`|\xi `$ which satisfies additional constraints:
$$\stackrel{~}{B}_k|\xi =|\xi \text{for each}k.$$
(40)
Note that $`B_k=ic_{2k}c_{2k+1}=12a_k^{}a_k`$, so $`|\xi `$ represents the fermionic state $`|0,\mathrm{},0`$. By a general argument, the qubit state $`|0\mathrm{}|0`$ can be transformed into $`|\xi `$ by applying $`O(u^2)`$ symplectic gates.
However, if we continue the speculations about fermions being possibly mimicked by bosons, the vacuum state of the bosonic system must absorb new degrees of freedom as the Universe expands. Certainly, this process should be reversible, i. e. it should also work when the space shrinks. In our model, shrinking the space corresponds to contracting some edges. More specifically, contracting an edge $`(j,k)`$ means removing it while identifying the vertices $`j`$ and $`k`$. If both $`j`$ and $`k`$ are connected to the same vertex $`l`$, a double edge between $`jk`$ and $`l`$ appears; it must be then reduced to a single one. We should be able to update our “vacuum state” $`|\xi `$ through these transformations. (The qubit being removed should come out in the state $`|0`$). One can show that a single contraction or reduction step requires $`O(d)`$ symplectic gates. This does not involve any nonlocality and can be done simultaneously in different places, which is quite consistent with the idea of adiabatic vacuum transformation in the expanding Universe.
## 9 Quantum codes by Majorana fermions
Some symplectic codes on qubits can be conveniently described in terms of Majorana fermions. We will show how it works for the Shor code . Whether this approach can help to find new codes still remains to be seen.
By inverting the formula (24), we can represent the operators $`a_k,a_k^{}`$ in terms of the Majorana operators: $`a_k=\frac{1}{2}(c_{2k}+ic_{2k+1})`$, $`a_k^{}=\frac{1}{2}(c_{2k}ic_{2k+1})`$. However, one can also introduce another set of annihilation and creation operators which will satisfy the same commutation relations:
$$b_k=\frac{1}{2}\left(c_{\tau (2k)}+ic_{\tau (2k+1)}\right),b_k^{}=\frac{1}{2}\left(c_{\tau (2k)}ic_{\tau (2k+1)}\right),$$
(41)
where $`\tau :\{0,\mathrm{},2m1\}\{0,\mathrm{},2m1\}`$ is arbitrary permutation on $`2m`$ elements. One can define a quantum code by fixing the occupation numbers of some of the new LFMs, e. g. by requiring that each codevector $`|\psi `$ satisfies $`b_k^{}b_k|\psi =0`$ for $`k=1\mathrm{}m1`$. In other words, the set of stabilizer operators is
$$X_k=ic_{\tau (2k)}c_{\tau (2k+1)},k=1,\mathrm{}m1.$$
(42)
The number of encoded qubits is $`m(m1)=1`$. The logical operators for this code are generated by $`c_{\tau (0)}`$ (the encoded $`\sigma ^x`$) and $`c_{\tau (1)}`$ (the encoded $`\sigma ^y`$).
Thus each permutation $`\tau :\{0,\mathrm{},2m1\}\{0,\mathrm{},2m1\}`$ defines a quantum code which encodes $`1`$ qubit into $`m`$ qubits. It turns out that these codes can have arbitrary large code distances. We are to define a family of such codes which can be considered as slightly modified Shor codes. Let $`l2`$ be an integer (the index of a code in the family), $`m=l^2`$. There will be two types of stabilizer operators:
$$\begin{array}{cccc}\hfill Z_k& =& ic_{2kl+1}c_{2(k+2)l2},\hfill & k=0\mathrm{}l2,\hfill \\ \hfill Y_{k,j}& =& ic_{2kl+2j+3}c_{2kl+2j},\hfill & k=0,\mathrm{}l1,j=0,\mathrm{}l2.\hfill \end{array}$$
(43)
They can be expressed in terms of the Pauli operators as follows :
$$Z_k=\sigma ^x[kl]\sigma ^x[(k+2)l1]\underset{s=kl+1}{\overset{(k+2)l2}{}}\sigma ^z[s],Y_{k,j}=\sigma ^y[kl+j]\sigma ^y[kl+j+1]$$
(44)
For example, let us consider the $`l=3`$ code. Its 9 qubits can be arranged into a $`3\times 3`$ array as follows
$$\begin{array}{ccc}0,& 1,& 2,\\ 3,& 4,& 5,\\ 6,& 7,& 8.\end{array}$$
Then the stabilizer operators of the first type become:
$$Z_0=\left(\begin{array}{ccc}\sigma ^x,& \sigma ^z,& \sigma ^z,\\ \sigma ^z,& \sigma ^z,& \sigma ^x,\\ I,& I,& I\end{array}\right),Z_1=\left(\begin{array}{ccc}I,& I,& I,\\ \sigma ^x,& \sigma ^z,& \sigma ^z,\\ \sigma ^z,& \sigma ^z,& \sigma ^x\end{array}\right).$$
(These are not matrices. We just mean that, for example, $`Z_0=\sigma ^x\sigma ^z\sigma ^z\sigma ^z\sigma ^z\sigma ^xIII`$). The stabilizer operators of the second type are
$$\begin{array}{c}Y_{0,0}=\left(\begin{array}{ccc}\sigma ^y,& \sigma ^y,& I,\\ I,& I,& I,\\ I,& I,& I\end{array}\right),Y_{0,1}=\left(\begin{array}{ccc}I,& \sigma ^y,& \sigma ^y,\\ I,& I,& I,\\ I,& I,& I\end{array}\right),Y_{1,0}=\left(\begin{array}{ccc}I,& I,& I,\\ \sigma ^y,& \sigma ^y,& I,\\ I,& I,& I\end{array}\right),\text{}\\ Y_{1,1}=\left(\begin{array}{ccc}I,& I,& I,\\ I,& \sigma ^y,& \sigma ^y,\\ I,& I,& I\end{array}\right),Y_{2,0}=\left(\begin{array}{ccc}I,& I,& I,\\ I,& I,& I,\\ \sigma ^y,& \sigma ^y,& I\end{array}\right),Y_{2,1}=\left(\begin{array}{ccc}I,& I,& I,\\ I,& I,& I,\\ I,& \sigma ^y,& \sigma ^y\end{array}\right).\end{array}$$
If one performs the cyclic permutation $`\sigma ^x\sigma ^y\sigma ^z\sigma ^x`$, the operators $`Y_{k,j}`$ turn into certain stabilizer operators of the Shor code. The operators $`Z_k`$ will not become exactly the same as in the Shor code. Each of them will contain two $`\sigma ^y`$ instead of $`\sigma ^x`$. However, the code distance of this code is the same, namely, $`3`$.
For arbitrary $`l`$, the code distance is $`d_l=l`$. (The proof is essentially the same as for the Shor code).
|
warning/0003/quant-ph0003043.html
|
ar5iv
|
text
|
# The Haroche-Ramsey experiment as a generalized measurement
## I Introduction
In the standard formalism of quantum mechanics a measurement corresponds to a self-adjoint operator yielding its eigenvalues as measurement results, with probabilities determined by the expectation values of the projections on its eigenvectors. It is by now well-known that this formalism is too restricted to encompass all possible experiments within the domain of application of quantum mechanics. Many experiments performed in actual practice are of the type of generalized measurements yielding probabilities determined by the expectation values of so-called positive operator-valued measures (POVMs) rather than projection-valued ones. The theory describing this generalization is referred to as the operational approach.
Such generalized measurements are important for two different reasons. First, from a fundamental point of view they are interesting because they enable to transcend the bounds imposed by the standard formalism on the notion of a quantum mechanical measurement. In particular, the projection postulate does not make sense in a generalized measurement because in general no projection operators are available. The inapplicability of this postulate in most realistic measurement procedures indeed demonstrates the restricted scope of the standard formalism. What is a quantum mechanical measurement, and which observable is measured by a particular measurement procedure must be determined by considering in detail the interaction of the microscopic object and the measuring instrument as a quantum mechanical process. In the second place generalized measurements are interesting from a practical point of view because they can yield more information on the state of the object than is provided by the measurement of a standard observable.
We should distinguish between the determinative and the preparative aspects of measurement . The former is related to the final state of the measuring instrument, and the information on the initial object state that is provided by the measurement. The latter refers to the preparation of the post-measurement state of the object. Unfortunately, these different aspects of measurement have been confounded ever since in the Copenhagen interpretation a measurement has been defined as a preparation of the object in a final state described by an eigenvector of the measured observable. It is important to note that most measurement procedures do not satisfy this criterion. Due to the ensuing confusion even recently the role of the Heisenberg uncertainty relations in quantum measurement has been a source of controversy . Whereas Storey et al. conclude that “the principle of complementarity is a consequence of the Heisenberg uncertainty relation,” Scully et al. observe that “The principle of complementarity is manifest although the position-momentum uncertainty relation plays no role.” Dürr et al. stress that quantum correlations due to the interaction of object and detector, rather than “classical” momentum transfer, enforces the loss of interference in a ‘which-way’ measurement. In their experiment momentum disturbance is not large enough to account for the loss of interference if the measurement arrangement is changed so as to yield ‘which-way’ information. These diverging statements can easily be reconciled if it is realized that the Heisenberg inequalities refer to the initial state of the object, and, as already stressed a long time ago by Ballentine , do not refer to the measurement procedure (although, of course, the post-measurement state of the object will once again satisfy the Heisenberg inequalities if measurements are performed in this latter state). It should be realized, however, that in general there need not exist a direct relation between the determinative properties of a measurement, yielding information on the initial state of the object, and its preparative ones, determining what is the final object state.
The notion of a generalized measurement is able to clarify the confusion with respect to the role of the Heisenberg uncertainty relations in quantum measurement . Whereas the standard formalism only allows the joint measurement of compatible (standard) observables, is the generalized formalism able to deal with incompatible ones. Moreover, from this formalism an inequality has been derived (equation (4) below) quantifying the mutual disturbance of the information on the initial state, that is a consequence of the incompatibility of the observables measured jointly, and that, contrary to the Heisenberg inequality, is a property of the measurement procedure. In recent years quite a few measurement procedures have been analyzed satisfying the characteristics of such joint measurements: for instance, eight-port optical homodyning , certain neutron interference experiments , Stern-Gerlach experiments , and a number of Mach-Zehnder interferometric procedures are of the generalized type, allowing an interpretation as a joint measurement of incompatible observables. Presumably also the experiment by Dürr et al. is of this type, the observables measured jointly being different from the position-momentum pair.
In this paper we want to consider a recent atomic beam experiment from the same point of view, and demonstrate that also this experiment has a generalized character. In the experiment, to be referred to in the following as the Haroche-Ramsey experiment, a Rb atom is sent through three cavities, $`R_1,C`$ and $`R_2`$ (figure 1), $`R_1`$ and $`R_2`$ being approximately resonant with a particular transition between two Rydberg states of the atom. Whereas the experiment without cavity $`C`$ is a pure interference experiment, already performed by Ramsey , the introduction of cavity $`C`$ provides the possibility to obtain also ‘which way’ information. The visibility of the interference fringes decreases as the field amplitude $`\gamma `$ increases, but it only vanishes in the limit $`\gamma \mathrm{}.`$ So, for finite $`\gamma `$ information on both interference and path can be obtained. In order to actually obtain ‘which way’ information, a measurement must be carried out on the field in cavity $`C`$ left behind by the atom. The Haroche-Ramsey experiment is very analogous to the neutron interference experiments performed by Summhammer et al. , in which an absorber is inserted into one of the paths, the absorber playing an analogous role as the cavity $`C`$ field. Consequently the present analysis is similar to the analysis of the neutron interference experiments performed in .
Recently there has been much interest in the problem of reconstructing the initial state of the object on the basis of information provided by quantum measurements . Such a reconstruction is impossible on the basis of a measurement of a single standard observable. As demonstrated by Vogel and Risken in quantum tomography this goal may be achieved by measuring a number of such observables. Generalized measurements provide the opportunity to obtain a comparable result using one single measurement arrangement. In general a generalized measurement need not allow a complete reconstruction of the initial state. It then is an interesting question which information is provided by the measurement . Also this subject will be discussed using the Haroche-Ramsey experiment as an example. The analysis is particularly suited to study the question of decoherence of the C field raised by Davidovich et al. . In order to experimentally study the decoherence a second atom is sent through the cavities, yielding information on the state of the cavity $`C`$ field as it was at the time of passage. From the experimental two-atom correlation found in the experiment it was concluded that a decoherence effect exists that cannot be explained by loss of photons from cavity $`C`$. From an analysis of the experiment as a generalized measurement it can be seen, however, that such a conclusion cannot be drawn, because as a measurement of the cavity $`C`$ field the measurement on the second atom can only provide non-ideal information on photon number.
The paper is organized as follows. In section II the theory of generalized measurements, and its application to joint non-ideal measurement of incompatible observables is briefly reviewed. In section III the Haroche-Ramsey experiment and the Davidovich-Haroche experiments are analyzed as generalized measurements. In section IV we demonstrate that the Davidovich-Haroche experiment is informationally equivalent to a Haroche-Ramsey experiment in which a measurement of cavity $`C`$ photon number is performed in coincidence with a determination of the final state of the atom. In this section the decoherence aspects of the Haroche-Ramsey experiment will be dealt with from the point of view of generalized measurements. In section V an alternative measurement procedure for the Haroche-Ramsey experiment is discussed, that can be interpreted as a joint measurement of incompatible observables having the complementary character of the “classical” double-slit experiments.
## II Generalized measurements
### A Operational approach
In the operational approach the experimental probabilities are calculated by treating the interaction between object and measuring instrument as a quantum mechanical process. Quantum mechanical measurement results are associated with pointer positions of the latter. If $`\widehat{\rho }`$ and $`\widehat{\rho }_a`$ are the initial density operators of object and measuring instrument, respectively, then the probability of a measurement result is obtained as the expectation value of the spectral representation $`\{\widehat{E}_m^{(a)}\}`$ of a pointer observable of the measuring instrument in the final state $`\widehat{\rho }_f=\widehat{U}\widehat{\rho }\widehat{\rho }_a\widehat{U}^{},\widehat{U}=exp(i\widehat{H}T_f/\mathrm{})`$ of the measurement. Thus, $`p_m=Tr_{oa}\widehat{\rho }_f\widehat{E}_m^{(a)}`$. This quantity can be interpreted as a property of the initial object state, $`p_m=Tr_o\widehat{\rho }\widehat{M}_m`$, with $`\widehat{M}_m=Tr_a\widehat{\rho }_a\widehat{U}^{}\widehat{E}_m^{(a)}\widehat{U}`$. Whereas in the standard formalism quantum mechanical probabilities $`p_m`$ are represented by the expectation values of mutually commuting projection operators ($`p_m=Tr\widehat{\rho }\widehat{E}_m,\widehat{E}_m^2=\widehat{E}_m,[\widehat{E}_m,\widehat{E}_m^{}]_{}=\widehat{O}`$), allows the generalized formalism to represent these probabilities by expectation values of operators $`\widehat{M}_m`$ that are not necessarily projection operators, and need not commute ($`\widehat{M}_m^2\widehat{M}_m,[\widehat{M}_m,\widehat{M}_m^{}]_{}\widehat{O}`$ in general). The operators $`\widehat{M}_m,\widehat{O}\widehat{M}_m\widehat{I},_m\widehat{M}_m=\widehat{I}`$ generate a POVM; the observables of the standard formalism are restricted to those POVMs of which the elements are mutually commuting projection operators (so-called projection-valued measures (PVM)).
### B Non-ideal measurements
In the generalized formalism it is possible to define a relation of partial ordering between observables, expressing that the measurement represented by one POVM can be interpreted as a non-ideal measurement of another one . Thus, a POVM $`\{\widehat{R}_m\}`$ is representing a non-ideal measurement of the (generalized or standard) observable $`\{\widehat{M}_m^{}\}`$ if the following relation holds between the elements of the POVMs:
$$\widehat{R}_m=\underset{m^{}}{}\lambda _{mm^{}}\widehat{M}_m^{},\lambda _{mm^{}}0,\underset{m}{}\lambda _{mm^{}}=1.$$
(1)
The matrix $`(\lambda _{mm^{}})`$ is the non-ideality matrix. It is a so-called stochastic matrix. Its elements $`\lambda _{mm^{}}`$ can be interpreted as conditional probabilities of finding measurement result $`m`$ if an ideal measurement had yielded measurement result $`m^{}`$. In case of an ideal measurement the non-ideality matrix $`(\lambda _{mm^{}})`$ reduces to the unit matrix $`(\delta _{mm^{}})`$.
Non-ideality relations of type (1) are well-known from the theory of transmission channels in the classical theory of stochastic processes , where the non-ideality matrix describes the crossing of signals between subchannels. It should be noted, however, that, notwithstanding the classical origin of the latter subject, the non-ideality relation (1) may be of a quantum mechanical nature. Shannon’s channel capacity can be used also in this latter case to quantify the deviation of a non-ideal measurement from ideality . Another useful measure of the departure of a non-ideality matrix from the unit matrix is the average row entropy of the non-ideality matrix $`(\lambda _{mm^{}})`$,
$$J_{(\lambda )}=\frac{1}{N}\underset{mm^{}}{}\lambda _{mm^{}}\mathrm{ln}\frac{\lambda _{mm^{}}}{_{m^{\prime \prime }}\lambda _{mm^{\prime \prime }}},$$
(2)
which (restricting to square $`N\times N`$ matrices) satisfies the following properties:
$`\begin{array}{c}0J_{(\lambda )}\mathrm{ln}N,\hfill \\ J_{(\lambda )}=0\text{ if }\lambda _{mm^{}}=\delta _{mm^{}},\hfill \\ J_{(\lambda )}=\mathrm{ln}N\text{ if }\lambda _{mm^{}}=\frac{1}{N}.\hfill \end{array}`$
Hence, the quantity $`J_{(\lambda )}`$ vanishes in case of an ideal measurement of observable $`\{\widehat{M}_m^{}\}`$, and obtains its maximal value if the measurement is uninformative (i.e. does not yield any information on the observable measured non-ideally; this is the case if in (1) $`\lambda _{mm^{}\text{ }}`$ is independent of $`m^{}`$, and, hence, $`\widehat{R}_m`$=$`\alpha _m\widehat{I},`$ $`\alpha _m`$ constants for all $`m`$) due to maximal disturbance of the measurement results. In the following we shall use the non-ideality measure (2).
### C Joint non-deal measurement of incompatible observables
Within the generalized formalism of POVMs it is possible to extend the notion of quantum mechanical measurement to the joint measurement of two (generalized) observables. Such a measurement is required to yield a bivariate joint probability distribution $`p_{mn}`$, satisfying $`p_{mn}0,_{mn}p_{mn}=1`$. Here $`m`$ and $`n`$ label the possible values of the two observables measured jointly, corresponding to pointer positions of two different pointers (one for each observable) being jointly read for each individual preparation of an object. It is assumed that, analogous to the case of single measurement, the probabilities $`p_{mn}`$ of finding the pair $`(m,n)`$ are represented in the formalism by the expectation values $`Tr\widehat{\rho }\widehat{R}_{mn}`$ of a bivariate POVM $`\{\widehat{R}_{mn}\},\widehat{R}_{mn}\widehat{O},_{mn}\widehat{R}_{mn}=\widehat{I}`$ in the initial state $`\widehat{\rho }`$ of the object. Then the marginal probabilities $`\{_np_{mn}\}`$ and $`\{_mp_{mn}\}`$ are expectation values of POVMs $`\{\widehat{M}_m=_n\widehat{R}_{mn}\}`$ and $`\{\widehat{N}_n=_m\widehat{R}_{mn}\}`$, respectively, which correspond to the (generalized) observables measured jointly. A measurement, represented by a bivariate POVM $`\{\widehat{R}_{mn}\}`$, can be interpreted as a joint non-ideal measurement of the observables $`\{\widehat{E}_m\}`$ and $`\{\widehat{F}_n\}`$ if the marginals $`\{_n\widehat{R}_{mn}\}`$ and $`\{_m\widehat{R}_{mn}\}`$ of the bivariate POVM $`\{\widehat{R}_{mn}\}`$ represent non-ideal measurements of observables $`\{\widehat{E}_m\}`$ and $`\{\widehat{F}_n\}`$, respectively. Then, in accordance with (1) two non-ideality matrices $`(\lambda _{mm^{}})`$ and $`(\mu _{nn^{}})`$ exist, such that
$$\begin{array}{c}_n\widehat{R}_{mn}=_m^{}\lambda _{mm^{}}\widehat{E}_m^{},\lambda _{mm^{}}0,_m\lambda _{mm^{}}=1,\hfill \\ _m\widehat{R}_{mn}=_n^{}\mu _{nn^{}}\widehat{F}_n^{},\mu _{nn^{}}0,_n\mu _{nn^{}}=1.\hfill \end{array}$$
(3)
It is possible that $`\{\widehat{E}_m\}`$ and $`\{\widehat{F}_n\}`$ are standard observables. It will be demonstrated in the following sections that the Haroche-Ramsey experiment satisfies the joint measurement scheme given above.
If $`\{\widehat{E}_m\}`$ and $`\{\widehat{F}_n\}`$ are standard observables, then the non-idealities expressed by the non-ideality matrices $`(\lambda _{mm^{}})`$ and $`(\mu _{nn^{}})`$ can be proven to satisfy the characteristic traits of the type of complementarity that is due to mutual disturbance of measurement results (or inaccuracy) in a joint measurement of incompatible observables. Quantifying the non-idealities of the non-ideality matrices $`(\lambda _{mm^{}})`$ and $`(\mu _{nn^{}})`$ by the average row entropy (2), it can be demonstrated that for a joint non-ideal measurement of two standard observables $`\widehat{A}=_ma_m\widehat{E}_m`$ and $`\widehat{B}=_nb_n\widehat{F}_n`$, with eigenvectors $`|a_m`$ and $`|b_n`$, respectively, the non-ideality measures $`J_{(\lambda )}`$ and $`J_{(\mu )}`$ obey the following inequality:
$$J_{(\lambda )}+J_{(\mu )}2\mathrm{ln}\{max_{mn}|a_m|b_n|\}.$$
(4)
It is evident that (4) is a nontrivial inequality (the right-hand side unequal to zero) if the two observables $`\widehat{A}`$ and $`\widehat{B}`$ are incompatible in the sense that the operators do not commute. Contrary to the Heisenberg inequality $`\mathrm{\Delta }\widehat{A}\mathrm{\Delta }\widehat{B}\frac{1}{2}Tr\rho [\widehat{A},\widehat{B}]_{}`$, inequality (4) does not refer to the preparation of the initial state but exclusively to the measurement process. Inequality (4) should also be clearly distinguished from the entropic uncertainty relation for the standard observables $`\widehat{A}=_ma_m\widehat{E}_m`$ and $`\widehat{B}=_nb_n\widehat{F}_n`$,
$$H_{\{\widehat{E}_m\}}(\widehat{\rho })+H_{\{\widehat{F}_n\}}(\widehat{\rho })2\mathrm{ln}\{max_{mn}|a_m|b_n|\},$$
(5)
in which $`H_{\{\widehat{E}_m\}}(\widehat{\rho })=_mp_m\mathrm{ln}p_m,p_m=Tr\widehat{\rho }\widehat{E}_m`$ (and analogously for $`\widehat{B}`$). Inequality (5), although quite similar to inequality (4), should be compared to the Heisenberg inequality, expressing a property of the initial state $`\widehat{\rho }`$, to be tested by means of separate measurements of observables $`\{\widehat{E}_m\}`$ and $`\{\widehat{F}_n\}`$.
### D Informational aspects
The operators $`\widehat{M}_m`$ of a POVM span a subspace $`_{\{\widehat{M}_m\}}`$ of the linear space of bounded operators. For finite-dimensional systems this can be the Hilbert-Schmidt space $`_{HS}`$ having inner product $`Tr\widehat{A}^{}\widehat{B}.`$ If the operators $`\widehat{M}_m`$ are linearly independent they constitute a (generally non-orthogonal) basis of the subspace. Within subspace $`_{\{\widehat{M}_m\}}`$ a bi-orthogonal system is defined by
$$\widehat{M}_m^{}^{}=\underset{m}{}\beta _{m^{}m}\widehat{M}_m,Tr\widehat{M}_m\widehat{M}_m^{}^{}=\delta _{mm^{}},\beta _{m^{}m}\text{real}.$$
(6)
In general the set of Hermitian operators $`\widehat{M}_m^{}^{}`$ constitutes another non-orthogonal basis of $`_{\{\widehat{M}_m\}}`$. The information the measurement of POVM $`\{\widehat{M}_m\}`$ yields on the initial state $`\widehat{\rho }`$ of the object can be represented by the projection $`\widehat{\rho }_{\{\widehat{M}_m\}}=𝒫_{\{\widehat{M}_m\}}\widehat{\rho }`$ of $`\widehat{\rho }`$ onto $`_{\{\widehat{M}_m\}}.`$ This projection is given by
$$𝒫_{\{\widehat{M}_m\}}\widehat{\rho }=\underset{m}{}(Tr\widehat{M}_m^{}\widehat{\rho })\widehat{M}_m=\underset{m}{}(Tr\widehat{M}_m\widehat{\rho })\widehat{M}_m^{}.$$
(7)
For complete measurements we have $`𝒫_{\{\widehat{M}_m\}}\widehat{\rho }=\widehat{\rho }.`$ Measurements are less complete as the subspace $`_{\{\widehat{M}_m\}}`$has a smaller dimension. It should be noted that, contrary to an assertion made in , for incomplete measurements $`\widehat{\rho }_{\{\widehat{M}_m\}}`$ need not be a density operator, even though $`Tr\widehat{\rho }_{\{\widehat{M}_m\}}=1.`$ In general $`\widehat{\rho }_{\{\widehat{M}_m\}}`$ is not a non-negative operator if the object Hilbert space has dimension greater than $`2`$ (in the Appendix it is proven that for two-dimensional Hilbert spaces $`\widehat{\rho }_{\{\widehat{M}_m\}}`$ is non-negative). Hence, $`\widehat{\rho }_{\{\widehat{M}_m\}}`$ should be compared to a description of a quantum state by means of the Wigner distribution, yielding $`Tr\widehat{\rho }\widehat{A}=Tr\widehat{\rho }_{\{\widehat{M}_m\}}\widehat{A}`$ for all operators $`\widehat{A}_{\{\widehat{M}_m\}},`$ but not containing any information on the part of $`\widehat{\rho }`$ that is in the orthogonal complement of $`_{\{\widehat{M}_m\}}.`$ In principle the subspace $`_{\{\widehat{M}_m\}}`$ completely determines the information on the initial density operator $`\widehat{\rho }`$ that can be retrieved by a measurement of POVM $`\{\widehat{M}_m\}`$.
## III Atomic beam interference experiments
### A The Ramsey experiment
In the Ramsey experiment a beam of Rb atoms is sent through two identical cavities $`R_1`$ and $`R_2`$. The relevant Hilbert space $``$ of a Rb atom is spanned by the orthogonal state vectors $`|e`$ and $`|g.`$ These correspond to circular Rydberg levels with principal quantum numbers $`n=51`$ and $`n=50,`$ respectively (transition frequency $`\omega _0=\omega _e\omega _g=32110^9`$ $`rad/s`$). The frequency of the classical microwave fields in the cavities is denoted by $`\omega `$, its amplitude by $`\mathrm{\Omega }`$ (the Rabi frequency), and the time needed by an atom to pass one cavity by $`T`$. The unitary transformation $`\widehat{U}_i`$ describing the evolution of the state of a Rb atom while passing cavity $`R_i`$ between $`t=t_i`$ and $`t_i+T`$ is given in the $`\{|e,|g\}`$-representation by the matrix
$$\widehat{U}_i=\left(\begin{array}{cc}e^{i\left(\frac{\nu }{2}+\omega _e\right)T}S_1\hfill & e^{i\left(\frac{\nu }{2}+\omega _e\right)Ti\omega t_i}S_2\hfill \\ e^{i\left(\frac{\nu }{2}\omega _g\right)T+i\omega t_i}S_2\hfill & e^{i\left(\frac{\nu }{2}\omega _g\right)T}S_1^{}\hfill \end{array}\right)$$
(8)
where $`S_1=\mathrm{cos}\frac{aT}{2}+\frac{i\nu }{a}\mathrm{sin}\frac{aT}{2}`$, $`S_2=\frac{i\mathrm{\Omega }}{a}\mathrm{sin}\frac{aT}{2}`$ , $`a=\sqrt{\nu ^2+\mathrm{\Omega }^2}`$, and $`\nu =\omega \omega _0`$ the detuning parameter. A derivation of (8) can be found in Ramsey and Paul . For all values of the parameters we have $`|S_1|^2+|S_2|^2=1`$. The Rb atom is said to undergo a $`\frac{\pi }{2}`$ pulse in cavity $`R_i`$ if $`|S_1|^2=|S_2|^2=\frac{1}{2}`$. We shall introduce a parameter $`\delta =|S_1|^2|S_2|^2`$, quantifying experimental deviation from the $`\frac{\pi }{2}`$ pulse condition. Note that satisfaction of this latter condition does not imply $`\nu =0.`$ The phase factor $`e^{i\omega t_i}`$ in (8) takes into account the phase of the microwave field at the moment the atom enters the cavity.
Let $`|\psi _{in}`$ be the initial state of the atom. If the standard observable $`\{|ee|,|gg|\}`$ is measured after the atom has passed cavity $`R_1,`$ then the probabilities $`p_e`$ and $`p_g`$ can be related to the initial state by means of the equalities
$$\begin{array}{c}p_e=\psi _{in}|\left(\widehat{U}_1^{}|ee|\widehat{U}_1\right)|\psi _{in}=\psi _{in}\left|\widehat{P}_+\right|\psi _{in},\\ p_g=\psi _{in}|\left(\widehat{U}_1^{}|gg|\widehat{U}_1\right)|\psi _{in}=\psi _{in}\left|\widehat{P}_{}\right|\psi _{in}.\end{array}$$
(9)
Due to the unitarity of $`\widehat{U}_1`$ this yields PVM $`\{\widehat{P}_+,\widehat{P}_{}\}`$ as the POVM of this experiment, with
$$\widehat{P}_+=|p_+p_+|=\widehat{I}\widehat{P}_{}\text{ };\text{ }|p_+=\left(\begin{array}{c}S_1^{}\\ S_2^{}e^{i\omega t_1}\end{array}\right),$$
(10)
in which $`|p_+`$ equals $`\widehat{U}_1^{}|e`$ up to a phase factor. Because of the analogy with neutron interference experiments the observable $`\{\widehat{P}_+,\widehat{P}_{}\}`$ will be referred to as the path observable, even though here the paths are not trajectories in configuration space (as it is in the double-slit experiment and the neutron interference experiments) but in the Hilbert space $``$ of the internal states of the Rb atom. Mathematically this does not constitute a difference, however.
The observable $`\{\widehat{P}_+,\widehat{P}_{}\}`$ is dependent on the initial phase of the microwave field. For this reason it is not allowed to ignore this phase if the experiment is intended to yield a measurement of the initial state of the atom. If the atoms are prepared at random phases, then a measurement of $`\{|ee|,|gg|\},`$ performed immediately after the atom has passed cavity $`R_1,`$ will yield probabilities obtained from (9) by phase averaging. The corresponding POVM $`\{\overline{\widehat{P}_+},\overline{\widehat{P}_{}}\}`$ is found according to
$`\overline{\widehat{P}_+}`$ $`=`$ $`|S_1|^2|ee|+|S_2|^2|gg|,`$
$`\overline{\widehat{P}_{}}`$ $`=`$ $`|S_2|^2|ee|+|S_1|^2|gg|,`$
and, hence, represents a non-ideal measurement of PVM $`\{|ee|,|gg|\}`$ in the sense of (1). Note that $`\{\overline{\widehat{P}_+},\overline{\widehat{P}_{}}\}`$ is uninformative in case the $`\frac{\pi }{2}`$ pulse condition is satisfied, since then its expectation values do not yield any information on $`|\psi _{in}`$.
The Ramsey set-up consists of two cavities $`R_i,i=1,2`$, entered by the atom at times $`t_i`$ (with $`t_2=t_1+T+\tau ,\tau >0`$). If the initial state of the Rb atom is
$$|\psi _{in}=\alpha |e+\beta |g,$$
(11)
then the final state at time $`t_2+T`$ is:
$`|\psi _f=\left[\alpha \left(E+F\right)+\beta \left(G+H\right)\right]|e+\left[\alpha \left(J+K\right)+\beta \left(L+M\right)\right]|g`$
where we have used the abbreviations
| $`E=S_1^2e^{i\left(\nu +2\omega _e\right)Ti\omega _e\tau }`$ | $`J=S_1S_2e^{i\nu \tau }e^{i\left(\nu 2\omega _g\right)Ti\omega _g\tau }^{+i\omega t_1}`$ |
| --- | --- |
| $`F=S_2^2e^{i\nu \tau }e^{i\left(\nu +2\omega _e\right)Ti\omega _e\tau }`$ | $`K=S_1^{}S_2e^{i\left(\nu 2\omega _g\right)Ti\omega _g\tau }^{+i\omega t_1}`$ |
| $`G=S_1S_2e^{i\left(\nu +2\omega _e\right)Ti\omega _e\tau i\omega t_1}`$ | $`L=S_2^2e^{i\nu \tau }e^{i\left(\nu 2\omega _g\right)Ti\omega _g\tau }`$ |
| $`H=S_1^{}S_2e^{i\nu \tau }e^{i\left(\nu +2\omega _e\right)Ti\omega _e\tau i\omega t_1}`$ | $`M=S_1^2e^{i\left(\nu 2\omega _g\right)Ti\omega _g\tau }.`$ |
(12)
In the Ramsey experiment the standard observable $`\{|ee|,|gg|\}`$ is measured after the atom has passed cavity $`R_2`$. The corresponding probabilities $`p_e`$ and $`p_g`$ can be related to the initial state according to
$`\begin{array}{c}p_e=\psi _f|ee|\psi _f=\psi _{in}\left|\widehat{Q}_e\right|\psi _{in},\\ p_g=\psi _f|gg|\psi _f=\psi _{in}\left|\widehat{Q}_g\right|\psi _{in},\end{array}`$
yielding
$$\widehat{Q}_e=|q_eq_e|=\widehat{I}\widehat{Q}_g\text{ ; }|q_e=\left(\begin{array}{c}\left(E^{}+F^{}\right)\\ \left(G^{}+H^{}\right)\end{array}\right).$$
(13)
The observable $`\{\widehat{Q}_e,\widehat{Q}_g\}`$ can be interpreted as the quantum mechanical observable measured in the Ramsey experiment if the initial phase of the microwave field is well-defined. It is easily verified that $`q_e|q_e=1`$. So $`\widehat{Q}_e`$ and $`\widehat{Q}_g`$ are projections and $`\{\widehat{Q}_e,\widehat{Q}_g\}`$ is a PVM. This PVM will be referred to as the interference observable, because, provided the atom velocity is sufficiently well-defined, its expectation values in the state $`|\psi _{in}`$ exhibit interference fringes if $`\omega `$ is varied. It is easily verified that if the $`\frac{\pi }{2}`$ pulse condition $`\delta =0`$ is satisfied, and the detuning parameter $`\nu `$ is taken to be zero, then the interference observable reduces to PVM $`\{|ee|,|gg|\}.`$ This is in agreement with the fact that under these conditions the Ramsey setup just interchanges the roles of the $`e`$ and $`g`$ states. In general the standard observables $`\{\widehat{P}_+,\widehat{P}_{}\}`$ and $`\{\widehat{Q}_e,\widehat{Q}_g\}`$ are incompatible.
Also $`\{\widehat{Q}_e,\widehat{Q}_g\}`$ is dependent on the initial phase of the microwave field. Averaging over this phase yields the POVM $`\{\overline{\widehat{Q}_e},\overline{\widehat{Q}_g}\},`$
$`\overline{\widehat{Q}_e}`$ $`=`$ $`|S_1^2+S_2^2e^{i\nu \tau }|^2|ee|+|S_2|^2|S_1+S_1^{}e^{i\nu \tau }|^2|gg|,`$
$`\overline{\widehat{Q}_g}`$ $`=`$ $`\widehat{I}\overline{\widehat{Q}_e},`$
which, once again, is a non-ideal measurement of $`\{|ee|,|gg|\}`$ (for $`\nu =\delta =0`$ it even is an ideal one). It is important to note that, nevertheless, the expectation values of $`\{\overline{\widehat{Q}_e},\overline{\widehat{Q}_g}\}`$ exhibit interference fringes if $`\omega `$ is varied. When in (11) $`\beta =0`$ (as was satisfied in the experiments that have actually been carried out), then the expectation values of $`\widehat{Q}_e`$ and $`\overline{\widehat{Q}_e}`$ coincide. For this special case it is possible to analyze the Ramsey experiment in terms of the standard formalism, even if the actually performed experiment is a generalized one, described by POVM $`\{\overline{\widehat{Q}_e},\overline{\widehat{Q}_g}\}`$.
Contrary to the phase-averaged experiment, in case of a well-defined initial phase of the field the measurement performed after the atom left $`R_1`$ is incompatible with the one performed after $`R_2`$ if $`\delta \pm 1`$. The two measurement arrangements are complementary. The experiments are analogous to double-slit experiments in which it either is directly measured which slit a particle has passed through (‘which way’ or path measurement), or the interference pattern is measured after the two partial beams have been allowed to interfere (interference experiment). The quantity $`\nu \tau `$ is the relative phase shift of the partial beams.
In standard quantum mechanics complementarity is interpreted as mutual exclusiveness of information, caused by the impossibility of having both experimental arrangements simultaneously. Of the two incompatible PVMs $`\{\widehat{P}_+,\widehat{P}_{}\}`$ and $`\{\widehat{Q}_e,\widehat{Q}_g\}`$ either one or the other can be measured. In the following we shall discuss a measurement arrangement that is intermediate between the two arrangements considered above, viz. the experiment reported by Brune et al. , to be referred to as the Haroche-Ramsey experiment. As a result the experiment may yield information on both the path and the interference observable.
### B The Haroche-Ramsey experiment
In the Haroche-Ramsey experiment a third cavity $`C`$ is placed between cavities $`R_1`$ and $`R_2`$, storing a coherent field $`|\gamma `$. The transition frequency $`\omega _0`$ of the Rb atom and the frequency of the cavity $`C`$ field are chosen to be off-resonance, so there is no exchange of energy when the atom passes $`C`$. Contrary to the microwave fields in cavities $`R_1`$ and $`R_2`$ the cavity $`C`$ field is treated quantum mechanically. The field in cavity $`C`$ merely undergoes a phase shift $`\mathrm{\Phi }`$ (single atom index effect) which depends on the state of the Rb atom in the following way :
$$|e|\gamma \stackrel{C}{}|e|\gamma e^{i\mathrm{\Phi }}\text{ };\text{ }|g|\gamma \stackrel{C}{}|g|\gamma e^{i\mathrm{\Phi }}.$$
(14)
The states $`|\gamma e^{i\mathrm{\Phi }}`$ and $`|\gamma e^{i\mathrm{\Phi }}`$ are also coherent states. This yields the following unitary transformation $`\widehat{U}_C`$ describing the evolution of the atom-field system when the atom is going from $`R_1`$ to $`R_2`$ :
$$\widehat{U}_C=e^{i\omega _e\tau }|ee|e^{i\mathrm{\Phi }\widehat{a}^{}\widehat{a}}+e^{i\omega _g\tau }|gg|e^{i\mathrm{\Phi }\widehat{a}^{}\widehat{a}},$$
(15)
where $`\widehat{a}^{}`$ and $`\widehat{a}`$ are the photon creation and annihilation operators of the cavity $`C`$ field mode. For the initial state $`|\mathrm{\Psi }_{in}=\left[\alpha |e+\beta |g\right]|\gamma `$ of the combined atom-field system we get as final state:
$$|\mathrm{\Psi }_f=\{\begin{array}{cc}\text{ }|e\hfill & \left[\left(\alpha E+\beta G\right)|\gamma e^{i\mathrm{\Phi }}+\left(\alpha F+\beta H\right)|\gamma e^{i\mathrm{\Phi }}\right]\hfill \\ +|g\hfill & \left[\left(\alpha J+\beta L\right)|\gamma e^{i\mathrm{\Phi }}+\left(\alpha K+\beta M\right)|\gamma e^{i\mathrm{\Phi }}\right],\hfill \end{array}$$
(16)
constants $`E,G,etc.`$ being given by (12).
After the atom has passed cavity $`C`$the field is containing path information that can be retrieved by a measurement of a well-chosen observable of the field. In the standard formalism this information is usually analyzed in terms of the inner product $`\gamma e^{i\mathrm{\Phi }}|\gamma e^{i\mathrm{\Phi }}`$ of the field states $`|\gamma e^{i\mathrm{\Phi }}`$ and $`|\gamma e^{i\mathrm{\Phi }}`$, determining their distinguishability. The possibility of interference is seen as a consequence of the indistinguishability of the paths. How distinguishable the paths are, depends on the values of the parameters $`\gamma `$ and $`\mathrm{\Phi }`$. If the states are identical, then the paths are completely indistinguishable. Ignoring the possibility $`\mathrm{\Phi }=m\pi `$ $`(mZZ),`$ this obtains if $`\gamma =0.`$ In this case the experiment cannot yield any path information. Complete distinguishability, corresponding to maximal path information, obtains if the field states are orthogonal. This only obtains in the limit $`\gamma \mathrm{}.`$ In the next sections this analysis will be corroborated on the basis of the generalized formalism.
Whereas for the limiting values of $`\gamma ,`$ considered above, the standard formalism is sufficient, is the generalized formalism necessary for experiments corresponding to intermediate values $`0<\gamma <\mathrm{}`$. This already holds true if no measurement of the cavity $`C`$ field is carried out at all. Thus, putting $`\mathrm{\Psi }_f`$ $`|ee|\mathrm{\Psi }_f=\psi _{in}|\widehat{R}_e|\psi _{in}`$ (and analogously for $`g)`$ we find the POVM $`\{\widehat{R}_e,\widehat{R}_g\}`$ of the Haroche-Ramsey measurement from (16). Restricting to $`\nu =\delta =0`$ we get
$$\widehat{R}_e=\frac{1}{2}\left(\begin{array}{cc}1C_1& e^{i\omega t_1}C_2\\ e^{i\omega t_1}C_2& 1+C_1\end{array}\right),\text{ }\widehat{R}_g=\widehat{I}\widehat{R}_e,$$
(17)
in which
$$C_1+iC_2=\gamma e^{i\mathrm{\Phi }}|\gamma e^{i\mathrm{\Phi }}=e^{\gamma ^2\left(1e^{2i\mathrm{\Phi }}\right)}.$$
(18)
It is easily verified that, unless $`C_1=\pm 1,C_2=0,`$ $`\{\widehat{R}_e,\widehat{R}_g\}`$ is not a PVM. Even in the limit $`\gamma \mathrm{}`$ (corresponding to $`C_1=C_2=0)`$ (17) is a POVM, be it an uninformative one. This is consistent with complementarity in the sense that in this limit no information on the interference observable is obtained, path information being obtainable by a measurement of an observable of the cavity $`C`$ field in the final state of this field.
Since the operators of POVM $`\{\overline{\widehat{R}_e},\overline{\widehat{R}_g}\},`$ obtained from (17) by phase-averaging, are diagonal, the Haroche-Ramsey experiment is just a non-ideal version of the Ramsey experiment if the initial phase of the microwave field is random. The non-ideality measure (2) is then given as
$$J^{\overline{HR}}=\frac{(1C_1)}{2}ln\frac{(1C_1)}{2}\frac{(1+C_1)}{2}ln\frac{(1+C_1)}{2}.$$
(19)
This quantity is a measure of the inaccuracy introduced in the observation of observable $`\{|ee|,|gg|\}`$ by the insertion of cavity $`C`$.
From the point of view of complementarity the experimental setup of the Haroche-Ramsey experiment is particularly interesting when also the information is exploited that is stored in the cavity $`C`$ field, because this may add ‘which way’ information to the (non-ideal) interference information obtained from the measurement of the final state of the atom. This will be discussed in the next sections.
### C The Davidovich-Haroche experiment
By Davidovich et al. a variation of the Haroche-Ramsey experiment has been proposed in which a second atom traverses the system some time after the first one has passed. In the reason for sending this second atom is to probe a possible decoherence of the field in cavity $`C`$. We shall discuss this aspect of the experiment in section IV A. Here we are interested in the possibility to consider the second atom as yielding information on the cavity $`C`$ field that might be useful for determining the path of atom $`1.`$ We shall demonstrate that the joint measurement of standard observables $`\{|e_1e_1|,|g_1g_1|\}`$ and $`\{|e_2e_2|,|g_2g_2|\}`$ in the final state of the atoms can be interpreted as a measurement of a POVM on the incoming state of atom $`1`$. We shall neglect decoherence here by taking a negligible time interval between the atoms. We also restrict here to the case $`\nu =\delta =0`$, for which $`E+F=0`$, and, hence, (13) is yielding $`\widehat{Q}_e=|gg|,\widehat{Q}_g=|ee|`$. Then, starting with atom $`2`$ in state $`|e_2,`$ and using rules (14) for both atoms, we find for an arbitrary initial state $`|\psi _{in1}=\alpha |e_1+\beta |g_1`$ of atom $`1`$ the final state
$`|\mathrm{\Psi }_f={\displaystyle \frac{1}{4}}\left[\begin{array}{c}|e_1e_2e^{2i\omega T}\left\{\alpha \left(|v_g^{}2|\gamma \right)i\beta e^{i\omega t_1}|v_e^{}\right\}+\\ |e_1g_2e^{i\omega _0\tau }\left\{i\alpha e^{i\omega t_1}|v_e^{}\beta \left(|v_g^{}+2|\gamma \right)\right\}+\\ |g_1e_2e^{i\omega _0\tau }\left\{i\alpha e^{i\omega t_1}|v_e^{}\beta \left(|v_g^{}2|\gamma \right)\right\}+\\ |g_1g_2e^{2i\omega T+2i\omega _0\tau +i\omega t_1}\left\{\alpha e^{i\omega t_1}\left(|v_g^{}+2|\gamma \right)+i\beta |v_e^{}\right\}\end{array}\right],`$
with $`|v_e^{}=|\gamma e^{i2\mathrm{\Phi }}|\gamma e^{i2\mathrm{\Phi }}`$ , $`|v_g^{}=|\gamma e^{i2\mathrm{\Phi }}+|\gamma e^{i2\mathrm{\Phi }}.`$
By putting $`\mathrm{\Psi }_f||e_1e_1||e_2e_2||\mathrm{\Psi }_f=`$ $`\psi _{in1}|\widehat{M}_{e_1e_2}`$ $`|\psi _{in1},`$ etc., the POVM of the Davidovich-Haroche experiment, interpreted as a measurement on atom $`1,`$ is straightforwardly found according to
$$\begin{array}{c}\widehat{M}_{e_1e_2}=\frac{1}{16}\left(\begin{array}{cc}|v_g^{}2|\gamma ^2& ie^{i\omega t_1}(v_g^{}|v_e^{}2\gamma |v_e^{})\\ ie^{i\omega t_1}(v_e^{}|v_g^{}2v_e^{}|\gamma )& |v_e^{}^2\end{array}\right),\\ \widehat{M}_{e_1g_2}=\frac{1}{16}\left(\begin{array}{cc}|v_e^{}^2& ie^{i\omega t_1}(v_e^{}|v_g^{}+2v_e^{}|\gamma )\\ ie^{i\omega t_1}(v_g^{}|v_e^{}+2\gamma |v_e^{})& |v_g^{}+2|\gamma ^2\end{array}\right),\\ \widehat{M}_{g_1e_2}=\frac{1}{16}\left(\begin{array}{cc}|v_e^{}^2& ie^{i\omega t_1}(v_e^{}|v_g^{}2v_e^{}|\gamma )\\ ie^{i\omega t_1}(v_g^{}|v_e^{}2\gamma |v_e^{})& |v_g^{}2|\gamma ^2\end{array}\right),\\ \widehat{M}_{g_1g_2}=\frac{1}{16}\left(\begin{array}{cc}|v_g^{}+2|\gamma ^2& ie^{i\omega t_1}(v_g^{}|v_e^{}+2\gamma |v_e^{})\\ ie^{i\omega t_1}(v_e^{}|v_g^{}+2v_e^{}|\gamma )& |v_e^{}^2\end{array}\right).\end{array}$$
(20)
From these expressions it is immediately clear that averaging over the initial phase $`\omega t_1`$ makes also the Davidovich-Haroche experiment a non-ideal measurement of observable $`\{|ee|,|gg|\}`$, the non-ideality measure (2) being given by
$`J^{\overline{DH}}=\begin{array}{c}(\frac{1C_1}{2}\frac{1C_1^{}}{8})ln(1\frac{(1C_1^{})}{4(1C_1)})\frac{1C_1^{}}{8}ln\frac{1C_1^{}}{4(1C_1)}+\\ (\frac{1+C_1}{2}\frac{1C_1^{}}{8})ln(1\frac{(1C_1^{})}{4(1+C_1)})\frac{1C_1^{}}{8}ln\frac{1C_1^{}}{4(1+C_1)},\end{array}`$
with
$`C_1^{}+iC_2^{}=\gamma e^{i2\mathrm{\Phi }}|\gamma e^{i2\mathrm{\Phi }}.`$
Comparing $`J^{\overline{DH}}`$ with the corresponding non-ideality measure (19) of the Haroche-Ramsey experiment, we find (cf. figure 2) that $`J^{\overline{DH}}<J^{\overline{HR}}`$ for $`\gamma 0.`$ Hence, by taking into account the extra information from the measurement of the cavity $`C`$ field the quality of the non-ideal measurement of the interference observable $`\{\widehat{Q}_e,\widehat{Q}_g\}`$ has been increased.
In the phase-averaged case the subspace $`_{\{\widehat{M}_m\}}`$ of Hilbert-Schmidt space spanned by the operators of POVM (20) is a two-dimensional one. From an informational point of view the Davidovich-Haroche experiment will be more interesting if it is possible to avoid the necessity of phase averaging, because in that case $`_{\{\widehat{M}_m\}}`$ is three-dimensional. Although, due to the equality $`\widehat{M}_{e_1e_2}+\widehat{M}_{g_1e_2}=(|v_g^{}2|\gamma ^2+|v_e^{}^2)/16\widehat{I},`$ the operators of the POVM are linearly dependent, and, hence, the measurement is not a complete one, it nevertheless is a generalized measurement, being interpretable, in the sense defined in sect. II C, as a joint non-ideal measurement of two incompatible observables. In order to see this the operators must be ordered in a bivariate way. Due to the uninformativeness of the marginal $`\{\widehat{M}_{e_1e_2}+\widehat{M}_{g_1e_2},\widehat{M}_{e_1g_2}+\widehat{M}_{g_1g_2}\}`$ the only interesting way to do so is according to
$`\widehat{R}_{mn}=\left(\begin{array}{cc}\widehat{M}_{e_1e_2}& \widehat{M}_{e_1g_2}\\ \widehat{M}_{g_1g_2}& \widehat{M}_{g_1e_2}\end{array}\right),`$
yielding marginals $`\{\mathrm{\Sigma }_n\widehat{R}_{mn}\}`$ and $`\{\mathrm{\Sigma }_m\widehat{R}_{mn}\}`$ with
$`\widehat{M}_{e_1e_2}+\widehat{M}_{e_1g_2}=\frac{1}{2}\left(\begin{array}{cc}1C_1& C_2\\ C_2& 1+C_1\end{array}\right)`$ and $`\widehat{M}_{e_1e_2}+\widehat{M}_{g_1g_2}=\frac{1}{4}\left(\begin{array}{cc}3+C_1^{}& C_2^{}\\ C_2^{}& 1C_1^{}\end{array}\right).`$ Here $`t_{1\text{ }}`$ is taken to be zero. Evidently, both marginals are depending on the parameter $`\mathrm{\Phi }`$ governing the distinguishability of the field states. In agreement with (3), for $`\gamma 0`$ these marginals can be interpreted as describing non-ideal measurements of two incompatible PVMs of atom $`1`$, with non-ideality matrices given by
$`\left(\lambda \right)`$ $`=`$ $`\left(\begin{array}{cc}\lambda & 1\lambda \\ 1\lambda & \lambda \end{array}\right),\left(\mu \right)=\left(\begin{array}{cc}\mu & 1\mu \\ 1\mu & \mu \end{array}\right),`$
$`\lambda `$ $`=`$ $`{\displaystyle \frac{1}{2}}(1+\sqrt{C_1^2+C_2^2}),\mu ={\displaystyle \frac{1}{2}}(1+{\displaystyle \frac{1}{2}}\sqrt{(C_1^{}+1)^2+C_2^2}),`$
yielding non-ideality measures (2) as
$`J_{(\lambda )}`$ $`=`$ $`\{\lambda \mathrm{ln}(\lambda )+(1\lambda )\mathrm{ln}(1\lambda )\},`$
$`J_{(\mu )}`$ $`=`$ $`\{\mu \mathrm{ln}(\mu )+(1\mu )\mathrm{ln}(1\mu )\}.`$
For the parameters $`\lambda `$ and $`\mu `$ we find
$`\lambda `$ $`=`$ $`{\displaystyle \frac{1}{2}}(1+e^{2\gamma ^2\mathrm{sin}^2\mathrm{\Phi }}),`$
$`\mu `$ $`=`$ $`{\displaystyle \frac{1}{2}}+{\displaystyle \frac{1}{4}}[\{1+e^{2\gamma ^2\mathrm{sin}^22\mathrm{\Phi }}\mathrm{cos}(\gamma ^2\mathrm{sin}4\mathrm{\Phi })\}^2+e^{4\gamma ^2\mathrm{sin}^22\mathrm{\Phi }}\mathrm{sin}^2(\gamma ^2\mathrm{sin}4\mathrm{\Phi })]^{1/2}.`$
We shall not bother to calculate the corresponding PVMs, because these do not admit a straightforward physical interpretation in terms of the interference and path observables defined above. From the non-ideality measures $`J_{(\lambda )}`$ and $`J_{(\mu )}`$ it can already be seen that the two PVMs measured jointly in the Davidovich-Haroche experiment do not constitute a canonically conjugate pair in the sense that, if the parameter $`\gamma `$ is varied, one measurement gets more accurate if the other one becomes more non-ideal. Thus, in both of the limits $`\gamma =0`$ and $`\gamma \mathrm{}`$ POVM (20) is representing a (non-)ideal measurement of the same PVM $`\{|ee|,|gg|\}.`$ From the plots of $`J_{(\lambda )}`$ and $`J_{(\mu )}`$ as functions of $`\gamma `$ and $`\mathrm{\Phi }`$ in figure 3 it is also seen that both non-ideality measures vanish in the limit $`\gamma 0`$. Hence, although the two PVMs measured jointly in this experiment do satisfy inequality (4) for all values of $`\gamma `$, this does not imply any complementarity for $`\gamma 0`$ because the right-hand side of the inequality is vanishing in this limit due to the fact that the two PVMs coincide. As will be demonstrated in section IV, the Davidovich-Haroche experiment, for general values of the parameters $`\nu `$ and $`\delta `$, is informationally equivalent to a measurement in which the second atom is replaced by a measurement of photon number in the final state of cavity $`C`$. The absence of information on the phase of the cavity $`C`$ field explains the somewhat non-complementary behavior observed here. In section V an alternative measurement procedure will be discussed, better satisfying the canonical notion of complementarity, in which it is proposed to perform a measurement of the cavity $`C`$ field also yielding phase information.
## IV Informational aspects of the Davidovich-Haroche experiment
### A Decoherence
The Haroche-Ramsey experiment was devised in the first place to probe decoherence in cavity $`C`$ following the passage of a Rb atom, entering in state $`|e`$. Hence $`\beta =0`$ in the final state (16). Restricting to $`\nu =\delta =0`$ it follows from (16) that, conditional on measurement result $`e`$ or $`g,`$ the cavity $`C`$ field is described by a superposition of coherent states $`(e|v_e=|\gamma e^{i\mathrm{\Phi }}|\gamma e^{i\mathrm{\Phi }},g|v_g=|\gamma e^{i\mathrm{\Phi }}+|\gamma e^{i\mathrm{\Phi }})`$. These states can be considered as Schrödinger cat states if $`\gamma `$ is sufficiently large. In it was proposed to probe, by sending after a time $`T`$ a second Rb atom through the system (like the first atom starting in state $`|e`$), whether a process of decoherence is active by which these superpositions could decay to a mixture of the coherent states $`|\gamma e^{i\mathrm{\Phi }}`$ and $`|\gamma e^{i\mathrm{\Phi }}`$. In it was concluded from the observed two-atom correlations that a decoherence effect obtains. In particular it was inferred from the experimental data that the decoherence time is much shorter than the decay time that can be attributed to loss of photons from cavity $`C`$. In the present section this latter conclusion is challenged. It is demonstrated that, in agreement with a result obtained by Vitali et al. , the second Rb atom can only yield information on cavity $`C`$’s photon number. Hence, any change of the measurement results obtained for this atom should be attributed to a change of photon number. Of course, this does not imply the absence of decoherence due to decay of pure phase correlations. However, this measurement is not sensitive to it.
In order to demonstrate this we have to determine which information is obtained on the cavity $`C`$ field by the measurement of the second atom. The corresponding POVM can be found by equating, for an arbitrary initial coherent state $`|\alpha `$ of the cavity field, the final state probabilities $`p_e`$ and $`p_g`$ to expectation values $`\alpha |\widehat{M}_e|\alpha `$ and $`\alpha |\widehat{M}_g`$ $`|\alpha `$, respectively. With $`|\mathrm{\Psi }_f=\frac{1}{2}|e_2(E|\alpha e^{i\mathrm{\Phi }}+F|\alpha e^{i\mathrm{\Phi }})+\frac{1}{2}|g_2(J|\alpha e^{i\mathrm{\Phi }}+K|\alpha e^{i\mathrm{\Phi }})`$ we find, once again restricting to $`\nu =\delta =0`$,
$`p_e`$ $`=`$ $`\mathrm{\Psi }_f|e_2e_2|\mathrm{\Psi }_f={\displaystyle \frac{1}{4}}|\alpha e^{i\mathrm{\Phi }}|\alpha e^{i\mathrm{\Phi }}^2={\displaystyle \frac{1}{4}}\alpha |2e^{2i\mathrm{\Phi }\widehat{a}^{}\widehat{a}}e^{2i\mathrm{\Phi }\widehat{a}^{}\widehat{a}}|\alpha ,`$
$`p_g`$ $`=`$ $`\mathrm{\Psi }_f|g_2g_2|\mathrm{\Psi }_f={\displaystyle \frac{1}{4}}|\alpha e^{i\mathrm{\Phi }}+|\alpha e^{i\mathrm{\Phi }}^2={\displaystyle \frac{1}{4}}\alpha |2+e^{2i\mathrm{\Phi }\widehat{a}^{}\widehat{a}}+e^{2i\mathrm{\Phi }\widehat{a}^{}\widehat{a}}|\alpha ,`$
from which we obtain
$`\widehat{M}_e=\mathrm{sin}^2\mathrm{\Phi }\widehat{a}^{}\widehat{a},\widehat{M}_g=\mathrm{cos}^2\mathrm{\Phi }\widehat{a}^{}\widehat{a}.`$
Note that this result holds independent of phase averaging. It is easily seen that POVM $`\{\widehat{M}_e,\widehat{M}_g\}`$ represents a non-ideal measurement of the number observable $`\widehat{N}=\widehat{a}^{}\widehat{a}=_{n=0}^{\mathrm{}}n|nn|`$ in the sense of definition (1):
$`\widehat{M}_e`$ $`=`$ $`{\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}\lambda _{en}|nn|,\widehat{M}_g={\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}\lambda _{gn}|nn|,`$
$`\lambda _{en}`$ $`=`$ $`\mathrm{sin}^2\mathrm{\Phi }n,\lambda _{gn}=\mathrm{cos}^2\mathrm{\Phi }n.`$
Using (6) it is possible to calculate the projection $`\widehat{\rho }_{\{\widehat{M}_m\}}`$ of the density operator $`\widehat{\rho }`$ on the subspace spanned by $`\widehat{M}_e`$ and $`\widehat{M}_g,`$ representing the information that is obtained by a measurement of POVM $`\{\widehat{M}_e,\widehat{M}_g\}.`$ Due to the infinite-dimensionality of the Hilbert space of the field this must be done with some care because the operators are not Hilbert-Schmidt operators then. For this reason the dimension must be truncated. Restricting to an arbitrary large but finite value $`D,`$ we get:
$`\widehat{M}_m^{}`$ $`=`$ $`{\displaystyle \underset{n=0}{\overset{D}{}}}[\beta _{me}\mathrm{sin}^2\mathrm{\Phi }n+\beta _{mg}\mathrm{cos}^2\mathrm{\Phi }n]|nn|,m=e,g,`$
$`\beta _{ee}`$ $`=`$ $`{\displaystyle \frac{rs}{(1+rs)(Drs)s^2}},\beta _{ge}=\beta _{eg}={\displaystyle \frac{s}{(1+rs)(Drs)s^2}},`$
$`\beta _{gg}`$ $`=`$ $`{\displaystyle \frac{Drs}{(1+rs)(Drs)s^2}},`$
with $`r=_{n=1}^D\mathrm{cos}^2\mathrm{\Phi }n,s=_{n=1}^D\mathrm{sin}^2\mathrm{\Phi }n\mathrm{cos}^2\mathrm{\Phi }n.`$ Then
$`\widehat{\rho }_{\{\widehat{M}_m\}}`$ $`=`$ $`\underset{D\mathrm{}}{lim}{\displaystyle \underset{n,n^{}=0}{\overset{D}{}}}[(\beta _{ee}\mathrm{sin}^2\mathrm{\Phi }n^{}+\beta _{ge}\mathrm{cos}^2\mathrm{\Phi }n^{})\mathrm{sin}^2\mathrm{\Phi }n+`$ (23)
$`+(\beta _{eg}\mathrm{sin}^2\mathrm{\Phi }n^{}+\beta _{gg}\mathrm{cos}^2\mathrm{\Phi }n^{})\mathrm{cos}^2\mathrm{\Phi }n]n^{}\left|\widehat{\rho }\right|n^{}|nn|.`$
Note that, although $`\beta _{mm^{}}0`$ if $`D\mathrm{},`$ yet $`\widehat{\rho }_{\{\widehat{M}_m\}}\widehat{O}.`$ Thus, it is easily verified that also in the limit $`D\mathrm{}`$ $`Tr\widehat{\rho }_{\{\widehat{M}_m\}}=1,`$ and $`Tr\widehat{\rho }_{\{\widehat{M}_m\}}\widehat{M}_m=Tr\widehat{\rho }\widehat{M}_m,m=e,g,`$ the latter equality explicitly demonstrating that $`\widehat{\rho }_{\{\widehat{M}_m\}}`$ contains the same information on the measurement results of POVM $`\{\widehat{M}_e,\widehat{M}_g\}`$ as does $`\widehat{\rho }.`$
Although a measurement of this POVM can distinguish between a mixture of the states $`|v_e`$ and $`|v_g`$ and a mixture of the states $`|\gamma e^{i\mathrm{\Phi }}`$ and $`|\gamma e^{i\mathrm{\Phi }},`$ this is only so because the probability distributions of the photon number observable are different in the two mixtures. Decoherence, not accompanied by a change of photon number, cannot be observed using the Davidovich-Haroche experiment.
### B Informational equivalence of second atom and number measurement
In this section it will be demonstrated that the Davidovich-Haroche experiment, in which a second atom is used as a probe of the cavity $`C`$ field, is informationally equivalent to a Haroche-Ramsey experiment in which the ‘which-way’ information is obtained by measuring photon number. This will be done by considering the informational aspects of the measurement, introduced in section II D. From the informational point of view the important feature is the structure of the subspaces $`_{\{\widehat{M}_m\}}`$ of Hilbert-Schmidt space, spanned by the operators $`\widehat{M}_m`$ generating the POVM, as a function of the experimental parameters. In order to be completely general, in this section we allow the different parameters to take arbitrary values. We first determine the POVM of the Haroche-Ramsey experiment in which cavity $`C`$ photon number is measured in coincidence with a determination of the final state of the atom. This POVM is found from (16) by the equalities
$`\begin{array}{c}p_{en}=\mathrm{\Psi }_f|ee||nn|\mathrm{\Psi }_f=\psi _{in}|\widehat{M}_{en}|\psi _{in},\\ p_{gn}=\mathrm{\Psi }_f|gg||nn|\mathrm{\Psi }_f=\psi _{in}|\widehat{M}_{gn}|\psi _{in},\end{array}`$
in which $`p_{en}`$ and $`p_{gn}`$ are the measured joint probabilities, and $`\{|nn|\}`$ is the PVM corresponding to the spectral measure of the photon number observable. With $`t_1=0`$ the operators $`\widehat{M}_{en}`$ and $`\widehat{M}_{gn}`$ are found as
$`\widehat{M}_{en}`$ $`=`$ $`{\displaystyle \frac{e^{\gamma ^2}\gamma ^{2n}}{n!}}\left(\begin{array}{cc}\mathrm{sin}^2\frac{\phi _n}{2}+\delta ^2\mathrm{cos}^2\frac{\phi _n}{2}& e^{i\psi }\frac{\sqrt{1\delta ^2}}{2}[\mathrm{sin}\phi _n+2i\delta \mathrm{cos}^2\frac{\phi _n}{2}]\\ e^{i\psi }\frac{\sqrt{1\delta ^2}}{2}[\mathrm{sin}\phi _n2i\delta \mathrm{cos}^2\frac{\phi _n}{2}]& (1\delta ^2)\mathrm{cos}^2\frac{\phi _n}{2}\end{array}\right),`$
$`\widehat{M}_{gn}`$ $`=`$ $`{\displaystyle \frac{e^{\gamma ^2}\gamma ^{2n}}{n!}}\left(\begin{array}{cc}(1\delta ^2)\mathrm{cos}^2\frac{\phi _n}{2}& e^{i\psi }\frac{\sqrt{1\delta ^2}}{2}[\mathrm{sin}\phi _n+2i\delta \mathrm{cos}^2\frac{\phi _n}{2}]\\ e^{i\psi }\frac{\sqrt{1\delta ^2}}{2}[\mathrm{sin}\phi _n2i\delta \mathrm{cos}^2\frac{\phi _n}{2}]& \mathrm{sin}^2\frac{\phi _n}{2}+\delta ^2\mathrm{cos}^2\frac{\phi _n}{2}\end{array}\right),`$
in which $`\phi _n=2n\mathrm{\Phi }+\nu \tau +2\psi ,\psi =\mathrm{arg}(S_1).`$
Excluding $`\delta =\pm 1`$, for which POVM $`\{\widehat{M}_{en},\widehat{M}_{gn}\}`$ reduces to a trivial refinement of PVM $`\{|ee|,|gg|\}`$, for most values of the parameters the operators $`\widehat{M}_{en}`$ and $`\widehat{M}_{gn},n=0,1..`$ span the whole Hilbert-Schmidt space of operators on a 2-dimensional Hilbert space. Hence, in general the measurement is a complete measurement. The parameter values for which the measurement is incomplete can be found by looking for Hermitian operators $`\widehat{T}`$ that are orthogonal to all $`\widehat{M}_{en}`$ and $`\widehat{M}_{gn}.`$ Thus, $`Tr\widehat{M}_{en}\widehat{T}=Tr\widehat{M}_{gn}\widehat{T}=0,n=0,..`$ We find
$`\mathrm{\Phi }=\frac{\pi }{2},\delta 0:`$
$`\widehat{T}=\left(\begin{array}{cc}\sqrt{1\delta ^2}\mathrm{tan}(\nu \tau +2\psi )& e^{i\psi }[1i\delta \mathrm{tan}(\nu \tau +2\psi )]\\ e^{i\psi }[1+i\delta \mathrm{tan}(\nu \tau +2\psi )]& \sqrt{1\delta ^2}\mathrm{tan}(\nu \tau +2\psi )\end{array}\right);`$
$`\mathrm{\Phi }\frac{\pi }{2},\delta =0:`$
$`\widehat{T}=\left(\begin{array}{cc}0& ie^{i\psi }\\ ie^{i\psi }& 0\end{array}\right);`$
$`\mathrm{\Phi }=\frac{\pi }{2},\delta =0:`$
$`\widehat{T}=\left(\begin{array}{cc}\mathrm{tan}(\nu \tau +2\psi )& e^{i\psi }\\ e^{i\psi }& \mathrm{tan}(\nu \tau +2\psi )\end{array}\right),\left(\begin{array}{cc}0& ie^{i\psi }\\ ie^{i\psi }& 0\end{array}\right).`$ Barring $`\delta =\pm 1`$, for all other values of the parameters no solution for $`\widehat{T}`$ can be found. Hence, specializing the parameters of the experiment to either $`\mathrm{\Phi }=\frac{\pi }{2}`$, or to the $`\frac{\pi }{2}`$ pulse condition $`\delta =0`$, reduces the dimension of the subspace spanned by the operators of the POVM to 3, the dimensionality being further reduced if both conditions are simultaneously satisfied. By determining in the same way the Hilbert-Schmidt operators that are orthogonal to the operators of POVM (20) it is straightforward to prove that the Davidovich-Haroche experiment, discussed in section III C, has exactly the same structure of subspaces, thus demonstrating the informational equivalence of these experiments for all values of the parameters.
The subspace structure is not essentially changed by taking the detuning parameter $`\nu =0.`$ Since then also $`\psi =0`$ the operators $`\widehat{T}`$, found above, are particularly simple, viz. $`\left(\begin{array}{cc}0& i\\ i& 0\end{array}\right)`$ and $`\left(\begin{array}{cc}0& 1\\ 1& 0\end{array}\right).`$ These are two orthogonal vectors, constituting together with the operators $`\left(\begin{array}{cc}1& 0\\ 0& 0\end{array}\right)`$ and $`\left(\begin{array}{cc}0& 0\\ 0& 1\end{array}\right)`$ an orthogonal basis of Hilbert-Schmidt space. Due to the uniqueness of the Hermitian projection operator $`𝒫_{\{\widehat{M}_m\}}`$ this makes it particularly easy to calculate the projected density operator $`𝒫_{\{\widehat{M}_m\}}\widehat{\rho }`$ representing the information about the density operator $`\widehat{\rho }`$ provided by the measurement. We find (with $`\nu =0):`$
$`\mathrm{\Phi }=\frac{\pi }{2},\delta 0:𝒫_{\{\widehat{M}_m\}}\widehat{\rho }=\left(\begin{array}{cc}\rho _{11}& \frac{\rho _{12}\rho _{21}}{2}\\ \frac{\rho _{12}\rho _{21}}{2}& \rho _{22}\end{array}\right);`$
$`\mathrm{\Phi }\frac{\pi }{2},\delta =0:𝒫_{\{\widehat{M}_m\}}\widehat{\rho }=\left(\begin{array}{cc}\rho _{11}& \frac{\rho _{12}+\rho _{21}}{2}\\ \frac{\rho _{12}+\rho _{21}}{2}& \rho _{22}\end{array}\right);`$
$`\mathrm{\Phi }=\frac{\pi }{2},\delta =0:𝒫_{\{\widehat{M}_m\}}\widehat{\rho }=\left(\begin{array}{cc}\rho _{11}& 0\\ 0& \rho _{22}\end{array}\right).`$
Note that all $`𝒫_{\{\widehat{M}_m\}}\widehat{\rho }`$ are non-negative (cf. Appendix). In the $`\mathrm{\Phi }=\frac{\pi }{2},\delta =0`$ case only information is obtained on the diagonal elements of $`\widehat{\rho }`$ because for these parameter values the measurement is a non-ideal measurement of PVM $`\{|ee|,|gg|\}.`$ It is clear from this that from the informational point of view our parameter choice in section III C was not completely appropriate for the purpose of reconstructing the initial density operator. By restricting to $`\frac{\pi }{2}`$ pulses the measurement cannot retrieve $`Im\rho _{12}`$. For a complete determination of $`\widehat{\rho }`$ it is necessary that $`\mathrm{\Phi }\frac{\pi }{2}`$ and $`\delta 0`$ (keeping $`\delta \pm 1`$).
One remark is in order here. Since the special parameter values $`\mathrm{\Phi }=\frac{\pi }{2}`$ and $`\delta =0`$ constitute sets of measure zero within the set of all possible values of the parameters, it might be thought that these special values are physically irrelevant because they cannot be attained in practice. In a strict sense this is correct. However, even though in practice the POVMs are complete, this does not mean that the subspace structure is unimportant. As a matter of fact, if the parameter values are near the special values given above, then the quantities $`Tr(\widehat{\rho }\widehat{\rho }_{\{\widehat{M}_m(\mathrm{\Phi }=\frac{\pi }{2},\delta =0)\}})\widehat{M}_m`$ will be very small. This means that the experimental error in the determination of these quantities is relatively large. Hence, the experimental probabilities will yield relatively poor information about the components of the Hilbert-Schmidt vector $`\widehat{\rho }`$ orthogonal to $`_{\{\widehat{M}_m(\mathrm{\Phi }=\frac{\pi }{2},\delta =0)\}}`$. Stated differently, sets $`\{\widehat{M}_m\}`$ constituting bases of Hilbert-Schmidt space need not be equivalent from an informational point of view, the quality of the information being largest for an orthogonal basis. A measure of this quality will be discussed elsewhere.
## V The Haroche-Ramsey experiment as a joint non-ideal measurement of interference and path observables
Rather than exploiting a second atom, or, equivalently, measuring observable $`\{\mathrm{sin}^2\mathrm{\Phi }\widehat{a}^{}\widehat{a},\mathrm{cos}^2\mathrm{\Phi }\widehat{a}^{}\widehat{a}\}`$ of the cavity $`C`$ field jointly with observable $`\{|ee|,|gg|\},`$ we consider here the field observable $`\{\widehat{M}_+,\widehat{M}_{}\}`$ defined by
$`\widehat{M}_+={\displaystyle \frac{1}{\pi }}{\displaystyle \underset{C^+}{}}d^2\alpha |\alpha \alpha \left|\text{ ; }\widehat{M}_{}={\displaystyle \frac{1}{\pi }}{\displaystyle \underset{C^{}}{}}d^2\alpha \right|\alpha \alpha |,`$
$`|\alpha `$ a coherent state, and the integrations being over the upper $`(C^+)`$ and lower $`(C^{})`$ complex half-planes, respectively. This observable is a coarsening of the observable $`\{\frac{1}{\pi }|\alpha \alpha |\}`$ measured in the eight-port homodyning detection method . For $`\gamma \mathrm{}`$ we have $`\gamma e^{i\mathrm{\Phi }}\left|\widehat{M}_m\right|\gamma e^{i\mathrm{\Phi }}\delta _{m+},\gamma e^{i\mathrm{\Phi }}\left|\widehat{M}_m\right|\gamma e^{i\mathrm{\Phi }}\delta _m`$. In this limit POVM $`\{\widehat{M}_+,\widehat{M}_{}\}`$ is evidently yielding information on the phase shift $`\mathrm{\Phi }`$ caused by the Rb atom, and, hence, is providing path information. In the Haroche-Ramsey experiment the value of $`\gamma `$ was finite ($`\gamma 3),`$ causing the distinguishability of the states to be only partial. As will be seen in the following, this loss of path information is compensated by the interference information obtained by measuring a quantity of the $`C`$ field yielding information on both number and phase.
In order to interpret the experiment as a measurement in the initial state $`|\psi _{in}`$ we put
$`p_{m\pm }=\mathrm{\Psi }_f|\left(|mm|\widehat{M}_\pm \right)|\mathrm{\Psi }_f=\psi _{in}\left|\widehat{M}_{m\pm }\right|\psi _{in},m=e,g,`$
yielding a POVM $`\{\widehat{M}_{e+},\widehat{M}_e,\widehat{M}_{g+},\widehat{M}_g\}`$ with elements given by
$$\begin{array}{c}\widehat{M}_{e\pm }=\frac{1}{2}[(1\pm AC_1)\left|S_1\right|^2\widehat{P}_++(1AC_1)\left|S_2\right|^2\widehat{P}_{}+C_1\widehat{Q}_e+C_2\widehat{S}],\\ \widehat{M}_{g\pm }=\frac{1}{2}[1A\delta ]\widehat{I}\widehat{M}_e,\delta =|S_1|^2|S_2|^2.\end{array}$$
(26)
Here $`\{\widehat{P}_+,\widehat{P}_{}\}`$ and $`\{\widehat{Q}_e,\widehat{Q}_g\}`$ are the path and interference observables (10) and (13) defined above. The constant $`A`$ is given by
$`A=erf(\gamma \mathrm{sin}\mathrm{\Phi }),`$
and $`C_1`$ and $`C_2`$ are given by (18). The operator $`\widehat{S}`$ is defined according to
$`\widehat{S}=ie^{i\nu \tau }S_1^{}S_2|p_+p_{}|\text{-}ie^{i\nu \tau }S_1^{}S_2|p_{}p_+|,`$
$`|p_\pm `$ being given by (10).
In the phase averaged case the experiment described by this POVM once again is a non-ideal measurement of PVM $`\{|ee|,|gg|\}.`$ Restricting to $`\nu =\delta =0`$ we find
$`\overline{\widehat{M}_{e\pm }}={\displaystyle \frac{1}{4}}(1C_1)|ee\left|+{\displaystyle \frac{1}{4}}(1+C_1)\right|gg|,\overline{\widehat{M}_{g\pm }}={\displaystyle \frac{1}{4}}(1+C_1)|ee\left|+{\displaystyle \frac{1}{4}}(1C_1)\right|gg|,`$
yielding for the non-ideality measure the same outcome (19) as obtained in the experiment in which no measurement is performed on the cavity $`C`$ field. Evidently, in the phase averaged case such a measurement does not improve the information. Indeed, the two measurements are equivalent in the sense defined in .
We shall now consider POVM (26) when no phase averaging is performed. We should exclude $`\delta =\pm 1`$ also here because then also POVM $`\{\widehat{M}_{e\pm },\widehat{M}_{g\pm }\}`$ reduces to a trivial refinement of PVM $`\{|ee|,|gg|\}`$ (this actually holds true for any choice of the observable of the cavity $`C`$ field). In the limit $`\gamma =0`$ the POVM reduces to $`\{\frac{1}{2}\widehat{Q}_e,\frac{1}{2}\widehat{Q}_e,\frac{1}{2}\widehat{Q}_g,\frac{1}{2}\widehat{Q}_g\},`$ representing a trivial refinement of the interference observable $`\{\widehat{Q}_e,\widehat{Q}_g\}`$. On the other hand, for $`\gamma \mathrm{}`$ the POVM reduces to the trivial refinement $`\{\left|S_1\right|^2\widehat{P}_+,\left|S_2\right|^2\widehat{P}_{},\left|S_2\right|^2\widehat{P}_+,\left|S_1\right|^2\widehat{P}_{}\}`$ of the path observable $`\{\widehat{P}_+,\widehat{P}_{}\}`$. We shall now demonstrate that if the $`\frac{\pi }{2}`$ pulse condition $`\delta =0`$ is satisfied ($`\nu `$ arbitrary) and $`\mathrm{\Phi }=\frac{\pi }{2},`$ the Haroche-Ramsey experiment can be interpreted, in the sense of section II C, as a joint non-ideal measurement of the incompatible observables $`\{\widehat{Q}_e,\widehat{Q}_g\}`$ and $`\{\widehat{P}_+,\widehat{P}_{}\}`$. To see this we define the bivariate POVM $`\{\widehat{M}_{mn}\}`$:
$$\{\widehat{M}_{mn}\}=\left(\begin{array}{cc}\widehat{M}_{e+}& \widehat{M}_{g+}\\ \widehat{M}_e& \widehat{M}_g\end{array}\right).$$
(27)
For the two marginals we find, respectively,
$$\left(\begin{array}{c}\widehat{M}_{e+}+\widehat{M}_{g+}\hfill \\ \widehat{M}_e+\widehat{M}_g\hfill \end{array}\right)=\stackrel{\left(\lambda _{mn}\right)}{\stackrel{}{\frac{1}{2}\left(\begin{array}{cc}1+erf(\gamma )\hfill & 1erf(\gamma )\hfill \\ 1erf(\gamma )\hfill & 1+erf(\gamma )\hfill \end{array}\right)}}\left(\begin{array}{c}\widehat{P}_+\hfill \\ \widehat{P}_{}\hfill \end{array}\right)$$
(28)
and
$$\left(\begin{array}{c}\widehat{M}_{e+}+\widehat{M}_e\hfill \\ \widehat{M}_{g+}+\widehat{M}_g\hfill \end{array}\right)=\stackrel{\left(\mu _{mn}\right)}{\stackrel{}{\frac{1}{2}\left(\begin{array}{cc}1+e^{2\gamma ^2}\hfill & 1e^{2\gamma ^2}\hfill \\ 1e^{2\gamma ^2}\hfill & 1+e^{2\gamma ^2}\hfill \end{array}\right)}}\left(\begin{array}{c}\widehat{Q}_e\hfill \\ \widehat{Q}_g\hfill \end{array}\right)\text{.}$$
(29)
In the limits $`\gamma \mathrm{}`$ and $`\gamma =0`$ the marginals describe ideal measurements of path and interference, respectively. The non-ideality measures $`J_{(\lambda )}`$ and $`J_{(\mu )}`$ corresponding to the non-ideality matrices $`\left(\lambda _{mn}\right)`$ and $`\left(\mu _{mn}\right)`$ are found according to
$`\begin{array}{c}J_{(\lambda )}=\frac{1}{2}\left[\left(1+erf(\gamma )\right)\mathrm{ln}\left(\frac{1+erf(\gamma )}{2}\right)+\left(1erf(\gamma )\right)\mathrm{ln}\left(\frac{1erf(\gamma )}{2}\right)\right]\text{,}\\ \text{ }\\ J_{(\mu )}=\frac{1}{2}\left[\left(1+e^{2\gamma ^2}\right)\mathrm{ln}\left(\frac{1+e^{2\gamma ^2}}{2}\right)+\left(1e^{2\gamma ^2}\right)\mathrm{ln}\left(\frac{1e^{2\gamma ^2}}{2}\right)\right]\text{.}\end{array}`$
For the special values of the parameters considered here we have $`|p_i|q_j|=1/\sqrt{2},i=\pm ,j=e,g.`$ Hence, for the observables $`\{\widehat{Q}_e,\widehat{Q}_g\}`$ and $`\{\widehat{P}_+,\widehat{P}_{}\}`$ the right-hand side of inequality (4) is non-vanishing, and it is impossible that both $`J_{(\lambda )}`$ and $`J_{(\mu )}`$ are equal to zero. In figure 4 $`J_{(\lambda )}`$ is plotted versus $`J_{(\mu )}`$ as a function of the parameter $`\gamma .`$ The resulting curve clearly exhibits the idea of complementarity expressed by inequality (4): the experiment constitutes a less accurate measurement of the interference observable as the path observable is determined more accurately by increasing $`\gamma `$ (and vice versa). It is impossible that both $`J_{(\lambda )}`$ and $`J_{(\mu )}`$ simultaneously have small values.
The nice feature of the $`\mathrm{\Phi }=\frac{\pi }{2}`$ condition is that the $`\gamma `$ dependence of the marginals is completely taken into account by the non-ideality matrices $`\left(\lambda _{mn}\right)`$ and $`\left(\mu _{mn}\right),`$ PVMs $`\{\widehat{Q}_e,\widehat{Q}_g\}`$ and $`\{\widehat{P}_+,\widehat{P}_{}\}`$ being independent of $`\gamma .`$ This feature is partly lost if we allow values $`\mathrm{\Phi }\frac{\pi }{2}`$. For general values of the parameters we can represent POVM (26) in the following way:
$`\widehat{M}_{e+}`$ $`=`$ $`{\displaystyle \frac{1}{2}}[(1+A)|S_1|^2|p_+p_+|+(1A)|S_2|^2|p_{}p_{}|+\{Ce^{i\nu \tau }S_1^{}S_2\}|p_+p_{}|+h.c.\}]`$
$`\widehat{M}_e`$ $`=`$ $`{\displaystyle \frac{1}{2}}[(1A)|S_1|^2|p_+p_+|+(1+A)|S_2|^2|p_{}p_{}|+\{Ce^{i\nu \tau }S_1^{}S_2\}|p_+p_{}|+h.c.\}]`$
$`\widehat{M}_{g+}`$ $`=`$ $`{\displaystyle \frac{1}{2}}[(1+A)|S_2|^2|p_+p_+|+(1A)|S_1|^2|p_{}p_{}|\{Ce^{i\nu \tau }S_1^{}S_2\}|p_+p_{}|+h.c.\}]`$
$`\widehat{M}_g`$ $`=`$ $`{\displaystyle \frac{1}{2}}[(1A)|S_2|^2|p_+p_+|+(1+A)|S_1|^2|p_{}p_{}|\{Ce^{i\nu \tau }S_1^{}S_2\}|p_+p_{}|+h.c.\}].`$
From this we find as one marginal
$`\left(\begin{array}{c}\widehat{M}_{e+}+\widehat{M}_{g+}\hfill \\ \widehat{M}_e+\widehat{M}_g\hfill \end{array}\right)=\stackrel{\left(\lambda _{mn}^{}\right)}{\stackrel{}{{\displaystyle \frac{1}{2}}\left(\begin{array}{cc}1+A\hfill & 1A\hfill \\ 1A\hfill & 1+A\hfill \end{array}\right)}}\left(\begin{array}{c}\widehat{P}_+\hfill \\ \widehat{P}_{}\hfill \end{array}\right),`$
$`A=erf(\gamma \mathrm{sin}\mathrm{\Phi }),`$ which still is a non-ideal measurement of the the path observable $`\{\widehat{P}_+,\widehat{P}_{}\}`$. However for the other marginal we get
$`\left(\begin{array}{c}\widehat{M}_{e+}+\widehat{M}_e\hfill \\ \widehat{M}_{g+}+\widehat{M}_g\hfill \end{array}\right)=\stackrel{\left(\mu _{mn}^{}\right)}{\stackrel{}{\left(\begin{array}{cc}\mu & 1\mu \\ 1\mu & \mu \end{array}\right)}}\left(\begin{array}{c}\widehat{Q}_e^{}\\ \widehat{Q}_g^{}\end{array}\right),`$
with $`\mu =(1\sqrt{1(1\delta ^2)(1|C|^2)})/2`$ and $`\{\widehat{Q}_e^{}`$,$`\widehat{Q}_g^{}\}`$ different from $`\{\widehat{Q}_e`$,$`\widehat{Q}_g\}.`$ Since PVM $`\{\widehat{Q}_e^{}`$,$`\widehat{Q}_g^{}\}`$ turns out to be dependent on $`\gamma `$ the experiment no longer is a joint measurement of one stable PVM pair when varying $`\gamma .`$ Nevertheless for each set of parameters PVM $`\{\widehat{Q}_e^{}`$,$`\widehat{Q}_g^{}\}`$ is incompatible with the path observable, and inequality (4) is satisfied by the non-ideality measures $`J_{(\lambda ^{})}`$ and $`J_{(\mu ^{})}`$. Since the parameters $`A`$ and $`|C|`$ depend on $`\gamma `$ and $`\mathrm{\Phi }`$ only as $`\gamma \mathrm{sin}\mathrm{\Phi },`$ this latter quantity, together with $`\delta `$, is determining the measure of complementarity of observables $`\{\widehat{Q}_e^{}`$,$`\widehat{Q}_g^{}\}`$ and $`\{\widehat{P}_+,\widehat{P}_{}\}`$. By comparing figures 5 and 3 it is seen that, contrary to the Davidovich-Haroche experiment, these observables are complementary for both $`\gamma =0`$ and $`\gamma \mathrm{},`$ complementarity being largest for $`\delta =0`$.
The measurement represented by POVM (26) is not a complete one. It can be verified that, if $`\gamma 0`$, the operator $`\widehat{T}=iCe^{i\nu \tau }S_1^{}S_2|p_+p_{}|+h.c.`$ is orthogonal to all operators of the POVM. However, for $`\delta \pm 1`$ no parameter values exist for which subspace $`_{\{\widehat{M}_m\}}`$ has dimension smaller than $`3`$. This demonstrates the informational superiority of the present measurement, based on homodyning, over the one measuring photon number. By refining the partition $`(𝐂^+,𝐂^{})`$ of the complex plane POVM (26) can easily be refined to one spanning the whole Hilbert-Schmidt space of $`2\times 2`$ matrices, allowing a complete determination of the incoming state $`\psi _{in}`$ of the atom.
## VI Summary and conclusions
In this paper we studied a number of atomic beam experiments related to the Ramsey experiment. Whereas this latter experiment is a pure interference measurement, in the experiments studied here also ‘which-way’ information can be obtained. This is achieved by inserting a third microwave cavity, $`C`$, between the ones already present in the Ramsey experiment, and measuring some observable of the cavity $`C`$ field after the atom has passed. Three different measurement arrangements were considered. In the first (referred to as the Davidovich-Haroche experiment) a second atom was used as a probe, in the second a measurement of photon number is performed instead. In the third arrangement homodyne optical detection of the cavity $`C`$ field is contemplated. The experiments were demonstrated to yield new examples of generalized measurements, to be described by positive operator-valued measures. POVMs were calculated explicitly for different values of the experimental parameters.
The experiments are interesting for two reasons. In the first place they can be interpreted as joint non-ideal measurements of incompatible observables, thus clarifying the notion of complementarity. It was found that, although all measurements satisfy inequality (4), only the last experiment exhibits complementarity in the sense that there exist two limiting values of the experimental parameters, for one of which the measurement is a pure interference measurement in which ‘which-way’ information is maximally disturbed, whereas in the other limit it is a pure ‘which-way’ measurement in which no interference can be observed. It was demonstrated that in general by the other measurement arrangements this “classical” type of complementarity need not be satisfied. In the second place, comparison of the different measurements can give insight into the question of which information is provided by a (generalized) quantum mechanical measurement. For this purpose the subspaces of Hilbert-Schmidt space, spanned by the operators of the POVM, were determined for different measurement arrangements and different values of the parameters. It was found that a measurement of the second atom in the Davidovich-Haroche experiment is equivalent to a non-ideal measurement of cavity $`C`$ photon number. Also with respect to measurement of the initial state of the atom this equivalence turns out to hold. In this respect the third arrangement, in which the photon number measurement is replaced by a measurement yielding also phase information, is shown to be superior.
An interesting aspect of the generalized measurements considered here, is that the measured probability distributions dependent on the initial phase of the microwave fields. This makes an experimental realization particularly challenging.
## Acknowledgment
The authors thank Maarten Jansen for his contribution to the calculations.
APPENDIX: POSITIVITY OF $`\widehat{\rho }_{\{\widehat{M}_m\}}`$ IN THE TWO-DIMENSIONAL CASE
In this appendix we prove that on a two-dimensional Hilbert space the operator $`\widehat{\rho }_{\{\widehat{M}_m\}}`$, obtained by projecting density operator $`\widehat{\rho }`$ according to (7), is a non-negative operator. We assume that the elements of POVM $`\{\widehat{M}_m\}`$ are linearly independent and consider the non-trivial situation $`\{\widehat{M}_m\}\{\widehat{I}\}`$ (if $`\{\widehat{M}_m\}`$ is uninformative, i.e. $`\{\widehat{M}_m\}=\{\widehat{I}\},`$ we get $`\widehat{\rho }_{\{\widehat{M}_m\}}=\frac{1}{2}\widehat{I}>\widehat{O}`$). Then the dimension of the subspace $`_{\{\widehat{M}_m\}}`$ spanned by the elements of $`\{\widehat{M}_m\}`$ is greater than $`1,`$ and it is easy to prove that $`_{\{\widehat{M}_m\}}`$ contains a maximal PVM $`\{\widehat{P}_n\}`$. Since the subspace $`_{\{\widehat{P}_n\}}`$ is a subspace of $`_{\{\widehat{M}_m\}},`$ the orthogonal projections $`𝒫_{\{\widehat{P}_n\}}`$ and $`𝒫_{\{\widehat{M}_m\}}`$ satisfy $`𝒫_{\{\widehat{P}_n\}}𝒫_{\{\widehat{M}_m\}}=𝒫_{\{\widehat{P}_n\}}`$, which implies $`Tr\widehat{\rho }_{\{\widehat{M}_m\}}\widehat{P}_n=Tr\widehat{\rho }\widehat{P}_n`$. So in the $`\{\widehat{P}_n\}`$-representation we get
$`\widehat{\rho }=\left(\begin{array}{cc}p& q\\ q^{}& 1p\end{array}\right)\text{ ; }\widehat{\rho }_{\{\widehat{M}_m\}}=\left(\begin{array}{cc}p& r\\ r^{}& 1p\end{array}\right)\text{.}`$
Because of the fact that $`𝒫_{\{\widehat{M}_m\}}`$ is an orthogonal projection onto $`_{\{\widehat{M}_m\}}`$ we should also have $`Tr\widehat{\rho }_{\{\widehat{M}_m\}}\widehat{\rho }_{\{\widehat{M}_m\}}^{}=0`$, with $`\widehat{\rho }_{\{\widehat{M}_m\}}^{}=\left(𝒫_{\{\widehat{M}_m\}}\right)\widehat{\rho }`$, implying $`Re\left(q^{}r\right)\left|r\right|^2=0`$, and hence
$$\left|q\right|\left|r\right|\text{.}$$
(30)
Denoting the eigenvalues of $`\widehat{\rho }_{\{\widehat{M}_m\}}`$ by $`\lambda _1`$ and $`\lambda _2`$, we find $`\lambda _1\lambda _2=p\left(1p\right)\left|r\right|^2`$. We already know that $`\lambda _1+\lambda _2=1`$. So both eigenvalues are non-negative if
$$p\left(1p\right)\left|r\right|^20.$$
(31)
But since $`\widehat{\rho }>\widehat{O}`$ implies $`p\left(1p\right)\left|q\right|^2>0`$ it directly follows from (30) that condition (31) is satisfied. Hence in the two-dimensional case $`\widehat{\rho }_{\{\widehat{M}_m\}}`$ is a non-negative operator.
|
warning/0003/cond-mat0003208.html
|
ar5iv
|
text
|
# Compressibility crossover and quantum opening of a gap for two dimensional disordered clusters with Coulomb repulsion
## 1 Introduction
The interplay of disorder and interactions in two dimensional ($`2d`$) electronic systems is currently a central problem in condensed matter physics loc99 . The importance of interactions is illustrated by several phenomena, notably the discovery of a metallic phase in various two-dimensional devices Kravchenko . Since the metal-insulator transition occurs at a Coulomb energy to Fermi energy ratio $`r_s10`$, Coulomb repulsion should be crucial in understanding this phenomenon. Compressibility measurements indicate that the high $`r_s`$ insulating phase is incompressible comp1 and spatially inhomogeneous comp2 .
In weakly disordered quantum dots, the inverse compressibility gives the spacing between adjacent conductance peaks. The peak spacing statistics, obtained in experiments in which $`r_s1`$, displays fluctuations larger than those predicted from Random Matrix Theory, characterized by a Gaussian-like distribution instead of a Wigner-Dyson distribution Sivan ; Marcus ; Simmel . This suggests a breakdown of the naive single particle picture, assumed in the constant interaction model.
In strongly disordered insulators, the long range nature of interactions leads to a soft Coulomb gap in the single particle density of states at the Fermi level ES , observed in tunneling spectroscopic experiments in $`3d`$ Lee and $`2d`$ Ashoori nonmetallic semiconductors. Assuming that single electron hoppings dominate the transport, the Coulomb gap leads to a crossover in the temperature dependence of the resistivity $`\rho (T)`$ from the Mott variable range hopping law ($`\rho (T)=\rho _M\mathrm{exp}(T_0/T)^{1/3}`$ in dimension $`d=2`$) to the Efros-Shklovskii behavior ($`\rho (T)=\rho _{ES}\mathrm{exp}(T_{ES}/T)^{1/2}`$) ES . This crossover has been reported khondaker not only decreasing temperature but also decreasing carrier density (around $`r_s1.7`$).
In this paper, we numerically investigate the physics of $`2d`$ Coulomb interacting spinless fermions in a random potential for relatively low values of the factor $`r_s`$ ($`3`$), when the single particle spectra display either Wigner-Dyson statistics for weak disorder (diffusion) or Poisson statistics for strong disorder (Anderson localization). As exact diagonalization studies are restricted to small system sizes paper1 , we use a numerical method CIphys1 ; CIphys2 familiar in quantum chemistry as the configuration interaction method (CIM) CIchem . The method consists in diagonalizing the Hamiltonian in an energetically truncated basis built of the low-energy states of the corresponding Hartree-Fock (HF) Hamiltonian. This method allows us to study the ground state and the lowest energy excitations for different system sizes $`L`$ at constant electronic density.
We show that the average inverse compressibility $`\mathrm{\Delta }_2`$ exhibits a smooth crossover from a $`1/L^2`$ towards a $`1/L`$ decay when $`r_s`$ increases from $`0`$ to $`3`$. At the same time, the distribution of $`\mathrm{\Delta }_2`$ evolves from the Wigner-Dyson distribution (or the Poisson distribution, if disorder is strong enough or $`L`$ large enough) towards a Gaussian-like shape. Therefore fluctuations in $`\mathrm{\Delta }_2`$ are determined by Coulomb repulsion rather than from single particle level fluctuations.
We have also studied the energy level spacing between the ground state and the first excitation for strongly disordered clusters: the average indicates a smooth opening of a gap when $`r_s`$ increases, while the distribution exhibits a sharp Poisson-Wigner-like transition at $`r_s=r_s^C1.2`$. The critical threshold $`r_s^C`$ is characterized by a scale invariant gap distribution, reminiscent of the one particle problem Shklovskii at a mobility edge. However, it is only the distribution of the first spacing which exhibits such a transition, the distributions of the next spacings remain Poissonian and are essentially unchanged when $`r_s`$ varies.
The paper is organized as follows: in Section 2 we introduce the model, in Section 3 we discuss the numerical technique used and its limits of validity; in Section 4 we present our compressibility data; in Section 5 we discuss the statistical properties of the lowest energy excitations; in Section 6 we discuss the implications of our results for hopping conductivity experiments; in Section 7 we present our conclusions.
## 2 The model
We consider a disordered square lattice with $`M=L^2`$ sites occupied by $`N`$ spinless fermions. The Hamiltonian reads
$$H=t\underset{<i,j>}{}c_i^{}c_j+\underset{i}{}v_in_i+U\underset{ij}{}\frac{n_in_j}{2r_{ij}},$$
(1)
where $`c_i^{}`$ ($`c_i`$) creates (destroys) an electron in the site $`i`$, the hopping term $`t`$ between nearest neighbours characterizes the kinetic energy, $`v_i`$ the site potentials taken at random inside the interval $`[W/2,+W/2]`$, $`n_i=c_i^{}c_i`$ is the occupation number at site $`i`$ and $`U`$ measures the strength of the Coulomb repulsion. The boundary conditions are periodic and $`r_{ij}`$ is the inter-particle distance for a $`2d`$ torus. If $`a_B^{}=\mathrm{}^2ϵ/(m^{}e^2)`$, $`m^{}`$, $`ϵ`$, $`a`$ and $`n_s=N/(aL^2)`$ denote respectively the effective Bohr radius, the effective mass, the dielectric constant, the lattice spacing and the carrier density, the factor $`r_s`$ is given by:
$$r_s=\frac{1}{\sqrt{\pi n_s}a_B^{}}=\frac{U}{2t\sqrt{\pi n_e}},$$
(2)
since in our units $`\mathrm{}^2/(2m^{}a^2)t`$, $`e^2/(ϵa)U`$ and $`n_e=N/L^2`$.
## 3 Configuration interaction method
A numerical study of the model (1) via exact diagonalization techniques for sparse matrices is possible only for small systems paper1 , and does not allow us to vary $`L`$ for a constant density. We are obliged to look for an approximate solution of the problem, using the Hartree-Fock orbitals, and to control the validity of the approximations. One starts from the HF Hamiltonian where the two-body part is reduced to an effective single particle Hamiltonian Kato ; Poilblanc ; Schreiber
$$\begin{array}{c}U(\underset{ij}{}\frac{1}{r_{ij}}n_in_j\underset{ij}{}\frac{1}{r_{ij}}c_i^{}c_jc_j^{}c_i),\end{array}$$
(3)
where $`\mathrm{}`$ stands for the expectation value with respect to the HF ground state, which has to be determined self-consistently. For large values of the interaction and large system sizes the single-particle problem (3) is still non-trivial, since the self-consistent iteration can be trapped in metastable states. This limits our study to small $`r_s`$ and forbids us to study by this method charge crystallization discussed in paper1 at a larger $`r_s^W10`$ .
Fig. 1 shows that the HF approximation gives a good estimate of the ground state energy. This approximation becomes worse for (1) larger interactions or (2) smaller disorder. The first characteristic is due to the fact that the ground state HF energy is exact in the limits $`U0`$ and $`U\mathrm{}`$, in which the ground state is given by a single Slater determinant, but deviations from this simple picture are expected at intermediate $`U`$ values. The latter feature can be understood as complicated many-body screening effects, which are effective up to a distance of the order of the localization length, cannot be reproduced within simple HF.
The main advantage of the HF approximation is that it reproduces well the single particle density of statesSchreiber , particularly the soft Coulomb gap at the Fermi energy ES . However, the approximations involved in the HF method are uncontrolled. The mean field HF results can be improved using the configuration interaction method CIphys1 ; CIphys2 ; CIchem . Once a complete orthonormal basis of HF orbitals has been calculated:
$$H_{HF}(|\psi _1,\mathrm{},|\psi _N)|\psi _\alpha =ϵ_\alpha |\psi _\alpha ,$$
(4)
with $`\alpha =1,2,\mathrm{},L^2`$, it is possible to build up a Slater determinants’ basis for the many-body problem which can be truncated to the $`N_H`$ first Slater determinants, ordered by increasing energies. The two-body Hamiltonian can be written as
$$H_{\mathrm{int}}=\frac{1}{2}\underset{\alpha ,\beta ,\gamma ,\delta }{}Q_{\alpha \beta }^{\gamma \delta }d_\alpha ^{}d_\beta ^{}d_\delta d_\gamma ,$$
(5)
with
$$Q_{\alpha \beta }^{\gamma \delta }=U\underset{ij}{}\frac{\psi _\alpha (i)\psi _\beta (j)\psi _\gamma (i)\psi _\delta (j)}{r_{ij}}$$
(6)
and $`d_\alpha ^{}=_j\psi _\alpha (j)c_j^{}|0`$. One gets the residual interaction subtracting Eq. 3 from Eq. 5. This keeps the two-body nature of the Coulomb interaction, and if $`N2`$ it is still possible to take advantage of the sparsity of the matrix and to diagonalize it via the Lanczos algorithm.
Fig. 2 compares HF and CIM results. Labelling the $`N`$-body levels $`E_i`$ ($`i=0,1,2,\mathrm{}`$) by increasing energy we have studied the first spacing $`\mathrm{\Delta }E_0=E_1E_0`$. When $`r_s>1`$, the residual interaction significantly reduces the mean gap, and slightly changes its distribution. Within HF approximation, the first excitation is a particle-hole excitation starting from the ground state: The mean gap reduction is due to correlation effects beyond the particle-hole interaction. For example, the electron (hole) polarizes its neighborhood as it creates fluctuations in the local charge density.
We have checked that CIM results agree with the results given from exact diagonalization with an accuracy of the order $`2\%`$ when one takes into account the $`N_H=10^3`$ lowest energy Slater determinants when $`N=4`$, $`L=8`$, $`W=15`$, and $`r_s=5`$. This means that a basis spanning only $`0.2\%`$ of the total Hilbert space is sufficient for studying the first excitations. For larger $`L`$, exact diagonalization is no longer possible, but one can look the variation of the results when $`N_H`$ increases. In the worst case considered ($`N=16`$, $`L=16`$, $`W=15`$, $`r_s=2.8`$) the accuracy in the first four spacings can be estimated as better than $`10\%`$ when $`N_H=2\times 10^3`$ (see Fig. 3).
Therefore the CIM method allows to study low energy level statistics for $`r_s<3`$. However, its accuracy is not sufficient to determine more sensitive quantities, like a small change of the ground state energy when the boundary conditions are twisted (i.e. the persistent currents).
## 4 Electronic compressibilty
In quantum dots experiments in the Coulomb blockade regime the spacing between consecutive conductance peaks is given by
$$\mathrm{\Delta }_2(N)=E_0(N+1)2E_0(N)+E_0(N1),$$
(7)
with $`E_0(N)`$ $`N`$-body ground state. This quantity is the discretized second derivative of the ground state energy with respect to the number of particles, i.e. the inverse compressibility. In the constant interaction model, which ignores fluctuations in the Coulomb energy
$$Q_{\alpha \beta }^{\gamma \delta }=\frac{e^2}{C}\delta _{\alpha \gamma }\delta _{\beta \delta },$$
(8)
where $`C`$ is the capacitance of the dot. This gives $`E_0(N)=e^2N(N1)/2C+_{k=1}^N\eta _k`$, where $`\eta _k`$ is the $`k`$-th single particle energy at $`U=0`$. Substituting this expression into Eq. (7) one finds $`\mathrm{\Delta }_2=e^2/C+\eta _{N+1}\eta _N`$. Random Matrix Theory describes the single particle level spacings when the disorder is weak. The peak spacing distribution is expected to follow the Wigner-Dyson distribution with fluctuations in $`\mathrm{\Delta }_2`$ whose standard deviation is given by $`\delta \mathrm{\Delta }_2\sqrt{<\mathrm{\Delta }_2^2><\mathrm{\Delta }_2>^2}=\sqrt{4/\pi 1}<\mathrm{\Delta }>0.52<\mathrm{\Delta }>`$, where $`<\mathrm{\Delta }>`$ is the single particle mean level spacing. However, experiments Sivan ; Marcus ; Simmel performed at $`r_s1`$ found a distribution which is Gaussian-like and has a larger width, up to $`7.5\mathrm{\Delta }`$ Simmel . This suggests that fluctuations in $`\mathrm{\Delta }_2`$ are dominated by electron-electron interactions. Several observed features of the peak spacing distributions have been reproduced using exact numerical diagonalization of the Hamiltonian model (1) Sivan , HF calculations Levit ; Walker ; Berkovits , and a random matrix model for interacting fermionic systems Alhassid .
In this paper we study the inverse compressibility at different system sizes, for a constant filling factor $`n_e=1/9`$. We consider $`N=4,9,16`$ particles on square lattices of size $`L=6,9,12`$ respectively. A Fermi golden rule approximation for the elastic scattering time gives Walker , for $`n_e1`$, $`k_Fl192\pi n_e(t/W)^2`$, $`k_F`$ and $`l`$ denoting the Fermi wave vector and the elastic mean free path, respectively. We consider $`W=5,8,15`$, corresponding to $`k_Fl2.7,1,0.3`$; the first case corresponds to a diffusive system ($`L>l/a2.3`$, with $`a`$ lattice spacing), the last one to a strongly localized system.
The following inverse compressibility data are obtained with the CIM. We have checked that the residual interaction does not change qualitatively HF results. This is consistent with Fig. 1: HF approximation gives a good estimate of the ground state energy and the inverse compressibility is a physical observable which depends only on the ground state energies at different number of particles. Neither a precise knowledge of the ground state wavefuction (as in the calculation of persistent currents) nor excited states energies (as in studies of spectral statistics) are required.
Fig. 4 shows the $`L`$-dependence of the average inverse compressibility, which is well fitted by the power law $`<\mathrm{\Delta }_2(r_s)>L^{\alpha (r_s)}`$, with $`\alpha (r_s)`$ going from $`2`$ to $`1`$ when $`r_s`$ goes from $`0`$ to $`3`$ approximately. The value $`\alpha =2`$ is expected without interaction ($`<\mathrm{\Delta }_2>=<\mathrm{\Delta }>1/L^2`$). The exponent $`\alpha =1`$ can be understood as follows: for $`r_s>1`$, $`\mathrm{\Delta }_2`$ is dominated by the charging energy and therefore $`<\mathrm{\Delta }_2>1/C1/L`$. In the strongly localized regime ($`W=15`$) the $`1/L`$ law is justified due to the Coulomb gap in the single particle density of states ES : According to Koopmans’ theorem note , assuming that the other charges are not reorganized by the addition of an extra charge,
$$\mathrm{\Delta }_2ϵ_{N+1}ϵ_N\frac{1}{L}$$
(9)
due to the Coulomb gap, where $`ϵ_k`$ is the energy of the $`k`$-th HF orbital at a fixed number $`N`$ of particles.
Fig. 5 shows the distribution of inverse compressibilities, for $`L=12`$, $`W=5`$ and $`W=15`$ (inset). At $`r_s=0`$, $`\mathrm{\Delta }_2`$ distributions are close to the Wigner surmise $`P_W(s=\mathrm{\Delta }_2/<\mathrm{\Delta }_2>)=(\pi s/2)\mathrm{exp}(\pi s^2/4)`$ for $`W=5`$ and to the Poisson distribution $`P_P(s)=\mathrm{exp}(s)`$ for $`W=15`$. In the latter case, deviations from the Poisson distribution at small $`s`$ values are due to the finite system size. In both cases at $`r_s=2.5`$ distributions show a Gaussian shape. This can be understood within the Koopmans’ theorem, as the HF energies are given by
$$ϵ_k=\psi _k|H_1|\psi _k+\underset{\alpha =1}{\overset{N}{}}\left(Q_{\alpha k}^{\alpha k}Q_{\alpha k}^{k\alpha }\right),$$
(10)
with $`H_1`$ one-body part of the Hamiltonian (1): Due to the small correlations of eigenfunctions in a random potential, one can reasonably invoke the central limit theorem (see Ref. Berkovits for a more detailed discussion).
Fig. 6 shows the $`r_s`$-dependence of the inverse compressibility standard deviation $`\delta \mathrm{\Delta }_2`$, normalized to the single particle mean level spacing $`<\mathrm{\Delta }>`$. One can see that $`\delta \mathrm{\Delta }_2/<\mathrm{\Delta }>`$ increases slowly for $`r_s<1`$ and then more rapidly. This results are in contrast with random phase approximation estimates BA ; Blanter , giving fluctuations of the order $`\mathrm{\Delta }`$. On the contrary, we have checked that $`\delta \mathrm{\Delta }_2/<\mathrm{\Delta }_2>`$ has a weak $`r_s`$ dependence for $`r_s>1`$, in agreement with findings of Ref. Sivan . This confirms that peak spacing fluctuations are dominated by capacitance fluctuations instead of single particle fluctuations.
We point out that our data are limited to spinless fermions. Interesting interaction-induced spin effects have been recently reported in experiments Ensslin and theoretically investigated Brouwer ; Baranger ; Jacquod .
## 5 First energy excitation
We have calculated the first $`N`$-body energy levels $`E_i`$, $`(i=0,1,2,\mathrm{})`$ for different sizes $`L`$, with a large disorder to hopping ratio $`W/t=15`$ imposed to have Anderson localization and Poissonian spectral statistics for the one particle levels at $`r_s=0`$ when $`L8`$. We studied $`N=4,9`$, and $`16`$ particles inside clusters of size $`L=8,12`$, and $`16`$ respectively. This corresponds to a constant low filling factor $`n_e=1/16`$. We studied an ensemble of $`10^4`$ disorder configurations.
The first average spacing $`<\mathrm{\Delta }E_0>`$ is given in Fig. 7. It exhibits a power law decay as $`L`$ increases, with an exponent $`\beta `$ given in the inset. One finds for the first spacing that $`\beta `$ linearly decreases from $`d=2`$ to $`1`$ when $`r_s`$ increases from $`0`$ to $`3`$. The next mean spacings $`<\mathrm{\Delta }E_i>=<E_{i+1}E_i>`$ depend more weakly on $`r_s`$, as shown in Fig. 7.
For $`r_s=0`$, the distribution of the first spacing $`s=\mathrm{\Delta }E_0/<\mathrm{\Delta }E_0>`$ becomes closer and closer to the Poisson distribution $`P_P(s)`$ when $`L`$ increases, as it should be for an Anderson insulator. For a larger $`r_s`$, the distribution seems to become close to the Wigner surmise $`P_W(s)`$ characteristic of level repulsion in Random Matrix Theory, as shown in Fig. 8 for $`r_s=2.8`$ and $`L=16`$. To study how this $`P(s)`$ goes from Poisson to a Wigner-like distribution when $`r_s`$ increases, we have calculated a parameter $`\eta `$ which decreases from $`1`$ to $`0`$ when $`P(s)`$ goes from Poisson to Wigner:
$$\eta =\frac{\text{var}(P(s))\text{var}(P_W(s))}{\text{var}(P_P(s))\text{var}(P_W(s))},$$
(11)
where $`\text{var}(P(s))`$ denotes the variance of $`P(s)`$, $`\text{var}(P_P(s))=1`$ and $`\text{var}(P_W(s))=0.273`$. In Fig. 9, one can see that the three curves $`\eta (r_s)`$ characterizing the first spacing for $`L=8,12,16`$ intersect at a critical value $`r_s^C1.2`$. Our data suggest that for $`r_s<r_s^C`$ the distribution tends to Poisson in the thermodynamic limit, while for $`r_s>r_s^C`$ it tends to a Wigner-like behavior glass . At the threshold $`r_s^C`$, there is a size-independent intermediate distribution shown in the inset of Fig. 8, exhibiting level repulsion at small $`s`$ followed by a $`\mathrm{exp}(as)`$ decay at large $`s`$ with $`a1.52`$. This Poisson-Wigner transition characterizes only the first spacing, the distributions of the next spacings being quite different. The inset of Fig. 9 does not show an intersection for the parameter $`\eta `$ calculated with the second spacing. The second excitation is less localized than the first one when $`r_s=0`$, since the one particle localization length weakly increases with energy. It is only for $`L=16`$ that the distribution of the second spacing becomes close to Poisson without interaction, and a weak level repulsion occurs as $`r_s`$ increases.
The observed transition, and the difference between the first spacing and the following ones is mainly an effect of the HF mean field. For the first spacing, the curves $`\eta `$ calculated with the HF data are qualitatively the same. At the mean field level the low energy excitations are particle-hole excitations starting from the ground state. The energy spacing between the first and the second excited states is given by the difference of two particle-hole excitations and a Poisson distribution follows if the low energy particle-hole excitations are uncorrelated.
Notice that the energy of an electron-hole pair is given by $`ϵ_jϵ_iU/r_{ij}`$ and the classical argument for the existence of a gap in the single particle density of states does not apply ES . For example, in the constant interaction model one pays a gap $`e^2/C`$ in the single particle density of states but not in the many-body excitation spectrum ($`\mathrm{\Delta }E_0=\mathrm{\Delta }E_0(U=0)1/L^2`$). Therefore the observed opening of a gap for the first energy excitation is a remarkable phenomenon beyond the predictions of the classical Coulomb gap model.
## 6 Hopping conductivity
In disordered insulators, a crossover ES in the temperature dependence of the resistivity $`\rho (T)`$ is induced by Coulomb repulsion from the Mott variable range hopping law ($`\rho (T)=\rho _M\mathrm{exp}(T_0/T)^{1/3}`$ in $`2d`$) to the Efros-Shklovskii behavior ($`\rho (T)=\rho _{\mathrm{ES}}\mathrm{exp}(T_{\mathrm{ES}}/T)^{1/2}`$).
The usual argument is to consider the length $`L(T)`$ which maximizes the probability $`P_{ij}`$ of a single particle hop from a state with energy $`ϵ_i`$ to a state with energy $`ϵ_j`$:
$$P_{ij}\mathrm{exp}\left(\frac{2L}{\xi }\frac{ϵ_jϵ_i}{kT}\right),$$
(12)
with $`\xi `$ localization length.
To measure possible delocalization effects, we have calculated for $`W=15`$ the number of occupied sites per particle $`\xi _s=N/_i\rho _i^2`$, where $`\rho _i=\mathrm{\Psi }_0|n_i|\mathrm{\Psi }_0`$ is the charge density of the ground state at the site $`i`$. Around $`r_s1.2`$ and after ensemble average, the maximum increase of $`\xi _s`$ compared to $`r_s=0`$ is negligibly small ($`2\%`$). Therefore noticeable effects come mainly from the mean and the distribution of the single particle density of states around the Fermi level.
At low temperatures, hopping conductivity involves states near the Fermi level. If one takes for the Coulomb gap its average value $`ϵ_{N+1}ϵ_N\mathrm{\Delta }_2(A+Br_s)/L^{\alpha (r_s)}`$ (see Fig. 4), one obtains for the hopping resistivity a smooth and continuous crossover from Mott to Efros-Shklovskii hopping, given by:
$$\rho (T)\mathrm{exp}\left(\frac{T(r_s)}{T}\right)^{1/(\alpha +1)},$$
(13)
where
$$T(r_s)\frac{A+Br_s}{k\xi ^\alpha }.$$
(14)
This prediction neglects the crossover from the Poisson distribution to a Gaussian distribution in the inverse compressibility (see Fig. 5), which could be included by considering for $`\mathrm{\Delta }_2`$ a more typical value than its average, for instance the value $`\mathrm{\Delta }_2(b)`$ for which $`_0^{\mathrm{\Delta }_2(b)}P(\mathrm{\Delta }_2^{^{}})𝑑\mathrm{\Delta }_2^{^{}}=b`$, with $`b=1/2`$ for example. This would introduce a crossover in $`T(r_s)`$ around $`r_s1`$.
The existence of a critical $`r_s`$ value for the opening of the Coulomb gap can be understood similarly to PD99 . The single particle density of states around the Fermi energy $`E_F`$ is given by ES $`\rho (E)|EE_F|/U^2`$ and the gap size $`\mathrm{\Delta }_g=|E_gE_F|`$ can be estimated from the condition $`\rho (E_g)\overline{\rho }`$, with $`\overline{\rho }1/W`$ mean density of states for $`Wt`$, obtaining $`\mathrm{\Delta }_gU^2/W`$. According to Fermi golden rule, the inverse lifetime of a Slater determinant built from electrons localized at given sites is $`\mathrm{\Gamma }_tt^2(1/W)(N/L^2)`$, with $`N/(WL^2)`$ density of states directly coupled by the hopping term of the Hamiltonian (1). This argument suggests that quantum fluctuations can melt the Coulomb gap for $`\mathrm{\Gamma }_t\mathrm{\Delta }_g`$, giving $`r_s1`$. Therefore a crossover from Efros-Shklovskii to Mott hopping conductivity is expected not only increasing temperature but also increasing carrier density, as observed in khondaker at $`r_s1.7`$.
Notice that in the above arguments we have considered only the excitations leading to the Efros-Shklovskii behavior for the hopping conductivity, that correspond to the withdrawal of one electron or its addition to the system. However, a single electron hop may reorganize the location of the other particles, inducing complex many particle excitations (see Fig. 2, showing that simple HF is unable to reproduce the mean first many-body energy spacing when $`r_s>1`$). Therefore one cannot exclude that correlated hops involving more than one particle affect the hopping conductivity ES .
## 7 Conclusions
In summary, we have analyzed the inverse compressibility and the first energy excitation statistics for spinless fermions in weakly and strongly disordered squared lattices when $`r_s`$ increases from $`0`$ to $`3`$. On one hand, we have found a crossover in the electronic inverse compressibility from a $`1/L^2`$ towards a $`1/L`$ decay; at the same time its distribution evolves towards a Gaussian shape. On the contrary, our data for the first many-body energy spacing are consistent with a smooth opening of a gap and a sharp interaction-induced Poisson-Wigner-like transition in its distribution at $`r_s^C1.2`$, with a scale invariant distribution at the critical point showing intermediate statistics. For those small values of $`r_s`$, the ground state energy is essentially given by an effective Hartree-Fock mean field. We underline that the intermediate quantum regime discussed in Ref. paper1 and charge crystallization occur for larger $`r_s`$ factors, where the mean-field approximation breaks down and the configuration interaction method ceases to efficiently converge. This does not allow us to study the regime characterizing the experiments reported in Refs. comp1 ; comp2 .
###### Acknowledgements.
Partial support from the TMR network “Phase Coherent Dynamics of Hybrid Nanostructures” of the European Union is gratefully acknowledged.
|
warning/0003/cond-mat0003068.html
|
ar5iv
|
text
|
# Possible critical point in phase diagrams of interlayer Josephson-vortex systems in high-𝑇_𝑐 superconductors
## Abstract
A critical value in the product of the anisotropy parameter and the magnetic field is observed in interlayer Josephson-vortex systems by extensive Monte Carlo simulations. Below/above this critical value the thermodynamic phase transition between the normal and the superconducting states upon temperature sweeping is first/second order. The origin is the intrinsic pinning effect of the layered structure of high-$`T_c`$ superconductors.
PACS numbers: 74.60.Ge, 74.25.Bt, 74.25.Dw, 74.20.De
In the superconducting state, an external magnetic field applied in the Cu-O plane of a high-$`T_c`$ superconductor induces the so-called Josephson vortices. The center of a Josephson vortex enters into a block layer, the layer between two superconducting Cu-O layers, in order to save the condensation energy of superconductivity . The thermodynamic phase transition and the lattice structure of interlayer Josephson vortices have been attracting considerable interests since the discovery of high-$`T_c`$ superconductivity. Using a London theory, the structure of Josephson-vortex lattice was derived as the compressed hexagons of triangular lattice pointing along the c axis by Ivlev, Kopnin and Pokrovsky . The interlayer shear modulus was shown to be exponentially small, and the shear deformation of a rhombic lattice might arise through a second-order phase transition. However, at higher temperatures fluctuations are more important, and the London theory is generally inaccurate for discussions about phase transitions. Considerable efforts have been made in order to clarify the thermodynamic phase transition in Josephson-vortex systems both experimentally and theoretically thereafter. Nevertheless, the understanding for the problem is still not satisfactory yet. Results obtained by different techniques even seem to be contrary to each other. The difficulty in approaching this problem is two-fold. On the experimental side, a small deviation of the direction of the magnetic field from the Cu-O plane can lead to a strong influence of the c-axis component of the magnetic field on thermodynamic properties of the systems, since all the high-$`T_c`$ superconductors are very anisotropic. On the theoretical side, one has to treat simultaneously anisotropic inter-vortex forces, the commensuration between the vortex alignment and the underlying layered structure, and thermal fluctuations.
In the present Letter, we report new results of extensive Monte Carlo (MC) simulations on the thermodynamic phase transition and the lattice structure of interlayer Josephson vortices. Our results suggest the existence of a critical value of product of the anisotropy parameter and the magnetic field, such that below/above this critical value the thermodynamic phase transition between the normal and the superconducting states is first/second order upon temperature sweeping.
The model Hamiltonian used for the present simulations is the so-called 3D anisotropic, frustrated XY model defined on the simple cubic lattice :
$$H=J\underset{<i,j>x\mathrm{axis}}{}\mathrm{cos}(\phi _i\phi _j)J\underset{<i,j>y\mathrm{axis}}{}\mathrm{cos}(\phi _i\phi _j)\frac{J}{\gamma ^2}\underset{<i,j>c\mathrm{axis}}{}\mathrm{cos}(\phi _i\phi _j\frac{2\pi }{\varphi _0}_i^jA_c𝑑r_c).$$
(1)
Here the y axis is along the external magnetic field, and $`𝐲𝐜𝐱`$. The unit length of the simple cubic lattice is the distance $`d`$ between the neighboring Cu-O layers in a cuprate. Therefore, the discreteness in the c axis comes from the underlying layered structure of cuprates, while that in the Cu-O planes is introduced merely for computer simulations. The coupling constant is given by $`J=\varphi _0^2d/16\pi ^3\lambda _{ab}^2`$. The anisotropy parameter is defined by $`\gamma =\lambda _c/\lambda _{ab}`$, and determines the ratio between the couplings in the Cu-O plane and along the c axis. In the present model, fluctuations in amplitudes of superconducting order parameters and in the magnetic induction are neglected.
Details of simulation technique are summarized as follows: The density of flux lines induced by the external magnetic field is $`f=Bd^2/\varphi _0`$. A Landau gauge is adopted so that $`A_x=0`$ and $`A_c=xB`$. The system size is $`L_x\times L_y\times L_c=384d\times 200d\times 20d`$, which is compatible with the filling factor $`f=1/32`$. There are 240 Josephson flux lines in the ground state. Periodic boundary conditions are applied on phase variables in all directions. A typical simulation process is started from a random configuration of the phase variables at a high temperature, such as $`T=1.5J/k_B`$. 30000 and 90000 MC sweeps are used for equilibration and statistics, respectively, at each temperature. The last configuration at a temperature is used as the initial configuration at a slightly lower temperature, where the temperature difference is $`\mathrm{\Delta }T=0.1J/k_B`$. Around the transition temperature, more than one million MC sweeps are adopted at each temperature, and meanwhile the cooling rate is reduced to $`\mathrm{\Delta }T=0.01J/k_B`$. Vortices are identified by counting phase differences around plaquettes.
In order to compare our simulation results with existing experimental observations, we choose to study first a system of anisotropy parameter $`\gamma =8`$, which is near to that of YBa<sub>2</sub>Cu<sub>3</sub>O<sub>7-δ</sub>. The magnetic field corresponding to $`f=1/32`$ in our simulations is much stronger than those in experiments, and we come back to this point later. The temperature dependence of the helicity modulus (a quantity proportional to the superfluid density) along the magnetic field and the specific heat is depicted in Fig. 1. There is a clearly observable $`\delta `$-function like peak in the specific heat at $`T_m0.96J/k_B`$, where the helicity modulus along the direction of magnetic field increase sharply from zero. Shown in the same figure is the temperature dependence of the intensities of Bragg peaks in diffraction patterns at $`𝐪_{xc}^{(1)}=(\pm \pi /8d,0)`$ and $`𝐪_{xc}^{(2)}=(\pm \pi /16d,\pm \pi /d)`$. Therefore, a thermodynamic first-order phase transition occurs at $`T_m`$, where the gauge symmetry and translation symmetry are broken simultaneously, corresponding to the realization of superconductivity and Josephson-vortex lattice respectively.
The lattice structure of Josephson vortices at low temperatures is shown in Fig. 2. The unit cell is rhombic with short axis along the c direction and of a length of $`2d`$, and the long axis along the x direction and of a length of $`32d`$. Josephson vortices are distributed in every block layer for the present parameters $`\gamma =8`$ and $`f=1/32`$. This structure is the same as that predicted by Ivlev, Kopnin and Pokrovsky .
The lattice structure in Fig. 2 is obviously the ground state for $`\gamma 8`$ when the filling factor is fixed at $`f=1/32`$. Therefore we can use it for investigations of thermodynamic properties for large anisotropy parameters by a heating process. The specific heats thus obtained are shown in Fig. 3 for anisotropy parameters $`\gamma =8`$, 9, and 10. The $`\delta `$-function peaks in the curves for $`\gamma =8`$ and 9, is suppressed for $`\gamma =10`$ . In Fig. 4 we display the temperature and anisotropy parameter dependence of the phase difference between nearest neighboring Cu-O layers $`\mathrm{cos}(\phi _n\phi _{n+1})`$. There is a jump in $`\mathrm{cos}(\phi _n\phi _{n+1})`$ for $`\gamma =8`$ and 9, which is smeared out for $`\gamma =10`$. As the jump in $`\mathrm{cos}(\phi _n\phi _{n+1})`$ is nothing but the jump in the Josephson energy in units of $`J/\gamma ^2`$, there exists a latent heat at the transition temperature for $`\gamma =8`$ and 9, but not for $`\gamma =10`$, consistently with the data for the specific heat. The value of the latent heat itself is too tiny, about $`\gamma ^2`$ times smaller than that in $`\mathrm{cos}(\phi _n\phi _{n+1})`$, to be detected directly. On the other hand, from a standard finite-size scaling theory for a first-order phase transition, the height of the $`\delta `$-function like peak in the specific heat is proportional to the system size . Therefore, by using a large system such as the one in our simulations, the $`\delta `$-function like peak in the specific heat becomes observable as in Figs. 2 and 3 for the first-order phase transitions.
We have performed simulations for anisotropy parameters $`\gamma =7,6,\mathrm{}`$, down to the isotropic case of $`\gamma =1`$ fixing the filling factor at $`f=1/32`$, and observed first-order phase transitions for all these anisotropy parameters. Therefore, the present simulation results indicate that there is a critical anisotropy parameter in between $`\gamma =9`$ and $`\gamma =10`$ for $`f=1/32`$, below/above which the phase transition is first/second order.
Now we look for the reason of the suppression of the first-order phase transition when the anisotropy parameter is increased. Suppose a complete commensuration is achieved between the alignment of the Josephson vortices shown in Fig. 2 and the underlying layered structure of high-$`T_c`$ cuprates. In other words, the Cu-O layers do not influence the lattice structure of Josephson vortices, but merely fix its position in the c direction. In such a case, the Josephson-vortex lattice should be rescaled into equilateral triangular lattice using the anisotropy parameter $`\gamma `$, and we have a relation as seen in Fig. 5:
$`(2d)^2=d^2+(d/2f\gamma )^2,`$
which results in
$$f\gamma =\frac{1}{2\sqrt{3}}.$$
(2)
Now we increase the anisotropy parameter from that determined by the above relation when the filling factor $`f`$ is fixed. Since the repulsive force between Josephson vortices in the c direction is reduced, the Josephson-vortex lattice would be compressed in this direction in order to achieve the energy minimum for the new anisotropy parameter. This reconstruction of Josephson-vortex lattice is forbidden by the underlying layered structure of the high-$`T_c`$ superconductor. Therefore, the above relation provides a criterion for onset of the intrinsic pinning effect of the layered structure on the formation of Josephson-vortex lattice. For anisotropy parameters larger than that evaluated by the above relation, the lattice structure of Josephson vortices is determined by both the inter-vortex repulsions and the pinning force of the underlying layered structure. The thermodynamic phase transition associated with the formation of the Josephson-vortex lattice can be different in the two regions divided by the above relation.
Numerically, the critical anisotropy parameter for the filling factor $`f=1/32`$ is evaluated as $`\gamma =16/\sqrt{3}9.24`$ by the relation (2). This estimate coincides well with our simulation results, since first-order phase transitions are observed for $`\gamma 9`$ but not for $`\gamma 10`$. We have also performed simulations for the filling factor $`f=1/25`$, and found the variation of phase transition from first to second order around $`\gamma =8`$. This observation is consistent with the relation (2), since for $`f=1/25`$ one has the critical anisotropy parameter $`\gamma 7.22`$. For $`f=1/36`$, we have observed a first-order phase transition even for $`\gamma =10`$, consistently with the critical value $`\gamma 10.39`$. Namely, our simulation results indicate clearly that the critical anisotropy parameter increases with decreasing filling factor, or magnetic field. Quantitatively, the simple relation (2) seems to give a reasonable estimate on the critical anisotropy parameter.
The same variation of the phase transition should be observed when the anisotropy parameter is fixed while the filling factor, or the strength of the magnetic field, is tuned. The relation (2) can be rewritten as
$$B=\frac{\varphi _0}{2\sqrt{3}\gamma d^2}.$$
(3)
For YBa<sub>2</sub>Cu<sub>3</sub>O<sub>7-δ</sub> with $`\gamma 8`$ and $`d=12\AA `$, the critical magnetic field is estimated as $`B50T`$. Therefore the phase transition in the Josephson-vortex systems in YBa<sub>2</sub>Cu<sub>3</sub>O<sub>7-δ</sub> is first order for magnetic fields available experimentally, according to our present study. For Bi<sub>2</sub>Sr<sub>2</sub>CaCu<sub>2</sub>O<sub>8+y</sub> with $`\gamma 150`$ and $`d=15\AA `$, the critical magnetic field is evaluated as $`B1.7T`$, which can be checked experimentally.
The phase transition of interlayer Josephson vortices in a layered superconductor has been addressed theoretically by Blatter, Ivlev and Rhyner and Balents and Nelson . In these theories, the Josephson-vortex lattice melts in-between the layers with the formation of a smectic-like vortex liquid, via a second-order phase transition (see also ). Therefore, for strong magnetic fields the theories give a reasonable scenario for the phenomena in Josephson-vortex systems observed in the present simulations.
The importance of thermal fluctuations in the phase transition should be stressed. For example, about one vortex-antivortex pair is thermally excited per Josephson flux at each xc section at $`T=0.8J/k_B`$, a temperature lower than the corresponding transition point, for $`\gamma =8`$ and $`f=1/32`$ . Energetically, an additional Josephson vortex only costs energy of order of $`J/\gamma ^2`$, which becomes very small when the anisotropy parameter $`\gamma `$ is large . Most of the thermally excited Josephson vortices and antivortices are confined in same block layers and form overhangs in flux lines, or closed loops. As the result, for large anisotropy parameters Josephson flux lines induced by the magnetic field collide with each other in same block layers even below the transition temperature . Therefore, these thermally excited Josephson vortices and antivortices play important roles in smearing out the first-order phase transition in a Josephson-vortex system in a layered superconductor with a large anisotropy parameter .
In summary, from extensive Monte Carlo simulations we have found a critical value in the product of the anisotropy parameter and magnetic field in interlayer Josephson-vortex systems in high-$`T_c`$ superconductors. Below/above this critical value, the thermodynamic phase transition between the normal state and the superconducting state is first/second order. According to the present results, the phase transition in Josephson-vortex systems in YBa<sub>2</sub>Cu<sub>3</sub>O<sub>7-δ</sub> is first order under magnetic fields up to $`B50T`$, while it varies from first to second order in Bi<sub>2</sub>Sr<sub>2</sub>CaCu<sub>2</sub>O<sub>8+y</sub> as the magnetic field is increased to across $`B1.7T`$.
The authors thank L. Bulaevskii, J. Clem, A. Sudbø and Y. Nonomura for useful discussions. The present simulations are performed on the Numerical Materials Simulator (SX-4) of National Research Institute for Metals (NRIM), Japan.
Fig. 1: Temperature dependence of the helicity modulus along the magnetic field, the specific heat and the intensities $`I_1`$ and $`I_2`$ for the Bragg peaks at $`𝐪_{xc}^{(1)}=(\pm \pi /8d,0)`$ and $`𝐪_{xc}^{(2)}=(\pm \pi /16d,\pm \pi /d)`$ respectively for $`\gamma =8`$ and $`f=1/32`$.
Fig. 2: Josephson vortex lattice for $`\gamma =8`$ and $`f=1/32`$ obtained by MC simulations of a cooling process from a random state at high temperatures.
Fig. 3: Temperature dependence of the specific heat for $`\gamma =8`$, 9 and 10 and $`f=1/32`$. Data for $`\gamma =8`$ and 9 are shifted by constants.
Fig. 4: Temperature dependence of the phase difference between nearest neighboring Cu-O layers for $`f=1/32`$. The lines are for eye-guide.
Fig. 5: Real-space unit cell of the Josephson-vortex lattice in a layered superconductor of a large anisotropy parameter.
|
warning/0003/hep-ph0003210.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
The flux of solar neutrinos was first measured in the Homestake mine (see and references therein) over thirty years ago. Since then, it was realized that the measured flux was significantly suppressed with respect to theoretical predictions. More recently, a handful of different experiments have also succeeded in measuring the solar neutrino flux . All experiments measure a neutrino flux which is significantly suppressed with respect to the theoretical predictions of the most recent version of the Standard Solar Model (SSM) . This thirty year old problem is what is referred to as the “solar neutrino puzzle.”
There are different types of solutions to the solar neutrino puzzle. At first sight, it appears natural to suspect that the SSM predictions for the solar neutrino flux are slightly off, and/or that the experiments have underestimated their systematic effects, given that detailed models of the Sun and neutrino experiments are highly non-trivial. However, SSM independent analyses of the neutrino data (see for a particularly nice and simple example), together with independent experimental evidence in favour of the SSM , seem to indicate that the above solution to the puzzle is strongly disfavoured.
The best solution to the solar neutrino puzzle involves extending the Standard Model of particle physics by assuming that the neutrinos have mass and that they mix, i.e., neutrino mass eigenstates are different from neutrino weak eigenstates. This possibility has become particularly natural in light of the recent strong evidence for $`\nu _\mu `$ oscillations from the atmospheric neutrino data at SuperKamiokande .
Nonetheless, in order to firmly establish that the solution to the solar neutrino puzzle involves physics beyond the Standard Model, it is necessary to come up with SSM independent, robust experimental evidence for, e.g., solar neutrino oscillations. Indeed, these “Smoking Gun” signatures of solar $`\nu _e\nu _{\mathrm{other}}`$ oscillations are among the present goals of the SuperKamiokande experiment, via the measurement of the day-night asymmetry of the solar neutrino data and the recoil electron energy spectrum , and the SNO experiment , via the measurement of the charged to neutral current ratio, the day-night asymmetry of the data, and the recoil electron kinetic energy spectrum.
Other goals of this and the next generation of neutrino experiments are, if solar neutrino oscillations are established, to determine neutrino oscillation modes and measure masses and mixing angles. The current data allow for different $`\nu _e`$ oscillation modes and a handful of disconnected regions in the mass–mixing-angle parameter space (see for two-flavour analyses and also for an extension to the “dark side” of the parameter space).
Experiments dedicated to measuring the flux of low-energy solar neutrinos ($`E_\nu =O(1001000)`$ keV) are going to be extremely useful, and perhaps crucial, in order to fully solve the solar neutrino puzzle. It was recently shown that future experiments (Borexino and, perhaps, KamLAND ) dedicated to measuring the flux of <sup>7</sup>Be neutrinos (produced by <sup>7</sup>Be+$`e^{}^7`$Li$`+\nu _e`$ inside the Sun) should be able to establish or exclude the “just-so” solution to the solar neutrino puzzle via the study of the seasonal variations of the neutrino flux , and establish or exclude the LOW MSW solution via the study of the zenith angle dependence of their data . Furthermore, the measurement of a sizable <sup>7</sup>Be neutrino flux would significantly disfavour the SMA MSW solution to the solar neutrino puzzle, especially in the case of $`\nu _e\nu _s`$ oscillations (where $`\nu _s`$ is a sterile neutrino, i.e., a standard model singlet), and significantly constrain the SSM independent analysis, which require the flux of <sup>7</sup>Be neutrinos to be virtually absent . Finally, we have shown that, in the advent that the background rates at Borexino and/or KamLAND are exceptionally low, it should be possible to measure a nonzero component of $`\nu _{\mu ,\tau }`$ in the solar neutrino flux by analysing the recoil electron kinetic energy spectrum.
Another exciting possibility is that of measuring the “fundamental” $`pp`$-neutrinos, which are produced in the interior of the Sun by proton-proton fusion ($`p+p^2`$H$`+e^++\nu _e`$) in a real time experiment. Future experiments, such as HELLAZ, HERON, LENS, etc (see for an overview) are being designed to do just that. The flux of $`pp`$-neutrinos is particularly constrained by the photon flux i.e., the Sun’s luminosity, which, of course, is very well measured on the Earth. These lowest energy solar neutrinos ($`E_\nu \begin{array}{c}<\hfill \\ \hfill \end{array}420`$ keV) are not only the most abundant ones, but also have the best known flux. Their energy spectrum is also very well known, since it is dictated by the particularly well studied $`p+p`$ nuclear fusion reaction. Among these proposed experiments, HELLAZ will be able to determine the incoming neutrino energy in an event-by-event basis and have the unique opportunity of studying the solar neutrino spectrum and the recoil electron kinetic energy spectrum separately. Similar to what was shown for <sup>7</sup>Be neutrinos , the authors of showed that HELLAZ may be able to measure a nonzero component of $`\nu _{\mu ,\tau }`$ in the solar $`pp`$-neutrino flux by analysing the recoil electron kinetic energy spectrum independent of the SSM prediction for the solar neutrino flux.
In this paper we extend the analysis done in , and study the flavour composition of the flux of $`pp`$ and <sup>7</sup>Be neutrinos using the recoil electron kinetic energy spectrum. In particular we will address the capability of future low-energy solar-neutrino experiments to see evidence for $`\nu _{\mu ,\tau }`$ coming from the Sun, and, in light of such evidence, exclude more “exotic” oscillation scenarios, such as $`\nu _e\nu _s`$ or $`\nu _e\overline{\nu }_{\mathrm{any}}`$ oscillations.
Our presentation is organised as follows: Sec. 2 describes the flavour dependent recoil kinetic energy distribution of events at Borexino and HELLAZ. Sec. 3 presents the technique for determining the presence of $`\nu _{\mu ,\tau }`$ coming from the Sun, independent of the SSM prediction for the neutrino flux. We present simulations for both Borexino and HELLAZ and show how such a determination can be improved once we take the SSM prediction for the neutrino flux into account. Sec. 4 describes how the same procedure can be used to exclude the presence of antineutrinos or sterile neutrinos in the solar neutrino flux. In Sec. 5, we conclude.
## 2 Recoil Electron Kinetic Energy Spectrum
In this section, we discuss the differences in the recoil electron kinetic energy spectra among different neutrino species. Low-energy solar neutrinos are detected via “$`\nu `$$`+e^{}`$$`\nu `$$`+e^{}`$ elastic scattering in the experiments which will be considered here. By “$`\nu `$” in the previous sentence, one actually means any of $`\nu _e`$, $`\nu _\mu `$, $`\nu _\tau `$, $`\overline{\nu }_e`$, $`\overline{\nu }_\mu `$, or $`\overline{\nu }_\tau `$. Because $`\nu _\mu `$ and $`\nu _\tau `$ are indistinguishable as far as the reaction above is concerned, we will refer to both as $`\nu _\mu `$.
The kinetic energy distribution of the recoil electrons, for a given incoming neutrino energy $`E_\nu `$ is very well known and given by
$$\frac{\mathrm{d}\sigma (T,E_\nu )}{\mathrm{d}T}=\frac{2G_F^2m_e}{\pi }\left[A^2+B^2\left(1\frac{T}{E_\nu }\right)^2AB\frac{m_eT}{(E_\nu )^2}\right],$$
(2.1)
where $`m_e`$ is the electron mass, $`T`$ is the kinetic energy of the recoil electron, and $`G_F`$ is the Fermi constant. The parameters $`A`$ and $`B`$ are given in Table 1. The sign difference in the term $`1/2`$ is a consequence of the presence (absence) of $`W`$-boson exchange and the interchange of $`A`$ and $`B`$ between neutrino and anti-neutrino cases is a consequence of the “handedness” of the weak interactions. Eq. (2.1) is a tree-level expression, but higher order corrections are known to be very small , especially for the neutrino energies of interest, and will be neglected throughout.
Borexino (under construction) is an ultra-pure liquid scintillator tank which detects the scintillating light produced by the recoil electron absorbed by the medium. For more details see . It is sensitive to recoil electron kinetic energies greater than 250 keV, and is therefore sensitive to the (almost) monochromatic <sup>7</sup>Be neutrinos with $`E_\nu =862`$ keV. The expected resolution for the kinetic energy measurement varies from roughly 12%, for $`T=T_{\mathrm{min}}=0.25`$ MeV, to 7% for $`T=T_{\mathrm{max}}=0.66`$ MeV . They expect 53 events/day in the SSM (BP95) together with 19 background events/day with the anticipated radiopurity of the scintillator of $`10^{16}`$g/g for U/Th, $`10^{18}`$g/g for <sup>40</sup>K, and <sup>14</sup>C$`/^{12}`$C$`=10^{18}`$ and no radon diffusion. It is remarkable, however, that the München group of Borexino achieved a radiopurity for an organic liquid (Phenyl-ortho-xylylethane) better than $`1.0\times 10^{17}`$g/g ; this is an upper bound on the contamination, limited by the sensitivity of the neutron activation measurement and hence the actual radiopurity may be even better. In this paper, we ignore the background to the <sup>7</sup>Be solar neutrino signal at Borexino. This is probably an overoptimistic assumption, but could be realised in future upgrades given the above-mentioned achievement.
HELLAZ (proposed) is a large time projection chamber (TPC) filled with roughly 2000 m<sup>3</sup> of cool helium gas ($`6`$ tons at 5 atmos, 77 K), which serves as the target for $`\nu `$-$`e`$ scattering. The recoil electron propagates in the gas medium before being absorbed, leaving a track of ionization electrons. These are then collected, yielding information about the kinetic energy and the flight direction of the recoil electron. HELLAZ is sensitive to recoil kinetic energies greater than $`50`$ keV, and can therefore “see” most of the $`pp`$-neutrino spectrum. Most importantly, since not only the recoil kinetic energy of scattered electrons is measured but also their direction, it is possible to reconstruct the incoming neutrino energy, given that the position of the Sun in the sky is known, via the simple kinematic relation
$$T=m_e\frac{2\mathrm{cos}^2\theta }{(1+m_e/E_\nu )^2\mathrm{cos}^2\theta },$$
(2.2)
where $`\theta `$ is the recoil electron scattering angle with respect to the incoming neutrino direction in the laboratory frame. Incidently, from Eq. (2.2) it is very easy to compute the maximum value of the recoil electron kinetic energy, $`T_{\mathrm{max}}=T(\theta =0)=E_\nu /(1+m_e/(2E_\nu ))`$. HELLAZ expects to measure the recoil electron kinetic energy with a resolution which varies roughly from 2% to 4% and the incoming neutrino energy with a resolution which varies between 5% and 12% . They expect around 7 events/day from $`pp`$ neutrinos in the SSM with negligible background. The major sources of background at HELLAZ are radioactive impurities from <sup>232</sup>Th and <sup>238</sup>U in the structure of the TPC. However, because of the detector’s total event reconstruction capabilities (including directional information), very efficient background rejection schemes are possible (see and references therein for further information).
The issue we would like to concentrate on is whether the shapes of the recoil electron kinetic energy distributions for different (anti)neutrino species are statistically different at Borexino and HELLAZ. With this in mind, Fig. 1 depicts the normalised distribution of events at HELLAZ (left) and Borexino (right).<sup>*</sup><sup>*</sup>*In addition to $`pp`$-neutrinos, HELLAZ is also sensitive to <sup>7</sup>Be neutrinos, as well as the $`pep`$-neutrinos and the neutrinos coming from the CNO-cycle. <sup>7</sup>Be neutrinos can be clearly separated from $`pp`$-neutrinos, while the number of $`pep`$ and CNO-cycle neutrino generated events is expected to be less than 10% that of $`pp`$-neutrinos. Borexino is sensitive to, in addition to <sup>7</sup>Be neutrinos (with $`E_\nu =862`$ keV), a fraction of the $`pep`$ and the CNO-cycle neutrinos, which produce approximately 10% as many events as <sup>7</sup>Be neutrinos. We assume throughout, for simplicity, that only $`pp`$ (<sup>7</sup>Be) neutrinos are detected at HELLAZ (Borexino). In the case of Borexino, the data is binned into ten kinetic energy bins, between 250 keV and 650 keV. In the case of HELLAZ, the data is binned into $`4\times 21`$ bins in $`E_\nu \times T`$. The bins have a width of 50 keV in the $`E_\nu `$ direction and central values of 245, 295, 345 and 395 keV, while in the $`T`$ direction they have a width of 10 keV in the range from 50 to 260 keV. The bin sizes have been chosen such that they are roughly the same as the resolution of both detectors. In order to integrate over the incoming neutrino energy at HELLAZ, the (normalised) BP98 $`pp`$-spectrum presented at was used.
Many important features of the recoil electron kinetic distributions are worthwhile to point out. First of all, it is quite clear that the spectrum produced by $`\overline{\nu }_e`$-$`e`$ is much steeper than all the other ones.Some of these features were pointed out in . Second, the $`\nu _e`$ and $`\nu _\mu `$ generated spectra have opposite slopes when the neutrino energy is small enough, while their shapes start to look more and more similar as the neutrino energy increases. Finally, the spectra produced by $`\nu _\mu `$-$`e`$ and $`\overline{\nu }_\mu `$-$`e`$ scattering are extremely similar, especially at very low energies.
All of these features can be readily understood from Eq. (2.1). First, it is convenient to write the expression for the normalised recoil electron kinetic energy distributions,
$$\frac{\mathrm{d}\overline{\sigma }}{\mathrm{d}y}=N[\frac{A}{B}+\frac{B}{A}(1y)^2\frac{m_e}{E_\nu }y,],$$
(2.3)
where $`y=T/E_\nu `$, and $`N`$ is a normalisation constant, such that $`\frac{\mathrm{d}\overline{\sigma }}{\mathrm{d}T}`$ from $`T_{\mathrm{min}}`$ to $`T_{\mathrm{max}}`$ equals unity.
For $`\nu _e`$, $`\frac{A}{B}3`$, while $`\frac{B}{A}1/3`$. For $`\overline{\nu }_e`$ the situation is reversed, while for both $`\nu _\mu `$ and $`\overline{\nu }_\mu `$, $`\frac{A}{B}\frac{B}{A}1`$.In this case, the normalisation constant $`N`$ is negative. Note, curiously enough, that the reason for the similarity between the $`\nu _\mu `$ and $`\overline{\nu }_\mu `$ cases is simply due to the accidental fact that $`\mathrm{sin}^2\theta _W`$ is close to $`1/4`$.<sup>§</sup><sup>§</sup>§The fact that there is a sign difference between $`g_L`$ and $`g_R`$ for muon-type (anti)neutrinos is irrelevant, since these coefficients either appear as squares or as the product $`g_Lg_R`$. This similarity is even more pronounced at very low energies, when the $`\frac{m_e}{E_\nu }`$ term dominates over the $`\frac{A}{B}`$ and $`\frac{B}{A}`$ terms.
Keeping in mind that $`0.59\begin{array}{c}<\hfill \\ \hfill \end{array}\frac{m_e}{E_\nu }\begin{array}{c}<\hfill \\ \hfill \end{array}2.3`$ in the neutrino energy range of interest and $`0<y(1+m_e/2E_\nu )^1<1`$, one may write approximate expressions
$`\left({\displaystyle \frac{\mathrm{d}\overline{\sigma }}{\mathrm{d}y}}\right)_{\nu _e}3{\displaystyle \frac{m_e}{E_\nu }}y,`$ (2.4)
$`\left({\displaystyle \frac{\mathrm{d}\overline{\sigma }}{\mathrm{d}y}}\right)_{\overline{\nu }_e}3(1y)^2{\displaystyle \frac{m_e}{E_\nu }}y,`$ (2.5)
$`\left({\displaystyle \frac{\mathrm{d}\overline{\sigma }}{\mathrm{d}y}}\right)_{\overline{\nu }_\mu }\left({\displaystyle \frac{\mathrm{d}\overline{\sigma }}{\mathrm{d}y}}\right)_{\nu _\mu }1+(1y)^2+{\displaystyle \frac{m_e}{E_\nu }}y.`$ (2.6)
In the limit $`\frac{m_e}{E_\nu }1`$, all three distributions are quite different (see, e.g., Fig. 1(B) in ). The $`\nu _e`$ case is roughly flat, the $`\overline{\nu }_e`$ case ranges from 3 at $`y=0`$ to 0 at $`y=1`$ and the $`\nu _\mu ,\overline{\nu }_\mu `$ case ranges from $`3/2`$ at $`y=0`$ to $`3/4`$ at $`y=1`$.
For $`\frac{m_e}{E_\nu }\begin{array}{c}>\hfill \\ \hfill \end{array}1`$, things are slightly more complicated, but still easy to understand. For example, the slope of the distributions for small values of $`y`$ are, up to normalisation factors, $`\frac{m_e}{E_\nu }`$, $`\left(6+\frac{m_e}{E_\nu }\right)`$ and $`+\left(\frac{m_e}{E_\nu }2\right)`$ for $`\nu _e`$, $`\overline{\nu }_e`$ and $`\nu _\mu ,\overline{\nu }_\mu `$ respectively. It is then easy to note that the $`\overline{\nu }_e`$ slope is significantly more negative than the other two, and that, in the case of $`\nu _\mu ,\overline{\nu }_\mu `$ the slope is actually positive if $`E_\nu `$ is small enough. This is indeed what one observes in Fig. 1.
As the incoming neutrino energy increases, the distributions generated by $`\nu _e`$ and $`\nu _\mu ,\overline{\nu }_\mu `$ look more and more similar. One hint of this behaviour is that the slope of the $`\nu _e`$ case increases (decreases in absolute value), while the slope of the $`\nu _\mu ,\overline{\nu }_\mu `$ decreases. One can easily estimate that for $`0.9`$ MeV$`\begin{array}{c}<\hfill \\ \hfill \end{array}E_\nu \begin{array}{c}<\hfill \\ \hfill \end{array}1.0`$ MeV the shapes of the $`\nu _e`$ and $`\nu _\mu `$ induced recoil kinetic energy distributions are most similar. Indeed, for <sup>7</sup>Be neutrino energies, one can already note that the difference between the $`\nu _e`$ and the $`\nu _\mu `$ cases is similar to the difference between the $`\nu _\mu `$ and the $`\overline{\nu }_\mu `$ cases.
## 3 Measuring a $`\nu _{\mu ,\tau }`$ Component in the Solar Neutrino Flux
In this section, we address the question whether the shapes of the recoil electron kinetic energy distributions presented in Sec. 2 are statistically different at Borexino or HELLAZ. In the affirmative case, there is hope that one may be sensitive to a “contamination” of other neutrino types in the solar neutrino flux by analysing the shape of the recoil kinetic energy spectrum. We consider this an “appearance experiment” of the “wrong” types of neutrinos from the Sun. In this section we will only consider the case of $`\nu _e\nu _\mu `$ oscillations.
In the advent of neutrino oscillations, a mixture of different neutrino weak eigenstates reaches the Earth. Given an electron-type neutrino survival probability $`P_{ee}`$, a fraction $`P_{ee}`$ of all the neutrinos arriving at the detector are $`\nu _e`$, while a fraction $`1P_{ee}`$ are $`\nu _\mu `$. The recoil electron kinetic energy distribution will, therefore, be given by
$$\frac{\mathrm{d}\sigma (T,E_\nu )}{\mathrm{d}T}=P_{ee}\times \left(\frac{\mathrm{d}\sigma (T,E_\nu )}{\mathrm{d}T}\right)_{\nu _e}+(1P_{ee})\times \left(\frac{\mathrm{d}\sigma (T,E_\nu )}{\mathrm{d}T}\right)_{\nu _\mu }.$$
(3.7)
Note that, in general, $`P_{ee}`$ is a function of the neutrino oscillation parameters (the mass-squared differences of the neutrino mass eigenstates and the neutrino mixing angles) and the neutrino energy.
We simulate “data” at Borexino and HELLAZ for different values of $`P_{ee}`$. We use the distributions presented in Sec. 2, while the flux of $`pp`$ and <sup>7</sup>Be neutrinos are taken from the SSM . In the case of Borexino, the energy dependence of $`P_{ee}`$ is irrelevant, given the monochromatic nature of <sup>7</sup>Be neutrinos. In the case of HELLAZ, we assume that $`P_{ee}`$ is constant inside each individual neutrino energy bin. Following the central idea presented in , we perform a $`\chi ^2`$ fit to the “data” using a linear combination of $`\nu _e`$-$`e`$ scattering and $`\nu _\mu `$-$`e`$ scattering with arbitrary coefficients,
$$C_e\times \left(\frac{\mathrm{d}\sigma (T,E_\nu )}{\mathrm{d}T}\right)_{\nu _e}+C_\mu \times \left(\frac{\mathrm{d}\sigma (T,E_\nu )}{\mathrm{d}T}\right)_{\nu _\mu },$$
(3.8)
i.e., we perform a two parameter ($`C_e`$ and $`C_\mu `$) fit to the data. This measurement procedure is independent of the SSM prediction for the neutrino flux. Therefore, if a nonzero coefficient of the $`\nu _\mu `$-$`e`$ scattering distribution is measured, one can claim to have detected evidence for neutrinos other than $`\nu _e`$ coming from the Sun. This “appearance” result certainly qualifies as a smoking gun signature for neutrino oscillations.
Fig. 2 (long, thin error bars) depicts the measured value of $`1P_{ee}=\frac{C_\mu }{C_e+C_\mu }`$ in each of the neutrino energy bins defined in Sec. 2 as a function of the input value of $`P_{ee}`$, for 5 years of simulated HELLAZ data. As was mentioned in the previous paragraph, the relevant information one should obtain from the plot is if the measured value of $`1P_{ee}C_\mu `$ is statistically different from zero.
Fig. 3 (right, long, thin error bars) depicts the measured value of $`1P_{ee}`$ as a function of the input value of $`P_{ee}`$, for two years of Borexino running. This is just a repetition of Fig. 2(A) in .<sup>*</sup><sup>*</sup>*In , a different variable, $`P1P_{ee}`$, was used. Both results are, of course, equivalent. Fig. 3 (left, long, thin error bars) depicts the result obtained at HELLAZ if all energy bins are used in the “data” analysis. This result is only meaningful if $`P_{ee}`$ is roughly constant for neutrino energies ranging from from 220 keV to 420 keV. This happens to be the case for most of the currently preferred regions of the two-neutrino oscillation parameter space, especially LMA, LOW and VAC solutions (see, e.g., ).For the SMA solution, there is a sharp drop in $`P_{ee}`$ at $`E_\nu 0.4`$ MeV, and the three lower bins can be combined without any problem. At HELLAZ, this will show up in the data, as the $`E_\nu `$ spectrum differs from the expected $`pp`$-neutrino spectrum, and hence is not a concern. Clearly, the significance of the measurement is better than the one obtained for individual energy bins (Fig. 2).
Next, the same analysis as above is repeated, except that the SSM prediction for the solar neutrino flux is included in the $`\chi ^2`$ analysis. An uncertainty of 20% (5%) was assumed for the <sup>7</sup>Be ($`pp`$) neutrino flux. The theoretical error was considered Gaussian for simplicity.This procedure follows the one used in . The readers are referred to this article for details. Note that the uncertainties assumed here are inflated with respect to the ones quoted in the SSM calculations (9% and 1%, respectively), in order to render the results very conservative. Since the rates are very high both at Borexino and HELLAZ, the error bars are dominated by the uncertainties in the fluxes and hence they can be approximately scaled according to the assigned flux uncertainties.
The results are presented in Figs. 2 and 3 (short, thick error bars). The significance of the measured value of $`1P_{ee}`$ improves significantly, especially at HELLAZ, because of the small assigned uncertainty on the $`pp`$-neutrino flux. After five years of HELLAZ running, for example, one should be be able to determine a 1-sigma-away-from-zero $`\nu _\mu `$ component in the $`pp`$-neutrino flux even for $`P_{ee}0.9`$. It is also noteworthy that in the case of the SMA MSW solution to the solar neutrino puzzle $`P_{ee}0`$ for <sup>7</sup>Be neutrinos, in which case a 4-sigma-away-from-zero evidence for $`\nu _\mu `$ in the solar neutrino flux can be established in only two years of Borexino running! It is important to emphasise that using SSM predictions for the solar neutrino flux is a reasonable thing to do, especially for $`pp`$-neutrinos. As mentioned before, the flux of $`pp`$-neutrinos is very well known because it is tightly related to the flux of light coming from the Sun. It is, therefore, the neutrino flux which is least sensitive to detailed modelling of the Sun’s innards.
Some comments are in order. First, only statistical uncertainties were considered, and there are no background events in our “data.” As discussed in Sec. 2, the assumption of a negligible background rate seems less than realistic at Borexino, but may be possible in future upgrades. It may, however, be a fair assumption in the case of HELLAZ. If the real experimental data contains a sizable number of background events, it is necessary to either subtract the background in a bin-by-bin basis or to somehow model the recoil kinetic energy distribution produced by background events. Analysing either of these procedures, however, is beyond the scope of this paper.
Second, the analysis which does not include the SSM flux predictions is completely model-independent (the only assumption being the electron recoil spectrum as predicted by the standard electroweak theory), while the one which includes the SSM flux predictions is model-dependent. Obviously, one obtains a much better determination of $`P_{ee}`$ with the additional input of the SSM flux predictions. For establishing the “wrong neutrino component” in the solar neutrino flux as a smoking gun signature of the solar neutrino oscillations, the former approach is desired. However, for the purpose of determining the oscillation parameters, the energy dependence of the survival probability, and excluding other neutrino oscillation modes, such as $`\nu _e\nu _s`$ or $`\nu _e\overline{\nu }_{e,\mu ,\tau }`$ (as will be discussed in Sec. 4), it is reasonable to include the SSM predictions in the analysis.
Finally, we point out that the results we obtained for HELLAZ are similar to the ones obtained by J. Séguinot et al . Indeed, we chose neutrino energy bins at HELLAZ which coincide with the ones used in . They also perform two different analyses of their simulated data, one which is independent of the SSM prediction for the solar neutrino flux, and one which assumes the SSM prediction for the flux. However, their data analysis procedure is somewhat different, and they do not take the theoretical uncertainty of the solar neutrino flux prediction into account.
## 4 Testing for the $`\nu _e\nu _s`$ or $`\overline{\nu }_{e,\mu ,\tau }`$ Hypotheses
Although it is most natural to assume that electron-type neutrinos oscillate into some linear combination of muon-type and tau-type neutrinos, there is a logical possibility that electron-type neutrinos might oscillate into standard model singlet sterile neutrinos , or, perhaps, into antineutrinos of all flavours<sup>*</sup><sup>*</sup>*The original neutrino oscillation paper by Bruno Pontecorvo did, after all, consider $`\nu _e\overline{\nu }_e`$! (see and references therein). In this section, we will address the issue of excluding these solar neutrino oscillation modes if the data collected at Borexino and HELLAZ are consistent with $`\nu _e\nu _\mu `$ oscillations.
One can already address these “exotic” oscillation modes with the current experimental data. The flux of electron-type anti-neutrinos from the Sun is particularly constrained by the SuperKamiokande and the LSD experiments : the 95% CL SuperKamiokande upper bound on the flux of $`\overline{\nu }_e`$ from the Sun with energies $`\begin{array}{c}>\hfill \\ \hfill \end{array}6.5`$ MeV is $`\mathrm{\Phi }_{\overline{\nu }_e}<1.8\times 10^5`$ cm<sup>-2</sup> s<sup>-1</sup>, or $`\mathrm{\Phi }_{\overline{\nu }_e}/\mathrm{\Phi }_{SSM}^{{}_{}{}^{8}\mathrm{B}}=0.035`$, where $`\mathrm{\Phi }_{SSM}^{{}_{}{}^{8}\mathrm{B}}`$ is the SSM prediction for the <sup>8</sup>B neutrino flux. KamLAND, being a dedicated detector for $`\overline{\nu }_e`$, will improve this bound further down to 0.1% of the SSM flux above reactor anti-neutrino energies ($`E_\nu \begin{array}{c}>\hfill \\ \hfill \end{array}8`$ MeV) after one year of running . There are, however, scenarios in which, for energies below the SuperKamiokande threshold, the $`\nu _e\overline{\nu }_e`$ mixing is quite large ( and references therein). Such a possibility can only be addressed by low-energy solar neutrino experiments.
Below, we discuss the exclusion of electron-type neutrino oscillations into sterile neutrinos or into one of the antineutrino types separately at Borexino and HELLAZ experiments.
### 4.1 $`\nu _e\nu _s`$
The $`\nu _e\nu _s`$ oscillation mode is allowed by the analysis of current solar neutrino data , even though in the case of the MSW solutions to the solar neutrino puzzle, only the equivalent of the SMA MSW solution exists at the 99% confidence level . It is curious to note that, in the case of atmospheric neutrinos, the $`\nu _\mu \nu _s`$ hypothesis is currently somewhat disfavoured .
In the case of $`\nu _e\nu _s`$ oscillations, one expects the recoil electron kinetic energy spectrum to be exactly the same as the one generated by $`\nu _e`$-$`e`$ scattering, since $`\nu _s`$ do not interact with electrons. The only effect of the neutrino oscillations would be to suppress the expected number of events, i.e., the hypothesis of $`\nu _e\nu _s`$ oscillations is identical to assuming that the solar neutrino flux is, somehow, suppressed. Therefore, we attempt to fit the “data” simulated according to Eq. (3.7) in Sec. 3 (remember that the “data” is consistent with $`\nu _e\nu _\mu `$ oscillations) to the trial function Eq. (3.8), where the piece which corresponds to $`C_\mu `$ vanishes identically. This is a one parameter $`\chi ^2`$ fit to $`C_e`$. Note that the only discrimination against $`\nu _s`$ is the recoil energy spectrum, because the rate can be always fitted with the free parameter $`C_e`$. The inclusion of the SSM flux prediction does not help in excluding the $`\nu _e\nu _s`$ oscillation because the free parameter $`C_e`$ makes the predicted flux irrelevant.One can “discover” $`\nu _s`$ by observing a nearly vanishing rate for <sup>7</sup>Be neutrinos. In the case of $`\nu _e\nu _{\mu ,\tau }`$ oscillations, neutral-current scattering guarantees at least (when $`P_{ee}=0`$) 21% of the SSM rate. For this purpose, one should rely on the SSM flux prediction. Note that <sup>7</sup>Be neutrinos are predicted to have almost completely oscillated to $`\nu _s`$ for the (only available) SMA solution: $`P_{ee}=0.009_{0.005}^{+0.244}`$ .
Fig. 4 depicts the value of $`\chi ^2`$ obtained when one attempts to fit the “data” to the sterile neutrino hypothesis, for 5 years of HELLAZ running (left) and 2 years of Borexino running (right). In the case of HELLAZ, we have assumed that $`P_{ee}`$ is constant over the entire $`pp`$-neutrino energy range. The value of $`\chi ^2`$ is determined using the philosophy employed in , and should be compared to $`N_{\mathrm{bins}}1`$ ($`N_{\mathrm{bins}}`$ is the number of “data” bins). After 5 years of HELLAZ running, one should be able to exclude sterile neutrinos coming form the Sun at more than 99.9% confidence level (CL) if all electron-type neutrino have turned into muon-type neutrinos ($`P_{ee}=0`$). After 2 years of Borexino running, the sterile neutrino hypothesis is only ruled out, at best, at the 89% CL.The situation does improve, of course, it more events are collected at Borexino. After 5 years of Borexino running, for example, one can exclude sterile neutrinos for $`P_{ee}\begin{array}{c}<\hfill \\ \hfill \end{array}0.1`$ at more than 95% CL. The explanation for this is the fact that the recoil electron kinetic energy spectra are very different when one compares the $`\nu _e`$-$`e`$ and the $`\nu _\mu `$-$`e`$ scattering cases at very low energies, i.e., $`pp`$-neutrinos, and similar at $`O`$(MeV) energies, i.e., <sup>7</sup>Be neutrinos, as discussed in Sec. 2. The exclusion CL decreases with increasing $`P_{ee}`$, and sterile neutrinos are excluded after 5 years of HELLAZ running only at the 77% CL for $`P_{ee}=0.4`$.
### 4.2 $`\nu _e\overline{\nu }`$, Model-independent Fit
In the case of $`\nu _e\overline{\nu }_{e,\mu }`$ oscillations, we perform a two parameter fit to the “data” simulated as in Sec. 3 to a linear combination of the $`\nu _e`$-$`e`$ and $`\overline{\nu }_{e,\mu }`$-$`e`$ scattering recoil kinetic energy distributions. Fig. 5 depicts the value of $`\chi ^2`$ obtained when such a fit is performed, for 5 years of Borexino and HELLAZ running. The value of $`\chi ^2`$ is to be compared to $`N_{\mathrm{bins}}2`$ to determine exclusion confidence levels. As advertised in Sec. 2, the $`\nu _\mu `$-$`e`$ and $`\overline{\nu }_\mu `$-$`e`$ scattering cases produce almost identical recoil kinetic energy spectra, and are almost undistiguishable at HELLAZ. At Borexino, however, the difference between $`\nu _\mu `$-$`e`$ and $`\overline{\nu }_\mu `$-$`e`$ scattering is similar to the difference between the $`\nu _\mu `$-$`e`$ and $`\nu _e`$-$`e`$ cases, as mentioned in Sec. 2 (see Fig. 1), and some discrimination seems possible. Furthermore, upon close inspection, one should note that the shape of the distribution produced due to $`\nu _e`$-$`e`$ scattering is more similar to the $`\nu _\mu `$-$`e`$ case than the $`\overline{\nu }_\mu `$-$`e`$. Therefore, any $`\overline{\nu }_\mu `$ component in the trial function makes the value of $`\chi ^2`$ larger, i.e., the minimum of $`\chi ^2`$ is obtained when the coefficient of the $`\overline{\nu }_\mu `$ component is zero.<sup>§</sup><sup>§</sup>§We only allow nonnegative coefficients of the distribution functions in the fits, for obvious reasons. This is exactly what happens in the case of $`\nu _e\overline{\nu }_e`$ oscillations, at both experiments. Any $`\overline{\nu }_e`$ component in the flux makes the agreement between the theoretical function and the “data” worse, and again the best value of $`\chi ^2`$ is obtained when the coefficient of the $`\overline{\nu }_e`$-$`e`$ scattering distribution is zero.
One can see from Fig. 5 that, after 5 years of HELLAZ data, $`\overline{\nu }_e`$ coming from the Sun can be ruled out at more than 95% CL if $`P_{ee}\begin{array}{c}<\hfill \\ \hfill \end{array}0.2`$, while $`\nu _e\overline{\nu }_\mu `$ oscillations are not constrained at all, even for $`P_{ee}=0`$. After 5 years of Borexino data, both $`\nu _e\overline{\nu }_\mu `$ and $`\nu _e\overline{\nu }_e`$ oscillations are ruled out at more than 95% CL if $`P_{ee}\begin{array}{c}<\hfill \\ \hfill \end{array}0.1`$.
Even if the $`\nu _e\overline{\nu }`$ hypothesis cannot be ruled out at some reasonable CL, one may still be able to place upper limits on the flux of anti-neutrinos coming from the Sun. In the case of $`\nu _e\overline{\nu }_e`$ oscillations, it is straight-forward to place upper bounds on the flux of electron-type antineutrinos at both HELLAZ and Borexino. The 95% CL upper bounds on the $`\overline{\nu }_e`$ flux are depicted in Fig. 6. Of course, for $`P_{ee}\begin{array}{c}<\hfill \\ \hfill \end{array}0.2`$ (0.1) at HELLAZ (Borexino) the upper bound on the flux is meaningless, since the hypothesis of $`\overline{\nu }_e`$ is already ruled out at more than 95% CL. Note that the upper bounds on the antineutrino fluxes are normalised by the SSM prediction for the $`pp`$-neutrino flux $`\mathrm{\Phi }_{SSM}^{pp}=5.94\times 10^{10}`$ cm<sup>-2</sup>s<sup>-1</sup> for the HELLAZ result, and the SSM prediction for the <sup>7</sup>Be neutrino flux $`\mathrm{\Phi }_{SSM}^{{}_{}{}^{7}\mathrm{Be}}=4.8\times 10^9`$ cm<sup>-2</sup>s<sup>-1</sup> for the Borexino result. For comparison, the 95% CL SuperKamiokande upper bound is 3.5% of the SSM flux, while KamLAND will improve it to 0.1% after one year of running. However, both of them are only for the <sup>8</sup>B neutrinos. Both the HELLAZ and (especially) the Borexino limits obtained from 5 years of “data” are competitive with the SuperKamiokande limit for lower energy neutrinos.
In the case of $`\nu _e\overline{\nu }_\mu `$ oscillations the situation is more ambiguous, especially at HELLAZ.The same is true at Borexino if one assumes the SSM prediction of the total neutrino flux, as will be described later. Not only are the minimum values of $`\chi ^2`$ very small, but in some cases (especially for small values of $`P_{ee}`$) a zero $`\overline{\nu }_\mu `$ flux is ruled out at more than 95% CL. In such cases, it seems that the reasonable thing to do is to measure the antineutrino flux, not determine upper limits! The only exception to this is the case $`P_{ee}=1`$, when the data looks exactly like the SSM prediction, without neutrino oscillations. Indeed, one can not only set upper limits on the antineutrino fluxes, but should also set limits to the $`\nu _\mu `$ flux. Such limits are presented in Table 2.
It is worthwhile to comment that the information contained in Figs. 5, 6, and in Table 2 is also valid for the case of any unknown source of solar antineutrinos of the electron and the muon-types, not only neutrino oscillations. This is because our “data” was analysed assuming that the total flux of solar neutrinos is unknown. We emphasise that $`P_{ee}`$ is the survival probability of electron neutrinos assuming that they oscillate into active neutrinos, i.e., $`\nu _e\nu _\mu `$ oscillations.
### 4.3 $`\nu _e\overline{\nu }`$, SSM-dependent Fit
Next, the same analysis can be repeated assuming that the solar neutrino flux is known within theoretical errors. Again, the value of $`\chi ^2`$ is computedThis procedure follows the one used in . The readers are referred to this article for details. and compared with $`N_{\mathrm{bins}}2+1`$ (the $`2`$ corresponds to the two coefficients that are varied during the minimisation procedure and the $`+1`$ corresponds to the solar neutrino flux constraint). Fig. 7 depicts the minimised values of $`\chi ^2`$ obtained with 5 years of HELLAZ (left) and Borexino (right) “data.” The theoretical uncertainty on the $`pp`$ (<sup>7</sup>Be) neutrino flux was taken to be 2% (20%); we inflated the theoretical errors by roughly a factor of two from those in BP98.
A comparison between Figs. 5 and 7 reveals that the exclusion confidence levels increase, sometimes significantly. For example, after 5 years of HELLAZ one can exclude $`\nu _e\overline{\nu }_e`$ oscillations for virtually all values of $`P_{ee}`$ at more than 99.9% CL. This is mostly because the $`\overline{\nu }_e`$ has a total cross section which is significantly larger than $`\nu _\mu `$, and the oscillation $`\nu _e\overline{\nu }_e`$ cannot account for the large suppression in the event rate in the “data” (due to $`\nu _e\nu _\mu `$). Even the elusive $`\nu _e\overline{\nu }_\mu `$ case can be excluded at Borexino at more than 95% CL for $`P_{ee}\begin{array}{c}<\hfill \\ \hfill \end{array}0.2`$. Note that at HELLAZ the ability to discriminate between $`\nu _\mu `$ and $`\overline{\nu }_\mu `$ is still quite limited. It is worthwhile to comment that, unlike in the case of model-independent fits in Sec. 4.2, the minimum value of $`\chi ^2`$ is in general obtained for a nonzero coefficient of the $`\overline{\nu }`$-$`e`$ scattering distribution. The reason for this is that, even though the shape of the $`\overline{\nu }`$-$`e`$ scattering recoil electron kinetic energy distribution is “more wrong,” the contribution to the overall cross section is smaller than the $`\nu _e`$-$`e`$ scattering case, and therefore one obtains values of the solar neutrino flux which are closer to the theoretical ones by having a finite $`\overline{\nu }`$ component, decreasing the value of $`\chi ^2`$.
Again, one may set upper limits on the antineutrino flux. As before, there is some ambiguity with regard to setting upper limits for the $`\overline{\nu }_\mu `$ flux, because for almost all values of $`P_{ee}1`$ at both experiments a zero flux is excluded at more than 95% CL. On the other hand, the $`\nu _e\overline{\nu }_e`$ oscillation hypothesis is almost completely ruled out by HELLAZ and the upper limits obtained at Borexino are not much better than the ones depicted in Fig. 6. For this reason, the equivalent of Fig. 6 in the case at hand is not presented.
Table 3 contains the obtained upper limits on the (anti)neutrino fluxes when $`P_{ee}=1`$, i.e., when the data agrees with the predictions of the SSM. Unlike the case of a free total flux analysis, the results presented in Table 3 assume that the total neutrino flux of neutrinos to be detected at HELLAZ and Borexino is the one predicted by the SSM, i.e., there is no “room” for other, yet unknown, low-energy solar neutrino sources. For this reason, of course, the bounds obtained are (in some cases) much more stringent.
Finally, as argued before, we emphasise that fixing the value of the solar neutrino flux to its SSM value is a reasonable thing to do, especially for $`pp`$-neutrinos. In these “exclusion analyses” such a procedure is even more natural, especially if one keeps in mind that a theoretical hypothesis, i.e., $`\nu _e\nu _\mu `$ oscillations plus the SSM computed values for the solar neutrino flux, has been “confirmed experimentally.”
## 5 Conclusions
In order to unambiguously solve the solar neutrino puzzle, and to establish the oscillations of solar neutrinos (if they occur), clear “smoking gun” signatures are required. Such signatures include a large day-night effect, anomalous seasonal variations, or an obvious distortion of the neutrino energy spectrum. Another unambiguous signature is a discrepancy between the number of charged current and neutral current events at SNO, which can be viewed as an “appearance” experiment of $`\nu _{\mu ,\tau }`$. However, SNO can look for this “appearance” signature only for <sup>8</sup>B neutrinos with $`E_\nu \begin{array}{c}>\hfill \\ \hfill \end{array}6.5`$ MeV and hence similar studies for lower energy neutrinos such as <sup>7</sup>Be and $`pp`$ neutrinos, which are less sensitive to details of the solar model, are important.
We have argued in this paper that a careful analysis of the recoil kinetic energy spectrum at Borexino and HELLAZ serves as another “smoking gun” signature, in the sense that one may be able to infer, independent of the SSM prediction for the solar neutrino flux, the existence of $`\nu _{\mu ,\tau }`$ coming from the Sun. It is worthwhile to emphasise that this is different from distortions in the incoming neutrino energy spectrum. In our case we are describing an “appearance” experiment, while the analysis of the neutrino energy spectrum is a (energy dependent) “disappearance” experiment.
It is important to point out that, in our simple simulations, no background events were included. While this is probably an oversimplification in the case of Borexino, it may well be a good approximation for HELLAZ. Moreover future upgrades of Borexino may reduce the background further according to recent encouraging progress . One should keep in mind that, even if the background rates are significant, the procedure we described may still be useful if the background can be successfully dealt with (one should not underestimate the ability and creativity of experimental physicists!).
We have also included in the analysis the SSM prediction of the flux of solar neutrinos. While the results obtained in this manner are model-dependent (they are not “smoking gun” signatures of neutrino oscillations), we found them very useful. This is a reasonable thing to do especially for $`pp`$-neutrinos, whose flux is constrained well by the solar luminosity. This additional input makes the measurement of the oscillation probability more precise.
Finally, we have argued that, if the data collected at Borexino and HELLAZ is consistent with $`\nu _e\nu _{\mu ,\tau }`$ oscillations, one can try to exclude other neutrino oscillation modes ($`\nu _e\nu _s`$ and $`\nu _e\overline{\nu }_{e,\mu ,\tau }`$) using the same procedure or, at least, to set upper limits on the flux of solar antineutrinos. Again we considered the possibility of constraining the solar neutrino flux to the SSM predicted value. The main result we obtained is that $`\nu _e\overline{\nu }_e`$ oscillations can, in general, be excluded, while the $`\nu _e\overline{\nu }_{\mu ,\tau }`$ case is much more elusive. Nonetheless, Borexino should be able to exclude $`\nu _e\overline{\nu }_{\mu ,\tau }`$ oscillations if the SMA MSW solution to the solar neutrino puzzle happens to be the correct one.
## Acknowledgements
AdG thanks Hiroshi Nunokawa for questioning whether or not the issue of antineutrinos in the solar flux could be addressed by this procedure during a seminar at the Instituto de Física Gleb Wataghin, UNICAMP, Brazil. The work of HM was supported in part by the Director, Office of Science, Office of High Energy and Nuclear Physics, Division of High Energy Physics of the U.S. Department of Energy under Contract DE-AC03-76SF00098 and in part by the National Science Foundation under grant PHY-95-14797 and also by the Alfred P. Sloan Foundation.
|
warning/0003/hep-ph0003144.html
|
ar5iv
|
text
|
# Axial vector current in an electromagnetic field and low-energy neutrino-photon interactions
## I Introduction
The study of neutrino-photon interactions began many years ago when Pontecorvo and Chiu and Morrison suggested that the process $`\gamma \gamma \nu \overline{\nu }`$ may be important in the analysis of stellar cooling. These interactions are of interest in astrophysics and cosmology.
It is well known that, in the standard model, neutrino-photon interactions appear at the one-loop level. In addition to the heavy gauge bosons which can immediately be integrated out at the desired scales leading to the effective four-fermion interaction, charged fermions (electrons) run in the loop; in particular, the coupling of the photons to these fermions is responsible for the $`\gamma \nu `$ interactions.
A variety of processes of physical relevance in different situations belong to this class of $`\gamma \nu `$ interactions, for which an effective-action description involving only neutrino currents and field strength tensors has partly been found. E.g., Dicus and Repko have recently derived an effective action for interactions between two neutrinos and three soft photons, which immediately allows for the calculation of all scattering amplitudes of this type. This effective action can be derived via the expectation value of the electromagnetic vector current $`j_\mu =e\overline{\mathrm{\Psi }}(x)\gamma _\mu \mathrm{\Psi }(x)`$; the latter can in turn be calculated very easily from the Heisenberg-Euler effective Lagrangian $`_{\text{HE}}`$ with the aid of the formula $`\overline{\mathrm{\Psi }}(x)\gamma _\mu \mathrm{\Psi }(x)=\delta _{\text{HE}}/e\delta A^\mu (x)`$. In , the effective action of Dicus and Repko was reproduced by a more direct diagrammatic approach.
Incidentally, processes with two photons, for example, neutrino-photon scattering $`\gamma \nu \gamma \nu `$, turn out to be highly suppressed in the standard model because, according to Yang’s theorem , two photons cannot couple to the $`J=1`$ state. As a result, typical cross sections are exceedingly small and suppressed by factors of $`1/m_W^2`$ (see, e.g., ).
The presence of a medium or external electromagnetic field drastically changes the situation. It induces an effective coupling between photons and neutrinos, even enhancing processes such as $`\nu \nu \gamma `$. Based on the same effective action of Dicus and Repko , it was shown in that in the presence of an external magnetic field, cross sections for neutrino-photon processes such as $`\gamma \gamma \nu \overline{\nu }`$ and $`\nu \gamma \nu \gamma `$ are amplified by the factor $`\left(m_W/m_e\right)^4\left(B/B_\text{c}\right)^2`$ for soft photon frequencies $`\omega m_e`$ and a weak magnetic field $`BB_\text{c}m^2/e`$ (for extensions to this result, see and ). The subject of the electromagnetic properties of neutrinos in media was comprehensively studied in (see also for neutrinos in magnetized media).
In general, an effective action describing low-energy neutrino-photon processes with an arbitrary number of photons and in the presence of strong external electromagnetic fields appears highly desirable. In other words, we are looking for the analogue of the Heisenberg-Euler effective action which turned out to be extremely useful in QED (see e.g., ).
For this, we start from the effective four-fermion interaction Lagrangian valid at energies very much smaller than the W- and Z-boson masses:
$$_{4f}=\frac{G_F}{\sqrt{2}}\overline{\nu }\gamma ^\mu \left(1+\gamma _5\right)\nu \overline{E}\gamma _\mu \left(g_V+g_A\gamma _5\right)E.$$
(1)
Here, $`E`$ denotes the electron field, $`\gamma _5=i\gamma ^0\gamma ^1\gamma ^2\gamma ^3`$, $`g_V=1\frac{1}{2}\left(14\mathrm{sin}^2\theta _W\right)`$ and $`g_A=1\frac{1}{2}`$ for $`\nu _e`$, where the first terms in $`g_V`$ and $`g_A`$ are the contributions from the W exchange diagram and the second one from the Z exchange diagram. Also, $`g_V=2\mathrm{sin}^2\theta _W\frac{1}{2}`$ and $`g_A=\frac{1}{2}`$ for $`\nu _{\mu ,\tau }`$.
Now concentrating on soft electromagnetic degrees of freedom with energies much smaller than the electron mass, we may integrate out the actual “heaviest” particle, i.e., the electron, and find
$$_{\text{eff}}=\frac{G_F}{\sqrt{2}}\frac{1}{e}\overline{\nu }\gamma ^\mu \left(1+\gamma _5\right)\nu \left(g_\text{V}j^\mu ^A+g_\text{A}j_5^\mu ^A\right),$$
(2)
where the expectation values of the currents are given in terms of the Green’s function in this field: e.g., $`j_5^\mu ^A=\text{i}e\text{tr}\left[\gamma ^\mu \gamma _5G(x,x|A)\right]`$.
Obviously, in order to obtain this effective Lagrangian, one must calculate the expectation values of vector and axial vector currents in a slowly varying electromagnetic field background. As mentioned above, the vector current expectation value can be easily obtained using the well-known Heisenberg-Euler Lagrangian $`_{\text{HE}}`$: $`\overline{E}(x)\gamma _\mu E(x)=\delta _{\text{HE}}/e\delta A^\mu (x)`$. In this way, one obtains a derivative expansion of the vector current around a strong field. For example, the term which is third order in the field and first order in derivatives was used by Dicus and Repko in their study of $`\nu \gamma \nu \gamma \gamma `$ and cross processes.
With regard to Eq. (2), the derivative expansion of the axial vector current around an arbitrarily strong field is finally required. To our surprise, we could not find such an expression in the vast literature on derivative expansions. Therefore, its derivation remains the final task of our present work.
It is convenient to choose a special gauge condition for the potential $`A_\mu (x)`$ without loss of generality: the so-called Fock–Schwinger gauge<sup>*</sup><sup>*</sup>*Especially for gradient expansions of the heat kernel or the Heisenberg-Euler Lagrangian, this gauge has proved extremely useful; see, e.g., .
$$(x^\mu y^\mu )A_\mu (x)=0,$$
(3)
which is satisfied by the series
$$A_\mu (x)=\frac{1}{2}(x^\lambda y^\lambda )F_{\lambda \mu }(y)+\frac{1}{3}(x^\lambda y^\lambda )(x^\sigma y^\sigma )_\sigma F_{\lambda \mu }(y)+\mathrm{}$$
(4)
Abbreviating the first term on the right-hand side with $`A_{\text{c}\mu }`$ and the second term with $`a_\mu `$, we may perform a perturbation expansion for the Green’s function with respect to the derivative term $`a_\mu `$:
$$G(x,x^{}|A)=G_c(x,x^{}|A_\text{c})+d^4wG_c(x,w|A_\text{c})ea_\mu (w)\gamma ^\mu G_c(w,x^{}|A_\text{c})+\mathrm{},$$
(5)
where $`G_c(x,x^{}|A_\text{c})`$ is the Green’s function of the electron in the constant field produced by $`A_\text{c}`$. Inserting the expansion (5) into the definition of $`j_5^\mu ^A`$, we obtain, to first order in derivatives,
$$j_5^\mu ^A=\frac{1}{3}_\sigma F_{\alpha \beta }\frac{^2}{k^\sigma k^\alpha }\left[\mathrm{\Pi }_5^{\beta \mu }(k)\right]|_{k=0},$$
(6)
where we encounter the vector–axial-vector amplitude (VA) $`\mathrm{\Pi }_5^{\beta \mu }`$, i.e., the axial analogue of the polarization tensor. This VA amplitude has been calculated very recently by one of the authors in ; an independent confirmation can be found in .
Inserting the presentation of into Eq. (6), we finally arrive at the first-order gradient expansion of the axial vector current for arbitrarily strong electromagnetic fields valid for slowly varying fields:
$`j_5^\mu ^A`$ $`=`$ $`\text{i}e\text{tr}\left[\gamma ^\mu \gamma _5G(x,x|A)\right]`$ (7)
$`=`$ $`\text{i}{\displaystyle \frac{e^2}{24\pi ^2}}_\sigma F_{\lambda \nu }{\displaystyle _0^{\mathrm{}}}{\displaystyle \frac{ds}{s}}e^{ism^2}{\displaystyle \frac{(eas)(ebs)}{\mathrm{sin}(ebs)\mathrm{sinh}(eas)}}`$ (10)
$`\{[C^{\nu \lambda }\left(C^2\right)^{\mu \sigma }+C^{\nu \sigma }\left(C^2\right)^{\mu \lambda }]N_1+[B^{\nu \lambda }\left(B^2\right)^{\mu \sigma }+B^{\nu \sigma }\left(B^2\right)^{\mu \lambda }]N_2`$
$`[3F^{\nu \mu }g^{\lambda \sigma }+F^{\nu \lambda }g^{\mu \sigma }+F^{\nu \sigma }g^{\mu \lambda }]N_3\},`$
where the $`N_i`$ are simple functions of the field strength:
$`N_1={\displaystyle \frac{2\mathrm{sin}(ebs)}{\mathrm{sinh}^2(eas)}}\left[\mathrm{cosh}(eas){\displaystyle \frac{\mathrm{sinh}(eas)}{eas}}\right]+2ebsN_3`$ (11)
$`N_2={\displaystyle \frac{2\mathrm{sinh}(eas)}{\mathrm{sin}^2(ebs)}}\left[\mathrm{cos}(ebs)+{\displaystyle \frac{\mathrm{sin}(ebs)}{ebs}}\right]+2easN_3`$ (12)
$`N_3={\displaystyle \frac{1}{easebs(a^2+b^2)}}\left[a^2{\displaystyle \frac{\mathrm{sin}(ebs)}{\mathrm{sinh}(eas)}}b^2{\displaystyle \frac{\mathrm{sinh}(eas)}{\mathrm{sin}(ebs)}}\right].`$ (13)
For the analysis, we employed a decomposition of the field strength tensor in linearly independent subspaces of electric and magnetic parts:
$$C_{\mu \nu }:=\frac{1}{a^2+b^2}\left(aF_{\mu \nu }+bF_{\mu \nu }^{}\right);B_{\mu \nu }:=\frac{1}{a^2+b^2}\left(bF_{\mu \nu }aF_{\mu \nu }^{}\right),$$
(14)
where the invariants $`a`$ and $`b`$ are defined by
$$a,b=\sqrt{(^2+𝒢^2)^{\frac{1}{2}}\pm };=\frac{1}{4}F_{\mu \nu }F^{\mu \nu },𝒢=\frac{1}{4}F_{\mu \nu }^{}F^{\mu \nu }.$$
(15)
As a consequence, we findWe always use the metric $`g=\mathrm{diag}(+)`$.
$$\left(C^2\right)_{\mu \nu }=\frac{1}{a^2+b^2}\left(F_{\mu \nu }^2+b^2g_{\mu \nu }\right);\left(B^2\right)_{\mu \nu }=\frac{1}{a^2+b^2}\left(F_{\mu \nu }^2a^2g_{\mu \nu }\right).$$
(16)
Concerning our central result for the axial vector current in Eq. (10), we observe its linearity in the derivative of $`F_{\mu \nu }`$ (by construction) which is multiplied by a finite proper-time integral containing the complete dependence on the strength of the fields. We would like to point out that our method of performing the derivative expansion in momentum space (cf. Eq. (6)) is much in the spirit of Ref. with the essential difference that we expand around an arbitrary value of field strength.
Our representation of $`j_5^\mu `$ in terms of the $`C_{\mu \nu }`$ and $`B_{\mu \nu }`$ tensors is very convenient for performing a weak-field expansion; to second order in $`1/m^2`$, it finally reads:
$`j_5^\mu ^A`$ $`=`$ $`{\displaystyle \frac{e^3}{24\pi ^2m^2}}\left(^\mu 𝒢+(^\alpha F_{\alpha \beta })F^{\beta \mu }\right)`$ (19)
$`+{\displaystyle \frac{e^5}{90\pi ^2m^6}}_\sigma F_{\alpha \beta }[𝒢(F^{\beta \alpha }g^{\mu \sigma }+F^{\beta \sigma }g^{\mu \alpha })+(F^{\beta \alpha }(F^2)^{\mu \sigma }+F^{\beta \sigma }(F^2)^{\mu \alpha })`$
$`(3F^{\mu \beta }g^{\alpha \sigma }+F^{\beta \alpha }g^{\mu \sigma }+F^{\beta \sigma }g^{\mu \alpha })].`$
As a cross-check, one can indeed prove that the first term can immediately be obtained from the order $`1/m^2`$ term of the famous triangle graph. As a side remark, it should be mentioned that the axial vector anomaly is not present in our result, since we computed the $`1/m^2`$-expansion of $`j_5^\mu `$, whereas the anomaly is mass independent.
Returning to the quest for the general effective action, we can easily insert Eq. (10) into Eq. (2) and a similar expression for the vector current $`j_\mu `$. The well-known latter expression can be derived via the Heisenberg-Euler Lagrangian as mentioned above or, alternatively, via the same method as here proposed for the axial vector current; for this, we would like to stress that Eq. (6) also holds for the vector current with the VA amplitude $`\mathrm{\Pi }_5^{\mu \nu }`$ replaced by the polarization tensor $`\mathrm{\Pi }^{\mu \nu }`$. This tensor in the presence of an arbitrarily constant electromagnetic field was first calculated in (an appropriate representation can be found in ). Details will be presented elsewhere. We observe that the axial vector current is of even order in the field, while the vector current is of odd order; this is a direct consequence of the Dirac algebra (Furry’s theorem).
To conclude, we have completed the search for a low-energy effective action for neutrino interactions with an arbitrary number of soft photons with wavelengths larger than the Compton wavelength. From a different perspective, we have found the generalization of the Heisenberg-Euler action to the case involving an axial vector coupling.
## Acknowledgments
We would like to thank Professor W. Dittrich for helpful discussions and for carefully reading the manuscript. This work was supported by Deutsche Forschungsgemeinschaft under grant DFG Di 200/5-1.
|
warning/0003/quant-ph0003018.html
|
ar5iv
|
text
|
# The Modal Interpretation of Algebraic Quantum Field Theory
## 1 The Modal Interpretation of Nonrelativistic Quantum Theory
The modal interpretation is actually a family of interpretations sharing the feature that a system’s density operator constrains the possibilities for assigning definite values to its observables . We start with a brief review of the basic idea behind these interpretations from an algebraic point of view .
Consider a ‘universe’ comprised of a quantum system $`U`$, with finitely many degrees of freedom, represented by a Hilbert space H$`{}_{U}{}^{}=_i𝖧_i`$. At any given time, $`U`$ will occupy a pure vector state $`x𝖧_U`$ that determines a reduced density operator $`D_S`$ on the Hilbert space H$`{}_{S}{}^{}=_{iS}𝖧_i`$ of any subsystem $`SU`$. Let $`(𝖧_S)`$ denote the algebra of all bounded operators on H<sub>S</sub>, and for any single operator or family of operators $`T(𝖧_S)`$, let $`T^{}`$ denote its commutant (i.e., all operators on $`𝖧_S`$ that commute with those in $`T`$). Let $`P_S`$ denote the projection onto the range of $`D_S`$, and consider the subalgebra of $`(𝖧_S)`$ given by the direct sum
$$_SP_S^{}(𝖧_S)P_S^{}+D_S^{\prime \prime }P_S.$$
(1)
When $`S`$ is not entangled with its environment $`\overline{S}`$ (represented by H$`{}_{\overline{S}}{}^{}=_{iS}`$H<sub>i</sub>), $`D_S`$ will itself be a pure state, induced by a unit vector $`y𝖧_S`$. In this case, $`_S`$ consists of all operators with $`y`$ as an eigenvector — the self-adjoint members of which are taken, in orthodox quantum theory, to be the observables of $`S`$ with definite values. On the other hand, when there is entanglement and $`D_S`$ is mixed — in particular, in the extreme case $`P_S=I`$ (where essentially every vector state $`y`$ H<sub>S</sub> is a component of the mixture) — then $`_S=D_S^{\prime \prime }`$, and $`_S`$ consists simply of all functions of $`D_S`$. In this case, orthodox quantum theory has nothing to say about the properties of $`S`$. Thus, when $`S`$ is Schrödinger’s cat entangled with some potentially cat killing device $`\overline{S}`$, we get the infamous measurement problem.
In non-atomic versions of the modal interpretation (, ), there is no preferred partition of the universe into subsystems. Any particular subsystem $`SU`$ is taken to have definite values for all the self-adjoint operators that lie in $`_S`$, and this applies whether or not $`D_S`$ is pure. The values of these observables are taken to be distributed according to the usual Born rule. Thus, the expectation of any observable $`A_S`$ is $`\mathrm{T}r(D_SA)`$, and the probability that $`A`$ possesses some particular value $`a_j`$ is $`\mathrm{T}r(D_SP^j)`$, where $`P^j_S`$ is the corresponding eigenprojection of $`A`$. As in orthodox quantum theory, which precise value for $`A`$ occurs on a given occasion (from amongst those with nonzero probability in state $`D_S`$) is not fixed by the interpretation. However, the occurrence of a value does not require that it be ‘measured’. And, unlike orthodox quantum theory, no miraculous collapse is needed to solve the measurement problem. Instead, after a typical unitary ‘measurement’ interaction between two parts of the universe — $`O`$ the ‘measured’ system, and $`A`$ the apparatus — decoherence induced by $`A`$’s coupling to the environment $`\overline{OA}`$ will force the density operator $`D_A`$ of the apparatus to diagonalize in a basis extremely close to one which diagonalizes the pointer observable of $`A`$ . Thus, after the measurement, the definite-valued observables in $`_A`$ will be such that the pointer points!
In atomic versions of the modal interpretation (, ), one does not tell a separate story about the definite-valued observables for each subsystem $`SU`$. Rather, $`S`$ is taken to inherit properties from those of its atomic components, represented by the individual Hilbert spaces in $`_{iS}𝖧_i`$. In the approach favoured by Dieks (cf. ), each atomic system $`i`$ possesses definite values for all the observables in $`_i`$, as determined by the corresponding atomic density operator $`D_i`$ in accordance with (1). The definite-valued observables of $`S`$ are then built up by embedding each $`_i`$ in $`(𝖧_S)`$ (via tensoring it with the identity on H$`_{\overline{S}}`$), and taking the von Neumann subalgebra of $`(𝖧_S)`$ generated by all these embeddings. The definite properties of $`S`$ will therefore include all projections $`_{iS}P_i^{j(i)}`$ that are tensor products of the spectral projections of the individual atomic density operators, and their joint probabilities are again taken to be given by the usual Born rule $`\mathrm{T}r[D_S(_{iS}P_i^{j(i)})]`$. In the presence of decoherence, the expectation is that atomic versions of the modal interpretation can yield essentially the same resolution of the measurement problem as do non-atomic versions .
In both versions of the modal interpretation, the observables with definite values must change over time as a function of the (generally, non-unitary) evolution of the reduced density operators of the systems involved. In principle, this evolution can be determined from the (unitary) Schrödinger evolution of the universal state vector $`x`$ H<sub>U</sub> . But the evolution of the precise values of the definite-valued observables themselves is not determined by the Schrödinger equation. Various more or less natural proposals have been made for ‘completing’ the modal interpretation with a dynamics for values (,). Unfortunately, it has been shown that the most natural proposals for a dynamics, particularly in the case of atomic modal interpretations, must break Lorentz-invariance . Most of Dieks’ recent Letter is concerned to address this dynamics problem by appropriating ideas from the decoherent histories approach to quantum theory (cf. ). However, we shall focus here entirely on the viability of Dieks’ new proposal for picking out definite-valued observables of a relativistic quantum field that are associated with approximately point-sized regions of Minkowski spacetime (, Sec. 5).
## 2 Critique of Dieks’ Proposal
Dieks’ stated aim is to see if the modal interpretation can achieve sensible results in the context of quantum field theory. For this purpose, he adopts the formalism of algebraic quantum field theory because of its generality . In the concrete ‘Haag-Araki’ approach, one supposes that a quantum field on Minkowski spacetime $`M`$ will associate to each bounded open region $`OM`$ a von Neumann algebra $`(O)`$ of observables measurable in that region, where the collection $`\{(O):OM\}`$ acts irreducibly on some fixed Hilbert space H. It is then natural to treat each open region $`O`$ and associated algebra $`(O)`$ as a quantum system in its own right. Given any (normal) state $`\rho `$ of the field (where $`\rho `$ is a state functional on $`(𝖧)`$), we can then ask which observables in $`(O)`$ are picked out as definite-valued by the restriction, $`\rho _O`$, of the state $`\rho `$ to $`(O)`$.
The difficulty for the modal interpretation (though Dieks himself does not put it this way) is that when $`O`$ has nonempty spacelike complement $`O^{}`$, $`(O)`$ will typically be a type III factor that contains no nonzero finite projections (, Sec. V.6; , Sec. 17.2). Because of this, $`(O)`$ cannot contain compact operators, like density operators, all of whose (non-null) spectral projections are finite-dimensional. As a result, there is no density operator in $`(O)`$ that can represent $`\rho _O`$. Moreover, if we try to apply the standard modal prescription based on Eq. (1) to a density operator in $`(𝖧)`$ that agrees with $`\rho _O`$, there is no guarantee that the resulting set of observables will pick out a subalgebra of $`(O)`$, and we will be left with nothing to say about which observables have definite values in $`O`$. The moral Dieks draws from this is that “We can therefore not take the open spacetime regions and their algebras as fundamental, if we want an interpretation in terms of (more or less) localized systems whose properties would specify an event” (, p. 322). We shall see in the next section that this conclusion is overly pessimistic.
In any case, Dieks’ strategy for dealing with the problem is to exploit the fact that, in most models of the axioms of algebraic quantum field theory, the local algebras associated with diamond shaped spacetime regions (i.e., regions given by interior of the intersection of the causal future and past of two spacetime points) have the split property. The property is that for any two concentric diamond shaped regions $`\mathrm{}_r`$ $`,\mathrm{}_{r+ϵ}`$ $``$ $`M`$, with radii $`r`$ and $`r+ϵ`$, there is a type I ‘interpolating’ factor $`𝒩_{r+ϵ}`$ such that $`(\mathrm{}_r)𝒩_{r+ϵ}(\mathrm{}_{r+ϵ})`$. Now since $`𝒩_{r+ϵ}`$ $`(\overline{𝖧})`$ for some Hilbert space $`\overline{𝖧}`$, we know there is always a unique density operator $`D_{r+ϵ}(\overline{𝖧})`$ that agrees with $`\rho `$on $`𝒩_{r+ϵ}`$, and therefore with $`\rho _\mathrm{}_r`$. The proposal is, then, to take both $`r`$ and $`ϵ`$ to be fixed small numbers and apply the prescription in Eq. (1) to this density operator $`D_{r+ϵ}`$, yielding a definite-valued subalgebra $`_{r+ϵ}𝒩_{r+ϵ}`$. This, according to Dieks, should give an approximation indication of which observables have definite values at the common origin of the two diamonds in “the classical limiting situation in which classical field and particle concepts become approximately applicable” (, p. 323). Thus, Dieks proposes to build up an atomic modal interpretation of the field as follows. (i) We subdivide (approximately) the whole of spacetime $`M`$ into a collection of non-overlapping diamond regions $`\mathrm{}_r`$, with some fixed small radius $`r`$. (ii) We choose some fixed small $`ϵ`$ and an interpolating factor $`𝒩_{r+ϵ}`$ for each diamond, using its $`\rho `$-induced density operator $`D_{r+ϵ}`$ and Eq. (1) to determine the definite-valued observables $`_{r+ϵ}`$ $`(\mathrm{}_{r+ϵ})`$ to be loosely associated with the origin. (iii) Finally, we build up definite-valued observables associated with collections of diamonds, and define their joint probabilities in the usual way via Born’s rule (defining transition probabilities between values of observables associated with timelike-separated diamonds using the familiar multi-time generalization of that rule employed in the decoherent histories approach).
As things stand, there is much arbitrariness in this proposal that enters into the stages (i) and (ii). Dieks himself recognizes the arbitrariness in the size of the partition of $`M`$ chosen. He also acknowledges that this arbitrariness cannot be eliminated by passing to the limit $`r,ϵ0`$, because the intersection of the algebras associated with any collection of concentric diamonds is always the trivial algebra $`I`$ . Indeed, one would have thought that this undermines any attempt to formulate an atomic modal interpretation in this context, because it forces the choice of ‘atomic diamonds’ in the partition of $`M`$ to be essentially arbitrary. Dieks appears to suggest that this arbitrariness will become unimportant in some classical limit of relativistic quantum field theory in which we should recover “the classical picture according to which field values are attached to spacetime points” (, p. 325). But it is not sufficient for the success of a proposal for interpreting a relativistic quantum field theory that the interpretation give sensible results in the limit of classical relativistic (or nonrelativistic) field theory. Indeed, the only relevant limit would appear to be the nonrelativistic limit; i.e., Galilean quantum field theory. But, there, one still needs to spatially smear “operator-valued” fields at each point to obtain a well-defined algebra of observables in a spatial region , so there will again be no natural choice to make for atomic spatial regions or algebras.
There is also another, more troubling, degree of arbitrariness at step (ii) in the choice of the type I interpolating factor $`𝒩_{r+ϵ}`$ about each origin point. For any fixed partition and fixed $`r,ϵ>0`$, we can always sub-divide the interval $`(r,r+ϵ)`$ further, and then the split property implies the existence of a pair of interpolating type I factors satisfying
$$(\mathrm{}_r)𝒩_{r+ϵ/2}(\mathrm{}_{r+ϵ/2})𝒩_{r+ϵ}(\mathrm{}_{r+ϵ}).$$
(2)
The problem is that we now face a nontrivial choice deciding which of these factors’ $`\rho `$-induced density operators to use to pick out the definite-valued observables in the state $`\rho `$ associated with the origin. If we pick $`D_{r+ϵ}𝒩_{r+ϵ}`$, then by (1) all observables $`A𝒩_{r+ϵ}`$ that share the same spectral projections as $`D_{r+ϵ}`$ will have definite values. However, no such $`A`$ can lie in $`_{r+ϵ/2}𝒩_{r+ϵ/2}`$, nor even in $`(\mathrm{}_{r+ϵ/2})`$. The reason is that $`A`$’s spectral projections are finite in $`𝒩_{r+ϵ}`$. So if those projections were also in the type III algebra $`(\mathrm{}_{r+ϵ/2})`$, they would have to be infinite in $`(\mathrm{}_{r+ϵ/2})`$, and therefore also infinite projections in $`𝒩_{r+ϵ}`$ — which is impossible.
Clearly we can sub-divide the interval $`(r,r+ϵ)`$ arbitrarily many times in this way and obtain a monotonically decreasing sequence of type I factors satisfying $`(\mathrm{}_{r+ϵ/2^{n+1}})𝒩_{r+ϵ/2^n}(\mathrm{}_{r+ϵ/2^n})`$ that all interpolate between $`(\mathrm{}_r)`$ and $`(\mathrm{}_{r+ϵ})`$. The sequence $`\{𝒩_{r+ϵ/2^n}\}_{n=0}^{\mathrm{}}`$ has no least member, and its greatest member, $`𝒩_{r+ϵ}`$, is arbitrary, because we could also further sub-divide the interval $`(r+ϵ/2,r+ϵ)`$ ad infinitum. Thus there is no natural choice of interpolating factor for picking out the observables definite at the origin, even if we restrict ourselves to a ‘nice’ decreasing sequence of interpolating factors of the form $`\{𝒩_{r+ϵ/2^n}\}_{n=0}^{\mathrm{}}`$.
On the other hand, it can actually be shown that $`(\mathrm{}_r)=_{n=0}^{\mathrm{}}𝒩_{r+ϵ/2^n}`$ (, pp. 12-3; , p. 426). Furthermore, suppose $`\{𝒩_{r+ϵ_n}\}_{n=0}^{\mathrm{}}`$ is any other decreasing type I sequence satisfying
$$(\mathrm{}_{r+ϵ_{n+1}})𝒩_{r+ϵ_n}(\mathrm{}_{r+ϵ_n}),ϵ_0=ϵ,ϵ_n>ϵ_{n+1},limϵ_n=0.$$
(3)
Then, since for any $`n`$ there will be a sufficiently large $`n^{}`$ such that $`𝒩_{r+ϵ_n^{}}𝒩_{r+ϵ/2^n}`$ (and vice-versa), clearly $`(\mathrm{}_r)=_{n=0}^{\mathrm{}}𝒩_{r+ϵ_n}`$, and this intersection will also be independent of $`ϵ`$. It would seem, then, that the natural way to avoid choosing between the myriad type I factors that interpolate between $`(\mathrm{}_r)`$ and $`(\mathrm{}_{r+ϵ})`$ is to take the observables definite-valued at the origin to be those in the intersection $`_{n=0}^{\mathrm{}}_{r+ϵ_n}(\mathrm{}_r)`$ (where, as before, $`_{r+ϵ_n}`$ is the modal subalgebra of $`𝒩_{r+ϵ_n}`$ determined via (1) by the density operator in $`𝒩_{r+ϵ_n}`$ that represents $`\rho `$). Indeed, Dieks’ suggestion appears to be that when we take successively smaller values for $`ϵ`$ (holding $`r`$ fixed), and choose a type I interpolating factor at each stage, we should be getting progressively better approximations to the set of observables that are truly definite at the origin. What better candidate for that set can there be than an intersection like $`_{n=0}^{\mathrm{}}_{r+ϵ_n}`$?
Unfortunately, we have no guarantee that this intersection, unlike $`_{n=0}^{\mathrm{}}𝒩_{r+ϵ_n}`$ itself, is independent of the particular sequence $`\{ϵ_n\}`$ or its starting value $`ϵ_0=ϵ`$. The reason any intersection of form $`_{n=0}^{\mathrm{}}𝒩_{r+ϵ_n}`$ is so independent is because $`𝒩_{r+ϵ_n}𝒩_{r+ϵ_{n+1}}`$ for all $`n`$. But this does not imply $`_{r+ϵ_n}_{r+ϵ_{n+1}}`$. To take just a trivial example: when $`\rho `$ is a pure state of $`𝒩_{r+ϵ_n}`$ that induces a mixed state on the proper subalgebra $`𝒩_{r+ϵ_{n+1}}`$of $`𝒩_{r+ϵ_n}`$, $`_{r+ϵ_{n+1}}`$ will contain observables with dispersion in the state $`\rho `$, but $`_{r+ϵ_n}`$ will not.
We conclude that there is little prospect of eliminating the arbitrariness in Dieks’ proposal and making it well-defined. One ought to look for another, intrinsic way to pick out the definite-valued observables in $`(\mathrm{}_r)`$ that does not depend on special assumptions such as the split property.
## 3 The Modal Interpretation for Arbitrary von Neumann Algebras
There are two salient features of the algebra $`_S`$ in Eq. (1) that make it an attractive set of definite-valued observables to modal interpreters. First, $`_S`$ is locally determined by the quantum state $`D_S`$ of system $`S`$ together with the structure of its algebra of observables. In particular, there is no need to add any additional structure to the standard formalism of quantum theory to pick out $`S`$’s properties. Second, the restriction of the state $`D_S`$ to the subalgebra $`_S`$ is a mixture of dispersion-free states (given by the density operators one obtains by renormalizing the (non-null) spectral projections of $`D_S`$). This second feature is what makes it possible to think of the observables in $`_S`$ as possessing definite values distributed in accordance with standard Born rule statistics . Let us see, then, whether we can generalize these two features to come up with a proposal for the definite-valued observables of a system described by an arbitrary von Neumann algebra $``$ (acting on some Hilbert space H) in an arbitrary state $`\rho `$ of $``$.
Generally, a state $`\rho `$ of $``$ will be a mixture of dispersion-free states on a subalgebra $`𝒮`$ just in case there is a probability measure $`\mu _\rho `$ on the space $`\mathrm{\Lambda }`$ of dispersion-free states of $`𝒮`$ such that
$$\rho (A)=_\mathrm{\Lambda }\omega _\lambda (A)𝑑\mu _\rho (\lambda ),\text{for all }A𝒮\text{,}$$
(4)
where $`\omega _\lambda (A^2)=\omega _\lambda (A)^2`$ for all self-adjoint elements $`A𝒮`$. This somewhat cumbersome condition turns out to be equivalent (, Prop. 2.2(ii)) to simply requiring that
$$\rho ([A,B]^{}[A,B])=0\text{for all }A,B𝒮.$$
(5)
In particular, $`\rho `$ can always be represented as a mixture of dispersion-free states on any abelian subalgebra $`𝒮`$. Conversely, if $`\rho `$ is a faithful state of $``$, i.e., $`\rho `$ maps no nonzero positive elements of $``$ to zero, then the only subalgebras that allow $`\rho `$ to be represented as a mixture of dispersion-free states are the abelian ones. There is now an easy way to pick out a subalgebra $`𝒮`$ with this property, using only $`\rho `$ and the algebraic operations available within $``$.
Consider the following two mathematical objects explicitly defined in terms of $``$ and $`\rho `$. First, the support projection of the state $`\rho `$ in $``$, defined by
$$P_{\rho ,}\{P=P^2=P^{}:\rho (P)=1\},$$
(6)
which is simply the smallest projection in $``$ that the state $`\rho `$ ‘makes true’. Second, there is the centralizer subalgebra of the state $`\rho `$ in $``$, defined by
$$𝒞_{\rho ,}\{A:\rho ([A,B])=0\text{for all }B\}.$$
(7)
For any von Neumann algebra $`𝒦`$, let $`𝒵`$($`𝒦`$)$`𝒦𝒦^{}`$, the center algebra of $`𝒦`$. Then it is reasonable for the modal interpreter to take as definite-valued all the observables that lie in the direct sum
$$𝒮=_{\rho ,}P_{\rho ,}^{}P_{\rho ,}^{}+𝒵(𝒞_{\rho ,})P_{\rho ,},$$
(8)
where the algebra in the first summand acts on the subspace $`P_{\rho ,}^{}𝖧`$ and that of the second acts on $`P_{\rho ,}𝖧.`$ The state $`\rho `$ is a mixture of dispersion-free states on $`_{\rho ,}`$, by (5), because $`\rho `$ maps all elements of the form $`P_{\rho ,}^{}P_{\rho ,}^{}`$ to zero, and the product of the commutators of any two elements of $`𝒵(𝒞_{\rho ,})P_{\rho ,}`$ also gets mapped to zero, for the trivial reason that $`𝒵(𝒞_{\rho ,})`$ is abelian.
The set $`_{\rho ,}`$ directly generalizes the algebra of Eq. (1) to the non-type I case where the algebra of observables of the system does not contain a density operator representative of the state $`\rho `$. Assuming the type I case, $``$ $``$ $`(\overline{𝖧})`$ for some Hilbert space $`\overline{𝖧}`$, $`\rho `$ is given by a density operator $`D`$ on $`\overline{𝖧}`$, $`P_{\rho ,}`$ is equivalent to the range projection of $`D`$, and $`𝒵(𝒞_{\rho ,})𝒵(𝒞_{D,(\overline{𝖧})})`$. So to show that $`_{\rho ,}`$ is isomorphic to the algebra of Eq. (1), it suffices to establish that $`𝒵(𝒞_{D,(\overline{𝖧})})=`$ $`D^{\prime \prime }`$. It is easy to see that $`𝒞_{D,(\overline{𝖧})}=D^{}`$ (invoking cyclicity and positive-definiteness of the trace), thus $`𝒵(𝒞_{D,(\overline{𝖧})})=`$ $`D^{}D^{\prime \prime }`$. However, since $`D^{}`$ always contains a maximal abelian subalgebra of $`(\overline{𝖧})`$ (viz., that generated by the projections onto any complete orthonormal basis of eigenvectors for $`D`$), we always have $`D^{\prime \prime }D^{}`$.
Choosing $`_{\rho ,}`$ is certainly not the only way to pick a subalgebra $`𝒮`$ that is definable in terms of $`\rho `$ and $``$ and allows $`\rho `$ to be represented as a mixture of dispersion-free states. There is the obvious orthodox alternative one can always consider, viz., the definite algebra of $`\rho `$ in $``$,
$$𝒮=𝒪_{\rho ,}\{A:\rho (AB)=\rho (A)\rho (B)\text{for all }B\},$$
(9)
which coincides with the complex span of all self-adjoint members of $``$ on which $`\rho `$ is dispersion-free (, p. 2445). Note, however, that we always have $`𝒪_{\rho ,}_{\rho ,}`$. Indeed the problem is that the orthodox choice $`𝒪_{\rho ,}`$ generally will contain far too few definite-valued observables to solve the measurement problem. For example, when $`\rho `$ is faithful — and there will always be a norm dense set of states of $``$ that are — we get just $`𝒪_{\rho ,}=I`$. Thus it is natural for a modal interpreter to require that the choice of $`𝒮`$ be maximal. In the case where $`\rho `$ is faithful, we now show that this singles out the choice $`𝒮=_{\rho ,}=𝒵(𝒞_{\rho ,})`$ uniquely (and we conjecture that a similar uniqueness result holds for the more general expression for $`_{\rho ,}`$ in Eq. (8), using the fact that an arbitrary state $`\rho `$ always renormalizes to a faithful state on $`P_{\rho ,}P_{\rho ,}`$).
Proposition 1 Let $``$ be a von Neumann algebra and $`\rho `$ a faithful normal state of $``$ with centralizer $`𝒞_{\rho ,}`$. Then $`𝒵(𝒞_{\rho ,})`$, the center of $`𝒞_{\rho ,}`$, is the unique subalgebra $`𝒮`$ such that:
1. The restriction of $`\rho `$ to $`𝒮`$ is a mixture of dispersion-free states.
2. $`𝒮`$ is definable solely in terms of $`\rho `$ and the algebraic structure of $``$.
3. $`𝒮`$ is maximal with respect to properties 1. and 2.
Proof: By 3., it suffices to show than any $`𝒮`$ satisfying 1. and 2. is contained in $`𝒵(𝒞_{\rho ,})`$. And for this, it suffices (because von Neumann algebras are generated by their projections) to show that $`𝒮`$, the subset of projections in $`𝒮`$, is contained in $`𝒵(𝒞_{\rho ,})`$. Recall also that, as a consequence of 1. and the faithfulness of $`\rho `$, $`𝒮`$ must be abelian. And in virtue of 2., any automorphism $`\sigma :`$ $``$ that preserves the state $`\rho `$ in the sense that $`\rho \sigma =\rho `$, must leave the set $`𝒮`$ (not necessarily pointwise) invariant, i.e., $`\sigma (𝒮)=𝒮`$.
$`𝒮`$ $`𝒞_{\rho ,}^{}`$. Any unitary operator $`U𝒞_{\rho ,}`$ defines an inner automorphism on $``$ that leaves $`\rho `$ invariant, therefore $`U𝒮U^1=`$ $`𝒮`$. Since $`𝒮`$ is abelian, $`[UPU^1,P]=0`$ for each $`P`$ $`𝒮`$ and all unitary $`U𝒞_{\rho ,}`$. By Lemma 4.2 of (with $`𝔙=𝒞_{\rho ,}^{\prime \prime }=𝒞_{\rho ,}`$), this implies that $`P𝒞_{\rho ,}^{}`$.
$`𝒮`$ $`𝒞_{\rho ,}`$. Since $`\rho `$ is faithful, there is a one-parameter group $`\{\sigma _t:t\}`$ of automorphisms of $``$ — the modular automorphism group of $``$ determined by $`\rho `$ (, Sec. 9.2) — leaving $`\rho `$ invariant. Since $`𝒞_{\rho ,}`$ consists precisely of the fixed points of the modular group (, Prop. 9.2.14), it suffices to show that it leaves the individual elements of $`𝒮`$ fixed. For this, we use the fact that the modular group satisfies the KMS condition with respect to $`\rho `$: for each $`A,B`$, there is a complex-valued function $`f`$, bounded and continuous on the strip $`\{z:0\mathrm{I}mz1\}`$ in the complex plane, and analytic on the interior of that strip, such that
$$f(t)=\rho (\sigma _t(A)B),f(t+i)=\rho (B\sigma _t(A)),t.$$
(10)
In fact, we shall need only one simple consequence of the KMS condition, viz., if $`f(t)=f(t+i)`$ for all $`t`$, then $`f`$ is constant (, p. 611).
Fix an arbitrary projection $`P`$ $`𝒮`$. Since the modular automorphism group must leave $`𝒮`$ as a whole invariant, and $`𝒮`$ is abelian, $`[\sigma _t(P),P^{}]=0`$ for all $`t`$. However, there exists a function $`f`$ with the above properties such that
$$f(t)=\rho (\sigma _t(P)P^{})=\rho (P^{}\sigma _t(P))=f(t+i),t,$$
so it follows that $`f`$ is constant. In particular, since $`\sigma _0(P)P^{}=PP^{}=0`$, $`f`$ is identically zero, and $`\rho (\sigma _t(P)P^{})=0`$ for all $`t`$. And since $`\sigma _t(P)`$ and $`P^{}`$ are commuting projections, their product is a (positive) projection, so that the faithfulness of $`\rho `$ requires that $`\sigma _t(P)P^{}=0`$, or equivalently $`\sigma _t(P)=\sigma _t(P)P`$, for all $`t`$. Running through the exact same argument, starting with $`P^{}`$ $`𝒮`$ in place of $`P`$, yields $`\sigma _t(P^{})=\sigma _t(P^{})P^{}`$, or equivalently, $`\sigma _t(P)P=P`$, for all $`t`$. Together with $`\sigma _t(P)=\sigma _t(P)P`$, this implies that $`\sigma _t(P)=P`$ for all $`t`$. $`QED.`$
The choice $`_{\rho ,}=𝒵(𝒞_{\rho ,})`$ has another feature that generalizes a natural consequence of the modal interpretation of nonrelativistic quantum theory. Suppose the universal state $`x`$H<sub>U</sub> defines a faithful state $`\rho _x`$ on both $`(𝖧_S)`$ and $`(𝖧_{\overline{S}})`$. This requires that $`dim𝖧_S`$ $`=dim𝖧_S=n`$ (possibly $`\mathrm{}`$), and, furthermore, that any Schmidt decomposition of the state vector $`x`$ relative to the factorization H$`{}_{U}{}^{}=𝖧_S𝖧_{\overline{S}}`$ takes the form
$$x=\underset{i=1}{\overset{n}{}}c_iv_iw_i,c_i0\text{for all }i=1\text{ to }n\text{,}$$
(11)
where the vectors $`v_i`$ and $`w_i`$ are complete orthonormal bases in their respective spaces. As is well-known, for each distinct eigenvalue $`\stackrel{~}{\lambda }_j`$ for $`D_S`$, the span of the vectors $`v_i`$ for which $`|c_i|^2=\stackrel{~}{\lambda }_j`$ coincides with the range of the $`\stackrel{~}{\lambda }_j`$-eigenprojection of $`D_S`$, and similarly for $`D_{\overline{S}}`$ . Consequently, there is a natural bijective correspondence between the properties represented by the projections in the two sets $`_S`$ $`=D_S^{\prime \prime }`$ and $`_{\overline{S}}`$ $`=D_{\overline{S}}^{\prime \prime }`$: any definite property $`S`$ happens to possess is strictly correlated to a unique property of its environment $`\overline{S}`$ that occurs with the same frequency. More formally, for any $`P_S`$, there is a unique $`\overline{P}_{\overline{S}}`$ satisfying
$$(x,P\overline{P}x)=(x,Px)=(x,\overline{P}x).$$
(12)
To see this, note that any $`P_S`$ (in this case, the set of all functions of $`D_S`$) is a sum of spectral projections of $`D_S`$. Let $`\overline{P}_{\overline{S}}`$ be the sum of the corresponding spectral projections of $`D_{\overline{S}}`$ for the same eigenvalues. Then it is evident from the form of the state expansion in (11) that $`\overline{P}`$ has the property in (12), and no other projection in $`_{\overline{S}}`$ does. This has led some non-atomic modal interpreters, such as Kochen , to interpret each property $`P`$ of $`S`$, not as a property that $`S`$ possesses absolutely, but only in relation to its environment $`\overline{S}`$ possessing the corresponding property $`\overline{P}`$.
For a general von Neumann factor $``$, $`(^{})^{\prime \prime }`$ $`=(𝖧)`$ need not be isomorphic to the tensor product $`^{}`$ (particularly when $``$ is type III, for then $`^{}`$ must be type III as well). Therefore, there is no direct analogue of a Schmidt decomposition for a pure state $`x𝖧`$ relative to the factorization $`(^{})^{\prime \prime }`$ of $`(𝖧)`$. Nevertheless, we show next that there is still the same strict correlation between definite properties in $`_{\rho _x,}`$ $`=𝒵(𝒞_{\rho _x,})`$ and $`_{\rho _x,^{}}`$ $`=𝒵(𝒞_{\rho _x,^{}})`$.
Proposition 2 Let $``$ be a von Neumann algebra acting on a Hilbert space $`𝖧`$, and suppose $`x𝖧`$ induces a state $`\rho _x`$ that is faithful on both $``$ and $`^{}`$. Then for any projection $`P𝒵(𝒞_{\rho _x,})`$, there is a unique projection $`\overline{P}𝒵(𝒞_{\rho _x,^{}})`$ such that $`(x,P\overline{P}x)=(x,Px)=(x,\overline{P}x)`$.
Proof: For any fixed $`A,`$ call an element $`B^{}`$ a double for $`A`$ (in state $`x`$) just in case $`Ax=Bx`$ and $`A^{}x=B^{}x`$ . By an elementary application of modular theory, Werner (, Sec. II) has shown that $`𝒞_{\rho _x,}`$ consists precisely of those elements of $``$ with doubles in $`^{}`$ (with respect to $`x`$). Moreover, the double of any element of $``$ clearly has to be unique, by the faithfulness of $`\rho _x`$ on $`^{}`$. Now it is easy to see (again using the faithfulness of $`\rho _x`$) that the double of any projection $`P𝒞_{\rho _x,}`$ is a projection $`\overline{P}𝒞_{\rho _x,^{}}`$ satisfying (12). We claim that whenever $`P𝒞_{x,}^{}`$, we have $`\overline{P}𝒞_{x,^{}}^{}`$. For this, it suffices to show $`P𝒞_{x,}^{}`$ implies that for arbitrary $`B`$ $`𝒞_{x,^{}}`$, $`[\overline{P},B]x=0`$ (and then $`[\overline{P},B]`$ itself is zero, since $`\rho _x`$ is faithful). Letting $`A`$ $`𝒞_{x,}`$ be the double of $`B`$ in $``$, we get
$$\overline{P}BxB\overline{P}x=\overline{P}AxBPx=A\overline{P}xPBx=APxPAx=0,$$
(13)
as required. Finally, were there another projection $`\stackrel{~}{P}𝒵(𝒞_{\rho _x,^{}})`$ satisfying (12), then by exploiting the fact that $`\overline{P}`$ is $`P`$’s double in $`^{}`$, we get $`(x,\overline{P}\stackrel{~}{P}x)=(x,\overline{P}x)=(x,\stackrel{~}{P}x)`$; or, equivalently,
$$(x,\overline{P}\stackrel{~}{P}^{}x)=(x,\overline{P}^{}\stackrel{~}{P}x)=0.$$
(14)
Since $`𝒵(𝒞_{\rho _x,^{}})`$ is abelian, both $`\overline{P}\stackrel{~}{P}^{}`$ and $`\overline{P}^{}\stackrel{~}{P}`$ are (positive) projections in $``$. But as $`\rho _x`$ is faithful on $``$, Eqs. (14) entail that $`\overline{P}\stackrel{~}{P}^{}=\overline{P}^{}\stackrel{~}{P}=0`$, which in turn implies that $`\overline{P}=\stackrel{~}{P}`$, as required for uniqueness. $`QED.`$
Let us return now to the problem of picking out a set of definite-valued observables localized in a diamond region with associated algebra $`(\mathrm{}_r)`$. Let $`\rho `$ be any pure state of the field that induces a faithful state on $`(\mathrm{}_r)`$; for example, $`\rho `$ could be the vacuum or any one of the dense set of states of a field with bounded energy (by the Reeh-Schlieder theorem — see , Thm. 1.3.1). By Proposition 1, the definite-valued observables in $`(\mathrm{}_r)`$ are simply those in the subalgebra $`𝒵(𝒞_{\rho ,(\mathrm{}_r)})`$. Note that this proposal yields observables all of which have an exact spacetime localization within the open set $`\mathrm{}_r`$ and are picked out intrinsically by the local algebra $`(\mathrm{}_r)`$ and the field state $`\rho `$. Contrary to Dieks’ pessimistic conclusion, we can take open spacetime regions as fundamental for determining the definite-valued observables. In fact, this proposal works independent of the size of $`r`$, and so could also be embraced by non-atomic modal interpreters not wishing to commit themselves to a particular partition of the field into subsystems (or to thinking from the outset in terms of approximately point-localized field observables). Finally, note that since the algebra of a diamond region $`\mathrm{}_r`$ satisfies duality with respect to the algebra of its spacelike complement $`\mathrm{}_r^{}`$, i.e., $`(\mathrm{}_r)^{}=(\mathrm{}_r^{})`$ (, p. 145), Proposition 2 tells us that there is a natural bijective correspondence between the properties in $`𝒵(𝒞_{\rho ,(\mathrm{}__r)})`$ and strictly correlated properties in $`𝒵(𝒞_{\rho ,(\mathrm{}_r^{})})`$ associated with the complement region.
## 4 A Potential Difficulty with Ergodic States
We have seen that there is, after all, a well-motivated and unambiguous prescription extending the standard modal interpretation of nonrelativistic quantum theory to the local algebras of quantum field theory. We also, now, have a natural standard of comparison with Galilean quantum field theory. At least in the case of free fields, it is possible to build up local algebras in $`M`$ from spatially smeared “field algebras” defined on spacelike hyperplanes in $`M`$. A diamond region corresponds to the domain of dependence of a spatial region in a hyperplane, and it can be shown that the algebra of that spatial region will also be type III and, indeed, coincide with its domain of dependence algebra (, Prop. 3.3.2, Thm. 3.3.4). These type III spatial algebras in $`M`$, and the definite-valued observables therein, are what should be compared, in the nonrelativistic limit, to the corresponding equal time spatial algebras defined on simultaneity slices of Galilean spacetime. Unfortunately, since the algebras in the Galilean case are invariably type I (, p. 35), this limit is bound to be mathematically singular, and its physical characterization needs to be dealt with carefully. But this is a problem for any would-be interpreter of relativistic quantum field theory, not just modal interpreters. All we should require of them, at this stage, is that they be able to say something sensible in the relativistic case about the local observables with definite values (which was, indeed, Dieks’ original goal). However, as we now explain, it is not clear whether even this goal can be attained.
If $``$ $``$ $`(\overline{𝖧})`$ is type I, it possesses at most one faithful state $`\rho `$ such that $`𝒵(𝒞_{\rho ,})=I`$. This is easy to see, because if $`D(\overline{𝖧})`$ represents $`\rho `$, $`𝒵(𝒞_{\rho ,})D^{\prime \prime }`$, and $`D^{\prime \prime }=I`$ implies that $`D`$ itself must be a multiple of the identity. So when $`\overline{𝖧}`$ is finite-dimensional, we must have $`D=I/dim\overline{𝖧}`$, the unique maximally mixed state, and in the infinite-dimensional case, no such density operator even exists. Elsewhere Dieks has argued convincingly that there is no problem when a system, occupying a maximally mixed state, possesses only trivial properties, because such states are rare and highly unstable under environmental decoherence (cf. , pp. 99-100). However, the situation is quite different for the local algebras of algebraic quantum field theory.
In all physically reasonable models of the axioms of the theory, every local algebra $`(O)`$ is isomorphic to the unique (up to isomorphism) hyperfinite type III<sub>1</sub> factor (, Sec. V.6; , Sec. 17.2). In that case, there is a novel way to obtain $`𝒵(𝒞_{\rho ,(O)})=I`$, namely, when the state $`\rho `$ of $`(O)`$ is an ergodic state , i.e., $`\rho `$ possesses a trivial centralizer in $`(O)`$. (Were $``$ a nonabelian type I factor, this would be impossible, since $`D^{}=I`$ implies $`D^{\prime \prime }=(\overline{𝖧})`$ $``$, which is patently false.) In fact, we have the following result.
Proposition 3 If $``$ is the hyperfinite type III<sub>1</sub> factor, there is a norm dense set of unit vectors in the Hilbert space H on which $``$ acts that induce faithful states on $``$ with trivial centralizers (i.e., ergodic states).
Proof: First recall the following facts provable from the axioms of algebraic quantum field theory: (i) the vacuum state of a field on $`M`$ has a trivial centralizer in the algebra of any Rindler wedge (, Sec. 16.1.1); (ii) the vacuum is faithful for any wedge algebra (by the Reeh-Schlieder theorem); and (iii) wedge algebras are hyperfinite type III<sub>1</sub> factors (, Ex. 16.2.14, pp. 426-7). Since being faithful and having a trivial centralizer are isomorphic invariants, it follows that any instantiation $``$ of the hyperfinite type III<sub>1</sub> factor possesses at least one faithful normal state $`\rho `$ with trivial centralizer (even when $``$ is the algebra of a bounded open region, like a diamond). Now since $``$ is type III, all its states are vector states (combine , Cor. 2.9.28 with , Thm. 7.2.3); in particular, $`\rho =\rho _x`$ for some unit vector $`x`$H. Furthermore, by the homogeneity of the state space of type III<sub>1</sub> factors (, Cor. 6), the set of all unit vectors of the form $`UU^{}x`$, with $`U`$ and $`U^{}^{}`$ unitary operators, lies dense in H. But clearly any such vector must again induce a faithful state on $``$ with trivial centralizer. $`QED.`$
Combining Propositions 1 and 3, there will be a whole host of states of any relativistic quantum field in which the modal interpreter is forced to assert that no nontrivial local observables have definite values! Note, however, that while the set of field states ergodic for any given type III<sub>1</sub> local algebra is always dense, this does not automatically imply that such states are typical or generic. Indeed, results of Summers and Werner (, particularly Cor. 2.4) imply that for any local diamond algebra $`(\mathrm{}ß_r)`$, there will also always be a dense set field states whose centralizers in $`(\mathrm{}_r)`$ contain the hyperfinite type II<sub>1</sub> factor, and so will not be trivial. Still, the modal interpreter needs to provide some physical reason for neglecting the densely many field states that do yield trivial definite-valued observables locally. Obviously instability under decoherence is no longer be relevant.
Perhaps one could try to bypass Proposition 1 by exploiting extra structure not contained in the particular field state and local algebra to pick out the definite-valued observables in a region. For example, one might try to exploit the field’s total energy-momentum operator, and, in particular, its generator of time evolution. In the context of the nonrelativistic modal interpretation, Bacciagaluppi et al. (cf. also , p. 1181) have successfully invoked the analytic properties of the time evolution of the spectral projections of a system’s reduced density operator $`D_S`$ to avoid discontinuities that occur in the definite-valued set $`_S`$ at moments of time where the multiplicity of the eigenvalues of $`D_S`$ changes. Their methods yield a natural dynamical way, independent of instability considerations, to avoid the trivial definite-valued sets determined by maximally mixed density operators. So one might hope that these same dynamical methods could be extended to type III<sub>1</sub> algebras so as to yield a richer set of properties in a local region than Proposition 1 allows for ergodic states. In any case, modal interpreters need to do more work to show that their interpretation yields sensible local properties in quantum field theory (even before one considers, with Dieks, how to define Lorentz invariant decoherent histories of properties).
Acknowledgement
The author wishes to thank Hans Halvorson (for supplying the argument immediately following Eq. (2) and suggesting the use of the KMS condition in the proof of Proposition 1), and Reinhard Werner (for helpful correspondence about centralizers and inspiring Proposition 3).
|
warning/0003/math0003108.html
|
ar5iv
|
text
|
# A Survey of Noncommutative Chern Characters
## 1 Introduction
### 1.1 Motivation
It is well-known the Chern characters
$$ch:K^{}\left(X\right)H_{DR}^{}\left(X\right)$$
as homomorphisms from cohomological K-groups to $`𝐙/\left(2\right)`$-graded de Rham cohomology groups. The first theory is related with index theory and the second theory - with integration over manifolds. It is also well-known that modulo torsions, they are isomorphisms. More precisely, without torsions, they become isomorphisms
$$ch_𝐐:K^{}\left(X\right)𝐐H_{DR}^{}(X;𝐐).$$
This means that we can consider the homology theory as some approximation of K-theory. And therefore, the indices of elliptic operators related with integration over manifolds. The discovery of Atiyah-Hirzebruch-Singer index formulla reflects this relation, expressing the indices of elliptic operators, which count the differences of dimensions of kernels and cokernels of Fredholm operators, with some integrals over manifolds, which evaluate some characteristic classes over the fundamental class of manifolds.
Chern characters play an important role in the problem of describing the structure of group C\*-algebras and quantum groups, particularly in the Atiyah-Hirzebruch-Singer index formula for Dirac operators on $`Spin^c`$-manifold,
$$Ind𝐃=2^{\frac{d\left(d1\right)}{2}}Td\left(X\right)Ch_E,\left[X\right],\text{ where }d=dim\left(X\right).$$
It is well-known that the algebras of functions on manifolds define the structure of manifolds. One can therefore define manifolds as some special class of commutative algebras. From this we have therefore a new look to the notion of manifolds: In the K-theory context this can be expressed as follows. Following a well-known Serre-Swan theorem, the topological K-theory of the topological compact $`X`$ is isomorphic to the algebraic K-theory $`K_{}\left(𝐂\left(X\right)\right)`$ of the Banach algebra $`𝐂\left(X\right)`$ of continuous fuctions on $`X`$. In the de Rham theory context, the cohomology of topological spaces is isomorphic to the de Rham homology of the algebras of smooth functions on manifolds. The Chern characters have sense as homomorphisms, and in many cases, isomorphisms between them. Our main goal is to do “the same” for more general (noncommutative) algebras. In this survey, three kinds of noncommutative algebras are considered: compact Lie group C\*-algebras, compact quantum group C\*-algebras and some classes of quantum algebras via deformation quantization:
The group C\*-algebras were introduced many year ago, but still it is difficult to describe the structure of the group C\*-algebras of noncompact locally compact groups. We introduce the new idea related with the Chern characters of noncommutative group C\*-algebras. We are going also to consider algebras of functions on the quantum groups. These algebras are obtained from some algebraic deformation as it was in the works of Vaksman and Soibelman. A rich variety of quantum algebras is the deformation quantization of Poisson structure. The idea simulizes what was done in Physics.
### 1.2 Tools and Ideas
Noncommutative geometry with two important tools: KK-theory and cyclic homology. A. Connes introduced the cyclic homology theory and then the entire cyclic periodic homology $`HE_{}\left(A\right)`$. He shows that this theory has all good properties of a homology theory like: homotopy invariant, Morita invariance and excision and constructed the Chern characters
$$ch:K_{}\left(A\right)HE_{}\left(A\right)$$
with the help of pairing
$$K_{}\left(A\right)\times HE^{}\left(A\right)𝐂.$$
This theory recovered the classical results when $`A=𝐂\left(X\right)`$ is a commutative Banach algebra. It is more difficult to do any computation for noncommutative algebras. We consider the case when the algebra $`A`$ can be presented as an inductive limits of good ideals $`I_\alpha `$ and restrict the pairing of Connes
$$K_{}\left(A\right)\times HE^{}\left(A\right)𝐂$$
to every ideal $`I_\alpha `$ with $`ad`$-invariant trace
$$K_{}\left(A\right)\times HE^{}\left(I_\alpha \right)𝐂$$
and then extend to
$$K_{}\left(A\right)\times _\alpha HE^{}\left(I_\alpha \right)𝐂.$$
This therefore gives us a homomorphisms
$$K_{}\left(A\right)\left(_\alpha HE^{}\left(I_\alpha \right)\right)^{}$$
as some generalized Chern character. The question is to introduc a modified entire periodic cyclic homology $`HE_{}\left(\underset{}{\mathrm{lim}}I_\alpha \right)`$ as some new $`HE_{}\left(A\right)`$ with the same properities. This modification, fall constructed gives us a possibility to compute the noncommutative Chern characters in the indicated 3 cases: the group C\*-algebras of compact Lie groups, the compact quantum group C\*-algebras and some quantum coadjoint orbits, appeared from deformation quantization. This review is closely related to and can be considered as the second part of our previous review\[D1\] and book\[D2\].
## 2 Compact Lie Group C\*-Algebras
Let us recall $`G`$ a compact Lie group, $`L^1\left(G\right)`$ involutive Banach algebra of functions with convolution product w. r. t. the natural Haar measure. The Banach $`L^1`$ norm is irregular, i.e. in general
$$aa^{}a^2.$$
One introduced, therfore, the regular norm
$$a_{C^{}\left(G\right)}:=\underset{\pi \widehat{G}}{sup}\pi \left(a\right).$$
The C\*-algebra is the completion with respect to this C\* -norm. It is well known for compact groups:
1. Every representations is unitarizable and every irreducible unitary representation is finite dimensional, namely of dimension $`n_i`$
2. The set of equivalent classes of irreducible unitary representations are not more than denombrable.
As consequence it was proven:
###### Theorem 2.1
The group C\*-algebras of compact groups can be expressed as topological Cartisean product of matix algebras:
$$C^{}\left(G\right)\left(\stackrel{~}{}\right)_{i=1}^{\mathrm{}}Mat_{n_i}\left(𝐂\right),$$
Let us recall that \[DT1\] $`\tau _\alpha `$ is an $`ad`$-invariant trace on an ideal $`I_\alpha `$ in a Banach algebra $`A`$ iff:
1. $`\tau _\alpha `$ is continuous linear, $`\tau _\alpha =1`$
2. $`\tau _\alpha `$ is positive $`\tau _\alpha \left(a^{}a\right)0,\alpha `$ and strictly positive $`\tau _\alpha \left(a^{}a\right)=0\text{ iff }a=0`$
3. $`\tau _\alpha `$ is $`ad`$-invariant, i.e. $`\tau _\alpha \left([a,x]\right)=0,xA,a\tau _\alpha .`$
###### Corollary 2.2
The C\*-algebra $`C^{}(G)`$ can be presented as inductive limit of ideal
$$C^{}\left(G\right)\underset{}{\mathrm{lim}}_NI_N,I_N:=\underset{i=1}{\overset{N}{}}Mat_{n_i}\left(𝐂\right).$$
## 3 Algebras of Functions on Compact Quantum Groups
For a compact Lie group $`G`$ with Lie algebra $`𝔤`$, define Hopf algebra $`_\epsilon \left(G\right)`$ of functions over the quantum group, which is dual to the quantized universal enveloping algebra $`U_\epsilon \left(𝔤\right)`$.
###### Definition 3.1
C\*-norm:
$$f:=\underset{\rho }{sup}\rho \left(f\right)\left(f_\epsilon \left(G\right)\right)$$
###### Theorem 3.2
$$C_\epsilon ^{}\left(G\right)𝐂\left(𝐓\right)\underset{ewW}{}_{W\times 𝐓}𝒦\left(𝐇_{w,t}\right)𝑑t,$$
###### Corollary 3.3
Compact quantum group C\*-algebras can be presented as inductive limits of ideals with $`ad`$-invariant trace.
## 4 C\*-Algebraic Noncommutative Chern Characters
### 4.1 K-Theory of Banach algebras
$`A`$ Banach algebra with unit 1
###### Theorem 4.1
For Banach \*-algebras,
$$KK(A,𝐂)K_{}^{top}\left(A\right)K_{}^{alg}\left(A\right)$$
Let us recall that the algebraic K\*-groups are defined as follows: $`K_0\left(A\right)`$ is the Abelian group envelop of the Grothendieck semi-group, generated by projective finitely generated $`A`$-modules. The $`K_1`$-groups are defined as the Whitehead groups, which are the abelized infinite general linear groups $`GL_{\mathrm{}}\left(A\right)/[GL_{\mathrm{}}\left(A\right),GL_{\mathrm{}}\left(A\right)]`$
### 4.2 Entire homology
Let us consider a Banach \*-algebra which can be presented as a direct limit of special ideal $`I_\alpha `$
$$A=\underset{}{\mathrm{lim}}_\alpha (I_\alpha ,\tau _\alpha ),$$
where $`\tau _\alpha `$ is an $`ad`$-invariant trace on $`I_\alpha `$, i.e. :
1. $`\tau _\alpha `$ is continuous linear, $`\tau _\alpha =1`$
2. $`\tau _\alpha `$ is positive $`\tau _\alpha \left(a^{}a\right)0,\alpha `$ and strictly positive $`\tau _\alpha \left(a^{}a\right)=0\text{ iff }a=0`$
3. $`\tau _\alpha `$ is $`ad`$-invariant, i.e. $`\tau _\alpha \left([a,x]\right)=0,xA,a\tau _\alpha .`$
We then have for every $`\alpha \mathrm{\Gamma }`$ a scalar product
$$a,b_\alpha :=\tau _\alpha \left(a^{}b\right)$$
and also an direct system $`\{I_\alpha ,\tau _\alpha \}_{\alpha \mathrm{\Gamma }}`$. Let $`\overline{I}_\alpha `$ be the completion of $`I_\alpha `$ with respect to the scalar product defined above and $`\stackrel{~}{\overline{I}_\alpha }`$ denote $`\overline{I}_\alpha `$ with formally adjointed unity element. Define $`C^n\left(\stackrel{~}{\overline{I}_\alpha }\right)`$ as the set of $`n+1`$-linear maps $`\phi :\left(\stackrel{~}{\overline{I}_\alpha }\right)^{\left(n+1\right)}𝐂`$. There exists a Hilbert structure on $`\left(\stackrel{~}{\overline{I}_\alpha }\right)^{\left(n+1\right)}`$ and we can identify $`C_n\left(\left(\stackrel{~}{\overline{I}_\alpha }\right)\right):=Hom(C^n\left(\overline{\stackrel{~}{I}}_\alpha \right),𝐂)`$ with $`C^n\left(\overline{\stackrel{~}{I}}_\alpha \right)`$ via an anti-isomorphism.
For $`I_\alpha I_\beta `$, we have a well-defined map
$$D_\alpha ^\beta :C^n\left(\stackrel{~}{\overline{I}_\alpha }\right)C^n\left(\stackrel{~}{\overline{I}_\beta }\right),$$
which makes $`\left\{C^n\left(\stackrel{~}{\overline{I}}_\alpha \right)\right\}`$ into a direct system. Write $`Q=\underset{}{\mathrm{lim}}C^n\left(\stackrel{~}{\overline{I}_\alpha }\right)`$ and observe that it admits a Hilbert space structure, see \[DT1\]-\[DT2\]. Let $`C_n\left(A\right):=Hom(\underset{}{\mathrm{lim}}C^n\left(\stackrel{~}{\overline{I}_\alpha }\right),𝐂)=Hom(\underset{}{\mathrm{lim}}C_n\left(\stackrel{~}{\overline{I}_\alpha }\right),𝐂)`$ which is anti-isomorphic to $`\underset{}{\mathrm{lim}}_\alpha C^n\left(\overline{\stackrel{~}{I}}_\alpha \right)`$. So we have finally
$$C_n\left(A\right)=\underset{}{\mathrm{lim}}C_n\left(\overline{\stackrel{~}{I}}_\alpha \right).$$
Let
$$b,b^{}:C^n\left(\stackrel{~}{\overline{I}}_\alpha \right)C^{n+1}\left(\stackrel{~}{\overline{I}}_\alpha \right),$$
$$N:C^n\left(\stackrel{~}{\overline{I}}_\alpha \right)C^n\left(\stackrel{~}{\overline{I}}_\alpha \right),$$
$$\lambda :C^n\left(\stackrel{~}{\overline{I}}_\alpha \right)C^n\left(\stackrel{~}{\overline{I}}_\alpha \right),$$
$$S:C^{n+1}\left(\stackrel{~}{\overline{I}}_\alpha \right)C^n\left(\stackrel{~}{\overline{I}}_\alpha \right)$$
be defined as in A. Connes \[Ca\]. We adopt the notations in \[Kh1\] . Denote by $`b^{},\left(b^{}\right)^{},N^{},\lambda ^{},S^{}`$ the corresponding adjoint operators. Note also that for each $`I_\alpha `$ we have the same formulae for adjoint operators for homology as Connes obtained for cohomology.
We now have a bi-complex
$`𝒞\left(A\right):\begin{array}{ccccccccc}& & \mathrm{}& & \mathrm{}& & \mathrm{}& & \\ & & \left(b^{}\right)^{}& & b^{}& & \left(b^{}\right)^{}& & & & \\ \mathrm{}& \stackrel{1\lambda ^{}}{}& \underset{}{\mathrm{lim}}_\alpha C_1\left(\overline{\stackrel{~}{I}}_\alpha \right)& \stackrel{N^{}}{}& \underset{}{\mathrm{lim}}_\alpha C_1\left(\overline{\stackrel{~}{I}}_\alpha \right)& \stackrel{1\lambda ^{}}{}& \underset{}{\mathrm{lim}}_\alpha C_1\left(\overline{\stackrel{~}{I}}_\alpha \right)& \stackrel{N^{}}{}& \mathrm{}\\ & & \left(b^{}\right)^{}& & b^{}& & \left(b^{}\right)^{}& & & & \\ \mathrm{}& \stackrel{1\lambda ^{}}{}& \underset{}{\mathrm{lim}}_\alpha C_0\left(\overline{\stackrel{~}{I}}_\alpha \right)& \stackrel{N^{}}{}& \underset{}{\mathrm{lim}}_\alpha C_0\left(\overline{\stackrel{~}{I}}_\alpha \right)& \stackrel{1\lambda ^{}}{}& \underset{}{\mathrm{lim}}_\alpha C_0\left(\overline{\stackrel{~}{I}}_\alpha \right)& \stackrel{N^{}}{}& \mathrm{}\end{array}`$
with $`d_v:=b^{}`$ in the even columns and $`d_v:=\left(b^{}\right)^{}`$ in the odd columns, $`d_h:=1\lambda ^{}`$ from odd to even columns and $`d_h:=N^{}`$ from even to odd columns, where * means the corresponding adjoint operator. Now we have
$$Tot\left(𝒞\left(A\right)\right)^{even}=Tot\left(𝒞\left(A\right)\right)^{odd}:=_{n0}C_n\left(A\right),$$
which is periodic with period two. Hence, we have
$$\begin{array}{ccc}& & \\ _{n0}C_n\left(A\right)& \begin{array}{c}\\ \end{array}& _{n0}C_n\left(A\right)\\ & & \end{array}$$
where $`=d_v+d_h`$ is the total differential.
###### Definition 4.2
Let $`HP_{}(A)`$ be the homology of the total complex $`(Tot𝒞(A))`$. It is called the periodic cyclic homology of $`A`$.
Note that this $`HP_{}\left(A\right)`$ is, in general different from the $`HP_{}\left(A\right)`$ of Cuntz-Quillen, because we used the direct limit of periodic cyclic homology of ideals. But in special cases, when the whole algebra $`A`$ is one of these ideals with $`ad_A`$-invariant trace, (e.g. the commutative algebras of complex-valued functions on compact spaces), we return to the Cuntz-Quillen $`HP_{}`$, which we shall use later.
###### Definition 4.3
An even (or odd) chain $`(f_n)_{n0}`$ in $`𝒞(A)`$ is called entire if the radius of convergence of the power series $`_n\frac{n!}{[\frac{n}{2}]!}f_nz^n`$, $`z𝐂`$ is infinite.
Let $`C_{}^e\left(A\right)`$ be the sub-complex of $`C\left(A\right)`$ consisting of entire chains. Then we have a periodic complex.
###### Theorem 4.4
Let
$$Tot\left(C_{}^e\left(A\right)\right)^{even}=Tot\left(C_{}^e\left(A\right)\right)^{odd}:=\underset{n0}{}C_n^e\left(A\right),$$
where $`C_n^e(A)`$ is the entire $`n`$-chain. Then we have a complex of entire chains with the total differential $`:=d_v+d_h`$
$$\begin{array}{ccc}& & \\ _{n0}C_n^e\left(A\right)& \begin{array}{c}\\ \end{array}& _{n0}C_n^e\left(A\right)\\ & & \end{array}$$
###### Definition 4.5
The homology of this complex is called also the entire homology and denoted by $`HE_{}(A)`$.
Note that this entire homology is defined through the inductive limits of ideals with ad-invariance trace.
In \[DT1\]-\[DT2\], the main properties of this theory, namely
* Homotopy invariance,
* Morita invariance and
* Excision,
were proven and hence $`HE_{}`$ is a generalized homology theory.
###### Lemma 4.6
If the Banach algebra $`A`$ can be presented as a direct limit $`\underset{}{\mathrm{lim}}_\alpha I_\alpha `$ of a system of ideals $`I_\alpha `$ with $`ad`$-invariant trace $`\tau _\alpha `$, then
$$K_{}\left(A\right)=\underset{}{\mathrm{lim}}_\alpha K_{}\left(I_\alpha \right),$$
$$HE_{}\left(A\right)=\underset{}{\mathrm{lim}}_\alpha HE_{}\left(I_\alpha \right).$$
The following result from K-theory is well-known:
###### Theorem 4.7
The entire homology of non-commutative de Rham currents admits the following stability property
$$K_{}\left(𝒦\left(𝐇\right)\right)K_{}\left(𝐂\right),$$
$$K_{}\left(A𝒦\left(𝐇\right)\right)K_{}\left(A\right),$$
where $`𝐇`$ is a separable Hilbert space and $`A`$ is an arbitrary Banach space.
The similar result is true for entire homology $`HE_{}`$ :
###### Theorem 4.8
The entire homology of non-commutative de Rham currents admits the following stability property
$$HE_{}\left(𝒦\left(𝐇\right)\right)HE_{}\left(𝐂\right),$$
$$HE_{}\left(A𝒦\left(𝐇\right)\right)HE_{}\left(A\right),$$
where $`𝐇`$ is a separable Hilbert space and $`A`$ is an arbitrary Banach space.
Let $`A`$ be an involutive Banach algebra. In this section, we construct a non-commutative character
$$ch:K_{}\left(A\right)HE\left(A\right)$$
and later show that when $`A=C^{}\left(G\right)`$, this Chern character reduces up to isomorphism to classical Chern character.
Let $`A`$ be an involutive Banach algebra with unity.
###### Theorem 4.9
There exists a Chern character
$$ch:K_{}\left(A\right)HE_{}\left(A\right).$$
We first recall that there exists a pairing
$$K_n\left(A\right)\times C^n\left(A\right)𝐂$$
due to A. Connes, see \[Co\]. Hence there exists a map $`K_n\left(A\right)\stackrel{C_n}{}Hom(C^n\left(A\right),𝐂).`$ So, by 1.1, we have for each $`\alpha \mathrm{\Gamma }`$ a pairing
$$K_n\left(A\right)\times C^n\left(\stackrel{~}{\overline{I}}_\alpha \right)𝐂$$
and hence a map $`K_n\left(A\right)\stackrel{C_n^\alpha }{}Hom(C^n\left(\stackrel{~}{\overline{I}_\alpha }\right),𝐂)`$ and hence a map $`K_n\left(A\right)\stackrel{C_n}{}Hom(C^n\left(\stackrel{~}{\overline{I}_\alpha }\right),𝐂).`$ We now show that $`C_n`$ induces the Chern map
$$ch:K_n\left(A\right)HE_n\left(A\right)$$
Now let $`e`$ be an idempotent in $`M_k\left(A\right)`$ for some $`k𝐍`$. It suffies to show that for $`n`$ even, if $`\phi =\psi `$, where $`\phi C^n\left(\stackrel{~}{\overline{I}_\alpha }\right)`$ and $`\psi C^{n+1}\left(\stackrel{~}{\overline{I}_\alpha }\right)`$, then
$$e,\phi =\underset{n=1}{\overset{\mathrm{}}{}}\frac{\left(1\right)^n}{n!}\phi (e,e,\mathrm{},e)=0.$$
However, this follows from Connes’ results in (\[Co\], Lemma 7).
The proof of the case for $`n`$ odd would also follow from \[Co\].
### 4.3 Compact Lie group C\*-algebra case
Our next result computes the Chern character in 2.1 for $`A=C^{}\left(G\right)`$ by reducing it to the classical case. Let us recall that the group C\*-algebras of compact locally compact groups can be presented as some inductive limits of $`ad`$-invariant ideals of type $`I_N=_{i=1}^NMat_{n_i}\left(𝐂\right)`$,
$$C^{}\left(G\right)=\underset{}{\mathrm{lim}}_NI_N,\text{ where }I_N=\underset{i=1}{\overset{N}{}}Mat_{n_i}\left(𝐂\right).$$
Let $`𝐓`$ be a fixed maximal torus of $`G`$ with Weyl group $`W:=𝒩_G\left(𝐓\right)/𝐓`$.
###### Theorem 4.10
Then the Chern character
$$K_{}^G\left(𝐂^{}\left(G\right)\right)HE_{}^G\left(C^{}\left(G\right)\right)$$
is an isomorphism, which can be identified with the classical Chern character
$$K_{}^W\left(𝐂\left(𝐓\right)\right)HE_{}^W\left(𝐂\left(𝐓\right)\right)$$
that is also an isomorphism.
Because of the pairing
$$K_{}\left(A\right)\times HE^{}\left(A\right)𝐂,$$
and because $`A_\alpha `$ are supposed to be ideals in $`A`$, we can extend it to the pairing
$$K_{}\left(A\right)\times \underset{}{\mathrm{lim}}_\alpha HE^{}\left(A_\alpha \right)𝐂,$$
and therefore
$$K_{}\left(A\right)\times \underset{\alpha }{}\underset{}{\mathrm{lim}}_\alpha HE^{}\left(A_\alpha \right)𝐂.$$
### 4.4 Quantum group C\*-algebra case
In this case of compact quantum group C\*-algebras we have also
$$C_\epsilon ^{}\left(G\right)𝐂\left(𝐓\right)_{𝒩_𝐓\left\{e\right\}\times 𝐓}^{}𝒦\left(𝐇_{w,t}\right)𝑑t$$
Our next result computes the Chern characters for $`A=C_\epsilon ^{}\left(G\right)`$ by reducing it to the classical case.
###### Theorem 4.11
Let $`𝐓`$ be a fixed maximal torus of $`G`$ with Weyl group $`W:=𝒩_𝐓/𝐓`$. Then the Chern character
$$ch_C^{}:K_{}\left(C_\epsilon ^{}\left(G\right)\right)HE_{}\left(C_\epsilon ^{}\left(G\right)\right)$$
is an isomorphism modulo torsion, i.e.
$$\begin{array}{ccc}ch_C^{}:K_{}\left(C_\epsilon ^{}\left(G\right)\right)𝐂& \stackrel{}{}& HE_{}\left(C_\epsilon ^{}\left(G\right)\right),\end{array}$$
which can be identified with the classical Chern character
$$\begin{array}{ccc}ch:K_{}\left(𝐂\left(𝒩_𝐓\right)\right)& & HE_{}\left(𝐂\left(𝒩_𝐓\right)\right)\end{array}$$
that is also an isomorphism modulo torsion, i.e.
$$\begin{array}{ccc}ch:K_{}\left(𝒩_𝐓\right)𝐂& \stackrel{}{}& H_{DR}^{}\left(𝒩_𝐓\right).\end{array}$$
## 5 Quantum Strata of Coadjoint Orbits
Let us now consider the third class of noncommutative algebras, arized from deformation quantization. In this section we expose the results obtained in \[D4\]. For locally compact groups, their C\*-algebras contain exhausted informations about the groups them-selves and their representations, see \[D1\], \[D2\]. In some sense \[R1\]-\[R2\], the group algebras can be considered as C\*-algebraic deformation quantization $`C_q^{}\left(G\right)`$ at the special value $`q=1`$.
In \[D1\] and \[D2\], the group C\*-algebras were described as repeated extensions of C\*-algebras of strata of coadjoint orbits. Quantum groups are group Hopf algebras, i.e. replace C\*-algebras by Hopf algebras. It is therefore interesting to ask whether we could describe quantum groups as some repeated extensions of some kind quantum strata of coadjoint orbits? We are attempting to give a positive answer to this question. It is not yet completely described but we obtained a reasonable answer. Let us describe the main ingredients of our approach.
For the good strata of coadjoint orbits, there exist always continuous fields of polarizations (in the sense of the representation theory), satisfying the L. Pukanszky irreducibility condition: for each orbit $`𝒪`$ and any point $`F_𝒪`$ in it, the affine subspace, orthogonal to polarizations with respect to the symplectic form is included in orbits themselves, i.e.
$$F_𝒪+𝔥_𝒪^{}𝒪$$
and
$$dim𝔥_𝒪=\frac{1}{2}dim𝒪.$$
We choose the the canonical Darboux coordinates with impulse $`p`$’s-coordinates, following a vector structure basis of $`𝔥^{}`$. From this we can deduce that in this kind of Darboux coordinates, the Kirillov form $`\omega _𝒪`$ locally are canonical and every element $`X𝔤=LieG`$ can be considered as a function $`\stackrel{~}{X}`$ on $`𝒪`$, linear on $`p`$’s-coordinates, i.e.
$$\stackrel{~}{X}=\underset{i=1}{\overset{n}{}}a_i\left(q\right)p_i+a_0\left(q\right).$$
This essential fact gives us a possibility to effectively write out the corresponding $``$-product of functions, and define quantum strata $`C_q(V,)`$. On the strata acts our Lie group of symmetry. It induces therefore an action on equivariant differential operators. Using the indicate fields of polarizations we prove some kind of Poincaré-Birkhoff-Witt theorem and then provide quantization with separation of variable in sense of Karabegov \[Ka1\]. We can then express the corresponding representations of the quantum strata $`C_q(V,)`$, where $`V_{dim𝒪=const}𝒪`$, through the Feynman path integrals etc…,see \[D3\].
### 5.1 Canonical coordinates on a stratum
Let us consider a connected and simply connected Lie group $`G`$ with Lie algebra $`𝔤`$. Denote the dual to $`𝔤`$ vector space by $`𝔤^{}`$. It is well-known that the action of $`G`$ on itself by conjugation
$$A\left(g\right):GG,$$
defined by $`A\left(g\right)\left(h\right):=ghg^1`$ keeps the identity element $`h=e`$ unmoved. This induces the tangent map $`Ad\left(g\right):=A\left(g\right)_{}:𝔤=T_eG𝔤,`$ defined by
$$Ad\left(g\right)X:=\frac{d}{dt}|_{t=0}A\left(g\right)\mathrm{exp}\left(tX\right)$$
and the co-adjoint action $`K\left(g\right):=A\left(g^1\right)^{}`$ maps the dual space $`𝔤^{}`$ into itself. The orbit space $`𝒪\left(G\right):=𝔤^{}/G`$ is in general a bad topological space, namely non-Hausdorff. Consider an arbitrary orbit $`\mathrm{\Omega }𝒪\left(G\right)`$ and an element $`F𝔤^{}`$ in it. The stabilizer is denote by $`G_F`$, its connected component by $`\left(G_F\right)_0`$ and its Lie algebra by $`𝔤_F:=Lie\left(G_F\right)`$. It is well-known that
$$\begin{array}{ccc}G_F& & G\\ & & \\ & & \mathrm{\Omega }_F\end{array}$$
is a principal bundle with the structural group $`G_F`$. Let us fix some connection in this principal bundle, i.e. some trivialization of this bundle. We want to construct representations in some cohomology spaces with coefficients in the sheaf of sections of some vector bundle associated with this principal bundle. It is well know that every vector bundle is an induced one with respect to some representation of the structural group in the typical fiber. It is natural to fix some unitary representation $`\stackrel{~}{\sigma }`$ of $`G_F`$ such that its kernel contains $`\left(G_F\right)_0`$, the character $`\chi _F`$ of the connected component of stabilizer
$$\chi _F\left(\mathrm{exp}X\right):=\mathrm{exp}\left(2\pi \sqrt{1}F,X\right)$$
and therefore the differential $`D\left(\stackrel{~}{\sigma }\chi _F\right)=\stackrel{~}{\rho }`$ is some representation of the Lie algebra $`𝔤_F`$. We suppose that the representation $`D\left(\stackrel{~}{\rho }\chi _F\right)`$ was extended to the complexification $`\left(𝔤_F\right)_𝐂`$. The whole space of all sections seems to be so large for the construction of irreducible unitary representations. One consider the invariant subspaces with the help of some so called polarizations, see \[D1\], \[D2\].
###### Definition 5.1
We say that a triple $`(𝔭_𝒪,\rho _𝒪,\sigma _{0,𝒪})`$ is a $`(\stackrel{~}{\sigma },F)`$-polarization of $`𝒪`$ iff :
1. $`𝔭_𝒪`$ is some complex sub-algebra of the complexified $`𝔤_𝐂`$, containing $`𝔤_{F_𝒪}`$.
2. The sub-algebra $`𝔭_𝒪`$ is invariant under the action of all the operators of type $`Ad_{𝔤_𝐂}x,xG_{F_𝒪}.`$
3. The vector space $`𝔭_𝒪+\overline{𝔭_𝒪}`$ is the complexification of some real Lie sub-algebra $`𝔪_𝒪:=\left(𝔭_𝒪+\overline{𝔭}_𝒪\right)𝔤.`$
4. All the subgroups $`M_{0,𝒪}`$, $`H_{0,𝒪}`$, $`M_𝒪`$, $`H_𝒪`$ are closed, where by definition, $`M_{0,𝒪}`$ (resp., $`H_{0,𝒪}`$) is the connected subgroup of $`G`$ with the Lie algebra $`𝔪_𝒪`$ (resp., $`𝔥_𝒪:=𝔭_𝒪𝔤`$) and the semi-direct products $`M:=M_{0,𝒪}G_{F_𝒪}`$, $`H_𝒪:=H_{0,𝒪}G_{F_𝒪}`$.
5. $`\sigma _{0,𝒪}`$ is an irreducible representation of $`H_{0,𝒪}`$ in some Hilbert space $`V_𝒪`$ such that: (1.) the restriction $`\sigma _{0,𝒪}|_{G_{F_𝒪}H_{0,𝒪}}`$ is some multiple of the restriction $`\chi _{F_𝒪}.\stackrel{~}{\sigma }_𝒪|_{G_{F_𝒪}H_{0,𝒪}}`$, where the character $`\chi _𝒪`$ is by definition, $`\chi _𝒪\left(\mathrm{exp}X\right)=\mathrm{exp}\left(2\pi \sqrt{1}F_𝒪,X\right)`$; (2.) under the action of $`G_{F_𝒪}`$ on the dual $`\widehat{H}_{0,𝒪}`$, the point $`\sigma _{0,𝒪}`$ is fixed.
6. $`\rho _𝒪`$ is a representation of the complex Lie algebra $`𝔭_𝒪`$ in the same $`V_𝒪`$, which satisfies the Nelson conditions for $`H_{0,𝒪}`$ and $`\rho _𝒪|_{𝔥_𝒪}D\sigma _{0,𝒪}`$.
There is a natural order in the set of all $`(\stackrel{~}{\sigma }_𝒪,F_𝒪)`$-polarizations by inclusion and from now on speaking about polarizations we means always the maximal ones. It is not hard to prove that the (maximal) polarizations are also the Lagrangian distributions and in particular the co-dimension of $`𝔥_𝒪`$ in $`𝔤`$ is a half of the dimension of the coadjoint orbit $`𝒪`$,
$$codim_𝔤𝔥_𝒪=\frac{1}{2}dim𝒪,$$
see e.g. \[D2\] or \[Ki1\].
Let us now recall the Pukanszky condition.
###### Definition 5.2
We say that the $`(\stackrel{~}{\sigma }_𝒪,F_𝒪)`$-polarization $`(𝔭_𝒪,\rho _𝒪,\sigma _{0,𝒪})`$ satisfies the Pukanszky condition, iff
$$F_𝒪+𝔥_𝒪^{}𝒪.$$
###### Remark 5.3
* Pukanszky conditions involve an inclusion of the Lagrangian affine subspace of $`p`$’s coordinates into the local Darboux coordinates.
* The partial complex structure on orbits let us to use smaller subspaces of section in the induction construction, as subspaces of partially invariant, partially holomorphic sections of the induced bundles.
###### Theorem 5.4
There exists on each coadjoint orbit a local canonical system of Darboux coordinates, in which the Hamiltonian function $`\stackrel{~}{X}`$, $`X𝔤`$, are linear on $`p`$’s impulsion coordinates and in theses coordinates,
$$\stackrel{~}{X}=\underset{i=1}{\overset{n}{}}a_i\left(q\right)p_i+a_0\left(q\right).$$
### 5.2 Poisson structure on strata
Let us first recall the construction of strata of coadjoint orbit from \[D1\]-\[D2\]. The orbit space $`𝒪\left(G\right)`$ is a disjoint union of $`\mathrm{\Omega }_{2n}`$, each of which is a union of the coadjoint orbits of dimension $`2n`$, $`02ndimG`$. Denote
$$V_{2n}=_{dim𝒪=2n}𝒪.$$
Then $`V_{2n}`$ is the set of points of fixed rank $`2n`$ of the Poisson structure bilinear function
$$\{X,Y\}=F,[X,Y].$$
Suppose that it is a foliation, at least for $`V_{2n}`$, where $`2n`$ is the maximal dimension possible in $`𝒪\left(G\right)`$. It can be shown that the foliation $`V_{2n}`$ can be obtained from the group action of $`𝐑^{2n}`$ on $`V_{2n}`$. Let us for this aim, consider a basis $`X_1,\mathrm{},X_{2n}`$ of the tangent space $`T_{F_𝒪}𝒪𝔤/𝔤_{F_𝒪}`$ at the point $`F_𝒪𝒪V_{2n}`$. We can define an action of $`𝐑^{2n}`$ on $`V_{2n}`$ as
$$(t_1,\mathrm{},t_{2n})\mathrm{exp}\left(t_1X_1\right)\mathrm{exp}\left(t_2X_2\right)\mathrm{}\mathrm{exp}\left(t_{2n}X_{2n}\right)F_𝒪.$$
We have therefore the Hamiltonian vector fields
$$\xi _k:=\frac{d}{dt}|_{t=0}\mathrm{exp}\left(t_kX_k\right)F,k=1,\mathrm{},2n$$
and their span $`_{2n}=\{\xi _1,\mathrm{},\xi _{2n}\}`$ provides a tangent distribution. It is easy to show that we have therefore a measurable (in sense of A. Connes) foliation. One can therefore define also the Connes C\*-algebra $`C^{}(V_{2n},_{2n})`$, $`02ndimG`$. By introducing some technical condition in \[D1\], it is easy to reduced these C\*-algebras to extensions of other ones, those are in form of tensor product $`C\left(X\right)𝒦\left(H\right)`$ of algebras of continuous functions on compacts and the elementary algebra $`𝒦\left(H\right)`$ of compact operators in a separable Hilbert space $`H`$. The strata of these kinds we means $`good`$ strata. Another kind of strata of coadjoint orbits are obtained from relation with the cases where the Gelfand-Kirillov conjecture was solve, for examples for connected and simply connected solvable Lie groups, see §5.
It is deduced from a result of Kontsevich that this Poisson structure can be quantized. This quantization however is formal. The question of convergence of the quantizing series is not clear. We show in this section the case of charts with the linear p’s canonical coordinates, the corresponding $``$-product is convergent.
In this kind of special local chart systems of Darboux coordinates it is easy to deduce existence local convergent $``$-products.
###### Theorem 5.5
Locally on each coadjoint orbit, there exist a $`star`$-product.
Let us denote by $`_p^1`$ the Fourier inverse transformation on variables $`p`$’s and by $`_p`$ the Fourier transformation. For two symbols $`f,g_p^1\left(PDO_G\left(𝒪\right)\right)=𝐂_q\left(𝒪\right)`$, their Fourier images $`_p\left(f\right)`$ and $`_p\left(g\right)`$ are operators and we define their oprator product and then take Fourier inverse, as $``$-product
$$fg:=_p^1(_p\left(f\right)._p\left(g\right)).$$
So this product is again a symbol in the same class $`𝐂_q\left(𝒪\right)`$.
Because of existence of special coordinate systems, linear on $`p`$’s coordinates we can treat for the good strata in the same way as in the cases of exponential groups. And the formal series of $``$-product is convergent. The proof could be also done in the same scheme as in exponential or compact cases, see \[AC1\]-\[AC2\].
Let us denote by $`\mathrm{\Gamma }=\pi _1\left(𝒪\right)`$ the fundamental group of the orbit. Our next step is to globally extend this kind of local $``$-products. Our idea is as follows. We lift this $``$-product to the universal covering of coadjoint orbits as some $`\mathrm{\Gamma }`$-invariant $``$-products, globally extend them in virtue of the monodromy theorem and then pushdown to our orbits. We start with the following lemmas
###### Lemma 5.6
There is one-to-one correspondence between $``$-products on Poisson manifolds and $`\mathrm{\Gamma }`$-invariant $``$-products on their the universal coverings.
We use this lemma to describe existence of a $``$-product on coadjoint orbits.
###### Lemma 5.7
On a universal covering, each local $``$-product can be uniquely extended to some global $``$-product on this covering.
For local charts, there exist deformation quantization, as said above, $`fOp\left(f\right)`$ by using the formulas of Fedosov quantization. Also in the intersection of two local charts of coordinates $`(q,p)`$ and $`(\stackrel{~}{q},\stackrel{~}{p})`$, there is a symplectomorphism, namely $`\phi `$ such that
$$(\stackrel{~}{q},\stackrel{~}{p})=\phi (q,p).$$
Using the local oscilatting integrals and by compensating the local Maslov’s index obstacles, one can exactly construct the unitary operator $`U`$ such that
$$Op\left(f\phi \right)=U.Op\left(f\right).U^1,$$
see for example Fedosov’s book \[F\]. Because the universal coverings are simply connected, the extensions can be therefore produced because of Monodromy Theorem.
###### Theorem 5.8
There exists a convergent $``$-product on each orbit, which is a symplectic leaf of the Poisson structure on each stratum of coadjoint orbits.
From the description of the canonical coordinates in the previous section, we see that there exist locally $``$-product. This local $``$-product then extended to a $`\mathrm{\Gamma }`$-invariant global one on the universal covering, which produces a convergent $``$-product on the coadjoint orbits, following Lemmas 5.6,5.7.
### 5.3 Star-product, Quantization and PBW Theorem
We use the constructed in the previous section $``$-product to provide an action of functions on the spaces of partially invariant partially holomorphic sections of the corresponding partially invariant holomorphically induced bundles associated with polarizations. It is possibles because the quantum induced bundles are locally trivial and the spaces of partially invariant partially holomorphic sections with section in Hilbert spaces are locally finitely generated modules over the algebras of quantizing functions, see \[D2\]. We use then the construction of Karabegov and Fedosov to obtain a Hopf \*-algebra of longitudinal pseudo-differential operators elliptic along the leaves of this measurable foliations. The main ingredient is that we use here the Poincaré-Birkhopf-Witt theorem to provide this quantization.
###### Theorem 5.9
There is a natural deformation quantization with separation of variables, corresponding to the Poincaré-Birkhoff-Witt Theorem for polarizations.
We have from the multidimensional quantization $`𝔤PDO^1\left(𝒪\right)`$, $`X\widehat{X}`$. More precisely, Let us denote by $`PDO\left(𝒪\right)`$ the algebra of right $`G`$-invariant pseudo-differential operators on $`𝒪`$, i.e. the continuous maps from $`C^{\mathrm{}}\left(𝒪\right)`$ into itself not extending support. We also denote $`PDO^1\left(𝒪\right)`$ the Lie algebras of right $`G`$-invariant first order pseudo-differential operators. By the procedure of multidimensional quantization, there is a homomorphism of Lie algebras
$$𝔤PDO_G^1\left(𝒪\right)PDO_G\left(𝒪\right).$$
Following the universal property of $`U\left(𝔤\right)`$, there is a unique homomorphism of associative algebras $`U\left(𝔤\right)PDO_G\left(𝒪\right)`$making the following diagram commutative
$$\begin{array}{ccc}𝔤& & PDO_G\left(𝒪\right)\\ & & & & \\ U\left(𝔤\right)& =& U\left(𝔤\right)\end{array}$$
###### Remark 5.10
Because of BKW, $`U(𝔤)U(𝔭_𝒪/𝔥_𝒪)U(𝔥_𝒪)U(\overline{𝔭}_𝒪/𝔥_𝒪)`$ and because of this our quantizing map is coincided with that one used by Karabegov in the quantization with separation of variables in case of coadjoint orbits with totally complex polarizations $`𝔤_𝐂=𝔪_𝐂=𝔭\overline{𝔭}`$.
Let us first describe the machinery applied in the method of Karabegov’s separation of variable. We use the idea about polarizations in multidimensional quantization, \[D2\].
Let us recall the root decomposition
$$𝔤=𝔫_{}𝔞𝔫_+.$$
If $`𝔭`$ is a complex polarization, then from definition we have $`𝔪_𝐂=𝔭\overline{𝔭},`$ $`𝔥_𝐂=𝔭\overline{𝔭},`$ where $`𝔪=𝔤\left(𝔭\overline{𝔭}\right)`$ and $`𝔥=𝔤\left(𝔭\overline{𝔭}\right).`$ The quotients subspaces $`𝔭/𝔥_𝐂`$ and $`\overline{𝔭}/𝔥_𝐂`$ are included in $`𝔪`$ as linear subspaces (not necessarily to be sub-algebras). Let us fix some (non-canonical) inclusions. Let us denote by $`U\left(𝔭/𝔥_𝐂\right)`$ (resp. $`U\left(\overline{𝔭}/𝔥_𝐂\right)`$) the sub-algebra, generated by elements from $`𝔭/𝔥_𝐂𝔪_𝐂`$ (resp. $`𝔭/𝔥_𝐂𝔪_𝐂`$) in the universal algebra $`U\left(𝔪\right)`$. We have therefore a analog of the well-known Poincaré-Birkhoff-Witt theorem.
###### Theorem 5.11 (Poincaré-Birkhoff-Witt Theorem)
If $`𝔭`$ is as polarization, then
$$U\left(𝔪_𝐂\right)U\left(𝔭/𝔥_𝐂\right)U\left(𝔥_𝐂\right)U\left(\overline{𝔭}/𝔥_𝐂\right)$$
Let us fix in a basis each of $`𝔭/𝔥_𝐂`$, $`𝔥_𝐂`$ and $`\overline{𝔭}/𝔥_𝐂`$. We have therefore a basis of $`𝔪_𝐂=𝔭/𝔥_𝐂𝔥_𝐂\overline{𝔭}/𝔥_𝐂`$. Our theorem is therefore deduced from the original Poincaré-Birkhoff-Witt Theorem \[P\].
It is easy to see that by this reason, on the variety $`M/H`$ there is a natural complex structure and therefore on coadjoint orbits exists some partial complex structure. We use this complex structure and this Poincaré-Birkhoff-Witt theorem to do a separation of variables on $`M`$ and apply the machinery of Karabegov.
###### Theorem 5.12
In the case of totally complex polarizable coadjoint orbits, the quantization map from the theorem 5.9 is coincided with the rule of Karabegov’s quantization with separation of variables.
### 5.4 Representations
Let us consider now a continuous fields of $`(\stackrel{~}{\sigma }_𝒪,F_𝒪)`$-polarizations $`(𝔭_𝒪,\rho _𝒪,\sigma _{0,𝒪})`$ satisfying the Pukanszky condition. On one hand side, we can use the multidimensional quantization procedure to obtain irreducible unitary representations $`\mathrm{\Pi }_𝒪=Ind(G;𝔭_𝒪,H_𝒪,\rho _𝒪,\sigma _{0,𝒪})`$ of $`G`$, \[D2\]. On other hand side, we can use $``$-product construction to provide the representations $`\mathrm{\Pi }_𝒪:\stackrel{~}{X|_\mathrm{\Omega }}\mathrm{}_X`$ of the quantum strata $`𝐂_q\left(\mathrm{\Omega }\right)`$. We’d like to show we have the same one.
###### Definition 5.13
Quantum coadjoint orbit $`𝐂_q(𝒪)`$ is defined as the Hopf algebra of symbols of differential operators $`U_{q,𝒪}(𝔤)=U(𝔤)|_𝒪`$. The homomorphism $`Q:U(𝔤)PDO_G(𝒪)`$ is defined to be the second quantization homomorphism.
###### Theorem 5.14
The representation of the Lie algebra obtained from $``$-product is equal to the representations obtained from the multidimensional quantization procedure.
Let us recall \[D2\], that
$$Lie_XInd(G;𝔭_𝒪,H_𝒪,\rho _𝒪,\sigma _{0,𝒪})\widehat{X}.$$
From the construction of quantization map
$$U\left(𝔤\right)PDO_G\left(𝒪\right)$$
as the map arising from the universal property the map $`𝔤PDO_G^1\left(𝒪\right),`$
$$\mathrm{}_X=Op\left(\stackrel{~}{X}\right)=\widehat{X},X𝔤=Lie\left(G\right).$$
The associated representation of $`𝐂_q\left(\mathrm{\Omega }\right)`$ is the solution of the Cauchy problem for the differential equation
$$\frac{}{t}U(t,q,p)=\mathrm{}_XU(t,q,p),$$
$$U(0,q,p)=Id.$$
The solution of this problem is uniquely defined.
### 5.5 Oscillating Fourier integrals
Let us in this section consider the good family of coadjoint orbits arising from the solved cases of Gelfand-Kirillov Conjecture. The results in this section is revised from the \[D3\].
Consider a connected and simply connected Lie group $`G`$ with Lie algebra $`𝔤`$.
###### Theorem 5.15
There exists a $`G`$-invariant Zariski open set $`\mathrm{\Omega }`$ and a covering $`\stackrel{~}{\mathrm{\Omega }}`$ of $`\mathrm{\Omega }`$, with natural action of $`G`$ such that for each continuous field of polarizations $`(𝔭_𝒪,H_𝒪,\rho _𝒪,\sigma _{0,𝒪})`$, $`𝒪\mathrm{\Omega }/G`$, the Lie derivative of the direct integral of representations arized from the multidimensional quantization procedure is equivalent to the tensor product of the Schrödinger representation $`\mathrm{\Pi }=Sch`$ of the Gelfand-Kirillov basis of the enveloping field and a continuous field of trivial representations $`\{V_𝒪\}`$ on $`\stackrel{~}{\mathrm{\Omega }}`$.
The proof is rather long and consists of several steps:
1. We apply the construction of Nghiem \[Ng5\] to the solvable radical $`{}_{}{}^{r}𝔤`$ of $`𝔤`$ to obtain the Zariski open set $`{}_{}{}^{r}\mathrm{\Omega }`$ in $`{}_{}{}^{r}𝔤_{}^{}`$ and its covering $`{}_{}{}^{r}\stackrel{~}{\mathrm{\Omega }}`$.
2. The general case is reduced to the semi-simple case $`{}_{}{}^{s}𝔤`$, see (\[Ng1\], Thm. C). Denote by $`{}_{}{}^{s}G`$ the corresponding analytic subgroup of $`G`$.
3. Take the Zariski open set $`𝒜_s𝒫(^sG)`$ of admissible and well-polarizable strongly regular functionals from $`{}_{}{}^{s}𝔤_{}^{}`$ and its covering $`_s(^sG)`$ via Duflo’s construction \[Du1\].
4. The desired $`G`$-invariant Zariski open set and its covering are the corresponding Cartesian products
$$\mathrm{\Omega }=𝒜_s𝒫(^sG)\times {}_{}{}^{r}\mathrm{\Omega }_{}^{0}\times {}_{}{}^{r}\mathrm{\Omega }_{}^{1}\times \mathrm{}\times {}_{}{}^{r}\mathrm{\Omega }_{}^{k}$$
$$\stackrel{~}{\mathrm{\Omega }}=_s(^sG)\times {}_{}{}^{r}\stackrel{~}{\mathrm{\Omega }^0}\times {}_{}{}^{r}\stackrel{~}{\mathrm{\Omega }^1}\times \mathrm{}\times {}_{}{}^{r}\stackrel{~}{\mathrm{\Omega }}_{}^{k}$$
5.
###### Lemma 5.16
There exists a continuous field of polarizations of type $`(𝔭_𝒪,H_𝒪,\rho _𝒪,\sigma _{0,𝒪})`$, for each $`𝒪\stackrel{~}{\mathrm{\Omega }}/G`$.
6.
###### Lemma 5.17
The Lie derivative commutes with direct integrals, i.e.
$$_{\stackrel{~}{\mathrm{\Omega }}/G}^{}Ind(G;𝔭_𝒪,H_𝒪,\rho _𝒪,\sigma _{0,𝒪})𝑑𝒪=_{\stackrel{~}{\mathrm{\Omega }}/G}^{}Ind(G;𝔭_𝒪,H_𝒪,\rho _𝒪,\sigma _{0,𝒪})𝑑𝒪.$$
7.
###### Lemma 5.18
The restriction of the Schrödinger representation to coadjoint orbits provides a continuous field of polarizations. In particular,
$$D\sigma _{0,𝒪}=\rho _𝒪|_{𝔭_𝒪𝔤}multSch_𝒪.$$
###### Theorem 5.19
1. There exists an operator-valued phase $`\mathrm{\Phi }(t,z,x)`$ and an operator-valued amplitude $`a(t,z,x,y,\xi )`$ extended from the expression
$$\mathrm{exp}\left(\mathrm{\Phi }(t,z,y)+\sqrt{1}\xi \left(g\left(t\right)xx\right)\right)$$
in such a fashion that the action of the representation
$$\mathrm{\Pi }_𝒪=Ind(G;𝔭_𝒪,H_𝒪,\rho _𝒪,\sigma _{0,𝒪})$$
can be expressed as an oscilatting Fourier integral
$$\mathrm{\Pi }_𝒪\left(g\left(t\right)\right)f(z,x)=c._{𝐑^M}_{𝐑^M}a(t,z,x,y,\xi )\mathrm{exp}\left(\sqrt{1}\xi \left(xy\right)\right)f(z,y)𝑑y𝑑\xi ,$$
where $`c`$ is some constant.
2. For each function $`\phi `$ of Schwartz class $`𝒮(G)`$ satisfying the compactness criteria \[D2\] in every induction step \[Li\], \[D2\], the operator $`\mathrm{\Pi }_𝒪(\phi )`$ is of trace class and its action can be expressed as the oscilatting Fourier integral
$$\mathrm{\Pi }_𝒪\left(\phi \right)f(z,x)=const._{𝐑^M}_{𝐑^M}\left(_𝐑^{\mathrm{}}a(t,z,x,y,\xi )\phi \left(g\left(t\right)\right)dt\right)\times $$
$$\times \mathrm{exp}\left(\sqrt{1}\xi \left(xy\right)\right)f(z,y)dyd\xi .$$
Hence, its trace is
$$tr\mathrm{\Pi }_𝒪\left(\phi \right)=const._{𝐑^M}_{𝐑^M}\left(_𝐑^{\mathrm{}}a(t,z,x,x,\xi )\phi \left(g\left(t\right)\right)𝑑t\right)𝑑x𝑑\xi $$
The proof also consists of several steps:
1. From \[D1\] \- \[D2\] and \[Du1\] it is easy to see obtain a slight unipotent modification( i.e. a reduction to the unipotent radical) of the multidimensional quantization procedure. We refer the reader to \[TrV\]-\[TrV2\], and \[TDV\] for a detailed exposition.
2. From the unitary representations $`\mathrm{\Pi }_𝒪=Ind(G;𝔭_𝒪,H_𝒪,\rho _𝒪,\sigma _{0,𝒪})`$ in the unipotent context and its differential $`\pi _𝒪=D\mathrm{\Pi }_𝒪(G;𝔭_𝒪,H_𝒪,\rho _𝒪,\sigma _{0,𝒪})`$, it is easy to select the constant term and the vector fields term, (see \[Ng4\] for the simplest case and notations),
$$\pi \left(L\right)=\pi _c\left(L\right)+\pi _d\left(L\right),\pi _s\left(L\right)=a_0(L,x),$$
$$\pi _d\left(L\right)=\left(L\right)=\underset{k}{}a_k\left(L\right)\frac{}{x_k}.$$
The operator-valued partial (i.e. depending on $`L`$) phase
$$\mathrm{\Phi }_L(t_L,z,x)=_0^{t_L}a_0(z,g\left(s\right)x)𝑑s.$$
The phase $`\mathrm{\Phi }(t,z,x)`$ is the sum of the partial phases, on which the one-parameter group $`g_L\left(s\right)`$ operates by translations for all the factors on the left of $`g_L\left(t_L\right)`$ in the ordered product
$$g\left(t\right)=\underset{L}{}g_L\left(t_L\right).$$
It is easy then to see that our induced representations $`\mathrm{\Pi }_𝒪`$ act as follows
$$\mathrm{\Pi }_𝒪\left(g\left(t\right)\right)f(z,x)=\sigma _{0,𝒪}(t,z,x)f(z,g\left(t\right)x).$$
3. It suffices now to apply the Fourier transforms
$$f(z,x)=\left(2\pi \right)^M_{𝐑^M}_{𝐑^M}\mathrm{exp}\left\{\sqrt{1}\xi \left(xy\right)\right\}f(z,y)𝑑y𝑑\xi $$
to the function $`\mathrm{\Pi }_𝒪\left(g\left(t\right)\right)f(z,x)`$ to obtain
$$\mathrm{\Pi }_𝒪\left(g\left(t\right)\right)f(z,x)=$$
$$\left(2\pi \right)^M_{𝐑^M}_{𝐑^M}a(t,z,x,y,\xi )\mathrm{exp}\left\{\sqrt{1}\xi \left(xy\right)\right\}f(z,y)𝑑y𝑑\xi ,$$
where the amplitude $`a(t,z,x,y,\xi )`$ is the natural extension of the expression
$$\mathrm{exp}\left\{\mathrm{\Phi }(t,z,y)+\sqrt{1}\xi \left(g\left(t\right)xx\right)\right\}$$
in the correspondence with the fields of polarizations. Hence for each $`\phi 𝒮\left(G\right)`$,
$$\mathrm{\Pi }_𝒪\left(\phi \right)f(z,x)=\left(2\pi \right)^M_{𝐑^M}_{𝐑^M}\left(_𝐑^{\mathrm{}}a(t,z,x,y\xi )\phi \left(g\left(t\right)\right)dt\right)\times $$
$$\times \mathrm{exp}\left\{\sqrt{1}\xi \left(xy\right)\right\}f(z,y)dyd\xi .$$
Remark that the integral $`_𝐑^{\mathrm{}}a(t,z,x,y,\xi )\phi \left(g\left(t\right)\right)𝑑t`$ is just a type of Feynman path integrals.
4.
###### Lemma 5.20
If in every repeated induction step, see \[Li\], \[D2\], $`\mathrm{\Pi }_𝒪(\phi )`$ satisfies the compactness criteria, then the operator $`\mathrm{\Pi }_𝒪(\phi )`$ is trace class and hence
$$tr\mathrm{\Pi }_𝒪\left(\phi \right)=\left(2\pi \right)^M_{𝐑^M}_{𝐑^M}tr(_𝐑^{\mathrm{}}a(t,z,x,x,\xi )\phi \left(g\left(t\right)\right)dtdxd\xi .$$
## 6 Examples
In this section, we illustrate the main ideas in examples.
### 6.1 $`\overline{MD}`$-groups
We expose in this subsection our joint works with Nguyen Viet Hai \[DH1\]-\[DH2\].
#### 6.1.1 The group of affine transformations of the real straight line
We refer the reader to the work \[DH1\] for a detailed exposition with complete proof.
Canonical coordinates on the upper half-planes. Recall that the Lie algebra $`𝔤=aff\left(𝐑\right)`$ of affine transformations of the real straight line is described as follows, see for example \[D2\]: The Lie group $`Aff\left(𝐑\right)`$ of affine transformations of type
$$x𝐑ax+b,\text{ for some parameters }a,b𝐑,a0.$$
It is well-known that this group $`Aff\left(𝐑\right)`$ is a two dimensional Lie group which is isomorphic to the group of matrices
$$Aff\left(𝐑\right)\left\{\left(\begin{array}{cc}a& b\\ 0& 1\end{array}\right)\right|a,b𝐑,a0\}.$$
We consider its connected component
$$G=Aff_0\left(𝐑\right)=\left\{\left(\begin{array}{cc}a& b\\ 0& 1\end{array}\right)\right|a,b𝐑,a>0\}$$
of identity element. Its Lie algebra is
$$𝔤=aff\left(𝐑\right)\left\{\left(\begin{array}{cc}\alpha & \beta \\ 0& 0\end{array}\right)\right|\alpha ,\beta 𝐑\}$$
admits a basis of two generators $`X,Y`$ with the only nonzero Lie bracket $`[X,Y]=Y`$, i.e.
$$𝔤=aff\left(𝐑\right)\left\{\alpha X+\beta Y\right|[X,Y]=Y,\alpha ,\beta 𝐑\}.$$
The co-adjoint action of $`G`$ on $`𝔤^{}`$ is given (see e.g. \[AC2\], \[Ki1\]) by
$$K\left(g\right)F,Z=F,Ad\left(g^1\right)Z,F𝔤^{},gG\text{ and }Z𝔤.$$
Denote the co-adjoint orbit of $`G`$ in $`𝔤`$, passing through $`F`$ by
$$\mathrm{\Omega }_F=K\left(G\right)F:=\left\{K\left(g\right)F\right|FG\}.$$
Because the group $`G=Aff_0\left(𝐑\right)`$ is exponential (see \[D2\]), for $`F𝔤^{}=aff\left(𝐑\right)^{}`$, we have
$$\mathrm{\Omega }_F=\{K\left(\mathrm{exp}\left(U\right)F\right|Uaff\left(𝐑\right)\}.$$
It is easy to see that
$$K\left(\mathrm{exp}U\right)F,Z=F,\mathrm{exp}\left(ad_U\right)Z.$$
It is easy therefore to see that
$$K\left(\mathrm{exp}U\right)F=F,\mathrm{exp}\left(ad_U\right)XX^{}+F,\mathrm{exp}\left(ad_U\right)YY^{}.$$
For a general element $`U=\alpha X+\beta Y𝔤`$, we have
$$\mathrm{exp}\left(ad_U\right)=\underset{n=0}{\overset{\mathrm{}}{}}\frac{1}{n!}\left(\begin{array}{cc}0& 0\\ \beta & \alpha \end{array}\right)^n=\left(\begin{array}{cc}1& 0\\ L& e^\alpha \end{array}\right),$$
where $`L=\alpha +\beta +\frac{\alpha }{\beta }\left(1e^\beta \right)`$. This means that
$$K\left(\mathrm{exp}U\right)F=\left(\lambda +\mu L\right)X^{}+\left(\mu e^\alpha \right)Y^{}.$$
From this formula one deduces \[D2\] the following description of all co-adjoint orbits of $`G`$ in $`𝔤^{}`$:
* If $`\mu =0`$, each point $`\left(x=\lambda ,y=0\right)`$ on the abscissa ordinate corresponds to a 0-dimensional co-adjoint orbit
$$\mathrm{\Omega }_\lambda =\left\{\lambda X^{}\right\},\lambda 𝐑.$$
* For $`\mu 0`$, there are two 2-dimensional co-adjoint orbits: the upper half-plane $`\left\{(\lambda ,\mu )\right|\lambda ,\mu 𝐑,\mu >0\}`$ corresponds to the co-adjoint orbit
$$\mathrm{\Omega }_+:=\{F=(\lambda +\mu L)X^{}+\left(\mu e^\alpha \right)Y^{}|\mu >0\},$$
(1)
and the lower half-plane $`\left\{(\lambda ,\mu )\right|\lambda ,\mu 𝐑,\mu <0\}`$ corresponds to the co-adjoint orbit
$$\mathrm{\Omega }_{}:=\{F=(\lambda +\mu L)X^{}+\left(\mu e^\alpha \right)Y^{}|\mu <0\}.$$
(2)
We shall work from now on for the fixed co-adjoint orbit $`\mathrm{\Omega }_+`$. The case of the co-adjoint orbit $`\mathrm{\Omega }_{}`$ is similarly treated. First we study the geometry of this orbit and introduce some canonical coordinates in it. It is well-known from the orbit method \[Ki1\] that the Lie algebra $`𝔤=aff\left(𝐑\right)`$, realized by the complete right-invariant Hamiltonian vector fields on co-adjoint orbits $`\mathrm{\Omega }_FG_FG`$ with flat (co-adjoint) action of the Lie group $`G=Aff_0\left(𝐑\right)`$. On the orbit $`\mathrm{\Omega }_+`$ we choose a fix point $`F=Y^{}`$. It is well-known from the orbit method that we can choose an arbitrary point $`F`$ on $`\mathrm{\Omega }_F`$. It is easy to see that the stabilizer of this (and therefore of any) point is trivial $`G_F=\left\{e\right\}`$. We identify therefore $`G`$ with $`G_Y^{}G`$. There is a natural diffeomorphism $`Id_𝐑\times \mathrm{exp}(.)`$ from the standard symplectic space $`𝐑^2`$ with symplectic 2-form $`dpdq`$ in canonical Darboux $`(p,q)`$-coordinates, onto the upper half-plane $`𝐇_+𝐑𝐑_+`$ with coordinates $`(p,e^q)`$, which is, from the above coordinate description, also diffeomorphic to the co-adjoint orbit $`\mathrm{\Omega }_+`$. We can use therefore $`(p,q)`$ as the standard canonical Darboux coordinates in $`\mathrm{\Omega }_Y^{}`$. There are also non-canonical Darboux coordinates $`(x,y)=(p,e^q)`$ on $`\mathrm{\Omega }_Y^{}`$. We show now that in these coordinates $`(x,y)`$, the Kirillov form looks like $`\omega _Y^{}(x,y)=\frac{1}{y}dxdy`$, but in the canonical Darboux coordinates $`(p,q)`$, the Kirillov form is just the standard symplectic form $`dpdq`$. This means that there are symplectomorphisms between the standard symplectic space $`𝐑^2,dpdq)`$, the upper half-plane $`(𝐇_+,\frac{1}{y}dxdy)`$ and the co-adjoint orbit $`(\mathrm{\Omega }_Y^{},\omega _Y^{})`$. Each element $`Z𝔤`$ can be considered as a linear functional $`\stackrel{~}{Z}`$ on co-adjoint orbits, as subsets of $`𝔤^{}`$, $`\stackrel{~}{Z}\left(F\right):=F,Z`$. It is well-known that this linear function is just the Hamiltonian function associated with the Hamiltonian vector field $`\xi _Z`$, which represents $`Z𝔤`$ following the formula
$$\left(\xi _Zf\right)\left(x\right):=\frac{d}{dt}f\left(x\mathrm{exp}\left(tZ\right)\right)|_{t=0},fC^{\mathrm{}}\left(\mathrm{\Omega }_+\right).$$
The Kirillov form $`\omega _F`$ is defined by the formula
$$\omega _F(\xi _Z,\xi _T)=F,[Z,T],Z,T𝔤=aff\left(𝐑\right).$$
(3)
This form defines the symplectic structure and the Poisson brackets on the co-adjoint orbit $`\mathrm{\Omega }_+`$. For the derivative along the direction $`\xi _Z`$ and the Poisson bracket we have relation $`\xi _Z\left(f\right)=\{\stackrel{~}{Z},f\},fC^{\mathrm{}}\left(\mathrm{\Omega }_+\right)`$. It is well-known in differential geometry that the correspondence $`Z\xi _Z,Z𝔤`$ defines a representation of our Lie algebra by vector fields on co-adjoint orbits. If the action of $`G`$ on $`\mathrm{\Omega }_+`$ is flat \[D2\], we have the second Lie algebra homomorphism from strictly Hamiltonian right-invariant vector fields into the Lie algebra of smooth functions on the orbit with respect to the associated Poisson brackets.
Denote by $`\psi `$ the indicated symplectomorphism from $`𝐑^2`$ onto $`\mathrm{\Omega }_+`$
$$(p,q)𝐑^2\psi (p,q):=(p,e^q)\mathrm{\Omega }_+$$
###### Proposition 6.1
1. Hamiltonian function $`f_Z=\stackrel{~}{Z}`$ in canonical coordinates $`(p,q)`$ of the orbit $`\mathrm{\Omega }_+`$ is of the form
$$\stackrel{~}{Z}\psi (p,q)=\alpha p+\beta e^q,\text{ if }Z=\left(\begin{array}{cc}\alpha & \beta \\ 0& 0\end{array}\right).$$
2. In the canonical coordinates $`(p,q)`$ of the orbit $`\mathrm{\Omega }_+`$, the Kirillov form $`\omega _Y^{}`$ is just the standard form $`\omega =dpdq`$.
Computation of generators $`\widehat{\mathrm{}}_Z`$ Let us denote by $`\mathrm{\Lambda }`$ the 2-tensor associated with the Kirillov standard form $`\omega =dpdq`$ in canonical Darboux coordinates. We use also the multi-index notation. Let us consider the well-known Moyal $``$-product of two smooth functions $`u,vC^{\mathrm{}}\left(𝐑^2\right)`$, defined by
$$uv=u.v+\underset{r1}{}\frac{1}{r!}\left(\frac{1}{2i}\right)^rP^r(u,v),$$
where
$$P^r(u,v):=\mathrm{\Lambda }^{i_1j_1}\mathrm{\Lambda }^{i_2j_2}\mathrm{}\mathrm{\Lambda }^{i_rj_r}_{i_1i_2\mathrm{}i_r}u_{j_1j_2\mathrm{}j_r}v,$$
with
$$_{i_1i_2\mathrm{}i_r}:=\frac{^r}{x^{i_1}\mathrm{}x^{i_r}},x:=(p,q)=(p_1,\mathrm{},p_n,q^1,\mathrm{},q^n)$$
as multi-index notation. It is well-known that this series converges in the Schwartz distribution spaces $`𝒮\left(𝐑^n\right)`$. We apply this to the special case $`n=1`$. In our case we have only $`x=(x^1,x^2)=(p,q)`$.
###### Proposition 6.2
In the above mentioned canonical Darboux coordinates $`(p,q)`$ on the orbit $`\mathrm{\Omega }_+`$, the Moyl $``$-product satisfies the relation
$$i\stackrel{~}{Z}i\stackrel{~}{T}i\stackrel{~}{T}i\stackrel{~}{Z}=i\stackrel{~}{[Z,T]},Z,Taff\left(𝐑\right).$$
Consequently, to each adapted chart $`\psi `$ in the sense of \[AC2\], we associate a $`G`$-covariant $``$-product.
###### Proposition 6.3 (see \[G\])
Let $``$ be a formal differentiable $``$-product on $`C^{\mathrm{}}(M,𝐑)`$, which is covariant under $`G`$. Then there exists a representation $`\tau `$ of $`G`$ in $`AutN[[\nu ]]`$ such that
$$\tau \left(g\right)\left(uv\right)=\tau \left(g\right)u\tau \left(g\right)v.$$
Let us denote by $`_pu`$ the partial Fourier transform of the function $`u`$ from the variable $`p`$ to the variable $`x`$, i.e.
$$_p\left(u\right)(x,q):=\frac{1}{\sqrt{2\pi }}_𝐑e^{ipx}u(p,q)𝑑p.$$
Let us denote by $`_p^1\left(u\right)(x,q)`$ the inverse Fourier transform.
###### Lemma 6.4
1. $`_p_p^1(p.u)=i_p^1(x.u)`$ ,
2. $`_p(v)=i_x_p(v)`$ ,
3. $`P^k(\stackrel{~}{Z},_p^1(u))=(1)^k\beta e^q\frac{^k_p^1(u)}{^kp},\text{ with }k2.`$
For each $`Zaff\left(𝐑\right)`$, the corresponding Hamiltonian function is $`\stackrel{~}{Z}=\alpha p+\beta e^q`$ and we can consider the operator $`\mathrm{}_Z`$ acting on dense subspace $`L^2(𝐑^2,\frac{dpdq}{2\pi })^{\mathrm{}}`$ of smooth functions by left $``$-multiplication by $`i\stackrel{~}{Z}`$, i.e. $`\mathrm{}_Z\left(u\right)=i\stackrel{~}{Z}u`$. It is then continuated to the whole space $`L^2(𝐑^2,\frac{dpdq}{2\pi })`$. It is easy to see that, because of the relation in Proposition (6.2), the correspondence $`Zaff\left(𝐑\right)\mathrm{}_Z=i\stackrel{~}{Z}.`$ is a representation of the Lie algebra $`aff\left(𝐑\right)`$ on the space $`N\left[\left[\frac{i}{2}\right]\right]`$ of formal power series in the parameter $`\nu =\frac{i}{2}`$ with coefficients in $`N=C^{\mathrm{}}(M,𝐑)`$, see e.g. \[G\] for more detail.
We study now the convergence of the formal power series. In order to do this, we look at the $``$-product of $`i\stackrel{~}{Z}`$ as the $``$-product of symbols and define the differential operators corresponding to $`i\stackrel{~}{Z}`$. It is easy to see that the resulting correspondence is a representation of $`𝔤`$ by pseudo-differential operators.
###### Proposition 6.5
For each $`Zaff(𝐑)`$ and for each compactly supported $`C^{\mathrm{}}`$ function $`uC_0^{\mathrm{}}(𝐑^2)`$, we have
$$\widehat{\mathrm{}}_Z\left(u\right):=_p\mathrm{}_Z_p^1\left(u\right)=\alpha \left(\frac{1}{2}_q_x\right)u+i\beta e^{q\frac{x}{2}}u.$$
###### Remark 6.6
Setting new variables $`s=q\frac{x}{2}`$, $`t=q+\frac{x}{2}`$, we have
$$\widehat{\mathrm{}}_Z\left(u\right)=\alpha \frac{u}{s}+i\beta e^su,$$
(4)
e.i.
$$\widehat{\mathrm{}}_Z=\alpha \frac{}{s}+i\beta e^s,$$
which provides a representation of the Lie algebra $`aff\left(𝐑\right)`$.
The associate irreducible unitary representations
Our aim in this section is to exponentiate the obtained representation $`\widehat{\mathrm{}}_Z`$ of the Lie algebra $`aff\left(𝐑\right)`$ to the corresponding representation of the Lie group $`Aff_0\left(𝐑\right)`$. We shall prove that the result is exactly the irreducible unitary representation $`T_{\mathrm{\Omega }_+}`$ obtained from the orbit method or Mackey small subgroup method applied to this group $`Aff\left(𝐑\right)`$. Let us recall first the well-known list of all the irreducible unitary representations of the group of affine transformation of the real straight line.
###### Theorem 6.7 (\[GN\])
Every irreducible unitary representation of the group $`Aff(𝐑)`$ of all the affine transformations of the real straight line, up to unitary equivalence, is equivalent to one of the pairwise nonequivalent representations:
* the infinite dimensional representation $`S`$, realized in the space $`L^2(𝐑^{},\frac{dy}{\left|y\right|})`$, where $`𝐑^{}=𝐑\left\{0\right\}`$ and is defined by the formula
$$\left(S\left(g\right)f\right)\left(y\right):=e^{iby}f\left(ay\right),\text{ where }g=\left(\begin{array}{cc}a& b\\ 0& 1\end{array}\right),$$
* the representation $`U_\lambda ^\epsilon `$, where $`\epsilon =0,1`$, $`\lambda 𝐑`$, realized in the 1-dimensional Hilbert space $`𝐂^1`$ and is given by the formula
$$U_\lambda ^\epsilon \left(g\right)=\left|a\right|^{i\lambda }\left(sgna\right)^\epsilon .$$
Let us consider now the connected component $`G=Aff_0\left(𝐑\right)`$. The irreducible unitary representations can be obtained easily from the orbit method machinery.
###### Theorem 6.8
The representation $`\mathrm{exp}(\widehat{\mathrm{}}_Z)`$ of the group $`G=Aff_0(𝐑)`$ is exactly the irreducible unitary representation $`T_{\mathrm{\Omega }_+}`$ of $`G=Aff_0(𝐑)`$ associated following the orbit method construction, to the orbit $`\mathrm{\Omega }_+`$, which is the upper half-plane $`𝐇𝐑𝐑^{}`$, i. e. +
$$\left(\mathrm{exp}\left(\widehat{\mathrm{}}_Z\right)f\right)\left(y\right)=\left(T_{\mathrm{\Omega }_+}\left(g\right)f\right)\left(y\right)=e^{iby}f\left(ay\right),fL^2(𝐑^{},\frac{dy}{\left|y\right|}),$$
where $`g=\mathrm{exp}Z=\left(\begin{array}{cc}a& b\\ 0& 1\end{array}\right).`$
By analogy, we have also
###### Theorem 6.9
The representation $`\mathrm{exp}(\widehat{\mathrm{}}_Z)`$ of the group $`G=Aff_0(𝐑)`$ is exactly the irreducible unitary representation $`T_\mathrm{\Omega }_{}`$ of $`G=Aff_0(𝐑)`$ associated following the orbit method construction, to the orbit $`\mathrm{\Omega }_{}`$, which is the lower half-plane $`𝐇𝐑𝐑^{}`$, i. e.
$$\left(\mathrm{exp}\left(\widehat{\mathrm{}}_Z\right)f\right)\left(y\right)=\left(T_\mathrm{\Omega }_{}\left(g\right)f\right)\left(y\right)=e^{iby}f\left(ay\right),fL^2(𝐑^{},\frac{dy}{\left|y\right|}),$$
where $`g=\mathrm{exp}Z=\left(\begin{array}{cc}a& b\\ 0& 1\end{array}\right).`$
###### Remark 6.10
1. We have demonstrated how all the irreducible unitary representation of the connected group of affine transformations could be obtained from deformation quantization. It is reasonable to refer to the algebras of functions on co-adjoint orbits with this $``$-product as quantum ones.
2. In a forthcoming work, we shall do the same calculation for the group of affine transformations of the complex straight line $`𝐂`$. This achieves the description of quantum $`\overline{MD}`$ co-adjoint orbits, see \[D2\] for definition of $`\overline{MD}`$ Lie algebras.
#### 6.1.2 The group of affine transformations of the complex straight line
Recall that the Lie algebra $`𝔤=aff\left(𝐂\right)`$ of affine transformations of the complex straight line is described as follows, see \[D\].
It is well-known that the group $`Aff\left(𝐂\right)`$ is a four (real) dimensional Lie group which is isomorphism to the group of matrices:
$$Aff\left(𝐂\right)\left\{\left(\begin{array}{cc}a& b\\ 0& 1\end{array}\right)\right|a,b𝐂,a0\}$$
The most easy method is to consider $`X`$,$`Y`$ as complex generators, $`X=X_1+iX_2`$ and $`Y=Y_1+iY_2`$. Then from the relation $`[X,Y]=Y`$, we get$`[X_1,Y_1][X_2,Y_2]+i\left(\left[X_1Y_2\right]+[X_2,Y_1]\right)=Y_1+iY_2`$. This mean that the Lie algebra $`aff\left(𝐂\right)`$ is a real 4-dimensional Lie algebra, having 4 generators with the only nonzero Lie brackets: $`[X_1,Y_1][X_2,Y_2]=Y_1`$; $`[X_2,Y_1]+[X_1,Y_2]=Y_2`$ and we can choose another basic noted again by the same letters to have more clear Lie brackets of this Lie algebra:
$$[X_1,Y_1]=Y_1;[X_1,Y_2]=Y_2;[X_2,Y_1]=Y_2;[X_2,Y_2]=Y_1$$
###### Remark 6.11
The exponential map
$$\mathrm{exp}:𝐂𝐂^{}:=𝐂\backslash \left\{0\right\}$$
giving by $`ze^z`$ is just the covering map and therefore the universal covering of $`C^{}`$ is $`\stackrel{~}{𝐂}^{}𝐂`$. As a consequence one deduces that
$$\stackrel{~}{Aff}\left(𝐂\right)𝐂𝐂\left\{(z,w)\right|z,w𝐂\}$$
with the following multiplication law:
$$(z,w)(z^{^{}},w^{^{}}):=(z+z^{},w+e^zw^{})$$
###### Remark 6.12
The co-adjoint orbit of $`\stackrel{~}{Aff}\left(𝐂\right)`$ in $`𝔤^{}`$ passing through $`F𝔤^{}`$ is denoted by
$$\mathrm{\Omega }_F:=K\left(\stackrel{~}{Aff}\left(𝐂\right)\right)F=\left\{K\left(g\right)F\right|g\stackrel{~}{Aff}\left(𝐂\right)\}$$
Then, (see \[D\]):
1. Each point $`(\alpha ,0,0,\delta )`$ is 0-dimensional co-adjoint orbit $`\mathrm{\Omega }_{(\alpha ,0,0,\delta )}`$
2. The open set $`\beta ^2+\gamma ^2`$ 0 is the single 4-dimensional co-adjoint orbit $`\mathrm{\Omega }_F=\mathrm{\Omega }_{\beta ^2+\gamma ^20}`$. We shall also use $`\mathrm{\Omega }_F`$ in form $`\mathrm{\Omega }_F𝐂\times 𝐂^{}`$.
###### Remark 6.13
Let us denote:
$$𝐇_k=\left\{w=q_1+iq_2𝐂\right|\mathrm{}<q_1<+\mathrm{};2k\pi <q_2<2k\pi +2\pi \};k=0,\pm 1,\mathrm{}$$
$$L=\left\{\rho e^{i\phi }𝐂\right|0<\rho <+\mathrm{};\phi =0\}\text{ and }𝐂_k=𝐂\backslash L$$
is a univalent sheet of the Riemann surface of the complex variable multi-valued analytic function $`Ln\left(w\right)`$, ($`k=0,\pm 1,\mathrm{}`$) Then there is a natural diffeomorphism $`w𝐇_ke^w𝐂_k`$ with each $`k=0,\pm 1,\mathrm{}.`$ Now consider the map:
$$𝐂\times 𝐂\mathrm{\Omega }_F=𝐂\times 𝐂^{}$$
$$(z,w)(z,e^w),$$
with a fixed $`k𝐙`$. We have a local diffeomorphism
$$\phi _k:𝐂\times 𝐇_k𝐂\times 𝐂_k$$
$$(z,w)(z,e^w)$$
This diffeomorphism $`\phi _k`$ will be needed in the all sequel.
On $`𝐂\times 𝐇_k`$ we have the natural symplectic form
$$\omega =\frac{1}{2}\left[dzdw+d\overline{z}d\overline{w}\right],$$
(5)
induced from $`𝐂^2`$. Put $`z=p_1+ip_2,w=q_1+iq_2`$ and $`(x^1,x^2,x^3,x^4)=(p_1,q_1,p_2,q_2)𝐑^4`$, then
$$\omega =dp_1dq_1dp_2dq_2.$$
The corresponding symplectic matrix of $`\omega `$ is
$$=\left(\begin{array}{cccc}0& 1& 0& 0\\ 1& 0& 0& 0\\ 0& 0& 0& 1\\ 0& 0& 1& 0\end{array}\right)\text{ and }^1=\left(\begin{array}{cccc}0& 1& 0& 0\\ 1& 0& 0& 0\\ 0& 0& 0& 1\\ 0& 0& 1& 0\end{array}\right)$$
We have therefore the Poisson brackets of functions as follows. With each $`f,g𝐂^{\mathrm{}}\left(\mathrm{\Omega }\right)`$
$$\{f,g\}=^{ij}\frac{f}{x^i}\frac{g}{x^j}=^{12}\frac{f}{p_1}\frac{g}{q_1}+^{21}\frac{f}{q_1}\frac{g}{p_1}+^{34}\frac{f}{p_2}\frac{g}{q_2}+^{43}\frac{f}{q_2}\frac{g}{p_2}=$$
$$=\frac{f}{p_1}\frac{g}{q_1}\frac{f}{q_1}\frac{g}{p_1}\frac{f}{p_2}\frac{g}{q_2}+\frac{f}{q_2}\frac{g}{p_2}=$$
$$=2\left[\frac{f}{z}\frac{g}{w}\frac{f}{w}\frac{g}{z}+\frac{f}{\overline{z}}\frac{g}{\overline{w}}\frac{f}{\overline{w}}\frac{g}{\overline{z}}\right]$$
###### Proposition 6.14
Fixing the local diffeomorphism $`\phi _k(k𝐙)`$, we have:
1. For any element $`Aaff\left(𝐂\right)`$, the corresponding Hamiltonian function $`\stackrel{~}{A}`$ in local coordinates $`(z,w)`$ of the orbit $`\mathrm{\Omega }_F`$ is of the form
$$\stackrel{~}{A}\phi _k(z,w)=\frac{1}{2}\left[\alpha z+\beta e^w+\overline{\alpha }\overline{z}+\overline{\beta }e^{\overline{w}}\right]$$
2. In local coordinates $`(z,w)`$ of the orbit $`\mathrm{\Omega }_F`$, the symplectic Kirillov form $`\omega _F`$ is just the standard form (1).
Computation of Operators $`\widehat{\mathrm{}}_A^{(k)}`$.
###### Proposition 6.15
With $`A,Baff(𝐂)`$, the Moyal $``$-product satisfies the relation:
$$i\stackrel{~}{A}i\stackrel{~}{B}i\stackrel{~}{B}i\stackrel{~}{A}=i\left[\stackrel{~}{A,B}\right]$$
(6)
For each $`A\text{aff}(𝐂`$), the corresponding Hamiltonian function is
$$\stackrel{~}{A}=\frac{1}{2}\left[\alpha z+\beta e^w+\overline{\alpha }\overline{z}+\overline{\beta }e^{\overline{w}}\right]$$
and we can consider the operator $`\mathrm{}_A^{\left(k\right)}`$ acting on dense subspace $`L^2(𝐑^2\times \left(𝐑^2\right)^{},\frac{dp_1dq_1dp_2dq_2}{\left(2\pi \right)^2})^{\mathrm{}}`$ of smooth functions by left $``$-multiplication by $`i\stackrel{~}{A}`$, i.e: $`\mathrm{}_A^{\left(k\right)}\left(f\right)=i\stackrel{~}{A}f`$. Because of the relation in Proposition 3.1, we have
###### Corollary 6.16
$$\mathrm{}_{[A,B]}^{\left(k\right)}=\mathrm{}_A^{\left(k\right)}\mathrm{}_B^{\left(k\right)}\mathrm{}_B^{\left(k\right)}\mathrm{}_A^{\left(k\right)}:=[\mathrm{}_A^{\left(k\right)},\mathrm{}_B^{\left(k\right)}]^{}$$
(7)
From this it is easy to see that, the correspondence $`Aaff\left(𝐂\right)\mathrm{}_A^{\left(k\right)}=`$i$`\stackrel{~}{A}`$. is a representation of the Lie algebra $`aff(𝐂`$) on the space N$`\left[\left[\frac{i}{2}\right]\right]`$ of formal power series, see \[G\] for more detail.
Now, let us denote $`_z`$(f) the partial Fourier transform of the function f from the variable $`z=p_1+ip_2`$ to the variable $`\xi =\xi _1+i\xi _2`$, i.e:
$$_z\left(f\right)(\xi ,w)=\frac{1}{2\pi }_{R^2}e^{iRe\left(\xi \overline{z}\right)}f(z,w)𝑑p_1𝑑p_2$$
Let us denote by
$$_z^1\left(f\right)(\xi ,w)=\frac{1}{2\pi }_{R^2}e^{iRe\left(\xi \overline{z}\right)}f(\xi ,w)𝑑\xi _1𝑑\xi _2$$
the inverse Fourier transform.
###### Lemma 6.17
Putting $`g=g(z,w)=_z^1(f)(z,w)`$ we have:
1. $$_zg=\frac{i}{2}\overline{\xi }g;_z^rg=\left(\frac{i}{2}\overline{\xi }\right)^rg,r=2,3,$$
2. $$_{\overline{z}}g=\frac{i}{2}\xi g;_{\overline{z}}^rg=\left(\frac{i}{2}\xi \right)^rg,r=2,3,$$
3. $$_z\left(zg\right)=2i_{\overline{\xi }}_z\left(g\right)=2i_{\overline{\xi }}f;_z\left(\overline{z}g\right)=2i_\xi _z\left(g\right)=2i_\xi f$$
4. $$_wg=_w\left(_z^1\left(f\right)\right)=_{z}^{}{}_{}{}^{1}\left(_wf\right);_{\overline{w}}g=_{\overline{w}}(_z^1\left(f\right)=_{z}^{}{}_{}{}^{1}\left(_{\overline{w}}f\right)$$
We also need another Lemma which will be used in the sequel.
###### Lemma 6.18
With $`g=_z^1`$$`(f)(`$$`z,w)`$, we have:
1. $$_z\left(P^0(\stackrel{~}{A},g)\right)=i\left(\alpha _{\overline{\xi }}+\overline{\alpha }_\xi \right)f+\frac{1}{2}\beta e^wf+\frac{1}{2}\overline{\beta }e^{\overline{w}}f.$$
2. $$_z\left(P^1(\stackrel{~}{A},g)\right)=\overline{\alpha }_{\overline{w}}f+\alpha _wf\overline{\beta }e^{\overline{w}}\left(\frac{i}{2}\xi \right)f\beta e^w\left(\frac{i}{2}\overline{\xi }\right)f.$$
3. $$_z\left(P^r(\stackrel{~}{A},g)\right)=\left(1\right)^r.2^{r1}\left[\overline{\beta }e^{\overline{w}}\left(\frac{i}{2}\xi \right)^r+\beta e^w\left(\frac{i}{2}\overline{\xi }\right)^r\right]fr2.$$
###### Proposition 6.19
For each $`A=\left(\begin{array}{cc}\alpha & \beta \\ 0& 0\end{array}\right)aff(𝐂)`$ and for each compactly supported $`C^{\mathrm{}}`$-function $`fC_0^{\mathrm{}}(𝐂\times 𝐇_k)`$, we have:
$$\mathrm{}_A^{\left(k\right)}f:=_z\mathrm{}_A^{\left(k\right)}_z^1\left(f\right)=[\alpha (\frac{1}{2}_w_{\overline{\xi }})f+\overline{\alpha }(\frac{1}{2}_{\overline{w}}_\xi )f+$$
(8)
$$+\frac{i}{2}(\beta e^{w\frac{1}{2}\overline{\xi }}+\overline{\beta }e^{\overline{w}\frac{1}{2}\xi })f]$$
###### Remark 6.20
Setting new variables u = $`w\frac{1}{2}\overline{\xi }`$;$`v=w+\frac{1}{2}\overline{\xi }`$ we have
$$\widehat{\mathrm{}}_A^{\left(k\right)}\left(f\right)=\alpha \frac{f}{u}+\overline{\alpha }\frac{f}{\overline{u}}+\frac{i}{2}\left(\beta e^u+\overline{\beta }e^{\overline{u}}\right)f|_{(u,v)}$$
(9)
i.e $`\widehat{\mathrm{}}_A^{\left(k\right)}=\alpha \frac{}{u}+\overline{\alpha }\frac{}{\overline{u}}+\frac{i}{2}\left(\beta e^u+\overline{\beta }e^{\overline{u}}\right)`$,which provides a ( local) representation of the Lie algebra aff(C).
The Irreducible Representations of $`\stackrel{~}{Aff}(𝐂)`$. Since $`\widehat{\mathrm{}}_A^{\left(k\right)}`$ is a representation of the Lie algebra $`\stackrel{~}{\text{Aff}}\left(𝐂\right)`$, we have:
$$\mathrm{exp}\left(\widehat{\mathrm{}}_A^{\left(k\right)}\right)=\mathrm{exp}\left(\alpha \frac{}{u}+\overline{\alpha }\frac{}{\overline{u}}+\frac{i}{2}\left(\beta e^u+\overline{\beta }e^{\overline{u}}\right)\right)$$
is just the corresponding representation of the corresponding connected and simply connected Lie group $`\stackrel{~}{Aff}\left(𝐂\right)`$.
Let us first recall the well-known list of all the irreducible unitary representations of the group of affine transformation of the complex straight line, see \[D\] for more details.
###### Theorem 6.21
Up to unitary equivalence, every irreducible unitary representation of $`\stackrel{~}{\text{Aff}}(𝐂)`$ is unitarily equivalent to one the following one-to-another nonequivalent irreducible unitary representations:
1. The unitary characters of the group, i.e the one dimensional unitary representation $`U_\lambda ,\lambda 𝐂`$, acting in $`𝐂`$ following the formula $`U_\lambda (z,w)=e^{i\mathrm{}\left(z\overline{\lambda }\right)},(z,w)\stackrel{~}{Aff}\left(𝐂\right),\lambda 𝐂.`$
2. The infinite dimensional irreducible representations $`T_\theta ,\theta 𝐒^1`$, acting on the Hilbert space $`L^2\left(𝐑\times 𝐒^1\right)`$ following the formula:
$$\left[T_\theta \right(z,w\left)f\right]\left(x\right)=\mathrm{exp}\left(i\right(\mathrm{}\left(wx\right)+2\pi \theta \left[\frac{\mathrm{}\left(x+z\right)}{2\pi }\right]\left)\right)f(xz),$$
(10)
Where $`(z,w)\stackrel{~}{Aff}\left(𝐂\right)`$ ; $`x𝐑\times 𝐒^1=𝐂\backslash \left\{0\right\};fL^2\left(𝐑\times 𝐒^1\right);`$
$$xz=Re\left(x+z\right)+2\pi i\left\{\frac{\mathrm{}\left(x+z\right)}{2\pi }\right\}$$
In this section we will prove the following important Theorem which is very interesting for us both in theory and practice.
###### Theorem 6.22
The representation $`\mathrm{exp}(\widehat{\mathrm{}}_A^{(k)})`$ of the group $`\stackrel{~}{Aff}(𝐂)`$ is the irreducible unitary representation $`T_\theta `$ of $`\stackrel{~}{Aff}(𝐂)`$ associated, following the orbit method construction, to the orbit $`\mathrm{\Omega }`$, i.e:
$$\mathrm{exp}\left(\widehat{\mathrm{}}_A^{\left(k\right)}\right)f\left(x\right)=\left[T_\theta \left(\mathrm{exp}A\right)f\right]\left(x\right),$$
where $`fL^2(𝐑\times 𝐒^1);A=\left(\begin{array}{cc}\alpha & \beta \\ 0& 0\end{array}\right)aff(𝐂);\theta 𝐒^1;k=0,\pm 1,`$
###### Remark 6.23
We say that a real Lie algebra $`𝔤`$ is in the class $`\overline{MD}`$ if every K-orbit is of dimension, equal 0 or dim $`𝔤`$. Further more, one proved that (\[D, Theorem 4.4\]) Up to isomorphism, every Lie algebra of class $`\overline{MD}`$ is one of the following:
1. Commutative Lie algebras.
2. Lie algebra $`aff\left(𝐑\right)`$ of affine transformations of the real straight line
3. Lie algebra $`aff\left(𝐂\right)`$ of affine transformations of the complex straight line.
Thus, by calculation for the group of affine transformations of the real straight line $`Aff\left(𝐑\right)`$ in \[DH\] and here for the group affine transformations of the complex straight line $`Aff\left(𝐂\right)`$ we obtained a description of the quantum $`\overline{MD}`$ co-adjoint orbits.
### 6.2 $`MD_4`$-groups
We refer the reader to the results of Nguyen Viet Hai \[H3\]-\[H4\] for the class of $`MD_4`$-groups (i.e. 4-dimensional solvable Lie groups, all the coadjoint of which are of dimension 0 or maximal). It is interesting that here he obtained the same exact computation for $``$-products and all representations.
### 6.3 $`SO(3)`$
As an typical example of compact Lie group, the author proosed Job A. Nable to consider the case of $`SO\left(3\right)`$. We refer the reader to the results of Job Nable \[Na1\]-\[Na3\]. In these examples, it is interesting that the $``$-products, in some how as explained in these papers, involved the Maslov indices and Monodromy Theorem.
### 6.4 Exponetial groups
Arnal-Cortet constructed star-products for this case \[AC1\]-\[AC2\].
### 6.5 Compact groups
We refer readers to the works of C. Moreno \[Mo\].
## 7 Algebraic Noncommutative Chern Characters
### 7.1 Modification for inductive limits
Let $`G`$ be a compact group, $`HP_{}\left(C^{}\left(G\right)\right)`$ the periodic cyclic homology introduced in §2. Since $`C^{}\left(G\right)=\underset{}{\mathrm{lim}}_N_{i=1}^NMat_{n_i}\left(𝐂\right)`$, $`HP_{}\left(C^{}\left(G\right)\right)`$ coincides with the $`HP_{}\left(C^{}\left(G\right)\right)`$ defined by J. Cuntz-D. Quillen \[CQ\].
###### Lemma 7.1
Let $`\{I_N\}_{N𝐍}`$ be the above defined collection of ideals in $`C^{}(G)`$. Then
$$K_{}\left(C^{}\left(G\right)\right)=\underset{}{\mathrm{lim}}_{N𝐍}K_{}\left(I_N\right)=K_{}\left(𝐂\left(𝐓\right)\right),$$
where $`𝐓`$ is the fixed maximal torus in $`G`$.
First note that the algebraic K-theory of C\*-algebras has the stability property
$$K_{}\left(AM_n\left(𝐂\right)\right)K_{}\left(𝐂\left(𝐓\right)\right).$$
Hence,
$$\underset{}{\mathrm{lim}}K_{}\left(I_{n_i}\right)K_{}\left(\underset{w=\text{ highest weight}}{}𝐂_w\right)K_{}\left(𝐂\left(𝐓\right)\right),$$
by Pontryagin duality.
### 7.2 An algebraic construction of noncommutative Chern characters
J. Cuntz and D. Quillen \[CQ\] defined the so called $`X`$-complexes of $`𝐂`$-algebras and then used some ideas of Fedosov product to define algebraic Chern characters. We now briefly recall the their definitions. For a (non-commutative) associate $`𝐂`$-algebra $`A`$, consider the space of even non-commutative differential forms $`\mathrm{\Omega }^+\left(A\right)RA`$, equipped with the Fedosov product
$$\omega _1\omega _2:=\omega _1\omega _2\left(1\right)^{\left|\omega _1\right|}d\omega _1d\omega _2,$$
see \[CQ\]. Consider also the ideal $`IA:=_{k1}\mathrm{\Omega }^{2k}\left(A\right)`$. It is easy to see that $`RA/IAA`$ and that $`RA`$ admits the universal property that any based linear map $`\rho :AM`$ can be uniquely extended to a derivation $`D:RAM`$. The derivations $`D:RAM`$ bijectively correspond to lifting homomorpisms from $`RA`$ to the semi-direct product $`RAM`$, which also bijectively correspond to linear map $`\overline{\rho }:\overline{A}=A/𝐂M`$ given by
$$a\overline{A}D\left(\rho a\right).$$
From the universal property of $`\mathrm{\Omega }^1\left(RA\right)`$, we obtain a bimodule isomorphism
$$RA\overline{A}RA\mathrm{\Omega }^1\left(RA\right).$$
As in \[CQ\], let $`\mathrm{\Omega }^{}A=_{k0}\mathrm{\Omega }^{2k+1}A`$. Then we have
$$\mathrm{\Omega }^{}ARA\overline{A}\mathrm{\Omega }^1\left(RA\right)_\mathrm{\#}:=\mathrm{\Omega }^1\left(RA\right)/\left[(\mathrm{\Omega }^1\left(RA\right),RA)\right].$$
J. Cuntz and D. Quillen proved
###### Theorem 7.2
(\[CQ\], Theorem1): There exists an isomorphism of $`𝐙/(2)`$-graded complexes
$$\mathrm{\Phi }:\mathrm{\Omega }A=\mathrm{\Omega }^+A\mathrm{\Omega }^{}ARA\mathrm{\Omega }^1\left(RA\right)_\mathrm{\#},$$
such that
$$\mathrm{\Phi }:\mathrm{\Omega }^+ARA,$$
is defined by
$$\mathrm{\Phi }(a_0da_1\mathrm{}da_{2n}=\rho \left(a_1\right)\omega (a_1,a_2)\mathrm{}\omega (a_{2n1},a_{2n}),$$
and
$$\mathrm{\Phi }:\mathrm{\Omega }^{}A\mathrm{\Omega }^1\left(RA\right)_\mathrm{\#},$$
$$\mathrm{\Phi }\left(a_0da_1\mathrm{}da_{2n+1}\right)=\rho \left(a_1\right)\omega (a_1,a_2)\mathrm{}\omega (a_{2n1},a_{2n})\delta \left(a_{2n+1}\right).$$
With respect to this identification, the product in $`RA`$ is just the Fedosov product on even differential forms and the differentials on the $`X`$-complex
$$X\left(RA\right):RA\mathrm{\Omega }^+A\mathrm{\Omega }^1\left(RA\right)_\mathrm{\#}\mathrm{\Omega }^{}ARA$$
become the operators
$$\beta =b\left(1+\kappa \right)d:\mathrm{\Omega }^{}A\mathrm{\Omega }^+A,$$
$$\delta =N_{\kappa ^2}b+B:\mathrm{\Omega }^+A\mathrm{\Omega }^{}A,$$
where $`N_{\kappa ^2}=_{j=0}^{n1}\kappa ^{2j}`$, $`\kappa (da_1\mathrm{}da_n):=da_n\mathrm{}da_1`$.
Let us denote by $`IARA`$ the ideal of even non-commutative differential forms of order $`2`$. By the universal property of $`\mathrm{\Omega }^1`$
$$\mathrm{\Omega }^1(RA/IA)=\mathrm{\Omega }^1RA/(\left(IA\right)\mathrm{\Omega }^1RA+\mathrm{\Omega }^1RA.\left(IA\right)+dIA).$$
Since $`\mathrm{\Omega }^1RA=\left(RA\right)dRA=dRA.\left(RA\right)`$, then $`\mathrm{\Omega }^1RA\left(IA\right)IA\mathrm{\Omega }^1RAmod[RA,\mathrm{\Omega }^1R].`$
$$\mathrm{\Omega }^1(RA/IA)_\mathrm{\#}=\mathrm{\Omega }^1RA/([RA,\mathrm{\Omega }^1RA]+IA.dRA+dIA).$$
For $`IA`$-adic tower $`RA/\left(IA\right)^{n+1}`$, we have the complex
$$𝒳(RA/\left(IA\right)^{n+1}):RA/IA^{n+1}\mathrm{\Omega }^1RA/([RA,\mathrm{\Omega }^1RA]+\left(IA\right)^{n+1}dRA+d\left(IA\right)^{n+1}).$$
Define
$$𝒳^{2n+1}(RA,IA):RA/\left(IA\right)^{n+1}\mathrm{\Omega }^1RA/([RA,\mathrm{\Omega }^1RA]+\left(IA\right)^{n+1}dRA+d\left(IA\right)^{n+1})$$
$$RA/\left(IA\right)^{n+1},$$
$$𝒳^{2n}(RA,IA):RA/(\left(IA\right)^{n+1}+[RA,IA^n])\mathrm{\Omega }^1RA/([RA,\mathrm{\Omega }^1RA]+d\left(IA\right)^ndRA)$$
$$RA/\left(\left(IA\right)^{n+1}+[RA,IA^n]\right).$$
One has
$$b\left(\left(IA\right)^ndIA\right)=[\left(IA\right)^n,IA]\left(IA\right)^{n+1},$$
$$d\left(IA\right)^{n+1}\underset{j=0}{\overset{n}{}}\left(IA\right)^jd\left(IA\right)\left(IA\right)^{nj}\left(IA\right)^ndIA+[RA,\mathrm{\Omega }^1RA].$$
and hence
$$𝒳^1(RA,IA=X(RA,IA),$$
$$𝒳^0(RA,IA)=\left(RA/IA\right)_\mathrm{\#}.$$
There is an sequence of maps between complexes
$$\mathrm{}X\left(RA/IA\right)𝒳^{2n+1}(RA,IA)𝒳^{2n}(RA,IA)X\left(RA/IA\right)\mathrm{}$$
We have the inverse limits
$$\widehat{X}(RA,IA):=\underset{}{\mathrm{lim}}X\left(RA/\left(IA\right)^{n+1}\right)=\underset{}{\mathrm{lim}}𝒳^n(RA,IA).$$
Remark that
$$𝒳^q=\mathrm{\Omega }A/F^q\mathrm{\Omega }A,$$
$$\widehat{X}\left(RA/IA\right)=\widehat{\mathrm{\Omega }}A.$$
We quote the second main result of J. Cuntz and D. Quillen (\[CQ\], Thm2), namely:
$$H_i\widehat{𝒳}(RA,IA)=\underset{i}{HP}\left(A\right).$$
We now apply this machinery to our case. First we have the following.
###### Lemma 7.3
$$\underset{}{\mathrm{lim}}\stackrel{}{HP}\left(I_N\right)\stackrel{}{HP}\left(𝐂\left(𝐓\right)\right).$$
By similar arguments as in the previous lemma 7.1. More precisely, we have
$$HP\left(I_{n_i}\right)=HP\left(\underset{w=\text{heighest weight}}{}𝐂_w\right)HP\left(𝐂\left(𝐓\right)\right)$$
by Pontryagin duality.
Now, for each idempotent $`eM_n\left(A\right)`$ there is an unique element $`xM_n\left(\widehat{RA}\right)`$. Then the element
$$\stackrel{~}{e}:=x+\left(x\frac{1}{2}\right)\underset{n1}{}\frac{2^n\left(2n1\right)!!}{n!}\left(xx^{2n}\right)^{2n}M_n\left(\widehat{RA}\right)$$
is a lifting of $`e`$ to an idempotent matrix in $`M_n\left(\widehat{RA}\right)`$. Then the map $`\left[e\right]tr\left(\stackrel{~}{e}\right)`$ defines the map $`K_0H_0\left(X\left(\widehat{RA}\right)\right)=HP_0\left(A\right)`$. To an element $`gGL_n\left(A\right)`$ one associates an element $`pGL\left(\widehat{RA}\right)`$ and to the element $`g^1`$ an element $`qGL_n\left(\widehat{RA}\right)`$ then put
$$x=1qp,\text{ and }y=1pq.$$
And finally, to each class $`\left[g\right]GL_n\left(A\right)`$ one associates
$$tr\left(g^1dg\right)=tr\left(1x\right)^1d\left(1x\right)=d\left(tr\left(log\left(1x\right)\right)\right)=tr\underset{n=0}{\overset{\mathrm{}}{}}x^ndx\mathrm{\Omega }^1\left(A\right)_\mathrm{\#}.$$
Then $`\left[g\right]tr\left(g^1dg\right)`$ defines the map $`K_1\left(A\right)HH_1\left(A\right)=H_1\left(X\left(\widehat{RA}\right)\right)=HP_1\left(A\right)`$.
###### Definition 7.4
Let $`HP(I_{n_i})`$ be the periodic cyclic cohomology defined by Cuntz-Quillen. Then the pairing
$$K_{}^{alg}\left(C^{}\left(G\right)\right)\times \underset{N}{}\stackrel{}{HP}\left(I_N\right)𝐂$$
defines an algebraic non-commutative Chern character
$$ch_{alg}:K_{}^{alg}\left(C^{}\left(G\right)\right)\underset{}{HP}\left(C^{}\left(G\right)\right),$$
which gives us a variant of non-commutative Chern characters with values in $`HP`$-groups.
We close this section with an algebraic analogue of theorem 3.2.
###### Theorem 7.5
Let $`G`$ be a compact group and $`𝐓`$ a fixed maximal compact torus of $`G`$. Then in the notations of 4.3, the Chern character
$$ch_{alg}:K_{}\left(C^{}\left(G\right)\right)\underset{}{HP}\left(C^{}\left(G\right)\right)$$
is an isomorphism, which can be identified with the classical Chern character
$$ch:K_{}\left(𝐂\left(𝐓\right)\right)HP\left(𝐂\left(𝐓\right)\right)$$
which is also an isomorphism.
## Acknowledgments
This survey is an extended version of invited lectures at the Operator Theory Seminar, the University of Iowa, September 14 and 21, 1999 and at the INFAS (Iowa-Nebraska Funtional Analysis Seminar), Drake University, October 30, 1999 and was completed during the stay of the author as a visiting mathematician at the Department of mathematics, The University of Iowa. The author would like to express the deep and sincere thanks to Professor Tuong Ton-That and his spouse, Dr. Thai-Binh Ton-That for their effective helps and kind attention they provided during the stay in Iowa, and also for a discussion about the PBW Theorem. The deep thanks are also addressed to the organizers of the Seminar on Mathematical Physics, Seminar on Operator Theory in Iowa and the Iowa-Nebraska Functional Analysis Seminar (INFAS), in particular the professors Raul Curto, Palle Jorgensen, Paul Muhly and Tuong Ton-That for the stimulating scientific atmosphere.
The author would like to thank the University of Iowa for the hospitality and the scientific support, the Alexander von Humboldt Foundation, Germany, for an effective support.
Institute of Mathematics, National Centre for Science and Technology, P. O. Box 631, Bo Ho, VN-10000, Hanoi, Vietnam
Email: dndiep@hn.vnn.vn
and
Department of Mathematics, The University of Iowa, 14 MacLean Hall, Iowa City, IA 52242-1419, U.S.A.
Email: ndiep@math.uiowa.edu
|
warning/0003/quant-ph0003015.html
|
ar5iv
|
text
|
# Atomic Quantum State Teleportation and Swapping
## Abstract
A set of protocols for atomic quantum state teleportation and swapping utilizing Einstein-Podolsky-Rosen light is proposed. The protocols are suitable for collective spin states of a macroscopic sample of atoms, i.e. for continuous atomic variables. Feasibility of experimental realization for teleportation of a gas sample of atoms is analyzed.
Quantum teleportation , has recently attracted considerable attention as the means of disembodied transfer of an unknown quantum state. Besides being a new fundamental concept in quantum physics, it is also relevant for such applications as quantum computing and quantum memory. Experimental realizations of quantum teleportation have so far been limited to teleportation of light , . Since quantum information processing involves material particles such as atoms and ions, teleportation of atomic states will be the next important benchmark on the way to obtaining a complete set of quantum information processing tools. Several proposals for teleportation of atoms onto atoms and atoms onto light have been published recently .
We propose here a set of protocols for teleportation of a collective spin state of a macroscopic atomic sample. Our teleportation procedure does not require strong coupling of a single atom with light, and does not therefore involve high-Q cavities. We also consider a variation of the procedure that provides swapping of the spin states of two atomic systems.
As in the e.-m. field quantum teleportation of Refs., the protocol for quantum teleportation of spin proposed here relies on the continuous entanglement of the Einstein-Podolsky-Rosen (EPR) -type output of an optical parametric oscillator (OPO) below threshold. For parametric gain $`r`$ ($`r>0`$) the quadratures of the two modes of the EPR state satisfy the following relations:
$`\widehat{X}_1+\widehat{X}_2`$ $`=`$ $`\sqrt{2}\mathrm{exp}(r)\widehat{X}_v,`$ (1)
$`\widehat{Y}_1\widehat{Y}_2`$ $`=`$ $`\sqrt{2}\mathrm{exp}(r)\widehat{Y}_v,`$ (2)
where $`\widehat{X}_v,\widehat{Y}_v`$ describe a vacuum mode.
As was pointed out in the original teleportation proposal , it is imperative to be able to perform joint measurements on the quantum state to be teleported and one of the EPR states. We propose to use off-resonant atom-photon interaction to achieve this goal. Such interaction as a probe of the quantum state for atoms with total ground state electronic angular momentum equal $`\mathrm{}/2`$ has been analyzed in Ref. , and recently has been shown to be suitable for spin quantum non-demolition (QND) measurements , as well as for generation of entangled samples of atoms . As shown in Refs. the unitary time-evolution operator corresponding to the interaction with light propagating along the $`z`$-axis has the following form:
$$\widehat{U}=\mathrm{exp}(ia\widehat{S}_z\widehat{F}_z),$$
(3)
where $`a`$ is given by $`a=\frac{\sigma }{AF}\frac{\gamma }{\mathrm{\Delta }}\alpha _v`$, $`\sigma `$ is the resonant absorption cross section for an unpolarized photon on an unpolarized atom of total spin $`F`$, $`A`$ is the area of the transverse cross section of the light beam, $`\gamma `$ is the spontaneous emission rate of the upper atomic level, $`\mathrm{\Delta }`$ is the detuning, $`\widehat{𝐅}=_\mu \widehat{𝐅}^{(\mu )}`$ (summed over all the atoms in the interaction region) is the operator of the collective ground-state atomic spin, $`\widehat{𝐒}`$ is the operator of the Stokes vector of the optical field integrated over the duration of the interaction. The dynamic vector polarizability $`\alpha _v=\pm 1`$ for the $`D_1`$ transition of alkali atoms, while $`\alpha _v=\frac{1}{2}`$ for the $`D_2`$ transition. Here the upper/lower sign is for the hyperfine sublevel with $`F=I\pm \frac{1}{2}`$, $`I`$ is the value of the nuclear spin.
Interaction (3) leads to rotation of the polarization of the field that is proportional to $`\widehat{F}_z`$. Subsequent optical polarization measurements provide the classical information in our teleportation protocol. In order to do that, each of the two EPR modes (polarized along the y-axis) is mixed on a polarizing beamsplitter with a strong x-polarized coherent pulse having $`n_1`$ and $`n_2`$ photons respectively. Then, for the Stokes vectors of the two resulting bright EPR fields $`\widehat{𝐒}^{(1)}`$ and $`\widehat{𝐒}^{(2)}`$ we obtain: $`\widehat{S}_z^{(1)}\sqrt{n_1}\widehat{X}_1,\widehat{S}_y^{(1)}\sqrt{n_1}\widehat{Y}_1,\widehat{S}_z^{(2)}\sqrt{n_2}\widehat{X}_2,\widehat{S}_y^{(2)}\sqrt{n_2}\widehat{Y}_2`$. That is, the spin components of the Stokes vectors of the bright EPR fields are given by the corresponding quadrature operators of the OPO output.
Teleportation of an atomic spin state onto polarization state of light. Let Alice pass her bright EPR beam 1 with the Stokes vector $`\widehat{𝐒}^{(1)}`$ through the cell containing polarized atomic vapor along the z-axis. The unitary time evolution operator of Eq.(3) in the case of small spin fluctuations in $`\widehat{𝐅}`$ and $`\widehat{𝐒}`$ ($`\sqrt{(\mathrm{\Delta }\widehat{F}_{z,y})^2}F`$, $`\sqrt{(\mathrm{\Delta }\widehat{S}_{z,y})^2}S)`$ leads to the following transformations:
$`\widehat{S}_y^{(1)(out)}`$ $``$ $`\widehat{S}_y^{(1)(in)}+{\displaystyle \frac{an_1}{2}}\widehat{F}_z^{(in)},`$ (4)
$`\widehat{F}_y^{(out)}`$ $``$ $`\widehat{F}_y^{(in)}+(aNF)\widehat{S}_z^{(1)(in)}.`$ (5)
These equations for interaction of the collective atomic spin with light resemble the beamsplitter relations used for teleportation of a mode of electromagnetic field , . An insightful analysis of why a linear device of a beamsplitter allows to perform joint measurements necessary for quantum teleportation has been given by Vaidman and Yoran . The connection to our atom-light interaction case takes root in the fact that small rotations of a spin on the Bloch sphere around some point are equivalent to displacements in the tangent plane of this point, which can be considered as motions in the phase plane of the harmonic oscillator.
The nonlinear atom-light interaction of Eq.(3) is analogous to the QND-type interaction for the optical quadratures $`\widehat{X}_a\widehat{X}_b`$. Unlike the beamsplitter, which mixes both the $`\widehat{X}`$ and the $`\widehat{Y}`$ quadratures of the two input beams in the same fashion, interaction $`\widehat{X}_a\widehat{X}_b`$mixes only one pair of quadratures ($`\widehat{Y}_a`$ and $`\widehat{Y}_b`$), leaving the other pair ($`\widehat{X}_a`$ and $`\widehat{X}_b)`$ unchanged. Relations (5) correspond to a 50:50 beamsplitter when
$$|a|\sqrt{FNn_1/2}=1.$$
(6)
Subsequent (destructive) measurement of $`\widehat{S}_y^{(1)(out)}`$ provides Alice with half of the classical information needed for implementation of ”feed-forward” on Bob’s bright EPR beam to complete the quantum teleportation. The other half must come from the measurement of $`\widehat{F}_y^{(out)}`$. Although it is possible in principle to use a completely destructive measurement (e.g., by means of photoionization), experimentally it may be more convenient to employ another QND measurement using an auxiliary coherent pulse. Let us rotate the spin by $`\pi /2`$ around the x-axis, so that a pulse propagating along the z-axis will measure $`\widehat{F}_y^{(out)}`$. The corresponding transformation is
$`\widehat{S}_z^{(coh)(out)}`$ $``$ $`\widehat{S}_z^{(coh)(in)}+{\displaystyle \frac{an_{coh}}{2}}\widehat{F}_y^{(out)}`$ (7)
$``$ $`\widehat{S}_z^{(coh)(in)}+{\displaystyle \frac{an_{coh}}{2}}\widehat{F}_y^{(in)}+{\displaystyle \frac{n_{coh}}{n_1}}\widehat{S}_z^{(1)(in)}.`$ (8)
The first term is negligible if $`n_{coh}/n_11`$. After Alice sends the results of measurements of $`\widehat{S}_y^{(1)(out)}`$ and $`\widehat{S}_z^{(coh)(out)}`$ to Bob, reconstruction of $`\widehat{F}^{(in)}`$ in $`\widehat{𝐒}^{(2)}`$ is achieved by Bob rotating the latter around the z-axis by the angle $`\varphi _1=\frac{2}{\sqrt{n_1n_2}}S_y^{(1)(out)}`$and around the y-axis by the angle $`\varphi _2=\frac{2\sqrt{n_1}}{n_{coh}\sqrt{n_2}}S_z^{(coh)(out)}`$ . The rotations are equivalent to displacements because the angles are small. We obtain using Eqs.(2)
$`\widehat{S}_y^{(2)(out)}`$ $``$ $`\sqrt{{\displaystyle \frac{n_2}{2FN}}}\widehat{F}_z^{(in)}+\sqrt{2n_2}\mathrm{exp}(r)\widehat{Y}_v,`$ (9)
$`\widehat{S}_z^{(2)(out)}`$ $``$ $`\sqrt{{\displaystyle \frac{n_2}{2FN}}}\widehat{F}_y^{(in)}+\sqrt{2n_2}\mathrm{exp}(r)\widehat{X}_v.`$ (10)
The last terms in these expressions are due to the extra noise introduced by the imperfect EPR state. This noise goes to zero as $`r`$ goes to infinity. These terms correspond to the ”quantum duty” of the quantum teleportation , as seen explicitly in the case of zero parametric gain, $`r=0`$. Eqs. (10) show that the Stokes vector $`\widehat{𝐒}_2^{out}`$ is identical to the initial collective spin vector of Alice’s atoms when
$$\frac{n_2}{2FN}=1.$$
(11)
The teleportation protocol described above may serve as a read-out for quantum memory with the atomic sample as a memory cell. On the other hand, if the process of mapping of the Stokes vector $`\widehat{𝐒}_2^{out}`$ onto another atomic spin follows the described atom-to-light teleportation, atom-to-atom teleportation is achieved.
Teleportation of atoms onto atoms. We now describe a protocol which performs the direct teleportation of the quantum state of Alice’s collective spin onto Bob’s collection of atoms without using light as an intermediate carrier of the quantum state. Suppose we have two macroscopic spin systems (Alice’s and Bob’s) in the initial states given by collective spin operators $`\widehat{F}_A,\widehat{F}_B`$ with mean polarizations $`\widehat{F}_{Ax}=\widehat{F}_{Bx}`$ and with the other two projections $`\widehat{F}_{Ay},\widehat{F}_{By},\widehat{F}_{Az},\widehat{F}_{Bz}.`$ We also have at our disposal the EPR source of light as described above, with each EPR beam mixed with a strong pulse containing equal photon numbers, $`n_1=n_2=n`$, polarized along the $`x`$-axis, similar to the previous section. Alice’s EPR beam is phase shifted by $`\pi /2`$ so that the Stokes operators are $`\widehat{S}_{Az}=\widehat{S}_{By},\widehat{S}_{Ay}=\widehat{S}_{Bz}`$. The protocol begins with Alice sending her EPR beam along the $`z`$-axis and measuring its $`y`$ -Stokes parameter with the detector $`D_{A1}`$, and Bob sending a coherent $`x`$-polarized pulse $`\widehat{S}_B^{coh}`$ containing $`n_c`$ photons along the $`y`$-axis and measuring its $`z`$-Stokes parameter with the detector $`D_{B1}`$ (see Fig. 1). The resulting atomic states of Alice and Bob are described by the following operators:
$`\widehat{F}_{Az}^{^{}}`$ $`=`$ $`\widehat{F}_{Az},\widehat{F}_{Ay}^{^{}}=\widehat{F}_{Ay}+\widehat{S}_{Az},`$ (12)
$`\widehat{F}_{By}^{^{}}`$ $`=`$ $`\widehat{F}_{By},\widehat{F}_{Bz}^{^{}}=\widehat{F}_{Bz}\widehat{S}_{By}^{coh}.`$ (13)
In these equations we combined conditions $`n_1=n_2=n`$ and (6),(11) to obtain unity coupling coefficients between $`\widehat{F}`$ and $`\widehat{S}`$ projections. The Stokes parameters measured by detectors $`D_{A1}`$ and $`D_{B1}`$ are
$`\widehat{d}_{A1}`$ $`=`$ $`\widehat{S}_{Ay}+\widehat{F}_{Az},`$ (14)
$`\widehat{d}_{B1}`$ $`=`$ $`\widehat{S}_{Bz}^{coh}{\displaystyle \frac{n_c}{n}}\widehat{F}_{By}{\displaystyle \frac{n_c}{n}}\widehat{F}_{By}`$ (15)
We assume $`n/n_c1`$. While the above measurements are performed, the second EPR beam sent by Alice to Bob begins its journey along the quantum channel.
Next Alice sends a coherent $`x`$-polarized probe containing $`n_c`$ photons along $`y`$ axis onto the detector $`D_{A2}`$. The detector measures
$$\widehat{d}_{A2}=\widehat{S}_{Az}^{coh}\frac{n_c}{n}\widehat{F}_{Ay}^{^{}}\frac{n_c}{n}\widehat{F}_{Ay}^{^{}}$$
(16)
Alice now sends $`d_{A1},\frac{n}{n_c}d_{A2}`$ to Bob along a classical channel.
When Bob receives the EPR beam from Alice he sends it along $`z`$ axis onto the detector $`D_{B2}`$ which reads
$$\widehat{d}_{B2}=\widehat{S}_{By}+\widehat{F}_{Bz}^{^{}}.$$
(17)
After that the state of Bob’s atoms is
$$\widehat{F}_{By}^{^{\prime \prime }}=\widehat{F}_{By}^{^{}}+\widehat{S}_{Bz},\widehat{F}_{Bz}^{^{^{\prime \prime }}}=\widehat{F}_{Bz}^{^{}}.$$
(18)
To complete the teleportation Bob now rotates his atomic state. We use displacements instead of rotations to simplify the expressions. Bob’s state is displaced along $`z`$ by $`\frac{n}{n_c}d_{A2}d_{B2}`$ and along $`y`$ by $`d_{A1}+\frac{n}{n_c}d_{B1}`$ (see Fig.1). The final state of his atoms, according to (14-18), is described by the following equations:
$`\widehat{F}_{Bz}^{tele}=\widehat{F}_{Bz}^{^{\prime \prime }}{\displaystyle \frac{n}{n_c}}d_{A2}d_{B2}=\widehat{F}_{Ay},`$ (19)
$`\widehat{F}_{By}^{tele}=\widehat{F}_{By}^{^{\prime \prime }}+d_{A1}+{\displaystyle \frac{n}{n_c}}d_{B1}=\widehat{F}_{Az}`$ (20)
and the teleportation of the unknown state of Alice’s collection of atomic spins onto Bob’s atoms is proven (within a rotation of $`\pi `$ around the $`x`$-axis). In the above equations we assumed perfect entanglement between the EPR beam components ($`r\mathrm{}`$ in Eqs.(2)).
Atomic quantum state swapping. We now describe a protocol which exchanges initial quantum states of two collections of atomic spins. Suppose Alice’s and Bob’s spin samples are in the initial state given by operators $`\widehat{F}_A,\widehat{F}_B`$ with mean polarizations $`\widehat{F}_{Ax}=\widehat{F}_{Bx}.`$ We again have at our disposal the EPR source of light as described above, with Stokes operators $`\widehat{S}_z,\widehat{S}_y`$ for one of the beams, and $`\widehat{S}_z,\widehat{S}_y`$ for the other. Both EPR beams are mixed with strong beams containing equal photon numbers, $`n_1=n_2=n`$, polarized along the $`x`$ -axis in a way similar to the previous section. One of the EPR beams is used for a joint measurement on the $`z`$ spin components of both samples by sending it through both of them along the $`z`$ -axis. The resulting states of atomic samples are:
$`\widehat{F}_{Az}^{^{}}`$ $`=`$ $`\widehat{F}_{Az};\widehat{F}_{Bz}^{^{}}=\widehat{F}_{Bz},`$ (21)
$`\widehat{F}_{Ay}^{^{}}`$ $`=`$ $`\widehat{F}_{Ay}+\widehat{S}_z;\widehat{F}_{By}^{^{}}=\widehat{F}_{By}+\widehat{S}_z;`$ (22)
and the state of the beam is
$$\widehat{S}_z^{^{}}=\widehat{S}_z;\widehat{S}_y^{^{}}=\widehat{S}_y+\widehat{F}_{Az}+\widehat{F}_{Bz}.$$
(23)
In the above equations we assumed conditions (6),(11) to be fulfilled. Next, the second EPR beam shifted in phase by $`\pi /2`$ is sent along the $`y`$ -axis of the two atomic samples to perform a joint measurement on the $`y`$ spin components. The beam is transformed by the phase shift and the change of the direction in the following way $`\widehat{S}_z,\widehat{S}_y\widehat{S}_z^r,\widehat{S}_y^r`$. The evolution operator in this new coordinate system is $`\widehat{U}=\mathrm{exp}(ia\widehat{S}_y^r\widehat{F}_y).`$ We obtain
$`\widehat{F}_{Az}^{^{\prime \prime }}`$ $`=`$ $`\widehat{F}_{Az}\widehat{S}_y^r=\widehat{F}_{Az}+\widehat{S}_y;\widehat{F}_{Bz}^{^{\prime \prime }}=\widehat{F}_{Bz}+\widehat{S}_y,`$ (24)
$`\widehat{F}_{Ay}^{^{\prime \prime }}`$ $`=`$ $`\widehat{F}_{Ay}^{^{}}=\widehat{F}_{Ay}+\widehat{S}_z;\widehat{F}_{By}^{^{\prime \prime }}=\widehat{F}_{By}^{^{}}=\widehat{F}_{By}+\widehat{S}_z,`$ (25)
and
$`\widehat{S}_z^r^{}`$ $`=`$ $`\widehat{S}_z^{^{}}=\widehat{S}_z+\widehat{F}_{Ay}^{^{}}+\widehat{F}_{By}^{^{}}=\widehat{F}_{Ay}+\widehat{F}_{By}+\widehat{S}_z,`$ (26)
$`\widehat{S}_y^r^{}`$ $`=`$ $`\widehat{S}_y^{^{}}=\widehat{S}_y.`$ (27)
Here we used Eqs. (21, 23). After interacting with the atoms the two EPR beams are detected by photodetectors $`D_1,D_2`$. The Stokes parameter $`S_y^{^{}}`$ (23) is measured for the first beam and the Stokes parameter $`S_z^r^{}`$ (26) for the second. The results of the measurements are used to rotate the atomic spins in order to achieve the swapping. The projections $`\widehat{F}_{Az}^{^{\prime \prime }}`$ and $`\widehat{F}_{Bz}^{^{\prime \prime }}`$ are displaced by the value $`S_y^{^{}}`$ obtained from $`D_1`$, and the projections $`\widehat{F}_{Ay}^{^{\prime \prime }}`$ and $`\widehat{F}_{By}^{^{\prime \prime }}`$ are displaced by the value $`S_z^r^{}`$. The results are $`\widehat{F}_{Az}^{swap}=\widehat{F}_z^{^{\prime \prime }}S_y^{^{}}=\widehat{F}_{Bz},\widehat{F}_{Ay}^{swap}=\widehat{F}_y^{^{\prime \prime }}S_z^r^{}=\widehat{F}_{By}`$. Similarly for the other atomic system $`\widehat{F}_{Bz}^{swap}=\widehat{F}_{Az}\widehat{F}_{By}^{swap}=\widehat{F}_{Ay}`$ and the initial quantum states of the two samples have been swapped.
Feasibility of experimental realization. The experimental attractiveness of the above protocols relies on the fact that they can be applied to the system as simple as a gas of atoms in a cell at room temperature. The only basic assumption is that the number of atoms in the teleported ensemble has to be much greater than one. The unity coupling conditions $`\frac{1}{2}an=aNF=1`$ yields $`n=2FN,|a|=|(\sigma \gamma \alpha _v)/(AF\mathrm{\Delta })|=2/n`$. The required optical depth of the samples has to fulfill the condition $`\alpha _\mathrm{\Delta }=\sigma N\gamma /A|\mathrm{\Delta }|=\gamma /|\mathrm{\Delta }|1`$ to secure the validity of interaction described by Eq.(3). Optimal cross section of the optical beams is then $`A=(\sigma n|\alpha _v|\alpha _\mathrm{\Delta })/(2F)`$, that shows that $`n1`$ is necessary to fulfill $`A\sigma \lambda ^2`$ (the condition of weak focusing, i.e. weak coupling of light to a single atom). If a collection of, say $`N=10^5,`$ $`Cs`$ atoms in $`F=4`$ state probed close to the $`6S_{1/2}6P_{3/2},\gamma =5MHz`$ is to be teleported, the following parameters will be close to optimal: the detuning of probes $`\mathrm{\Delta }=800MHz`$, the cell size $`10\times 10\times 200\mu m^3`$, and the atomic density $`n_A=5\times 10^{12}cm^3.`$ The cell size can vary with the optimal size scaling roughly as $`\sqrt{N}`$. The number of photons in EPR pulses should be around $`n=8\times 10^5`$ with the pulse duration longer than $`100n\mathrm{sec}`$ to avoid saturation effects. With a typical finite bandwidth non-classical light generated by an optical parametric oscillator the pulse duration will also have to be longer than the inverse OPO bandwidth, $`\mathrm{\Gamma }_{OPO}^110n\mathrm{sec}`$ to preserve quantum correlations. A buffer gas can be added into the cell and/or its walls can be coated to extend the lifetime of the ground state spin state .
Summary. We propose a set of quantum state teleportation and swapping protocols based on an off-resonant interaction of EPR light with macroscopic atomic spin ensembles. The three protocols proposed include (a) teleportation of an atomic state onto a state of light suitable for quantum memory read out (b) teleportation of an atomic sample onto another atomic sample and (c) quantum state swapping of two atomic samples. The macroscopic character of the collective spin teleported unites this proposal with the previous work on quantum state processing with continuous quantum variables of light and atoms ,, and experimental approaches towards creating multiatom entangled states ,,. The absence of high-Q optical cavities may become an attractive feature simplifying an experimental realization. One possible realization involving a gas of long lived spins of alkali atoms is shown to be accessible. Teleportation between small cells filled with atoms should provide a suitable playground for studies of quantum memory and continuous quantum information processing.
We gratefully acknowledge useful discussions with B. Julsgaard, I. Cirac, P. Zoller, K. Mølmer and I. Walmsley. ESP acknowledges funding from the Danish Research Council and Thomas B. Thriges Center for Quantum Information.
Present address: NEC Research Institute, Inc., 4 Independence Way, Princeton, NJ 08540-6634.
|
warning/0003/hep-th0003085.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
The study of D-branes as probes of geometry and topology of space-time has by now been of wide practice (cf. e.g. ). In particular, the analysis of the moduli space of gauge theories, their matter content, superpotential and $`\beta `$-function, as world-volume theories of D-branes sitting at geometrical singularities is still a widely pursued topic. Since the pioneering work in , where the moduli and matter content of D-branes probing ALE spaces had been extensively investigated, much work ensued. The primary focus on (Abelian) orbifold singularities of the type $`\text{ }\mathrm{C}^2/\mathrm{ZZ}_n`$ was quickly generalised using McKay’s Correspondence, to arbitrary (non-Abelian) orbifold singularities $`\text{ }\mathrm{C}^2/(\mathrm{\Gamma }SU(2))`$, i.e., to arbitrary ALE spaces, in .
Several directions followed. With the realisation that these singularities provide various horizons, was quickly generalised to a treatment for arbitrary finite subgroups $`\mathrm{\Gamma }SU(N)`$, i.e., to generic Gorenstein singularities, by . The case of $`SU(3)`$ was then promptly studied in using this technique and a generalised McKay-type Correspondence was proposed in . Meanwhile, via T-duality transformations, certain orbifold singularities can be mapped to type II brane-setups in the fashion of . The relevant gauge theory data on the world volume can thereby be conveniently read from configurations of NS-branes, D-brane stacks as well as orientifold planes. For $`\text{ }\mathrm{C}^2`$ orbifolds, the $`A`$ and $`D`$ series have been thus treated , whereas for $`\text{ }\mathrm{C}^3`$ orbifolds, the Abelian case of $`\mathrm{ZZ}_k\times \mathrm{ZZ}_k^{}`$ has been solved by the brane box models . First examples of non-Abelian $`\text{ }\mathrm{C}^3`$ orbifolds have been addressed in as well as recent works in .
Thus rests the status of orbifold theories. What we note in particular is that once we specify the properties of the orbifold in terms of the algebraic properties of the finite group, the gauge theory information is easily extracted. Of course, orbifolds are a small subclass of algebro-geometric singularities. This is where we move on to toric varieties. Inspired by the linear $`\sigma `$-model approach of , which provides a rich structure of the moduli space, especially in connexion with various geometrical phases of the theory, the programme of utilising toric methods to study the behaviour of the gauge theory on D-branes which live on and hence resolve certain singularities was initiated in . In this light, toric methods provide a powerful tool for studying the moduli space of the gauge theory. In treating the F-flatness and D-flatness conditions for the SUSY vacuum in conjunction, these methods show how branches of the moduli space and hence phases of the theory may be parametrised by the algebraic equations of the toric variety. Recent developments in “brane diamonds,” as an extension of the brane box rules, have been providing great insight to such a wider class of toric singularities, especially the generalised conifold, via blown-up versions of the standard brane setups . Indeed, with toric techniques much information could be extracted as we can actually analytically describe patches of the moduli space.
Now Abelian orbifolds have toric descriptions and the above methodolgy is thus immediately applicable thereto. While bearing in mind that though non-Abelian orbifolds have no toric descriptions, a single physical D-brane has been placed on various general toric singularities. Partial resolutions of $`\text{ }\mathrm{C}^3/(\mathrm{ZZ}_2\times \mathrm{ZZ}_2)`$, such as the conifold and the suspended pinched point have been investigated in and brane setups giving the field theory contents are constructed by . Groundwork for the next family, coming from the toric orbifold $`\text{ }\mathrm{C}^3/(\mathrm{ZZ}_3\times \mathrm{ZZ}_3)`$, such as the del Pezzo surfaces and the zeroth Hirzebruch, has been laid in . Essentially, given the gauge theory data on the D-brane world volume, the procedure of transforming this information (F and D terms) into toric data which parametrises the classical moduli space is by now well-established.
One task is therefore immediately apparent to us: how do we proceed in the reverse direction, i.e., when we probe a toric singularity with a D-brane, how do we know the gauge theory on its world-volume? We recall that in the case of orbifold theories, devised a general method to extract the gauge theory data (matter content, superpotential etc.) from the geometry data (the characters of the finite group $`\mathrm{\Gamma }`$), and vice versa given the geometry, brane-setups for example, conveniently allow us to read out the gauge theory data. The same is not true for toric singularities, and the second half of the above bi-directional convenience, namely, a general method which allows us to treat the inverse problem of extracting gauge theory data from toric data is yet pending, or at least not in circulation.
The reason for this shortcoming is, as we shall see later, that the problem is highly non-unique. It is thus the purpose of this writing to address this inverse problem: given the geometry data in terms of a toric diagram, how does one read out (at least one) gauge theory data in terms of the matter content and superpotential? We here present precisely this algorithm which takes the matrices encoding the singularity to the matrices encoding a good gauge theory of the D-brane which probes the said singularity.
The structure of the paper is as follows. In Section 2 we review the procedure of proceeding from the gauge theory data to the toric data, while establishing nomenclature. In Subsection 3.1, we demonstrate how to extract the matter content and F-terms from the charge matrix of the toric singularity. In Subsection 3.2, we exemplify our algorithm with the well-known suspended pinched point before presenting in detail in Subsection 3.3, the general algorithm of how to obtain the gauge theory information from the toric data by the method of partial resolutions. In Subsection 3.4, we show how to integrate back to obtain the actual superpotential once the F-flatness equations are extracted from the toric data. Section 4 is then devoted to the illustration of our algorithm by tabulating the D-terms and F-terms of D-brane world volume theory on the toric del Pezzo surfaces and Hirzebruch zero. We finally discuss in Section 5, the non-uniqueness of the inverse problem and provide, through the studying of two types of ambiguities, ample examples of rather different gauge theories flowing to the same toric data. Discussions and future prospects are dealt with in Section 6.
## 2 The Forward Procedure: Extracting Toric Data From Gauge Theories
We shall here give a brief review of the procedures involved in going from gauge theory data on the D-brane to toric data of the singularity, using primarily the notation and concepts from . In the course thereof special attention will be paid on how toric diagrams, SUSY fields and linear $`\sigma `$-models weave together.
A stack of $`n`$ D-brane probes on algebraic singularities gives rise to SUSY gauge theories with product gauge groups resulting from the projection of the $`U(n)`$ theory on the original stack by the geometrical structure of the singularity. For orbifolds $`\text{ }\mathrm{C}^k/\mathrm{\Gamma }`$, we can use the structure of the finite group $`\mathrm{\Gamma }`$ to fabricate product $`U(n_i)`$ gauge groups . For toric singularities, since we have only (Abelian) $`U(1)`$ toroidal actions, we are so far restricted to product $`U(1)`$ gauge groups<sup>2</sup><sup>2</sup>2 Proposals toward generalisations to D-brane stacks have been made .. In physical terms, we have a single D-brane probe. Extensive work has been done in to see how the geometrical structure of the variety can be thus probed and how the gauge theory moduli may be encoded. The subclass of toric singularities, namely Abelian orbifolds, has been investigated to great detail and we shall make liberal usage of their properties throughout.
Now let us consider the world-volume theory on the D-brane probe on a toric singularity. Such a theory, as it is a SUSY gauge theory, is characterised by its matter content and interactions. The former is specified by quiver diagrams which in turn give rise to D-term equations; the latter is given by a superpotential, whose partial derivatives with respect to the various fields are the so-called F-term equations. F and D-flatness subsequently describe the (classical) moduli space of the theory. The basic idea is that the D-term equations together with the FI-parametres, in conjunction with the F-term equations, can be concatenated together into a matrix which gives the vectors forming the dual cone of the toric variety which the D-branes probe. We summarise the algorithm of obtaining the toric data from the gauge theory in the following, and to illuminate our abstraction and notation we will use the simple example of the Abelian orbifold $`\text{ }\mathrm{C}^3/(\mathrm{ZZ}_2\times \mathrm{ZZ}_2)`$ as given in Figure 1.
1. Quivers and D-Terms:
1. The bi-fundamental matter content of the gauge theory can be conveniently encoded into a quiver diagram $`𝒬`$, which is simply the (possibly directed) graph whose adjacency matrix $`a_{ij}`$ is precisely the matrix of the bi-fundamentals. In the case of an Abelian orbifold<sup>3</sup><sup>3</sup>3This is true for all orbifolds but of course only Abelian ones have known toric description. prescribed by the group $`\mathrm{\Gamma }`$, this diagram is the McKay Quiver (i.e., for the irreps $`R_i`$ of $`\mathrm{\Gamma }`$, $`a_{ij}`$ is such that $`RR_i=_ja_{ij}R_j`$ for some fundamental representation $`R`$). We denote the set of nodes as $`𝒬_0:=\{v\}`$ and the set of the edges, $`𝒬_1:=\{a\}`$. We let the number of nodes be $`r`$; for Abelian orbifolds, $`r=|\mathrm{\Gamma }|`$ (and for generic orbifolds $`r`$ is the number of conjugacy classes of $`\mathrm{\Gamma }`$). Also, we let the number of edges be $`m`$; this number depends on the number of supersymmetries which we have. The adjacency matrix (bi-fundamentals) is thus $`r\times r`$ and the gauge group is $`\underset{j=1}{\overset{r}{}}SU(w_j)`$. For our example of $`\mathrm{ZZ}_2\times \mathrm{ZZ}_2`$, $`r=4`$, indexed as 4 gauge groups $`U(1)_A\times U(1)_B\times U(1)_C\times U(1)_D`$ corresponding to the 4 nodes, while $`m=4\times 3=12`$, corresponding to the 12 arrows in Figure 1. The adjacency matrix for the quiver is $`\left(\begin{array}{cccc}0& 1& 1& 1\\ 1& 0& 1& 1\\ 1& 1& 0& 1\\ 1& 1& 1& 0\end{array}\right)`$. Though for such simple examples as Abelian orbifolds and conifolds, brane setups and specify the values of $`w_j`$ as well as $`a_{ij}`$ completely<sup>4</sup><sup>4</sup>4For arbitrary orbifolds, $`\underset{j}{}w_in_i=|\mathrm{\Gamma }|`$ where $`n_i`$ are the dimensions of the irreps of $`\mathrm{\Gamma }`$; for Abelian case, $`n_i=1`$. , there is yet no discussion in the literature of obtaining the matter content and gauge group for generic toric varieties in a direct and systematic manner and a partial purpose of this note is to present a solution thereof.
2. From the $`r\times r`$ adjacency matrix, we construct a so-called $`r\times m`$ incidence matrix $`d`$ for $`𝒬`$; this matrix is defined as $`d_{v,a}:=\delta _{v,head(a)}\delta _{v,tail(a)}`$ for $`v𝒬_0`$ and $`a𝒬_1`$. Because each column of $`d`$ must contain a 1, a $`1`$ and the rest 0’s by definition, one row of $`d`$ is always redundant; this physically signifies the elimination of an overall trivial $`U(1)`$ corresponding to the COM motion of the branes. Therefore we delete a row of $`d`$ to define the matrix $`\mathrm{\Delta }`$ of dimensions $`(r1)\times m`$; and we could always extract $`d`$ from $`\mathrm{\Delta }`$ by adding a row so as to force each column to sum to zero. This matrix $`\mathrm{\Delta }`$ thus contains almost as much information as $`a_{ij}`$ and once it is specified, the gauge group and matter content are also, with the exception that precise adjoints (those charged under the same gauge group factor and hence correspond to arrows that join a node to itself) are not manifest. For our example the $`4\times 12`$ matrix $`d`$ is as follows and $`\mathrm{\Delta }`$ is the top 3 rows:
$$d=\left(\begin{array}{ccccccccccccc}& X_{AD}& X_{BC}& X_{CB}& X_{DA}& X_{AB}& X_{BA}& X_{CD}& X_{DC}& X_{AC}& X_{BD}& X_{CA}& X_{DB}\\ A& 1& 0& 0& 1& 1& 1& 0& 0& 1& 0& 1& 0\\ B& 0& 1& 1& 0& 1& 1& 0& 0& 0& 1& 0& 1\\ C& 0& 1& 1& 0& 0& 0& 1& 1& 1& 0& 1& 0\\ & & & & & & & & & & & & \\ D& 1& 0& 0& 1& 0& 0& 1& 1& 0& 1& 0& 1\end{array}\right)$$
3. The moment maps, arising in the sympletic-quotient language of the toric variety, are simply $`\mu :=d|x(a)|^2`$ where $`x(a)`$ are the affine coordinates of the $`\text{ }\mathrm{C}^r`$ for the torus $`(\text{ }\mathrm{C}^{})^r`$ action. Physically, $`x(a)`$ are of course the bi-fundamentals in chiral multiplets (in our example they are $`X_{ij\{A,B,C,D\}}`$ as labelled above) and the D-term equations for each $`U(1)`$ group is
$$D_i=e^2(\underset{a}{}d_{ia}|x(a)|^2\zeta _i)$$
with $`\zeta _i`$ the FI-parametres. In matrix form we have $`\mathrm{\Delta }|x(a)|^2=\stackrel{}{\zeta }`$ and see that D-flatness gives precisely the moment map. These $`\zeta `$-parametres will encode the resolution of the toric singularity as we shall shortly see.
2. Monomials and F-Terms:
1. From the super-potential $`W`$ of the SUSY gauge theory, one can write the F-Term equation as the system $`\frac{}{X_j}W=0`$. The remarkable fact is that we could solve the said system of equations and express the $`m`$ fields $`X_i`$ in terms of $`r+2`$ parametres $`v_j`$ which can be summarised by a matrix $`K`$.
$$X_i=\underset{j}{}v_j^{K_{ij}},i=1,2,..,m;j=1,2,..,r+2$$
(2.1)
This matrix $`K`$ of dimensions $`m\times (r+2)`$ is the analogue of $`\mathrm{\Delta }`$ in the sense that it encodes the F-terms and superpotential as $`\mathrm{\Delta }`$ encodes the D-terms and the matter content. In the language of toric geometry $`K`$ defines a cone<sup>5</sup><sup>5</sup>5 We should be careful in this definition. Strictly speaking we have a lattice $`𝐌=\mathrm{ZZ}^{r+2}`$ with its dual lattice $`𝐍\mathrm{ZZ}^{r+2}`$. Now let there be a set of $`\mathrm{ZZ}_+`$-independent vectors $`\{\stackrel{}{k}_i\}𝐌`$ and a cone is defined to be generated by these vectors as $`\sigma :=\{_ia_i\stackrel{}{k}_i|a_i\mathrm{IR}_0\}`$; Our $`𝐌_+`$ should be $`𝐌\sigma `$. In much of the literature $`𝐌_+`$ is taken to be simply $`𝐌_+^{}:=\{_ia_i\stackrel{}{k}_i|a_i\mathrm{ZZ}_0\}`$ in which case we must make sure that any lattice point contained in $`𝐌_+`$ but not in $`𝐌_+^{}`$ must be counted as an independent generator and be added to the set of generators $`\{\stackrel{}{k}_i\}`$. After including all such points we would have $`𝐌_+^{}=𝐌_+`$. Throughout our analyses, our cone defined by $`K`$ as well the dual cone $`T`$ will be constituted by such a complete set of generators. $`𝐌_+`$ : a non-negative linear combination of $`m`$ vectors $`\stackrel{}{K_i}`$ in an integral lattice $`\mathrm{ZZ}^{r+2}`$.
For our example, the superpotential is
$$\begin{array}{c}W=X_{AC}X_{CD}X_{DA}X_{AC}X_{CB}X_{BA}+X_{CA}X_{AB}X_{BC}X_{CA}X_{AD}X_{DC}\\ +X_{BD}X_{DC}X_{CB}X_{BD}X_{DA}X_{AB}X_{DB}X_{BC}X_{CD},\end{array}$$
giving us 12 F-term equations and with the manifold of solutions parametrisable by $`4+2`$ new fields, whereby giving us the $`12\times 6`$ matrix (we here show the transpose thereof, thus the horizontal direction corresponds to the original fields $`X_i`$ and the vertical, $`v_j`$):
$$K^t=\left(\begin{array}{ccccccccccccc}& X_{AC}& X_{BD}& X_{CA}& X_{DB}& X_{AB}& X_{BA}& X_{CD}& X_{DC}& X_{AD}& X_{BC}& X_{CB}& X_{DA}\\ & & & & & & & & & & & & \\ v_1=X_{AC}& 1& 0& 0& 1& 1& 0& 0& 1& 0& 0& 0& 0\\ v_2=X_{BD}& 0& 1& 1& 0& 1& 0& 0& 1& 0& 0& 0& 0\\ v_3=X_{BA}& 0& 0& 0& 0& 0& 1& 0& 1& 0& 1& 0& 1\\ v_4=X_{CD}& 0& 0& 0& 0& 1& 0& 1& 0& 0& 1& 0& 1\\ v_5=X_{AD}& 0& 0& 1& 1& 0& 0& 0& 0& 1& 1& 0& 0\\ v_6=X_{CB}& 0& 0& 1& 1& 0& 0& 0& 0& 0& 0& 1& 1\end{array}\right).$$
For example, the third column reads $`X_{CA}=v_2v_5^1v_6`$, i.e., $`X_{AD}X_{CA}=X_{BD}X_{CB}`$, which the the F-flatness condition $`\frac{W}{X_{DC}=0}`$. The details of obtaining $`W`$ and $`K`$ from each other are discussed in and Subsection 3.4.
2. We let $`T`$ be the space of (integral) vectors dual to $`K`$, i.e., $`KT0`$ for all entries; this gives an $`(r+2)\times c`$ matrix for some positive integer $`c`$. Geometrically, this is the definition of a dual cone $`𝐍_+`$ composed of vectors $`\stackrel{}{T_i}`$ such that $`\stackrel{}{K}\stackrel{}{T}0`$. The physical meaning for doing so is that $`K`$ may have negative entries which may give rise to unwanted singularities and hence we define a new set of $`c`$ fields $`p_i`$ (a priori we do not know the number $`c`$ and we present the standard algorithm of finding dual cones in the Appendix). Thus we reduce (2.1) further into
$$v_j=\underset{\alpha }{}p_\alpha ^{T_{j\alpha }}$$
(2.2)
whereby giving $`X_i=_jv_j^{K_{ij}}=_\alpha p_\alpha ^{_jK_{ij}T_{j\alpha }}`$ with $`_jK_{ij}T_{j\alpha }0`$. For our $`\mathrm{ZZ}_2\times Z_2`$ example, $`c=9`$ and
$$T_{j\alpha }=\left(\begin{array}{cccccccccc}& p_1& p_2& p_3& p_4& p_5& p_6& p_7& p_8& p_9\\ & & & & & & & & & \\ X_{AC}& 1& 1& 0& 0& 0& 0& 0& 0& 1\\ X_{BD}& 0& 1& 1& 0& 0& 0& 0& 0& 1\\ X_{BA}& 0& 0& 1& 1& 1& 0& 0& 0& 0\\ X_{CD}& 0& 0& 1& 0& 1& 1& 0& 0& 0\\ X_{AD}& 0& 0& 0& 0& 0& 1& 1& 0& 1\\ X_{CB}& 0& 0& 0& 0& 0& 1& 1& 1& 0\end{array}\right)$$
3. These new variables $`p_\alpha `$ are the matter fields in Witten’s linear $`\sigma `$-model. How are these fields charged? We have written $`r+2`$ fields $`v_j`$ in terms of $`c`$ fields $`p_\alpha `$, and hence need $`c(r+2)`$ relations to reduce the independent variables. Such a reduction can be done via the introduction of the new gauge group $`U(1)^{c(r+2)}`$ acting on the $`p_i`$’s so as to give a new set of D-terms. The charges of these fields can be written as $`Q_{k\alpha }`$. The gauge invariance condition of $`v_i`$ under $`U(1)^{c(r+2)}`$, by (2.2), demands that the $`(cr2)\times c`$ matrix $`Q`$ is such that $`_\alpha T_{j\alpha }Q_{k\alpha }=0`$. This then defines for us our charge matrix $`Q`$ which is the cokernel of $`T`$:
$$TQ^t=(T_{j\alpha })(Q_{k\alpha })^t=0,j=1,..,r+2;\alpha =1,..,c;k=1,..,(cr2)$$
For our example, the charge matrix is $`(942)\times 9`$ and one choice is $`Q_{k\alpha }=\left(\begin{array}{ccccccccc}0& 0& 0& 1& 1& 1& 1& 0& 0\\ 0& 1& 0& 0& 0& 0& 1& 1& 1\\ 1& 1& 1& 0& 1& 0& 0& 0& 0\end{array}\right)`$.
4. In the linear $`\sigma `$-model language, the F-terms and D-terms can be treated in the same footing, i.e., as the D-terms (moment map) of the new fields $`p_\alpha `$; with the crucial difference being that the former must be set exactly to zero<sup>6</sup><sup>6</sup>6Strictly speaking, we could have an F-term set to a non-zero constant. An example of this situation could be when there is a term $`a\varphi +\varphi \stackrel{~}{Q}Q`$ in the superpotential for some chargeless field $`\varphi `$ and charged fields $`\stackrel{~}{Q}`$ and $`Q`$. The F-term for $`\varphi `$ reads $`\stackrel{~}{Q}Q=a`$ and not 0. However, in our context $`\varphi `$ behaves like an integration constant and for our purposes, F-terms are set exactly to zero. while the latter are to be resolved by arbitrary FI-parameters.
Therefore in addition to finding the charge matrix $`Q`$ for the new fields $`p_\alpha `$ coming from the original F-terms as done above, we must also find the corresponding charge matrix $`Q_D`$ for the $`p_i`$ coming from the original D-terms. We can find $`Q_D`$ in two steps. Firstly, we know the charge matrix for $`X_i`$ under $`U(1)^{r1}`$, which is $`\mathrm{\Delta }`$. By (2.1), we transform the charges to that of the $`v_j`$’s, by introducing an $`(r1)\times (r+2)`$ matrix $`V`$ so that $`VK^t=\mathrm{\Delta }`$. To see this, let the charges of $`v_j`$ be $`V_{lj}`$ then by (2.1) we have $`\mathrm{\Delta }_{li}=\underset{j}{}V_{lj}K_{ij}=VK^t`$. A convenient $`V`$ which does so for our $`\mathrm{ZZ}_2\times \mathrm{ZZ}_2`$ example is $`\left(\begin{array}{cccccc}1& 0& 1& 0& 1& 0\\ 0& 1& 1& 0& 0& 1\\ 1& 0& 0& 1& 0& 1\end{array}\right)_{\left(41\right)\times \left(4+2\right)}`$. Secondly, we use (2.2) to transform the charges from $`v_j`$’s to our final variables $`p_\alpha `$’s, which is done by introducing an $`(r+2)\times c`$ matrix $`U_{j\alpha }`$ so that $`UT^t=\mathrm{Id}_{(r+2)\times (r+2)}`$. In our example, one choice for $`U`$ is $`U_{j\alpha }=\left(\begin{array}{ccccccccc}1& 0& 0& 0& 0& 0& 0& 0& 0\\ 1& 1& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 1& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 1& 1& 0& 0\\ 0& 1& 0& 0& 0& 0& 0& 0& 1\\ 0& 0& 0& 0& 0& 0& 0& 1& 0\end{array}\right)_{\left(4+2\right)\times 9}`$.
Threfore, combining the two steps, we obtain $`Q_D=VU`$ and in our example, $`\left(VU\right)_{l\alpha }=\left(\begin{array}{ccccccccc}1& 1& 0& 1& 0& 0& 0& 0& 1\\ 1& 1& 0& 1& 0& 0& 0& 1& 0\\ 1& 0& 0& 0& 0& 1& 1& 1& 0\end{array}\right)`$.
3. Thus equipped with the information from the two sides: the F-terms and D-terms, and with the two required charge matrices $`Q`$ and $`VU`$ obtained, finally we concatenate them to give a $`(c3)\times c`$ matrix $`Q_t`$. The transpose of the kernel of $`Q_t`$, with (possible repeated columns) gives rise to a matrix $`G_t`$. The columns of this resulting $`G_t`$ then define the vertices of the toric diagram describing the polynomial corresponding to the singularity on which we initially placed our D-branes. Once again for our example, $`Q_t=\left(\begin{array}{cccccccccc}0& 0& 0& 1& 1& 1& 1& 0& 0& 0\\ 0& 1& 0& 0& 0& 0& 1& 1& 1& 0\\ 1& 1& 1& 0& 1& 0& 0& 0& 0& 0\\ & & & & & & & & & \\ 1& 1& 0& 1& 0& 0& 0& 0& 1& \zeta _1\\ 1& 1& 0& 1& 0& 0& 0& 1& 0& \zeta _2\\ 1& 0& 0& 0& 0& 1& 1& 1& 0& \zeta _3\end{array}\right)`$ and $`G_t=\left(\begin{array}{ccccccccc}0& 1& 0& 0& 1& 0& 1& 1& 1\\ 1& 1& 1& 0& 1& 0& 1& 0& 0\\ 1& 1& 1& 1& 1& 1& 1& 1& 1\end{array}\right)`$. The columns of $`G_t`$, up to repetition, are precisely marked in the toric diagram for $`\mathrm{ZZ}_2\times \mathrm{ZZ}_2`$ in Figure 1.
Thus we have gone from the F-terms and the D-terms of the gauge theory to the nodes of the toric diagram. In accordance with , $`G_t`$ gives the algebraic variety whose equation is given by the maximal ideal in the polynomial ring
$`\text{ }\mathrm{C}[YZ,XYZ,Z,X^1YZ,XY^1Z,XZ]`$ (the exponents $`(i,j,k)`$ in $`X^iY^jZ^k`$ are exactly the columns), which is $`uvw=s^2`$, upon defining $`u=(YZ)(XYZ)^2(Z)(XZ)^2;v=(YZ)^2(Z)^2(X^1YZ)^2;w=(Z)^2(XY^1Z)(XZ)^2`$ and
$`s=(YZ)^2(XYZ)(Z)^2(X^1YZ)(XY^1Z)(XZ)^2`$; this is precisely $`\text{ }\mathrm{C}^3/(\mathrm{ZZ}_2\times \mathrm{ZZ}_2)`$. In physical terms this equation parametrises the moduli space obtained from the F and D flatness of the gauge theory.
We remark two issues here. In the case of there being no superpotential we could still define $`K`$-matrix. In this case, with there being no F-terms, we simply take $`K`$ to be the identity. This gives $`T=`$Id and $`Q=0`$. Furthermore $`U`$ becomes Id and $`V=\mathrm{\Delta }`$, whereby making $`Q_t=\mathrm{\Delta }`$ as expected because all information should now be contained in the D-terms. Moreover, we note that the very reason we can construct a $`K`$-matrix is that all of the equations in the F-terms we deal with are in the form $`\underset{i}{}X_i^{a_i}=\underset{j}{}X_j^{b_j}`$; this holds in general if every field $`X_i`$ appears twice and precisely twice in the superpotential. More generic situations would so far transcend the limitations of toric techniques.
Schematically, our procedure presented above at length, what it means is as follows: we begin with two pieces of physical data: (1) matrix $`d`$ from the quiver encoding the gauge groups and D-terms and (2) matrix $`K`$ encoding the F-term equations. From these we extract the matrix $`G_t`$ containing the toric data by the flow-chart:
$$\begin{array}{ccccccc}\text{Quiver}d& & \mathrm{\Delta }& & & & \\ & & & & & & \\ \text{F-Terms}K& \stackrel{VK^t=\mathrm{\Delta }}{}& V& & & & \\ & & & & & & \\ T=\mathrm{Dual}(K)& \stackrel{UT^t=\mathrm{Id}}{}& U& & VU& & \\ & & & & & & \\ Q=[\mathrm{Ker}(T)]^t& & & & Q_t=\left(\begin{array}{c}Q\\ VU\end{array}\right)& & G_t=[\mathrm{Ker}(Q_t)]^t\end{array}$$
## 3 The Inverse Procedure: Extracting Gauge Theory Information from Toric Data
As outlined above we see that wherever possible, the gauge theory of a D-brane probe on certain singularities such as Abelian orbifolds, conifolds, etc., can be conveniently encoded into the matrix $`Q_t`$ which essentially concatenates the information contained in the D-terms and F-terms of the original gauge theory. The cokernel of this matrix is then a list of vectors which prescribes the toric diagram corresponding to the singularity. It is natural to question ourselves whether the converse could be done, i.e., whether given an arbitrary singularity which affords a toric description, we could obtain the gauge theory living on the D-brane which probes the said singularity. This is the inverse problem we projected to solve in the introduction.
### 3.1 Quiver Diagrams and F-terms from Toric Diagrams
Our result must be two-fold: first, we must be able to extract the D-terms, or in other words the quiver diagram which then gives the gauge group and matter content; second, we must extract the F-terms, which we can subsequently integrate back to give the superpotential. These two pieces of data then suffice to specify the gauge theory. Essentially we wish to trace the arrows in the above flow-chart from $`G_t`$ back to $`\mathrm{\Delta }`$ and $`K`$. The general methodology seems straightforward:
1. Read the column-vectors describing the nodes of the given toric diagram, repeat the appropriate columns to obtain $`G_t`$ and then set $`Q_t=\mathrm{Coker}(G_t)`$;
2. Separate the D-term ($`VU`$) and F-term ($`Q_t`$) portions from $`Q_t`$;
3. From the definition of $`Q`$, we obtain<sup>7</sup><sup>7</sup>7As mentioned before we must ensure that such a $`T`$ be chosen with a complete set of $`\mathrm{ZZ}_+`$-independent generators; $`T=\mathrm{ker}(Q)`$.
4. Farka’s Theorem guarantees that the dual of a convex polytope remains convex whence we could invert and have $`K=\mathrm{Dual}(T^t)`$; Moreover the duality theorem gives that Dual(Dual($`K))=K`$, thereby facilitating the inverse procedure.
5. Definitions $`UT^t=\mathrm{Id}`$ and $`VK^t=\mathrm{\Delta }`$ $``$ $`(VU)(T^tK^t)=\mathrm{\Delta }`$.
We see therefore that once the appropriate $`Q_t`$ has been found, the relations
$$K=\mathrm{Dual}(T^t)\mathrm{\Delta }=(VU)(T^tK^t)$$
(3.3)
retrieve our desired $`K`$ and $`\mathrm{\Delta }`$. The only setback of course is that the appropriate $`Q_t`$ is NOT usually found. Two ambiguities are immediately apparent to us: (A) In step 1 above, there is really no way to know a priori which of the vectors we should repeat when writing into the $`G_t`$ matrix; (B) In step 2, to separate the D-terms and the F-terms, i.e., which rows constitute $`Q`$ and which constitute $`VU`$ within $`Q_t`$, seems arbitrary. We shall in the last section discuss these ambiguities in more detail and actually perceive it to be a matter of interest. Meanwhile, in light thereof, we must find an alternative, to find a canonical method which avoids such ambiguities and gives us a consistent gauge theory which has such well-behaved properties as having only bi-fundamentals etc.; this is where we appeal to partial resolutions.
Another reason for this canonical method is compelling. The astute reader may question as to how could we guarantee, in our mathematical excursion of performing the inverse procedure, that the gauge theory we obtain at the end of the day is one that still lives on the world-volume of a D-brane probe? Indeed, if we naïvely traced back the arrows in the flow-chart, bearing in mind the said ambiguities, we have no a fortiori guarantee that we have a brane theory at all. However, the method via partial resolution of Abelian orbifolds (which are themselves toric) does give us assurance. When we are careful in tuning the FI-parametres so as to stay inside cone-partitions of the space of these parametres (and avoid flop transitions) we do still have the resulting theory being physical . Essentially this means that with prudence we tune the FI-parametres in the allowed domains from a parent orbifold theory, thereby giving a subsector theory which still lives on the D-brane probe and is well-behaved. Such tuning we shall practice in the following.
The virtues of this appeal to resolutions are thus twofold: not only do we avoid ambiguities, we are further endowed with physical theories. Let us thereby present this canonical mathod.
### 3.2 A Canonical Method: Partial Resolutions of Abelian Orbifolds
Our programme is standard : theories on the Abelian orbifold singularity of the form $`\text{ }\mathrm{C}^k/\mathrm{\Gamma }`$ for $`\mathrm{\Gamma }(k,n)=\mathrm{ZZ}_n\times \mathrm{ZZ}_n\times \mathrm{}\mathrm{ZZ}_n`$ ($`k1`$ times) are well studied. The complete information (and in particular the full $`Q_t`$ matrix) for $`\mathrm{\Gamma }(k,n)`$ is well known: $`k=2`$ is the elliptic model, $`k=3`$, the Brane Box, etc. In the toric context, $`k=2`$ has been analysed in great detail by , $`k=3,n=2`$ in e.g. , $`k=3,n=3`$ in . Now we know that given any toric diagram of dimension $`k`$, we can embed it into such a $`\mathrm{\Gamma }(k,n)`$-orbifold for some sufficiently large $`n`$; and we choose the smallest such $`n`$ which suffices. This embedding is always possible because the toric diagram for the latter is the $`k`$-simplex of length $`n`$ enclosing lattice points and any toric diagram, being a collection of lattice points, can be obtained therefrom via deletions of a subset of points. This procedure is known torically as partial resolutions of $`\mathrm{\Gamma }(k,n)`$. The crux of our algorithm is that the deletions in the toric diagram corresponds to the turning-on of the FI-parametres, and which in turn induces a method to determine a $`Q_t`$ matrix for our original singularity from that of $`\mathrm{\Gamma }(n,k)`$.
We shall first turn to an illustrative example of the suspended pinched point singularity (SPP) and then move on to discuss generalities. The SPP and conifold as resolutions of $`\mathrm{\Gamma }(3,2)=\mathrm{ZZ}_2\times \mathrm{ZZ}_2`$ have been extensively studied in . The SPP, given by $`xy=zw^2`$, can be obtained from the $`\mathrm{\Gamma }(3,2)`$ orbifold, $`xyz=w^2`$, by a single $`\mathrm{IP}^1`$ blow-up. This is shown torically in Figure 2. Without further ado let us demonstrate our procedure.
1. Embedding into $`\mathrm{ZZ}_2\times \mathrm{ZZ}_2`$: Given the toric diagram $`D`$ of SPP, we recognise that it can be embedded minimally into the diagram $`D^{}`$ of $`\mathrm{ZZ}_2\times \mathrm{ZZ}_2`$. Now information on $`D^{}`$ is readily at hand , as presented in the previous section. Let us re-capitulate:
$$Q_t^{}:=\left(\begin{array}{cccccccccc}p_1& p_2& p_3& p_4& p_5& p_6& p_7& p_8& p_9& \\ 0& 0& 0& 1& 1& 1& 1& 0& 0& 0\\ 0& 1& 0& 0& 0& 0& 1& 1& 1& 0\\ 1& 1& 1& 0& 1& 0& 0& 0& 0& 0\\ 1& 1& 0& 1& 0& 0& 0& 0& 1& \zeta _1\\ 1& 1& 0& 1& 0& 0& 0& 1& 0& \zeta _2\\ 1& 0& 0& 0& 0& 1& 1& 1& 0& \zeta _3\end{array}\right),$$
and
$$G_t^{}:=\mathrm{coker}\left(Q_t^{}\right)=\left(\begin{array}{ccccccccc}p_1& p_2& p_3& p_4& p_5& p_6& p_7& p_8& p_9\\ 0& 1& 0& 0& 1& 0& 1& 1& 1\\ 1& 1& 1& 0& 1& 0& 1& 0& 0\\ 1& 1& 1& 1& 1& 1& 1& 1& 1\end{array}\right),$$
which is drawn in Figure 1. The fact that the last row of $`G_t`$ has the same number (i.e., these three-vectors are all co-planar) ensures that $`D^{}`$ is Calabi-Yau . Incidentally, it would be very helpful for one to catalogue the list of $`Q_t`$ matrices of $`\mathrm{\Gamma }(3,n)`$ for $`n=2,3\mathrm{}`$ which would suffice for all local toric singularities of Calabi-Yau threefolds.
In the above definition of $`Q_t^{}`$ we have included an extra column $`(0,0,0,\zeta _1,\zeta _2,\zeta _3)`$ so as to specify that the first three rows of $`Q_t^{}`$ are F-terms (and hence exactly zero) while the last three rows are D-terms (and hence resolved by FI-parametres $`\zeta _{1,2,3}`$). We adhere to the notation in and label the columns (linear $`\sigma `$-model fields) as $`p_1\mathrm{}p_9`$; this is shown in Figure 2.
2. Determining the Fields to Resolve by Tuning $`\zeta `$: We note that if we turn on a single FI-parametre we would arrive at the SPP; this is the resolution of $`D^{}`$ to $`D`$. The subtlety is that one may need to eliminate more than merely the 7th column as there is more than one field attributed to each node in the toric diagram and eliminating column 7 some other columns corresponding to the adjacent nodes (namely out of 4,6,8 and 9) may also be eliminated. We need a judicious choice of $`\zeta `$ for a consistent blowup. To do so we must solve for fields $`p_{1,..,9}`$ and tune the $`\zeta `$-parametres such that at least $`p_7`$ acquires non-zero VEV (and whereby resolved). Recalling that the D-term equations are actually linear equations in the modulus-squared of the fields, we shall henceforth define $`x_i:=|p_i|^2`$ and consider linear-systems therein. Therefore we perform Gaussian row-reduction on $`Q^{}`$ and solve all fields in terms of $`x_7`$ to give: $`\stackrel{}{x}=\{x_1,x_2,x_1+\zeta _2+\zeta _3,\frac{2x_1x_2+x_7\zeta _1+\zeta _2}{2},2x_1x_2+\zeta _2+\zeta _3,\frac{2x_1x_2+x_7+\zeta _1+\zeta _2+2\zeta _3}{2},x_7,\frac{x_2+x_7\zeta _1\zeta _2}{2},\frac{x_2+x_7+\zeta _1+\zeta _2}{2}\}`$.
The nodes far away from $`p_7`$ are clearly unaffected by the resolution, thus the fields corresponding thereto continue to have zero VEV. This means we solve the above set of solutions $`\stackrel{}{x}`$ once again, setting $`x_{5,1,3,2}=0`$, with $`\zeta _{1,2,3}`$ being the variables, giving upon back substitution, $`\stackrel{}{x}=\{0,0,0,\frac{x_7\zeta _1\zeta _3}{2},0,\frac{x_7+\zeta _1+\zeta _3}{2},x_7,\frac{x_7\zeta _1+\zeta _3}{2},\frac{x_7+\zeta _1\zeta _3}{2}\}`$. Now we have an arbitrary choice and we set $`\zeta _3=0`$ and $`x_7=\zeta _1`$ to make $`p_4`$ and $`p_8`$ have zero VEV. This makes $`p_{6,7,9}`$ our candidate for fields to be resolved and seems perfectly reasonable observing Figure 2. The constraint on our choice is that all solutions must be $`0`$ (since the $`x_i`$’s are VEV-squared).
3. Solving for $`G_t`$: We are now clear what the resolution requires of us: in order to remove node $`p_7`$ from $`D^{}`$ to give the SPP, we must also resolve 6, 7 and 9. Therefore we immediately obtain $`G_t`$ by directly removing the said columns from $`G_t^{}`$:
$$G_t:=\mathrm{coker}\left(Q_t\right)=\left(\begin{array}{cccccc}p_1& p_2& p_3& p_4& p_5& p_8\\ 0& 1& 0& 0& 1& 1\\ 1& 1& 1& 0& 1& 0\\ 1& 1& 1& 1& 1& 1\end{array}\right),$$
the columns of which give the toric diagram $`D`$ of the SPP, as shown in Figure 2.
4. Solving for $`Q_t`$: Now we must perform linear combination on the rows of $`Q_t^{}`$ to obtain $`Q_t`$ so as to force columns 6, 7 and 9 zero. The following constraints must be born in mind. Because $`G_t`$ has 6 columns and 3 rows and is in the null space of $`Q_t`$, which itself must have $`93`$ columns (having eliminated $`p_{6,7,9}`$), we must have $`63=3`$ rows for $`Q_t`$. Also, the row containing $`\zeta _1`$ must be eliminated as this is precisely our resolution chosen above (we recall that the FI-parametres are such that $`\zeta _{2,3}=0`$ and are hence unresolved, while $`\zeta _1>0`$ and must be removed from the D-terms for SPP).
We systematically proceed. Let there be variables $`\{a_{i=1,..,6}\}`$ so that $`y:=_ia_i\mathrm{row}_i(Q_t^{})`$ is a row of $`Q_t`$. Then (a) the 6th, 7th and 9th columns of $`y`$ must be set to 0 and moreover (b) with these columns removed $`y`$ must be in the nullspace spanned by the rows of $`G_t`$. We note of course that since $`Q_t^{}`$ was in the nullspace of $`G_t^{}`$ initially, that the operation of row-combinations is closed within a nullspace, and that the columns to be set to 0 in $`Q_t^{}`$ to give $`Q_t`$ are precisely those removed in $`G_t^{}`$ to give $`G_t`$, condition (a) automatically implies (b). This condition (a) translates to the equations $`\{a_1+a_6=0,a_1+a_2a_6=0,a_2+a_4=0\}`$ which afford the solution $`a_1=a_6;a_2=a_4=0`$. The fact that $`a_4=0`$ is comforting, because it eliminates the row containing $`\zeta _1`$. We choose $`a_1=1`$. Furthermore we must keep row 5 as $`\zeta _2`$ is yet unresolved (thereby setting $`a_5=1`$). This already gives two of the 3 anticipated rows of $`Q_t`$: row<sub>5</sub> and row<sub>1</sub> \- row<sub>6</sub>. The remaining row must corresponds to an F-term since we have exhausted the D-terms, this we choose to be the only remaining variable: $`a_3=1`$. Consequently, we arrive at the matrix
$$Q_t=\left(\begin{array}{ccccccc}p_1& p_2& p_3& p_4& p_5& p_8& \\ 1& 1& 1& 0& 1& 0& 0\\ 1& 1& 0& 1& 0& 1& \zeta _2\\ 1& 0& 0& 1& 1& 1& \zeta _3\end{array}\right).$$
5. Obtaining $`K`$ and $`\mathrm{\Delta }`$ Matrices: The hard work is now done. We now recognise from $`Q_t`$ that $`Q=(1,1,1,0,1,0)`$, giving
$$T_{j\alpha }:=\mathrm{ker}\left(Q\right)=\left(\begin{array}{cccccc}0& 0& 0& 0& 0& 1\\ 1& 0& 0& 0& 1& 0\\ 0& 0& 0& 1& 0& 0\\ 1& 0& 1& 0& 0& 0\\ 1& 1& 0& 0& 0& 0\end{array}\right);K^t:=\mathrm{Dual}\left(T^t\right)=\left(\begin{array}{cccccc}1& 0& 0& 0& 0& 0\\ 0& 0& 1& 0& 1& 0\\ 0& 1& 0& 0& 0& 0\\ 0& 0& 1& 1& 0& 0\\ 0& 0& 0& 1& 0& 1\end{array}\right).$$
Subsequently we obtain $`T^tK^t=\left(\begin{array}{cccccc}0& 0& 0& 0& 1& 1\\ 0& 0& 0& 1& 0& 1\\ 0& 0& 1& 1& 0& 0\\ 0& 1& 0& 0& 0& 0\\ 0& 0& 1& 0& 1& 0\\ 1& 0& 0& 0& 0& 0\end{array}\right),`$ which we do observe indeed to have every entry positive semi-definite. Furthermore we recognise from $`Q_t`$ that $`VU=\left(\begin{array}{cccccc}1& 1& 0& 1& 0& 1\\ 1& 0& 0& 1& 1& 1\end{array}\right),`$ whence we obtain at last, using (3.3),
$$\mathrm{\Delta }=\left(\begin{array}{cccccc}1& 1& 0& 1& 1& 0\\ 1& 1& 1& 0& 0& 1\end{array}\right)d=\left(\begin{array}{ccccccc}& X_1& X_2& X_3& X_4& X_5& X_6\\ U\left(1\right)_A& 1& 1& 0& 1& 1& 0\\ U\left(1\right)_B& 1& 1& 1& 0& 0& 1\\ & & & & & & \\ U\left(1\right)_C& 0& 0& 1& 1& 1& 1\end{array}\right),$$
giving us the quiver diagram (included in Figure 3 for reference), matter content and gauge group of a D-brane probe on SPP in agreement with . We shall show in the ensuing sections that the superpotential we extract has similar accordance.
### 3.3 The General Algorithm for the Inverse Problem
Having indulged ourselves in this illustrative example of the SPP, we proceed to outline the general methodology of obtaining the gauge theory data from the toric diagram.
1. Embedding into $`\text{ }\mathrm{C}^k/(\mathrm{ZZ}_n)^{k1}`$: We are given a toric diagram $`D`$ describing an algebraic variety of complex dimension $`k`$ (usually we are concerned with local Calabi-Yau singularities of $`k=2,3`$ so that branes living thereon give $`𝒩=2,1`$ gauge theories). We immediately observe that $`D`$ could always be embedded into $`D^{}`$, the toric diagram of the orbifold $`\text{ }\mathrm{C}^k/(\mathrm{ZZ}_n)^{k1}`$ for some sufficiently large integer $`n`$. The matrices $`Q_t^{}`$ and $`G_t^{}`$ for $`D^{}`$ are standard. Moreover we know that the matrix $`G_t`$ for our original variety $`D`$ must be a submatrix of $`G_t^{}`$. Equipped with $`Q_t^{}`$ and $`G_t^{}`$ our task is to obtain $`Q_t`$; and as an additional check we could verify that $`Q_t`$ is indeed in the nullspace of $`G_t`$.
2. Determining the Fields to Resolve by Tuning $`\zeta `$: $`Q_t^{}`$ is a $`k\times a`$ matrix<sup>8</sup><sup>8</sup>8We henceforth understand that there is an extra column of zeroes and $`\zeta `$’s. (because $`D^{}`$ and $`D`$ are dimension $`k`$) for some $`a`$; $`G_t^{}`$, being its nullspace, is thus $`(ak)\times a`$. $`D`$ is a partial resolution of $`D^{}`$. In the SPP example above, we performed a single resolution by turning on one FI-parametre, generically however, we could turn on as many $`\zeta `$’s as the embedding permits. Therefore we let $`G_t`$ be $`(ak)\times (ab)`$ for some $`b`$ which depends on the number of resolutions. Subsequently the $`Q_t`$ we need is $`(kb)\times (ab)`$.
Now $`b`$ is determined directly by examining $`D^{}`$ and $`D`$; it is precisely the number of fields $`p`$ associated to those nodes in $`D^{}`$ we wish to eliminate to arrive at $`D`$. Exactly which $`b`$ columns are to be eliminated is determined thus: we perform Gaussian row-reduction on $`Q_t^{}`$ so as to solve the $`k`$ linear-equations in $`a`$ variables $`x_i:=|p_i|^2`$, with F-terms set to 0 and D-terms to FI-parametres. The $`a`$ variables are then expressed in terms of the $`\zeta _i`$’s and the set $`B`$ of $`x_i`$’s corresponding to the nodes which we definitely know will disappear as we resolve $`D^{}D`$. The subtlety is that in eliminating $`B`$, some other fields may also acquire non zero VEV and be eliminated; mathematically this means that Order$`(B)<b`$.
Now we make a judicious choice of which fields will remain and set them to zero and impose this further on the solution $`x_{i=1,..,a}=x_i(\zeta _j;B)`$ from above until Order$`(B)=b`$, i.e., until we have found all the fields we need to eliminate. We know this occurs and that our choice was correct when all $`x_i0`$ with those equaling 0 corresponding to fields we do not wish to eliminate as can be observed from the toric diagram. If not, we modify our initial choice and repeat until satisfaction. This procedure then determines the $`b`$ columns which we wish to eliminate from $`Q_t^{}`$.
3. Solving for $`G_t`$ and $`Q_t`$: Knowing the fields to eliminate, we must thus perform linear combinations on the $`k`$ rows of $`Q_t^{}`$ to obtain the $`kb`$ rows of $`Q_t`$ based upon the two constraints that (1) the $`b`$ columns must be all reduced to zero (and thus the nodes can be removed) and that (2) the $`kb`$ rows (with $`b`$ columns removed) are in the nullspace of $`G_t`$. As mentioned in our SPP example, condition (1) guarantees (2) automatically.
In other words, we need to solve for $`k`$ variables $`\{x_{i=1,..,k}\}`$ such that
$$\underset{i=1}{\overset{k}{}}x_i(Q_t^{})_{ij}=0\mathrm{for}j=p_1,p_2,\mathrm{}p_bB.$$
(3.4)
Moreover, we immediately obtain $`G_t`$ by eliminating the $`b`$ columns from $`G_t^{}`$. Indeed, as discussed earlier, (3.4) implies that $`\underset{i=1}{\overset{k}{}}\underset{jp_{1\mathrm{}b}}{}x_i(Q_t^{})_{ij}(G_t)_{mj}=0`$ for $`m=1,\mathrm{},ak`$ and hence guarantees that the $`Q_t`$ we obtain is in the nullspace of $`G_t`$.
We could phrase equation (3.4) for $`x_i`$ in matrix notation and directly evaluate
$$Q_t=NullSpace(W)^t\stackrel{~}{Q_t^{}}$$
(3.5)
where $`\stackrel{~}{Q_t^{}}`$ is $`Q_t^{}`$ with the appropriate columns ($`p_{1\mathrm{}b}`$) removed and $`W`$ is the matrix constructed from the deleted columns.
4. Obtaining the $`K`$ Matrix (F-term): Having obtained the $`(kb)\times (ab)`$ matrix $`Q_t`$ for the original variety $`D`$, we proceed with ease. Reading from the extraneous column of FI-parametres, we recognise matrices $`Q`$ (corresponding to the rows that have zero in the extraneous column) and $`VU`$ (corresponding to those with combinations of the unresolved $`\zeta `$’s in the last column). We let $`VU`$ be $`c\times (ab)`$ whereby making $`Q`$ of dimension $`(kbc)\times (ab)`$. The number $`c`$ is easily read from the embedding of $`D`$ into $`D^{}`$ as the number of unresolved FI-parametres.
From $`Q`$, we compute the kernel $`T`$, a matrix of dimensions $`(ab)(kbc)\times (ab)=(ak+c)\times (ab)`$ as well as the matrix $`K^t`$ of dimensions $`(ak+c)\times d`$ describing the dual cone to that spanned by the columns of $`T`$. The integer $`d`$ is uniquely determined from the dimensions of $`T`$ in accordance with the algorithm of finding dual cones presented in the Appendix. From these two matrices we compute $`T^tK^t`$, of dimension $`(ab)\times d`$.
5. Obtaining the $`\mathrm{\Delta }`$ Matrix (D-term): Finally, we use (3.3) to compute $`(VU)(T^tK^t)`$, arriving at our desired matrix $`\mathrm{\Delta }`$ of dimensions $`c\times d`$, the incidence matrix of our quiver diagram. The number of gauge groups we have is therefore $`c+1`$ and the number of bi-fundamentals, $`d`$.
Of course one may dispute that finding the kernel $`T`$ of $`Q`$ is highly non-unique as any basis change in the null-space would give an equally valid $`T`$. This is indeed so. However we note that it is really the combination $`T^tK^t`$ that we need. This is a dot-product in disguise, and by the very definition of the dual cone, this combination remains invariant under basis changes. Therefore this step of obtaining the quiver $`\mathrm{\Delta }`$ from the charge matrix $`Q_t`$ is a unique procedure.
### 3.4 Obtaining the Superpotential
Having noticed that the matter content can be conveniently obtained, we proceed to address the interactions, i.e., the F-terms, which require a little more care. The matrix $`K`$ which our algorithm extracts encodes the F-term equations and must at least be such that they could be integrated back to a single function: the superpotential.
Reading the possible F-flatness equations from $`K`$ is ipso facto straight-forward. The subtlety exists in how to find the right candidate among many different linear relations. As mentioned earlier, $`K`$ has dimensions $`m\times (r2)`$ with $`m`$ corresponding to the fields that will finally manifest in the superpotential, $`r2`$, the fields that solve them according to (2.1) and (2.2); of course, $`mr2`$. Therefore we have $`r2`$ vectors in $`\mathrm{ZZ}^m`$, giving generically $`mr+2`$ linear relations among them. Say we have $`\mathrm{row}_1+\mathrm{row}_3\mathrm{row}_7=0`$, then we simply write down $`X_1X_3=X_7`$ as one of the candidate F-terms. In general, a relation $`\underset{i}{}a_iK_{ij}=0`$ with $`a_i\mathrm{ZZ}`$ implies an F-term $`\underset{i}{}X_i^{a_i}=1`$ in accordance with (2.1). Of course, to find all the linear relations, we simply find the $`\mathrm{ZZ}`$-nullspace of $`K^t`$ of dimension $`mr+2`$.
Here a great ambiguity exists, as in our previous calculations of nullspaces: any linear combinations therewithin may suffice to give a new relation as a candidate F-term<sup>9</sup><sup>9</sup>9Indeed each linear relation gives a possible candidate and we seek the correct ones. For the sake of clarity we shall call candidates “relations” and reserve the term “F-term” for a successful candidate.. Thus educated guesses are called for in order to find the set of linear relations which may be most conveniently integrated back into the superpotential. Ideally, we wish this back-integration procedure to involve no extraneous fields (i.e., integration constants<sup>10</sup><sup>10</sup>10By constants we really mean functions since we are dealing with systems of partial differential equations.) other than the $`m`$ fields which appear in the K-matrix. Indeed, as we shall see, this wish may not always be granted and sometimes we must include new fields. In this case, the whole moduli space of the gauge theory will be larger than the one encoded by our toric data and the new fields parametrise new branches of the moduli in the theory.
Let us return to the SPP example to enlighten ourselves before generalising. We recall from subsection 3.2, that $`K=\left(\begin{array}{ccccccc}& X_1& X_2& X_3& X_4& X_5& X_6\\ & & & & & & \\ v_1& 1& 0& 0& 0& 0& 0\\ v_2& 0& 0& 1& 0& 1& 0\\ v_3& 0& 1& 0& 0& 0& 0\\ v_4& 0& 0& 1& 1& 0& 0\\ v_5& 0& 0& 0& 1& 0& 1\end{array}\right)`$ from which we can read out only one relation $`X_3X_6X_4X_5=0`$ using the rule described in the paragraph above. Of course there can be only one relation because the nullspace of $`K^t`$ is of dimension $`65=1`$.
Next we must calculate the charge under the gauge groups which this term carries. We must ensure that the superpotential, being a term in a Lagrangian, be a gauge invariant, i.e., carries no overall charge under $`\mathrm{\Delta }`$. From $`d=\left(\begin{array}{ccccccc}& X_1& X_2& X_3& X_4& X_5& X_6\\ & & & & & & \\ U\left(1\right)_A& 1& 1& 0& 1& 1& 0\\ U\left(1\right)_B& 1& 1& 1& 0& 0& 1\\ U\left(1\right)_C& 0& 0& 1& 1& 1& 1\end{array}\right)`$ we find the charge of $`X_3X_6`$ to be $`(q_A,q_B,q_C)=(0+0,1+(1),(1)+1)=(0,0,0)`$; of course by our very construction, $`X_4X_5`$ has the same charge. Now we have two choices: (a) to try to write the superpotential using only the six fields; or (b) to include some new field $`\varphi `$ which also has charge $`(0,0,0)`$. For (a) we can try the ansatz $`W=X_1X_2(X_3X_6X_4X_5)`$ which does give our F-term upon partial derivative with respect to $`X_1`$ or $`X_2`$. However, we would also have a new F-term $`X_1X_2X_3=0`$ by $`\frac{}{X_6}`$, which is inconsistent with our $`K`$ since columns 1, 2 and 3 certainly do not add to 0.
This leaves us with option (b), i.e., $`W=\varphi (X_3X_6X_4X_5)`$ say. In this case, when $`\varphi =0`$ we not only obtain our F-term, we need not even correct the matter content $`\mathrm{\Delta }`$. This branch of the moduli space is that of our original theory. However, when $`\varphi 0`$, we must have $`X_3=X_4=X_5=X_6=0`$. Now the D-terms read $`|X_1|^2|X_2|^2=\zeta _1=\zeta _2`$, so the moduli space is: $`\{\varphi \text{ }\mathrm{C},X_1\text{ }\mathrm{C}\}`$ such that $`\zeta _1+\zeta _2=0`$ for otherwise there would be no moduli at all. We see that we obtain another branch of moduli space. As remarked before, this is a general phenomenon when we include new fields: the whole moduli space will be larger than the one encoded by the toric data. As a check, we see that our example is exactly that given in , after the identification with their notation, $`Y_{12}X_6,X_{24}X_3,Z_{23}X_1,Z_{32}X_2,Y_{34}X_4,X_{13}X_5,Z_{41}\varphi `$ and $`(X_1X_2\varphi )\varphi `$. We note that if we were studying a non-Abelian extension to the toric theory, as by brane setups (e.g. ) or by stacks of probes (in progress from ), the chargeless field $`\varphi `$ would manifest as an adjoint field thereby modifying our quiver diagram. Of course since the study of toric methods in physics is so far restricted to product $`U(1)`$ gauge groups, such complexities do not arise. To avoid confusion we shall henceforth mark only the bi-fundamentals in our quiver diagrams but will write the chargeless fields explicit in the superpotential.
Our agreement with the results of is very reassuring. It gives an excellent example demonstrating that our canonical resolution technique and the inverse algorithm do indeed, in response to what was posited earlier, give a theory living on a D-brane probing the SPP (T-dual to the setup in ). However, there is a subtle point we would like to mention. There exists an ambiguity in writing the superpotential when the chargeless field $`\varphi `$ is involved. Our algorithm gives $`W=\varphi (X_3X_6X_4X_6)`$ while gives $`W=(X_1X_2\varphi )(X_3X_6X_4X_6)`$. Even though they have identical moduli, it is the latter which is used for the brane setup. Indeed, the toric methods by definition (in defining $`\mathrm{\Delta }`$ from $`a_{ij}`$) do not handle chargeless fields and hence we have ambiguities. Fortunately our later examples will not involve such fields.
The above example of the SPP was a naïve one as we need only to accommodate a single F-term. We move on to a more complicated example. Suppose we are now given $`d=\left(\begin{array}{ccccccccccc}& X_1& X_2& X_3& X_4& X_5& X_6& X_7& X_8& X_9& X_{10}\\ A& 1& 0& 0& 1& 0& 0& 0& 1& 0& 1\\ B& 1& 1& 0& 0& 0& 1& 0& 0& 1& 0\\ C& 0& 0& 1& 0& 1& 0& 1& 1& 1& 1\\ D& 0& 1& 1& 1& 1& 1& 1& 0& 0& 0\end{array}\right)`$ and $`K=\left(\begin{array}{cccccccccc}X_1& X_2& X_3& X_4& X_5& X_6& X_7& X_8& X_9& X_{10}\\ 1& 0& 1& 0& 0& 0& 1& 0& 0& 0\\ 0& 1& 1& 0& 0& 0& 0& 1& 0& 0\\ 1& 0& 0& 1& 0& 0& 0& 0& 1& 0\\ 0& 1& 0& 1& 0& 1& 0& 0& 0& 0\\ 0& 0& 1& 0& 1& 0& 1& 0& 0& 0\\ 0& 0& 0& 0& 0& 1& 1& 0& 0& 1\end{array}\right)`$. We shall see in the next section, that these arise for the del Pezzo 1 surface. Now the nullspace of $`K`$ has dimension $`106=4`$, we could obtain a host of relations from various linear combinations in this space. One relation is obvious: $`X_2X_7X_3X_6=0`$. The charge it carries is $`(q_A,q_B,q_C,q_D)=(0+0,1+0,0+1,1+(1))=(0,1,1,0)`$ which cancels that of $`X_9`$. Hence $`X_9(X_2X_7X_3X_6)`$ could be a term in $`W`$. Now $`\frac{}{X_2}`$ thereof gives $`X_7X_9`$ and from $`K`$ we see that $`X_7X_9X_1X_5X_{10}=0`$, therefore, $`X_1X_2X_5X_{10}`$ could be another term in $`W`$. We repeat this procedure, generating new terms as we proceed and introducing new fields where necessary. We are fortunate that in this case we can actually reproduce all F-terms without recourse to artificial insertions of new fields: $`W=X_2X_7X_9X_3X_6X_9X_4X_8X_7X_1X_2X_5X_{10}+X_3X_4X_{10}+X_1X_5X_6X_8`$.
Enlightened by these examples, let us return to some remarks upon generalities. Making all the exponents of the fields positive, the F-terms can then be written as
$$\underset{i}{}X_i^{a_i}=\underset{j}{}X_j^{b_j},$$
(3.6)
with $`a_i,b_j\mathrm{ZZ}^+`$. Indeed if we were to have another field $`X_k`$ such that $`k\{i\},\{j\}`$ then the term $`X_k\left(\underset{i}{}X_i^{a_i}\underset{j}{}X_j^{b_j}\right)`$, on the condition that $`X_k`$ appears only this once, must be an additive term in the superpotential $`W`$. This is because the F-flatness condition $`\frac{W}{X_k}=0`$ implies (3.6) immediately. Of course judicious observations are called for to (A) find appropriate relations (3.6) and (B) find $`X_k`$ among our $`m`$ fields. Indeed (B) may not even be possible and new fields may be forced to be introduced, whereby making the moduli space of the gauge theory larger than that encodable by the toric data.
In addition, we must ensure that each term in $`W`$ be chargeless under the product gauge groups. What this means for us is that for each of the terms $`X_k\left(\underset{i}{}X_i^{a_i}\underset{j}{}X_j^{b_j}\right)`$ we must have $`\mathrm{Charge}_s(X_k)+\underset{i}{}a_i\mathrm{Charge}_s(X_i)=0`$ for $`s=1,..,r`$ indexing through our $`r`$ gauge group factors (we note that by our very construction, for each gauge group, the charges for $`\underset{i}{}X_i^{a_i}`$ and for $`\underset{j}{}X_j^{b_j}`$ are equal). If $`X_k`$ in fact cannot be found among our $`m`$ fields, it must be introduced as a new field $`\varphi `$ with appropriate charge. Therefore with each such relation (3.6) read from $`K`$, we iteratively perform this said procedure, checking $`\mathrm{\Delta }_{sk}+\underset{i}{}a_i\mathrm{\Delta }_{si}=0`$ at each step, until a satisfactory superpotential is reached. The right choices throughout demands constant vigilance and astuteness.
## 4 An Illustrative Example: the Toric del Pezzo Surfaces
As the $`\text{ }\mathrm{C}^3/(\mathrm{ZZ}_2\times \mathrm{ZZ}_2)`$ resolutions were studied in great detail in , we shall use the data from to demonstrate the algorithm of finding the gauge theory from toric diagrams extensively presented in the previous section.
The toric diagram of the dual cone of the (parent) quotient singularity $`\text{ }\mathrm{C}^3/(\mathrm{ZZ}_3\times \mathrm{ZZ}_3)`$ as well as those of its resolution to the three toric del Pezzo surface are presented in Figure 4.
del Pezzo 1: Let us commence our analysis with the first toric del Pezzo surface<sup>11</sup><sup>11</sup>11Now some may identify the toric diagram of del Pezzo 1 as given by nodes (using the notation in Figure 4) $`(1,1,1)`$, $`(2,1,0)`$, $`(1,1,1)`$, $`(0,0,1)`$ and $`(1,0,2)`$ instead of the one we have chosen in the convention of , with nodes $`(0,1,2)`$, $`(0,0,1)`$, $`(1,1,1)`$, $`(1,0,0)`$ and $`(0,1,0)`$. But of course these two $`G_t`$ matrices describe the same algebraic variety. The former corresponds to $`\mathrm{Spec}\left(\text{ }\mathrm{C}[XY^1Z,X^2Y^1,X^1YZ,Z,X^1Z^2]\right)`$ while the latter corresponds to $`\mathrm{Spec}\left(\text{ }\mathrm{C}[Y^1Z^2,Z,X^1YZ,X,Y]\right)`$. The observation that $`(X^2Y^1)=(X)(X^1YZ)^1(Z)`$, $`(XY^1Z)=(X)(Y)^1(Z)`$ and $`(X^1Z^2)=(Y^1Z^2)(Y)(X^1)`$ for the generators of the polynomial ring gives the equivalence. In other words, there is an $`SL(5,\mathrm{ZZ})`$ transformation between the 5 nodes of the two toric diagrams.. From its toric diagram, we see that the minimal $`\mathrm{ZZ}_n\times \mathrm{ZZ}_n`$ toric diagram into which it embeds is $`n=3`$. As a reference, the toric diagram for $`\text{ }\mathrm{C}^3/(\mathrm{ZZ}_3\times \mathrm{ZZ}_3)`$ is given in Figure 4 and the quiver diagram, given later in the convenient brane-box form, in Figure 5. Luckily, the matrices $`Q_t^{}`$ and $`G_t^{}`$ for this Abelian quotient is given in . Adding the extra column of FI-parametres we present these matrices below<sup>12</sup><sup>12</sup>12In , a canonical ordering was used; for our purposes we need not belabour this point and use their $`Q_{total}^{}`$ as $`Q_t^{}`$. This is perfectly legitimate as long as we label the columns carefully, which we have done.:
$$\begin{array}{c}G_t^{}=(\begin{array}{cccccccccccccccccccccccc}p_1& p_2& p_3& p_4& p_5& p_6& p_7& p_8& p_9& p_{10}& p_{11}& p_{12}& p_{13}& p_{14}& p_{15}& p_{16}& p_{17}& p_{18}& p_{19}& p_{20}& p_{21}& p_{22}& p_{23}& p_{24}\\ 0& 0& 0& 1& 0& 0& 0& 1& 1& 1& 1& 0& 0& 0& 0& 1& 0& 0& 0& 0& 0& 0& 1& 0\\ 0& 0& 0& 1& 1& 0& 0& 1& 0& 1& 0& 0& 1& 0& 0& 1& 0& 0& 0& 1& 0& 0& 1& 0\\ 1& 1& 1& 1& 2& 1& 1& 1& 2& 3& 2& 1& 2& 1& 1& 1& 1& 1& 1& 2& 1& 1& 1& 1\end{array}\mathrm{}\mathrm{}\hfill \\ \\ \\ \mathrm{}\mathrm{}\begin{array}{ccccccccccccccccccc}p_{25}& p_{26}& p_{27}& p_{28}& p_{29}& p_{30}& p_{31}& p_{32}& p_{33}& p_{34}& p_{35}& p_{36}& p_{37}& p_{38}& p_{39}& p_{40}& p_{41}& p_{42}& \\ 0& 1& 1& 1& 1& 0& 0& 0& 0& 0& 0& 2& 1& 0& 0& 0& 1& 1\\ 0& 0& 1& 1& 2& 0& 0& 0& 0& 0& 0& 1& 0& 1& 1& 1& 0& 0\\ 1& 2& 1& 1& 0& 1& 1& 1& 1& 1& 1& 0& 0& 0& 0& 0& 0& 0\end{array})\hfill \end{array}$$
and
$$\begin{array}{c}Q_t^{}=(\begin{array}{ccccccccccccccccccccccccc}p_1& p_2& p_3& p_4& p_5& p_6& p_7& p_8& p_9& p_{10}& p_{11}& p_{12}& p_{13}& p_{14}& p_{15}& p_{16}& p_{17}& p_{18}& p_{19}& p_{20}& p_{21}& p_{22}& p_{23}& p_{24}& \\ 1& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 1& 0& 0& 0& 0& 0& 0& 1& 0& 0& 0& 0\\ 0& 1& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 1& 0& 0& 0& 0& 0& 0& 1& 0& 0& 0& 0\\ 0& 0& 1& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 1& 0& 0& 0& 0& 0& 0& 1& 0& 0& 0& 0\\ 0& 0& 0& 1& 0& 0& 0& 0& 0& 0& 0& 0& 0& 1& 0& 0& 0& 0& 0& 0& 1& 0& 0& 0& 0\\ 0& 0& 0& 0& 1& 0& 0& 0& 0& 0& 0& 0& 0& 1& 0& 0& 0& 0& 0& 0& 1& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 1& 0& 0& 0& 0& 0& 0& 0& 1& 0& 0& 0& 0& 0& 0& 1& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 1& 0& 0& 0& 0& 0& 0& 1& 0& 0& 0& 0& 0& 0& 1& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 1& 0& 0& 0& 0& 0& 1& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 1& 0& 0& 0& 0& 1& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 1& 0& 0& 0& 1& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 1& 0& 0& 1& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 1& 0& 1& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 1& 1& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 1& 0& 0& 0& 0& 0& 1& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 1& 0& 0& 0& 0& 1& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 1& 0& 0& 0& 1& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 1& 0& 0& 1& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 1& 0& 1& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 1& 1& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 1& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 1& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 1& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 1& 1& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& \zeta _1\\ 0& 1& 1& 1& 2& 1& 0& 1& 0& 1& 0& 0& 0& 0& 1& 1& 0& 0& 0& 0& 0& 0& 0& 0& \zeta _2\\ 0& 1& 0& 0& 0& 1& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& \zeta _3\\ 1& 2& 1& 0& 2& 2& 0& 1& 0& 1& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& \zeta _4\\ 1& 1& 1& 1& 0& 1& 0& 1& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& \zeta _5\\ 0& 0& 0& 1& 2& 1& 0& 1& 0& 1& 0& 0& 0& 0& 0& 1& 0& 0& 0& 0& 0& 1& 0& 0& \zeta _6\\ 0& 1& 1& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& \zeta _7\\ 0& 1& 1& 0& 2& 1& 0& 1& 0& 1& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 1& 0& 0& \zeta _8\end{array}\mathrm{}\mathrm{}\hfill \\ \\ \\ \mathrm{}\mathrm{}\begin{array}{ccccccccccccccccccc}p_{25}& p_{26}& p_{27}& p_{28}& p_{29}& p_{30}& p_{31}& p_{32}& p_{33}& p_{34}& p_{35}& p_{36}& p_{37}& p_{38}& p_{39}& p_{40}& p_{41}& p_{42}& \\ 1& 0& 0& 0& 0& 0& 0& 0& 0& 1& 1& 0& 0& 0& 1& 1& 1& 1& 0\\ 1& 0& 0& 0& 0& 0& 0& 0& 1& 0& 1& 0& 0& 0& 1& 1& 1& 1& 0\\ 1& 0& 0& 0& 0& 0& 0& 1& 2& 0& 1& 0& 0& 0& 2& 2& 2& 2& 0\\ 1& 0& 0& 0& 0& 0& 0& 1& 1& 1& 1& 0& 0& 0& 2& 1& 2& 1& 0\\ 2& 0& 0& 0& 0& 0& 0& 1& 2& 1& 2& 0& 0& 0& 3& 2& 2& 2& 0\\ 2& 0& 0& 0& 0& 0& 0& 0& 2& 0& 1& 0& 0& 0& 1& 1& 1& 1& 0\\ 2& 0& 0& 0& 0& 0& 0& 0& 1& 1& 1& 0& 0& 0& 1& 1& 1& 1& 0\\ 1& 0& 0& 0& 0& 0& 0& 0& 1& 0& 0& 0& 0& 0& 0& 1& 0& 1& 0\\ 1& 0& 0& 0& 0& 0& 0& 0& 0& 1& 1& 0& 0& 0& 1& 1& 0& 1& 0\\ 1& 0& 0& 0& 0& 0& 0& 1& 1& 0& 1& 0& 0& 0& 2& 1& 1& 2& 0\\ 1& 0& 0& 0& 0& 0& 0& 1& 2& 1& 0& 0& 0& 0& 1& 1& 1& 2& 0\\ 0& 0& 0& 0& 0& 0& 0& 1& 1& 0& 0& 0& 0& 0& 1& 1& 1& 1& 0\\ 0& 0& 0& 0& 0& 0& 0& 1& 0& 0& 0& 0& 0& 0& 1& 0& 1& 1& 0\\ 1& 0& 0& 0& 0& 0& 0& 0& 1& 0& 0& 0& 0& 0& 1& 1& 1& 1& 0\\ 1& 0& 0& 0& 0& 0& 0& 0& 1& 1& 1& 0& 0& 0& 1& 2& 0& 1& 0\\ 1& 0& 0& 0& 0& 0& 0& 1& 0& 1& 1& 0& 0& 0& 1& 1& 0& 0& 0\\ 1& 0& 0& 0& 0& 0& 0& 1& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 1& 1& 0& 0& 0& 1& 1& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 1& 0& 0& 0& 0& 0& 1& 0& 0& 0\\ 1& 0& 0& 0& 0& 0& 0& 0& 0& 1& 1& 0& 0& 0& 1& 1& 0& 0& 0\\ 1& 0& 0& 0& 0& 0& 0& 0& 1& 0& 1& 0& 0& 0& 1& 0& 1& 0& 0\\ 1& 0& 0& 0& 0& 0& 0& 0& 1& 1& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 1& 0& 0& 0& 0& 0& 1& 0& 0& 1& 0& 0& 0& 1& 1& 0& 1& 0\\ 0& 0& 1& 0& 0& 0& 0& 1& 1& 1& 0& 0& 0& 0& 0& 1& 0& 1& 0\\ 0& 0& 0& 1& 0& 0& 0& 0& 1& 1& 1& 0& 0& 0& 0& 1& 1& 0& 0\\ 0& 0& 0& 0& 1& 0& 0& 0& 0& 0& 0& 0& 0& 0& 1& 1& 1& 0& 0\\ 0& 0& 0& 0& 0& 1& 0& 1& 0& 1& 1& 0& 0& 0& 1& 1& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 1& 1& 1& 0& 1& 0& 0& 0& 1& 1& 1& 1& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 1& 0& 0& 0& 1& 1& 1& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 1& 0& 1& 1& 0& 1& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 1& 1& 0& 1& 1& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& \zeta _1\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& \zeta _2\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& \zeta _3\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& \zeta _4\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& \zeta _5\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& \zeta _6\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& \zeta _7\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& \zeta _8\end{array})\hfill \end{array}$$
According to our algorithm, we must perform Gaussian row-reduction on $`Q_t^{}`$ to solve for 42 variables $`x_i`$. When this is done we find that we can in fact express all variables in terms of 3 $`x_i`$’s together with the 8 FI-parametres $`\zeta _i`$. We choose these three $`x_i`$’s to be $`x_{10,29,36}`$ corresponding to the 3 outer vertices which we know must be resolved in going from $`\text{ }\mathrm{C}^3/(\mathrm{ZZ}_3\times \mathrm{ZZ}_3)`$ to del Pezzo 1.
Next we select the fields which must be kept and set them to zero in order to determine the range for $`\zeta _i`$. Bearing in mind the toric diagrams from Figure 4, these fields we judiciously select to be: $`p_{13,8,37,38}`$. Setting $`x_{13,8,37,38}=0`$ gives us the solution $`\{\zeta _6=0;x_{29}=\zeta _7=\zeta _3=\zeta _1\zeta _5;x_{10}=\zeta _4+\zeta _5+\zeta _3;x_{36}=\zeta _7\zeta _8\}`$, which upon back-substitution to the solutions $`x_i`$ we obtained from $`Q_t^{}`$, gives zero for $`x_{13,8,37,38}`$ (which we have chosen by construction) as well as $`x_{7,14,17,32}`$; for all others we obtain positive values. This means precisely that all the other fields are to be eliminated and these 8 columns { 13, 8, 37, 38, 7,14,17,32 } are to be kept while the remaining 42-8=34 are to be eliminated from $`Q_t^{}`$ upon row-reduction to give $`Q_t`$. In other words, we have found our set $`B`$ to be {1,2,3,4,5,6,9,10,11,12,15,16,18,19,20,21,22,23,24,25, 26,27,28,29,30,31,33,34,35,36,39,40,41,42} and thus according to (3.5) we immediately obtain
$$Q_t=\left(\begin{array}{ccccccccc}p_7& p_8& p_{13}& p_{14}& p_{17}& p_{32}& p_{37}& p_{38}& \\ 1& 0& 0& 0& 0& 1& 0& 0& \zeta _2+\zeta _8\\ 0& 0& 0& 0& 1& 1& 0& 0& \zeta _6\\ 1& 0& 0& 1& 0& 0& 0& 0& \zeta _1+\zeta _3+\zeta _5\\ 0& 0& 1& 1& 0& 1& 0& 1& 0\\ 1& 1& 1& 1& 1& 0& 1& 0& 0\end{array}\right).$$
We note of course that 5 out of the 8 FI-parametres have been eliminated automatically; this is to be expected since in resolving $`\text{ }\mathrm{C}^3/(\mathrm{ZZ}_3\times \mathrm{ZZ}_3)`$ to del Pezzo 1, we remove precisely 5 nodes. Obtaining the D-terms and F-terms is now straight-forward. Using (3.3) and re-inserting the last row we obtain the D-term equations (incidence matrix) to be
$$d=\left(\begin{array}{cccccccccc}X_1& X_2& X_3& X_4& X_5& X_6& X_7& X_8& X_9& X_{10}\\ 1& 0& 0& 1& 0& 0& 0& 1& 0& 1\\ 1& 1& 0& 0& 0& 1& 0& 0& 1& 0\\ 0& 0& 1& 0& 1& 0& 1& 1& 1& 1\\ 0& 1& 1& 1& 1& 1& 1& 0& 0& 0\end{array}\right)$$
From this matrix we immediately observe that there are 4 gauge groups, i.e., $`U(1)^4`$ with 10 matter fields $`X_i`$ which we have labelled in the matrix above. In an equivalent notation we rewrite $`d`$ as the adjacency matrix of the quiver diagram (see Figure 5) for the gauge theory:
$$a_{ij}=\left(\begin{array}{cccc}0& 0& 2& 0\\ 1& 0& 1& 0\\ 0& 0& 0& 3\\ 1& 2& 0& 0\end{array}\right).$$
The K-matrix we obtain to be:
$$K^t=\left(\begin{array}{cccccccccc}X_1& X_2& X_3& X_4& X_5& X_6& X_7& X_8& X_9& X_{10}\\ 1& 0& 1& 0& 0& 0& 1& 0& 0& 0\\ 0& 1& 1& 0& 0& 0& 0& 1& 0& 0\\ 1& 0& 0& 1& 0& 0& 0& 0& 1& 0\\ 0& 1& 0& 1& 0& 1& 0& 0& 0& 0\\ 0& 0& 1& 0& 1& 0& 1& 0& 0& 0\\ 0& 0& 0& 0& 0& 1& 1& 0& 0& 1\end{array}\right)$$
which indicates that the original 10 fields $`X_i`$ can be expressed in terms of 6. This was actually addressed in the previous section and we rewrite that pleasant superpotential here:
$$W=X_2X_7X_9X_3X_6X_9X_4X_8X_7X_1X_2X_5X_{10}+X_3X_4X_{10}+X_1X_5X_6X_8.$$
del Pezzo 2: Having obtained the gauge theory for del Pezzo 1, we now repeat the above analysis for del Pezzo 2. Now we have the FI-parametres restricted as $`\{p_{36}=\zeta _2=0;\zeta _3=\zeta _4;x_{29}=\zeta _4+\zeta _6;x_{10}=\zeta _1+\zeta _4\}`$, making the set to be eliminated as $`B=\{`$ 1, 2, 3, 5, 6, 10, 11, 13, 16, 17, 19, 20, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 38, 39, 40, 41, 42 $`\}`$. Whence, we obtain
$$Q_t=\left(\begin{array}{cccccccccccc}p_4& p_7& p_8& p_9& p_{12}& p_{14}& p_{15}& p_{18}& p_{21}& p_{36}& p_{37}& \\ 0& 1& 0& 0& 0& 0& 0& 0& 1& 0& 0& \zeta _4+\zeta _6+\zeta _8\\ 1& 1& 1& 0& 0& 1& 0& 0& 0& 0& 0& \zeta _7\\ 0& 1& 0& 0& 0& 1& 0& 0& 0& 0& 0& \zeta _1+\zeta _3+\zeta _5\\ 1& 1& 1& 0& 0& 1& 1& 0& 1& 0& 0& \zeta _2\\ 0& 1& 0& 1& 0& 0& 1& 0& 0& 0& 1& 0\\ 0& 1& 1& 1& 0& 1& 0& 0& 1& 1& 0& 0\\ 1& 1& 1& 0& 0& 1& 1& 1& 0& 0& 0& 0\\ 1& 1& 1& 0& 1& 0& 0& 0& 0& 0& 0& 0\end{array}\right),$$
and observe that 4 D-terms have been resolved, as 4 nodes have been eliminated from $`\text{ }\mathrm{C}^3/(\mathrm{ZZ}_3\times \mathrm{ZZ}_3)`$. From this we easily extract (see Figure 5)
$$d=\left(\begin{array}{ccccccccccccc}X_1& X_2& X_3& X_4& X_5& X_6& X_7& X_8& X_9& X_{10}& X_{11}& X_{12}& X_{13}\\ 1& 0& 0& 1& 0& 1& 0& 1& 0& 0& 0& 1& 1\\ 0& 0& 1& 0& 1& 1& 0& 0& 0& 1& 0& 0& 0\\ 0& 0& 1& 0& 1& 0& 1& 1& 1& 0& 1& 1& 1\\ 1& 1& 0& 0& 0& 0& 0& 0& 1& 1& 0& 0& 0\\ 0& 1& 0& 1& 0& 0& 1& 0& 0& 0& 1& 0& 0\end{array}\right);$$
moreover, we integrate the F-term matrices
$$K^t=\left(\begin{array}{ccccccccccccc}X_1& X_2& X_3& X_4& X_5& X_6& X_7& X_8& X_9& X_{10}& X_{11}& X_{12}& X_{13}\\ 0& 1& 1& 0& 0& 0& 0& 1& 0& 0& 0& 1& 0\\ 1& 0& 1& 0& 0& 0& 0& 0& 0& 0& 1& 1& 0\\ 1& 0& 0& 1& 0& 1& 0& 0& 1& 0& 0& 0& 0\\ 0& 1& 0& 1& 0& 1& 0& 0& 0& 1& 0& 0& 0\\ 0& 1& 1& 1& 1& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 1& 0& 1& 0& 1& 0& 0& 0& 1& 0& 0\\ 0& 0& 0& 1& 1& 0& 0& 0& 1& 0& 0& 0& 1\end{array}\right)$$
to obtain the superpotential
$$\begin{array}{c}W=X_2X_9X_{11}X_9X_3X_{10}X_4X_8X_{11}X_1X_2X_7X_{13}+X_{13}X_3X_6\\ X_5X_{12}X_6+X_1X_5X_8X_{10}+X_4X_7X_{12}.\end{array}$$
del Pezzo 3: Finally, we shall proceed to treat del Pezzo 3. Here we have the range of the FI-parametres to be $`\{\zeta _1=\zeta _6=\zeta _6=0;x_{29}=\zeta _3=\zeta _5;x_{10}=\zeta _4;\zeta _2=x_{36};\zeta _8=\zeta _2\zeta _{10}\}`$, which gives the set $`B`$ as {1, 2, 3, 10, 11, 13, 16, 17, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 31, 32, 33, 34, 35, 36, 39, 40, 41, 42}, and thus according to (3.5) we immediately obtain
$$Q_t=\left(\begin{array}{cccccccccccccc}p_4& p_5& p_6& p_7& p_8& p_9& p_{12}& p_{14}& p_{15}& p_{18}& p_{30}& p_{37}& p_{38}& \\ 0& 0& 0& 1& 0& 0& 0& 0& 0& 0& 1& 0& 0& \zeta _2+\zeta _4+\zeta _8\\ 1& 0& 0& 1& 1& 0& 0& 1& 0& 0& 0& 0& 0& \zeta _7\\ 1& 0& 0& 1& 1& 0& 0& 1& 1& 0& 1& 0& 0& \zeta _6\\ 0& 0& 1& 0& 0& 0& 0& 1& 0& 0& 0& 0& 0& \zeta _3+\zeta _5\\ 0& 0& 1& 1& 0& 0& 0& 0& 0& 0& 0& 0& 0& \zeta _1\\ 0& 1& 1& 0& 0& 0& 0& 0& 0& 0& 1& 0& 1& 0\\ 1& 1& 1& 0& 0& 0& 0& 1& 1& 0& 0& 1& 0& 0\\ 1& 0& 0& 1& 1& 0& 0& 1& 1& 1& 0& 0& 0& 0\\ 1& 0& 0& 1& 1& 0& 1& 0& 0& 0& 0& 0& 0& 0\\ 1& 1& 1& 1& 0& 1& 0& 1& 0& 0& 0& 0& 0& 0\end{array}\right)$$
We note indeed that 3 out of the 8 FI-parametres have been automatically resolved, as we have removed 3 nodes from the toric diagram for $`\text{ }\mathrm{C}^3/(\mathrm{ZZ}_3\times \mathrm{ZZ}_3)`$. The matter content (see Figure 5) is encoded in
$$d=\left(\begin{array}{cccccccccccccc}X_1& X_2& X_3& X_4& X_5& X_6& X_7& X_8& X_9& X_{10}& X_{11}& X_{12}& X_{13}& X_{14}\\ 1& 0& 0& 0& 1& 0& 0& 1& 1& 0& 0& 1& 1& 0\\ 0& 0& 1& 1& 0& 1& 0& 0& 0& 0& 0& 0& 1& 0\\ 1& 1& 0& 1& 0& 0& 0& 0& 0& 0& 0& 0& 0& 1\\ 0& 0& 1& 0& 0& 0& 0& 1& 0& 1& 1& 0& 0& 0\\ 0& 0& 0& 0& 1& 1& 1& 0& 0& 1& 0& 1& 0& 1\\ 0& 1& 0& 0& 0& 0& 1& 0& 1& 0& 1& 0& 0& 0\end{array}\right),$$
and from the F-terms
$$K^t=\left(\begin{array}{cccccccccccccc}X_1& X_2& X_3& X_4& X_5& X_6& X_7& X_8& X_9& X_{10}& X_{11}& X_{12}& X_{13}& X_{14}\\ 1& 0& 0& 0& 0& 1& 1& 1& 0& 0& 0& 0& 0& 0\\ 0& 1& 0& 0& 0& 1& 0& 1& 0& 0& 0& 1& 0& 0\\ 1& 0& 0& 0& 0& 0& 0& 0& 1& 0& 0& 0& 1& 1\\ 0& 1& 0& 1& 0& 0& 0& 0& 1& 0& 0& 0& 1& 0\\ 0& 1& 1& 0& 0& 1& 0& 0& 1& 0& 0& 0& 0& 0\\ 0& 0& 1& 0& 0& 1& 1& 0& 0& 0& 1& 0& 0& 0\\ 0& 0& 1& 0& 1& 0& 0& 0& 1& 0& 0& 0& 0& 1\\ 0& 0& 0& 0& 0& 1& 1& 1& 0& 1& 0& 0& 0& 0\end{array}\right)$$
we integrate to obtain the superpotential
$$\begin{array}{c}W=X_3X_8X_{13}X_8X_9X_{11}X_5X_6X_{13}X_1X_3X_4X_{10}X_{12}\\ +X_7X_9X_{12}+X_1X_2X_5X_{10}X_{11}+X_4X_6X_{14}X_2X_7X_{14}.\end{array}$$
Note that we have a quintic term in $`W`$; this is an interesting interaction indeed.
del Pezzo 0: Before proceeding further, let us attempt one more example, viz., the degenerate case of the del Pezzo 0 as shown in Figure 4. This time we note that the ranges for the FI-parametres are $`\{\zeta _5=x_{29}+\zeta _6A;\zeta _6=x_{29}B;x_{29}=B+C;\zeta _8=x_{36}+B;x_{36}=B+C+D;x_{10}=A+E\}`$ for some positive $`A,B,C,D`$ and $`E`$, that $`B=`$ { 1, 2, 3, 4, 5, 6, 9, 10, 11, 12, 15, 16, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 38, 39, 40, 41, 42 } and whence the charge matrix is
$$Q_t=\left(\begin{array}{ccccccc}p_7& p_8& p_{13}& p_{14}& p_{17}& p_{37}& \\ 1& 0& 0& 0& 1& 0& \zeta _2+\zeta _6+\zeta _8\\ 1& 0& 0& 1& 0& 0& \zeta _1+\zeta _3+\zeta _5\\ 1& 1& 1& 1& 1& 1& 0\end{array}\right).$$
We extract the matter content (see Figure 5) as $`d=\left(\begin{array}{ccccccccc}X_1& X_2& X_3& X_4& X_5& X_6& X_7& X_8& X_9\\ 1& 0& 1& 0& 1& 0& 1& 1& 1\\ 0& 1& 0& 1& 0& 1& 1& 1& 1\\ 1& 1& 1& 1& 1& 1& 0& 0& 0\end{array}\right),`$ and the F-terms as $`K^t=\left(\begin{array}{ccccccccc}X_1& X_2& X_3& X_4& X_5& X_6& X_7& X_8& X_9\\ 1& 1& 0& 0& 0& 0& 1& 0& 0\\ 1& 0& 1& 0& 1& 0& 0& 0& 0\\ 0& 1& 0& 1& 0& 1& 0& 0& 0\\ 0& 0& 1& 1& 0& 0& 0& 1& 0\\ 0& 0& 0& 0& 1& 1& 0& 0& 1\end{array}\right),`$ and from the latter we integrate to obtain the superpotential
$$W=X_1X_4X_9X_4X_5X_7X_2X_3X_9X_1X_6X_8+X_2X_5X_8+X_3X_6X_7.$$
Of course we immediately recognise the matter content (which gives a triangular quiver which we shall summarise below in Figure 5) as well as the superpotential from equations (4.7-4.14) of ; it is simply the theory on the Abelian orbifold $`\text{ }\mathrm{C}^3/\mathrm{ZZ}_3`$ with action $`(\alpha \mathrm{ZZ}_3):(z_1,z_2,z_3)(e^{\frac{2\pi i}{3}}z_1,e^{\frac{2\pi i}{3}}z_2,e^{\frac{2\pi i}{3}}z_3)`$. Is our del Pezzo 0 then $`\text{ }\mathrm{C}^3/\mathrm{ZZ}_3`$? We could easily check from the $`G_t`$ matrix (which we recall is obtained from $`G_t^{}`$ of $`\text{ }\mathrm{C}^3/(\mathrm{ZZ}_3\times \mathrm{ZZ}_3)`$ by eliminating the columns corresponding to the set $`B`$):
$$G_t=\left(\begin{array}{cccccc}0& 1& 0& 0& 0& 1\\ 0& 1& 1& 0& 0& 0\\ 1& 1& 2& 1& 1& 0\end{array}\right).$$
These columns (up to repeat) correspond to monomials $`Z,X^1YZ,Y^1Z^2,X`$ in the polynomial ring $`\text{ }\mathrm{C}[X,Y,Z]`$. Therefore we need to find the spectrum (set of maximal ideals) of the ring $`\text{ }\mathrm{C}[Z,X^1YZ,Y^1Z^2,X]`$, which is given by the minimal polynomial relation: $`(X^1YZ)(Y^1Z^2)X=(Z)^3`$. This means, upon defining $`p=X^1YZ;q=Y^1Z^2;r=X`$ and $`s=Z`$, our del Pezzo 0 is described by $`pqr=s^3`$ as an algebraic variety in $`\text{ }\mathrm{C}^4(p,q,r,s)`$, which is precisely $`\text{ }\mathrm{C}^3/\mathrm{ZZ}_3`$. Therefore we have actually come through a full circle in resolving $`\text{ }\mathrm{C}^3/(\mathrm{ZZ}_3\times \mathrm{ZZ}_3)`$ to $`\text{ }\mathrm{C}^3/\mathrm{ZZ}_3`$ and the validity of our algorithm survives this consistency check beautifully. Moreover, since we know that our gauge theory is exactly the one which lives on a D-brane probe on $`\text{ }\mathrm{C}^3/\mathrm{ZZ}_3`$, this gives a good check for physicality: that our careful tuning of FI-parametres via canonical partial resolutions does give a physical D-brane theory at the end. We tabulate the matter content $`a_{ij}`$ and the superpotential $`W`$ for the del Pezzo surfaces below, and the quiver diagrams, in Figure 5.
$$\begin{array}{cccc}& & & \\ & \text{del Pezzo 1}& \text{del Pezzo 2}& \text{del Pezzo 3}\\ & & & \\ \text{Matter }a_{ij}=& \left(\begin{array}{cccc}0& 0& 2& 0\\ 1& 0& 1& 0\\ 0& 0& 0& 3\\ 1& 2& 0& 0\end{array}\right)& \left(\begin{array}{ccccc}0& 1& 0& 1& 1\\ 0& 0& 2& 0& 0\\ 3& 0& 0& 1& 0\\ 0& 1& 0& 0& 1\\ 0& 0& 2& 0& 0\end{array}\right)& \left(\begin{array}{cccccc}0& 0& 1& 1& 0& 1\\ 0& 0& 0& 1& 1& 0\\ 0& 1& 0& 0& 0& 1\\ 1& 0& 0& 0& 1& 0\\ 2& 0& 1& 0& 0& 0\\ 0& 0& 0& 1& 1& 0\end{array}\right)\\ & & & \\ \text{Superpotential }W=& \begin{array}{c}X_2X_7X_9X_3X_6X_9\\ X_4X_8X_7X_1X_2X_5X_{10}\\ +X_3X_4X_{10}+X_1X_5X_6X_8\end{array}& \begin{array}{c}X_2X_9X_{11}X_9X_3X_{10}\\ X_4X_8X_{11}X_1X_2X_7X_{13}\\ +X_{13}X_3X_6X_5X_{12}X_6\\ +X_1X_5X_8X_{10}+X_4X_7X_{12}\end{array}& \begin{array}{c}X_3X_8X_{13}X_8X_9X_{11}\\ X_5X_6X_{13}X_1X_3X_4X_{10}X_{12}\\ +X_7X_9X_{12}+X_1X_2X_5X_{10}X_{11}\\ +X_4X_6X_{14}X_2X_7X_{14}\end{array}\end{array}$$
$$\begin{array}{ccc}& & \\ & \text{del Pezzo 0}\mathrm{C}^3/\mathrm{ZZ}_3& \text{Hirzebruch 0}\mathrm{IP}^1\times \mathrm{IP}^1:=F_0=E_1\\ & & \\ \text{Matter }a_{ij}& \left(\begin{array}{ccc}0& 3& 0\\ 0& 0& 3\\ 3& 0& 0\end{array}\right)& \left(\begin{array}{cccc}0& 2& 0& 2\\ 0& 0& 2& 0\\ 4& 0& 0& 0\\ 0& 0& 2& 0\end{array}\right)\\ & & \\ \text{Superpotential }W& \begin{array}{c}X_1X_4X_9X_4X_5X_7\\ X_2X_3X_9X_1X_6X_8\\ +X_2X_5X_8+X_3X_6X_7\end{array}& \begin{array}{c}X_1X_8X_{10}X_3X_7X_{10}\\ X_2X_8X_9X_1X_6X_{12}\\ +X_3X_6X_{11}+X_4X_7X_9\\ +X_2X_5X_{12}X_4X_5X_{11}\end{array}\end{array}$$
Upon comparing Figure 4 and Figure 5, we notice that as we go from del Pezzo 0 to 3, the number of points in the toric diagram increases from 4 to 7, and the number of gauge groups (nodes in the quiver) increases from 3 to 6. This is consistent with the observation for $`𝒩=1`$ theories that the number of gauge groups equals the number of perimetre points (e.g., for del Pezzo 1, the four nodes 13, 8, 37 and 38) in the toric diagram. Moreover, as discussed in , the rank of the global symmetry group ($`E_i`$ for del Pezzo $`i`$) which must exist for these theories equals the number of perimetre point minus 3; it would be an intereting check indeed to see how such a symmetry manifests itself in the quivers and superpotentials.
Hirzebruch 0: Let us indulge ourselves with one more example, namely the 0th Hirzebruch surface, or simply $`\mathrm{IP}^1\times \mathrm{IP}^1:=F_0:=E_1`$. The toric diagram is drawn in Figure 4. Now the FI-parametres are $`\{\zeta _4=x_{29}x_{36}\zeta _5\zeta _8A;\zeta _5=AB;\zeta _7=x_{10}+x_{29}+x_{36}+\zeta _8C;\zeta _8=x_{10}x_{29}x_{36}+D;D=A+B;C=A+B;A=x_{10}E;x_{10}=E+F;x_{29}=B+G\}`$ for positive $`A,B,C,D,E,F`$ and $`G`$. Moreover, $`B=`$ { 1, 2, 3, 5, 6, 10, 11, 13, 16, 17, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 38, 39, 40, 41, 42 }. We note that this can be obtained directly by partial resolution of fields 21 and 36 from del Pezzo 2 as is consistent with Figure 4. Therefrom we obtain the charge matrix
$$Q_t=\left(\begin{array}{cccccccccc}p_4& p_7& p_8& p_9& p_{12}& p_{14}& p_{15}& p_{18}& p_{37}& \\ 1& 2& 1& 0& 0& 1& 1& 0& 0& \zeta _2+\zeta _4+\zeta _6+\zeta _8\\ 1& 1& 1& 0& 0& 1& 0& 0& 0& \zeta _7\\ 0& 1& 0& 0& 0& 1& 0& 0& 0& \zeta _1+\zeta _3+\zeta _5\\ 0& 1& 0& 1& 0& 0& 1& 0& 1& 0\\ 1& 1& 1& 0& 0& 1& 1& 1& 0& 0\\ 1& 1& 1& 0& 1& 0& 0& 0& 0& 0\end{array}\right),$$
from which we have the matter content $`d=\left(\begin{array}{cccccccccccc}X_1& X_2& X_3& X_4& X_5& X_6& X_7& X_8& X_9& X_{10}& X_{11}& X_{12}\\ 1& 0& 1& 0& 1& 0& 1& 1& 1& 0& 1& 1\\ 0& 1& 0& 1& 1& 0& 0& 0& 1& 0& 0& 0\\ 0& 1& 0& 1& 0& 1& 1& 1& 0& 1& 1& 1\\ 1& 0& 1& 0& 0& 1& 0& 0& 0& 1& 0& 0\end{array}\right)`$ the quiver for which is presented in Figure 5. The F-terms are
$$K^t=\left(\begin{array}{cccccccccccc}X_1& X_2& X_3& X_4& X_5& X_6& X_7& X_8& X_9& X_{10}& X_{11}& X_{12}\\ 1& 1& 0& 0& 0& 0& 1& 0& 0& 0& 1& 0\\ 1& 0& 1& 0& 1& 0& 0& 0& 1& 0& 0& 0\\ 1& 1& 1& 1& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 1& 0& 1& 0& 1& 0& 0& 0& 1& 0& 0\\ 0& 0& 1& 1& 0& 0& 0& 1& 0& 0& 0& 1\\ 0& 0& 0& 0& 1& 1& 1& 1& 0& 0& 0& 0\end{array}\right),$$
from which we obtain
$$W=X_1X_8X_{10}X_3X_7X_{10}X_2X_8X_9X_1X_6X_{12}+X_3X_6X_{11}+X_4X_7X_9+X_2X_5X_{12}X_4X_5X_{11},$$
a perfectly acceptable superpotential with only cubic interactions. We include these results with our table above.
## 5 Uniqueness?
In our foregoing discussion we have constructed in detail an algorithm which calculates the matter content encoded by $`\mathrm{\Delta }`$ and superpotential encoded in $`K`$, given the toric diagram of the singularity which the D-branes probe. As abovementioned, though this algorithm gives one solution for the quiver and the $`K`$-matrix once the matrix $`Q_t`$ is determined, the general inverse process of going from toric data to gauge theory data, is highly non-unique and a classification of all possible theories having the same toric description would be interesting<sup>13</sup><sup>13</sup>13We thank R. Plesser for pointing this issue out to us.. Indeed, by the very structure of our algorithm, in immediately appealing to the partial resolution of gauge theories on $`\mathrm{ZZ}_n\times \mathrm{ZZ}_n`$ orbifolds which are well-studied, we have granted ourselves enough extraneous information to determine a unique $`Q_t`$ and hence the ability to proceed with ease (this was the very reason for our devising the algorithm).
However, generically we do not have any such luxury. At the end of subsection 3.1, we have already mentioned two types of ambiguities in the inverse problem. Let us refresh our minds. They were (A) the F-D ambiguity which is the inability to decide, simply by observing the toric diagram, which rows of the charge matrix $`Q_t`$ are D-terms and which are F-terms and (B) the repetition ambiguity which is the inability to decide which columns of $`G_t`$ to repeat once having read the vectors from the toric diagram. Other ambiguities exist, such as in each time when we compute nullspaces, but we shall here discuss to how ambiguities (A) and (B) manifest themselves and provide examples of vastly different gauge theories having the same toric description. There is another point which we wish to emphasise: as mentioned at the end of subsection 3.1, the resolution method guarantees, upon careful tuning of the FI-parametres, that the resulting gauge theory does originate from the world-volume of a D-brane probe. Now of course, by taking liberties with experimentation of these ambiguities we are no longer protected by physicality and in general the theories no longer live on the D-brane. It would be a truly interesting exercise to check which of these different theories do.
F-D Ambiguity: First, we demonstrate type (A) by returning to our old friend the SPP whose charge matrix we had earlier presented. Now we write the same matrix without specifying the FI-parametres:
$$Q_t=\left(\begin{array}{cccccc}1& 1& 1& 0& 1& 0\\ 1& 1& 0& 1& 0& 1\\ 1& 0& 0& 1& 1& 1\end{array}\right)$$
We could apply the last steps of our algorithm to this matrix as follows.
1. If we treat the first row as $`Q`$ (the F-terms) and the second and third as $`VU`$ (the D-terms) we obtain the gauge theory as discussed in subsection 3.3 and in .
2. If we treat the second row as $`Q`$ and first with the third as $`VU`$, we obtain $`d=\left(\begin{array}{cccccc}1& 0& 1& 1& 1& 0\\ 1& 0& 0& 1& 2& 1\\ 0& 0& 1& 0& 1& 1\end{array}\right)`$ which is an exotic theory indeed with a field ($`p_5`$) charged under three gauge groups.
Let us digress a moment to address the stringency of the requirements upon matter contents. By the very nature of finite group representations, any orbifold theory must give rise to only adjoints and bi-fundamentals because its matter content is encodable by an adjacency matrix due to tensors of representations of finite groups. The corresponding incidence matrix $`d`$, has (a) only 0 and $`\pm 1`$ entries specifying the particular bi-fundamentals and (b) has each column containing precisely one 1, one $`1`$ and with the remaining entries 0. However more exotic matter contents could arise from more generic toric singularities, such as fields charged under 3 or more gauge group factors; these would then have $`d`$ matrices with conditions (a) and (b) relaxed<sup>14</sup><sup>14</sup>14Note that we still require that each column sums to 0 so as to be able to factor out an overall $`U(1)`$.. Such exotic quivers (if we could even call them quivers still) would give interesting enrichment to those well-classified families as discussed in .
Moreover we must check the anomaly cancellation conditions. These could be rather involved; even though for $`U(1)`$ theories they are a little simpler, we still need to check trace anomalies and cubic anomalies. In a trace-anomaly-free theory, for each node in the quiver, the number of incoming arrows must equal the number of outgoing (this is true for a $`U(1)`$ theory which is what toric varieties provide; for a discussion on this see e.g. ). In matrix language this means that each row of $`d`$ must sum to 0.
Now for a theory with only bi-fundamental matter with $`\pm 1`$ charges, since $`(\pm 1)^3=\pm 1`$, the cubic is equal to the trace anamaly; therefore for these theories we need only check the above row-condition for $`d`$. For more exotic matter content, which we shall meet later, we do need to perform an independent cubic-anomaly check.
Now for the above $`d`$, the second row does not sum to zero and whence we do unfortunately have a problematic anomalous theory. Let us push on to see whether we have better luck in the following.
3. Treating row 3 as the F-terms and the other two as the D-terms gives
$`d=\left(\begin{array}{cccccc}0& 1& 1& 1& 1& 0\\ 0& 1& 0& 1& 2& 1\\ 0& 0& 1& 0& 1& 1\end{array}\right)`$ which has the same anomaly problem as the one above.
4. Now let rows 1 and 2 as the F-terms and the 3rd, as the D-terms, we obtain $`d=\left(\begin{array}{ccccc}X_1& X_2& X_3& X_4& X_5\\ 0& 1& 1& 1& 1\\ 0& 1& 1& 1& 1\end{array}\right)`$, which is a perfectly reasonable matter content. Integrating $`K=\left(\begin{array}{ccccc}1& 0& 1& 0& 0\\ 0& 1& 0& 1& 0\\ 1& 0& 0& 1& 0\\ 0& 0& 1& 0& 1\end{array}\right)`$ gives the superpotential $`W=\varphi (X_1X_2X_5X_3X_4)`$ for some field $`\varphi `$ of charge $`(0,0)`$ (which could be an adjoint for example; note however that we can not use $`X_1`$ even though it has charge $`(0,0)`$ for otherwise the F-terms would be altered). This theory is perfectly legitimate. We compare the quiver diagrams of theories (a) (which we recall from Figure 3) and this present example in Figure 6. As a check, let us define the gauge invariant quantities: $`a=X_2X_4`$, $`b=X_2X_5`$, $`c=X_3X_4`$, $`d=X_3X_5`$ and $`e=X_1`$. Then we have the algebraic relations $`ad=bc`$ and $`eb=c`$, from which we immediately obtain $`ad=eb^2`$, precisely the equation for the SPP.
5. As a permutation on the above, treating rows 1 and 3 as the F-terms gives a theory equivalent thereto.
6. Furthermore, we could let rows 2 and 3 be $`Q`$ giving us $`d=\left(\begin{array}{ccccc}0& 1& 1& 1& 1\\ 0& 1& 1& 1& 1\end{array}\right)`$, but this again gives an anomalous matter content.
7. Finally, though we cannot treat all rows as F-terms, we can however treat all of them as D-terms in which $`Q_t`$ is simply $`\mathrm{\Delta }`$ as remarked at the end of Section 2 before the flow chart. In this case we have the matter content $`d=\left(\begin{array}{cccccc}1& 1& 1& 0& 1& 0\\ 1& 1& 0& 1& 0& 1\\ 1& 0& 0& 1& 1& 1\\ 1& 0& 1& 0& 0& 0\end{array}\right)`$ which clearly is both trace-anomaly free (each row adds to zero) and cubic-anomaly-free (the cube-sum of the each row is also zero). The superpotential, by our very choice, is of course zero. Thus we have a perfectly legitimate theory without superpotential but with an exotic field (the first column) charged under 4 gauge groups.
We see therefore, from our list of examples above, that for the simple case of the SPP we have 3 rather different theories (a,d,g) with contrasting matter content and superpotential which share the same toric description.
Repetition Ambiguity: As a further illustration, let us give one example of type (B) ambiguity. First let us eliminate all repetitive columns from the $`G_t`$ of SPP, giving us:
$$G_t=\left(\begin{array}{ccccc}1& 0& 0& 1& 1\\ 1& 1& 0& 1& 0\\ 1& 1& 1& 1& 1\end{array}\right),$$
which is perfectly allowed and consistent with Figure 2. Of course many more possibilities for repeats are allowed and we could redo the following analyses for each of them. As the nullspace of our present choice of $`G_t`$, we find $`Q_t`$, and we choose, in light of the foregoing discussion, the first row to represent the D-term:
$$Q_t=\left(\begin{array}{cccccc}1& 1& 1& 0& 1& \zeta \\ 1& 2& 0& 1& 0& 0\end{array}\right).$$
Thus equipped, we immediately retrieve, using our algorithm,
$$d=\left(\begin{array}{ccccc}X_1& X_2& X_3& X_4& X_5\\ 1& 1& 1& 1& 0\\ 1& 1& 1& 1& 0\end{array}\right)K^t=\left(\begin{array}{ccccc}1& 0& 0& 0& 0\\ 0& 0& 2& 0& 1\\ 0& 1& 0& 0& 0\\ 0& 0& 1& 1& 1\end{array}\right)T=\left(\begin{array}{ccccc}0& 0& 0& 0& 1\\ 1& 0& 0& 1& 0\\ 0& 0& 1& 0& 0\\ 2& 1& 0& 0& 0\end{array}\right).$$
We see that $`d`$ passes our anomaly test, with the same bi-fundamental matter content as theory (d). The superpotential can be read easily from $`K`$ (since there is only one relation) as $`W=\varphi (X_5^2X_3X_4)`$. As a check, let us define the gauge invariant quantities: $`a=X_1X_2`$, $`b=X_1X_4`$, $`c=X_3X_2`$, $`d=X_3X_4`$ and $`e=X_5`$. These have among themselves the algebraic relations $`ad=bc`$ and $`e^2=d`$, from which we immediately obtain $`bc=ae^2`$, the equation for the SPP. Hence we have yet another interesting anomaly free theory, which together with our theories (a), (d) and (g) above, shares the toric description of the SPP.
Finally, let us indulge in one more demonstration. Now let us treat both rows of our $`Q_t`$ as D-terms, whereby giving a theory with no superpotential and the exotic matter content $`d=\left(\begin{array}{ccccc}1& 1& 1& 0& 1\\ 1& 2& 0& 1& 0\\ 0& 1& 1& 1& 1\end{array}\right)`$ with a field (column 2) charged under 3 gauge groups. Indeed though the rows sum to 0 and trace-anomaly is avoided, the cube-sum of the second row gives $`1^3+1^3+(2)^3=6`$ and we do have a cubic anomaly.
In summary, we have an interesting phenomenon indeed! Taking so immediate an advantage of the ambiguities in the above has already produced quite a few examples of vastly different gauge theories flowing in the IR to the same universality class by having their moduli spaces identical. The vigilant reader may raise two issues. First, as mentioned earlier, one may take the pains to check whether these theories do indeed live on a D-brane. Necessary conditions such as that the theories may be obtained from an $`𝒩=4`$ theory must be satisfied. Second, the matching of moduli spaces may not seem so strong since they are on a classical level. However, since we are dealing with product $`U(1)`$ gauge groups (which is what toric geometry is capable to dealing with so far), the classical moduli receive no quantum corrections<sup>15</sup><sup>15</sup>15We thank K. Intriligator for pointing this out.. Therefore the matching of the moduli for these various theories do persist to the quantum regime, which hints at some kind of “duality” in the field theory. We shall call such a duality toric duality. It would be interesting to investigate how, with non-Abelian versions of the theory (either by brane setups or stacks of D-brane probes), this toric duality may be extended.
## 6 Conclusions and Prospects
The study of resolution of toric singularities by D-branes is by now standard. In the concatenation of the F-terms and D-terms from the world volume gauge theory of a single D-brane at the singularity, the moduli space could be captured by the algebraic data of the toric variety. However, unlike the orbifold theories, the inverse problem where specifying the structure of the singularity specifies the physical theory has not yet been addressed in detail.
We recognise that in contrast with D-brane probing orbifolds, where knowing the group structure and its space-time action uniquely dictates the matter content and superpotential, such flexibility is not shared by generic toric varieties due to the highly non-unique nature of the inverse problem. It has been the purpose and main content of the current writing to device an algorithm which constructs the matter content (the incidence matrix $`d`$) and the interaction (the F-term matrix $`K`$) of a well-behaved gauge theory given the toric diagram $`D`$ of the singularity at hand.
By embedding $`D`$ into the Abelian orbifold $`\text{ }\mathrm{C}^k/(\mathrm{ZZ}_n)^{k1}`$ and performing the standard partial resolution techniques, we have investigated how the induced action upon the charge matrices corresponding to the toric data of the latter gives us a convenient charge matrix for $`D`$ and have constructed a programmatic methodology to extract the matter content and superpotential of one D-brane world volume gauge theory probing $`D`$. The theory we construct, having its origin from an orbifold, is nicely behaved in that it is anomaly free, with bi-fundamentals only and well-defined superpotentials. As illustrations we have tabulated the results for all the toric del Pezzo surfaces and the zeroth Hirzebruch surface.
Directions of further work are immediately clear to us. From the patterns emerging from del Pezzo surfaces 0 to 3, we could speculate the physics of higher (non-toric) del Pezzo cases. For example, we expect del Pezzo $`n`$ to have $`n+3`$ gauge groups. Moreover, we could attempt to fathom how our resolution techniques translate as Higgsing in brane setups, perhaps with recourse to diamonds, and realise the various theories on toric varieties as brane configurations.
Indeed, as mentioned, the inverse problem is highly non-unique; we could presumably attempt to classify all the different theories sharing the same toric singularity as their moduli space. In light of this, we have addressed two types of ambiguity: that in having multiple fields assigned to the same node in the toric diagram and that of distinguishing the F-terms and D-terms in the charge matrix. In particular we have turned this ambiguity to a matter of interest and have shown, using our algorithm, how vastly different theories, some with quite exotic matter content, may have the same toric description. This commonality would correspond to a duality wherein different gauge theories flow to the same universality class in the IR. We call this phenomenon toric duality. It would be interesting indeed how this duality may manifest itself as motions of branes in the corresponding setups. Without further ado however, let us pause here awhile and leave such investigations to forthcoming work.
## Appendix: Finding the Dual Cone
Let us be given a convex polytope $`C`$, with the edges specifying the faces of which given by the matrix $`M`$ whose columns are the vectors corresponding to these edges. Our task is to find the dual cone $`\stackrel{~}{C}`$ of $`C`$, or more precisely the matrix $`N`$ such that
$$N^tM0\text{for all entries.}$$
There is a standard algorithm, given in . Let $`M`$ be $`n\times p`$, i.e., there are $`p`$ $`n`$-dimensional vectors spanning $`C`$. We note of course that $`pn`$ for convexity. Out of the $`p`$ vectors, we choose $`n1`$. This gives us an $`n\times (n1)`$ matrix of co-rank 1, whence we can extract a 1-dimensional null-space (as indeed the initial $`p`$ vectors are all linearly independent) described by a single vector $`u`$.
Next we check the dot product of $`u`$ with the remaining $`p(n1)`$ vectors. If all the dot products are positive we keep $`u`$, and if all are negative, we keep $`u`$, otherwise we discard it.
We then select another $`n1`$ vectors and repeat the above until all combinations are exhausted. The set of vectors we have kept, $`u`$’s or $`u`$’s then form the columns of $`N`$ and span the dual cone $`\stackrel{~}{C}`$.
We note that this is a very computationally intensive algorithm, the number of steps of which depends on $`\left(\begin{array}{c}p\\ n1\end{array}\right)`$ which grows exponentially.
A subtle point to remark. In light of what we discussed in a footnote in the paper on the difference between $`𝐌_+=𝐌\sigma `$ and $`𝐌_+^{}`$, here we have computed the dual of $`\sigma `$. We must ensure that $`\mathrm{ZZ}_+`$-independent lattice points inside the cones be not missed.
## Acknowledgements
We would like to extend our sincere gratitude to the CTP of MIT for her gracious patronage as well as the Institute for Theoretical Phyics at UCSB for her warm hospitality and for hosting the “Program on Supersymmetric Gauge Dynamics and String Theory.” Furthermore, we thank K. Intriligator and J. S. Song for insightful comments. AH is grateful to M. Aganagic, D.-E. Diaconescu, A. Karch, D. Morrison and R. Plesser for valuable discussions. BF thanks A. Uranga and R. von Unge for helpful insights. YHH acknowledges V. Rodoplu of Stanford University and S. Wu of the Dept of Mathematics, UCSB for enlightening discussions and is ever indebted to M. R. Warden for inspiration and emotional support.
|
warning/0003/hep-ph0003119.html
|
ar5iv
|
text
|
# References
TECHNION-PH-00-25
EFI-2000-8
hep-ph/0003119
March 2000
THE ROLE OF $`B_sK\pi `$ IN DETERMINING THE WEAK PHASE $`\gamma `$ <sup>1</sup><sup>1</sup>1To be submitted to Physics Letters B.
Michael Gronau
Technion – Israel Institute of Technology, 32000 Haifa, Israel
and
Jonathan L. Rosner
Enrico Fermi Institute and Department of Physics
University of Chicago, Chicago, IL 60637
ABSTRACT
> The decay rates for $`B^0K^+\pi ^{}`$, $`B^+K^0\pi ^+`$, and the charge-conjugate processes were found to provide information on the weak phase $`\gamma \mathrm{Arg}(V_{ub}^{})`$ when the ratio $`r`$ of weak tree and penguin amplitudes was taken from data on $`B\pi \pi `$ or semileptonic $`B\pi `$ decays. We show here that the rates for $`B_sK^{}\pi ^+`$ and $`\overline{B}_sK^+\pi ^{}`$ can provide the necessary information on $`r`$, and estimate the statistical accuracy of forthcoming measurements at the Fermilab Tevatron.
PACS codes: 12.15.Hh, 12.15.Ji, 13.25.Hw, 14.40.Nd
The measurement of phases of the Cabibbo-Kobayashi-Maskawa (CKM) matrix is a crucial test of our understanding of CP violation. Various aspects of the decays $`BK\pi `$, in particular, have been shown to provide information on the weak phase $`\gamma \mathrm{Arg}(V_{ub}^{})`$ . In Ref. we showed that ratios of partial decay rates for charged and neutral $`B`$ mesons to $`K\pi `$ final states yielded $`\gamma `$ when supplemented with information on the ratio $`r`$ of weak tree and penguin amplitudes. It was necessary to extract $`r`$ either from data on $`B\pi \pi `$ or from semileptonic $`B\pi `$ decays.
In the present Letter we show that the decays $`B_sK^{}\pi ^+`$ and $`\overline{B}_sK^+\pi ^{}`$, which are related by U-spin to the processes $`B^0K^+\pi ^{}`$ and $`\overline{B}^0K^{}\pi ^+`$, respectively, can provide the necessary information on $`r`$. We describe the constraints available by including information on these processes, and estimate the statistical power of forthcoming measurements at the Fermilab Tevatron.
We use the same procedure as Ref. , which may be consulted for further details. However, for convenience, we shall decompose our amplitudes into “tree” and “penguin” contributions in a somewhat different manner.
We define the following charge-averaged ratios:
$$R\frac{\mathrm{\Gamma }(B^0K^+\pi ^{})+\mathrm{\Gamma }(\overline{B}^0K^{}\pi ^+)}{\mathrm{\Gamma }(B^+K^0\pi ^+)+\mathrm{\Gamma }(B^{}\overline{K}^0\pi ^{})},$$
(1)
$$R_s\frac{\mathrm{\Gamma }(B_sK^{}\pi ^+)+\mathrm{\Gamma }(\overline{B}_sK^+\pi ^{})}{\mathrm{\Gamma }(B^+K^0\pi ^+)+\mathrm{\Gamma }(B^{}\overline{K}^0\pi ^{})},$$
(2)
and CP-violating rate (pseudo-)asymmetries:
$$A_0\frac{\mathrm{\Gamma }(B^0K^+\pi ^{})\mathrm{\Gamma }(\overline{B}^0K^{}\pi ^+)}{\mathrm{\Gamma }(B^+K^0\pi ^+)+\mathrm{\Gamma }(B^{}\overline{K}^0\pi ^{})},$$
(3)
$$A_s\frac{\mathrm{\Gamma }(B_sK^{}\pi ^+)\mathrm{\Gamma }(\overline{B}_sK^+\pi ^{})}{\mathrm{\Gamma }(B^+K^0\pi ^+)+\mathrm{\Gamma }(B^{}\overline{K}^0\pi ^{})}.$$
(4)
The amplitudes for $`BK\pi `$ were expressed in Ref. in terms of “tree” and “penguin” contributions involving CKM factors $`V_{ub}^{}V_{us}`$ and $`V_{tb}^{}V_{ts}`$, respectively. Using the unitarity of the CKM matrix, it is more convenient in our case to decompose the amplitudes into terms containing $`V_{ub}^{}V_{us}`$ and $`V_{cb}^{}V_{cs}`$. This decomposition is in accordance with the structure of the $`\mathrm{\Delta }B=1,\mathrm{\Delta }C=0,\mathrm{\Delta }S=1`$ effective Hamiltonian
$$_{\mathrm{eff}}=\frac{G_F}{\sqrt{2}}\left[V_{ub}^{}V_{us}\left(\underset{1}{\overset{2}{}}c_iQ_i^{us}+\underset{3}{\overset{10}{}}c_iQ_i^s\right)+V_{cb}^{}V_{cs}\left(\underset{1}{\overset{2}{}}c_iQ_i^{cs}+\underset{3}{\overset{10}{}}c_iQ_i^s\right)\right],$$
(5)
where the flavor structure of the various short-distance operators is $`Q_{1,2}^{qs}\overline{b}q\overline{q}s,Q_{3,..,6}^s\overline{b}s\overline{q}^{}q^{},Q_{7,..,10}^s\overline{b}se_q^{}\overline{q}^{}q^{}`$.
The amplitude of $`B_sK^{}\pi ^+`$ is obtained from the one for $`B^0K^+\pi ^{}`$ by a U-spin transformation, $`sd`$, acting both on the effective Hamiltonian and on the initial and final hadronic states. The ratios of the corresponding two CKM factors occuring in these amplitudes are $`V_{ub}^{}V_{ud}/V_{ub}^{}V_{us}=1/\stackrel{~}{\lambda }`$ and $`V_{cb}^{}V_{cd}/V_{cb}^{}V_{cs}=\stackrel{~}{\lambda }`$, where $`\stackrel{~}{\lambda }=|V_{us}/V_{ud}|=|V_{cd}/V_{cs}|=\mathrm{tan}\theta _c0.226`$.
In a convention where the coefficient of the strangeness-changing penguin amplitude in $`B^+K^0\pi ^+`$ is taken to be real and positive, we then have
$$A(B^+K^0\pi ^+)=A(B^{}\overline{K}^0\pi ^{})=P+𝒪(\frac{1}{2}\stackrel{~}{\lambda }^2),$$
(6)
$$A(B^0K^+\pi ^{})=[P+Te^{i(\delta +\gamma )}],$$
$$A(\overline{B}^0K^{}\pi ^+)=[P+Te^{i(\delta \gamma )}],$$
(7)
$$A(B_sK^{}\pi ^+)=\stackrel{~}{\lambda }P(1/\stackrel{~}{\lambda })Te^{i(\delta +\gamma )},$$
$$A(\overline{B}_sK^+\pi ^{})=\stackrel{~}{\lambda }P(1/\stackrel{~}{\lambda })Te^{i(\delta \gamma )},$$
(8)
Here $`\delta `$ is the strong phase difference between the tree and penguin amplitudes. The $`𝒪(\frac{1}{2}\stackrel{~}{\lambda }^2)`$ term in Eq. (6) is the relative contribution of the $`V_{ub}^{}V_{us}`$ term in comparison with the dominant $`V_{cb}^{}V_{cs}`$ term. The effects of this term could be amplified if rescattering is important. Various estimates consider this possibility to be unlikely, but it can be checked by measuring the CP-violating rate difference between $`B^+K^0\pi ^+`$ and $`B^{}\overline{K}^0\pi ^{}`$ or by improving bounds on the charge-averaged rate of the U-spin related decay $`B^\pm K^0(\overline{K}^0)K^\pm `$. Also, when using isospin symmetry in Eqs. (6) and (7) to assume equal penguin amplitudes in $`B^+`$ and $`B^0`$ decays, we have neglected color-suppressed electroweak penguin contributions . Their effects on determining $`\gamma `$ were studied in , and ways for controlling these small terms were discussed in .
In the above equations we have taken $`P`$ and $`T`$ to be real but of indeterminate sign. Calculations based on the factorization hypothesis or free-quark estimates suggest $`T>0`$, $`P<0`$, $`\delta 0`$. We define the ratio $`rT/P`$. One then expects $`r<0,\delta 0`$ in the factorization limit. Note that this definition differs from the one in , since the present definition of $`T`$ contains a contribution from the $`V_{ub}^{}V_{uq}`$ term ($`q=d,s`$) in the penguin amplitude of Ref. .
The charge-averaged ratios and charge asymmetries are now given by
$$R=1+r^2+2r\mathrm{cos}\delta \mathrm{cos}\gamma ,$$
(9)
$$R_s=\stackrel{~}{\lambda }^2+(r/\stackrel{~}{\lambda })^22r\mathrm{cos}\delta \mathrm{cos}\gamma ,$$
(10)
$$A_0=A_s=2r\mathrm{sin}\gamma \mathrm{sin}\delta .$$
(11)
The expressions for $`R`$ and $`A_0`$ are those given in Ref. , with a sign change in the term proportional to $`r`$ corresponding to our different convention for labeling amplitudes. The expressions for $`R_s`$ and $`A_s`$ provide new information.
Notice that since both $`R_s`$ and $`A_s`$ involve the ratios of strange and nonstrange $`B`$ partial widths, their measurement in a hadronic experiment will demand a better estimate of the relative production fraction of strange and nonstrange $`B`$ mesons. (The CDF detector at the Tevatron has measured this ratio to be $`f_s/(f_u+f_d)=0.213\pm 0.068`$ .) One way to accomplish this will be to compare same-sign and opposite-sign lepton pair production .
The equation for $`R_s`$ provides an opportunity to learn the magnitude of $`r`$, related to a quantity estimated previously with the help of $`B\pi \pi `$ or $`B\pi \mathrm{}\nu _{\mathrm{}}`$ decays. Adding Eqs. (9) and (10), we find
$$R+R_s=(1+\stackrel{~}{\lambda }^2)(1+\frac{r^2}{\stackrel{~}{\lambda }^2}),\mathrm{or}|r|=\stackrel{~}{\lambda }\left[\frac{R+R_s}{1+\stackrel{~}{\lambda }^2}1\right]^{1/2}.$$
(12)
Once $`|r|`$ is known we can establish a useful bound on $`\gamma `$ independent of any possible CP-violating charge-asymmetry. We write
$$(R1r^2)^2=4r^2\mathrm{cos}^2\delta \mathrm{cos}^2\gamma 4r^2\mathrm{cos}^2\gamma ,$$
(13)
implying
$$|\mathrm{cos}\gamma |\frac{|R1r^2|}{2|r|},$$
(14)
or
$$|\mathrm{sin}\gamma |\frac{[\lambda (1,R,r^2)]^{1/2}}{2|r|},\lambda (a,b,c)a^2+b^2+c^22ab2ac2bc.$$
(15)
The limiting cases of an equality, in which $`\gamma `$ is determined up to a twofold ambiguity, $`\gamma \pi \gamma `$, correspond to $`\delta =0`$ or $`\delta =\pi `$. Assuming that the strong phase difference between $`T`$ and $`P`$ is small, as indicated by perturbative QCD arguments , measurements of $`R`$ and $`R_s`$ already determine $`\gamma `$. Note that we expect $`r<0`$ for a vanishing final-state phase $`\delta `$, in which case one has
$$\mathrm{cos}\gamma =\frac{1+r^2R}{2|r|}.$$
(16)
In case that the strong phase is large, one expects to measure nonzero asymmetries $`A_0=A_s`$, which together with $`R`$ and $`R_s`$ would determine $`\gamma `$. As quoted previously , by eliminating $`\delta `$ from the equations for $`R`$ and $`A_0`$, one finds
$$R=1+r^2\pm \sqrt{4r^2\mathrm{cos}^2\gamma A_0^2\mathrm{cot}^2\gamma },$$
(17)
which can then be solved for $`\mathrm{sin}^2\gamma `$ up to a two-fold ambiguity when $`A_00`$.
We have treated the systematic effects associated with theoretical uncertainties in our method in Ref. . There it was shown that an error on $`\gamma `$ of about $`10^{}`$ could be achieved if one obtained an error of about 10% on $`r`$ (defined in a somewhat different manner) for the value favored there, $`r=0.16`$. When obtaining $`r`$ using the measurement of $`R`$ and $`R_s`$, the statistically limiting measurement is likely to be that of $`R_s`$, since the $`B_s`$ is expected to be produced less copiously than the $`B^0`$, and the tree amplitude $`T/\stackrel{~}{\lambda }`$ dominating the decay $`B_sK^{}\pi ^+`$ is expected to be smaller than the amplitude $`P`$ dominating the decay $`B^0K^+\pi ^{}`$. In one estimate , the ratio of produced $`B_sK^{}\pi ^+`$ : $`B^0K^+\pi ^{}`$ events is expected to be 1:8.
As one example, we take $`R=1`$ (the present experimental value is $`1.01\pm 0.30`$ ), and find that $`r=0.16\pm 0.016`$ corresponds to $`R_s=0.58_{0.10}^{+0.11}`$. Thus, one must measure $`R_s`$ to about $`\pm 18\%`$ in order to achieve a 10% measurement of $`r`$. This would require about 30 events of $`B_sK^{}\pi ^+`$ or $`\overline{B}_sK^+\pi ^{}`$, whereas a sample of about 2000 to 4000 untagged events, two orders of magnitude larger, is envisioned for Run II at the Tevatron .
It is sufficient to use untagged events since the rapid $`B_s`$$`\overline{B}_s`$ mixing ensures that the effective production rates for $`B_sK^{}\pi ^+`$ and $`\overline{B}_sK^+\pi ^{}`$ will be very nearly identical at each value of rapidity. While this is not so for $`B^0K^+\pi ^{}`$ and $`\overline{B}^0K^{}\pi ^+`$, it is true in proton-antiproton collisions if one averages over rapidity in a symmetric detector. Otherwise, one must know the individual production rates for $`B^0`$ and $`\overline{B}^0`$ before mixing. These can be measured, for example using self-tagging processes such as $`B^0J/\psi K^0`$ where the CP-violating asymmetry is expected to be small, or with the help of an assumption of isospin invariance and a corresponding measurement of $`B^+`$ and $`B^{}`$ production using self-tagging modes with small CP asymmetry such as $`B^\pm J/\psi K^\pm `$ .
In order to estimate the systematic error associated with our flavor SU(3) approximation, we adopt the often used assumption of factorized tree and penguin amplitudes. In this approximation, SU(3) breaking introduces in the ratio $`A(B_sK^{}\pi ^+)/A(B^0K^+\pi ^{})`$ a relative factor $`f_1=F_{B_sK}(m_\pi ^2)f_\pi /F_{B\pi }(m_K^2)f_K`$, involving the product of corresponding ratios of form factors and decay constants describing $`B_sK\pi `$ and $`BK\pi `$. Although $`B`$ decay form factors to $`K`$ and $`\pi `$ have not yet been measured, their ratio is expected to be somewhat larger than one; quark model and QCD sum rule calculations find $`F_{B_sK}(m_\pi ^2)/F_{B\pi }(m_K^2)1.16`$. With $`f_K/f_\pi =1.22`$, SU(3) breaking is expected to introduce a systematic uncertainty of only a few percent, $`f_10.95`$.
An actual check of the SU(3) (or U-spin) approximation relies on the asymmetry prediction $`A_s=A_0`$, or, in fact, on the prediction of equal CP-violating rate differences in $`B^0K\pi `$ and $`B_sK\pi `$. Including SU(3) breaking, one expects
$$\mathrm{\Gamma }(B_sK^{}\pi ^+)\mathrm{\Gamma }(\overline{B}_sK^+\pi ^{})=f_1^2[\mathrm{\Gamma }(B^0K^+\pi ^{})\mathrm{\Gamma }(\overline{B}^0K^{}\pi ^+)].$$
(18)
With 2000 events of $`B_sK^{}\pi ^+`$ or $`\overline{B}_sK^+\pi ^{}`$, one should be able to check this prediction to a fractional accuracy of $`(2000)^{1/2}2.2\%`$ which seems adequate for a precise determination of $`f_1`$. This factor can then be included in the analysis leading to a determination of $`\gamma `$. Note, however, that a very small strong phase difference $`\delta `$ may not permit useful asymmetry measurements. (The strong phases in $`B^0K\pi `$ and $`B_sK\pi `$ differ in broken SU(3), in spite of the fact that the final states in both decays are $`K^\pm \pi ^{}`$. This is similar to the case of $`K\pi `$ states produced in Cabibbo-favored and double-Cabibbo-suppressed $`D`$ decays .)
The processes studied here, $`B^0K^+\pi ^{}`$ and $`B_sK^{}\pi ^+`$, are related by a spectator quark transformation, $`ds`$, to another pair of U-spin related processes $`B_sK^+K^{}`$ and $`B^0\pi ^+\pi ^{}`$. The amplitudes of these two pairs of processes are equal, respectively, in the limit of flavor SU(3) symmetry and neglecting smaller “exchange” and “penguin annihilation” amplitudes . These contributions are expected to be very small unless rescattering is important . This can be tested by measuring the rates for $`B^0K^+K^{}`$ or $`B_s\pi ^+\pi ^{}`$ which are due solely to exchange and penguin annihilation amplitudes. Including an SU(3) breaking factor $`f_2=F_{B_sK}(m_K^2)/F_{B\pi }(m_K^2)F_{B_sK}(m_\pi ^2)/F_{B\pi }(m_\pi ^2)1.16`$, as expected in the factorization approximation, one has
$$A(B_sK^+K^{})=f_2A(B^0K^+\pi ^{}),$$
$$A(B^0\pi ^+\pi ^{})=f_2^1A(B_sK^{}\pi ^+).$$
(19)
The rates of these four processes can be used to check the factorization assumption and to determine the SU(3) breaking factor $`f_2`$. A study of $`\gamma `$ through time-integrated rates of $`B_sK^+K^{}`$, $`B^0\pi ^+\pi ^{}`$ and their charge-conjugates is very similar and complementary to the one using $`B^0/\overline{B}^0K^\pm \pi ^{}`$ and $`B_s/\overline{B}_sK^{}\pi ^\pm `$.
The present proposal for determining $`\gamma `$ is also complementary to the method comparing time-dependent asymmetry measurements in $`B^0\pi ^+\pi ^{}`$ and in $`B_sK^+K^{}`$ . In that method, one measures both an oscillation amplitude and an oscillation phase for both channels, i.e., four quantities, and thereby extracts four unknowns: the ratio of tree and penguin amplitudes (equivalent to our $`r`$), a strong final-state phase $`\delta `$ between tree and penguin assumed to be the same for $`\pi ^+\pi ^{}`$ and $`K^+K^{}`$, and two weak phases $`\beta `$ and $`\gamma `$. One must tag the flavor of the neutral mesons at time of production.
We comment briefly on the relation between the two methods. The present method has the following advantages:
(1) It requires neither time-dependent measurements nor flavor tagging for $`B^0`$ and $`B_s`$.
(2) The kinematic peaks associated with $`B^0K\pi `$ and $`B_sK\pi `$ are well-separated from one another and reasonably well separated from those corresponding to $`B^0\pi ^+\pi ^{}`$ and $`B_sK^+K^{}`$ (which lie between them and are nearly on top of each other), even without particle identification . To see this, we note that in the limit of relativistic pions or kaons, the mass $`m_{21}`$ of a $`\pi \pi `$, $`K\pi `$, or $`K\overline{K}`$ system is given by
$$m_{12}=\left[m_1^2\left(1+\frac{p_2}{p_1}\right)+m_2^2\left(1+\frac{p_1}{p_2}\right)+2p_1p_2(1\mathrm{cos}\theta _{12})\right]^{1/2},$$
(20)
where $`p_1`$ and $`p_2`$ are laboratory momenta of the two final-state tracks and $`\theta _{12}`$ is the laboratory angle between them. If we were to call both tracks pions and neglect $`m_1^2`$ and $`m_2^2`$, we would thus estimate
$$(m_{12}^{\mathrm{est}})^2=2p_1p_2(1\mathrm{cos}\theta _{12})=m_{12}^2m_1^2\left(1+\frac{p_2}{p_1}\right)+m_2^2\left(1+\frac{p_1}{p_2}\right).$$
(21)
For the case of symmetric momenta $`p_1=p_2`$ corresponding to the Jacobian peak of the distribution, we would thus estimate
$$m_{12}^{\mathrm{est}}(m_{12}2m_1^22m_2^2)^{1/2},$$
(22)
or about 45 MeV below the $`B`$ for $`BK\pi `$, $`M_B`$ for both $`B\pi \pi `$ and $`B_sK\overline{K}`$, and about 45 MeV above the $`B`$ for $`B_sK\pi `$.
The CDF Detector at Run II of the Tevatron will have a time-of-flight system for particle identification, whose capabilities of distinguishing kaons from pions in $`B`$ decay will be minimal. It will also make use of $`dE/dx`$ discrimination. Later-generation detectors, such as LHC-b at CERN and the proposed BTeV detector at Fermilab, will have better capabilities for particle identification, allowing for cleaner separation of $`\pi \pi `$, $`K\pi `$, and $`K\overline{K}`$ final states of $`B`$ and $`B_s`$ decay.
The method based on $`B^0\pi ^+\pi ^{}`$ and $`B_sK^+K^{}`$ has the following points in its favor:
(1) It involves no small corrections from rescattering and color-suppressed electroweak penguin contributions. Also, certain ratios of form factors and decay constants cancel one another in the factorization limit, since one is only discussing asymmetries.
(2) Although the peaks for $`B\pi \pi `$ and $`B_sK\overline{K}`$ lie nearly on top of one another in the absence of particle identification, the time-dependent CP-violating rate differences in the two cases can be separated from one another (as long as they both are nonzero) by a frequency analysis. For the two asymmetries one would still need particle identification in order to distinguish between the denominators defining $`B^0`$ and $`B_s`$ asymmetries.
(3) The analysis also allows for a lifetime difference between CP-even and CP-odd $`B_s`$ mass eigenstates , which may allow resolution of some discrete ambiguities present in our scheme .
To conclude, we have found that simultaneous measurement of rates for $`B^0K^+\pi ^{}`$, $`B_sK^{}\pi ^+`$, and their charge-conjugates can provide useful information on the weak phase $`\gamma `$. This measurement appears feasible at the Fermilab Tevatron and is an appealing possibility for other facilities envisioning $`B_s`$ production, such as the HERA-B detector at DESY, the BTeV detector at Fermilab, and the LHC-b detector at CERN. Effects of SU(3) breaking and rescattering appear to be controllable and to some extent testable, while electroweak penguin effects are expected to be small enough that a determination of $`\gamma `$ to better than $`10^{}`$ appears feasible. Our method, which can also be applied to time-integrated rates of $`B^0\pi ^+\pi ^{}`$ and $`B_sK^+K^{}`$, appears complementary to that based on time-dependent asymmetries in the latter processes, for which comparable accuracy in $`\gamma `$ is expected . Whereas the latter determination of $`\gamma `$ could be affected by new physics contributions to $`B_s\overline{B}_s`$ mixing , our method is free of such modifications.
We thank S. Stone and F. Würthwein for useful discussions. M. G. would like to acknowledge the very kind hospitality of the Enrico Fermi Institute at the University of Chicago. This work was supported in part by the United States – Israel Binational Science Foundation under Research Grant Agreement 98-00237 and by the United States Department of Energy under Contract No. DE-FG02-90-ER40560.
|
warning/0003/hep-th0003147.html
|
ar5iv
|
text
|
# References
HUTP-00/A007, CITUSC/00-013
CALT68-2265, hep-th/0003147
Consistency Conditions for Holographic Duality
Vijay Balasubramanian<sup>1</sup><sup>*</sup><sup>*</sup>*vijayb@pauli.harvard.edu, Eric Gimon<sup>2</sup>egimon@theory.caltech,edu, and Djordje Minic<sup>3</sup>minic@physics.usc.edu
<sup>1</sup>Jefferson Laboratory of Physics, Harvard University,
Cambridge, MA 02138, USA.
<sup>2</sup> CIT-USC Center for Theoretical Physics
California Institute of Technology, Pasadena, CA 91125.
<sup>3</sup> CIT-USC Center for Theoretical Physics
Department of Physics and Astronomy, University of Southern California
Los Angeles, CA 90089-0484, USA.
## Abstract
We show that if the beta functions of a field theory are given by the gradient of a certain potential on the space of couplings, a gravitational background in one more dimension can express the renormalization group (RG) flow of the theory. The field theory beta functions and the gradient flow constraint together reconstruct the second order spacetime equations of motion. The RG equation reduces to the conventional gravitational computation of the spacetime quasilocal stress tensor, and a c-theorem holds true as a consequence of the Raychaudhuri equation. Conversely, under certain conditions, if the RG evolution of a field theory possesses a monotonic c-function, the flow of couplings can be expressed in terms of a higher dimensional gravitational background.
Introduction
The holographic principle states that the degrees of freedom describing quantum gravity in some volume can be encoded at fixed density on a “screen” or surface that bounds that volume . A particularly simple realization of this principle appears to occur for asymptotically anti-de Sitter (AdS) spaces where $`(d+1)`$-dimensional spacetime dynamics is conjectured to be encoded in a local quantum field theory (QFT) on the timelike, $`d`$-dimensional AdS boundary .
In this correspondence, phenomena occuring closer to the AdS boundary are related to local, ultraviolet physics in the field theory, while infrared and non-local data encode the deep interior of the space . This suggests that the semiclassical structure called spacetime can sometimes be generated from a QFT via renormalization group (RG) flow.<sup>1</sup><sup>1</sup>1In fact, it is known that the string field theory equations of motion are closely related to the Wilsonian RG flow of the string sigma model .
Here, we show that if the beta functions of a field theory in $`4`$ dimensions are given by the gradient of a certain potential on the space of couplings, the RG flow of the theory admits a description in terms of $`5`$ dimensional gravity coupled to scalar fields. For such flows, the second order spacetime equations of motion can be reconstructed from the field theory beta functions and the gradient flow constraint. The field theory RG equation is realized as the conventional gravitational computation of the trace of the quasilocal stress tensor. A c-theorem is satisfied by the RG flow as a consequence of the Raychaudhuri equation for the $`5`$ dimensional gravitational background. Conversely, if a QFT has a c-function that evolves monotonically down an RG trajectory, under certain conditions the flow of couplings may be expressed in terms of a $`5`$ dimensional gravitational background.
Gradient beta functions
Consider a $`4`$-dimensional QFT on a Ricci-flat manifold with an interaction Lagrangian:
$$S_{int}=d^4x\varphi ^IO_I.$$
(1)
On a flat manifold there is no conformal anomaly, and the RG equation is simply given by the conformal Ward identity for the trace of the stress tensor:
$$TT_i^i\frac{d\mathrm{\Gamma }}{d\mathrm{log}\lambda }=\beta ^I\frac{\mathrm{\Gamma }}{\varphi ^I}.$$
(2)
$`\mathrm{\Gamma }`$ is quantum effective action of the field theory, the beta functions are the scale derivatives of the couplings $`\beta ^I=\frac{d\varphi ^I}{d\mathrm{log}\lambda }`$, and $`\lambda =\mathrm{\Lambda }/\mathrm{\Lambda }_0`$ in terms of an energy cutoff $`\mathrm{\Lambda }`$ and a reference scale $`\mathrm{\Lambda }_0`$.
Now consider a QFT in which the beta fuctions can be derived as gradients of a particular potential on the couplings:
$$\beta ^I=G^{IJ}(\varphi )\frac{}{\varphi ^J}\mathrm{log}(aT+U(\varphi ))$$
(3)
$`G^{IJ}`$ is a positive, symmetric function of the couplings $`\varphi ^I`$, $`a`$ is a constant of length dimension 4, and $`U(\varphi )`$ is a function of couplings that will be related to the norm of the beta functions. We will show that near a conformal point the inverse metric $`G_{IJ}`$ must be related to the normalization of the 2-point function $`O_IO_J`$. At a conformal point the beta functions vanish. Then (3) implies that $`\frac{(T+U)}{\varphi ^I}`$ also vanishes, so that $`(T+U)`$ is extremized as a function of the couplings.
The QFT RG flow can be related to the equations of motion of 5d gravity if, in addition, the norm of the beta functions is:
$$\frac{1}{4}G_{IJ}\beta ^I\beta ^J=1\frac{V(\varphi )}{(aT+U(\varphi ))^2}0.$$
(4)
$`V(\varphi )`$ will be related to a c-function for the RG flow. The ultraviolet limit of the theory, if it exists, is conformal. In this limit, $`T0`$, and we can choose the normalization $`U(\varphi )1`$. Furthermore, at any conformal point, the vanishing of the beta functions requires that $`V(\varphi )T+U(\varphi )`$, implying that $`V(\varphi )`$ is extremized. We will show that the first order RG flow of the $`d`$-dimensional QFT can be represented as the second order equations of motion of $`5`$-dimensional gravity coupled to scalars $`\varphi ^I`$, with a potential $`V(\varphi )`$ and a “boundary” cosmological constant related to $`U(\varphi )`$.
At first glance, there are three unknowns in (3) and (4): the metric $`G^{IJ}`$ and the potentials $`U(\varphi )`$ and $`V(\varphi )`$. Since there are only two equations, this suggests that any RG flow is expressible in this manner by a suitable choice of metric and potentials. Near a conformal point, we will show that these quantities relate to other dynamical data such as correlation functions. But away from a conformal point, despite the constraints implied by the positivity and symmetry of the metric, (3) and (4) seem to be rather weak restrictions. However, we will argue that any flow that can be written as (3) and (4) possesses a c-function related to the potentials $`U`$ and $`V`$, provided a certain positive energy condition is satisfied. The inequality in (4) will be equivalent to a statement that field theories obeying (3) have a c-function that decreases monotonically during RG flow. So we will be forced to conclude either that a surprisingly general class of field theories possesses a monotonic c-function, or that the conditions (3) and (4) are much more restrictive than they appear.
Gravitational description
Einstein gravity coupled to scalars on a 5-dimensional manifold $``$ has an action
$`S_h`$ $`=`$ $`{\displaystyle \frac{1}{16\pi G_N}}{\displaystyle _{}}d^4x𝑑r\sqrt{h}\left[R_h+{\displaystyle \frac{1}{2}}g_{IJ}h^{\mu \nu }(_\mu \alpha ^I)(_\nu \alpha ^J){\displaystyle \frac{12}{\mathrm{}^2}}v(\alpha )\right]`$ (5)
$``$ $`{\displaystyle \frac{1}{8\pi G_N}}{\displaystyle _{}}d^4x\sqrt{\gamma }\left[\mathrm{\Theta }+L_{c.t.}\right]`$
$`R_h`$ is the Ricci scalar of the spacetime metric $`h_{\mu \nu }`$, $`g_{IJ}`$ is the metric on the space of scalars $`\alpha ^I`$ and $`\mathrm{}`$ is a length scale. By choice, the potential $`v`$ has negative extrema, with at least one extremum at $`v=1`$. Placing the scalars at these points induces a negative cosmological constant in the space. We will be interested in solutions whose potentials approach $`v=1`$ as $`r\mathrm{}`$. The boundary extrinsic curvature $`\theta `$ makes the equations of motion well-defined, and $`L_{\mathrm{c}.\mathrm{t}.}`$ is a counterterm Lagrangian constructed from intrinsic invariants of the induced metric $`\gamma _{ij}`$ on the spacetime boundary. When the scalars are at the $`v(\alpha )=1`$ extremum, the gravitational part of the action can be rendered finite by setting $`L_{\mathrm{c}.\mathrm{t}.}=\frac{3}{\mathrm{}}\left(1\frac{\mathrm{}^2}{12}R_\gamma \right)`$ . We require a counterterm scheme that yields a finite gravitational action for any solution that asymptotically approaches an extremum of the scalar potential. One such scheme is
$$L_{\mathrm{c}.\mathrm{t}.}=\frac{3}{\mathrm{}}\left(u(\alpha )\frac{\mathrm{}^2}{12u(\alpha )}R_\gamma \right),$$
(6)
subject to the requirement that $`u(\alpha )^2v(\alpha )`$ when the $`\alpha ^I`$ approach any extremum of $`v`$.<sup>2</sup><sup>2</sup>2The arguments of $`u(\alpha )`$ are the boundary values of the scalars. Also see for discussions for boundary counterterms in theories with scalars. In effect, $`u(\alpha )`$ serves as a cosmological constant on the boundary of the space. Similarly, counterterms may be added to cancel divergences in the total action for the scalar fields .<sup>3</sup><sup>3</sup>3As we will discuss later, there is a large scheme dependence in this counterterm prescription, in parallel with scheme dependences in field theory RG flows.
The most general $`4`$-dimensional Poincaré invariant solution of this action can be put in the form
$$ds^2=e^{2A(r)}\eta _{ij}dx^idx^j+dr^2,$$
(7)
with $`\eta _{ij}`$ the flat metric in $`4`$ dimensions and the scalars $`\varphi `$ chosen as functions of $`r`$ only. The second order equations of motion that follow from varying (5) with respect to the metric and requiring a solution of the form (7) are :
$`{\displaystyle \frac{d^2A}{dr^2}}`$ $`=`$ $`{\displaystyle \frac{1}{6}}g_{IJ}{\displaystyle \frac{d\alpha ^I}{dr}}{\displaystyle \frac{d\alpha ^J}{dr}},`$ (8)
$`\left({\displaystyle \frac{dA}{dr}}\right)^2`$ $`=`$ $`{\displaystyle \frac{1}{\mathrm{}^2}}v(\alpha )+{\displaystyle \frac{1}{24}}g_{IJ}{\displaystyle \frac{d\alpha ^I}{dr}}{\displaystyle \frac{d\alpha ^J}{dr}}.`$ (9)
The second equation is simply the statement of $`r`$-reparametrization invariance of the action (5) – i.e., it is the Hamiltonian constraint. The scalar equations of motion are
$$g_{KL}\frac{d^2\alpha ^L}{dr^2}+4g_{KL}\frac{d\alpha ^L}{dr}\frac{dA}{dr}=\frac{12}{\mathrm{}^2}\frac{v}{\alpha ^K}+\frac{1}{2}\frac{g_{IJ}}{\alpha ^K}\frac{d\alpha ^I}{dr}\frac{d\alpha ^J}{dr}\frac{g_{KL}}{\alpha ^I}\frac{d\alpha ^I}{dr}\frac{d\alpha ^L}{dr}.$$
(10)
So long as $`d\alpha ^I/dr0`$, we can treat $`A^{}dA/dr`$ and $`\alpha ^I^{}d\alpha ^I/dr`$ as functions of $`\alpha `$, to write (8) as<sup>4</sup><sup>4</sup>4The analysis below will always hold piecewise in domains where $`d\alpha ^I/dr0`$. When the scalars are at an extremum of $`v`$, $`\alpha ^I^{}=0`$ and $`A^{}`$ is a constant. This will correspond to a conformal point in the field theory.
$`{\displaystyle \frac{A^{}}{\alpha ^I}}{\displaystyle \frac{d\alpha ^I}{dr}}`$ $`=`$ $`{\displaystyle \frac{1}{6}}g_{IJ}{\displaystyle \frac{d\alpha ^I}{dr}}{\displaystyle \frac{d\alpha ^J}{dr}}.`$ (11)
Solutions to (11) are obtained by setting
$$\frac{d\alpha ^I}{dr}=6g^{IJ}\frac{A^{}}{\alpha ^J}.$$
(12)
Computing $`A^{}(\alpha )`$ from (12) and (9), it is easy to show that the scalar equations (10) are automatically satisfied.<sup>5</sup><sup>5</sup>5To show this, differentiate (12) with respect to $`r`$ and use the $`\alpha ^I`$ derivative of (9). Thus, starting from (12) and (9) one can derive the complete equations of motions for a five-dimensional lagrangian of the form (5).
Integrating the solutions for $`\varphi ^I^{}`$ and $`A^{}`$ yields a trajectory $`\alpha ^I(r)`$ in the N-dimensional space of scalar fields. The $`N`$ first-order equations (11) and the the N-dimensional first-order equation (9) produce $`2N`$ integration constants . Along with the specification of the integration bound $`r_0`$ and $`A(r_0)`$, these constants reproduce the $`2N+2`$ expected initial conditions of the scalar and gravity equations of motion .
Now consider the total action as a functional of the induced “boundary” metric $`\gamma _{ij}`$ on a surface of fixed $`r`$. Following Brown and York , and including the counterterms (6, the quasilocal stress tensor of the spacetime on a fixed-$`r`$ surface is the response of the action to variations of $`\gamma _{ij}`$:
$`\tau _{ij}`$ $`=`$ $`{\displaystyle \frac{2}{\sqrt{\gamma }}}{\displaystyle \frac{\delta S_h}{\delta \gamma ^{ij}}}`$ (13)
$`=`$ $`{\displaystyle \frac{1}{8\pi G_N}}\left[\theta _{ij}\theta \gamma _{ij}{\displaystyle \frac{3u(\alpha )}{\mathrm{}}}\gamma _{ij}{\displaystyle \frac{\mathrm{}}{2u(\alpha )}}𝒢_{ij}\right].`$ (14)
Here $`\theta _{ij}`$ is the extrinsic curvature of the fixed-$`r`$ surface, $`\theta `$ is the trace of $`\theta _{ij}`$, and $`𝒢_{ij}`$ is the Einstein tensor constructed from the boundary metric. From the Hamilton-Jacobi perspective, $`\tau _{ij}`$ is simply the variable conjugate to the boundary metric $`\gamma _{ij}`$, given the action (5) as a functional of boundary data .
For solutions of the form (7), the trace of $`\tau _{ij}`$ is
$$\tau \gamma ^{ij}\tau _{ij}=\frac{3}{2\pi G_N\mathrm{}}\left[\mathrm{}\frac{dA}{dr}u(\alpha )\right].$$
(15)
We found solutions to the equations of motion by assuming $`d\alpha ^I/dr0`$, so that $`A^{}`$ and $`\alpha ^I^{}`$ could be expressed as functions of $`\alpha ^I`$. Under these conditions, the trace of the Hamilton-Jacobi expression for the Brown-York stress tensor is proportional to
$$\frac{\delta S_h}{\delta A}=\frac{\delta S_h}{\delta \alpha ^I}\frac{\delta \alpha ^I}{\delta A}.$$
(16)
(Recall that $`\tau =(D/\sqrt{\gamma })(\delta S/\delta A)`$).
In the connection between CFTs and gravity on AdS spaces, position in the radial direction is known to play the role of the scale in dual field theory computations. It is difficult to map the radial coordinate directly into field theory scales because of the possibility of radial reparametrization. However, there is an invariant way to study radial positions in classical solutions of the form (7) – we will associate the warp factor in the metric ($`A`$), with the $`\mathrm{log}`$ of $`\lambda `$, the scale in the definition of the beta functions in (2). The map between the field theory RG equations and the gravitational equations of motion is then:
| $`G^{IJ}6g^{IJ}`$ | $`\varphi ^I\alpha ^I`$ | $`a\frac{2\pi G_N\mathrm{}}{3}`$ | $`\mathrm{log}\lambda =A`$ |
| --- | --- | --- | --- |
| $`T\tau `$ | $`V(\varphi )v(\alpha )`$ | $`U(\varphi )u(\alpha )`$ | |
Using these substitutions and the fact that $`A^{}=(a\tau +u(\alpha ))/\mathrm{}`$ (15), it easy to show that the equations of motion (12) and (9) are exactly the gradient beta function equation (3) and the potential equation (4) respectively. As discussed, simultaneous solution of (12) and (9) is sufficient to solve the complete coupled second-order equations for the scalars and the spacetime metric subject to the 4d Poincaré invariant ansatz (7). Identifying variations of $`\mathrm{\Gamma }`$, the QFT effective action as a function of couplings, with variations of $`S`$, the spacetime action as a function of boundary data, the expression for the trace of Brown-York stress tensor (16) becomes equivalent to the field theory RG equation (2).
We have mapped the boundary values of the spacetime scalar fields and the metric to field theory couplings. That accounts for half the integration constants of the equations of motion (8) – (10). The remaining constants are the radial derivatives of bulk fields which, in the AdS/CFT context, were related to expectation values of operators in the CFT . A similar relationship holds here; for example, $`T=\tau dA/dr`$ according to (15). Our analysis treats RG flow in the vacuum state where operator expectation values have been set to zero. From the gravitational perspective, this should correspond to fixing the radial derivatives of bulk fields to yield the lowest energy given the boundary values.
In perturbative string theory, a central consistency condition is the requirement of world-sheet conformal invariance, which implies that strings can consistently propagate only in backgrounds which satisfy the string equation of motion. In other words, the condition that the beta function is zero is equivalent to the target space equation of motion . Equation (3) is an analogue of this statement in the context of holographic duality. Supplemented by the requirement (4) on the norm of the beta functions, (3) yields the full equations of motion of the spacetime fields. As we shall see, the potential in the gradient beta function equation is intimately related to a c-function which decreases monotonically along our RG flows. After the identifications above, (4) became the condition of $`r`$-reparametrization invariance of the 5-dimensional spacetime. This suggests that it should be directly related to invariance of the field theory under redefinitions of the floating scale.
C-functions and an analogue Zamolodchikov metric
Any field theory satisfying the constraint (3) has a c-function that decreases monotonically during RG-flow. The basic reason for this is that $`T`$ is identified with the trace of $`\tau _{ij}`$, the quasilocal stress tensor of spacetime given in (14). Since the intrinsic curvatures of the boundaries at fixed $`r`$ vanish for metrics of the form (7), the trace of the stress tensor depends only on the extrinsic curvature:
$$T\tau =\frac{3}{8\pi G}\left(\theta +\frac{4u(\alpha )}{\mathrm{}}\right).$$
(17)
The Raychaudhuri equation implies a monotonic radial flow of $`\theta `$,
$$\frac{d\theta }{dr}0,$$
(18)
so long as a form of the weak positive energy condition is satisfied by the scalar fields .<sup>6</sup><sup>6</sup>6Monotonicity follows from the Raychaudhuri equation describing evolution of the expansion $`\theta `$ along null curves, applied to static metrics of the form (7). If $`𝒯_{ab}`$ is the matter stress tensor and $`𝒯`$ is its trace, the required energy condition is that $`𝒯_{ab}k^ak^b0`$ for all null $`k^a`$. In , the weak energy condition is defined with respect to timelike $`\xi ^a`$: $`𝒯_{ab}\xi ^a\xi ^b0`$. Both this, and the strong positive energy condition in , $`𝒯_{ab}\xi ^a\xi ^b(1/2)𝒯`$, imply the positivity of $`𝒯_{ab}k^ak^b`$.
This fact, coupled with dimensional analysis and Bousso’s covariant entropy formula , has led Sahakian to propose a candidate gravitational analogue of a c-function:<sup>7</sup><sup>7</sup>7We are specializing the proposal to metrics of the form (7).
$$c\frac{1}{G_N\theta ^3}$$
(19)
(See for interesting examples and related proposals.) Translating into field theory variables, the candidate c-function is
$$c=c_0(aT+U(\varphi ))^3;c_0=\frac{b\mathrm{}^3}{G_N}$$
(20)
with $`b`$ a dimensionless numerical constant. The Raychaudhuri equation automatically implies monotonicity of this c-function. Accepting this proposal, the gradient beta function equation (3) may be rewritten as
$$\beta ^I=\frac{1}{3}G^{IJ}\frac{\mathrm{log}c}{\varphi ^J}\stackrel{~}{G}^{IJ}\frac{c}{\varphi ^J}.$$
(21)
This formula gives a useful consistency check on the identifications we have been performing between field theory and gravitational quantities. We will show that $`\stackrel{~}{G}^{IJ}=G^{IJ}/3c`$ determines the normalization of the field theory 2-point function in the vicinity of a conformal point.
To see this, recall that perturbing around a conformal point by some marginal operators $`O_I`$, an analogue of the c-function in two dimensions can be defined as
$$T(x)\beta ^IO_I;T(k)T(k)=\stackrel{~}{c}k^4$$
(22)
in terms of the trace of the stress tensor $`T=T_i^i`$. At a conformal point, the 2-point function of $`O_I`$ is given by $`O_I(x)O_J(0)=\frac{G_{IJ}^Z}{x^8}`$, so that in the vicinity of this point
$$O_I(k)O_J(k)=k^4(\mathrm{const}.+G_{IJ}^Z\mathrm{log}\frac{k}{\mathrm{\Lambda }_0}+\mathrm{})G_{IJ}^Zf(k)+\mathrm{}$$
(23)
Here $`G_{IJ}^Z`$ is an analogue of the Zamolodchikov metric defined in two dimensions. The last two equations together imply that
$$\stackrel{~}{c}=\mathrm{const}.+\beta ^I\beta ^JG_{IJ}^Z\mathrm{log}\frac{k}{\mathrm{\Lambda }_0}+\mathrm{}$$
(24)
An alternative expression for $`c`$ can be obtained by expanding the couplings around the conformal point as $`\varphi ^I=\varphi _0^I+\beta ^I\mathrm{log}\frac{k}{\mathrm{\Lambda }_0}+\mathrm{}`$ and expanding $`c(\varphi )`$ around $`\varphi _0`$:
$$\stackrel{~}{c}=\mathrm{const}.+\frac{c}{\varphi ^I}\beta ^I\mathrm{log}\frac{k}{\mathrm{\Lambda }_0}+\mathrm{}$$
(25)
The two expressions for $`\stackrel{~}{c}`$ are equal close to a conformal point if
$$\beta ^I=G_Z^{IJ}\frac{\stackrel{~}{c}}{\varphi ^J},$$
(26)
$`G_Z^{IJ}`$ being the inverse of $`G_{IJ}^Z`$. In two dimensions, this formula is exactly true by Zamolodchikov’s theorem , while the above is an approximate derivation in four dimensions, in the vicinity of a conformal point. In fact, various candidates have been proposed for a c-function in 4-dimensional field theory . The coefficient $`\stackrel{~}{c}`$ in (22) is one of these, and is related to the coefficient of the Weyl tensor squared term in the conformal anomaly . Another candidate is the coefficient of the Euler invariant in the anomaly. The material point for us is that equation (26) holds true near a conformal point for either candidate c-function (see and references therein).
Comparing (21) and (26) shows that consistency of the identification of the c-function in the former requires that $`\stackrel{~}{G}_{IJ}`$ determine the 2-point function of $`O_I`$ at the conformal point. At the UV conformal point, the beta functions and $`T`$ are zero, while $`u(\varphi )=1`$; we then expect that $`O_I(k)O_J(k)=f(k)\stackrel{~}{G}_{IJ}=3cf(k)G_{IJ}`$. This serves as a consistency check on the dictionary between field theory and gravity quantities that we are developing. We equated the spacetime action $`S`$, as a functional of boundary data, to the field theory quantum effective action $`\mathrm{\Gamma }`$, as a functional of couplings. So the field theory 2-point function relates directly to the spacetime propagator between boundary points for the scalars $`\alpha _I`$<sup>8</sup><sup>8</sup>8Note the lower index.. We would like to see that this is proportional to $`f(k)G_{IJ}(G_N/\mathrm{}^3)`$.
At the ultraviolet conformal point, the potential $`v(\alpha )`$ in (5) was normalized to $`1`$, and the equations of motion are solved to give pure anti-de Sitter space. So the scalar propagator in spacetime is given by the standard computation in the AdS/CFT correspondence and is proportional in momentum space to $`f(k)=k^4\mathrm{log}k`$. The normalization of the scalar fields in (5) implies that the scalar propagator for $`\alpha _I`$ is proportional to $`G_Ng_{IJ}`$. Since $`G_N`$ has length dimension three and $`\mathrm{}`$ is only remaning length scale, the scalar propagator must yield, via the QFT-gravity dictionary,
$$O_IO_Jf(k)G_{IJ}\frac{G_N}{\mathrm{}^3}+\mathrm{}$$
(27)
in the vicinity of the UV conformal point. In the identification of the RG equations and the spacetime equations of motion, only the combination of parameters $`a=2\pi G_N\mathrm{}/3`$ appeared. Now we see that this data can combine with the normalization of the 2-point function of $`O_I`$ to separately yield the Newton constant ($`G_N`$) and the spacetime curvature scale ($`\mathrm{}`$). Analysis of higher point functions would give further consistency conditions.
The results we have accumulated suffice to show that the RG flow of any theory with a monotonic c-function can be be rewritten in terms of a 5d gravity background under some conditions. Suppose we are given as data the beta functions, stress tensor and monotonic c-function of a 4-dimensional QFT. The relation (21) serves to define a metric $`G^{IJ}`$, which we require to be positive and symmetric. Equation (20) defines the potential $`U(\varphi )`$, and thereby the gradient beta function equation (3). Finally, consider the scale variation of the c-function:
$$\frac{d\mathrm{log}c}{d\mathrm{log}\lambda }=\beta ^I\frac{\mathrm{log}c}{\varphi ^I}$$
(28)
Inverting (21) to get an equation for $`\mathrm{log}c/\varphi ^J`$, we find:
$$\frac{d\mathrm{log}c}{d\mathrm{log}\lambda }G_{IJ}\beta ^I\beta ^J$$
(29)
The right hand side of this equation defines the potential $`V(\varphi )`$ in (4). We now see that $`V(\varphi )`$ is related to the deviation of the c-function away from its value at the conformal point. In summary, given the beta functions, stress tensor and monotonic c-function of a 4d QFT, we have defined a metric and two potentials. As we have shown, these are the elements of a five-dimensional gravitational Lagrangian whose equations of motion reproduce the field theory RG equations.
Discussion
The Hamiltonian constraint (9) expresses radial reparametrization invariance of solutions to the action (5). If radial positions are mapped to field theory scales, we would expect to extract a “holographic” RG equation from this constraint. In de Boer, Verlinde and Verlinde achieve this goal by separating the five-dimensional bulk spacetime action $`S_h`$ bounded at a given radial position into one piece which is local in the boundary data ($`S_\mathrm{l}`$), and another which is non-local ($`S_{\mathrm{nl}}`$). The Hamiltonian constraint, i.e. the Hamilton-Jacobi equation for the total spacetime action $`S_h`$, is rewritten as a first order RG equation of the Callan-Symanzik type for the non-local action $`S_{\mathrm{nl}}`$. The beta-functions are implicitly determined by the form of the local action $`S_\mathrm{l}`$ since they are defined as ratios of $`\frac{\delta S_\mathrm{l}}{\delta A}`$ and $`\frac{\delta S_\mathrm{l}}{\delta \varphi ^I}`$. This definition agrees with the beta function derived from the conformal Ward identity, i.e., the trace of the Brown-York stress tensor. Therefore, knowledge of $`S_\mathrm{l}`$ completely specifies the form of the beta function. Combined with the original Hamiltonian constraint, this determines all the equations of motion, as we have shown above. Furthermore, when the space is $`3+1`$ Poincaré invariant, the local action $`S_\mathrm{l}`$ is determined by the vacuum energy density of the boundary field theory, which is proportional to the trace of the stress-energy tensor. We have derived an RG equation from gravity by directly computing the trace of the quasilocal stress tensor. The resulting beta function agrees exactly with the one discussed in . The Hamiltonian constraint (9) can be rewritten as (4) which, as we have seen (29), simply defines the flow of the $`c`$ function in our formalism.
We have shown that if the beta functions of a 4d field theory are given by gradients of a certain potential, then the RG flow can be expressed as a classical solution of 5 dimensional gravity coupled to scalar fields.<sup>9</sup><sup>9</sup>9By construction, the examples of holographic RG flows developed in all fit our formalism. The Raychaudhuri equation in 5 dimensions automatically guaranteed a c-theorem for such flows. Equivalently, a 4d RG flow relates to 5d gravity when the beta functions are given by the gradient of a monotonic c-function (21), while the c-function is related to the trace of the stress tensor as in (20). Given the beta functions, stress tensor and c-function of a field theory we can always rewrite RG flow in terms of 5d gravity so long as the symmetric, positive metric $`G^{IJ}`$ in (21) can be defined. Similar results may be derived for RG flows of field theories in other dimensions. Famously, beta functions of any renormalizable theory in two dimensions can be expressed as gradients of a c-function with a positive symmetric metric appearing in (21). It has been suggested that a holographic c-function appropriate to 2-dimensional theories will satisfy $`c1/(aT+U(\varphi ))`$ . Using this to define $`U(\varphi )`$, and (4) to define $`V(\varphi )`$, we expect that any such RG flow is expressible in terms of a gravity background.
What characterizes a theory which realizes gradient beta functions of the form discussed in this paper? By construction, the large $`N`$, conformal theories appearing in the AdS/CFT correspondence have the requisite traits. However, the symmetries and properties guaranteeing (3) (or a monotonic c-function) have not been generally understood. Based on the AdS/CFT experience, we expect that a theory might have an RG flow satisfying (3) in some limit of parameters, but that deviations from the limit produce systematic corrections. These should be compared to modified spacetime equations of motion arising from higher derivative terms added to the action (5). In string theory, such terms originate in propagation of the excited states of string and in loop corrections.
Only some field theory RG schemes can be expected to have holographic descriptions in gravity. Within our analysis, there is a scheme dependence in the definition of the stress tensor $`T`$ and the potentials $`U`$ and $`V`$ in field theory. However, these stress tensor ambiguities can be precisely matched by modifications in the counterterm scheme (6) for the gravitational stress tensor.<sup>10</sup><sup>10</sup>10One such ambiguity, affecting the definition of the trace anomaly on a curved manifold is discussed in . More generally, different RG schemes involve different ways of imposing a cutoff and integrating out modes or, alternatively, different choices of counterterms. The AdS/CFT correspondence suggests that an appropriate class of RG schemes coarsens field variables by convolving them against a smearing kernel . Even this cannot be entirely sufficient as the field theory approaches the deep infrared. In this limit, the gravity description involves a large bubble of essentially flat space at the center of an AdS spacetime. Since the field theory is in the deep infrared, it appears that the physics of homogeneous modes describes the flat space region. In the AdS/CFT case, these modes constitute the quantum mechanics of a large matrix, in an echo of the M(atrix) model of M-theory . This suggests that recovery of the equations of motion of a flat space region in AdS requires implementation of a “matrix renormalization group” relating the physics of $`SU(N)`$ to $`SU(Nk)`$.
Holographic realizations of the renormalization group associate field theory scales with radial positions in a higher dimensional spacetime. The full set of 5-dimensional diffeomorphisms can put “bumps” in surfaces at fixed radial positions. Recovering such transformations from 4-dimensional field theory will certainly involve a local notion of the renormalization group where the coarsening scale varies from point to point. This automatically requires consideration of theories with spatially varying couplings. It will be illuminating to elucidate the relation between local RG invariance in field theory and higher dimensional diffeomorphism invariance.
Acknowledgements
We have benefitted from discussions with M. Douglas, P. Hořava, P. Kraus, S. Shenker, E. Verlinde, H. Verlinde, N. Warner, E. Witten, and, particularly, J. de Boer and R. Gopakumar. V.B. was supported by the Harvard Society of Fellows, the Milton Fund of Harvard University, and NSF grant NSF-PHY-9802709. E.G. was supported by DOE grant DE-FG03-92ER40701. D.M. was supported by DOE grant DE-FG03-84ER40168. V.B. is grateful to the University of Chicago and UCLA where part of this work was completed. D.M. enjoyed the hospitality of the University of Illinois at Chicago during this project.
Recent work which overlaps with the content of this paper appears in .
|
warning/0003/gr-qc0003026.html
|
ar5iv
|
text
|
# On product spacetime with 2-sphere of constant curvature
## Abstract
If we consider the spacetime manifold as product of a constant curvature 2-sphere (hypersphere) and a 2-space, then solution of the Einstein equation requires that the latter must also be of constant curvature. There exist only two solutions for classical matter distribution which are given by the Nariai (anti) metric describing an Einstein space and the Bertotti - Robinson (anti) metric describing a uniform electric field. These two solutions are transformable into each other by letting the timelike convergence density change sign. The hyperspherical solution is anti of the spherical one and the vice -versa. For non classical matter, we however find a new solution, which is electrograv dual to the flat space, and describes a cloud of string dust of uniform energy density. We also discuss some interesting features of the particle motion in the Bertotti - Robinson metric.
PACS numbers : 04.20,04.60,98.80Hw
We consider the 4-dimensional spacetime as a product manifold of two spaces $`R^2\times S^2`$ with $`S^2`$ having constant curvature. The latter condition implies that $`R_2^2=R_3^3=const.`$ and $`R_0^0=R_1^1`$ where the coordinates are designated as $`t=0,z=1,\theta =2,\phi =3`$. Each space is a 2-space intersecting the other orthogonally and hence each would be specified by a single curvature, which would be $`R^{23}_{23}`$ for $`S^2`$ and $`R^{01}_{01}`$ for $`R^2`$. The Einstein equation would hence be a relation between them and it turns out that the only classical matter distribution it can sustain is the Einstein space ($`R_{ab}=\mathrm{\Lambda }g_{ab},\rho +p=0`$) or the uniform electric field, with $`R=0`$.
Let us define the the energy density by $`\rho =T_{ab}u^au^b`$, timelike convergence density by $`\rho _t=(T_{ab}1/2Tg_{ab})u^au^b`$ and the null convergence density by $`\rho _n=T_{ab}k^ak^b`$ where $`u^au_a=1,k^ak_a=0`$ . Note that for the product manifold, $`\rho =R_2^2,\rho _n=R_0^0R_1^1=0,\rho _t=R_0^0`$. For the classical matter distribution, there exist the only two possibilities; (i) $`\rho +\rho _t=0`$, the Einstein space given by the Nariai metric and (ii) $`\rho \rho _t=0`$, the uniform electric field given by the Bertotti - Robinson metric , and $`\rho _n=0`$ always. The only other possibility could be of the two curvatures one could vanish; i.e. either $`\rho =0`$ or $`\rho _t=0`$. The former cannot vanish because it defines the constant curvature of the 2-sphere (hypersphere), the construction we began with. Even for the non classical matter, the only possibility is $`\rho _t=0`$, which is the equation of state for the cosmological defects, string dust, global monopole and global texture . The new solution that we shall obtain could describe a string dust of constant energy density. The solution could be shown to be the electrograv dual to the flat space .
For defining the electrogravity duality, we decompose, in analogy with the Maxwell theory, the Riemann curvature into electric and magnetic parts relative to a timelike unit vector. Since the Riemann curvature is a double 2-form, and hence it would always be the double projection. The active electric part is given by $`E_{ab}=R_{acbd}u^cu^d`$, the passive electric part by $`\stackrel{~}{E}_{ab}=Racbdu^cu^d`$ and the magnetic part by $`H_{ab}=R_{acbd}u^cu^d`$ where $`R_{abcd}=1/4\eta _{abmn}\eta _{cdpq}R^{mnpq}`$ and $`\eta _{abcd}`$ is the 4-volume element. In terms of the electromagnetic parts, the Ricci curvature is given by
$$R_{ab}=E_{ab}+\stackrel{~}{E}_{ab}+(E+\stackrel{~}{E})u_au_b\stackrel{~}{E}g_{ab}+u^cH^{mn}(\eta _{acmn}u_b+\eta _{bcmn}u_a)$$
(1)
where $`E=E_a^a,\stackrel{~}{E}_a^a=\stackrel{~}{E}`$, and $`\rho =\stackrel{~}{E},\rho _t=E`$.
By the electrogravity duality we mean ,
$$E_{ab}\stackrel{~}{E}_{ab},H_{ab}H_{ab}$$
(2)
That is the interchange of active and passive electric parts. Note that under duality, the Ricci and the Einstein tensors interchange , which is because contraction of the Riemann is Ricci while its double dual is Einstein.
Clearly under the duality transformation, $`\rho \rho _t,\rho _n\rho _n`$ which means the Einstein space would be anti dual while the uniform electric field solution would be self dual. We would further show that the new solution describing a string dust would be dual to the flat space.
We write the metric for the product spacetime as
$$ds^2=c^2dt^2a^2dz^2\frac{1}{\lambda ^2}(d\theta ^2+sin^2\theta d\phi ^2)$$
(3)
where $`\lambda `$ is a constant with dimension of inverse length, and $`c,a`$ are in general functions of $`z`$ and $`t`$ which run from $`\mathrm{}`$ to $`\mathrm{}`$. There are only two independent components of the Ricci tensor which read as
$$\rho _t=R_0^0=R_1^1=\frac{1}{c^2}(\frac{\ddot{a}}{a}\frac{\dot{a}\dot{c}}{ac})\frac{1}{a^2}(\frac{c^{\prime \prime }}{c}\frac{a^{}c^{}}{ac}),\rho =R_2^2=R_3^3=\lambda ^2$$
(4)
The two classical matter solutions would be obtained by solving the single equation $`\rho =\lambda ^2=\pm \rho _t`$. For its solution we will have to choose the one of $`a`$ and $`c`$ or a relation between them. There could exist both static and non static solutions but one could always be transformed into the other. We shall give the both in the gauge $`ac=1`$. The solutions are given as follows.
I. $`\rho +\rho _t=0`$: The Nariai metric for the Einstein space,
$$c^2=a^2=(1+\lambda ^2t^2)^1,1\lambda ^2z^2$$
(5)
which gives $`\rho =\lambda ^2=p`$.
II. $`\rho =\rho _t=\lambda ^2`$: The Bertotti - Robinson metric for the uniform electric field,
$$c^2=a^2=(1\lambda ^2t^2)^1,1+\lambda ^2z^2$$
(6)
The electric field would be along the $`z`$-axis, $`F_{01}=\pm \sqrt{2}ac\lambda `$, which would imply
$$(F^{ij}\sqrt{g})_{,j}=0$$
(7)
where a coma denotes ordinary derivative. Thus the electric field is uniform and is given by $`\pm \sqrt{2}\lambda `$. This equation indicates absence of charges, which would be lying at $`z=\pm \mathrm{}`$ to produce a uniform field.
Let us note some of the interesting properties of the two solutions:
(a) $`NBR`$ when $`\rho _t\rho _t`$, where N stands for the Nariai metric and BR for the Bertotti - Robinson metric. That means changing the sign of $`\lambda ^2`$ only in $`c`$.
(b) The anti metrics of N and BR would be given by letting sphere ($`sin\theta `$) to go to hypersphere ($`sinh\theta `$) in the metric. That is, in N and BR let both $`\rho \rho ,\rho _t\rho _t`$ to get to their anti counterparts.
(c) The anti metrics are transformable into each-other as in (a) above by letting $`\rho _t\rho _t`$.
(d) Under the duality transformation, $`\rho \rho _t`$, the Einstein space is anti dual, and the Nariai and anti Nariai are dual of each - other. On the other hand, the trace free matter field is self dual, and the BR and anti BR are dual of each other with both $`\rho `$ and $`\rho _t`$ changing sign. This is because under the duality gravitational constant changes sign .
Now let us turn to the only remaining possibility of a curved spacetime for the metric (3),e.g. $`\rho _t=R_0^0=0`$. That is $`R^2`$ is flat which would mean $`c=a=1`$. The metric (3) would still be curved giving rise to the stresses $`\rho =T_0^0=T_1^1=\lambda ^2`$ and the rest being zero. The equation of state, $`\rho _t=0`$ is characteristic of topological defects; string dust, global monopole and global texture . In the present case it is the string dust that has constant energy density. The anti string dust metric would result if we let $`sin\theta sinh\theta ,spherehypersphere`$ in the metric (3) with $`a=c=1`$. This will have $`\rho =\lambda ^2`$.
The string dust solution can be obtained from the equation,
$$E_{ab}=0,\stackrel{~}{E}_b^a=\lambda ^2g_1^ag_b^1$$
(8)
which is dual to the effective flat space equation,
$$\stackrel{~}{E}_{ab}=0,E_b^a=\lambda ^2g_1^ag_b^1$$
(9)
for the metric (3). For a general spherically symmetric metric (replace $`z`$ by $`r`$, $`1/\lambda `$ by $`b`$ and take $`a,b,c`$ as functions of $`r`$ and $`t`$) it can be shown after considerable manipulation , that the above equation characterizes flat space. This shows that the string dust which is the solution of the dual set (8) is dual to flat space which is the solution of the set (9).
The stress tensor for string dust is given by $`T^{ab}=ϵ\sigma ^{ac}\sigma _c^b/\gamma ^{1/2}`$ where $`ϵ`$ is the proper energy density of the string cloud, $`\gamma _{ab}`$ is the 2-dimensional metric on the string world sheet, $`\sigma ^{ab}`$ is the bivector associated with the world sheet: $`\sigma ^{ab}=ϵ^{AB}x^a/\lambda ^Ax^b/\lambda ^B`$. Here $`ϵ^{AB}`$ is the 2-D Levi-Civita tensor ($`ϵ^{01}=ϵ^{10}=1`$) and $`\lambda ^A=(\lambda ^0,\lambda ^1)`$, where $`\lambda ^0`$ and $`\lambda ^1`$ are timelike and spacelike parameters on the string world sheet. In Ref. the stress tensor for a spherically symmetric metric has been computed and shown that it has $`T_0^0=T_1^11/r^2`$ which in our case would be proportional to a constant because the 2-sphere (hypersphere) has constant curvature. The matter distribution represented by the solution is indeed a cloud of string dust. As in other cases its anti version would have $`sin\theta `$ being replaced by $`sinh\theta `$, and $`\rho \rho `$.
We shall next consider some interesting features of particle motion in the Bertotti - Robinson metric as given by the following form,
$$ds^2=(1+\lambda ^2z^2)dt^2(1+\lambda ^2z^2)^1dz^2\frac{1}{\lambda ^2}(d\theta ^2+sin^2\theta d\phi ^2)$$
(10)
Here the red-shifted proper acceleration is given by $`\lambda z`$ which would be attractive/repulsive for $`z>0(z<0)`$. The most pertinent motion in this spacetime is the $`z`$-motion and it would clearly be simple harmonic about the stable state of rest $`z=0`$. A particle sitting at $`z=0`$, which could be chosen anywhere on the axis freely, would remain stay put there for ever. That would mean that a particle at rest would always remain at rest at any $`z`$ while one in motion would execute simple harmonic oscillation about some appropriate $`z=0`$ location. This happens because as it moves on the positive $`z`$, electrostatic energy lying behind it keeps on building up to pull it back, then it turns back and the same happens on the other side. This is an interesting simple harmonic oscillator which can be set up anywhere and is entirely maintained for ever by gravity (Fig. 1). Of course the catch is that the total energy of the system is infinite. It is though not very realistic, yet it is an interesting and novel example of a relativistic analogue of a simple classical situation.
What would happen if we consider motion of a charged particle in this spacetime? The effective potential for the pertinent motion would then be given by
$$V=q\lambda z+(1+\lambda ^2z^2)^{1/2}$$
(11)
where $`q`$ is the charge per unit mass. Without loss of generality, we can take $`q0`$, because negative $`q`$ would only imply reflection in $`z`$. Here there would be three cases corresponding to $`q^2<1`$, $`q^2>1`$ and $`q^2=1`$.
Case (i): For $`q^2<1`$, the potential would have the minimum at $`\lambda z=q/\sqrt{1q^2}`$ and $`V_{min}=\sqrt{1q^2}`$. Particle would have oscillatory motion like the neutral particle with $`V_{min}`$ being lowered and its location shifted on the negative $`z`$-axis. However the potential would not be symmetric but would be wider on the left of the minimum (Fig. 1). A particle sitting at the minimum of the potential would remain so for ever. That is, a charged particle in a uniform electric field is being kept at stable state of rest by gravity.
Case (ii): For $`q^2>1`$, there will be no minimum and $`V=0`$ at $`\lambda z=1/\sqrt{q^21}`$. It increases monotonically from negative large values to positive large values (Fig. 1). Negative energy orbits were first encountered in the Kerr spacetime and they were confined to ergosphere with its outer boundary at $`r=2M`$. By bringing in electromagnetic interaction, the effective ergosphere could be extended but its extent would always remain bounded . In here we have a situation where occurrence of negative energy extends upto infinity in the negative $`z`$ direction. Particles with both positive and negative energy can have a bounce and go back to infinity. This unphysical feature is due to infinite energy of the spacetime.
Case (iii): For $`q^2=1`$, there again occurs no minimum and it asymptotically tends to zero for negative $`z`$ (Fig. 1).
With the constant curvature sphere (hypersphere) as one part of the product space, there are only three possible solutions. The two are the known solutions; Nariai cosmological metric for the Einstein space representing a fluid with $`\rho +p=0`$ and the Bertotti - Robinson metric for uniform electric field. The third one is new and represents non classical matter of string dust of constant energy density. The Nariai metric has six parameter symmetry group and is not conformally flat. It is shown to be created by quantum polarization of vacuum and asymptotically it decays into the de Sitter and the Kasner spacetimes . The Bertotti - Robinson metric is conformally flat and has also been used to calculate one-loop quantum gravitational corrections induced by fields of large mass . It would be worthwhile to carry out such calculations for the new string dust solution, which could in a way be viewed as ”minimally” curved because it is completely free of the Newtonian gravity .
Apart from the new solution, the novel feature of the paper is to expose the inter - transformability of the Nariai and the Bertotti - Robinson metrics, the role of electrogravity duality in finding the new solution and some interesting features of particle motion in the spacetime of uniform electric field. It is interesting that a particle at rest in this spacetime would remain at rest for ever while the one in motion along the axis of the field would execute simple harmonic oscillation for ever.
Acknowledgement: I wish to thank Varun Sahni and Ramseh Tikekar for enlightening and useful discussions, and Ivor Robinson for reading the manuscript.
|
warning/0003/hep-th0003080.html
|
ar5iv
|
text
|
# Twist decomposition of nonlocal light-cone operators II: General tensors of 2nd rank II. General tensors of 2nd rank
## 1 Introduction
Any careful quantum field theoretical analysis of the experimental data for the different light–cone dominated QCD–processes (for recent experiments and analyses, see ) necessarily requires a rigorous twist decomposition of the various nonlocal light–ray operators appearing in the theoretical set up of these processes. The matrix elements of these special operators are related to the distribution amplitudes (or the hadronic wave functions) being necessary for the phenomenological description of the above mentioned processes. Moreover, their anomalous dimensions determine the $`Q^2`$–evolution of the distribution amplitudes.
The notion of twist has been originally introduced for local operators by Gross and Treiman as a geometric quantity, twist $`\tau `$ = (canonical) dimension $`d`$ – spin $`j`$, which is directly related to the irreducible representations of the Lorentz group.<sup>1</sup><sup>1</sup>1 Obviously, if $`d`$ is taken as scale dimension of the operators it is not related to the generators of the Lorentz group but to those of the subgroup $`SO(2h)_+`$ of the conformal group in $`2h`$-dimensions. Thus a local operator with definite twist is a irreducible finite dimensional representation of the group $`SO(2h)_+`$. The twist decomposition of the local operators of some well-defined tensor structure corresponds directly to their decomposition into irreducible tensor representations of the Lorentz group. These representations are uniquely (up to equivalence) determined by their symmetry classes, i.e., Young frames, and by the (anti)symmetrization defined by corresponding Young operators, i.e., specific Young tableaux.
In the case of nonlocal tensor operators, being given as (infinite) towers of local tensor operators of growing rank, the definition of (geometric) twist is more subtle. For that reason another notion of (dynamical) twist, being related to the LC–quantization in the infinite momentum frame , has been introduced by counting powers of $`1/Q`$ which, from a phenomenological point of view, is quite advantageous. However, this different notion of twist has serious theoretical disadvantages: It is not Lorentz invariant and, therefore, not immediately related to geometric twist; furthermore, it is process–dependent because it is only applicable to matrix elements of the corresponding LC–operators and not to the operators itself.
In order to be able to disentangle the twist content of nonlocal LC–operators which are relevant for various hard QCD processes a systematic group theoretical study has been started in an earlier paper , thereafter cited as I. There, bilocal quark operators up to (antisymmetric) second rank have been considered. Here, we continue this study for general tensor operators of 2nd rank. We carry out the complete twist decomposition of all relevant LC–operators, e.g., bilocal gluon operators, trilocal quark–gluon correlation operators (related to so-called Shuryak-Vainshtein operators) and four–fermion operator, as well as multilocal quark and gluon operators. Thereby, generalized harmonic polynomials being tensors up to rank 2 are introduced and the interior derivative on the light–cone is used.
The paper is organized as follows. In Chapt. 2 we shortly review the general method introduced in Part I. In Chapt. 3 we give the complete twist decomposition of the various gluonic bilocal light–ray tensor operators of second rank being specified by different symmetry classes; their twist content ranges from 2 up to 6. In Chapt. 4 we extend these results with minor modifications to the trilocal light–ray operators mentioned above. Any of these nonlocal operators consists of an infinite tower of local operators which are characterized by equal twist. Appendix A contains some material about the tensorial harmonic functions.
## 2 General procedure
Let us now shortly review the general procedure, introduced in Part I, of how to decompose arbitrary bilocal LC–operators into harmonic operators of definite twist. Those nonlocal operators which are relevant for the virtual Compton scattering are obtained by the nonlocal LC expansion of the (renormalized) time-ordered product of two electromagnetic hadronic currents:
$$T(J^\mu (y+\frac{x}{2})J^\nu (y\frac{x}{2})S)\underset{x^20}{}\mathrm{d}^2\underset{¯}{\kappa }C_\mathrm{\Gamma }^{\mu \nu }(x^2,\underset{¯}{\kappa })T\left(𝒪^\mathrm{\Gamma }(y+\kappa _1\stackrel{~}{x},y+\kappa _2\stackrel{~}{x})S\right)+\mathrm{},$$
where $`\mathrm{\Gamma }=\{1,\gamma _\mu ,\sigma _{\mu \nu };\gamma _5,\gamma _\mu \gamma _5,\sigma _{\mu \nu }\gamma _5\}`$ indicates the tensor structure of the nonlocal quark operators and $`\stackrel{~}{x}`$ is a light-like vector related to $`x`$, cf. . The unrenormalized nonlocal (flavour singlet) operators at $`y=0`$ are given by
$$𝒪^\mathrm{\Gamma }(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=\overline{\psi }(\kappa _1\stackrel{~}{x})\mathrm{\Gamma }U(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})\psi (\kappa _2\stackrel{~}{x})$$
(2.1)
with the path ordered phase factor
$$U(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=𝒫\mathrm{exp}\left\{\mathrm{i}g_{\kappa _1}^{\kappa _2}d\tau \stackrel{~}{x}^\mu A_\mu (\tau \stackrel{~}{x})\right\},$$
(2.2)
ensuring gauge invariance. As is well-known, under renormalization these operators mix with appropriate chiral-even (or chiral-odd) bilocal gluonic tensor operators, eventually multiplied by appropriate tensors built up from $`\stackrel{~}{x}`$:
$`G_{\alpha \beta }(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`=F_\alpha ^\rho (\kappa _1\stackrel{~}{x})U(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})F_{\beta \rho }(\kappa _2\stackrel{~}{x}),`$ (2.3)
$`\stackrel{~}{G}_{\alpha \beta }(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`=F_\alpha ^\rho (\kappa _1\stackrel{~}{x})U(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})\stackrel{~}{F}_{\beta \rho }(\kappa _2\stackrel{~}{x}),`$ (2.4)
where $`F_{\alpha \beta }`$ and $`\stackrel{~}{F}_{\alpha \beta }=\frac{1}{2}ϵ_{\alpha \beta \mu \nu }F^{\mu \nu }`$ are the gluon field strength and its dual, respectively. Here, the phase factors are to be taken in the adjoint representation.
In general, the twist decomposition of an arbitrary bilocal light–ray operator may be formulated for any space-time dimension $`D=2h`$. Let us denote such operators for arbitrary values of $`x`$ as follows:
$$𝒪^\mathrm{\Gamma }(\kappa _1x,\kappa _2x)=\mathrm{\Phi }^{}(\kappa _1x)\mathrm{\Gamma }U(\kappa _1x,\kappa _2x)\mathrm{\Phi }(\kappa _2x)$$
(2.5)
where, suppressing any indices indicating the group representations, $`\mathrm{\Phi }`$ generically denotes the various local fields, e.g., scalars ($`d=h1`$), Dirac spinors ($`d=h\frac{1}{2}`$) as well as gauge field strength ($`d=h`$). $`A_\mu `$ is the gauge potential having dimension $`d=h1`$. Furthermore, $`\mathrm{\Gamma }`$ labels the tensor structure as well as additional quantum numbers, if necessary.
Now, the twist decomposition of operators (2.5) consists of the following steps:
(1) Taylor expansion of the nonlocal operators for arbitrary values of $`x`$ at the point $`y=0`$ into an infinite series of local tensor operators having definite rank $`n`$ and canonical dimension $`d`$:
$`𝒪^\mathrm{\Gamma }(\kappa _1x,\kappa _2x)={\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{1}{n!}}x^{\mu _1}\mathrm{}x^{\mu _n}\left[\mathrm{\Phi }^{}(y)\mathrm{\Gamma }\stackrel{}{D}_{\mu _1}(\kappa _1,\kappa _2)\mathrm{}\stackrel{}{D}_{\mu _n}(\kappa _1,\kappa _2)\mathrm{\Phi }(y)\right]_{y=0}`$ (2.6)
with the generalized covariant derivatives
$`\stackrel{}{D}_\mu (\kappa _1,\kappa _2)`$ $``$ $`\kappa _1\stackrel{}{D}_\mu +\kappa _2\stackrel{}{D}_\mu ,`$ (2.7)
$`\stackrel{}{D}_\mu =\stackrel{}{_\mu ^y}+\mathrm{i}gA_\mu (y),`$ $`\stackrel{}{D}_\mu =\stackrel{}{_\mu ^y}\mathrm{i}gA_\mu (y).`$
(2) Decomposition of local operators with respect to irreducible tensor representations of the Lorentz group $`SO(2h1,1)`$ or, equivalently, the orthogonal group $`SO(2h)`$.<sup>2</sup><sup>2</sup>2 The corresponding representations are related through analytic continuation, cf. These representations are built up by traceless tensors of rank $`m`$ whose symmetry class is determined by some (normalized) Young operators $`𝒴_{[m]}=(f_{[m]}/m!)𝒬𝒫`$, where $`[m]=(m_1,m_2,\mathrm{}m_r)`$ with $`m_1m_2\mathrm{}m_r`$ and $`_{i=1}^rm_i=m`$ denotes the corresponding Young pattern. $`𝒫`$ and $`𝒬`$, as usual, denote symmetrization and antisymmetrization with respect to that pattern. The allowed Young patterns for $`SO(2h)`$, which because of the tracelessness are restricted by $`\mathrm{}_1+\mathrm{}_22h`$ ($`\mathrm{}`$: length of columns of $`[m]`$), are for a fixed value of $`m`$:
1.
2.
3.
4.
$`\mathrm{}`$
(Here, we depicted only those patterns which appear for $`h=2`$ where $`\mathrm{}_1+\mathrm{}_24`$.) Loosely speaking, the lower spins in the above series correspond to the various trace terms which define, in general, reducible higher twist operators. These higher twist operators again have to be decomposed according to the considered symmetry classes.
Now, two remarks are in order. At first, this decomposition of the local operators appearing in (2.6) into irreducible components is independent of the special values of $`\kappa _1,\kappa _2`$. Therefore, we may choose $`\kappa _1=0,\kappa _2=\kappa `$ and factor out $`\kappa ^n`$ for every term in eq. (2.6), thus simplifying the explicit computations below. At the end of the computations the general values of $`\kappa _i`$ may be re-installed. Furthermore, if these irreducible local tensor operators are multiplied by $`x^{\mu _1}\mathrm{}x^{\mu _n}`$, as it is necessary according to eq. (2.6), then harmonic (tensorial) polynomials of order $`n`$ appear whose remaining tensor structure results from $`\mathrm{\Gamma }`$. For a detailed exposition of these objects, see Appendix A.
(3) Resummation of the local operators $`𝒪_{\mathrm{\Gamma }n}^\tau (x)`$ belonging to the same symmetry class (for any $`n`$) and having equal twist $`\tau `$. That infinite tower of local operators creates a nonlocal operator of definite twist,
$$𝒪_\mathrm{\Gamma }^\tau (0,\kappa x)=\underset{n=0}{\overset{\mathrm{}}{}}\frac{\kappa ^n}{n!}𝒪_{\mathrm{\Gamma }n}^\tau (x).$$
Obviously, from a group theoretical point of view, this nonlocal operator is built up as direct sum of irreducible local tensor operators. The nonlocal operators $`𝒪_\mathrm{\Gamma }^\tau (0,\kappa x)`$ are harmonic tensor functions. Now, putting together all contributions, i.e., including the higher twist contributions resulting from the trace terms, we get the (infinite) twist decomposition of the nonlocal operator we started with:
$$𝒪_\mathrm{\Gamma }(0,x)=\underset{\tau =\tau _{\mathrm{min}}}{\overset{\mathrm{}}{}}𝒪_\mathrm{\Gamma }^\tau (0,x).$$
(4) Finally, the projection onto the light–cone, $`x\stackrel{~}{x}`$, leads to the required light–cone operator with well defined geometric twist. However, since the harmonic tensor polynomials essentially depend on (infinite) sums of powers of $`x^2`$ and $`\mathrm{}`$ as well as some specific differential operators in front of it, in that limit only a finite number of terms survive:
$$𝒪_\mathrm{\Gamma }(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=\underset{\tau =\tau _{\mathrm{min}}}{\overset{\tau _{\mathrm{max}}}{}}𝒪_\mathrm{\Gamma }^\tau (\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}).$$
The resulting light–ray operators of definite twist, which have been rewritten for general arguments, are tensor functions on the light–cone which fulfil another kind of tracelessness conditions to be formulated with the help of the interior (on the cone) derivative. — Let us remark that step (3) and (4) can be interchanged without changing the result.
The advantages of our method are: First, these nonlocal operators of definite twist are Lorentz covariant tensors (contrary to the phenomenological concept of twist). Second, this twist decomposition is unique, process– and model–independent. Furthermore, the twist decomposition is independent from the dimension $`2h`$ of space-time.
## 3 Twist decomposition of a general 2nd rank <br>tensor operator
This Chapter is devoted to the twist decomposition of a general 2nd rank tensor operator. For simplicity, we restrict ourselves to four dimensional space-time ($`h=2`$). Furthermore, since the twist decomposition is independent from the chirality we demonstrate it only for the chiral-even gluon operator (2.3). In addition, we make the special choice $`\kappa _1=0,\kappa _2=\kappa `$, thereby also simplifying the generalized covariant derivatives, eq. (2.7), to the usual ones:
$`G_{\alpha \beta }(0,\kappa x)=`$ $`F_\alpha ^\rho (0)U(0,\kappa x)F_{\beta \rho }(\kappa x)={\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{\kappa ^n}{n!}}x^{\mu _1}\mathrm{}x^{\mu _n}G_{\alpha \beta (\mu _1\mathrm{}\mu _n)},`$ (3.1)
$`\mathrm{with}`$ $`G_{\alpha \beta (\mu _1\mathrm{}\mu _n)}=F_\alpha ^\rho (y)D_{(\mu _1}^y\mathrm{}D_{\mu _n)}^yF_{\beta \rho }(y)|_{y=0};`$
here the symbol $`(\mathrm{})`$ denotes symmetrization of the enclosed indices.
The local tensor operators $`G_{\alpha \beta (\mu _1\mathrm{}\mu _n)}`$ decompose according to the Young patterns (i) to (iv) and possible higher ones (i.e., for $`m_1+m_2=n+2,m_23`$):
$`G_{\alpha \beta (\mu _1\mathrm{}\mu _n)}`$ $`=`$ $`G_{\alpha \beta (\mu _1\mathrm{}\mu _n)}^{(\mathrm{i})}+\alpha _{n+1}G_{\alpha \beta (\mu _1\mathrm{}\mu _n)}^{(\mathrm{ii})}`$ (3.2)
$`+\beta _nG_{\alpha \beta (\mu _1\mathrm{}\mu _n)}^{(\mathrm{iii})}+\gamma _nG_{\alpha \beta (\mu _1\mathrm{}\mu _n)}^{(\mathrm{iv})}+\mathrm{},`$
with the (nontrivial) normalization coefficients $`f_{[m]}/m!`$ of the Young operators given by $`\alpha _{n+1}=2(n+1)/(n+2)`$ for $`[m]=(n+1,1)`$, $`\beta _n=3n/(n+2)`$ for $`[m]=(n,1,1)`$ and $`\gamma _n=4(n1)/(n+1)`$ for $`[m]=(n,2)`$. The corresponding Clebsch-Gordan series in terms of representations $`(j_1,j_2)`$ of the Lorentz group is
$`(\frac{1}{2},\frac{1}{2})(\frac{1}{2},\frac{1}{2})\left((\frac{n}{2},\frac{n}{2})(\frac{n2}{2},\frac{n2}{2})\mathrm{}\right)`$
$`=`$ $`(\frac{n+2}{2},\frac{n+2}{2})\left((\frac{n+2}{2},\frac{n}{2})(\frac{n}{2},\frac{n+2}{2})\right)(\frac{n}{2},\frac{n}{2})`$
$`\left((\frac{n+2}{2},\frac{n2}{2})(\frac{n2}{2},\frac{n+2}{2})\right)`$
$`\left((\frac{n}{2},\frac{n2}{2})(\frac{n2}{2},\frac{n}{2})\right)(\frac{n2}{2},\frac{n2}{2})\mathrm{};`$
the corresponding tensor spaces will be denoted by $`𝐓(j_1,j_2)`$. In the last line of eq. (3) such representations are listed down which correspond to higher twist contributions contained in the trace terms of symmetry classes (i) – (iv). The canonical (or scale) dimension $`d`$ of the local operator (3.2) is $`n+4`$, and the spin of the various contributions in (3) ranges from $`n+2`$ up to $`1`$ or $`0`$ if $`n`$ is even or odd, respectively; therefore, the local operators with well-defined twist are irreducible tensors of the Lorentz group or, equivalently, of the group $`SL(2,)_+`$, having scale dimension $`d`$.
In general, according to the spin content and the rank of the corresponding local tensor operators, the nonlocal operator (3.1) for arbitrary $`x`$ contains contributions of twist ranging from $`\tau =2`$ until $`\tau =\mathrm{}`$. However, after projection onto the light–cone, $`x\stackrel{~}{x}`$, this infinite series terminates at least at $`\tau =6`$ since higher order terms are proportional to $`x^2`$.
In order to be more explicit, as well as for later use, we introduce the (anti)symmetrization with respect to $`\alpha `$ and $`\beta `$ and we define the following nonlocal operators:
$`G_{\alpha \beta }^\pm (0,\kappa x)`$ $`:=`$ $`\frac{1}{2}\left(G_{\alpha \beta }(0,\kappa x)\pm G_{\beta \alpha }(0,\kappa x)\right),`$ (3.4)
$`G_\alpha ^\pm (0,\kappa x)`$ $`:=`$ $`x^\beta G_{\alpha \beta }^\pm (0,\kappa x)G_\alpha ^\pm (0,\kappa x),`$ (3.5)
$`G(0,\kappa x)`$ $`:=`$ $`x^\alpha x^\beta G_{\alpha \beta }^+(0,\kappa x).`$ (3.6)
In fact, the twist decomposition of these bilocal light–ray operators reads:<sup>3</sup><sup>3</sup>3 We use the notation $`A_{(\alpha \beta )}\frac{1}{2}(A_{\alpha \beta }+A_{\beta \alpha })`$ and $`A_{[\alpha \beta ]}\frac{1}{2}(A_{\alpha \beta }A_{\beta \alpha })`$.
$`G_{\alpha \beta }^+(0,\kappa \stackrel{~}{x})=`$ $`G_{(\alpha \beta )}^{\mathrm{tw2}}(0,\kappa \stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw3}}(0,\kappa \stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw4}}(0,\kappa \stackrel{~}{x})`$
$`+G_{(\alpha \beta )}^{\mathrm{tw5}}(0,\kappa \stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw6}}(0,\kappa \stackrel{~}{x}),`$ (3.7)
$`G_{\alpha \beta }^{}(0,\kappa \stackrel{~}{x})=`$ $`G_{[\alpha \beta ]}^{\mathrm{tw3}}(0,\kappa \stackrel{~}{x})+G_{[\alpha \beta ]}^{\mathrm{tw4}}(0,\kappa \stackrel{~}{x})+G_{[\alpha \beta ]}^{\mathrm{tw5}}(0,\kappa \stackrel{~}{x}),`$ (3.8)
$`G_\alpha ^+(0,\kappa \stackrel{~}{x})=`$ $`G_{(\alpha )}^{\mathrm{tw2}}(0,\kappa \stackrel{~}{x})+G_{(\alpha )}^{\mathrm{tw3}}(0,\kappa \stackrel{~}{x})+G_{(\alpha )}^{\mathrm{tw4}}(0,\kappa \stackrel{~}{x}),`$ (3.9)
$`G_\alpha ^{}(0,\kappa \stackrel{~}{x})=`$ $`G_{[\alpha ]}^{\mathrm{tw3}}(0,\kappa \stackrel{~}{x})+G_{[\alpha ]}^{\mathrm{tw4}}(0,\kappa \stackrel{~}{x}),`$ (3.10)
$`G(0,\kappa \stackrel{~}{x})=`$ $`G^{\mathrm{tw2}}(0,\kappa \stackrel{~}{x}).`$ (3.11)
The various twist contributions individually decompose further according to the symmetry classes which may contribute. The explicit expressions for generic tensor operators are given in the following subsections, where it will be shown that Young patterns (i) and (ii) as well as (iv) contribute to the symmetric tensor operators and related vector and scalar operators, and Young patterns (ii) and (iii) contribute to the antisymmetric tensor operators and related vector operators. Special tensor operators are considered in Chapt. 4.
### 3.1 Tensor operators of symmetry class (i)
#### 3.1.1 Construction of nonlocal symmetric class-(i) operators of definite twist
Let us start with the simplest case of the totally symmetric traceless tensors, and their contractions with $`x`$, which have twist $`\tau =2`$ and are contained in the tensor space $`𝐓(\frac{n+2}{2},\frac{n+2}{2})`$. They have symmetry class (i) and are uniquely characterized by the following standard tableau (with normalizing factor 1):
The irreducible local twist–2 operator reads
$$G_{(\alpha \beta \mu _1\mathrm{}\mu _n)}^{\mathrm{tw2}(\mathrm{i})}=F_{(\alpha }^\rho (0)D_{\mu _1}\mathrm{}D_{\mu _n}F_{\beta )\rho }(0)\mathrm{trace}\mathrm{terms}.$$
Let us postpone the determination of the trace terms and make the resummation to the corresponding nonlocal operator in advance. This is obtained, at first, by contracting with $`x^{\mu _1}\mathrm{}x^{\mu _n}`$ and rewriting in the following form:
$`G_{(\alpha \beta )n}^{\mathrm{tw2}(\mathrm{i})}(x)=x^{\mu _1}\mathrm{}x^{\mu _n}G_{(\alpha \beta \mu _1\mathrm{}\mu _n)}^{\mathrm{tw2}(\mathrm{i})}=\frac{1}{\left(n+2\right)\left(n+1\right)}_\alpha _\beta \stackrel{}{G}_{n+2}(x),`$
with the denotation
$`\stackrel{}{G}_{n+2}(x)`$ $`=`$ $`G_{n+2}(x)\mathrm{trace}\mathrm{terms},`$
$`G_{n+2}(x)`$ $``$ $`x^\mu x^\nu F_\mu ^\rho (0)(xD)^nF_{\nu \rho }(0).`$
Here, $`\stackrel{}{G}_{n+2}(x)`$ is a harmonic polynomial of order $`n+2`$, cf. eq. (A.5), Appendix A:
$$\stackrel{}{G}_{n+2}(x)=H_{n+2}^{(4)}\left(x^2|\mathrm{}\right)G_{n+2}(x)\underset{k=0}{\overset{[\frac{n+2}{2}]}{}}\frac{(n+2k)!}{(n+2)!k!}\left(\frac{x^2}{4}\right)^k\mathrm{}^kG_{n+2}(x).$$
Then, using $`((n+2)(n+1))^1=_0^1d\lambda (1\lambda )\lambda ^n`$ and the integral representation of Euler’s beta function we obtain the nonlocal twist–2 tensor operator:<sup>4</sup><sup>4</sup>4 Here, and in the following, we omit the indication of the symmetry class of the nonlocal twist–2 operators because only totally symmetric tensors contribute. However, the trace terms being of higher twist must be classified according to their symmetry type.
$$G_{(\alpha \beta )}^{\mathrm{tw2}}(0,\kappa x)=\underset{n=0}{\overset{\mathrm{}}{}}\frac{\kappa ^n}{n!}G_{(\alpha \beta )n}^{\mathrm{tw2}(\mathrm{i})}(x)=_\alpha _\beta _0^1d\lambda (1\lambda )\stackrel{}{G}(0,\kappa \lambda x),$$
(3.12)
with
$$\stackrel{}{G}(0,\kappa x)=G(0,\kappa x)+\underset{k=1}{\overset{\mathrm{}}{}}_0^1dtt\left(\frac{x^2}{4}\right)^k\frac{\mathrm{}^k}{k!(k1)!}\left(\frac{1t}{t}\right)^{k1}G(0,\kappa tx).$$
(3.13)
The operator $`G_{(\alpha \beta )}^{\mathrm{tw2}}(0,\kappa x)`$ satisfies the conditions of a harmonic tensor function:
$$g^{\alpha \beta }G_{(\alpha \beta )}^{\mathrm{tw2}}(0,\kappa x)=0,\mathrm{}G_{(\alpha \beta )}^{\mathrm{tw2}}(0,\kappa x)=0,^\alpha G_{(\alpha \beta )}^{\mathrm{tw2}}(0,\kappa x)=0.$$
(3.14)
#### 3.1.2 Reduction to vector and scalar operators
Now, we specify to the nonlocal vector and scalar operators. Multiplying eq. (3.12) by $`x^\beta `$ and observing the equality $`(x_x)\stackrel{}{G}(0,\kappa \lambda x)=(\lambda _\lambda +2)\stackrel{}{G}(0,\kappa \lambda x)`$ gives the twist–2 vector operator,
$$G_{(\alpha )}^{\mathrm{tw2}}(0,\kappa x)=_\alpha _0^1d\lambda \lambda \stackrel{}{G}(0,\kappa \lambda x),$$
(3.15)
which satisfies the conditions
$$\mathrm{}G_{(\alpha )}^{\mathrm{tw2}}(0,\kappa x)=0,^\alpha G_{(\alpha )}^{\mathrm{tw2}}(0,\kappa x)=0.$$
(3.16)
Multiplying with $`x^\alpha x^\beta `$ gives the twist–2 scalar operator,
$$G^{\mathrm{tw2}}(0,\kappa x)=\stackrel{}{G}(0,\kappa x),$$
(3.17)
which, by definition, satisfies the condition
$$\mathrm{}G^{\mathrm{tw2}}(0,\kappa x)=0.$$
(3.18)
The expression (3.13) for the scalar operator already has been given by Balitsky and Braun .
Comparing eqs. (3.12), (3.15) and (3.17) we may recognize that in the case of symmetry class (i) the tensor and vector operators are obtained from the scalar operator by very simple operations. Furthermore, we observe how in the case of the scalar operator the trace terms – being proportional to $`x^2`$ – are to be subtracted from $`G(0,\kappa x)`$ in order to make that operator traceless. In the case of vector and tensor operators such subtraction, because of the appearance of the derivatives, is more complicated.
#### 3.1.3 Projection onto the light–cone
Let us now project onto the light–cone and, at the same time, also extend to the case of general values $`(\kappa _1,\kappa _2)`$. Because of the derivatives $`_\alpha `$ and $`_\beta `$, appearing in eq. (3.12), only the terms with $`k=1,2`$ in eq. (3.13) contribute. The final expression for the symmetric twist–2 light–cone tensor operator is given by
$`G_{(\alpha \beta )}^{\mathrm{tw2}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`_\alpha _\beta {\displaystyle _0^1}d\lambda (1\lambda )G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}G_{(\alpha \beta )}^{>(\mathrm{i})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$ (3.19)
$`G_{(\alpha \beta )}^{>(\mathrm{i})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`{\displaystyle _0^1}\mathrm{d}\lambda \{(1\lambda +\lambda \mathrm{ln}\lambda )(\frac{1}{2}g_{\alpha \beta }+x_{(\alpha }_{\beta )})\mathrm{}`$ (3.20)
$`+\frac{1}{4}(2(1\lambda )+(1+\lambda )\mathrm{ln}\lambda )x_\alpha x_\beta \mathrm{}^2\}G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}.`$
The operator $`G_{(\alpha \beta )}^{>(\mathrm{i})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ contains the higher twist contributions which have to be subtracted from the first term, eq. (3.19), in order to make the whole expression traceless. In order to disentangle the different terms of well-defined twist we observe that the tensor operator (3.20) consists of scalar and vector parts. Here, an operator being contained in the trace terms is called a scalar and vector part, respectively, if the expression multiplied by $`g_{\alpha \beta }`$, $`x_\alpha `$ or $`x_\beta `$ is a scalar, like $`\mathrm{}G(\kappa _1x,\kappa _2x)`$ or $`^\mu G_\mu (\kappa _1x,\kappa _2x)`$, and a vector, like $`_\beta G(\kappa _1x,\kappa _2x)`$ or $`_\beta ^\mu G_\mu (\kappa _1x,\kappa _2x)`$, respectively.<sup>5</sup><sup>5</sup>5 Strictly speaking, $`x_\alpha `$ and $`x_\beta `$ as well as $`g_{\alpha \beta }`$ which are necessary for the dimension and tensor structure of the whole expression do not belong to the higher twist operator itself. The scalar operators, $`\mathrm{}G(\kappa _1x,\kappa _2x)|_{x=\stackrel{~}{x}}`$ and $`\mathrm{}^2G(\kappa _1x,\kappa _2x)|_{x=\stackrel{~}{x}}`$, occurring in eq. (3.20) already have well-defined twist $`\tau =4`$ and $`\tau =6`$, respectively. To ensure that the vector operator $`_\beta \mathrm{}G(\kappa _1x,\kappa _2x)|_{x=\stackrel{~}{x}}`$ also obtains well-defined twist, it is necessary to subtract its own trace terms which are of twist $`\tau =6`$.
The higher twist operators contained in the trace terms of the twist–2 tensor operator are the following:
$`G_{(\alpha \beta )}^{>(\mathrm{i})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{i})\mathrm{a}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{i})\mathrm{b}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$
$`+G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{i})\mathrm{a}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{i})\mathrm{b}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$ (3.21)
with
$`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{i})\mathrm{a}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{2}g_{\alpha \beta }\mathrm{}{\displaystyle _0^1}d\lambda (1\lambda +\lambda \mathrm{ln}\lambda )G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{i})\mathrm{b}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`x_{(\alpha }_{\beta )}\mathrm{}{\displaystyle _0^1}d\lambda (1\lambda +\lambda \mathrm{ln}\lambda )G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{i})\mathrm{b}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$
and
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{i})\mathrm{a}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{4}x_\alpha x_\beta \mathrm{}^2{\displaystyle _0^1}d\lambda \left(2(1\lambda )+(1+\lambda )\mathrm{ln}\lambda \right)G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{i})\mathrm{b}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{4}x_\alpha x_\beta \mathrm{}^2{\displaystyle _0^1}d\lambda (\frac{\left(1\lambda \right)^2}{\lambda }\frac{1\lambda ^2}{2\lambda }\lambda \mathrm{ln}\lambda )G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}.`$
Obviously, by inspection of the spin content of the symmetry type (i), the local twist–4 and twist–6 operators carry the representation $`𝐓(\frac{n}{2},\frac{n}{2})`$ and $`𝐓(\frac{n2}{2},\frac{n2}{2})`$, respectively.
Finally, completing the twist decomposition of the symmetric nonlocal tensor operator, we write down the symmetric twist–4 light–cone tensor operator contained in the expression (3.20)
$`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{i})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`=G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{i})\mathrm{a}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{i})\mathrm{b}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$
$`={\displaystyle _0^1}\mathrm{d}\lambda \{(1\lambda +\lambda \mathrm{ln}\lambda )(\frac{1}{2}g_{\alpha \beta }+x_{(\alpha }_{\beta )})\mathrm{}`$ (3.22)
$`\frac{1}{4}(\frac{\left(1\lambda \right)^2}{\lambda }\frac{1\lambda ^2}{2\lambda }\lambda \mathrm{ln}\lambda )x_\alpha x_\beta \mathrm{}^2\}G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
whereas the symmetric twist–6 light–cone operator contained in (3.20) reads
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{i})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`=G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{i})\mathrm{a}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{i})\mathrm{b}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ (3.23)
$`=\frac{1}{4}x_\alpha x_\beta \mathrm{}^2{\displaystyle _0^1}d\lambda \left(\frac{1\lambda ^2}{2\lambda }+\mathrm{ln}\lambda \right)G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}.`$
#### 3.1.4 Vector and scalar light–ray operators
Contracting eq. (3.19) with $`\stackrel{~}{x}^\beta `$, making use of formula $`(x_x)f(\lambda x)=\lambda _\lambda f(\lambda x)`$ and performing the partial integrations we obtain the final version of the twist–2 light–cone vector operator,
$$G_{(\alpha )}^{\mathrm{tw2}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=_0^1d\lambda \lambda \left[_\alpha +\frac{1}{2}(\mathrm{ln}\lambda )x_\alpha \mathrm{}\right]G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},$$
(3.24)
which already has been used in , and the twist–4 light–cone vector operator,
$$G_{(\alpha )}^{\mathrm{tw4}(\mathrm{i})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=\frac{1}{2}x_\alpha \mathrm{}_0^1d\lambda \lambda (\mathrm{ln}\lambda )G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}.$$
(3.25)
In order to obtain the scalar twist–2 light–ray operator we multiply (3.24) by $`\stackrel{~}{x}^\alpha `$. Then, the twist–4 part vanishes and the remaining twist–2 operator restores the scalar operator (compare eq. (3.11)),
$$G^{\mathrm{tw2}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=G(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}).$$
(3.26)
Let us point to the fact that the trace of the original gluon tensor, $`g^{\alpha \beta }G_{\alpha \beta }(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$, is a twist–4 scalar operator. It is contained in $`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{i})\mathrm{a}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ as well as similar expressions occurring below, cf. also eq. (3.83).
#### 3.1.5 Condition of tracelessness on the cone
Finally, it should be remarked, that the conditions (3.14), (3.16) and (3.18), if translated into the corresponding ones containing derivatives with respect to $`\stackrel{~}{x}`$, no longer hold for the light–cone operators. This is clear because, by projecting onto the light–cone, part of the original structure of the operators has been lost. Nevertheless, the conditions of tracelessness of the light–cone operators may be formulated by using the interior derivative on the light–cone which has been used extensively for the construction of local conformal operators . In four dimensions it is given by
$$\mathrm{d}_\alpha \left(1+\stackrel{~}{x}\stackrel{~}{}\right)\stackrel{~}{}_\alpha \frac{1}{2}\stackrel{~}{x}_\alpha \stackrel{~}{}^2\text{with}\stackrel{~}{}_\alpha \frac{}{\stackrel{~}{x}^\alpha },$$
and has the following properties
$$\mathrm{d}^2=0,[\mathrm{d}_\alpha ,\mathrm{d}_\beta ]=0\text{and}\mathrm{d}_\alpha \stackrel{~}{x}^2=\stackrel{~}{x}^2\left(\mathrm{d}_\alpha +2\stackrel{~}{}_\alpha \right).$$
Then, the conditions of tracelessness simplify, namely, they read:
$`g^{\alpha \beta }G_{(\alpha \beta )}^{\mathrm{tw2}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=0,\mathrm{d}^\alpha G_{(\alpha \beta )}^{\mathrm{tw2}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=0,`$
as well as
$`\mathrm{d}^\alpha G_{(\alpha )}^{\mathrm{tw2}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=0.`$
Analogous conditions hold for the light–cone operators of definite twist in the case of symmetry classes (ii) – (iv) obtained below.
In passing we remark that the interior derivative may be used in defining more directly totally symmetric local light–cone operators . For example it holds
$`G_{(\alpha \beta )n}^{\mathrm{tw2}(\mathrm{i})}(\stackrel{~}{x})`$ $`=\frac{1}{\left(n+2\right)^2\left(n+1\right)^2}\mathrm{d}_\alpha \mathrm{d}_\beta G_{n+2}(\stackrel{~}{x}),`$
$`G_{(\alpha )n}^{\mathrm{tw2}(\mathrm{i})}(\stackrel{~}{x})`$ $`=\frac{1}{\left(n+2\right)^2}\mathrm{d}_\alpha G_{n+2}(\stackrel{~}{x}).`$
However, in case of symmetry types (ii) – (iv) the corresponding expressions should be more involved.
### 3.2 Tensor operators of symmetry class (ii)
#### 3.2.1 Construction of nonlocal (anti)symmetric class-(ii) operators of definite twist: Young tableau A
Now we consider tensor operators, and their contractions with $`x`$, having symmetry class (ii) and whose local twist–3 parts are contained in $`𝐓(\frac{n+2}{2},\frac{n}{2})𝐓(\frac{n}{2},\frac{n+2}{2})`$. Contrary to the totally symmetric case we have different possibilities to put the tensor indices into the corresponding Young pattern. Without presupposing any symmetry of indices $`\alpha `$ and $`\beta `$ we should start with the following Young tableau:
with normalizing factor $`\alpha _{n+1}`$. (The tableau with $`\alpha \beta `$ will be considered thereafter.) Denoting this symmetry behaviour by (iiA) we may write the local twist-3 tensor operator as follows:
$`G_{\alpha \beta \mu _1\mathrm{}\mu _n}^{\mathrm{tw3}(\mathrm{iiA})}`$ $`=\frac{1}{n+2}\{F_\alpha ^\rho (0)D_{(\mu _1}\mathrm{}D_{\mu _n)}F_{\beta \rho }(0)+F_\alpha ^\rho (0)D_\beta D_{(\mu _2}\mathrm{}D_{\mu _n}F_{\mu _1)\rho }(0)`$
$`+{\displaystyle \underset{l=2}{\overset{n}{}}}F_\alpha ^\rho (0)D_{(\mu _1}\mathrm{}D_{\mu _{l1}}D_{|\beta |}D_{\mu _{l+1}}\mathrm{}D_{\mu _n}F_{\mu _l)\rho }(0)`$
$`(\alpha \mu _1)\}\mathrm{trace}\mathrm{terms}.`$
Proceeding in the same manner as in the last Subsection we multiply by $`x^{\mu _1}\mathrm{}x^{\mu _n}`$ and obtain:
$`G_{\alpha \beta n}^{\mathrm{tw3}(\mathrm{iiA})}(x)`$ $`=\frac{1}{n\left(n+2\right)}\left(\delta _\alpha ^\mu (x)x^\mu _\alpha \right)_\beta \stackrel{}{G}_{\mu |n+1}(x)`$
with
$`\stackrel{}{G}_{\mu |n+1}(x)`$ $`x^\nu \stackrel{}{G}_{\mu \nu |n}(x)`$
$`=\left\{\delta _\mu ^\alpha \frac{1}{\left(n+2\right)^2}\left[x_\mu ^\alpha (x)\frac{1}{2}x^2_\mu ^\alpha \right]\right\}H_{n+1}^{(4)}(x^2|\mathrm{})G_{\alpha |n+1}(x).`$
Here, $`\stackrel{}{G}_{\mu |n+1}(x)`$ is the harmonic vector polynomial of order $`n+1`$ (see, Eq. (A), Appendix A) and the symmetry behaviour of class (ii) is obtained through the differential operator in front of it. Now, using $`((n+2)n)^1=_0^1d\lambda (1\lambda ^2)\lambda ^n/(2\lambda )`$ and Euler’s beta function we sum up to obtain the following nonlocal twist–3 tensor operator:
$`G_{\alpha \beta }^{\mathrm{tw3}(\mathrm{iiA})}(0,\kappa x)`$ $`={\displaystyle _0^1}d\lambda \frac{1\lambda ^2}{2\lambda }\left(\delta _\alpha ^\mu (x)x^\mu _\alpha \right)_\beta \stackrel{}{G}_\mu (0,\kappa \lambda x),`$ (3.27)
with the nonlocal traceless (vector) operator
$`\stackrel{}{G}_\alpha (0,\kappa x)=`$ $`G_\alpha (0,\kappa x)+{\displaystyle \underset{k=1}{\overset{\mathrm{}}{}}}{\displaystyle _0^1}dt\left({\displaystyle \frac{x^2}{4}}\right)^k{\displaystyle \frac{\mathrm{}^k}{k!(k1)!}}\left({\displaystyle \frac{1t}{t}}\right)^{k1}G_\alpha (0,\kappa tx)`$
$`\left[x_\alpha ^\mu (x)\frac{1}{2}x^2_\alpha ^\mu \right]`$ (3.28)
$`\times {\displaystyle \underset{k=0}{\overset{\mathrm{}}{}}}{\displaystyle _0^1}\mathrm{d}\tau \tau {\displaystyle _0^1}\mathrm{d}tt\left({\displaystyle \frac{x^2}{4}}\right)^k{\displaystyle \frac{\mathrm{}^k}{k!k!}}\left({\displaystyle \frac{1t}{t}}\right)^kG_\mu (0,\kappa \tau tx).`$
The following decomposition into symmetric and antisymmetric part is useful for the further calculations
$$G_{\alpha \beta }^{\mathrm{tw3}(\mathrm{iiA})}(0,\kappa x)=G_{[\alpha \beta ]}^{\mathrm{tw3}(\mathrm{iiA})}(0,\kappa x)+G_{(\alpha \beta )}^{\mathrm{tw3}(\mathrm{iiA})}(0,\kappa x),$$
(3.29)
with
$`G_{[\alpha \beta ]}^{\mathrm{tw3}(\mathrm{iiA})}(0,\kappa x)`$ $`={\displaystyle _0^1}d\lambda \lambda \delta _{[\alpha }^\mu _{\beta ]}\stackrel{}{G}_\mu (0,\kappa \lambda x),`$ (3.30)
$`G_{(\alpha \beta )}^{\mathrm{tw3}(\mathrm{iiA})}(0,\kappa x)`$ $`={\displaystyle _0^1}d\lambda \frac{1}{\lambda }\left(\delta _{(\alpha }^\mu _{\beta )}\frac{1\lambda ^2}{2}_\alpha _\beta x^\mu \right)\stackrel{}{G}_\mu (0,\kappa \lambda x)`$ (3.31)
$`={\displaystyle _0^1}d\lambda \frac{1\lambda ^2}{2\lambda }\left(\delta _{(\alpha }^\mu (x)x^\mu _{(\alpha }\right)_{\beta )}\stackrel{}{G}_\mu (0,\kappa \lambda x).`$ (3.32)
Again, these operators are harmonic tensor functions. The conditions of tracelessness for the tensor operator are
$`g^{\alpha \beta }G_{\alpha \beta }^{\mathrm{tw3}(\mathrm{iiA})}(0,\kappa x)=0,`$ $`\mathrm{}G_{\alpha \beta }^{\mathrm{tw3}(\mathrm{iiA})}(0,\kappa x)=0,`$
$`^\alpha G_{\alpha \beta }^{\mathrm{tw3}(\mathrm{iiA})}(0,\kappa x)=0,`$ $`^\beta G_{\alpha \beta }^{\mathrm{tw3}(\mathrm{iiA})}(0,\kappa x)=0.`$ (3.33)
#### 3.2.2 Reduction to vector operators
The corresponding twist–3 vector operator is obtained from eq. (3.27) by multiplication with $`x^\beta `$:
$$G_\alpha ^{\mathrm{tw3}(\mathrm{iiA})}(0,\kappa x)=_0^1d\lambda \lambda \left(\delta _\alpha ^\mu (x)x^\mu _\alpha \right)\stackrel{}{G}_\mu (0,\kappa \lambda x).$$
(3.34)
Obviously, a corresponding scalar operator does not exist.
This vector operator fulfils the following conditions of tracelessness
$`\mathrm{}G_\alpha ^{\mathrm{tw3}(\mathrm{iiA})}(0,\kappa x)=0,^\alpha G_\alpha ^{\mathrm{tw3}(\mathrm{iiA})}(0,\kappa x)=0.`$ (3.35)
#### 3.2.3 Projection onto the light–cone
(a) The calculation of the antisymmetric tensor operator $`G_{[\alpha \beta ]}^{\mathrm{tw3}(\mathrm{iiA})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ on the light–cone is similar to that of $`M_{[\alpha \beta ]}^{\mathrm{tw2}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ in Part I; for the details we refer to it. The resulting expression is:
$`G_{[\alpha \beta ]}^{\mathrm{tw3}(\mathrm{iiA})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`=`$ $`{\displaystyle _0^1}d\lambda \lambda \delta _{[\alpha }^\mu _{\beta ]}G_\mu (\kappa _1\lambda \stackrel{~}{x},\kappa _2\lambda \stackrel{~}{x})|_{x=\stackrel{~}{x}}G_{[\alpha \beta ]}^{>(\mathrm{iiA})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$ (3.36)
$`G_{[\alpha \beta ]}^{>(\mathrm{iiA})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`=`$ $`\frac{1}{2}{\displaystyle _0^1}\mathrm{d}\lambda (1\lambda )\{(2x_{[\alpha }_{\beta ]}^\mu x_{[\alpha }\delta _{\beta ]}^\mu \mathrm{})G_\mu (\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}`$ (3.37)
$`(1\lambda +\lambda \mathrm{ln}\lambda )x_{[\alpha }_{\beta ]}\mathrm{}G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}\}.`$
$`G_{[\alpha \beta ]}^{>(\mathrm{iiA})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ contains twist–4 and twist–5 contributions, but twist–6 contributions do not appear due to $`x_{[\alpha }x_{\beta ]}=0`$. The higher twist operator (3.37) contains two vector operators multiplied by $`x_\alpha `$ and $`x_\beta `$, respectively. For their twist decomposition one has to take into account Young pattern (i) as well as (ii). The procedure is analogous to the decomposition of the vector operator $`O_\alpha (\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ made in Part I. After a straightforward calculation we obtain
$`G_{[\alpha \beta ]}^{>(\mathrm{iiA})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`G_{[\alpha \beta ]}^{\mathrm{tw4}(\mathrm{iiA})\mathrm{a}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{[\alpha \beta ]}^{\mathrm{tw4}(\mathrm{iiA})\mathrm{b}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$
$`+G_{[\alpha \beta ]}^{\mathrm{tw5}(\mathrm{iiA})\mathrm{a}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ (3.38)
with
$`G_{[\alpha \beta ]}^{\mathrm{tw4}(\mathrm{iiA})\mathrm{a}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`=`$ $`\frac{1}{2}x_{[\alpha }_{\beta ]}^\mu {\displaystyle _0^1}d\lambda \frac{1\lambda ^2}{\lambda }G_\mu (\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{[\alpha \beta ]}^{\mathrm{tw4}(\mathrm{iiA})\mathrm{b}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`=`$ $`\frac{1}{2}x_{[\alpha }_{\beta ]}\mathrm{}{\displaystyle _0^1}d\lambda \left(\frac{1\lambda ^2}{2\lambda }+\lambda \mathrm{ln}\lambda \right)G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{[\alpha \beta ]}^{\mathrm{tw5}(\mathrm{iiA})\mathrm{a}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`=`$ $`x_{[\alpha }\left(\delta _{\beta ]}^\mu (x)x^\mu _{\beta ]}\right)\mathrm{}{\displaystyle _0^1}d\lambda \frac{\left(1\lambda \right)^2}{4\lambda }G_\mu (\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}.`$
(b) Now, we determine the symmetric tensor operator $`G_{(\alpha \beta )}^{\mathrm{tw3}(\mathrm{iiA})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ on the light–cone. Putting eq. (3.28) into (3.31), after some lengthy but straightforward calculation (taking into account only the relevant terms of the $`k`$-summation and performing some partial integrations), we get the following result:
$`G_{(\alpha \beta )}^{\mathrm{tw3}(\mathrm{iiA})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`{\displaystyle _0^1}d\lambda \frac{1\lambda ^2}{2\lambda }\left(\delta _{(\alpha }^\mu (x)x^\mu _{(\alpha }\right)_{\beta )}G_\mu (\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}`$
$`G_{(\alpha \beta )}^{>(\mathrm{iiA})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$ (3.39)
with the higher twist contributions of the trace terms
$`G_{(\alpha \beta )}^{>(\mathrm{iiA})}(\kappa _1\stackrel{~}{x},`$ $`\kappa _2\stackrel{~}{x})={\displaystyle }_0^1\mathrm{d}\lambda \{[\frac{1\lambda ^2}{2\lambda }g_{\alpha \beta }^\mu +\frac{1\lambda }{2\lambda }\delta _{(\alpha }^\mu x_{\beta )}\mathrm{}+(1\lambda )x_{(\alpha }_{\beta )}^\mu `$
$`\frac{\left(1\lambda \right)^2}{4\lambda }x_\alpha x_\beta ^\mu \mathrm{}]G_\mu (\kappa _1\lambda \stackrel{~}{x},\kappa _2\lambda \stackrel{~}{x})[\frac{1}{2}(\frac{1\lambda ^2}{2\lambda }+\lambda \mathrm{ln}\lambda )g_{\alpha \beta }\mathrm{}`$
$`+\left(\frac{1}{2}(1\lambda )+\frac{1\lambda ^2}{4\lambda }+\lambda \mathrm{ln}\lambda \right)x_{(\alpha }_{\beta )}\mathrm{}`$
$`+\frac{1}{4}(\frac{1\lambda ^2}{2\lambda }\frac{\left(1\lambda \right)^2}{\lambda }+\lambda \mathrm{ln}\lambda )x_\alpha x_\beta \mathrm{}^2]G(\kappa _1\lambda x,\kappa _2\lambda x)\left\}\right|_{x=\stackrel{~}{x}}.`$ (3.40)
It is obvious that eq. (3.2.3) contains scalar and vector operators. Again, using Young patterns (i) and (ii) and subtracting the trace terms, we can decompose the vector part appearing in eq. (3.2.3) into twist–4, twist–5 and twist–6 operators. This procedure is analogous to the twist decomposition of the (vector) quark operator $`O_\alpha (\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ in Part I. From the twist–4 and twist–5 vector part we have to subtract their trace terms being of twist–6 and add it to the other twist–6 scalar operator. Let us recall that the scalar twist–4 and twist–6 operators are already traceless on the cone. In that manner we get the following decomposition:
$`G_{(\alpha \beta )}^{>(\mathrm{iiA})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`=G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iiA})\mathrm{a}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iiA})\mathrm{b}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iiA})\mathrm{c}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$
$`+G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iiA})\mathrm{d1}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iiA})\mathrm{d2}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iiA})\mathrm{e}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$
$`+G_{(\alpha \beta )}^{\mathrm{tw5}(\mathrm{iiA})\mathrm{d}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iiA})\mathrm{a}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iiA})\mathrm{b}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$
$`+G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iiA})\mathrm{c}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iiA})\mathrm{d1}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iiA})\mathrm{d2}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$
$`+G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iiA})\mathrm{d3}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iiA})\mathrm{d4}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iiA})\mathrm{e}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$ (3.41)
with
$`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iiA})\mathrm{a}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{2}g_{\alpha \beta }\mathrm{}{\displaystyle _0^1}d\lambda \left(\frac{1\lambda ^2}{2\lambda }+\lambda \mathrm{ln}\lambda \right)G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iiA})\mathrm{b}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{2}x_{(\alpha }_{\beta )}\mathrm{}{\displaystyle _0^1}\mathrm{d}\lambda ((1\lambda )+\frac{1\lambda ^2}{2\lambda }`$
$`+2\lambda \mathrm{ln}\lambda )G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iiA})\mathrm{b}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$
$`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iiA})\mathrm{d1}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{2}x_{(\alpha }_{\beta )}\mathrm{}{\displaystyle _0^1}d\lambda \left(\frac{1\lambda }{\lambda }+\frac{\mathrm{ln}\lambda }{\lambda }\right)G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iiA})\mathrm{d1}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$
and
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iiA})\mathrm{a}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{4}x_\alpha x_\beta \mathrm{}^2{\displaystyle _0^1}d\lambda (\frac{\left(1\lambda \right)^2}{\lambda }\frac{1\lambda ^2}{2\lambda }\lambda \mathrm{ln}\lambda )G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iiA})\mathrm{b}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{4}x_\alpha x_\beta \mathrm{}^2{\displaystyle _0^1}\mathrm{d}\lambda (\frac{\left(1\lambda \right)^2}{2\lambda }\frac{3}{2}\frac{1\lambda ^2}{2\lambda }`$
$`\lambda \mathrm{ln}\lambda \frac{\mathrm{ln}\lambda }{2\lambda })G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iiA})\mathrm{d1}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{4}x_\alpha x_\beta \mathrm{}^2{\displaystyle _0^1}d\lambda \left(\frac{1\lambda }{\lambda }+\frac{\mathrm{ln}\lambda }{\lambda }+\frac{\mathrm{ln}^2\lambda }{2\lambda }\right)G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iiA})\mathrm{d3}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{2}x_\alpha x_\beta \mathrm{}^2{\displaystyle _0^1}d\lambda \left(\frac{1\lambda }{\lambda }+\frac{\mathrm{ln}\lambda }{\lambda }+\frac{\mathrm{ln}^2\lambda }{2\lambda }\right)G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
being related to the scalar operator $`G(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$, as well as
$`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iiA})\mathrm{c}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{2}g_{\alpha \beta }^\mu {\displaystyle _0^1}d\lambda \frac{1\lambda ^2}{\lambda }G_\mu (\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iiA})\mathrm{d2}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`x_{(\alpha }_{\beta )}^\mu {\displaystyle _0^1}d\lambda \left(\frac{1\lambda }{\lambda }+\frac{\mathrm{ln}\lambda }{\lambda }\right)G_\mu (\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iiA})\mathrm{d2}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$
$`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iiA})\mathrm{e}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`x_{(\alpha }_{\beta )}^\mu {\displaystyle _0^1}d\lambda (1\lambda )G_\mu (\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iiA})\mathrm{e}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$
and
$`G_{(\alpha \beta )}^{\mathrm{tw5}(\mathrm{iiA})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{2}x_{(\alpha }\left(\delta _{\beta )}^\mu (x)x^\mu _{\beta )}\right)\mathrm{}`$
$`\times {\displaystyle _0^1}\mathrm{d}\lambda (\frac{1\lambda }{\lambda }+\frac{\mathrm{ln}\lambda }{\lambda })G_\mu (\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iiA})\mathrm{d3}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iiA})\mathrm{d4}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$
and
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iiA})\mathrm{c}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{4}x_\alpha x_\beta \mathrm{}^\mu {\displaystyle _0^1}d\lambda \frac{\left(1\lambda \right)^2}{\lambda }G_\mu (\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iiA})\mathrm{d2}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{2}x_\alpha x_\beta \mathrm{}^\mu {\displaystyle _0^1}d\lambda (\frac{1\lambda }{\lambda }+\frac{\mathrm{ln}\lambda }{\lambda }+\frac{\mathrm{ln}^2\lambda }{2\lambda })G_\mu (\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iiA})\mathrm{d4}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{2}x_\alpha x_\beta \mathrm{}^\mu {\displaystyle _0^1}d\lambda (\frac{1\lambda }{\lambda }+\frac{\mathrm{ln}\lambda }{\lambda })G_\mu (\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iiA})\mathrm{e}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{4}x_\alpha x_\beta \mathrm{}^\mu {\displaystyle _0^1}d\lambda \frac{\left(1\lambda \right)^2}{\lambda }G_\mu (\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
being related to the vector operator $`G_\mu (\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$.
#### 3.2.4 Contributions of Young tableau B
Let us now consider the other Young tableau which is obtained from the former one by exchanging $`\alpha `$ and $`\beta `$:
This symmetry behaviour will be denoted by (iiB). The corresponding nonlocal twist–3 operator which we obtain after analogous calculations is (with self-explaining denotations):
$$G_{\alpha \beta }^{\mathrm{tw3}(\mathrm{iiB})}(0,\kappa x)=_0^1d\lambda \frac{1\lambda ^2}{2\lambda }\left(\delta _\beta ^\mu (x)x^\mu _\beta \right)_\alpha \stackrel{}{G}_\mu (0,\kappa \lambda x).$$
(3.42)
Its decomposition into symmetric and antisymmetric part yields
$$G_{\alpha \beta }^{\mathrm{tw3}(\mathrm{iiB})}(0,\kappa x)=G_{[\alpha \beta ]}^{\mathrm{tw3}(\mathrm{iiB})}(0,\kappa x)+G_{(\alpha \beta )}^{\mathrm{tw3}(\mathrm{iiB})}(0,\kappa x),$$
(3.43)
with
$`G_{[\alpha \beta ]}^{\mathrm{tw3}(\mathrm{iiB})}(0,\kappa x)`$ $`={\displaystyle _0^1}d\lambda \lambda \delta _{[\beta }^\mu _{\alpha ]}\stackrel{}{G}_\mu (0,\kappa \lambda x),`$ (3.44)
$`G_{(\alpha \beta )}^{\mathrm{tw3}(\mathrm{iiB})}(0,\kappa x)`$ $`={\displaystyle _0^1}d\lambda \frac{1\lambda ^2}{2\lambda }\left(\delta _{(\alpha }^\mu (x)x^\mu _{(\alpha }\right)_{\beta )}\stackrel{}{G}_\mu (0,\kappa \lambda x).`$ (3.45)
The projection onto the light–cone and the calculation of the higher twist operators contained in the trace terms is completely analogous to case (A) and should be omitted here.
#### 3.2.5 Construction of the complete twist–3 tensor operators on the light–cone
In order to obtain the complete twist–3 operator it is necessary to add both twist–3 operators (3.29) and (3.43) resulting from the Young patterns (iiA) and (iiB). Since no further Young pattern contributes to twist–3 operators we omit the index (ii). After projection onto the light–cone the final result is
$$G_{\alpha \beta }^{\mathrm{tw3}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=G_{[\alpha \beta ]}^{\mathrm{tw3}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw3}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),$$
(3.46)
with the antisymmetric twist–3 light–cone tensor operator<sup>6</sup><sup>6</sup>6 Here, we used the abbreviation $`G_\mu ^{}x^\nu G_{[\mu \nu ]}`$ instead of $`G_{[\mu ]}`$ in order not to come into conflict with the antisymmetrization of the indices $`\alpha `$ and $`\beta `$.
$`G_{[\alpha \beta ]}^{\mathrm{tw3}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`2{\displaystyle _0^1}d\lambda \lambda _{[\alpha }G_{\beta ]}^{}(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}`$
$`G_{[\alpha \beta ]}^{\mathrm{tw4}(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})G_{[\alpha \beta ]}^{\mathrm{tw5}(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$ (3.47)
where
$`G_{[\alpha \beta ]}^{\mathrm{tw4}(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`=x_{[\alpha }_{\beta ]}^\mu {\displaystyle _0^1}d\lambda \frac{1\lambda ^2}{\lambda }G_\mu ^{}(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$ (3.48)
$`G_{[\alpha \beta ]}^{\mathrm{tw5}(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`=x_{[\alpha }\left(\delta _{\beta ]}^\mu (x)x^\mu _{\beta ]}\right)\mathrm{}{\displaystyle _0^1}d\lambda \frac{\left(1\lambda \right)^2}{2\lambda }G_\mu ^{}(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}.`$ (3.49)
The local twist–4 operators are contained in the tensor space $`𝐓(\frac{n}{2},\frac{n}{2})`$ and the local twist–5 operators are contained in the tensor space $`𝐓(\frac{n}{2},\frac{n2}{2})𝐓(\frac{n2}{2},\frac{n}{2})`$.
In turn, let us remark that the bilocal light–ray operator $`G_{[\alpha \beta ]}^{\mathrm{tw3}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ is the same as the antisymmetric tensor operator which obtains from the Young tableau . That tableau has been used in Part I for the computation of $`M_{[\alpha \beta ]}^{\mathrm{tw2}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$. Furthermore, let us note that the local operators contained in $`M_{[\alpha \beta ]}^{\mathrm{tw2}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ and in $`G_{[\alpha \beta ]}^{\mathrm{tw3}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ are in complete agreement with the local operators on the light–cone determined by Dobrev and Ganchev by means of the interior derivative.
From the above point of view our result for the antisymmetric part of twist–3 operator (3.46) could be obtained much easier. However, the symmetric twist–3 light–cone tensor operator is much more involved. It is given by<sup>7</sup><sup>7</sup>7 Similarly, we use the notation $`G_\mu ^+`$ instead of $`G_{(\mu )}`$.
$`G_{(\alpha \beta )}^{\mathrm{tw3}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`{\displaystyle _0^1}d\lambda \frac{1\lambda ^2}{2\lambda }\left(\delta _{(\alpha }^\mu (x)x^\mu _{(\alpha }\right)_{\beta )}G_\mu ^+(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}`$
$`G_{(\alpha \beta )}^{>(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$ (3.50)
where $`G_{(\alpha \beta )}^{>(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ includes all the trace terms having higher twist; these operators are given by
$`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{ii})\mathrm{a}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`g_{\alpha \beta }\mathrm{}{\displaystyle _0^1}d\lambda \left(\frac{1\lambda ^2}{2\lambda }+\lambda \mathrm{ln}\lambda \right)G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{ii})\mathrm{b}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`x_{(\alpha }_{\beta )}\mathrm{}{\displaystyle _0^1}d\lambda ((1\lambda )+\frac{1\lambda ^2}{2\lambda }+2\lambda \mathrm{ln}\lambda )G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{ii})\mathrm{b}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$
$`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{ii})\mathrm{d1}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`x_{(\alpha }_{\beta )}\mathrm{}{\displaystyle _0^1}d\lambda (\frac{1\lambda }{\lambda }+\frac{\mathrm{ln}\lambda }{\lambda })G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{ii})\mathrm{d1}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$
and
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{ii})\mathrm{a}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{2}x_\alpha x_\beta \mathrm{}^2{\displaystyle _0^1}d\lambda \left(\frac{\left(1\lambda \right)^2}{\lambda }\frac{1\lambda ^2}{2\lambda }\lambda \mathrm{ln}\lambda \right)G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{ii})\mathrm{b}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{2}x_\alpha x_\beta \mathrm{}^2{\displaystyle _0^1}\mathrm{d}\lambda (\frac{\left(1\lambda \right)^2}{2\lambda }\frac{3}{2}\frac{1\lambda ^2}{2\lambda }`$
$`\lambda \mathrm{ln}\lambda \frac{\mathrm{ln}\lambda }{2\lambda })G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{ii})\mathrm{d1}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{2}x_\alpha x_\beta \mathrm{}^2{\displaystyle _0^1}d\lambda \left(\frac{1\lambda }{\lambda }+\frac{\mathrm{ln}\lambda }{\lambda }+\frac{\mathrm{ln}^2\lambda }{2\lambda }\right)G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{ii})\mathrm{d3}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`x_\alpha x_\beta \mathrm{}^2{\displaystyle _0^1}d\lambda \left(\frac{1\lambda }{\lambda }+\frac{\mathrm{ln}\lambda }{\lambda }+\frac{\mathrm{ln}^2\lambda }{2\lambda }\right)G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
being related to the scalar operator $`G(\kappa _1x,\kappa _2x)`$, as well as
$`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{ii})\mathrm{c}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`g_{\alpha \beta }^\mu {\displaystyle _0^1}d\lambda \frac{1\lambda ^2}{\lambda }G_\mu ^+(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{ii})\mathrm{d2}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\mathrm{\hspace{0.17em}2}x_{(\alpha }_{\beta )}^\mu {\displaystyle _0^1}d\lambda \left(\frac{1\lambda }{\lambda }+\frac{\mathrm{ln}\lambda }{\lambda }\right)G_\mu ^+(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{ii})\mathrm{d2}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$
$`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{ii})\mathrm{e}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\mathrm{\hspace{0.17em}2}x_{(\alpha }_{\beta )}^\mu {\displaystyle _0^1}d\lambda (1\lambda )G_\mu ^+(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{ii})\mathrm{e}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$
and
$`G_{(\alpha \beta )}^{\mathrm{tw5}(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`x_{(\alpha }\left(\delta _{\beta )}^\mu (x)x^\mu _{\beta )}\right)\mathrm{}`$
$`\times {\displaystyle _0^1}\mathrm{d}\lambda (\frac{1\lambda }{\lambda }+\frac{\mathrm{ln}\lambda }{\lambda })G_\mu ^+(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{ii})\mathrm{d3}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{ii})\mathrm{d4}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$
and
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{ii})\mathrm{c}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{2}x_\alpha x_\beta \mathrm{}^\mu {\displaystyle _0^1}d\lambda \frac{\left(1\lambda \right)^2}{\lambda }G_\mu ^+(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{ii})\mathrm{d2}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`x_\alpha x_\beta \mathrm{}^\mu {\displaystyle _0^1}d\lambda (\frac{1\lambda }{\lambda }+\frac{\mathrm{ln}\lambda }{\lambda }+\frac{\mathrm{ln}^2\lambda }{2\lambda })G_\mu ^+(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{ii})\mathrm{d4}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`x_\alpha x_\beta \mathrm{}^\mu {\displaystyle _0^1}d\lambda \left(\frac{1\lambda }{\lambda }+\frac{\mathrm{ln}\lambda }{\lambda }\right)G_\mu ^+(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{ii})\mathrm{e}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{2}x_\alpha x_\beta \mathrm{}^\mu {\displaystyle _0^1}d\lambda \frac{\left(1\lambda \right)^2}{\lambda }G_\mu ^+(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
being related to the (symmetric) vector operator $`G_\mu ^+(\kappa _1x,\kappa _2x)`$.
#### 3.2.6 Determination of complete light–cone vector operators
Now we are able to determine the twist–3 vector operator resulting from the tensor operator (3.46) by multiplying with $`x^\beta `$. Because neither symmetry type (iii) nor (iv) may contribute to the vector operator this will be the final expression. The twist–3 light–cone vector operator consists of two parts, one originating from the symmetric and the other from the antisymmetric tensor operator,
$`G_\alpha ^{\mathrm{tw3}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=G_{(\alpha )}^{\mathrm{tw3}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{[\alpha ]}^{\mathrm{tw3}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$ (3.51)
where
$`G_{(\alpha )}^{\mathrm{tw3}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`={\displaystyle _0^1}d\lambda \lambda \left[\delta _\alpha ^\mu (x)x^\mu _\alpha x_\alpha (^\mu +\mathrm{ln}\lambda \mathrm{}x^\mu )\right]G_\mu ^+(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$ (3.52)
$`G_{[\alpha ]}^{\mathrm{tw3}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`={\displaystyle _0^1}d\lambda \lambda \left[\delta _\alpha ^\mu (x)x^\mu _\alpha x_\alpha ^\mu \right]G_\mu ^{}(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$ (3.53)
and the twist–4 vector operator which is contained in the trace terms of that twist–3 vector operator is given by
$`G_\alpha ^{\mathrm{tw4}(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`=G_{(\alpha )}^{\mathrm{tw4}(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{[\alpha ]}^{\mathrm{tw4}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$ (3.54)
where
$`G_{(\alpha )}^{\mathrm{tw4}(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`=x_\alpha {\displaystyle _0^1}d\lambda \lambda \left[^\mu +(\mathrm{ln}\lambda )\mathrm{}x^\mu \right]G_\mu ^+(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$ (3.55)
$`G_{[\alpha ]}^{\mathrm{tw4}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`=x_\alpha {\displaystyle _0^1}d\lambda \lambda ^\mu G_\mu ^{}(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}.`$ (3.56)
The antisymmetric part of the twist–4 vector operator is already complete. The complete twist–4 light–cone vector operator is given by
$`G_\alpha ^{\mathrm{tw4}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`=G_{(\alpha )}^{\mathrm{tw4}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{[\alpha ]}^{\mathrm{tw4}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$
$`=x_\alpha {\displaystyle _0^1}d\lambda \lambda \left[^\mu +\frac{\mathrm{ln}\lambda }{2}\mathrm{}x^\mu \right]G_\mu (\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$ (3.57)
where the symmetric twist–4 vector operator obtains by adding together expressions (3.25) and (3.54),
$`G_{(\alpha )}^{\mathrm{tw4}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`=G_{(\alpha )}^{\mathrm{tw4}(\mathrm{i})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha )}^{\mathrm{tw4}(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ (3.58)
$`=x_\alpha {\displaystyle _0^1}d\lambda \lambda \left[^\mu +\frac{\mathrm{ln}\lambda }{2}\mathrm{}x^\mu \right]G_\mu ^+(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}.`$
Furthermore, putting together expressions (3.24), (3.52), (3.58) and (3.53), (3.56), respectively, we obtain for the complete decomposition of the light–cone vector operator the following results anticipated by eqs. (3.9) and (3.10), respectively:
$`G_{(\alpha )}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`=G_{(\alpha )}^{\mathrm{tw2}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha )}^{\mathrm{tw3}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha )}^{\mathrm{tw4}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ (3.59)
$`G_{[\alpha ]}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`=G_{[\alpha ]}^{\mathrm{tw3}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{[\alpha ]}^{\mathrm{tw4}(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ (3.60)
### 3.3 Tensor operators of symmetry class (iii)
#### 3.3.1 Construction of the nonlocal antisymmetric class-(iii) operators of definite twist
Now we consider the twist–4 and twist–5 contributions originating from the (unique) Young tableau related to the symmetry class (iii) (with the normalizing factor $`\beta _n`$):
The corresponding traceless local operator having twist $`\tau =4`$ and being contained in the tensor space $`𝐓(\frac{n}{2},\frac{n}{2})`$ is given by:
$`G_{\alpha \beta \mu _1\mathrm{}\mu _n}^{\mathrm{tw4}(\mathrm{iii})}`$ $`=`$ $`\frac{n}{n+2}(F_{[\alpha |}^\rho (0)D_{(\mu _1}\mathrm{}D_{\mu _n)}F_{|\beta ]\rho }(0)F_{[\alpha |}^\rho (0)D_{(\beta }{}_{\mu _2}{}^{}\mathrm{}D_{\mu _n)}F_{|\mu _1]\rho }(0)`$
$`F_{[\mu _1|}^\rho (0)D_{(\alpha }D_{\mu _2}\mathrm{}D_{\mu _n)}F_{|\beta ]\rho }(0))\mathrm{trace}\mathrm{terms}.`$
Now, contracting this expression with $`x^{\mu _1}\mathrm{}x^{\mu _n}`$ we obtain:
$`G_{[\alpha \beta ]n}^{\mathrm{tw4}(\mathrm{iii})}(x)`$ $`=`$ $`x^{\mu _1}\mathrm{}x^{\mu _n}G_{[\alpha \beta ]\mu _1\mathrm{}\mu _n}^{\mathrm{tw4}}`$
$`=`$ $`\frac{1}{n+2}\left((x)\delta _{[\beta }^\nu 2x^\nu _{[\beta }\right)\delta _{\alpha ]}^\mu \stackrel{}{G}_{[\mu \nu ]n}(x).`$
with the (antisymmetric) tensorial harmonic polynomials of order $`n`$ (see eq. (A)):
$`\stackrel{}{G}_{[\alpha \beta ]n}(x)=`$ $`\{\delta _\alpha ^\mu \delta _\beta ^\nu +\frac{1}{\left(n+1\right)n}(2x_{[\alpha }\delta _{\beta ]}^{[\mu }^{\nu ]}(x)x^2_{[\alpha }\delta _{\beta ]}^{[\mu }^{\nu ]})`$
$`\frac{2}{\left(n+2\right)\left(n+1\right)n}x_{[\alpha }_{\beta ]}x^{[\mu }^{\nu ]}\left\}H^{(4)}_n\right(x^2|\mathrm{})G_{[\mu \nu ]n}(x).`$
Resumming these local terms gives the nonlocal twist–4 operator as follows
$`G_{[\alpha \beta ]}^{\mathrm{tw4}(\mathrm{iii})}(0,\kappa x)={\displaystyle _0^1}d\lambda \lambda \delta _{[\alpha }^\mu \left((x)\delta _{\beta ]}^\nu 2x^\nu _{\beta ]}\right)\stackrel{}{G}_{[\mu \nu ]}(0,\kappa \lambda x),`$ (3.61)
where, using the integral representation of the additional factor $`1/n`$ (the remaining factors of the denominator are taken together with $`1/n!`$ to get $`1/(n+2)!`$) and of the beta function, we introduced:
$`\stackrel{}{G}_{[\alpha \beta ]}(0,\kappa x)=G_{[\alpha \beta ]}(0,\kappa x)+{\displaystyle \underset{k=1}{\overset{\mathrm{}}{}}}{\displaystyle _0^1}{\displaystyle \frac{\mathrm{d}t}{t}}({\displaystyle \frac{x^2}{4}})^k{\displaystyle \frac{\mathrm{}^k}{k!(k1)!}}({\displaystyle \frac{1t}{t}})^{k1}G_{[\alpha \beta ]}(0,\kappa tx)`$ (3.62)
$`+{\displaystyle _0^1}{\displaystyle \frac{\mathrm{d}\tau }{\tau }}\{(2x_{[\alpha }\delta _{\beta ]}^{[\mu }^{\nu ]}(x)x^2_{[\alpha }\delta _{\beta ]}^{[\mu }^{\nu ]}){\displaystyle \underset{k=0}{\overset{\mathrm{}}{}}}{\displaystyle _0^1}\mathrm{d}t({\displaystyle \frac{x^2}{4}})^k{\displaystyle \frac{\mathrm{}^k}{k!k!}}({\displaystyle \frac{1t}{t}})^k`$
$`\mathrm{\hspace{0.17em}2}x_{[\alpha }_{\beta ]}x^{[\mu }^{\nu ]}{\displaystyle \underset{k=0}{\overset{\mathrm{}}{}}}{\displaystyle _0^1}\mathrm{d}tt({\displaystyle \frac{x^2}{4}})^k{\displaystyle \frac{\mathrm{}^k}{(k+1)!k!}}({\displaystyle \frac{1t}{t}})^{k+1}\}G_{[\mu \nu ]}(0,\kappa \tau tx).`$
Then, by construction, eq. (3.61) defines a nonlocal operator of twist–4. As is easily seen by partial integration it fulfils the following relation
$$G_{[\alpha \beta ]}^{\mathrm{tw4}(\mathrm{iii})}(0,\kappa x)=\stackrel{}{G}_{[\alpha \beta ]}(0,\kappa x)G_{[\alpha \beta ]}^{\mathrm{tw3}}(0,\kappa x)$$
(3.63)
and, because of eq. (3.2.1) and the properties of (3.62), it is a harmonic tensor operator:
$$\mathrm{}G_{[\alpha \beta ]}^{\mathrm{tw4}(\mathrm{iii})}(0,\kappa x)=0,^\alpha G_{[\alpha \beta ]}^{\mathrm{tw4}(\mathrm{iii})}(0,\kappa x)=0.$$
(3.64)
#### 3.3.2 Projection onto the light–cone
Let us now project onto the light–cone. After the same calculations as has been carried out for the operator $`M_{[\alpha \beta ]}^{\mathrm{tw3}}(0,\kappa \stackrel{~}{x})`$ in Part I we obtain
$`G_{[\alpha \beta ]}^{\mathrm{tw4}(\mathrm{iii})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`{\displaystyle _0^1}d\lambda \lambda \delta _{[\alpha }^\mu \left((x)\delta _{\beta ]}^\nu 2x^\nu _{\beta ]}\right)G_{[\mu \nu ]}(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}`$ (3.65)
$`G_{[\alpha \beta ]}^{\mathrm{tw5}(\mathrm{iii})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$
where the twist–5 part is determined by the trace terms, namely
$`G_{[\alpha \beta ]}^{\mathrm{tw5}(\mathrm{iii})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`{\displaystyle _0^1}\mathrm{d}\lambda \frac{1\lambda ^2}{\lambda }\{x_{[\alpha }(\delta _{\beta ]}^{[\mu }(x)x^{[\mu }_{\beta ]})^{\nu ]}`$
$`x_{[\alpha }\delta _{\beta ]}^{[\mu }x^{\nu ]}\mathrm{}\}G_{[\mu \nu ]}(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}.`$ (3.66)
Obviously, there exist no vector (and scalar) operators of symmetry type (iii).
#### 3.3.3 Determination of the complete antisymmetric light–cone tensor operator
This finishes the computation of twist contributions of the antisymmetric light–cone tensor operator. The complete antisymmetric twist–4 light–cone tensor operator obtains from eq. (3.65) and the twist–4 trace terms of the twist–3 operator, eq. (3.48):
$`G_{[\alpha \beta ]}^{\mathrm{tw4}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=G_{[\alpha \beta ]}^{\mathrm{tw4}(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{[\alpha \beta ]}^{\mathrm{tw4}(\mathrm{iii})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$
$`={\displaystyle _0^1}\mathrm{d}\lambda \{\lambda ((x)\delta _{[\beta }^\nu 2x^\nu _{[\beta })\delta _{\alpha ]}^\mu \frac{1\lambda ^2}{\lambda }(x_{[\alpha }_{\beta ]}x^{[\mu }^{\nu ]}`$
$`+x_{[\alpha }(\delta _{\beta ]}^{[\mu }(x)x^{[\mu }_{\beta ]})^{\nu ]}x_{[\alpha }\delta _{\beta ]}^{[\mu }x^{\nu ]}\mathrm{})\}G_{[\mu \nu ]}(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}.`$
Furthermore, the complete antisymmetric twist–5 light–cone tensor operator obtains from eqs. (3.49) and (3.3.2) as follows:
$`G_{[\alpha \beta ]}^{\mathrm{tw5}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=G_{[\alpha \beta ]}^{\mathrm{tw5}(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{[\alpha \beta ]}^{\mathrm{tw5}(\mathrm{iii})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$
$`={\displaystyle _0^1}d\lambda \frac{1\lambda }{\lambda }\left\{x_{[\alpha }\delta _{\beta ]}^{[\mu }x^{\nu ]}\mathrm{}2x_{[\alpha }\left(\delta _{\beta ]}^{[\mu }(x)x^{[\mu }_{\beta ]}\right)^{\nu ]}\right\}G_{[\mu \nu ]}(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}.`$
Together with the twist-3 part, eq. (3.2.5), we finally obtain the complete decomposition of the antisymmetric light–cone tensor operator (compare eq. (3.8)):
$$G_{[\alpha \beta ]}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=G_{[\alpha \beta ]}^{\mathrm{tw3}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{[\alpha \beta ]}^{\mathrm{tw4}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{[\alpha \beta ]}^{\mathrm{tw5}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}).$$
(3.69)
### 3.4 Tensor operators of symmetry class (iv)
#### 3.4.1 Construction of the nonlocal symmetric class-(iv) operator of definite twist
Now let us investigate the symmetric tensor operators of the symmetry class (iv) which are given by the following Young tableau (with normalizing factor $`\gamma _n`$):
These local tensor operators have twist–4 and transform according to the representation $`𝐓(\frac{n+2}{2},\frac{n2}{2})𝐓(\frac{n2}{2},\frac{n+2}{2})`$. They are given by:
$`G_{\alpha \beta \mu _1\mathrm{}\mu _n}^{\mathrm{tw4}(\mathrm{iv})}=`$ $`\frac{1}{\left(n+1\right)n}(G_{(\alpha \beta )(\mu _1\mathrm{}\mu _n)}G_{(\alpha \mu _2)(\mu _1\beta \mu _3\mathrm{}\mu _n)}G_{(\mu _1\beta )(\alpha \mu _2\mathrm{}\mu _n)}`$
$`+G_{(\mu _1\mu _2)(\alpha \beta \mu _3\mathrm{}\mu _n)})\mathrm{trace}\mathrm{terms},`$
with $`G_{(\alpha \beta )(\mu _1\mathrm{}\mu _n)}F_{(\alpha |}^\rho (0)D_{(\mu _1}\mathrm{}D_{\mu _n)}F_{|\beta )\rho }(0)`$. Multiplying with $`x^{\mu _1}\mathrm{}x^{\mu _n}`$ we obtain:
$`G_{(\alpha \beta )n}^{\mathrm{tw4}(\mathrm{iv})}(x)`$ $`=x^{\mu _1}\mathrm{}x^{\mu _n}G_{\alpha \beta \mu _1\mathrm{}\mu _n}^{\mathrm{tw4}(\mathrm{iv})}`$
$`=\frac{1}{\left(n+1\right)n}\left(\delta _\alpha ^\mu \delta _\beta ^\nu x^\rho (x)_\rho 2x^\mu (x)\delta _{(\alpha }^\nu _{\beta )}+x^\mu x^\nu _\alpha _\beta \right)\stackrel{}{G}_{(\mu \nu )n}(x).`$
Here, $`\stackrel{}{G}_{(\mu \nu )n}(x)`$ is the harmonic symmetric tensor polynomial of order $`n`$ (cf. Eq. (A), Appendix A):
$`\stackrel{}{G}_{(\alpha \beta )n}(x)=`$ $`\{\delta _\alpha ^\mu \delta _\beta ^\nu +\frac{n}{\left(n+2\right)^2}g_{\alpha \beta }x^{(\nu }^{\mu )}`$
$`\frac{1}{\left(n+2\right)\left(n+1\right)}\left(2nx_{(\alpha }\delta _{\beta )}^{(\nu }^{\mu )}x^2\delta _{(\alpha }^{(\nu }_{\beta )}^{\nu )}\right)`$
$`\frac{1}{\left(n+2\right)^2\left(n+1\right)}\left(2nx_{(\alpha }_{\beta )}x^{(\nu }^{\mu )}x^2_\alpha _\beta x^{(\nu }^{\mu )}\right)`$
$`+\frac{\left(n+3\right)n}{\left(n+2\right)^2\left(n+1\right)^2}\left(x_\alpha x_\beta \frac{1}{2}x^2g_{\alpha \beta }\right)^\mu ^\nu `$
$`\frac{1}{\left(n+2\right)^2\left(n+1\right)^2}(2x^2_{(\alpha }x_{\beta )}+\frac{1}{4}x^4_\alpha _\beta )^\mu ^\nu \}\stackrel{˘}{G}_{(\mu \nu )n}(x).`$
Resumming all local operators of symmetry class (iv) and using the integral representation $`((n+1)n)^1=_0^1d\lambda \lambda ^n(1\lambda )/\lambda `$ gives
$`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iv})}(0,\kappa x)`$ $`=\left(\delta _\alpha ^\mu \delta _\beta ^\nu x^\rho (x)_\rho 2x^\mu (x)\delta _{(\beta }^\nu _{\alpha )}+x^\mu x^\nu _\alpha _\beta \right)`$
$`\times {\displaystyle _0^1}\mathrm{d}\lambda \frac{1\lambda }{\lambda }\stackrel{}{G}_{(\mu \nu )}(0,\kappa \lambda x),`$ (3.70)
where $`\stackrel{}{G}_{(\mu \nu )}(0,\kappa x)`$ is given by
$`\stackrel{}{G}_{(\alpha \beta )}(0,\kappa x)=\stackrel{˘}{G}_{(\alpha \beta )}(0,\kappa x)+{\displaystyle \underset{k=1}{\overset{\mathrm{}}{}}}{\displaystyle _0^1}{\displaystyle \frac{\mathrm{d}t}{t}}({\displaystyle \frac{x^2}{4}})^k{\displaystyle \frac{\mathrm{}^k}{k!(k1)!}}({\displaystyle \frac{1t}{t}})^{k1}\stackrel{˘}{G}_{(\alpha \beta )}(0,\kappa tx)`$ (3.71)
$`\left[2x_{(\alpha }\delta _{\beta )}^{(\mu }^{\nu )}(x)x^2_{(\alpha }\delta _{\beta )}^{(\mu }^{\nu )}\right]`$
$`\times {\displaystyle \underset{k=0}{\overset{\mathrm{}}{}}}{\displaystyle _0^1}\mathrm{d}tt({\displaystyle \frac{x^2}{4}})^k{\displaystyle \frac{\mathrm{}^k}{k!(k+1)!}}({\displaystyle \frac{1t}{t}})^{k+1}\stackrel{˘}{G}_{(\mu \nu )}(0,\kappa tx)`$
$`\{[2x_{(\alpha }_{\beta )}x^{(\mu }^{\nu )}(x)g_{\alpha \beta }x^{(\mu }^{\nu )}(x)(x+1)x^2_\alpha _\beta x^{(\mu }^{\nu )}]{\displaystyle _0^1}\mathrm{d}\lambda \lambda `$
$`[(x_\alpha x_\beta \frac{x^2}{2}g_{\alpha \beta })^\mu ^\nu (x)(x+3)(2x^2_{(\alpha }x_{\beta )}+\frac{x^4}{4}_\alpha _\beta )^\mu ^\nu ]{\displaystyle _0^1}\mathrm{d}\lambda (1\lambda )\}`$
$`\times {\displaystyle \underset{k=0}{\overset{\mathrm{}}{}}}{\displaystyle _0^1}\mathrm{d}tt({\displaystyle \frac{x^2}{4}})^k{\displaystyle \frac{\mathrm{}^k}{(k+1)!k!}}({\displaystyle \frac{1t}{t}})^{k+1}\stackrel{˘}{G}_{(\mu \nu )}(0,\kappa \lambda tx).`$
Of course, eq. (3.4.1) we may be rewritten as follows
$`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iv})}(0,\kappa x)=`$ $`\stackrel{}{G}_{(\alpha \beta )}(0,\kappa x)2{\displaystyle _0^1}d\lambda \frac{1}{\lambda }_{(\alpha }\stackrel{}{G_{\beta )}^+}(0,\kappa \lambda x)+{\displaystyle _0^1}d\lambda \frac{1\lambda }{\lambda }_\alpha _\beta \stackrel{}{G}(0,\kappa \lambda x),`$
which will be used in the further considerations. Contrary to any of the formerly obtained symmetric twist–4 tensor operators, this operator is the ‘true tensor part’, i.e., not being proportional to $`g_{\alpha \beta }`$ or $`x_\alpha `$ and $`x_\beta `$. Obviously, this twist–4 operator satisfies the conditions of tracelessness
$$\mathrm{}G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iv})}(0,\kappa x)=0,^\alpha G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iv})}(0,\kappa x)=0,g^{\alpha \beta }G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iv})}(0,\kappa x).$$
(3.72)
#### 3.4.2 Projection onto the light–cone
The symmetric twist–4 tensor operator of symmetry type (iv) on the light–cone is obtained as follows
$`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iv})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\left(\delta _\alpha ^\mu \delta _\beta ^\nu x^\rho (x)_\rho 2x^\mu (x)\delta _{(\beta }^\nu _{\alpha )}+x^\mu x^\nu _\alpha _\beta \right)`$ (3.73)
$`\times {\displaystyle _0^1}\mathrm{d}\lambda \frac{1\lambda }{\lambda }G_{(\mu \nu )}(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}G_{(\alpha \beta )}^{>(\mathrm{iv})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$
where the trace terms of higher twist are given by
$`G_{(\alpha \beta )}^{>(\mathrm{iv})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})={\displaystyle _0^1}\mathrm{d}\lambda \{(2\lambda x_{(\alpha }\delta _{\beta )}^\nu ^\mu (1\lambda )x_\alpha x_\beta ^\mu ^\nu )G_{(\mu \nu )}(\kappa _1\lambda x,\kappa _2\lambda x)`$ (3.74)
$`\left(\frac{1}{\lambda }g_{\alpha \beta }^\mu +\frac{1\lambda ^2}{\lambda }x_{(\alpha }\delta _{\beta )}^\mu \mathrm{}\frac{\left(1\lambda \right)^2}{2\lambda }x_\alpha x_\beta \mathrm{}^\mu \right)G_\mu ^+(\kappa _1\lambda x,\kappa _2\lambda x)`$
$`+\left(\frac{1\lambda }{2\lambda }g_{\alpha \beta }\mathrm{}+\frac{\left(1\lambda \right)^2}{2\lambda }x_{(\alpha }_{\beta )}\mathrm{}\frac{1}{2}\left(\frac{1\lambda ^2}{2\lambda }+\mathrm{ln}\lambda \right)x_\alpha x_\beta \mathrm{}^2\right)G(\kappa _1\lambda x,\kappa _2\lambda x)`$
$`+(\frac{1}{2\lambda }g_{\alpha \beta }(x)\lambda x_{(\alpha }_{\beta )}\frac{\lambda \mathrm{ln}\lambda }{2}x_\alpha x_\beta \mathrm{})G_\rho ^\rho (\kappa _1\lambda x,\kappa _2\lambda x)\left\}\right|_{x=\stackrel{~}{x}}.`$
The explicit twist decomposition of this operator by means of the Young tableaux (i) and (ii) gives
$`G_{(\alpha \beta )}^{>(\mathrm{iv})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`=G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iv})\mathrm{a}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iv})\mathrm{b}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iv})\mathrm{c}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$
$`+G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iv})\mathrm{d1}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iv})\mathrm{d2}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iv})\mathrm{e1}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$
$`+G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iv})\mathrm{e2}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iv})\mathrm{f}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iv})\mathrm{h}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$
$`+G_{(\alpha \beta )}^{\mathrm{tw5}(\mathrm{iv})\mathrm{d}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw5}(\mathrm{iv})\mathrm{e}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$
$`+G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{a}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{b}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{c}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$
$`+G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{d1}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{d2}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{d3}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$
$`+G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{d4}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{e1}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{e2}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$
$`+G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{e3}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{e4}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{f}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$
$`+G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{h}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{k}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$ (3.75)
with the following operators of well-defined twist
$`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iv})\mathrm{a}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`g_{\alpha \beta }\mathrm{}{\displaystyle _0^1}d\lambda \frac{1\lambda }{2\lambda }G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iv})\mathrm{b}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`x_{(\alpha }_{\beta )}\mathrm{}{\displaystyle _0^1}d\lambda \frac{\left(1\lambda \right)^2}{2\lambda }G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{b}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$
$`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iv})\mathrm{d1}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`x_{(\alpha }_{\beta )}\mathrm{}{\displaystyle _0^1}d\lambda \left(\frac{1\lambda ^2}{2\lambda }+\frac{\mathrm{ln}\lambda }{\lambda }\right)G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{d1}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$
and
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{a}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{2}x_\alpha x_\beta \mathrm{}^2{\displaystyle _0^1}d\lambda \left(\frac{1\lambda ^2}{2\lambda }+\mathrm{ln}\lambda \right)G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{b}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{2}x_\alpha x_\beta \mathrm{}^2{\displaystyle _0^1}d\lambda \left(\frac{1\lambda }{\lambda }\frac{1\lambda ^2}{4\lambda }+\frac{\mathrm{ln}\lambda }{2\lambda }\right)G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{d1}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{4}x_\alpha x_\beta \mathrm{}^2{\displaystyle _0^1}d\lambda \left(\frac{1\lambda ^2}{2\lambda }+\frac{\mathrm{ln}\lambda }{\lambda }+\frac{\mathrm{ln}^2\lambda }{\lambda }\right)G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{d3}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{2}x_\alpha x_\beta \mathrm{}^2{\displaystyle _0^1}d\lambda \left(\frac{1\lambda ^2}{2\lambda }+\frac{\mathrm{ln}\lambda }{\lambda }+\frac{\mathrm{ln}^2\lambda }{\lambda }\right)G(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
being related to the scalar operator, as well as
$`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iv})\mathrm{c}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`g_{\alpha \beta }^\mu {\displaystyle _0^1}d\lambda \frac{1}{\lambda }G_\mu ^+(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iv})\mathrm{d2}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`2x_{(\alpha }_{\beta )}^\mu {\displaystyle _0^1}d\lambda \left(\frac{1\lambda ^2}{2\lambda }+\frac{\mathrm{ln}\lambda }{\lambda }\right)G_\mu ^+(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{d2}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$
$`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iv})\mathrm{e1}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`x_{(\alpha }_{\beta )}^\mu {\displaystyle _0^1}d\lambda \frac{1\lambda ^2}{\lambda }G_\mu ^+(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{e1}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$
and
$`G_{(\alpha \beta )}^{\mathrm{tw5}(\mathrm{iv})\mathrm{d}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`x_{(\alpha }\left(\delta _{\beta )}^\mu (x)x^\mu _{\beta )}\right)\mathrm{}`$
$`\times {\displaystyle _0^1}\mathrm{d}\lambda (\frac{1\lambda ^2}{2\lambda }+\frac{\mathrm{ln}\lambda }{\lambda })G_\mu ^+(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{d3}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{d4}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$ (3.76)
and
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{c}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{2}x_\alpha x_\beta \mathrm{}^\mu {\displaystyle _0^1}d\lambda \frac{\left(1\lambda \right)^2}{\lambda }G_\mu ^+(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{d2}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{2}x_\alpha x_\beta \mathrm{}^\mu {\displaystyle _0^1}d\lambda \left(\frac{1\lambda ^2}{2\lambda }+\frac{\mathrm{ln}\lambda }{\lambda }+\frac{\mathrm{ln}^2\lambda }{\lambda }\right)G_\mu ^+(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{d4}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`x_\alpha x_\beta \mathrm{}^\mu {\displaystyle _0^1}d\lambda \left(\frac{1\lambda ^2}{2\lambda }+\frac{\mathrm{ln}\lambda }{\lambda }\right)G_\mu ^+(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{e1}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{2}x_\alpha x_\beta \mathrm{}^\mu {\displaystyle _0^1}d\lambda \left(\frac{1\lambda ^2}{2\lambda }+\frac{\mathrm{ln}\lambda }{\lambda }\right)G_\mu ^+(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{e3}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`x_\alpha x_\beta \mathrm{}^\mu {\displaystyle _0^1}d\lambda \left(\frac{1\lambda ^2}{2\lambda }+\frac{\mathrm{ln}\lambda }{\lambda }\right)G_\mu ^+(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
being related to the symmetric vector operator,
$`G_{(\alpha \beta )}^{\mathrm{tw5}(\mathrm{iv})\mathrm{e}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`x_{(\alpha }\left(\delta _{\beta )}^\nu (x)x^\nu _{\beta )}\right)^\mu {\displaystyle _0^1}d\lambda \frac{1\lambda ^2}{\lambda }G_{(\mu \nu )}(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{e3}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{e4}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{e4}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`x_\alpha x_\beta ^\mu ^\nu {\displaystyle _0^1}d\lambda \frac{1\lambda ^2}{\lambda }G_{(\mu \nu )}(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{f}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`x_\alpha x_\beta ^\mu ^\nu {\displaystyle _0^1}d\lambda (1\lambda )G_{(\mu \nu )}(\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
being related to the symmetric tensor operator and, furthermore,
$`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iv})\mathrm{f}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`g_{\alpha \beta }(x){\displaystyle _0^1}d\lambda \frac{1}{2\lambda }G_\rho ^\rho (\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iv})\mathrm{h}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`x_{(\alpha }_{\beta )}{\displaystyle _0^1}d\lambda \lambda G_\rho ^\rho (\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{h}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$
$`G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iv})\mathrm{e2}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`x_{(\alpha }_{\beta )}{\displaystyle _0^1}d\lambda \frac{1\lambda ^2}{\lambda }G_\rho ^\rho (\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{e2}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$
and
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{h}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{4}x_\alpha x_\beta \mathrm{}{\displaystyle _0^1}d\lambda \frac{1\lambda ^2}{\lambda }G_\rho ^\rho (\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{e2}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{2}x_\alpha x_\beta \mathrm{}{\displaystyle _0^1}d\lambda \left(\frac{1\lambda ^2}{2\lambda }+\frac{\mathrm{ln}\lambda }{\lambda }\right)G_\rho ^\rho (\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`G_{(\alpha \beta )}^{\mathrm{tw6}(\mathrm{iv})\mathrm{k}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`\frac{1}{2}x_\alpha x_\beta \mathrm{}{\displaystyle _0^1}d\lambda \lambda (\mathrm{ln}\lambda )G_\rho ^\rho (\kappa _1\lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
being related to the trace of the symmetric tensor operator.
Here, it should be remarked that, contrary to the former cases (i) – (iii), the trace terms (3.74) contain operators having the same twist, $`\tau =4`$, as the primary operator of symmetry class (iv). However, these operators being multiplied by either $`g_{\alpha \beta }`$ or $`x_\alpha `$ and $`x_\beta `$ are scalar or vector operators corresponding to symmetry type (i). Whereas the local operators resulting from the twist–4 tensor operator of symmetry class (iv) are contained in the tensor space $`𝐓(\frac{n+2}{2},\frac{n2}{2})𝐓(\frac{n2}{2},\frac{n+2}{2})`$, the local operators of the twist–4 scalar and vector operators are totally symmetric traceless tensors and, therefore, contained in the tensor space $`𝐓(\frac{n}{2},\frac{n}{2})`$.
#### 3.4.3 Determination of the complete symmetric light–cone tensor operators
Now, having finished the decomposition of a general 2nd rank tensor operator, let us sum up the remaining symmetric tensor operators of twist greater than 3, which appear in the trace terms of the symmetric twist operators with Young symmetry (i), (ii) as well as (iv), to complete twist–4, twist–5 and twist–6 operators.
The ‘scalar part’ of the twist–4 light–ray tensor operator is given by
$`G_{(\alpha \beta )}^{\mathrm{tw4},\mathrm{s}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=g_{\alpha \beta }\{\frac{1}{2}\mathrm{}{\displaystyle _0^1}\mathrm{d}\lambda \lambda (\mathrm{ln}\lambda )G(\kappa _1\lambda x,\kappa _2\lambda x)`$ (3.77)
$`(x){\displaystyle _0^1}\mathrm{d}\lambda \frac{1}{2\lambda }G_\rho ^\rho (\kappa _1\lambda x,\kappa _2\lambda x)+^\mu {\displaystyle _0^1}\mathrm{d}\lambda \lambda G_\mu ^+(\kappa _1\lambda x,\kappa _2\lambda x)\left\}\right|_{x=\stackrel{~}{x}},`$
and the ‘vector part’ of the twist–4 light–ray tensor operator reads
$`G_{(\alpha \beta )}^{\mathrm{tw4},\mathrm{v}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=\stackrel{~}{x}_{(\alpha }\times `$
$`\{{\displaystyle _0^1}\mathrm{d}\lambda \{(\frac{1\lambda ^2}{2\lambda }+\lambda \mathrm{ln}\lambda )_{\beta )}+\frac{1}{4}x_{\beta )}(\frac{1\lambda ^2}{\lambda }+\lambda \mathrm{ln}\lambda +\frac{\mathrm{ln}\lambda }{\lambda })\mathrm{}\}\mathrm{}G(\kappa _1\lambda x,\kappa _2\lambda x)`$
$`+{\displaystyle _0^1}\frac{\mathrm{d}\lambda }{\lambda }\left\{_{\beta )}+\frac{1}{2}x_{\beta )}(\mathrm{ln}\lambda )\mathrm{}\right\}G_\rho ^\rho (\kappa _1\lambda x,\kappa _2\lambda x)`$
$`{\displaystyle _0^1}\frac{\mathrm{d}\lambda }{\lambda }\{2(1\lambda ^2)_{\beta )}+\frac{1}{2}x_{\beta )}(1\lambda ^2+2\mathrm{ln}\lambda )\mathrm{}\}^\mu G_\mu ^+(\kappa _1\lambda x,\kappa _2\lambda x)\left\}\right|_{x=\stackrel{~}{x}}.`$
Furthermore, the contributions to the complete ‘vector part’ of the twist–5 light–ray tensor operator are constructed by means of the Young tableau (ii), and their local operators are antisymmetric<sup>8</sup><sup>8</sup>8 Note that antisymmetry is not in $`\alpha `$ and $`\beta `$ but results from the differential operator! traceless tensors contained in the space $`𝐓(\frac{n}{2},\frac{n2}{2})𝐓(\frac{n2}{2},\frac{n}{2})`$. This complete twist–5 tensor operator is given by
$`G_{(\alpha \beta )}^{\mathrm{tw5},\mathrm{v}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=\stackrel{~}{x}_{(\alpha }\times `$
$`\{{\displaystyle _0^1}\mathrm{d}\lambda \{\frac{\left(1\lambda \right)^2}{2\lambda }(\delta _{\beta )}^\nu (x)x^\nu _{\beta )}x_{\beta )}^\nu )\mathrm{}(\frac{1\lambda }{\lambda }\frac{1\lambda ^2}{4\lambda }+\frac{1}{2\lambda }\mathrm{ln}\lambda )x_{\beta )}\mathrm{}^2x^\nu \}`$
$`\times G_\nu ^+(\kappa _1\lambda x,\kappa _2\lambda x)`$
$`{\displaystyle _0^1}d\lambda \left\{\frac{1\lambda ^2}{\lambda }\left(\delta _{\beta )}^\nu (x)x^\nu _{\beta )}x_{\beta )}^\nu \right)^\mu \left(\frac{1\lambda ^2}{2\lambda }+\frac{\mathrm{ln}\lambda }{\lambda }\right)x_{\beta )}\mathrm{}^\mu x^\nu \right\}`$
$`\times G_{(\mu \nu )}(\kappa _1\lambda x,\kappa _2\lambda x)\left\}\right|_{x=\stackrel{~}{x}}.`$ (3.79)
The ‘scalar part’ of the twist–6 light–ray tensor operator has Young symmetry (i). Thus, the local twist–6 operators are totally symmetric traceless tensors lying in the space $`𝐓(\frac{n2}{2},\frac{n2}{2})`$. This nonlocal twist–6 operator is given by
$`G_{(\alpha \beta )}^{\mathrm{tw6},\mathrm{s}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=\stackrel{~}{x}_\alpha \stackrel{~}{x}_\beta `$ $`\{\frac{1}{2}\mathrm{}^2{\displaystyle _0^1}\mathrm{d}\lambda (\frac{1\lambda }{\lambda }+\frac{1+\lambda }{2\lambda }\mathrm{ln}\lambda )G(\kappa _1\lambda x,\kappa _2\lambda x)`$
$`\mathrm{}{\displaystyle _0^1}d\lambda \frac{1\lambda ^2}{2\lambda }G_\rho ^\rho (\kappa _1\lambda x,\kappa _2\lambda x)`$
$`+\mathrm{}{\displaystyle _0^1}d\lambda \frac{\left(1\lambda \right)^2}{2\lambda }^\mu G_\mu ^+(\kappa _1\lambda x,\kappa _2\lambda x)`$
$`^\mu ^\nu {\displaystyle _0^1}\mathrm{d}\lambda \frac{1\lambda }{\lambda }G_{(\mu \nu )}(\kappa _1\lambda x,\kappa _2\lambda x)\left\}\right|_{x=\stackrel{~}{x}}.`$ (3.80)
Now, we may pick up all the contributions to the symmetric (traceless) tensor operator. Together with the twist–2 part, eq. (3.19), the twist–3 part, eq. (3.2.5), and the twist–4 part, eq. (3.73), we finally obtain the following complete decomposition of the symmetric light–cone tensor operator (compare eq. (3)):
$`G_{(\alpha \beta )}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`=G_{(\alpha \beta )}^{\mathrm{tw2}(\mathrm{i})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw3}(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iv})}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$
$`+G_{(\alpha \beta )}^{\mathrm{tw4},\mathrm{s}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw4},\mathrm{v}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$
$`+G_{(\alpha \beta )}^{\mathrm{tw5},\mathrm{v}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{(\alpha \beta )}^{\mathrm{tw6},\mathrm{s}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x}).`$ (3.81)
Let us remark that by construction the terms of the first line are traceless, whereas the traces of the third line vanish because of antisymmetry and $`\stackrel{~}{x}^2=0`$ on the light–cone, respectively. However, the traces of the second line do not vanish but restore the trace of the tensor operator,
$`g^{\alpha \beta }G_{\alpha \beta }(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=g^{\alpha \beta }G_{(\alpha \beta )}^{\mathrm{tw4},\mathrm{s}}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+g^{\alpha \beta }G_{(\alpha \beta )}^{\mathrm{tw4},\mathrm{v}}((\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})).`$ (3.82)
This may be proven explicitly by taking the trace of eqs. (3.77) and (3.4.3), using $`(x)\lambda _\lambda `$, performing partial integrations and observing the equality
$`\left(g^{\alpha \beta }2^\alpha x^\beta +(1/2)\mathrm{}x^\alpha x^\beta \right)G_{(\alpha \beta )}(\kappa _1\lambda x,\kappa _2\lambda x)|_{\lambda =0}=0.`$ (3.83)
Herewith, the twist decomposition of a generic bilocal 2nd rank light–ray tensor operator is completed. In the next Chapter we show how this general formalism may be applied also to tri– and multi–local light–ray operators.
## 4 Extension to trilocal tensor operators
The procedure reviewed in Chapt. 2 for bilocal operators may be extended to trilocal operators, too. These operators occur in higher twist contributions to light–cone dominated processes and, additionally, as counterparts in the renormalization of such higher twist operators which have been obtained in Chapt. 3. In the following we consider various trilocal operators of minimal twist–3 and twist–4. The local versions of such higher twist operators have been considered already in the early days of QCD in systematic studies of deep inelastic scattering and on behalf of giving some parton interpretation for the distribution amplitudes . Various studies of local twist–4 operators determined their anomalous dimension matrices and the behaviour of their structure functions . The relevance of local twist–3 operators for the structure function $`g_2`$ of polarized deep inelastic scattering is also well-known . If written nonlocal all these operators are necessarily multilocal. However, nonlocal higher twist operators have been extensively used only in the study of the structure function $`g_2`$ and of the photon and vector meson wave functions .
Here, our main interest is not to give an exhaustive study of the various scattering processes but to present the twist decomposition of some characteristic multilocal operators.
### 4.1 General trilocal tensor operators: Quark–antiquark–gluon, four–quark and four–gluon operators
This Subsection is devoted to the consideration of trilocal operators which share the same twist decomposition as the gluon operators $`G_{\alpha \beta }`$ (or their (anti)symmetric parts). We call them unconstrained because their truncation with $`x^\alpha x^\beta `$ does not vanish identically. For compact notations, we will introduce the following abbreviations:
$`\mathrm{\Gamma }^i=\{1,\gamma _5\},\mathrm{\Gamma }_\alpha ^i=\{\gamma _\alpha ,\gamma _5\gamma _\alpha \},\mathrm{\Gamma }_{\alpha \beta }^i=\{\sigma _{\alpha \beta },\gamma _5\sigma _{\alpha \beta }\},F_{\mu \nu }^i=\{F_{\mu \nu },\stackrel{~}{F}_{\mu \nu }\}.`$
#### 4.1.1 Quark–antiquark–gluon operators
First, we consider the following quark–antiquark–gluon operators,<sup>9</sup><sup>9</sup>9 For simplicity we restrict to the nonsinglet case and suppress flavour matrices.
$`V_{\alpha \beta }^{ij}(\kappa _1x,\tau x,\kappa _2x)`$ $`=\overline{\psi }(\kappa _1x)U(\kappa _1x,\tau x)\mathrm{\Gamma }_\alpha ^{i\rho }F_{\beta \rho }^j(\tau x)U(\tau x,\kappa _2x)\psi (\kappa _2x),`$ (4.1)
and
$`W_{[\alpha \beta ]}^{ij}(\kappa _1x,\tau x,\kappa _2x)`$ $`=\overline{\psi }(\kappa _1x)U(\kappa _1x,\tau x)\mathrm{\Gamma }^iF_{\alpha \beta }^j(\tau x)U(\tau x,\kappa _2x)\psi (\kappa _2x),`$ (4.2)
whose minimal twist is $`\tau _{\mathrm{min}}=3`$ and $`4`$, respectively; they have the same symmetry as the gluon tensor operators $`G_{\alpha \beta }`$ and $`G_{[\alpha \beta ]}`$, respectively.
In order to be able to apply the general procedure of Chapt. 2 as well as the results of Chapt. 3 we have to verify the local structure of these trilocal operators. Let us study, for example, the first of the above quark–gluon operators, $`V_{\alpha \beta }^{11}`$, in some detail and Taylor expand its three local fields around $`y=0`$:
$`V_{\alpha \beta }^{11}(\kappa _1x,\tau x,\kappa _2x)=`$
$`=\overline{\psi }(\kappa _1x)U(\kappa _1x,0)U(0,\tau x)\sigma _\alpha ^\rho F_{\beta \rho }(\tau x)U(\tau x,0)U(0,\kappa _2x)\psi (\kappa _2x)`$
$`={\displaystyle \underset{n_1,m,n_2=0}{\overset{\mathrm{}}{}}}\frac{\kappa _1^{n_1}\tau ^m\kappa _2^{n_2}}{n_1!m!n_2!}\left[\overline{\psi }(y)\left(\stackrel{}{D}_yx\right)^{n_1}\right]\sigma _\alpha ^\rho [\left(xD_z\right)^mF_{\beta \rho }(z)][\left(x\stackrel{}{D}_y\right)^{n_2}\psi (y)]|_{y=z=0}`$
$`={\displaystyle \underset{N=0}{\overset{\mathrm{}}{}}}\frac{1}{N!}\overline{\psi }(0)({\displaystyle \underset{n=0}{\overset{N}{}}}{\displaystyle \underset{\mathrm{}=0}{\overset{n}{}}}\left(\genfrac{}{}{0pt}{}{N}{n}\right)\left(\genfrac{}{}{0pt}{}{n}{\mathrm{}}\right)(\kappa _1\stackrel{}{D}x)^n\mathrm{}\sigma _\alpha ^\rho [\left(\tau xD\right)^{Nn}F_{\beta \rho }(0)](\kappa _2x\stackrel{}{D})^{\mathrm{}})\psi (0),`$
where the left and right derivatives are given by eqs. (2.7), and
$$D_\mu F_{\alpha \beta }=_\mu F_{\alpha \beta }+\mathrm{i}g[A_\mu ,F_{\alpha \beta }].$$
(4.3)
This leads to the following expansion into local tensor operators
$$V_{\alpha \beta }^{11}(\kappa _1x,\tau x,\kappa _2x)=\underset{N=0}{\overset{\mathrm{}}{}}\frac{1}{N!}x^{\mu _1}\mathrm{}x^{\mu _N}V_{\alpha \beta \mu _1\mathrm{}\mu _N}^{11}(\kappa _1,\tau ,\kappa _2),$$
(4.4)
where
$$V_{\alpha \beta \mu _1\mathrm{}\mu _N}^{11}(\kappa _1,\tau ,\kappa _2)\overline{\psi }(0)\sigma _\alpha ^\rho 𝐃_{[\beta \rho ]\mu _1\mathrm{}\mu _N}(\kappa _1,\tau ,\kappa _2)\psi (0),$$
(4.5)
and the field strength–dependent generalized covariant derivative of $`N`$th order is given by
$`𝐃_{[\beta \rho ]\mu _1\mathrm{}\mu _N}(\kappa _1,\tau ,\kappa _2)=`$ (4.6)
$`=N!{\displaystyle \underset{n=0}{\overset{N}{}}}{\displaystyle \underset{\mathrm{}=0}{\overset{n}{}}}\frac{\kappa _1^n\mathrm{}}{\left(n\mathrm{}\right)!}\stackrel{}{D}_{\mu _1}\mathrm{}\stackrel{}{D}_{\mu _n\mathrm{}}\frac{\tau ^{Nn}}{\left(Nn\right)!}\left[D_{\mu _{n+1}}\mathrm{}D_{\mu _N}F_{\beta \rho }(0)\right]\frac{\kappa _2^{\mathrm{}}}{\mathrm{}!}\stackrel{}{D}_{\mu _{n\mathrm{}+1}}\mathrm{}\stackrel{}{D}_{\mu _n}.`$
Obviously, if the field strength were not present this operation would reduce to the product $`\stackrel{}{D}_{\mu _1}(\kappa _1,\kappa _2)\mathrm{}\stackrel{}{D}_{\mu _N}(\kappa _1,\kappa _2)`$ of generalized derivatives (2.7) introduced in Chapt. 2. Furthermore, for $`\tau =0`$ we are left with a more simple expression which, however, cannot be reduced to the $`N`$th power of some extended derivative because $`F`$ is equipped with some matrix structure. Anyway, the local tensor operators (4.5) of rank $`N+2`$ with canonical dimension $`d=N+5`$, which are given as a sum of $`\frac{1}{2}(N+1)(N+2)`$ terms, have to be decomposed according to their geometric twist. In principle, this has to be done term-by-term. But, because of the linearity of that procedure we are not required to do this for any term explicitly.
Due to the same general local structure of $`V_{\alpha \beta }^{ij}`$ and $`W_{\alpha \beta }^{ij}`$ – which is governed by Taylor expansions completely analogous to eq. (4.4) – their decomposition into terms of definite twist leads to the same expressions as we obtained in Chapt. 3 but with $`G_{\alpha \beta }`$ and $`G_{[\alpha \beta ]}`$ exchanged by $`V_{\alpha \beta }^{ij}`$ and $`W_{\alpha \beta }^{ij}`$, respectively. The only difference consists in a shift of any twist by one unit, $`\tau \tau +1`$, in the various inputs of Table 1 and 2 below.
In contrast to the operators $`V`$ and $`W`$ the quark–antiquark–gluon operators
$`𝒪_{\alpha \beta }^{ij}(\kappa _1x,\tau x,\kappa _2x)`$ $`=x^\rho \overline{\psi }(\kappa _1x)U(\kappa _1x,\tau x)\mathrm{\Gamma }_\alpha ^iF_{\beta \rho }^j(\tau x)U(\tau x,\kappa _2x)\psi (\kappa _2x)`$ (4.7)
and
$`𝒪_{}^{}{}_{\alpha \beta }{}^{ij}(\kappa _1x,\tau x,\kappa _2x)`$ $`=x^\rho x^\sigma \overline{\psi }(\kappa _1x)U(\kappa _1x,\tau x)\mathrm{\Gamma }_{\alpha \rho }^iF_{\beta \sigma }^j(\tau x)U(\tau x,\kappa _2x)\psi (\kappa _2x)`$ (4.8)
are special, namely, despite of their arbitrary symmetry with respect to $`\alpha `$ and $`\beta `$ they vanish identically if multiplied by $`x^\beta `$ and $`x^\alpha `$ or $`x^\beta `$, respectively. Therefore, they show some peculiarities which will be treated in Subsection 4.2.
#### 4.1.2 Four–quark and four–gluon operators
Let us now consider trilocal operators which are built up from four quark or four gluon fields. We denote them by $`{}_{}{}^{I}𝒬_{\alpha \beta }^{ij}`$ with $`I=0`$ and $`I=2`$, respectively (here, $`I`$ counts the number of external $`x`$’s of the operators):
$`{}_{}{}^{I=0}𝒬_{\alpha \beta }^{ij}(\kappa _1x,\tau x,\kappa _2x)=`$ (4.9)
$`=\left(\overline{\psi }(\kappa _1x)\mathrm{\Gamma }_\alpha ^iU(\kappa _1x,\tau x)\psi (\tau x)\right)\left(\overline{\psi }(\tau x)U(\tau x,\kappa _2x)\mathrm{\Gamma }_\beta ^j\psi (\kappa _2x)\right)`$
as well as
$`{}_{}{}^{I=2}𝒬_{\alpha \beta }^{ij}(\kappa _1x,\tau x,\kappa _2x)=x^\rho x^\sigma {}_{}{}^{I=2}𝒬_{\alpha \rho \sigma \beta }^{ij}(\kappa _1x,\tau x,\kappa _2x),`$ (4.10)
$`=x^\rho x^\sigma \left(F_{\alpha \mu }(\kappa _1x)U(\kappa _1x,\tau x)F_\rho ^{i\mu }(\tau x)\right)\left(F_\sigma ^{j\nu }(\tau x)U(\tau x,\kappa _2x)F_{\beta \nu }(\kappa _2x)\right).`$
Both sets of operators have minimal twist $`\tau _{\mathrm{min}}=4`$. Their local versions have been already considered by various authors, see e.g. . The explicit form of the local operators will be given below. Because there are no restrictions concerning the free indices $`\alpha `$ and $`\beta `$ the Young patterns (i) – (iv) are involved. The twist decomposition may be performed along the lines of Chap. 3 and, as in the foregoing cases, the outcome will be almost the same as for $`G_{\alpha \beta }`$. Again, the difference is that now the value of twist raises by two units, $`\tau \tau +2`$, relative to the gluon operators shown in Table 1 and 2 of the Conclusions. In addition, because of the change in the external $`x`$–factors (not being accompanied by $`\lambda `$) the measures of the $`\lambda `$–integrations have to be changed according to (see also Subsection 4.2)
$$\mathrm{d}\lambda \mathrm{d}\lambda \lambda ^I.$$
(4.11)
Open, up to now, is the structure of the local operators. Let us study in detail the first of the four–quark operators, $`{}_{}{}^{I=0}𝒬_{\alpha \beta }^{11}`$, whose Taylor expansion may be written in two equivalent ways:
$`{}_{}{}^{I=0}𝒬_{\alpha \beta }^{11}(\kappa _1x,\tau x,\kappa _2x)=`$
$`=\left(\overline{\psi }(\kappa _1x)\gamma _\alpha U(\kappa _1x,0)U(0,\tau x)\psi (\tau x)\right)\left(\overline{\psi }(\tau x)U(\tau x,0)U(0,\kappa _2x)\gamma _\beta \psi (\kappa _2x)\right)`$
$`={\displaystyle \underset{n_1,m,n_2=0}{\overset{\mathrm{}}{}}}\frac{\kappa _1^{n_1}\tau ^m\kappa _2^{n_2}}{n_1!m!n_2!}[\overline{\psi }(y)\left(\stackrel{}{D}_yx\right)^{n_1}]\gamma _\alpha [\left(xD_z\right)^m\left(\psi (z)\overline{\psi }(z)\right)]\gamma _\beta [\left(x\stackrel{}{D}_y\right)^{n_2}\psi (y)]|_{y=z=0}`$
$`={\displaystyle \underset{n_1=0}{\overset{\mathrm{}}{}}}\frac{1}{n_1!}\left(\overline{\psi }(y)\left(x\stackrel{}{D}_y(\kappa _1,\tau )\right)^{n_1}\gamma _\alpha \psi (y)\right)|_{y=0}{\displaystyle \underset{n_2=0}{\overset{\mathrm{}}{}}}\frac{1}{n_2!}\left(\overline{\psi }(z)\left(x\stackrel{}{D}_z(\tau ,\kappa _2)\right)^{n_2}\gamma _\beta \psi (z)\right)|_{z=0},`$
where the two possibilities also use different generalized covariant derivatives. The left, right and left-right derivatives are given by eqs. (2.7), and, treating $`\psi (0)\overline{\psi }(0)`$ as a matrix in the group algebra, another form of the derivative obtains:
$`D_\mu \left(\psi (0)\overline{\psi }(0)\right)`$ $`=_\mu \left(\psi (0)\overline{\psi }(0)\right)+\mathrm{i}g[A_\mu ,\left(\psi (0)\overline{\psi }(0)\right)]`$
$`=\left(\stackrel{}{D}_\mu \psi (0)\right)\overline{\psi }(0)+\psi (0)\left(\overline{\psi }(0)\stackrel{}{D}_\mu \right).`$
This leads to the following expansion into local tensor operators
$${}_{}{}^{I=0}𝒬_{\alpha \beta }^{ij}(\kappa _1x,\tau x,\kappa _2x)=\underset{N=0}{\overset{\mathrm{}}{}}\frac{1}{N!}x^{\mu _1}\mathrm{}x^{\mu _N}{}_{}{}^{I=0}𝒬_{\alpha \beta \mu _1\mathrm{}\mu _N}^{ij}(\kappa _1,\tau ,\kappa _2),$$
(4.12)
where
$${}_{}{}^{I=0}𝒬_{\alpha \beta \mu _1\mathrm{}\mu _N}^{ij}(\kappa _1,\tau ,\kappa _2)\overline{\psi }(0)\gamma _\alpha 𝐃_{\mu _1\mathrm{}\mu _N}(\kappa _1,\tau ,\kappa _2)\gamma _\beta \psi (0),$$
(4.13)
and the quark field–dependent generalized covariant derivative of $`N`$th order is given by
$`𝐃_{\mu _1\mathrm{}\mu _N}(\kappa _1,\tau ,\kappa _2)={\displaystyle \underset{n=0}{\overset{N}{}}}\left(\genfrac{}{}{0pt}{}{N}{n}\right)\stackrel{}{D}_{\mu _1}\mathrm{}\stackrel{}{D}_{\mu _n}\psi (0)\times \overline{\psi }(0)\stackrel{}{D}_{\mu _{n+1}}\mathrm{}\stackrel{}{D}_{\mu _N}`$ (4.14)
$`={\displaystyle \underset{n=0}{\overset{N}{}}}N!{\displaystyle \underset{\mathrm{}=0}{\overset{n}{}}}\stackrel{}{D}_{\mu _1}\mathrm{}\stackrel{}{D}_{\mu _n\mathrm{}}\frac{\kappa _1^n\mathrm{}\tau ^{\mathrm{}}}{\left(n\mathrm{}\right)!\mathrm{}!}\stackrel{}{D}_{\mu _{n\mathrm{}+1}}\mathrm{}\stackrel{}{D}_{\mu _n}\psi (0)`$
$`\times {\displaystyle \underset{k}{\overset{Nn}{}}}\overline{\psi }(0)\stackrel{}{D}_{\mu _{n+1}}\mathrm{}\stackrel{}{D}_{\mu _{n+k}}\frac{\tau ^k\kappa _2^{Nnk}}{k![Nnk)!}\stackrel{}{D}_{\mu _{n+k+1}}\mathrm{}\stackrel{}{D}_{\mu _N}.`$
Analogous generalized covariant derivatives occur for the four–gluon operators. The only difference will be that instead of the quark fields corresponding gluon field strengths appear in (4.14), and that the derivatives are to be taken in the adjoint representation. Despite having a complicated structure this does not matter in the twist decomposition of the trilinear operators $`{}_{}{}^{I}𝒬_{\alpha \beta }^{ij}(\kappa _1x,\tau x,\kappa _2x)`$.
Let us finish this Subsection with a short remark concerning so-called four-particle operators which also have been considered in the literature for local operators. These operators have the general structure, cf.
$$(\overline{\psi }U\psi )(\overline{\psi }U\psi ),(\overline{\psi }UFUFU\psi )\mathrm{or}(FUFUFUF),$$
with any field being located at a different point $`\kappa _ix,i=1,\mathrm{},4`$. The expansion into local tensors is obtained in the same way as above, leading to a general expression of the following form:
$`{}_{}{}^{I}_{\alpha \beta }^{}(\kappa _ix)={\displaystyle \underset{N=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{1}{N!}}x^{\mu _1}\mathrm{}x^{\mu _N}x^\rho \mathrm{}x^\sigma {}_{}{}^{I}_{\alpha \beta \rho \mathrm{}\sigma \mu _1\mathrm{}\mu _N}^{}(\kappa _i),`$ (4.15)
where the terms with equal value of $`N`$ are given as well-defined sums of local tensor operators. Any of these operators decompose in the same manner into irreducible representations of the Lorentz group. The latter result is immediately related to the fact that all indices $`\mu _1\mathrm{}\mu _N`$ are to be symmetrized, i.e., lying in the first row of any relevant Young tableau! Different symmetry classifications solely depend on the distribution of the remaining indices $`\alpha \beta \rho \mathrm{}\sigma `$ – the others being somehow truncated – to the various Young tableaux. As long as only up to two free indices $`\alpha \beta `$ are relevant the twist decomposition takes place according to the results of Chapt. 3, eventually modified by the power $`\lambda ^I`$ which is related to the number of external $`x`$’s in $`{}_{}{}^{I}`$. In addition, the twist of the various components may be shifted by some (equal) amount.
### 4.2 Constrained trilocal operators: Shuryak-Vainshtein– and three–gluon operators
Now we consider the special quark–antiquark–gluon operator $`𝒪_{\alpha \beta }^{ij}`$, eq. (4.7), and related three–gluon operators both having minimal twist–3; they will be denoted by $`{}_{}{}^{I}𝒪_{\alpha \beta }^{ij}(\kappa _1x,\tau x,\kappa _2x)`$ with $`I=0`$ and $`I=1`$, respectively. The quark–antiquark–gluon operators are given by
$`{}_{}{}^{I=0}𝒪_{\alpha \beta }^{ij}(\kappa _1x,\tau x,\kappa _2x)`$ $`=x^\rho {}_{}{}^{I=0}𝒪_{\alpha \beta \rho }^{ij}(\kappa _1x,\tau x,\kappa _2x)`$ (4.16)
$`=x^\rho \overline{\psi }(\kappa _1x)U(\kappa _1x,\tau x)\mathrm{\Gamma }_\alpha ^iF_{\beta \rho }^j(\tau x)U(\tau x,\kappa _2x)\psi (\kappa _2x),`$
where again some possible flavour structure has been suppressed. These operators are related to the following generalizations of the so-called Shuryak-Vainshtein operators ,
$`S_\beta ^\pm (\kappa _1x,\tau x,\kappa _2x)=x^\alpha \left({}_{}{}^{I=1}𝒪_{\alpha \beta }^{11}(\kappa _1x,\tau x,\kappa _2x)\pm \mathrm{i}{}_{}{}^{I=1}𝒪_{\alpha \beta }^{22}(\kappa _1x,\tau x,\kappa _2x)\right)x^\alpha ,`$
which, in the flavour singlet case, mix with the following three–gluon operators:
$`{}_{}{}^{I=1}𝒪_{\alpha \beta }^{i}(\kappa _1x,\tau x,\kappa _2x)=x^\rho x^\sigma {}_{}{}^{I=1}𝒪_{\alpha \beta \rho \sigma }^{i}(\kappa _1x,\tau x,\kappa _2x)`$ (4.17)
$`=x^\rho x^\sigma F_{\alpha \nu }^a(\kappa _1x)U^{ab}(\kappa _1x,\tau x)F_{\beta \rho }^{ibc}(\tau x)U^{cd}(\tau x,\kappa _2x)F_\sigma ^{d\nu }(\kappa _2x),`$
where $`F_{\alpha \beta }^{ab}(x)f^{acb}F_{\alpha \beta }^c(x)`$ and the phase factors are taken in the adjoint representation.
Because of their construction these special trilocal operators have the property
$`x^\beta {}_{}{}^{I}𝒪_{\alpha \beta }^{ij}(\kappa _1x,\tau x,\kappa _2x)0.`$ (4.18)
They contain twist–3 up to twist–6 contributions which, because of the antisymmetry of the gluon field strength, have to be determined by means of the Young tableaux (ii) – (iv). Because of (4.18) the scalar operators $`x^\alpha x^\beta {}_{}{}^{I}𝒪_{\alpha \beta }^{ij}(\kappa _1x,\tau x,\kappa _2x)`$ vanish identically.
The Taylor expansion of the operator $`{}_{}{}^{I=0}𝒪_{\alpha \beta }^{11}(\kappa _1x,\tau x,\kappa _2x)`$ reads
$${}_{}{}^{I=0}𝒪_{\alpha \beta }^{11}(\kappa _1x,\tau x,\kappa _2x)=\underset{N=0}{\overset{\mathrm{}}{}}\frac{1}{N!}x^{\mu _1}\mathrm{}x^{\mu _N}x^\rho {}_{}{}^{I=0}𝒪_{\alpha \beta \rho \mu _1\mathrm{}\mu _N}^{11}(\kappa _1,\tau ,\kappa _2),$$
(4.19)
with
$${}_{}{}^{I=0}𝒪_{\alpha \beta \rho \mu _1\mathrm{}\mu _N}^{11}(\kappa _1,\tau ,\kappa _2)\overline{\psi }(0)\gamma _\alpha 𝐃_{[\beta \rho ]\mu _1\mathrm{}\mu _N}(\kappa _1,\tau ,\kappa _2)\psi (0),$$
(4.20)
where the field strength–dependent derivative already has been given by (4.6). The Taylor expansion of the related three–gluon operator obtains as follows:
$`{}_{}{}^{I=1}𝒪_{\alpha \beta }^{11}(\kappa _1x,\tau x,\kappa _2x)=`$ $`{\displaystyle \underset{N=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{1}{N!}}x^{\mu _1}\mathrm{}x^{\mu _N}x^\rho x^\sigma {}_{}{}^{I=1}𝒪_{\alpha \beta \rho \sigma \mu _1\mathrm{}\mu _N}^{A}(\kappa _1,\tau ,\kappa _2)`$ (4.21)
where
$`{}_{}{}^{I=1}𝒪_{\alpha \beta \rho \sigma \mu _1\mathrm{}\mu _N}^{11}(\kappa _1,\tau ,\kappa _2)`$ $`F_{\alpha \nu }^a(0)𝐃_{[\beta \rho ]\mu _1\mathrm{}\mu _N}^{ab}(\kappa _1,\tau ,\kappa _2)F_\sigma ^{b\nu }(0),`$ (4.22)
with the same field strength dependent generalized covariant derivative (4.6), but now in the adjoint representation. The generalization arbitrary values $`i`$ and $`j`$ is obvious.
As it became obvious by the above considerations all the nonlocal three–particle operators $`{}_{}{}^{I}𝒪_{\alpha \beta }^{ij}`$ have the same general local structure and, therefore, decompose according to the same symmetry patterns. They only differ in the rank of the local tensor operators which is due to the external powers of $`x`$. Again, this must be taken into account by a change of the integration measure according to<sup>10</sup><sup>10</sup>10 Observe, that because of notational simplicity in the explicit expressions below we shifted the variable $`II1`$.
$$\mathrm{d}\lambda \mathrm{d}\lambda \lambda ^{I+1}.$$
From now on we omit the indices $`i`$ and $`j`$.
In order to determine the twist–3 light–cone operators $`{}_{}{}^{I}𝒪_{\alpha \beta }^{}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ for $`I=0`$ and $`I=1`$, let us consider the following Young tableaux of symmetry pattern (iiB) and , respectively:
$`{}_{}{}^{I}𝒪_{\alpha \beta }^{\mathrm{tw3}}(\kappa _1x,\tau x,\kappa _2x)`$ $`=\frac{1}{2}{\displaystyle _0^1}d\lambda \lambda ^I\left[\left(1+\lambda ^2\right)\delta _\beta ^\mu _\alpha +\left(1\lambda ^2\right)\delta _\alpha ^\mu _\beta \right]{}_{}{}^{I}\stackrel{}{𝒪}_{\mu }^{}(\kappa _1\lambda x,\tau \lambda x,\kappa _2\lambda x)`$
$`={}_{}{}^{I}𝒪_{[\alpha \beta ]}^{\mathrm{tw3}}(\kappa _1x,\tau x,\kappa _2x)+{}_{}{}^{I}𝒪_{(\alpha \beta )}^{\mathrm{tw3}}(\kappa _1x,\tau x,\kappa _2x);`$ (4.23)
here, and in the following, we write $`𝒪_\mu x^\rho 𝒪_{\rho \mu }`$. Let us remark that, first, because of property (4.18), only the above mentioned Young tableaux contribute and, second, the difference between the expressions (4.2) and (3.42) result also from that property, namely the truncation by $`x^\rho `$ in the second term of the integrand, $`x^\mu _\alpha _\beta {}_{}{}^{I}\stackrel{}{𝒪}_{\mu }^{}(\kappa _1x,\tau x,\kappa _2x)=2\delta _{(\alpha }^\mu _{\beta )}{}_{}{}^{I}\stackrel{}{𝒪}_{\mu }^{}(\kappa _1x,\tau x,\kappa _2x)`$, after partial integrations leads to the above expression (4.2). In the case $`I=0`$ this operator is in agreement with the expression given by Balitsky and Braun (cf. eq. (5.14) of ). However, their operator lacks to be really of twist–3 since it is not traceless; therefore it contains also twist–4, twist–5 as well as twist–6 operators resulting from the trace terms.
Now, let us to project onto the light–cone. As in the case of the gluon tensor only the first terms, $`k=0,1,2`$, in the expansion of $`{}_{}{}^{I}\stackrel{}{𝒪}_{\mu }^{}(\kappa _1\lambda x,\tau \lambda x,\kappa _2\lambda x)`$ are to be taken into account. For the antisymmetric part of the twist–3 operator one gets
$${}_{}{}^{I}𝒪_{[\alpha \beta ]}^{\mathrm{tw3}}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})=_0^1d\lambda \lambda ^{2+I}_{[\alpha }{}_{}{}^{I}𝒪_{\beta ]}^{}(\kappa _1\lambda \stackrel{~}{x},\tau \lambda \stackrel{~}{x},\kappa _2\lambda \stackrel{~}{x}){}_{}{}^{I}𝒪_{[\alpha \beta ]}^{>(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})$$
(4.24)
where
$`{}_{}{}^{I}𝒪_{[\alpha \beta ]}^{>(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})=\frac{1}{2}`$ $`{\displaystyle _0^1}d\lambda \lambda ^{1+I}(1\lambda )\left(2x_{[\alpha }_{\beta ]}^\mu x_{[\alpha }\delta _{\beta ]}^\mu \mathrm{}\right)`$
$`\times {}_{}{}^{I}𝒪_{\mu }^{}(\kappa _1\lambda x,\tau \lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}.`$ (4.25)
The decomposition of $`{}_{}{}^{I}𝒪_{[\alpha \beta ]}^{>(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ arises as
$`{}_{}{}^{I}𝒪_{[\alpha \beta ]}^{>(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})={}_{}{}^{I}𝒪_{[\alpha \beta ]}^{\mathrm{tw4}(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})+{}_{}{}^{I}𝒪_{[\alpha \beta ]}^{\mathrm{tw5}(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$ (4.26)
with
$`{}_{}{}^{I}𝒪_{[\alpha \beta ]}^{\mathrm{tw4}(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`=`$ $`\frac{1}{2}x_{[\alpha }_{\beta ]}^\mu {\displaystyle _0^1}d\lambda \lambda ^I(1\lambda ^2){}_{}{}^{I}𝒪_{\mu }^{}(\kappa _1\lambda x,\tau \lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$
$`{}_{}{}^{I}𝒪_{[\alpha \beta ]}^{\mathrm{tw5}(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`=`$ $`\frac{1}{4}x_{[\alpha }\left(\delta _{\beta ]}^\mu (x)x^\mu _{\beta ]}\right)\mathrm{}`$
$`\times {\displaystyle _0^1}\mathrm{d}\lambda \lambda ^I(1\lambda )^2{}_{}{}^{I}𝒪_{\mu }^{}(\kappa _1\lambda x,\tau \lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}.`$
The symmetric part reads
$${}_{}{}^{I}𝒪_{(\alpha \beta )}^{\mathrm{tw3}}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})=_0^1d\lambda \lambda ^I_{(\alpha }{}_{}{}^{I}𝒪_{\beta )}^{}(\kappa _1\lambda \stackrel{~}{x},\tau \lambda \stackrel{~}{x},\kappa _2\lambda \stackrel{~}{x}){}_{}{}^{I}𝒪_{(\alpha \beta )}^{>(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})$$
(4.27)
where
$`{}_{}{}^{I}𝒪_{(\alpha \beta )}^{>(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})={\displaystyle _0^1}\mathrm{d}\lambda \lambda ^I\{\frac{1}{2}(1\lambda ^2)g_{\alpha \beta }^\mu +\frac{1}{2}(1\lambda )\delta _{(\alpha }^\mu x_{\beta )}\mathrm{}`$
$`+\lambda (1\lambda )x_{(\alpha }_{\beta )}^\mu \frac{1}{4}(1\lambda )^2x_\alpha x_\beta ^\mu \mathrm{}\}^I𝒪_\mu (\kappa _1\lambda x,\tau \lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}.`$
Let us remark that the conditions of tracelessness for the light-cone operators are:
$`g^{\alpha \beta }{}_{}{}^{I}𝒪_{(\alpha \beta )}^{\mathrm{tw3}}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})=0,\mathrm{d}^\alpha {}_{}{}^{I}𝒪_{(\alpha \beta )}^{\mathrm{tw3}}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`=`$ $`0,`$
$`\mathrm{d}^\alpha {}_{}{}^{I}𝒪_{[\alpha \beta ]}^{\mathrm{tw3}}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`=`$ $`0.`$
Now we can calculate the twist–3 vector operator from the twist–3 tensor operator. The twist–3 light–cone vector operator reads
$`{}_{}{}^{I}𝒪_{\alpha }^{\mathrm{tw3}}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})={\displaystyle _0^1}d\lambda \lambda ^{2+I}\left[\delta _\alpha ^\mu (x)x^\mu _\alpha x_\alpha ^\mu \right]{}_{}{}^{I}𝒪_{\mu }^{}(\kappa _1\lambda x,\tau \lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}`$
$`={}_{}{}^{I}𝒪_{\alpha }^{}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})\stackrel{~}{x}_\alpha {\displaystyle _0^1}d\lambda \lambda ^{2+I}^\mu {}_{}{}^{I}𝒪_{\mu }^{}(\kappa _1\lambda x,\tau \lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}},`$ (4.29)
and the twist–4 vector operator which is contained in the trace terms of the twist–3 vector operator
$${}_{}{}^{I}𝒪_{\alpha }^{\mathrm{tw4}(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})=\stackrel{~}{x}_\alpha _0^1d\lambda \lambda ^{2+I}^\mu {}_{}{}^{I}𝒪_{\mu }^{}(\kappa _1\lambda x,\tau \lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}.$$
(4.30)
The determination of the antisymmetric twist–4 operator having symmetry type (iii) obtains from Young tableaux and , respectively. Again, making use of property (4.18), the result is
$`{}_{}{}^{I}𝒪_{[\alpha \beta ]}^{\mathrm{tw4}(\mathrm{iii})}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ $`{\displaystyle _0^1}d\lambda \lambda ^{2+I}\delta _{[\alpha }^\mu ((x)\delta _{\beta ]}^\nu 2x^\nu _{\beta ]}){}_{}{}^{I}𝒪_{[\mu \nu ]}^{}(\kappa _1\lambda x,\tau \lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}`$
$`{}_{}{}^{I}𝒪_{[\alpha \beta ]}^{\mathrm{tw5}(\mathrm{iii})}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$ (4.31)
twist–5 part is determined by the trace namely with
$`{}_{}{}^{I}𝒪_{[\alpha \beta ]}^{\mathrm{tw5}(\mathrm{iii})}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa \stackrel{~}{x})=`$ $`{\displaystyle _0^1}\mathrm{d}\lambda \lambda ^I(1\lambda ^2)\{x_{[\alpha }(\delta _{\beta ]}^{[\mu }(x)x^{[\mu }_{\beta ]})^{\nu ]}`$
$`x_{[\alpha }\delta _{\beta ]}^{[\mu }x^{\nu ]}\mathrm{}\}^I𝒪_{[\mu \nu ]}(\kappa _1\lambda x,\tau \lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}.`$ (4.32)
The antisymmetric twist–5 tensor operator is given by:
$`{}_{}{}^{I}𝒪_{[\alpha \beta ]}^{\mathrm{tw5}}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa \stackrel{~}{x})={}_{}{}^{I}𝒪_{[\alpha \beta ]}^{\mathrm{tw5}(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa \stackrel{~}{x})+{}_{}{}^{I}𝒪_{[\alpha \beta ]}^{\mathrm{tw5}(\mathrm{iii})}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa \stackrel{~}{x})`$
$`=\frac{1}{4}x_{[\alpha }\left(\delta _{\beta ]}^\mu (x)x^\mu _{\beta ]}\right)\mathrm{}{\displaystyle _0^1}d\lambda \lambda ^I(1\lambda )^2{}_{}{}^{I}𝒪_{\mu }^{}(\kappa _1\lambda x,\tau \lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}`$
$`{\displaystyle _0^1}\mathrm{d}\lambda \lambda ^I(1\lambda ^2)\{x_{[\alpha }(\delta _{\beta ]}^{[\mu }(x)x^{[\mu }_{\beta ]})^{\nu ]}`$
$`x_{[\alpha }\delta _{\beta ]}^{[\mu }x^{\nu ]}\mathrm{}\}^I𝒪_{[\mu \nu ]}(\kappa _1\lambda x,\tau \lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}.`$ (4.33)
Finally, we obtain the complete decomposition of the antisymmetric tensor operator:
$`{}_{}{}^{I}𝒪_{[\alpha \beta ]}^{}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`={}_{}{}^{I}𝒪_{[\alpha \beta ]}^{\mathrm{tw3}(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})+{}_{}{}^{I}𝒪_{[\alpha \beta ]}^{\mathrm{tw4}(\mathrm{iii})}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})`$
$`+{}_{}{}^{I}𝒪_{[\alpha \beta ]}^{\mathrm{tw4}(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})+{}_{}{}^{I}𝒪_{[\alpha \beta ]}^{\mathrm{tw5}}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x}).`$ (4.34)
The determination of the symmetric twist–4 operator having symmetry type (iv) obtains from Young tableaux and , respectively. Making use of property (4.18) the result is
$`{}_{}{}^{I}𝒪_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iv})}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ (4.35)
$`{\displaystyle _0^1}d\lambda \lambda ^I(1\lambda )\left(\delta _\alpha ^\mu \delta _\beta ^\nu x^\rho (x)_\rho 2x^\mu (x)\delta _{(\beta }^\nu _{\alpha )}\right){}_{}{}^{I}𝒪_{(\mu \nu )}^{}(\kappa _1\lambda x,\tau \lambda x,\kappa _2\lambda x)|_{x=\stackrel{~}{x}}`$
$`{}_{}{}^{I}𝒪_{(\alpha \beta )}^{>(\mathrm{iv})}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x}),`$
where
$`{}_{}{}^{I}𝒪_{(\alpha \beta )}^{>(\mathrm{iv})}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})=`$ (4.36)
$`{\displaystyle _0^1}\mathrm{d}\lambda \{(2\lambda ^{2+I}x_{(\alpha }\delta _{\beta )}^\nu ^\mu \lambda ^{1+I}(1\lambda )x_\alpha x_\beta ^\mu ^\nu ){}_{}{}^{I}𝒪_{(\mu \nu )}^{}(\kappa _1\lambda x,\tau \lambda x,\kappa _2\lambda x)`$
$`\frac{1}{2}\left(\lambda ^Ig_{\alpha \beta }^\mu +\lambda ^I(1\lambda ^2)x_{(\alpha }\delta _{\beta )}^\mu \mathrm{}\frac{1}{2}\lambda ^I(1\lambda )^2x_\alpha x_\beta \mathrm{}^\mu \right){}_{}{}^{I}𝒪_{\mu }^{}(\kappa _1\lambda x,\tau \lambda x,\kappa _2\lambda x)`$
$`+(\frac{1}{2}\lambda ^Ig_{\alpha \beta }(x)\lambda ^{2+I}x_{(\alpha }_{\beta )}\frac{1}{2}\lambda ^{2+I}(\mathrm{ln}\lambda )x_\alpha x_\beta \mathrm{}){}_{}{}^{I}𝒪_{\rho }^{\rho }(\kappa _1\lambda x,\tau \lambda x,\kappa _2\lambda x)\left\}\right|_{x=\stackrel{~}{x}}.`$
Now, we use the property (4.18) and sum up the symmetric higher twist operators, which appear in the trace terms of the symmetric twist operators with Young symmetry (ii) and (iv), to a complete twist–4, twist–5 and twist–6 operator. The complete twist–4 operators are constructed by means of the Young tableau (i). The ‘scalar part’ of the twist–4 tensor operator is given by
$`{}_{}{}^{I}𝒪_{(\alpha \beta )}^{\mathrm{tw4},\mathrm{s}}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})=\frac{1}{2}g_{\alpha \beta }\{`$ $`(x){\displaystyle _0^1}d\lambda \lambda ^I{}_{}{}^{I}𝒪_{\rho }^{\rho }(\kappa _1\lambda x,\tau \lambda x,\kappa _2\lambda x)`$ (4.37)
$`^\mu {\displaystyle _0^1}\mathrm{d}\lambda \lambda ^{2+I}{}_{}{}^{I}𝒪_{\mu }^{}(\kappa _1\lambda x,\tau \lambda x,\kappa _2\lambda x)\left\}\right|_{x=\stackrel{~}{x}}`$
and the twist–4 ‘vector part’ reads
$`{}_{}{}^{I}𝒪_{(\alpha \beta )}^{\mathrm{tw4},\mathrm{v}}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})=\stackrel{~}{x}_{(\alpha }\{{\displaystyle _0^1}\mathrm{d}\lambda \lambda ^I\{_{\beta )}+\frac{1}{2}x_{\beta )}(\mathrm{ln}\lambda )\mathrm{}\}{}_{}{}^{I}𝒪_{\rho }^{\rho }(\kappa _1\lambda x,\tau \lambda x,\kappa _2\lambda x)`$
$`{\displaystyle _0^1}\mathrm{d}\lambda \lambda ^I\{(1\lambda ^2)_{\beta )}+\frac{1}{4}x_{\beta )}(1\lambda ^2+2\mathrm{ln}\lambda )\mathrm{}\}^\mu {}_{}{}^{I}𝒪_{\mu }^{}(\kappa _1\lambda x,\tau \lambda x,\kappa _2\lambda x)\left\}\right|_{x=\stackrel{~}{x}}.`$ (4.38)
Furthermore, the complete twist–5 vector operator is again constructed by means of the Young tableau (ii). The ‘vector part’ of the twist–5 tensor operator is given by
$`{}_{}{}^{I}𝒪_{(\alpha \beta )}^{\mathrm{tw5},\mathrm{v}}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})=\stackrel{~}{x}_{(\alpha }\times `$ (4.39)
$`\{\frac{1}{4}{\displaystyle _0^1}\mathrm{d}\lambda \lambda ^I(1\lambda )^2(\delta _{\beta )}^\nu (x)x^\nu _{\beta )}x_{\beta )}^\nu )\mathrm{}{}_{}{}^{I}𝒪_{\nu }^{}(\kappa _1\lambda x,\tau \lambda x,\kappa _2\lambda x)`$
$`{\displaystyle _0^1}d\lambda \lambda ^I\left\{\left(1\lambda ^2\right)\left(\delta _{\beta )}^\nu (x)x^\nu _{\beta )}x_{\beta )}^\nu \right)^\mu \left(\frac{1}{2}\left(1\lambda ^2\right)+\mathrm{ln}\lambda \right)x_{\beta )}\mathrm{}^\mu x^\nu \right\}`$
$`\times {}_{}{}^{I}𝒪_{(\mu \nu )}^{}(\kappa _1\lambda x,\tau \lambda x,\kappa _2\lambda x)\left\}\right|_{x=\stackrel{~}{x}}.`$
The full twist–6 scalar operator has Young symmetry (i). The ‘scalar part’ of the twist–6 tensor operator is given by
$`{}_{}{}^{I}𝒪_{(\alpha \beta )}^{\mathrm{tw6},\mathrm{s}}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})=\stackrel{~}{x}_\alpha \stackrel{~}{x}_\beta \{\frac{1}{2}\mathrm{}{\displaystyle _0^1}\mathrm{d}\lambda \lambda ^I(1\lambda ^2)\mathrm{ln}(\lambda ){}_{}{}^{I}𝒪_{\rho }^{\rho }(\kappa _1\lambda x,\tau \lambda x,\kappa _2\lambda x)`$
$`\frac{1}{4}\mathrm{}^\mu {\displaystyle _0^1}d\lambda \lambda ^I\left(1\lambda \right)^2{}_{}{}^{I}𝒪_{\mu }^{}(\kappa _1\lambda x,\tau \lambda x,\kappa _2\lambda x)`$
$`+^\mu ^\nu {\displaystyle _0^1}\mathrm{d}\lambda \lambda ^I(1\lambda ){}_{}{}^{I}𝒪_{(\mu \nu )}^{}(\kappa _1\lambda x,\tau \lambda x,\kappa _2\lambda x)\left\}\right|_{x=\stackrel{~}{x}}.`$ (4.40)
Thus, together with the twist–3 part, eq. (4.27), and the twist–4 part, eq. (4.35), we finally obtain the complete decomposition of the symmetric tensor operator:
$`{}_{}{}^{I}𝒪_{(\alpha \beta )}^{}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})={}_{}{}^{I}𝒪_{(\alpha \beta )}^{\mathrm{tw3}(\mathrm{ii})}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})+{}_{}{}^{I}𝒪_{(\alpha \beta )}^{\mathrm{tw4}(\mathrm{iv})}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ (4.41)
$`+{}_{}{}^{I}𝒪_{(\alpha \beta )}^{\mathrm{tw4},\mathrm{s}}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})+{}_{}{}^{I}𝒪_{(\alpha \beta )}^{\mathrm{tw4},\mathrm{v}}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})+{}_{}{}^{I}𝒪_{(\alpha \beta )}^{\mathrm{tw5},\mathrm{v}}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x})`$
$`+{}_{}{}^{I}𝒪_{(\alpha \beta )}^{\mathrm{tw6},\mathrm{s}}(\kappa _1\stackrel{~}{x},\tau \stackrel{~}{x},\kappa _2\stackrel{~}{x}).`$
## 5 Conclusion
Let us shortly summarize our results. Making use of the tensorial harmonic polynomials up to rank 2 we determined for the first time and in a systematic way the various contributions of definite twist for a generic bilocal 2nd rank tensor operator $`G_{\alpha \beta }(\kappa _1x,\kappa _2x)`$ as well as its symmetric and antisymmetric parts $`G_{(\alpha \beta )}(\kappa _1x,\kappa _2x)`$ and $`G_{[\alpha \beta ]}(\kappa _1x,\kappa _2x)`$, respectively. In addition, we determined the related vector and scalar operators which occur by truncating with $`x^\beta `$ and $`x^\alpha x^\beta `$ or $`g^{\alpha \beta }`$, respectively. All these operators are harmonic tensor functions. By projection onto the light–cone we obtained the decomposition of a generic bilocal light–ray tensor operator of 2nd rank – and its reductions to (anti)symmetric tensors as well as the related vector and scalar operators – into operators of definite twist. In order to make the main results quite obvious we summarize them in Table 1 and 2. There, we indicate the different twist contributions to $`G_{\alpha \beta }(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})=G_{(\alpha \beta )}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})+G_{[\alpha \beta ]}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ by their symmetry type (i) up to (iv) and the number of the corresponding equations of Chapt. 3; in the second half of the tables the contributions from the trace terms are shown.
These expressions also apply to generic trilocal light–ray operators. The only difference is that the respective values of twist increase by a definite amount which is equal for any given operator and possibly, depending on the number $`I`$ of external $`x`$–factors, by a change in the integration measure, $`\mathrm{d}\lambda \mathrm{d}\lambda \lambda ^I`$. The applicability of our procedure which has been described in Chapt. 2 for bilocal operators rests upon the fact that the general form of the Taylor expansion (around $`y=0`$) for any of the nonlocal operators is the same: It is given by an infinite sum of terms of $`N`$th order being multiplied by $`x^{\mu _1}x^{\mu _2}\mathrm{}x^{\mu _N}`$ where each term consists of a finite sum of local operators having exactly the same tensor structure. In principle, this works also for multi-local operators of higher order.
However, there are special trilocal operators which do not immediately fit into the general scheme of Chapt. 3. An example of such operators has been considered in Chapt. 4 on the same line of reasoning as for the generic case. Its twist decomposition is obtained by restricting the expressions of the generic case to the special conditions under consideration. The results are classified in Table 3 and 4, again by indicating their symmetry type and the number of the corresponding equation. In the same manner one should proceed for other kinds of special multilocal operators like that of eq. (4.8).
The present study made also obvious how difficult an exact treatment of higher twist contributions will be. There is a complicate interplay between the higher twists resulting from different symmetry types and the corresponding trace terms. The results also suffer from the equations of motion (EOM) which are frequently used in the applications because they do not contribute between physical states . The twist decomposition of these EOM operators
$`{}_{}{}^{\text{EOM}}𝒪_{\mathrm{\Gamma }}^{}(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ $`=\overline{\psi }(\kappa _1\stackrel{~}{x})\mathrm{\Gamma }U(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})[\mathrm{i}\stackrel{}{D}/m]\psi (\kappa _2\stackrel{~}{x})`$
$`\overline{\psi }(\kappa _1\stackrel{~}{x})[\mathrm{i}\stackrel{}{D}/m]\mathrm{\Gamma }U(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})\psi (\kappa _2\stackrel{~}{x}),\mathrm{\Gamma }=\{1,\gamma _\alpha ,\sigma _{\alpha \beta }\}`$
is analogous to the corresponding quark operators $`𝒪_\mathrm{\Gamma }(\kappa _1\stackrel{~}{x},\kappa _2\stackrel{~}{x})`$ which is given in . The only difference is that for the equation of motion operators the canonical dimension and the twist raises by one unit.
The various tensor operators which we considered here are nonlocal generalizations of local operators already considered earlier in the literature for twist–3, cf. , and twist–4, cf. . There, as far as possible an explicit twist decomposition has been circumvented by truncating any of the (constant) totally symmetrized tensor indices of the local operators by some light–like vector $`n^\mu `$, i.e., by reducing to the scalar case. To the best of our knowledge the only work where the twist decomposition of (scalar) bilocal operators has been considered is that of Balitsky and Braun . However, their work is based on the external field formalism and does not have an obvious group theoretical systematics.
In principle, our procedure may be extended also to tensors of arbitrary high rank. Despite being defined by an obvious algorithm, its application to the next step, the twist decomposition of a generic 3rd rank tensor operator, will be very cumbersome. First of all, it would be necessary to determine the projection operators onto all traceless 3rd rank tensorial harmonic polynomials and, secondly, also additional Young patterns had to be taken into account. Fortunately, such kind of nonlocal operators – at least in the near future – may not be of physical relevance. Therefore, any further study in that direction seems to be reasonable only from a group theoretical point of view.
However, another observation deserves mentioning. This is the appearance of the inner derivation on the cone and its relation to the conformal group. The inner derivative not only may be used in defining the property of tracelessness of the nonlocal light–ray operators. In some cases it is possible to construct the nontrivial tensor operators by applying (products of) the inner derivative on the scalar light–ray operators. Therefore, it seems to be of immediate value for a more direct determination of the complicated expressions obtained here and for a simplification of our procedure.
Acknowledgement
The authors are very much indebted to D. Robaschik and S. Neumeier for stimulating discussion. M.L. acknowledges financial support by Graduate College “Quantum field theory” at Center for Theoretical Sciences of Leipzig University.
## Appendix A Harmonic tensor polynomials
In this Appendix we derive the operators projecting onto the traceless part of
\- completely symmetric tensors of rank $`n`$, $`T_{(\mu _1\mathrm{}\mu _n)}`$,
\- tensors $`T_{\alpha (\mu _1\mathrm{}\mu _n)}`$ of rank $`n+1`$ being symmetric in $`n`$ of its indices,
\- tensors $`T_{[\alpha \beta ](\mu _1\mathrm{}\mu _n)}`$ of rank $`n+2`$ being symmetric in $`n`$ and antisymmetric in
\- two indices, and
\- tensors $`T_{(\alpha \beta )(\mu _1\mathrm{}\mu _n)}`$ of rank $`n+2`$ being symmetric in $`n`$ and, independently,
\- symmetric in two indices.
In order to construct vector and tensor polynomials corresponding to $`T_{\alpha (\mu _1\mathrm{}\mu _n)}`$, $`T_{[\alpha \beta ](\mu _1\mathrm{}\mu _n)}`$, and $`T_{(\alpha \beta )(\mu _1\mathrm{}\mu _n)}`$ we generalize the homogeneous polynomial technique which is well-known in constructing irreducible representations of the orthogonal groups and which has been used for two-point functions by Todorov et. al. and in the scalar case by Nachtmann . To achieve this we use the fact that, after contracting the indices of the symmetric part with some vector $`x`$, the resulting scalar, vector and (anti)symmetric tensor polynomials of order $`n`$ obey the following conditions of tracelessness:
$`\mathrm{}\stackrel{}{T}_n(x)`$ $`=`$ $`0,`$ (A.1)
$`\mathrm{}\stackrel{}{T}_{\alpha n}(x)`$ $`=`$ $`0,^\alpha \stackrel{}{T}_{\alpha n}(x)=0,`$ (A.2)
$`\mathrm{}\stackrel{}{T}_{[\alpha \beta ]n}(x)`$ $`=`$ $`0,^\alpha \stackrel{}{T}_{[\alpha \beta ]n}(x)=0,`$ (A.3)
$`\mathrm{}\stackrel{}{T}_{(\alpha \beta )n}(x)`$ $`=`$ $`0,^\alpha \stackrel{}{T}_{(\alpha \beta )n}(x)=0,g^{\alpha \beta }\stackrel{}{T}_{(\alpha \beta )n}(x)=0.`$ (A.4)
A solution of the first three sets of equations already have been given in Part I. Below, for the readers convenience we list them for arbitrary dimensions $`D=2h`$. In Chapt. 3 they will be used for $`D=4`$. In addition, we also solve the last set of equations; these symmetric harmonic tensor polynomials are wanted in the consideration of symmetry type (iv) of Chapter 3.4.
(i) The solutions of eq. (A.1) are the (scalar) harmonic polynomials of order $`n`$ corresponding to symmetric traceless tensors of rank $`n`$. They are given by (see e.g. , Chapter IX)
$`\stackrel{}{T}_n(x)`$ $`=`$ $`H_n^{(2h)}(x^2|\mathrm{})T_n(x)`$ (A.5)
with the harmonic projection operator
$`H_n^{(2h)}\left(x^2|\mathrm{}\right)`$ $`=`$ $`{\displaystyle \underset{k=0}{\overset{[\frac{n}{2}]}{}}}{\displaystyle \frac{(1)^k(n+hk2)!}{4^kk!(n+h2)!}}x^{2k}\mathrm{}^k.`$ (A.6)
(ii) The solutions of eqs. (A.2) are the harmonic vector polynomials $`T_{\alpha n}(x)`$ of order $`n`$ corresponding to the traceless tensors $`T_{\alpha (\mu _1\mathrm{}\mu _n)}`$, i.e., $`D`$ harmonic polynomials transforming as a vector under $`SO(D)`$. They are given by
$`\stackrel{}{T}_{\alpha n}(x)=`$ $`\left\{\delta _\alpha ^\beta \frac{1}{\left(h+n1\right)\left(2h+n3\right)}(h2+x)\left[x_\alpha ^\beta \frac{1}{2}x^2_\alpha ^\beta \right]\right\}`$
$`\times H_n^{(2h)}\left(x^2|\mathrm{}\right)T_{\beta n}(x).`$ (A.7)
After contraction with $`x^\alpha `$ the scalar harmonic polynomial $`\stackrel{}{T}_{n+1}(x)`$ obtains.
(iii) The solutions of eqs. (A.3) are the antisymmetric harmonic tensor polynomials $`\stackrel{}{T}_{[\alpha \beta ]n}(x)`$ of order $`n`$. They are given by
$`\stackrel{}{T}_{[\alpha \beta ]n}(x)=`$ $`\{\delta _{[\alpha }^\mu \delta _{\beta ]}^\nu +\frac{2}{\left(h+n1\right)\left(2h+n3\right)}(h2+x)(x_{[\alpha }\delta _{\beta ]}^{[\mu }^{\nu ]}\frac{1}{2}x^2_{[\alpha }\delta _{\beta ]}^{[\mu }^{\nu ]})`$
$`\frac{2}{\left(h+n1\right)\left(2h+n4\right)\left(2h+n2\right)}x_{[\alpha }_{\beta ]}x^{[\mu }^{\nu ]}\left\}H^{(4)}_n\right(x^2|\mathrm{})T_{[\mu \nu ]n}(x).`$ (A.8)
(iv) Now we want to determine the symmetric harmonic tensor polynomials $`\stackrel{}{T}_{(\alpha \beta )n}(x)`$. Its general structure is as follows:
$`\stackrel{}{T}_{(\alpha \beta )n}=`$ $`\{\delta _\alpha ^\mu \delta _\beta ^\nu +a_ng_{\alpha \beta }x^{(\nu }^{\mu )}+b_nx_{(\alpha }\delta _{\beta )}^{(\nu }^{\mu )}+c_nx_{(\alpha }x^{(\nu }^{\mu )}_{\beta )}`$
$`+d_nx^2\delta _{(\alpha }^{(\nu }_{\beta )}^{\mu )}+e_nx^2x^{(\nu }^{\mu )}_\alpha _\beta +f_nx^2x_{(\alpha }_{\beta )}^\mu ^\nu `$
$`+g_nx_\alpha x_\beta ^\mu ^\nu +k_nx^2g_{\alpha \beta }^\mu ^\nu +l_nx^4_\alpha _\beta ^\mu ^\nu \}\stackrel{˘}{T}_{(\mu \nu )n}(x)`$ (A.9)
with the partially traceless polynomials
$$\stackrel{˘}{T}_{(\mu \nu )n}(x)=\left(\delta _\mu ^\rho \delta _\nu ^\sigma \frac{1}{2h}g_{\mu \nu }g^{\rho \sigma }\right)H_n^{(2h)}\left(x^2|\mathrm{}\right)T_{(\rho \sigma )n}(x).$$
(A.10)
It holds by construction
$$\mathrm{}\stackrel{˘}{T}_{(\mu \nu )n}(x)=0,g^{\mu \nu }\stackrel{˘}{T}_{(\mu \nu )n}(x)=0.$$
(A.11)
From eqs. (A.4) the following system of linear equations for the nine unknown coefficients results:
$`0`$ $`=`$ $`(n1)c_n+b_n+2ha_n`$
$`0`$ $`=`$ $`d_n+g_n+(n2)f_n+2hk_n`$
$`0`$ $`=`$ $`2(h+n2)d_n+c_n+b_n`$
$`0`$ $`=`$ $`2(h+n2)e_n+c_n`$
$`0`$ $`=`$ $`2(h+n2)f_n+c_n+2g_n`$
$`0`$ $`=`$ $`2(h+n2)k_n+a_n+g_n`$
$`0`$ $`=`$ $`4(h+n3)l_n+f_n+e_n`$
$`0`$ $`=`$ $`(2h+n)b_n+2(n1)d_n+2a_n+2`$
$`0`$ $`=`$ $`(2h+n)c_n+2d_n+2a_n+4(n2)e_n`$
$`0`$ $`=`$ $`(2h+n)f_n+2k_n+d_n+8(n3)l_n+2e_n`$
$`0`$ $`=`$ $`2(2h+n1)g_n+4k_n+c_n+2(n2)f_n+b_n.`$
Its unique solution is
$`a_n`$ $`=`$ $`{\displaystyle \frac{h+n2}{(h1)(h+n)(2h+n2)}},`$
$`b_n`$ $`=`$ $`2{\displaystyle \frac{h(h+n1)n+1}{(h1)(h+n)(h+n1)(2h+n2)}},`$
$`c_n`$ $`=`$ $`2{\displaystyle \frac{h+n2}{(h1)(h+n)(h+n1)(2h+n2)}},`$
$`d_n`$ $`=`$ $`{\displaystyle \frac{h(h+n1)n+2}{(h1)(h+n)(h+n1)(2h+n2)}},`$
$`e_n`$ $`=`$ $`{\displaystyle \frac{1}{(h1)(h+n)(h+n1)(2h+n2)}},`$
$`f_n`$ $`=`$ $`{\displaystyle \frac{h(h+n4)2(n3)}{(h1)(h+n)(h+n1)(2h+n2)(2h+n3)}},`$
$`g_n`$ $`=`$ $`{\displaystyle \frac{\left(h(h+n1)n+3\right)(h+n2)}{(h1)(h+n)(h+n1)(2h+n2)(2h+n3)}},`$
$`k_n`$ $`=`$ $`{\displaystyle \frac{h(h+n3)2n+3+\frac{1}{2}(h+n)(h+n1)}{(h1)(h+n)(h+n1)(2h+n2)(2h+n3)}},`$
$`l_n`$ $`=`$ $`{\displaystyle \frac{h3}{(h1)(h+n)(h+n1)(2h+n2)(2h+n3)}}.`$
In the case $`D=4`$ these coefficients simplify; they are given as follows:
$`a_n`$ $`={\displaystyle \frac{n}{(n+2)^2}},`$ $`b_n=2{\displaystyle \frac{n(n+3)}{(n+1)(n+2)^2}},`$
$`c_n`$ $`=2{\displaystyle \frac{n}{(n+1)(n+2)^2}},`$ $`d_n={\displaystyle \frac{4+n}{(n+1)(n+2)^2}},`$
$`e_n`$ $`={\displaystyle \frac{1}{(n+1)(n+2)^2}},`$ $`f_n=2{\displaystyle \frac{1}{(n+1)^2(n+2)^2}},`$
$`g_n`$ $`={\displaystyle \frac{n(n+3)}{(n+1)^2(n+2)^2}},`$ $`k_n={\displaystyle \frac{1}{2}}{\displaystyle \frac{n^2+3n+4}{(n+1)^2(n+2)^2}},`$
$`l_n`$ $`={\displaystyle \frac{1}{4}}{\displaystyle \frac{1}{(n+1)^2(n+2)^2}}.`$
This finishes the determination of the harmonic tensor polynomials being necessary for the present study. In principle, the extension of the procedure to arbitrary tensors of higher order is obvious. However, the explicit computation, without any additional information about their general properties, will be quite complicated. Let us remark that, to the best of our knowledge, such quantities have not been considered in the mathematical literature.
In contrast to the harmonic scalar polynomials which carry irreducible representations of $`SO(2h)`$, the harmonic vector and tensor polynomials are not irreducible; they are only traceless. In order to obtain irreducible harmonic vector and tensor polynomials one has to (anti)symmetrize according to the possible Young patterns as we have done to construct operators with well-defined twist.
Finally, we remark that another method exists for the construction of scalar harmonic polynomials which seems to be well adapted to the problem of twist decomposition of light-cone operators. This method uses the space of homogeneous polynomials of degree $`n`$ on the complex cone as the carrier space for symmetric traceless tensors and its harmonic extensions . However, no general theoretical frame exists for the vector and tensor case.
|
warning/0003/gr-qc0003051.html
|
ar5iv
|
text
|
# Poincaré invariance in the ADM Hamiltonian approach to the general relativistic two-body problem
## Acknowledgments
We thank L. Blanchet for informing us, before completion of our and his work, that he and G. Faye had succeeded in determining $`\omega _{\text{kinetic}}`$. He communicated to us the numerical value $`\omega _{\text{kinetic}}1.71`$. P.J. gratefully acknowledges useful discussions with Piotr Bizoń and Prof. Andrzej Staruszkiewicz. P.J. and G.S. thank the Institut des Hautes Études Scientifiques for hospitality during the realization of this work. This work was supported in part by the KBN Grant No. 2 P03B 094 17 (to P.J.) and the Max-Planck-Gesellschaft Grant No. 02160-361-TG74 (to G.S.).
|
warning/0003/gr-qc0003013.html
|
ar5iv
|
text
|
# Asymptotic singular behavior of Gowdy spacetimes in string theory
## I Introduction
The singularity theorem gives us some sufficient conditions for the existence of spacetime singularities (a spacetime with incomplete geodesics) . These conditions fall into four categories : (1) the strong energy condition; (2) the generic condition; (3) the chronology condition and (4) the existence of trapped regions. In addition, after the discovery of the singularity theorem, various generalization by replacing these conditions have been suggested by some authors (see e.g. Refs.). So, although, we have new many variations of the singularity theorem, none of them gives information on the nature of the singularity.
To the best of our knowledge, the first research about the nature of the singularity was done by Belinskii, Khalatnikov and Lifshitz (BKL) . They investigated how the spacetime evolves into the singularity and conjectured that the dynamics of nearby observers would decouple near singularities. (the BKL conjecture.) BKL described a vacuum and spatially homogeneous (mainly Bianchi IX) spacetime singularity as an infinite sequence of Kasner epochs (“oscillatory approach to the singularity”). Independently, Misner showed the same behavior in terms of exponentially growing “potential” terms constructed by the spatial curvature. The spacetime is bounced by the potentials a infinite number of times. Such behavior is called the Mixmaster dynamics . Then, BKL further speculated that a generic singularity should be locally behaved like a Mixmaster type. To date, the BKL conjecture has neither been validated nor invalidated in general by rigorous arguments although Ringström has recently have a result which supports the validity of the BKL conjecture, that is, he has shown rigorously that the Bianchi type IX solutions converge to an attractor consisting of the closure of the vacuum type II orbits.
There is a special case of the BKL conjecture called the asymptotically velocity-term dominated (AVTD) singularity. It is not described by a infinite sequence of Kasner epochs but by only one epoch (i.e., “no oscillatory approach to the singularity”) . In brief, the characteristic feature of the AVTD singular behavior is that all spatial derivative terms of the field equations are negligible sufficiently close to the singularity. Hence, Kasner spacetimes are necessarily (A)VTD. It is possible to make rigorous arguments on the nature of the AVTD singularity since the AVTD solutions are simpler than the Mixmaster solutions in the sense that there is no complicated oscillation. Our interest is to verify BKL conjecture and clarify whether spacetimes is AVTD or not in non-vacuum and/or spatially inhomogeneous cases.
In order to attack the above issue, we will consider Gowdy spacetimes . They are spatially compact spacetimes which have $`U(1)\times U(1)`$ symmetry and vanishing twist. Gowdy spacetimes are adopted in various investigations because they are one of the most manageable inhomogeneous spacetimes and can give essential features of inhomogeneity in spite of its simpleness. The behavior near the singularity in Gowdy spacetime have also been discussed in various situations. In the vacuum case, Isenberg and Moncrief have shown that polarized Gowdy spacetimes are AVTD . Grubišić and Moncrief and Kichenassamy and Rendall have shown that non-polarized $`T^3`$ Gowdy spacetimes have AVTD singularities in generic in the sense that the AVTD singular solutions to Einstein’s equations depend on the maximal number of arbitrary functions if the low velocity condition is satisfied. When one of these functions is constant (i.e. non-generic), it was also shown without the low velocity condition that non-polarized $`T^3`$ Gowdy spacetimes are AVTD . Using numerical method, Berger, Garfinkle and Moncrief found that non-polarized $`T^3`$ Gowdy spacetimes have AVTD singularities even in the situation that the low velocity condition is broken initially . Recently, Garfinkle has shown that $`S^2\times S^1`$ Gowdy spacetimes have AVTD singularities numerically .
AVTD solutions with matter fields present have been found. Such examples include polarized Gowdy spacetimes admitting scalar fields minimally coupled with the Maxwell field and the Einstein-dilaton-axion (EDA) system in polarized Gowdy spacetimes . It has been shown numerically, however, that magnetic Gowdy spacetimes are not AVTD but Mixmaster . Hence, it is nontrivial to determine whether the spacetime is Mixmaster or AVTD type. Because of these facts, we shall study the non-vacuum and spatially inhomogeneous case.
There is an another point which needs to be clarified. Belinskii and Khalatnikov have suggested that the existence of massless scalar fields suppresses Mixmaster behavior by examining the Bianchi I spacetimes . It is due to the fact that the scalar field is algebraically equivalent to the stiff matter in some cases. Recently, Berger found some confirmation for this suggestion by numerical analysis . In both papers, however, it was suggested that exponential coupling of the scalar field to the Maxwell field restores Mixmaster behavior which was suppressed initially by the scalar field . It is important to note that such matter fields arises naturally from low energy effective superstring theory , i.e. the Einstein-Maxwell-dilaton-axion (EMDA) system. The EMDA system has been discussed actively in the context of the black hole solutions and singularities, and gives characteristic features different from the Einstein-Maxwell (EM) system .
Our purpose in this paper is to explore the nature of singularities in the EMDA system with Gowdy spacetimes on $`T^3\times 𝐑`$. We make use of the Fuchsian algorithm developed by Kichenassamy and Rendall . Although the system treated here is very complicated, the Fuchsian algorithm is powerful enough for the system since this algorithm is independent of non-linearity, number of functions and dimension of space. We will show that there exist families of AVTD singular solutions of the EMDA field equations, which are generic in the sense that they depend on the maximal number of arbitrary functions. Hence, our result shows that exponential coupling of the scalar field to the Maxwell field does not necessarily lead the Mixmaster behavior.
Organization of this paper is as follows. In Sec. II, an appropriate definition of AVTD is given. In Sec. III, we discuss the importance of studying the EMDA system. Sec. IV is devoted to introducing $`T^3`$ Gowdy spacetimes in the EMDA system. Sec. V presents a brief review of the Fuchsian algorithm. Our main results are given in Sec. VI. Finally, Sec. VII is a summary.
## II Asymptotically velocity-term dominated (AVTD) singularity
In this section, we give precise definition of AVTD. In contrast to the Mixmaster singularity which is complicated, the AVTD singularity is called simple or quiescent . An essential difference between Mixmaster and AVTD singularities is whether spatial curvature terms in Einstein’s equations becomes dominant or not as the system approaches the singularities. As mentioned in the previous section, the “potential” terms constructed by the spatial curvatures grow exponentially in the Mixmaster case. These “potential” terms give rise to a infinite number of reflection and the spacetime will approach an oscillating state, i.e. Mixmaster singularities described as an infinite sequence of Kasner epochs. On the other hand, such a “potential” term does not exist in the AVTD case. Therefore, reflection by the potential does not occur and the spacetime will approach a single Kasner epoch.
Now, let us define AVTD. Suppose that a four-dimensional spacetime $`(M,g)`$ with the signature $`(+++)`$ satisfies Einstein’s equations, $`G_{\mu \nu }=T_{\mu \nu }`$, where $`G_{\mu \nu }^{(4)}R_{\mu \nu }\frac{1}{2}^{(4)}Rg_{\mu \nu }`$ is the Einstein tensor and $`T_{\mu \nu }`$ is the energy-momentum tensor. For any $`3+1`$ decomposition, we can obtain the constraint equations,
$${}_{}{}^{(3)}RK^{ab}K_{ab}+(trK)^2=2T_0^0,$$
(1)
$${}_{}{}^{(3)}_{a}^{}K_b^a^{(3)}_b(trK)=T_b^0,$$
(2)
and the evolution equations,
$$_th_{ab}=2NK_{ab}+_𝐍h_{ab},$$
(3)
$$_tK_b^a=N\left[{}_{}{}^{(3)}R_{b}^{a}+(trK)K_b^a\right]^{(3)}_b^{(3)}^aN+_𝐍K_b^a+2N\left[T_b^a+\frac{1}{2}h_b^a(T_0^0T_c^c)\right],$$
(4)
where $`h_{ab}`$, $`K_{ab}`$ and $`(trK)`$ are the first and second fundamental form and mean curvature of the $`3`$-dimensional spacelike hypersurface, respectively. Also, $`{}_{}{}^{(3)}`$, $`{}_{}{}^{(3)}R_{ab}^{}`$ and $`{}_{}{}^{(3)}R`$ are the spatial covariant derivative, spatial Ricci and scalar curvature, respectively. $`N`$ and $`𝐍`$ are the lapse function and shift vector. $`_𝐍`$ is the Lie derivative along the vector field $`𝐍`$. From the full Einstein equations, we define the velocity term dominated (VTD) equations as follows:
$$K^{ab}K_{ab}+(trK)^2=2\stackrel{~}{T}_0^0,$$
(5)
$${}_{}{}^{(3)}_{a}^{}K_b^a^{(3)}_b(trK)=\stackrel{~}{T}_b^0,$$
(6)
$$_th_{ab}=2NK_{ab},$$
(7)
$$_tK_b^a=N\left[(trK)K_b^a\right]+2N\left[\stackrel{~}{T}_b^a+\frac{1}{2}h_b^a(\stackrel{~}{T}_0^0\stackrel{~}{T}_c^c)\right],$$
(8)
where $`\stackrel{~}{}`$ is modification of $``$. (Generally, the spatial derivatives are removed.)
In this setting we can give a definition of an AVTD spacetime and its singularity.
###### Definition 1 (Isenberg and Moncrief )
A spacetime $`(M,g)`$ is AVTD if
1. it is a solution to Einstein’s equations;
2. there exists another $`(M,\stackrel{~}{g})`$ and there exists a foliation $`i_t:\mathrm{\Sigma }^3M`$ such that
1. $`(M,\stackrel{~}{g})`$ obeys the VTD equations relative to $`i_t`$;
2. metric $`g`$ asymptotically approaches $`\stackrel{~}{g}`$ as $`t0`$ in the sense that, for an appropriate norm $``$ on the space of $`3+1`$ quantities $``$, for any $`ϵ>0`$ there exists $`T>0`$ such that $`\stackrel{~}{}<ϵ`$ for all $`t<T`$.
Furthermore, a spacetime $`(M,g)`$ has the AVTD singularity if
1. it contains a spacelike singularity at $`t0`$;
2. it is AVTD.
To sum up, a singularity is called AVTD if all spatial curvature and spatial derivative terms in Einstein’s equations can be neglected near the singularity $`t0`$. Then, near the AVTD singularity, AVTD solutions can be interpreted as a different spatially homogeneous cosmology at each point in space.
In order to see an example of the AVTD singularity, let us consider vacuum Bianchi I (Kasner) spacetimes. Since these spacetimes are spatially homogeneous, there is no spatial derivative terms. Furthermore, they are spatially flat, i.e. $`{}_{}{}^{(3)}R=0`$. Therefore, the singularity is necessarily AVTD. On the other hand, Bianchi IX spacetimes are not AVTD but Mixmaster since the spatial curvature does not vanish and the “potential” term constructed by the spatial curvature cannot be neglected as approaching to the singularity (See Sec. III).
## III Influence of matter fields
BKL assumed the vacuum spacetimes originally in the investigation of the singularity. The reason for putting such assumption is illustrated as follows. Consider class A Bianchi spacetimes with the metric
$$ds^2=dt^2+e^{2\mathrm{\Omega }}(e^{2\beta })_{ij}\sigma ^i\sigma ^j,$$
where $`e^\mathrm{\Omega }`$ is the averaged scale factor, $`\beta =`$ diag$`(2\beta _+`$, $`\beta _++\sqrt{3}\beta _{}`$, $`\beta _+\sqrt{3}\beta _{})`$ is a traceless matrix that determines the anisotropy and $`d\sigma ^i=ϵ_{}^{i}{}_{ik}{}^{}\sigma ^j\sigma ^k`$. For simplicity, take $`\sigma ^i=dx^i`$, i.e. Bianchi I spacetimes. Then, $`e^{3\mathrm{\Omega }}=t`$ and $`\mathrm{\Omega }^2=\beta _+^2+\beta _{}^2`$. The spacetimes have singularities at $`\mathrm{\Omega }\mathrm{}`$. The square of shear $`\sigma `$ is estimated as $`\sigma ^2t^2e^{6\mathrm{\Omega }}.`$ Further, consider a perfect fluid whose equation of state is $`p=(\gamma 1)\rho `$, where $`\rho `$ and $`p`$ are the energy density and the pressure, respectively. The contribution from the shear is more dominant than that of the energy density $`\rho e^{3\mathrm{\Omega }\gamma }`$ near singularities ($`\mathrm{\Omega }\mathrm{}`$) if $`1\gamma <2`$ for natural matters such as a dust ($`\gamma =1`$) and a radiation ($`\gamma =4/3`$). Thus, the effect of the matter field is negligible for the behavior of the spacetime near singularities, and we have only to consider the vacuum case. Indeed, the known Bianchi perfect fluid solutions have similar behavior as the vacuum solutions near the singularities .
We must, however, look more carefully into the influence of the energy density in the case $`\gamma =2`$, where time dependence of the energy density and shear are contribute equally. The matter with $`\gamma =2`$ is a stiff matter , and is not realistic matter in the relativistic sense. However, it is known that a scalar field $`\varphi `$ is algebraically equivalent to a stiff matter if $`\varphi `$ is timelike , where $``$ is covariant derivative with respect to $`g`$. Therefore, the existence of scalar fields will give strong influence on the nature of singularities.
Actually, Belinskii and Khalatnikov found that a massless scalar field can suppress Mixmaster oscillations . The result obtained by Belinskii and Khalatnikov is that all of Kasner exponents can be positive in Bianchi I spacetimes if the scalar field exists. Therefore, complicated permutation in the direction of anisotropic contraction toward singularities does not appear. Numerically, Berger confirmed Belinskii and Khalatnikov’s result in Bianchi IX spacetimes and furthermore, she showed that Mixmaster oscillations are suppressed in non-polarized $`U(1)`$-symmetric and magnetic Gowdy spacetimes which are Mixmaster if there exist no scalar fields .
On the contrary, there is a possibility that the existence of exponential potentials of scalar fields drastically changes the behavior of singularities, i.e. the Mixmaster dynamics may come back. Recall the mechanism of the Mixmaster dynamics. The Hamiltonian $`_{IX}`$ for Bianchi IX spacetimes is
$$2_{IX}=p_\mathrm{\Omega }^2+p_+^2+p_{}^2+V(\mathrm{\Omega },\beta _\pm ),$$
where $`p_\mathrm{\Omega }`$ and $`p_\pm `$ are conjugate momenta of $`\mathrm{\Omega }`$ and $`\beta _\pm `$, respectively. The “potential” term $`V(\mathrm{\Omega },\beta _\pm )`$ is given by
$$V(\mathrm{\Omega },\beta _\pm )=e^{4\mathrm{\Omega }}\left[e^{8\beta _+}+e^{4\beta _++4\sqrt{3}\beta _{}}+e^{4\beta _+4\sqrt{3}\beta _{}}2\left(e^{4\beta _+}+e^{2\beta _+2\sqrt{3}\beta _{}}+e^{2\beta _++2\sqrt{3}\beta _{}}\right)\right].$$
In the Bianchi I spacetimes there is no such “potential” term. It is known that Bianchi IX spacetimes have a singularity as $`\mathrm{\Omega }\mathrm{}`$. If the spacetimes are AVTD, the “potential” term must vanish as $`\mathrm{\Omega }\mathrm{}`$. When $`V(\mathrm{\Omega },\beta _\pm )=0`$, we have $`\beta _\pm =\beta _\pm ^0+v_\pm |\mathrm{\Omega }|`$, $`p_\mathrm{\Omega }=`$const. and $`p_\pm =`$const. by varying the Hamiltonian $`_{IX}`$, where $`v_\pm p_\pm /|p_\mathrm{\Omega }|`$ and $`\beta _\pm ^0`$ are constant. The Hamiltonian constraint $`_{IX}=0`$ is $`v_+^2+v_{}^2=1`$. Then,
$$Ve^{4|\mathrm{\Omega }|(1+2\mathrm{cos}\theta )}+e^{4|\mathrm{\Omega }|(1\mathrm{cos}\theta \sqrt{3}\mathrm{sin}\theta )}+e^{4|\mathrm{\Omega }|(1\mathrm{cos}\theta +\sqrt{3}\mathrm{sin}\theta )},$$
where we parameterize $`v_+=\mathrm{cos}\theta `$ and $`v_{}=\mathrm{sin}\theta `$. For generic $`\theta `$ (except for $`\theta =0,2\pi /3,4\pi /3`$), the term $`V`$ exponentially grows as $`\mathrm{\Omega }\mathrm{}`$. This is a contradiction to vanish the “potential” term as $`\mathrm{\Omega }\mathrm{}`$. Thus, Bianchi IX spacetimes are not AVTD since the “potential” term becomes dominant near the singularity. As a result, reflections by the potential cause the Mixmaster dynamics.
As we have seen in the above discussion, the essential point is the existence of exponentially growing “potential” terms. Then, Belinskii and Khalatnikov have claimed that a scalar field $`\varphi `$ exponentially coupled with the Maxwell field restores Mixmaster oscillations in homogeneous spacetimes . Berger also suggested that if there exists a “potential”
$$V(\varphi )=𝒜^2e^{\alpha \varphi }+^2e^{\beta \varphi },\alpha ,\beta >0,$$
(9)
where $`𝒜`$ and $``$ are functions, then Mixmaster oscillations can be retained in both homogeneous and inhomogeneous spacetimes .
We wish to explore how “potential” terms such as those in Eq. (9) produced by matter fields might change the nature of the approach to the singularity. As a first example we consider matter coupled to the Gowdy spacetime. It has been shown mathematically and numerically that vacuum Gowdy spacetimes are AVTD . It is known numerically that certain types of magnetic fields can ruin this behavior . While heuristically one would not expect this to happen for the matter fields we study here, no rigorous demonstration that Gowdy coupled to matter fields including exponential potentials remains AVTD has been possible until now.
It should be noted that this form of potential arises from the low energy effective superstring theory. It contains the dilaton which is non-minimally coupled with the Maxwell and axion field. From the unified theoretical point of view, it is believed that the distinction between gravity and matter fields cannot be maintained near spacetime singularities and the superstring theory is the most promising candidate of such unified theories . Also, it is pointed out that the EMD system with a positive cosmological constant has new mechanism forming singularities . Hence, it is important to investigate the behavior around the singularity in the EMDA system.
## IV Einstein-Maxwell-Dilaton-Axion system in Gowdy spacetimes on $`T^3\times R`$
In this section, we introduce a standard model with additional assumptions. First, we adopt Gowdy spacetimes. Gowdy proposed spatially compact and inhomogeneous spacetimes with the following metric ;
$$ds^2=e^{\lambda (t,\theta )/2}t^{1/2}(dt^2+d\theta ^2)+R(t,\theta )\left[e^{Z(t,\theta )}\left(dy+X(t,\theta )dz\right)^2+e^{Z(t,\theta )}dz^2\right].$$
(10)
Gowdy spacetimes have two twist free spacelike Killing vectors $`/y`$ and $`/z`$. The spatial topology of the spacetimes is classified into three types, $`T^3`$, $`S^2\times S^1`$ and $`S^3`$. We will consider only the simplest case, $`T^3`$, since the isometry group action has no degenerate orbits and therefore Einstein’s equations have no corresponding singularities . Gowdy spacetimes are called polarized if $`X0`$, and non-polarized if this condition is not meet.
Properties of the metric (10) depend on whether $`R`$ is timelike, spacelike or null. When the metric Eq. (10) describes a cosmological model, i.e. $`R`$ is globally timelike and the spatial topology is $`T^3`$ (periodic in $`\theta `$), one can take the function $`R(t,\theta )=t`$ without loss of generality by Gowdy’s corner theorem if the spacetime is vacuum . This fact be seen from Einstein’s equations,
$$G_{tt}G_{\theta \theta }=\ddot{R}R^{\prime \prime }=0,$$
(11)
where dot and prime denote $`t`$ and $`\theta `$ derivatives, respectively. Also in the EDA system, Eq. (11) holds . In the EM system, Eq. (11) is not satisfied generically. However in the case that the field strength $`F_{\mu \nu }=_\mu A_\nu _\nu A_\mu `$ has only the following components ;
$$F_{ty}=\dot{\omega },F_{\theta y}=\omega ^{},F_{tz}=\dot{\chi },F_{\theta z}=\chi ^{},$$
Eq. (11) is satisfied. This situation does not change even if the dilaton field is exponentially coupled with the Maxwell field. Thus, in the EMDA system, we can put $`R(t,\theta )=t`$. Hereafter, we will choose this gauge. In this case, Gowdy spacetimes have spacelike singularities at $`t=0`$ .
The action of the EMDA theory is
$`S`$ $`=`$ $`{\displaystyle d^4x\sqrt{g}\left[^{(4)}R+e^{2a\varphi }F^2+2\left(\varphi \right)^2+\frac{1}{3}e^{4a\varphi }H^2\right]}`$ (12)
$`=`$ $`{\displaystyle \frac{1}{2}}t\left[\ddot{\lambda }\lambda ^{\prime \prime }+\dot{Z}^2Z^2+e^{2Z}\left(\dot{X}^2X^2\right)\right]`$ (15)
$`2e^{2a\varphi }\left[\left(e^ZX^2+e^Z\right)\left(\dot{\omega }^2\omega ^2\right)+e^Z\left(\dot{\chi }^2\chi ^2\right)2Xe^Z\left(\dot{\omega }\dot{\chi }\omega ^{}\chi ^{}\right)\right]`$
$`2t\left(\dot{\varphi }^2\varphi ^2\right){\displaystyle \frac{1}{2}}e^{4a\varphi }t\left(\dot{\kappa }^2\kappa ^2\right),`$
where $`\varphi =\varphi (t,\theta )`$ is the dilaton field, $`H=dB\frac{1}{2}e^{4a\varphi }d\kappa (t,\theta )`$ is the three-index antisymmetric tensor field dual to the axion field $`\kappa `$, and $`a`$ is a coupling constant . The functions are periodic in $`\theta `$ ($`0\theta 2\pi `$) because the spatial topology is $`T^3`$. \[However, the spatial compactness is not relevant to our analysis.\] In the case of the effective superstring theory, $`a=1`$. If we neglect the axion field, $`a=\sqrt{3}`$ gives Kaluza-Klein theory. If $`a=0`$, the dilaton and axion fields are decoupled with the gravity.
Note that the metric function $`\lambda `$ is decoupled with other functions, and that $`\lambda `$ appears only in the Hamiltonian and momentum constraints. This is the primary advantage of Gowdy spacetimes. Now, let us focus on evolution equations. We can calculate the metric function $`\lambda `$ by evaluating the integral of $`\lambda ^{}`$ from $`0`$ to $`2\pi `$ after obtaining other functions.
## V Review of the Fuchsian algorithm
The above system is very complicated and gives rise to highly nonlinear partial differential equations which are nontrivially coupled with the matter fields. We would like to construct AVTD singular solutions of such complicated equations which are parametrized by as many arbitrary functions as possible. In order to do this, we adopt the Fuchsian algorithm developed by Kichenassamy and Rendall . One of the advantages of this algorithm is applicable to general (nonlinear and/or singular) partial differential equations (PDEs). Another advantage of the Fuchsian algorithm is applicable to arbitrary spatial dimension. \[Spatial dimension in Gowdy spacetimes is one since these spacetimes have two spacelike Killing vector fields.\] This fact suggests that the Fuchsian algorithm is applicable to spacetimes with fewer or no symmetry .
In this section, we briefly review the Fuchsian algorithm . Let us consider a PDE system
$$F\left[u(t,x^\alpha )\right]=0.$$
(16)
Generically, $`u`$ can have any number of components. Here, we will assume that the PDE is singular with respect to the argument $`t`$. The Fuchsian algorithm consists of three steps as follows:
Step 1 Identify the leading (singular) terms $`u_0(t,x)`$ which are parts of the desired expansion for $`u`$. This means that the most singular terms cancel each other when $`u_0(t,x)`$ is substituted in Eq. (16).
Step 2 Introduce a renormalized unknown function $`v(t,x)`$, which is given by
$$u=u_0+t^mv.$$
(17)
If $`u_0t^k`$, we should set $`m=k+\epsilon `$, where $`\epsilon >0`$. Thus, $`v`$ is a regular part of the desired expansion for $`u`$.
Step 3 Obtain a Fuchsian system for $`v`$ by substituting Eq. (17) in Eq. (16). That is,
$$(D+A)v=t^ϵf(t,x,v,_xv),$$
(18)
where $`Dt_t`$ and $`A`$ is a matrix which is independent of $`t`$ and $`ϵ>0`$. $`f`$ can be assumed to be analytic in all of arguments except $`t`$ and continuous in $`t`$. Note that Eq. (18) is a (singular) PDE system for the regular function $`v`$.
Roughly speaking, the Fuchsian algorithm is a transformation from finding singular solutions of the original equations (16) into finding regular solutions of the Fuchsian equations (18) which may be singular even if the original equations are regular. Once we have the Fuchsian system, we can show the existence of a unique solution for prescribed singular part $`u_0`$ by the following theorem.
###### Theorem 1 (Theorem 3 of Ref. )
Let us consider a system Eq. (18), where $`A`$ is an analytic matrix near $`x=0`$, such that $`\sigma ^AC`$ for $`0<\sigma <1`$ (boundedness condition) and $`f`$ is analytic in space $`x`$ and continuous in time $`t`$. Then the system (18) has exactly one solution which is defined near $`x=0`$ and $`t=0`$, and which is analytic in space and continuous in time, and tend to zero as $`t0`$. $`\mathrm{}`$
Applying the Fuchsian algorithm to our problem of Gowdy spacetimes in the EMDA system, the procedure is summarized as follows:
1. Solve VTD equations as ordinary differential equations with respect to $`t`$. Solutions obtained such a way are the leading (singular) terms of the desired formal solutions.
2. Substitute the formal solutions into full field equations. Here, make integral constants arbitrary functions of spatial arguments $`x`$.
3. Obtain a Fuchsian system for unknown functions and evaluate every eigenvalues of $`A`$.
4. Apply Theorem 1 to the system.
If every eigenvalues of $`A`$ are non-negative, the boundedness condition in Theorem 1 holds. Then, renormalized unknown functions must vanish as $`t0`$ by Theorem 1. Therefore, the only singular terms which are solutions to VTD equations remain and they are solutions to full field equations at $`t=0`$. Thus, we have AVTD singular solutions. Essentially, the above argument is a singular version of the Cauchy-Kowalevskaya theorem .
## VI Asymptotic behavior in Gowdy spacetimes
We are now ready to examine the singularity in the EMDA system. The EMDA system with Gowdy symmetry, however, has six functions ($`Z`$, $`X`$, $`\omega `$, $`\chi `$, $`\varphi `$, $`\kappa `$), and it is very complicated to treat all of them at the same time. Hence, we will reduce the system by selecting four functions from them for simplicity. As a selection rule, it is important to leave exponential coupling between the dilaton and other fields since, as mentioned in Sec.III, an essential part of our analysis is whether or not “potential” terms in Einstein’s equations exponentially grow as $`t0`$ . Therefore, we will consider three cases, which are the non-polarized EMD, the non-polarized EDA and the polarized EMDA.
### A non-polarized EMD ($`\chi =0`$, $`\kappa =0`$)
First, let us consider the case that the dilaton field is exponentially coupled with the Maxwell field. Varying the action (12) with respect to ($`Z`$, $`X`$, $`\omega `$, $`\chi `$), we obtain evolution equations,
$`D^2Zt^2Z^{\prime \prime }+t^2e^{2Z}(\dot{X}^2X^2)+2te^{2a\varphi }(e^ZX^2e^Z)(\dot{\omega }^2\omega ^2)=0,`$ (19)
$`D^2Xt^2X^{\prime \prime }2t^2(\dot{X}\dot{Z}X^{}Z^{})4te^{2a\varphi +Z}X(\dot{\omega }^2\omega ^2)=0,`$ (20)
$`D^2\omega t^2\omega ^{\prime \prime }t^2\left[2a\dot{\varphi }+{\displaystyle \frac{1}{t}}{\displaystyle \frac{(2\dot{X}\dot{Z}X)Xe^Z+\dot{Z}e^Z}{e^ZX^2+e^Z}}\right]\dot{\omega }`$ (21)
$`+t^2\left[2a\varphi ^{}{\displaystyle \frac{(2X^{}Z^{}X)Xe^Z+Z^{}e^Z}{e^ZX^2+e^Z}}\right]\omega ^{}=0,`$ (22)
$`D^2\varphi t^2\varphi ^{\prime \prime }+ate^{2a\varphi }\left(e^ZX^2+e^Z\right)(\dot{\omega }^2\omega ^2)=0.`$ (23)
Following the procedure explained in the previous section, we have VTD equations by dropping spatial derivative terms from Eqs.(19)-(23).
$`D^2Z+t^2e^{2Z}\dot{X}^2+2te^{2a\varphi }(e^ZX^2e^Z)\dot{\omega }^2=0,`$ (24)
$`D^2X2t^2\dot{X}\dot{Z}4te^{2a\varphi +Z}X\dot{\omega }^2=0,`$ (25)
$`D^2\omega t^2\left\{2a\dot{\varphi }+{\displaystyle \frac{1}{t}}{\displaystyle \frac{(2\dot{X}\dot{Z}X)Xe^Z+\dot{Z}e^Z}{e^ZX^2+e^Z}}\right\}\dot{\omega }=0,`$ (26)
$`D^2\varphi +ate^{2a\varphi }(e^ZX^2+e^Z)\dot{\omega }^2=0.`$ (27)
According to Definition 1, solutions to these equations are AVTD if they exist and are also solutions to full Einstein’s equations as $`t0`$.
Since we expect Kasner-like solutions, $`ZZ_0\mathrm{ln}t`$, $`XX_0`$ and $`\varphi \varphi _0\mathrm{ln}t`$ are chosen as the leading terms for $`Z`$, $`X`$ and $`\varphi `$, respectively. Substituting them into Eq. (26), we have a leading term of $`\omega `$. Thus, we can construct a family of formal solutions to full Einstein’s equations as,
$`Z=Z_0(\theta )\mathrm{ln}t+Z_1(\theta )+t^ϵ\alpha (t,\theta ),`$ (28)
$`X=X_0(\theta )+t^{2Z_0}\left(X_1(\theta )+\beta (t,\theta )\right),`$ (29)
$`\omega =\omega _0(\theta )+t^{Z_0+2a\varphi _0+1}\left(\omega _1(\theta )+\gamma (t,\theta )\right),`$ (30)
$`\varphi =\varphi _0(\theta )\mathrm{ln}t+\varphi _1(\theta )+t^ϵ\delta (t,\theta ),`$ (31)
where $`Z_0>0`$, $`Z_0+2a\varphi _0+1>0`$, $`ϵ>0`$. The first and second inequalities are given by Eqs. (25) and (26). $`\alpha `$, $`\beta `$, $`\gamma `$ and $`\delta `$ are renormalized unknown functions. Note that when we put $`\alpha =\beta =\gamma =\delta =0`$, Eqs. (28)-(31) are exact solutions to VTD equations as $`t0`$. Substituting the formal solutions (28)-(31) into full Einstein’s equations (19)-(23), we can obtain the following system,
$$(D+A)\stackrel{}{u}=\stackrel{}{f}(t,\theta ,\stackrel{}{u},\stackrel{}{u}^{}),$$
(32)
where
$`\stackrel{}{u}`$ $`=`$ $`(\alpha ,D\alpha ,t\alpha ^{},\beta ,D\beta ,t\beta ^{},\gamma ,D\gamma ,t\gamma ^{},\delta ,D\delta ,t\delta ^{})^T.`$ (33)
The matrix $`A`$ has the $`12\times 12`$ components,
$$A=\left(\begin{array}{cccc}A_Z& 0& 0& 0\\ 0& A_X& 0& 0\\ 0& 0& A_\omega & 0\\ 0& 0& 0& A_\varphi \end{array}\right),$$
where,
$$A_Z=\left(\begin{array}{ccc}0& 1& 0\\ ϵ^2& 2ϵ& 0\\ 0& 0& 0\end{array}\right),A_X=\left(\begin{array}{ccc}0& 1& 0\\ 0& 2Z_0& 0\\ 0& 0& 0\end{array}\right),$$
$$A_\omega =\left(\begin{array}{ccc}0& 1& 0\\ 0& Z_0+2a\varphi _0+1& 0\\ 0& 0& 0\end{array}\right),A_\varphi =\left(\begin{array}{ccc}0& 1& 0\\ ϵ^2& 2ϵ& 0\\ 0& 0& 0\end{array}\right).$$
We can easily verify that eigenvalues of $`A`$ are $`0`$, $`2Z_0`$, $`Z_0+2a\varphi _0+1`$ and $`ϵ`$, which are non-negative. Then, we conclude that the boundedness condition of Theorem 1 holds. Next, we examine the leading parts in components of the vector $`\stackrel{}{f}`$.
$$\stackrel{}{f}=\left(\begin{array}{c}0\\ X_0^2e^{2(Z_1+t^ϵ\alpha )}t^{22Z_0ϵ}+2X_0^2\omega _0^2e^{(2a(\varphi _1+t^ϵ\delta )+Z_1+t^ϵ\alpha )}t^{1Z_02a\varphi _0ϵ}+\mathrm{}\\ t(\alpha +D\alpha )^{}\\ 0\\ 2X_0^{}(Z_0^{}\mathrm{ln}t+Z_1^{}+t^ϵ\alpha )t^{22Z_0}4X_0\omega _0^2e^{2a(\varphi _1+t^ϵ\delta )+Z_1+t^ϵ\alpha }t^{1Z_02a\varphi _0}+\mathrm{}\\ t(\beta +D\beta )^{}\\ 0\\ X_0^2e^{Z_1+t^ϵ\alpha }\{\omega _0^{\prime \prime }[2a(\varphi _0^{}\mathrm{ln}t+\varphi _1^{}+t^ϵ\delta ^{})+1]\omega _0^{}\}t^{1Z_02a\varphi _0}+\mathrm{}\\ t(\gamma +D\gamma )^{}\\ 0\\ aX_0^2\omega _0^2e^{(2a(\varphi _1+t^ϵ\delta )+Z_1+t^ϵ\alpha )}t^{1Z_02a\varphi _0ϵ}+\mathrm{}\\ t(\delta +D\delta )^{}\end{array}\right).$$
The dots represent the terms which are lower order with respect to $`t`$ and are not important for our discussion since the power of these terms is necessarily positive if conditions $`Z_0>0`$ and $`Z_0+2a\varphi _0+1>0`$ hold. Positivity of the power of $`t`$ in components of the vector $`\stackrel{}{f}`$ gives us sufficient conditions to obtain the AVTD solutions. Finally, we apply Theorem 1 to this system and then, we obtain the following theorem.
###### Theorem 2
Let $`Z_i(\theta )`$, $`X_i(\theta )`$, $`\omega _i(\theta )`$, $`\varphi _i(\theta )`$, where $`i=0`$, $`1`$, be real analytic functions and $`0<Z_0<1`$ and $`1<Z_0+2a\varphi _0<1`$ for $`0\theta 2\pi `$. Then, for field equations (19)-(23) in the EMD system, there exists a unique solution with the form (28)-(31), where $`\alpha `$, $`\beta `$, $`\gamma `$ and $`\delta `$ tend to zero as $`t0`$. $`\mathrm{}`$
Theorem 2 means that $`T^3`$ Gowdy spacetimes in the EMD system have AVTD singularities in “general” in the sense that the singular solutions have maximal number (i.e. eight) of arbitrary functions ($`Z_0`$, $`X_0`$, $`\omega _0`$, $`\varphi _0`$, $`Z_1`$, $`X_1`$, $`\omega _1`$, $`\varphi _1`$).
In components of the vector $`\stackrel{}{f}`$, terms $`t^{1Z_02a\varphi _0}`$ and $`t^{22Z_0}`$ give us upper bound for $`Z_0`$ and $`Z_0+2a\varphi _0`$ in Theorem 2. However, since these terms are always multiplied by $`X_0^{}`$ and $`\omega _0^{}`$, the upper bound can be removed if $`X_0^{}=0`$ and $`\omega _0^{}=0`$. Hence, we obtain the following corollary:
###### Corollary 1
Let $`Z_i(\theta )`$, $`X_i(\theta )`$, $`\omega _i(\theta )`$, $`\varphi _i(\theta )`$, where $`i=0`$, $`1`$, be real analytic functions, and assume $`0<Z_0`$, $`1<Z_0+2a\varphi _0`$ and $`X_0^{}=\omega _0^{}=0`$ for $`0\theta 2\pi `$. Then, for field equations (19)-(23) in the EMD system, there exists a unique solution with the form (28)-(31), where $`\alpha `$, $`\beta `$, $`\gamma `$ and $`\delta `$ tend to zero as $`t0`$. $`\mathrm{}`$
Note that this is a “non-generic” case in the sense that two of arbitrary eight functions are constant in $`\theta `$.
### B non-polarized EDA ($`\omega =0`$, $`\chi =0`$)
Now, we will consider the case that the axion field is exponentially coupled with the dilaton field. In the similar way to Sec.VI A, we have evolution equations by varying the action (12) with respect to $`(Z,X,\varphi ,\kappa )`$.
$`D^2Zt^2Z^{\prime \prime }+t^2e^{2Z}(\dot{X}^2X^2)=0,`$ (34)
$`D^2Xt^2X^{\prime \prime }2t^2(\dot{X}\dot{Z}X^{}Z^{})=0,`$ (35)
$`D^2\varphi t^2\varphi ^{\prime \prime }{\displaystyle \frac{a}{2}}t^2e^{4a\varphi }(\dot{\kappa }^2\kappa ^2)=0,`$ (36)
$`D^2\kappa t^2\kappa ^{\prime \prime }+4at^2(\dot{\varphi }\dot{\kappa }\varphi ^{}\kappa ^{})=0.`$ (37)
Note that a pair of PDEs for gravity (34) and (35) is decoupled with those for scalar fields (36) and (37). Furthermore, Eqs. (34) and (35) have equivalent form to Eqs. (36) and (37) as simultaneous PDEs. These facts are known as “mirror images” in string cosmologies and used to construct new nontrivial solutions .
Dropping the spatial derivative terms from Eqs. (34)-(37), VTD equations are obtained as follows:
$`D^2Z+t^2e^{2Z}\dot{X}^2=0,`$ (38)
$`D^2X2t^2\dot{X}\dot{Z}=0,`$ (39)
$`D^2\varphi {\displaystyle \frac{a}{2}}t^2e^{4a\varphi }\dot{\kappa }^2=0,`$ (40)
$`D^2\kappa +4at^2\dot{\varphi }\dot{\kappa }=0.`$ (41)
We can easily solve these equations as $`t0`$ and find formal solutions to full Einstein’s equations.
$`Z=Z_0(\theta )\mathrm{ln}t+Z_1(\theta )+t^ϵ\alpha (t,\theta ),`$ (42)
$`X=X_0(\theta )+t^{2Z_0}(X_1(\theta )+\beta (t,\theta )),`$ (43)
$`\varphi =\varphi _0(\theta )\mathrm{ln}t+\varphi _1(\theta )+t^ϵ\gamma (t,\theta ),`$ (44)
$`\kappa =\kappa _0(\theta )+t^{4a\varphi _0}(\kappa _1(\theta )+\delta (t,\theta )),`$ (45)
where $`Z_0>0`$, $`2a\varphi _0>0`$ and $`ϵ>0`$. These restrictions for $`Z_0`$ and $`\varphi _0`$ given by Eqs. (39) and (41), respectively.
The system for regular functions $`\alpha `$, $`\beta `$, $`\gamma `$ and $`\delta `$ in the EDA system is given by Eqs. (32) and (33) by replacing the $`12\times 12`$ matrix $`A`$ by
$$A=\left(\begin{array}{cccc}A_Z& 0& 0& 0\\ 0& A_X& 0& 0\\ 0& 0& A_\varphi & 0\\ 0& 0& 0& A_\kappa \end{array}\right),$$
where
$$A_\kappa =\left(\begin{array}{ccc}0& 1& 0\\ 16a^2\varphi _0^2& 4a\varphi _0& 0\\ 0& 0& 0\end{array}\right).$$
Eigenvalues of $`A`$ are $`0`$, $`2Z_0,4a\varphi _0`$ and $`ϵ`$, and all of them are non-negative. These facts imply that the boundedness condition in Theorem 1 holds. After long calculation, the leading parts in components of the vector $`\stackrel{}{f}`$ is obtained as follows.
$$\stackrel{}{f}=\left(\begin{array}{c}0\\ X_0^2e^{2(Z_1+t^ϵ\alpha )}t^{22Z_0ϵ}+\mathrm{}\\ t(\alpha +D\alpha )^{}\\ 0\\ \left[X_0^{\prime \prime }2X_0^{}(Z_0^{}\mathrm{ln}t+Z_1^{}+t^ϵ\alpha ^{})\right]t^{22Z_0}+\mathrm{}\\ t(\beta +D\beta )^{}\\ 0\\ \frac{a}{2}\kappa _0^2e^{4a(\varphi _1+t^ϵ\gamma )}t^{2+4a\varphi _0ϵ}+\mathrm{}\\ t(\gamma +D\gamma )^{}\\ 0\\ \left[\kappa _0^{\prime \prime }+4a\kappa _0^{}(\varphi _0^{}\mathrm{ln}t+\varphi _1^{}+t^ϵ\gamma ^{})\right]t^{2+4a\varphi _0}+\mathrm{}\\ t(\delta +D\delta )^{}\end{array}\right).$$
Since the power of $`t`$ of other terms in $`\stackrel{}{f}`$ is positive under conditions $`Z_0>0`$ and $`2a\varphi _0>0`$, and then they are not needed here. The AVTD behavior requires that the power of the leading parts of $`t`$ is positive. Thus, we have the following theorem which states that non-polarized Gowdy spacetimes in the EDA system are AVTD in general.
###### Theorem 3
Let $`Z_i(\theta )`$, $`X_i(\theta )`$, $`\varphi _i(\theta )`$, $`\kappa _i(\theta )`$, where $`i=0`$, $`1`$, be real analytic and $`0<Z_0<1`$ and $`0<2a\varphi _0<1`$ for $`0\theta 2\pi `$. Then, for field equations (34)-(37) in the EDA system, there exists a unique solution of the form (42)- (45), where $`\alpha `$, $`\beta `$, $`\gamma `$ and $`\delta `$ tend to zero as $`t0`$. $`\mathrm{}`$
As seen from the vector $`\stackrel{}{f}`$, the upper bound for $`Z_0`$ and $`\varphi _0`$ in Theorem 3 are derived from terms multiplied by $`X_0^{}`$ and $`\kappa _0^{}`$. Thus, we obtain the following corollary in the “non-generic” case.
###### Corollary 2
Let $`Z_i(\theta )`$, $`X_i(\theta )`$, $`\varphi _i(\theta )`$, $`\kappa _i(\theta )`$, where $`i=0`$, $`1`$, be real analytic, and assume $`0<Z_0`$, $`0<2a\varphi _0`$ and $`X_0^{}=\kappa _0^{}=0`$, for $`0\theta 2\pi `$. Then, for field equations (34)-(37) in the EDA system, there exists a unique solution of the form (42)-(45), where $`\alpha `$, $`\beta `$, $`\gamma `$ and $`\delta `$ tend to zero as $`t0`$. $`\mathrm{}`$
### C polarized EMDA ($`X=0`$, $`\chi =0`$)
Finally, we will study the case of polarized Gowdy spacetimes in the EMDA system. As mentioned in Sec. III, the existence of “potential” like Eq. (9) might lead to Mixmaster oscillations. Such a potential appears in the EMDA system \[See Eq. 48\]. On the other hand, it has been shown that vacuum polarized $`T^3`$ Gowdy spacetimes are AVTD . Which behavior is realized in the EMDA system, AVTD or Mixmaster?
Varying the action (12) with respect to four functions $`(Z,\omega ,\varphi ,\kappa )`$, we have evolution equations as follows:
$`D^2Zt^2Z^{\prime \prime }2te^{2a\varphi +Z}(\dot{\omega }^2\omega ^2)=0,`$ (46)
$`D^2\omega +t^2\left(\dot{Z}2a\dot{\varphi }{\displaystyle \frac{1}{t}}\right)\dot{\omega }t^2\omega ^{\prime \prime }t^2(2a\varphi ^{}+Z^{})\omega ^{}=0,`$ (47)
$`D^2\varphi t^2\varphi ^{\prime \prime }+ate^{2a\varphi +Z}(\dot{\omega }^2\omega ^2)+{\displaystyle \frac{1}{2}}at^2e^{4a\varphi }(\dot{\kappa }^2\kappa ^2)=0,`$ (48)
$`D^2\kappa t^2\kappa ^{\prime \prime }+4at^2(\dot{\kappa }\dot{\varphi }\kappa ^{}\varphi ^{})=0.`$ (49)
Dropping the spatial derivative terms from Eqs. (46)-(49), following VTD equations are obtained as follows:
$`D^2Z2te^{2a\varphi +Z}\dot{\omega }^2=0,`$ (50)
$`D^2\omega +t^2\left(\dot{Z}2a\dot{\varphi }{\displaystyle \frac{1}{t}}\right)\dot{\omega }=0,`$ (51)
$`D^2\varphi +ate^{2a\varphi +Z}\dot{\omega }^2+{\displaystyle \frac{1}{2}}at^2e^{4a\varphi }\dot{\kappa }^2=0,`$ (52)
$`D^2\kappa +4at^2\dot{\kappa }\dot{\varphi }=0.`$ (53)
Similarly to the case of the non-polarized EDA system, we can solve these equations as $`t0`$. Thus, a family of formal solutions of this system can be obtained as follows:
$`Z=Z_0(\theta )\mathrm{ln}t+Z_1(\theta )+t^ϵ\alpha (t,\theta ),`$ (54)
$`\omega =\omega _0(\theta )+t^{2a\varphi _0Z_0+1}(\omega _1(\theta )+\beta (t,\theta )),`$ (55)
$`\varphi =\varphi _0(\theta )\mathrm{ln}t+\varphi _1(\theta )+t^ϵ\gamma (t,\theta ),`$ (56)
$`\kappa =\kappa _0(\theta )+t^{4a\varphi _0}(\kappa _1(\theta )+\delta (t,\theta )),`$ (57)
where $`2a\varphi _0Z_0+1>0`$, $`4a\varphi _0>0`$ and $`ϵ>0`$. Eqs. (51) and (53) restrict the values of $`Z_0`$ and $`\varphi _0`$. By substituting Eqs. (54)-(57) into Eqs. (46)-(49), the system in the polarized EMDA system is given by Eqs. (32) and (33) with $`A`$ given by
$$A=\left(\begin{array}{cccc}A_Z& 0& 0& 0\\ 0& A_\omega & 0& 0\\ 0& 0& A_\varphi & 0\\ 0& 0& 0& A_\kappa \end{array}\right),$$
where
$$A_\omega =\left(\begin{array}{ccc}0& 1& 0\\ 0& 2a\varphi _0Z_0+1& 0\\ 0& 0& 0\end{array}\right),A_\kappa =\left(\begin{array}{ccc}0& 1& 0\\ 0& 4a\varphi _0& 0\\ 0& 0& 0\end{array}\right).$$
We can easily evaluate the eigenvalues of the matrix $`A`$ as $`0`$, $`2a\varphi _0Z_0+1`$, $`4a\varphi _0`$ and $`ϵ`$. Since they are all non-negative, the boundedness condition in Theorem 1 hold. The leading parts in components of the vector $`\stackrel{}{f}`$ are as follows:
$$\stackrel{}{f}=\left(\begin{array}{c}0\\ 2\omega _0^2e^{2a(\varphi _1+t^ϵ\gamma )+Z_1+t^ϵ\alpha }t^{12a\varphi _0+Z_0ϵ}+\mathrm{}\\ t(\alpha +D\alpha )^{}\\ 0\\ \{\omega _0^{\prime \prime }+\omega _0^{}[2a(\varphi _0^{}\mathrm{ln}t+\varphi _1^{}+t^ϵ\gamma ^{})+Z_0^{}\mathrm{ln}t+Z_1^{}+t^ϵ\alpha ^{}]\}t^{12a\varphi _0+Z_0}+\mathrm{}\\ t(\beta +D\beta )^{}\\ 0\\ a\omega _0^2e^{2a(\varphi _1+t^ϵ\gamma )+Z_1+t^ϵ\alpha }t^{12a\varphi _0+Z_0ϵ}+\frac{a}{2}\kappa _0^2e^{4a(\varphi _1+t^ϵ)}t^{2+4a\varphi _0ϵ}+\mathrm{}\\ t(\gamma +D\gamma )^{}\\ 0\\ \left[\kappa _0^{\prime \prime }+4a\kappa _0^{}(\varphi _0^{}\mathrm{ln}t+\varphi _1^{}+t^ϵ\gamma ^{})\right]t^{2+4a\varphi _0}+\mathrm{}\\ t(\delta +D\delta )^{}\end{array}\right).$$
We do not need other terms since their power is positive under the conditions $`2a\varphi _0Z_0+1>0`$ and $`4a\varphi _0>0`$. Thus, we have again the theorem which states that polarized Gowdy spacetimes in the EMDA system are AVTD in general even if there exists a “potential” as Eq. (9) in the system. This result is a negative answer of the question in Sec. III.
###### Theorem 4
Let $`Z_i(\theta )`$, $`\omega _i(\theta )`$, $`\varphi _i(\theta )`$, $`\kappa _i(\theta )`$, where $`i=0`$, $`1`$, be real analytic, and $`1<2a\varphi _0Z_0<1`$ and $`1<2a\varphi _0<0`$ for $`0\theta 2\pi `$. Then, for field equations (46)-(49) in the EMDA system, there exists a unique solution of the form (54)-(57), where $`\alpha `$, $`\beta `$, $`\gamma `$ and $`\delta `$ tend to zero as $`t0`$. $`\mathrm{}`$
The leading terms in components of the vector $`\stackrel{}{f}`$ always include factors $`\omega _0^{}`$ and/or $`\kappa _0^{}`$. Then, we can be weaken the conditions for $`Z_0`$ and $`\varphi _0`$ as similar to the cases of the EMD and the EDA systems if we put $`\omega _0^{}=\kappa _0^{}=0`$.
###### Corollary 3
Let $`Z_i(\theta )`$, $`\omega _i(\theta )`$, $`\varphi _i(\theta )`$, $`\kappa _i(\theta )`$, where $`i=0`$, $`1`$, be real analytic, and assume $`\omega _0^{}=\kappa _0^{}=0`$, $`1<2a\varphi _0Z_0`$ and $`2a\varphi _0<0`$ for $`0\theta 2\pi `$. Then, for field equations (46)-(49) in the EMDA system, there exists a unique solution of the form (54)- (57), where $`\alpha `$, $`\beta `$, $`\gamma `$ and $`\delta `$ tend to zero as $`t0`$. $`\mathrm{}`$
## VII Summary
Theorem 2, 3 and 4 state that $`T^3`$ Gowdy spacetimes in the EMDA system have AVTD singularities in “general” in the sense that these AVTD singular solutions depend on the maximal number of singular data. In the “non-generic” cases where some functions in the singular data are constant in $`\theta `$, we have obtained corollary 1, 2 and 3. These results support BKL conjecture, that is, the dynamics of spatially different points effectively are decoupled from each other. However the spacetimes in our system do not show an oscillation of the Mixmaster spacetime near cosmological singularities for some parameter regions. In this sense, our system is a special case.
When we take the matter fields into account, Theorem 4 and corollary 3 give an answer, “NO”, to the question in Sec. III, i.e. the existence of a “potential” Eq. (9) does not necessarily restore the Mixmaster behavior. We must, however, impose another condition on the initial matter field in addition to the low-velocity condition to realize the AVTD behavior by the effect of the “potential”. Thus, we can say that the AVTD behavior is controlled by the “potential”.
We can verify that the Kretschemann invariant $`R^{\mu \nu \lambda \sigma }R_{\mu \nu \lambda \sigma }`$ of all of our AVTD solutions blows up as $`t^{(Z_0^2+2\varphi _0^2+3)}`$. Thus, the AVTD singularity in the EMDA system is the curvature singularity. This supports the validity of the strong cosmic censorship for $`T^3`$ Gowdy spacetimes in the EMDA system since the spacetimes cannot be extended beyond the singularity similar to the vacuum case .
Let us compare our results with those obtained by Grubišić-Moncrief and Kichenassamy-Rendall in the (non-) polarized and vacuum case . The low-velocity condition derived by Grubišić-Moncrief and Kichenassamy-Rendall are equivalent to our condition $`0<Z_0<1`$. If spacetimes are vacuum, then non-polarized $`T^3`$ Gowdy spacetimes are AVTD under this condition. Also, when $`X0`$ (polarized), the spacetimes are AVTD without the low-velocity condition. In any case, our results clearly include Grubišić-Moncrief and Kichenassamy-Rendall’s when the matter fields vanish.
We should mention about some recent works related to our analysis. Weaver, Isenberg and Berger have shown numerically that a Gowdy spacetime in the EM system (a magnetic Gowdy spacetime) has the Mixmaster behavior . The magnetic field they considered couples the metric function $`\lambda `$ through its amplitude. Our Maxwell fields, however, do not have such character (cf. ), so Gowdy spacetimes in our EM system are able to show the AVTD behavior. It is expected that if Gowdy spacetimes in the EM system such as Weaver’s model have Mixmaster behavior, they cannot show the AVTD even if we put the dilaton and the axion fields.
Andersson and Rendall have shown by using the Fuchsian algorithm that a generic cosmological spacetime in the Einstein-scalar field system has AVTD property . They have imposed the Gaussian coordinate conditions, $`g_{00}=1`$ and $`g_{0a}=0`$ on the spacetime. As it is understood from its definition, AVTD is defined under a coordinate condition. Then, choice of gauge or coordinate conditions is important. Clearly, our gauge conditions (see Sec IV) are different from Andersson-Rendall’s. Furthermore, the equations of matter fields used in Ref. are linear PDEs. Contrary, our field equations in the EMDA system are nonlinear. These nonlinear PDEs cannot be came to linear ones. Thus, their and our results can be complements each of the other.
Damour and Henneaux have shown under the Gaussian coordinate conditions that a generic cosmological singularity of low energy effective superstring theory is the Mixmaster . Their model is closely related to our model in the sense that it contains the matter fields with exponential potentials. However, there are some different points beside the gauge conditions : (i) the spacetime dimension (it is 11 or 10 in Ref. ), (ii) an assumption to find formal solutions, and (iii) components of $`p`$-forms (i.e. Maxwell and axion) fields. It is known that the existence of the Mixmaster behavior strongly depends on the spacetime dimension. Also, the contributions of the $`p`$-form fields are neglected in the field equations when formal solutions, i.e. VTD solutions, are constructed. Then, these formal solutions are different from ours since the velocity terms of any fields are not neglected in our analysis. Furthermore, thanks to Gowdy symmetry in our analysis, components of $`p`$-form fields are restricted, that is, the Maxwell field strength can have only four components (see Sec. IV), although there is no such restrictions in Ref. .
To avoid complicated calculation we focused only on four functions among the six functions ($`Z`$, $`X`$, $`\omega `$, $`\chi `$, $`\varphi `$ and $`\kappa `$) in the present paper. For a next step, we would like to investigate the full system where that all of six functions do not vanish. Another subject to consider is asymptotic behavior of $`U(1)`$-symmetric spacetimes which have only one spacelike Killing vector. Although Einstein’s equations for $`U(1)`$-symmetric spacetimes are more complicated than those of $`T^3`$ Gowdy spacetime, the Fuchsian algorithm is applicable to such system since this algorithm is independent on dimension of space. Recently, the Fuchsian algorithm is used for vacuum polarized $`U(1)`$-symmetric spacetimes . Our interest is to know whether non-polarized $`U(1)`$-symmetric and generic cosmological spacetimes in the EMDA system are AVTD or not. Although the procedure of the Fuchsian algorithm is routine, finding out plausible formal solutions to VTD equations and obtaining the Fuchsian system are algebraically complicated, and not so easy. Therefore, these subjects are devoted to the future investigation.
Acknowledgment
We are grateful to Akio Hosoya, Hideki Ishihara, Hideo Kodama, Kei-ichi Maeda and Shigeaki Yahikozawa for useful discussions. We also thank Michael Ashworth for his careful reading of our manuscript. This work is partially supported by Scientific Research Fund of the Ministry of Education, Science, Sports, and Culture, by the Grant-in-Aid for JSPS (No. 199704162 (T. T.) and No. 199906147 (K. M.)).
|
warning/0003/hep-ph0003125.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
A characteristic prediction of QCD in the non-perturbative sector concerns the existence of mesons with gluons as valence constituents, either purely gluonic mesons, “glueballs”, or “hybrids”, the bound states of quark, antiquark and gluon.
The search for glueballs is an important issue since the early days of QCD but their very existence has not been definitely established until today. Whereas there is general agreement that the lightest glueball should have quantum numbers $`J^{PC}=0^{++}`$ its mass remains controversial. Moreover, as $`q\overline{q}`$ mesons exist with the same quantum numbers the identification is difficult, in particular, as there could be mixing between the different states as well.
The important way to distinguish between the possibilities is the study of production rates and decay branching ratios which are expected to be sensitive to the constituent structure. Glueballs are expected to be produced preferentially in a gluon rich environment such as $`J/\psi \gamma +X`$, central production in hadron hadron collisions by double Pomeron exchange and $`p\overline{p}`$ annihilation into mesons near threshold. On the other hand, glueball production should be suppressed in $`\gamma \gamma `$ collisions (see for survey). All these reactions have been explored with increasing accuracy in the past.
Here we want to discuss another pair of reactions where glueball or hybrid production could be enhanced or suppressed, respectively. In the fragmentation region of the quark jet the usual $`q\overline{q}`$ hadrons which carry the initial quark as valence quark are enhanced, gluonic mesons are expected to be suppressed. In analogy, one might expect gluonic mesons to be enhanced in the fragmentation region of gluon jets and $`q\overline{q}`$ states to be suppressed. Correspondingly, one may study the relevant particle systems ($`\pi \pi ,K\overline{K},4\pi \mathrm{})`$ either at large momentum fraction (Feynman $`x`$), or with some advantages, in the head of the jet beyond a large rapidity gap (length $`\mathrm{\Delta }y`$). Of advantage are the similar kinematic conditions for both kinds of jets and often the possibility of a measurement in the same experiment. The own identity of gluon jets produced in hard collisions has been established in various details according to the expectations from perturbative QCD calculations (for review, see ). There are plenty of gluon jets in the $`e^+e^{}q\overline{q}g`$ final states and in hard $`pp`$ collisions which makes this type of studies promising.
Experimental studies along this line have not been carried out so far although there is some history to the idea to look for gluonic mesons in gluon jets. Peterson and Walsh suggested a non-perturbative model with a leading isoscalar cluster in the gluon jet which could be composed of ordinary $`q\overline{q}`$mesons like $`\eta ,\eta ^{},\omega ,\varphi `$ or glueballs. An enhanced production of $`\eta ^{}`$ in gluon jets because of its intrinsic $`gg`$ component has been suggested by Fritzsch in the explanation of the large decay rate of $`B`$ into $`\eta ^{}`$. More recently, Roy and Sridhar suggested the search for glueballs in gluon jets at large $`x`$ and considered a model with the $`ggb`$ fragmentation being similar to $`q\pi `$ fragmentation.
Whereas an earlier study at LEP presented evidence for an enhanced $`\eta `$ production in gluon jets at large $`x`$ the more recent studies of $`\eta ,\eta ^{}`$ mesons did not indicate any enhanced production of isoscalar mesons in gluon jets. No study has been published yet for glueball or hybrid candidates.
Whether the suggested systematics is realized in nature depends on the (nonperturbative) colour neutralization mechanism and could be quite different for ordinary $`q\overline{q}`$mesons, hybrids and glueballs. We distinguish the colour triplet and colour octet neutralization and suggest studies which should reveal the relative importance of both mechanisms independently of the existence of glueballs. In any case, we expect the enhanced production of hybrids in gluon jets – if they really exist (see also ).
## 2 Status of glueballs and hybrids
Lattice calculations in quenched approximation (without sea quarks) put the lightest glueball with quantum numbers $`J^{PC}=0^{++}`$ near a mass of 1600 MeV (for a review, see ). First results from unquenched calculations are not yet conclusive but may indicate a decrease of the glueball mass with decreasing quark mass.
On the other hand, within the QCD sum rule approach, a lighter gluonic component in the $`0^{++}`$ spectrum is required near 1000 MeV (earlier results in ). Also the early calculations in the MIT bag model prefer the low mass glueball around 1000 MeV.
In a recent phenomenological analysis it has been proposed that the lightest $`0^{++}`$ glueball should be identified with the very broad state
$$gb(0^{++}):M=1000\mathrm{MeV},\mathrm{\Gamma }=5001000\mathrm{MeV}$$
(1)
which corresponds to both states $`f_0(4001200)`$ and $`f_0(1370)`$ listed in the particle data table . Whereas in some experimental mass spectra there is a separate peak near 1370 MeV the phase shift analyses of elastic and inelastic $`\pi \pi `$ scattering do not indicate the phase movement corresponding to an ordinary Breit Wigner resonance centered at this mass value. Rather – after removal of the narrow $`f_0(980)`$ – the $`\pi \pi `$ elastic scattering phase shifts below 1500 MeV are slowly rising and pass through $`90^\mathrm{o}`$ near 1000 MeV; there is no indication of a second resonance – neither in the elastic nor in the inelastic channels .
Recent data on central production of particle pairs by the WA102 and GAMS collaborations show again peaks in the mass spectra near $`f_0(1370)`$; the measured angular distribution moments should now allow a judgement in favour of the Breit-Wigner resonance phase movement or against it from the interference with $`f_2(1270)`$; the same applies also to peripheral $`\pi ^0\pi ^0`$ production in $`\pi p`$ production by the E852 collaboration . Relative phases may also be obtained in the $`4\pi `$ final states measured by WA102 . Such analyses have the potential to further clarify whether $`f_0(1370)`$ is a reasonably narrow ($`\mathrm{\Gamma }200300`$ MeV) Breit-Wigner resonance or – as we suggested – a component of a yet broader object.
The broad state $`gb(1000)`$ in (1) is found consistent with most expectations for production and decay of a glueball and is therefore a respectable candidate. The main decay mode is $`\pi \pi `$, but also $`K\overline{K}`$ and $`\eta \eta `$ above the respective thresholds and possibly $`\pi \pi e^+e^{}`$ . The identification of $`f_0(1500)`$ as lightest glueball is not supported by the finding of opposite signs for the decay amplitude into $`K\overline{K}`$ and $`\eta \eta `$ .
There is even less experimental information about $`0^+`$ and $`2^{++}`$ glueballs. Possible candidates according to the analysis of ref. are $`\eta (1440)`$ and $`f_J(1710)`$ with $`J=2`$.
In the sector of hybrid mesons the spin-exotic state $`1^+`$ is of particular interest. Lattice results yield mass values around 1900 MeV for this hybrid composed of light quarks and gluon in quenched approximation and also in the first attempt including sea quarks .
There is experimental evidence for resonances with exotic $`1^+`$ quantum numbers in the $`\eta \pi `$ system at masses around 1400 and 1600 MeV (for a recent summary, see Chung ). This result is lower than the lattice expectation by around 500 MeV. This discrepency could be due to an incomplete calculation or the observed states are not hybrid but other exotic states, for example $`q\overline{q}q\overline{q}`$ states.
## 3 Hadron production and colour neutralization
An energetic quark or gluon emerging from a hard collision process will generate a parton cascade by subsequent gluon bremsstrahlung and quark pair production. The extension of such a cascade for a 100 GeV jet in space can be rather large and exceed 100 f (see, for example ). The formation of colour singlet systems should proceed during the evolution whenever the separation of colour charges exceeds the confinement length $`R_c1`$f. Two types of neutralization processes are possible (see Fig. 1).
1. Colour triplet neutralization.
Consider the production of a massive $`q\overline{q}`$ pair in a colour singlet state in its restframe corresponding to separating colour triplet charges (possibly accompanied by secondary partons from bremsstrahlung). The colour field between the primary quarks can be neutralized eventually by the (nonperturbative) production of a soft quark antiquark pair. This process of triplet neutralization may repeat itself for the various branchings inside the quark gluon cascade until the total energy is carried by $`q\overline{q}`$ hadrons. Also possible is the formation of hybrid mesons from the gluons at the end of the perturbative cascade and the soft $`q\overline{q}`$. An example for a quantitative description of soft nonperturbative $`q\overline{q}`$ production is given by the chromoelectric flux tube model which considers pair production in a static field. A systematic treatment for relativistic particles connecting to perturbation theory has not yet been achieved.
2. Colour octet neutralization.
In case of a primary colour singlet $`gg`$ pair the field between the separating colour octet charges can be neutralized at the confinement distance either by the production of a gluon pair or by the subsequent production of two quark pairs. Only the mechanism with gluon pair production could yield pure gluonic bound states at the end.
We introduce the relative probability of triplet and octet neutralization by a phenomenological parameter
$$R_3=P_3/P_8.$$
(2)
In the flux tube model of ref. the octet process was considered negligible assuming a large effective gluon mass $`>1`$ GeV in expectation of a heavy glueball. We consider here the possibility of a light glueball and also of a non-negligable soft $`gg`$ production rate.
It is not obvious to what extent the two types of neutralization mechanisms are realized in a particular process. Common collision processes at particle accelerators are initiated by quarks either as constituents of external hadrons or after production through electromagnetic or weak gauge bosons. A successive triplet neutralization by quark pairs is then always possible even in a gluon dominated final state. The octet neutralization could be an overall rare process enhanced for particular kinematic configurations.
Here we suggest such a configuration: consider the production of a hard gluon which travels without gluon radiation for a while forming a jet with a large rapidity gap empty of hadrons. The probability for such events decreases exponentially with the rapidity gap according to the Sudakov formfactor . In this case the hard isolated gluon builds up an octet field to the remaining partons which is not distorted by multiple gluon emissions and related colour neutralization processes of smaller rapidity range, so the $`gg`$ octet mechanism may be enhanced and will become clearly visible if it exists.
The experimental test can be carried out independently of the existence of glueballs. In case of $`gg`$ neutralization the total charge of the head particles beyond the gap should be $`Q=0`$; on the other hand, the charge distribution in case of double triplet neutralization should have a component with charges $`Q=\pm 1`$, a situation which is also met for an equal mixture of quark and antiquark jets with a gap. Therefore, if the colour octet mechanism exists, the charge distribution for the gluon gap events should be a mixture of both components according to the probability $`R_3`$ in (2).
In this test the gap $`\mathrm{\Delta }y`$ should be large enough so that the leading charges have approached a limiting distribution; only the charges which correspond to the combination of the leading parton and the soft neutralizing parton (or partons) are present. In this limit multiple exchanges through the gap and leading charges $`|Q|2`$ should be absent.
If the enhanced neutral component from octet neutralization is not observed we would not expect a preferred source of glueballs in gluon jets either. In this case glueballs would be produced through their mixing with quarkonium states in any collision process and not preferentially through their valence glue component.
## 4 Limiting charge distribution beyond the gap
For illustration we give an estimate of the limiting charge distribution for sufficiently large $`\mathrm{\Delta }y`$ in the head of the jet beyond the gap for the case of triplet neutralization. We assume that the charge is determined in the quark jet by the charge of the leading quark and an average soft antiquark and in case of gluon jet by the average charge of the soft quarks and antiquarks. In our estimate we assume that soft $`u,d,s`$ quarks are produced with relative probabilities $`\gamma ,\gamma ,\gamma _s`$ and we take $`\gamma =0.4,\gamma _s=12\gamma =0.2`$ as in ref. in the example (i.e. the production ratio $`s/u=0.5`$). More generally, $`\gamma `$ could depend on the energy scale of the separating partons and we compare also with the extreme cases $`\gamma =1/2(s/u=0)`$ and $`\gamma =1/3(s/u=1)`$.
The charge distribution in the head of the gluon jet is then given by
$$p_g(Q)=P_3p_3^g(Q)+P_8p_8^g(Q)$$
(3)
where the charge distributions for triplet and octet neutralization $`p_3^g(Q)`$ and $`p_8^g(Q)`$ are given in Table 1. As seen from the table, the probabilities vary within 10% for the different assumptions on the strange quark ratio $`s/u`$. From the measurement of the charge distribution (3) one can determine the ratio $`R_3`$ in (2), given the parameter $`\gamma `$.
The corresponding analysis can be done as well in quark jets with a gap to check the overall picture and study the parameter $`\gamma `$. The charge distribution in the head of the jet for different quarks and the quark-antiquark average relevant to $`e^+e^{}`$ events is given in Table 2. For a superposition of different quark (anti-quark) jets with probabilities $`\widehat{P}_q`$ the charge distribution is given by
$$\widehat{P}_{<q>}(Q)=\underset{q}{}\widehat{P}_qp_3^q(Q).$$
(4)
There are some differences between the gluon jet with triplet neutralization and the $`q\overline{q}`$ average jet, but both are clearly separated from the octet case which can therefore easily be identified with this method.
## 5 Fragmentation Region of Gluon and Quark Jets
The inclusive spectra of hadrons and other observables are well described by the corresponding quantities at the parton level provided the perturbative evolution is continued down to small values of the $`k_T`$ cutoff in the cascade $`Q_0\mathrm{\Lambda }_{QCD}`$ (this phenomenological approach is called “local parton hadron duality”, for reviews, see ). In this picture the $`x`$-distribution of hadrons in the gluon jet is steeper than in the quark jet because of the stronger semi-soft gluon radiation in the gluon jet.
As to the fragmentation region of the quark jet it is by now well established that the hadrons which carry the primary quark as a valence quark dominate at large momentum fraction $`x`$ as expected in the parton model . The leading hadron follows the distribution of the primary quark if it is color-neutralized by a soft anti-quark.
In the gluon jet, in case of triplet neutralization, the leading particles are the hadrons formed by the primary gluon and the soft $`q\overline{q}`$ pairs. If there are hybrid mesons they may be formed in the fast colour singlet $`q\overline{q}g`$ system, alternatively, the fast hadrons are ordinary $`q\overline{q}`$ mesons formed at the end of the parton cascade after having absorbed all gluon energy. If the octet mechanism is at work the leading gluon may also form a glueball. On the other hand, in the quark jet neither the hybrid nor the glueball will be leading if – as we assume – the leading quark is only neutralized in colour by a soft antiquark. These properties are summarized in Table 3.
The constituent nature of a particular particle or resonance according to Table 3 can best be studied by comparing its production rate in the fragmentation region of quark and gluon jets. One possibility is the study of the respective mass spectra at large $`x`$ (see also ); there may be considerable non-resonant background. Another possibility is the study of particles beyond the rapidity gap. For large gaps when the limiting charge distribution is reached the background should be small. As argued above the rapidity cut may enhance the possibility for the octet mechanism.
The comparison between quark and gluon jet is best carried out for similar kinematic configuration: the same length $`\delta y`$ for the head of the jet beyond the gap in a frame with the nearest jets in angle $`90^\mathrm{o}`$.
For the gluon jet in $`e^+e^{}3`$ jets such a preferred frame is reached after transforming first into the restframe of $`q\overline{q}`$ and then boosting along the $`q\overline{q}`$ direction until the $`g`$ jet is perpendicular (see Fig. 2).
In order to correct for the different inclusive parton distributions in quark and gluon jets and also in case of different available phase space it is best to normalize the rates to a well known $`q\overline{q}`$ hadron $`h_{norm}`$ ($`\rho ,f_2\mathrm{}`$) of comparable mass. Along this way we define the gluon factor $`F_g`$ for a hadron $`h`$
$$F_g(h/h_{norm})=\frac{\sigma (gh)/\sigma (gh_{norm})}{\sigma (qh)/\sigma (qh_{norm})}.$$
(5)
Apparently, for a $`q\overline{q}`$ hadron the gluon factor $`F_g1`$ and for a gluonic meson $`F_g1`$.<sup>2</sup><sup>2</sup>2 This quantity may be compared to the observable called “stickiness” , the ratio of radiative $`J/\psi `$ decay and $`\gamma \gamma `$ production for a given hadron. In that case the very different phase space and barrier factors in both processes are taken into account in the definition using an approximation appropriate for small masses; for high masses this approximation becomes unreliable and yields rather large unrealistic ratios.
According to our hypothesis about $`gb(1000)`$ as being the lightest binary glueball an effect $`F_g>1`$ should be seen for the mass spectrum of $`\pi ^+\pi ^{}`$ or $`\pi ^0\pi ^0`$ around and below 1 GeV, i.e. in the simplest case already for a neutral pair of charged particles. It may be compared with $`\rho (770),f_2(1270)`$. Other enigmatic states like the $`f_0(980)`$ can be tested for their gluonic components as well. Heavier glueball candidates like $`f_0(1500)`$ and $`f_J(1710)`$ may be studied through their $`K\overline{K}`$ decay modes. Also the mass spectrum of all particles with total charge $`Q=0`$ in the head of the gluon jet beyond the rapidity gap would be an interesting possibility for inclusive glueball hunting.
## 6 Conclusions
The limiting charge distribution in the leading cluster of quark and gluon jets with sufficiently large rapidity gap should reveal the relative importance of the octet neutralization mechanism. The comparison of such events in quark and gluon jets could then provide some clues about the constituent nature of candidate gluonic mesons with low background. Inclusive particle spectra did not prove any unusual behaviour in both kinds of jets so far, but particle correlations could do. Interesting tests are already possible with charged particle pairs.
|
warning/0003/cond-mat0003259.html
|
ar5iv
|
text
|
# Untitled Document
Macroscopic Evolution of Particle Systems
with Short and Long Range Interactions
Giambattista Giacomin <sup>1</sup> Dipartimento di Matematica, Università di Milano, via C. Saldini 50, 20133 MI, Italy. E–mail: giambattista.giacomin@mat.unimi.it, Joel L. Lebowitz <sup>2</sup> Departments of Mathematics and Physics, Hill Center. 110 Frelinghuysen Rd, Rutgers University, Piscataway, NJ 088954–8019, USA. E–mail: lebowitz@math.rutgers.edu and Rossana Marra <sup>3</sup> Dipartimento di Fisica and Unità INFM, Università di Roma Tor Vergata, Via della Ricerca Scientifica, 00133 Roma, Italy. E–mail: rossana.marra@roma2.infn.it
Abstract: We consider a lattice gas with general short range interactions and a Kac potential $`J_\gamma (r)`$ of range $`\gamma ^1`$, $`\gamma >0`$, evolving via particles hopping to nearest neighbor empty sites with rates which satisfy detailed balance with respect to the equilibrium measure. Scaling space like $`\gamma ^1`$ and time like $`\gamma ^2`$, we prove that in the limit $`\gamma 0`$ the macroscopic density profile $`\rho (r,t)`$ satisfies the equation
$$\frac{}{t}\rho (r,t)=[\sigma _s\left(\rho \right)\frac{\delta \left(\rho \right)}{\delta \rho \left(r\right)}].()$$
Here $`\sigma _s(\rho )`$ is the mobility of the reference system, the one with $`J0`$, and $`(\rho )=[f_s(\rho (r))\frac{1}{2}\rho (r)J(rr^{})\rho (r^{})drdr^{}]`$, where $`f_s(\rho )`$ is the (strictly convex) free energy density of the reference system. Beside a regularity condition on $`J`$, the only requirement for this result is that the reference system satisfy the hypotheses of the Varadhan–Yau Theorem \[VY\] leading to ($``$) for $`J0`$. Therefore ($``$) holds also if $``$ achieves its minimum on non constant density profiles and this includes the cases in which phase segregation occurs. Using the same techniques we also derive hydrodynamic equations for the densities of a two component A-B mixture with long range repulsive interactions between A and B particles. The equations for the densities $`\rho _A`$ and $`\rho _B`$ are of the form ($``$). They describe, at low temperatures, the demixing transition in which segregation takes place via vacancies, i.e. jumps to empty sites. In the limit of very few vacancies the problem becomes similar to phase segregation in a continuum system in the so called incompressible limit \[EP\], \[DG\].
2000 MSC: 82C22, 60K35
Key Words: Interacting particle systems, Kac potential, hydrodynamic limits, phase segregation, non–gradient particle models, Einstein relation, vacancy dynamics.
The state of a (one component) macroscopic system in equilibrium can be characterized by two numbers, the temperature $`T(=\beta ^1)`$ and the chemical potential $`\lambda `$. When $`T`$ and $`\lambda `$ correspond to a single phase (i.e. there is a unique Gibbs measure) then the particle density $`\overline{\rho }(T,\lambda )`$ is constant, i.e. spatially uniform. Given a microscopic dynamics for which this Gibbs measure is attractive, a disturbance in this uniform density corresponding to a profile $`\rho _0(r)`$, $`r`$ the space coordinate, at some time $`t_0`$ is expected to relax towards the uniform density $`\overline{\rho }`$. In certain types of systems (when the variations in temperature and hydrodynamical flows can be neglected, see below) the relaxation of the density profile $`\rho (t,r)`$ will occur via the diffusion equation
$$\frac{\rho }{t}=\left[D\rho \right],t>t_0,$$
$`(1.1)`$
where the bulk diffusion constant $`D=D(\rho )`$ will generally depend also on the temperature $`T`$, assumed constant and therefore omitted in the notation.
Equation (1.1) is a continuity equation for $`\rho `$ corresponding to a mass current given by Fick’s law
$$j=D(\rho )\rho .$$
$`(1.2)`$
To derive (1.1) from microscopic models it is most convenient to write Fick’s law in its Onsagerian form,
$$j=\sigma \lambda ,$$
$`(1.3)`$
where $`\sigma `$ is the conductivity, or mobility, and $`\lambda (\rho )`$ is the local chemical potential; $`\lambda `$ and $`\rho `$ are related as in the uniform equilibrium system, i.e. we are in a situation of local equilibrium. Comparing (1.2) and (1.3) gives
$$D=\sigma /\chi $$
$`(1.4),`$
where
$$\chi (\rho )=\left(\frac{\lambda }{\rho }\right)^1=\rho \frac{\rho }{p}$$
$`(1.5)`$
is the compressibility ($`p`$ being the pressure). In (1.5) we have expressed the chemical potential at equilibrium as a function of the density $`\rho `$. The relation (1.4) is sometimes referred to as the Einstein relation who first used it to relate the diffusion constant of a Brownian particle to its steady state mobility in an external field \[E, Sp\].
Mathematically rigorous derivations of (1.2) – (1.5), with $`\sigma `$ given by a Green–Kubo formula, have been achieved, via the use of the hydrodynamical (diffusive) scaling limit for a variety of microscopic models with fixed short (microscopic) range interactions evolving via stochastic dynamics (in which particle number is the only conserved quantity), see \[KL\]<sup>(1)</sup> There also exists a derivation for one system with a Hamiltonian evolution, namely a non–interacting gas of point particles moving among fixed periodic scatterers – the Sinai billiard with finite horizons \[BS\], \[ Sp\].. In all these cases the temperature $`T`$ is in the uniqueness region of the phase diagram, i.e. there is a unique phase for all values of $`\lambda `$ (or $`\rho `$).
The situation becomes much trickier when we consider temperatures where there is, for some value of $`\lambda `$, a coexistence of phases with two (or more) different densities, $`\rho _1`$ and $`\rho _2`$, corresponding to liquid and vapor or to fluid and solid. In such cases the macroscopic equilibrium system, with a fixed total number of particles corresponding to an average density in the interval $`(\rho _1,\rho _2)`$, will not have a uniform density. Instead it will be segregated into macroscopic regions of density $`\rho _1`$ and $`\rho _2`$ with shapes minimizing the free energy of the surface between them. An equation of form (1.1) is clearly not appropriate now, in fact, for any density $`\rho (\rho _1,\rho _2)`$, $`\chi (\rho )`$ in (1.4) will be infinite and therefore $`D`$ is formally zero (unless $`\sigma `$ is also infinite, which can be proven in some cases not to happen). Results in these directions have been proven for very particular systems, see the remarkable result in \[R\] where degenerate diffusion in the coexistence region is proven, but the general case of the evolution of phase domains for systems with interactions is a real challenge for the moment.
To get around this obstacle and derive macroscopic equations for a system undergoing phase segregation, some authors studied the time evolution of the macroscopic density profile in particle systems interacting via long range (compared to the interparticle spacing) Kac potentials, \[LOP\], \[GL1\]. The microscopic model considered in \[GL1\] is a lattice gas evolving under a particles hopping (Kawasaki exchange) dynamics which satisfies detailed balance (is reversible) with respect to the Gibbs canonical (fixed particle number) equilibrium measure with Hamiltonian $`H`$ at temperature $`\beta ^1`$. $`H`$ consisted of a sum of two terms, a short range part $`H_s`$ which may be thought of as a nearest neighbor interaction and a Kac potential $`H_\gamma `$ characterized by a range parameter $`\gamma ^1`$, namely
$$H_\gamma =\frac{1}{2}\underset{x,y}{}\gamma ^dJ(\gamma |xy|)\eta (x)\eta (y),$$
$`(1.6)`$
where $`x,y\mathrm{\Lambda }_\gamma Z^d`$ (say a torus of diameter $`[\gamma ^1]`$) and $`\eta (x)=0,1`$ specifies the occupancy of site $`x`$. They argued that in the diffusive hydrodynamical scaling limit, corresponding to $`\gamma 0`$ with space and time scaling like $`\gamma ^1`$ and $`\gamma ^2`$ respectively, the density profile $`\rho (r,t)`$ would satisfy a parabolic integro–differential equation of the form,
$$\frac{\rho }{t}=\left[\sigma _s(\rho )\frac{\delta (\rho )}{\delta \rho }\right]\left[\sigma _s(\rho )\lambda (\rho )\right],$$
$`(1.7)`$
where
$$=f_s(\rho (r))𝑑r\frac{1}{2}J(rr^{})\rho (r)\rho (r^{})𝑑r𝑑r^{}$$
$`(1.8)`$
$`f_s`$ is the free energy density and $`\sigma _s`$ is the mobility of the system with only short range interactions.
Equations (1.5) and (1.7) were proven in \[GL1\] for the case where the short range interactions $`H_s`$ consisted only of the hard core exclusion, i.e. no more than one particle per site, in which case $`f_s`$ and $`\sigma _s`$ reduce to $`f_s^0`$ and $`\sigma _s^0`$, with
$$f_s^0(\rho )=\beta [\rho \mathrm{log}\rho +(1\rho )\mathrm{log}(1\rho )],$$
$`(1.9)`$
and
$$\sigma _s^0(\rho )=\beta \rho (1\rho ).$$
$`(1.10)`$
The validity of (1.7) and (1.8) for the case of nontrivial short range interactions was conjectured in \[§3,GL1\] for values of $`\beta ^1`$ above the critical temperature of the reference (short range) system, when $`f_s(\rho )`$ is a strictly convex function of $`\rho `$. They gave heuristic arguments for the validity of (1.7) with $`\lambda `$ the local chemical potential of the system with the Kac potential, but with the mobility $`\sigma `$ being the same as what it would be in the system without the slowly varying Kac potential. This generalized a similar conjecture by Spohn \[Sp\] for the case of an external long range interaction. Besides the case in \[GL1\], the conjecture was shown to be valid also for some other special cases \[BL\],\[MM\],\[AX\]. In this paper we prove these conjectures for the general short range interactions in the case of Ising spins, that is in the case of a system with Hamiltonian $`H^\gamma =H_s+H_\gamma `$. In fact the methods we use here extend in a straightforward manner to general systems for which a diffusion equation can be proven in the absence of $`H_\gamma `$. We illustrate this by deriving in Section 5 integro–differential equations for a binary mixture which may undergo a demixing transition. Our results extend to the case of a system on which a weak external force is acting, characterized by the Hamiltonian
$$H_s(\eta )+\underset{x}{}V(\gamma x)\eta (x)$$
and also to the case of a weak (of order $`\gamma `$) external force which is not the gradient of a potential, like a costant force in a torus.
We note that solutions to (1.7) corresponding to interface dynamics are considered in \[GL2\]. There is also a review of various results about these models, including the case in which there is no conservation law (Glauber dynamics) \[GLP\].
Informal Description of Model. To be concrete, we restrict ourselves to the case $`H_s(\eta )=K_{x,y:|xy|=1}\eta (x)\eta (y)`$. The Kawasaki dynamics is defined in terms of Poisson jump rates depending on the energy differences. An example of such rates is
$$c_{(x,y)}^\gamma (\eta )=\mathrm{exp}\left(\beta \left[H^\gamma (\eta ^{x,y})H^\gamma (\eta )\right]/2\right),$$
$`(1.11)`$
where $`x,y`$ are nearest neighbor sites and $`\eta ^{x,y}`$ is the configuration in which the sites $`x`$ and $`y`$ exchange their occupation numbers. The microscopic current $`w_{x,y}`$ through the bond $`(x,y)`$ is the rate at which a particle jumps from $`x`$ to $`y`$ minus the rate at which a particle jumps from $`y`$ to $`x`$, namely
$$w_{x,y}^\gamma (\eta )=\left[\eta (x)c_{(x,y)}^\gamma (\eta )\eta (y)c_{(x,y)}^\gamma (\eta )\right].$$
$`(1.12)`$
Equation (1.12) will determine the form of the macroscopic current in the hydrodynamic equation.
A key ingredient in our analysis is that the dynamics with $`J0`$ is a weak perturbation of the $`J0`$ dynamics. This can be seen both at the level of the rates and of the current. In particular for the current some straightforward expansions with respect to the small parameter $`\gamma `$ (see Section 4) lead to
$$\begin{array}{cc}\hfill w_{x,x+e_i}^\gamma (\eta )=& c_{x,x+e_i}^0(\eta )\left[\eta (x)\eta (x+e_i)\right]\hfill \\ & +\frac{\gamma \beta }{2}c_{x,x+e_i}^0\left[\eta (x)\eta (x+e_i)\right]^2\left(_iJ\eta \right)(x)+O(\gamma ^2),\hfill \end{array}$$
$`(1.13)`$
where $`e_i`$ is the unit vector on the lattice in the direction $`i`$, $``$ denotes spatial discrete convolution and the superscript $`0`$ on $`c`$ and $`w`$ denotes the case in which there is no long range force ($`J0`$). Equation (1.13) indicates clearly that the dynamics with $`J0`$ is a weak perturbation of the reference ($`J0`$) dynamics. The case studied in \[GL1\] corresponds to $`c_{x,y}^0=1`$ in (1.13). The local equilibrium expectation of $`w_{x,x+e_i}^0`$ gives the macroscopic flux $`j`$ for this simplified model in the form of a term due to the exclusion dynamics and one due to the mean field force:
$$j=\rho +\sigma (\rho )J\rho ,$$
$`(1.14)`$
where $``$ denotes the (standard) spatial convolution, $`J\rho (r)=J(rr^{})\rho (r^{})dr^{}`$, and the mobility $`\sigma (\rho =\beta \rho (1\rho )`$ is just the expectation of $`(\beta /2)\left[\eta (x)\eta (x+e_i)\right]^2`$.
The naive extension of the above argument to the general case, i.e. just applying local equilibrium to (1.13), would give a macroscopic current of the form
$$j=D\rho +\stackrel{~}{\sigma }(\rho )J\rho ,$$
$`(1.15)`$
where $`D`$ is, as before, the diffusion coefficient found in \[VY\], but $`\stackrel{~}{\sigma }`$ would not satisfy the Einstein relation (1.4). This apparent contradiction is explained by the fact that the microscopic current of the reference system in the general case, i.e. the first term in the right–hand side of (1.13), is in not a lattice gradient of a function of the configuration, i.e. there is no function $`h`$ such that $`w_{x,y}^0(\eta )=h(\tau _y\eta )h(\tau _x\eta )`$, where $`\tau _x`$ is the translation in configuration space, unless $`K=0`$ (i.e. $`H_s0`$). The non gradient nature of the dynamics is responsible for the presence in (1.15) of a third term which, combined with the second term, gives the correct expression for the macroscopic current.
The main ingredient in our derivation is the recent work of Varadhan and Yau \[VY\] who proved (1.1) for a lattice gas with general short range interactions at small $`\beta `$. This is a major achievement since previous derivations all required that the dynamics be either of gradient type, or that the invariant measure be of product type, i.e. independent occupation values at different sites, see \[KL\] for a complete treatment and extended bibliography on this topic. The class of non–gradient models with invariant measures of product type includes the $`n`$–color simple exclusion process, which is a standard simple exclusion process but each particle is colored with one of $`n`$ possible colors. Exchanges only occur between occupied and empty sites, so that two sites with different colors cannot exchange. Our arguments do apply to this case too. This can be viewed as a multi–species system and the effect of long range interactions is of interest in its own right and we discuss it in Section 5 of this paper, limiting ourselves to the $`n=2`$ case, corresponding to a binary alloy in which exchanges take place only through vacancies.
The paper is organized as follows: in section 2 we give the precise definition of our model and we state the main results. In section 3 we recall some important results of \[VY\] that we need for our proof. Section 4 contains the proofs which are based on the control of the Radon-Nykodim derivative of the full process with respect to the reference process and on the \[GPV\]–\[V\] method. Finally in section 5 we consider the example, that we just mentioned, of a dynamics with two conservation laws, for which we prove a result analogous to the one in section 2. Moreover, in this case we get a stronger convergence result, since uniqueness of the weak solution for the limit equation does hold, under suitable hypotheses.
2. The model and the main result.
We work in the discrete torus of dimension $`d`$ and diameter $`\gamma ^1`$, $`\gamma >0`$, that we denote by $`\mathrm{\Lambda }_\gamma `$. Associated to $`\mathrm{\Lambda }_\gamma `$ there is a natural notion of $`\mathrm{\Lambda }_\gamma ^{}`$, the set of (non directed) bonds, i.e. the couples of nearest neighbor points of $`\mathrm{\Lambda }_\gamma `$. The configuration space on $`\mathrm{\Lambda }_\gamma `$ is $`\{0,1\}^{\mathrm{\Lambda }_\gamma }\mathrm{\Omega }_{\mathrm{\Lambda }_\gamma }`$. All these spaces are endowed with the discrete topology: when later we will deal with $`\{0,1\}^\mathrm{\Lambda }`$, $`\mathrm{\Lambda }`$ countable, we will use the product topology.
2.1 The Gibbsian reference measure. Let $`F`$ be a local isotropic function from $`\mathrm{\Omega }_{\mathrm{ZZ}^d}`$ to $`\mathrm{IR}`$. By isotropic we mean that $`F(\theta \eta )=F(\eta )`$ for every reflection $`\theta `$ with respect of an axis as well any lattice rotation $`\theta `$. For any $`x\mathrm{ZZ}^d`$ set $`F_x(\eta )=F(\tau _x\eta )`$, where $`\tau _x`$ is the translation of $`x\mathrm{ZZ}^d`$ in $`\mathrm{\Omega }_{\mathrm{ZZ}^d}`$:
$$\tau _x\eta (y)=\eta (x+y),\mathrm{for}\mathrm{every}y\mathrm{ZZ}^d.$$
And $`\tau `$ will also be the translation operator acting on functions $`F`$ of the configuration: $`\tau _xF(\eta )=F(\tau _x\eta )`$.
Let us consider the formal Hamiltonian
$$H(\eta )=\underset{x}{}F_x(\eta ).$$
$`(2.1)`$
It is well know that if the inverse temperature $`\kappa `$ is sufficiently small, given any chemical potential $`\lambda `$, there exists a unique extremal Gibbs measure $`\mu ^\rho `$, $`\rho [0,1]`$, which satisfies $`\rho =\mathrm{IE}^{\mu ^\rho }[\eta (x)]`$ for every $`x\mathrm{ZZ}^d`$. We will always work in this uniqueness regime: this guaranties also that $`\mu ^\rho `$ is translation invariant and that it has some mixing properties. However in some proofs in \[VY\] the authors require a stronger mixing condition (exponential mixing condition, \[Assumption A, VY\]), given in terms of finite volume grand–canonical measures. Since we are using this assumption only indirectly we do not give it explicitly: we just stress that our results are proven only if $`\kappa `$ is smaller then a certain $`\kappa _0>0`$, which depends on the dimension $`d`$ and on the interaction $`F`$. In some cases it can be shown that $`\kappa _0=\kappa _c`$, the inverse of the critical temperature of the reference system.
We will deal with the $`F`$–interaction also in the case of $`\mathrm{\Lambda }_\gamma `$: for $`\gamma `$ sufficiently small we can keep the very same infinite volume definitions by lifting $`\mathrm{\Lambda }_\gamma `$ to the whole of $`\mathrm{ZZ}^d`$ in the natural way. This way the notion of translation $`\tau _x`$ is unchanged. It is however often more natural to look at the periodic case as a finite volume case, and that’s what we will do. The same applies to every local function on $`\mathrm{\Omega }_{𝐙^d}`$, which will be viewed as a function on $`\mathrm{\Omega }_{\mathrm{\Lambda }_\gamma }`$ without notational changes.
2.2 The equilibrium measure for the full system. We consider a the probability measure on $`\mathrm{\Omega }_{\mathrm{\Lambda }_\gamma }`$ defined by
$$\mu _\gamma ^{\beta ,\kappa ,\lambda }(\eta )=\frac{1}{Z_\gamma (\beta ,\kappa ,\lambda )}\mathrm{exp}\left\{\kappa \underset{x}{}F_x\frac{\beta }{2}\underset{x,y}{}J_\gamma (x,y)\eta (x)\eta (y)+\lambda \underset{x}{}\eta (x)\right\},$$
$`(2.2)`$
$$\beta >0,\gamma >0,\lambda 𝐑,$$
where $`J_\gamma (x,y)=\gamma ^dJ(\gamma (xy))`$, $`JC^2(\mathrm{IT}^d)\mathrm{IR}`$ ($`\mathrm{IT}^d`$ is the d-dimensional torus of diameter $`1`$) such that
$$J(r)=J(r),_{\mathrm{IT}^d}J(r)dr=1,$$
$`(2.3)`$
and $`Z_\gamma (\beta ,\kappa ,\lambda )`$ is the normalization factor.
2.3 The dynamics. This will be introduced, for the reference system, both in the case $`\mathrm{\Lambda }=\mathrm{\Lambda }_\gamma `$ and in the case $`\mathrm{\Lambda }=𝐙^d`$. Let $`\mathrm{\Phi }C^2(\mathrm{IR};\mathrm{IR}^+)`$ be of the form<sup>1</sup> This is equivalent to the more customary \[p. 163, Sp\] detailed balance condition $`\mathrm{\Phi }(E)=\mathrm{\Phi }(E)\mathrm{exp}(E)`$, $`E\mathrm{IR}`$.
$$\begin{array}{cc}& \mathrm{\Phi }(E)=\mathrm{exp}(E/2)\varphi (E),\hfill \\ \hfill \varphi :\mathrm{IR}\mathrm{IR}^+,& \varphi (E)=\varphi (E)\mathrm{for}\mathrm{every}E\mathrm{IR}\mathrm{and}\varphi (0)=1.\hfill \end{array}$$
$`(2.4)`$
For $`b=(x,y)\mathrm{\Lambda }^{}`$ we define
$$c_b^0(\eta )=\mathrm{\Phi }(\kappa \mathrm{\Delta }_bH(\eta )),$$
$`(2.5)`$
where $`\mathrm{\Delta }_bH(\eta )=H(\eta ^b)H(\eta )`$ and
$$\eta ^b(z)=\{\begin{array}{cc}\eta (x),\hfill & \text{if }z=y\text{,}\hfill \\ \eta (y),\hfill & \text{if }z=x\text{,}\hfill \\ \eta (z),\hfill & \text{otherwise.}\hfill \end{array}$$
$`(2.6)`$
In the infinite volume case $`H`$ is not well defined, but $`\mathrm{\Delta }_bH`$ is taken, by definition, to be equal to $`lim_R\mathrm{}\mathrm{\Delta }_b_{|x|R}F_x(\eta )`$. To make the notation a bit lighter, if $`b=(x,y)`$ appears as a subscript, we will often drop the brackets. Given $`f:\mathrm{\Omega }_\mathrm{\Lambda }\mathrm{IR}`$ for $`b=(x,y)`$ we set
$$_b^0=c_b^0(\eta )\left[f(\eta ^b)f(\eta )\right].$$
$`(2.7)`$
A Markov pregenerator is then defined by setting
$$^0f(\eta )=\underset{b\mathrm{\Lambda }^{}}{}_b^0f(\eta ),$$
$`(2.8)`$
where $`f`$ is assumed to be local in the case $`\mathrm{\Lambda }=\mathrm{ZZ}^d`$. If $`\mathrm{\Lambda }=\mathrm{\Lambda }_\gamma `$, $`^0`$ is actually a generator and it is easy to construct a unique process in Skorohod space $`\{\eta _t\}_{t\mathrm{IR}^+}D(\mathrm{IR}^+;\mathrm{\Omega }_{\mathrm{\Lambda }_\gamma })`$ associated to it, once an initial condition is given. The law of this process will be denoted by $`𝐏^{\gamma ,0}`$ or $`𝐏_{\mu _\gamma }^{\gamma ,0}`$ if there is the need to stress the initial condition $`\mu _\gamma `$. It is immediate to verify that, by (2.4), $`^0`$ viewed as an operator in $`L^2(\mu _\gamma ^{0,\kappa ,\lambda })`$ is self–adjoint for every $`\lambda \mathrm{IR}`$. In the infinite volume setting the construction of the process is more delicate. We refer to \[Li\] for this construction: there the process $`\{\eta _t\}_{t\mathrm{IR}^+}D(\mathrm{IR}^+;\mathrm{\Omega }_{\mathrm{ZZ}^d})`$ associated to $`^0`$ is constructed. We remark also that, for every $`\rho [0,1]`$, $`^0`$ can be extended to a self–adjoint operator on $`L^2(\mu ^\rho )`$. The law of the process $`\{\eta _t\}_{t\mathrm{IR}^+}`$ will be denoted by $`𝐏^0`$. We will sometimes stress the chosen initial condition, say $`\mu 𝒫_1(\mathrm{\Omega }_\mathrm{\Lambda })`$, by writing $`𝐏_\mu ^0`$. Here we used $`𝒫_1()`$ to denote the probability measures on $``$.
The full dynamics is considered only in the case $`\mathrm{\Lambda }=\mathrm{\Lambda }_\gamma `$. We define
$$c_b^\gamma (\eta )=\mathrm{\Phi }\left(\mathrm{\Delta }_b\left[\kappa \underset{x}{}F_x\beta \underset{x,y}{}J_\gamma (x,y)\eta (x)\eta (y)\right]\right),$$
$`(2.9)`$
and
$$_\gamma f(\eta )=\underset{b\mathrm{\Lambda }^{}}{}c_b^\gamma (\eta )\left[f(\eta ^b)f(\eta )\right].$$
$`(2.10)`$
As before, associated to the finite dimensional operator $`_\gamma `$ there is a process with trajectories in $`D(\mathrm{IR}^+;\mathrm{\Omega }_{\mathrm{\Lambda }_\gamma })`$. $`_\gamma `$ is self–adjoint in $`L^2(\mu _\gamma ^{\beta ,\kappa ,\lambda })`$ for every $`\lambda \mathrm{IR}^+`$. The law of this process will be denoted by $`𝐏^\gamma =𝐏_{\mu _\gamma }^\gamma `$, where $`\mu _\gamma 𝒫(\mathrm{\Omega }_{\mathrm{\Lambda }_\gamma })`$ is the initial condition.
2.4 The main result. The compressibility $`\chi `$ for the system is defined as
$$\chi (\rho )=\underset{x\mathrm{ZZ}^d}{}\mathrm{cov}^{\mu ^\rho }(\eta (0),\eta (x)),$$
$`(2.11)`$
in terms of the local interaction alone. The corresponding diffusion matrix $`D`$ can be expressed via a variational formula. It is the $`\rho `$ dependent symmetric matrix defined by
$$v,Dv_{\mathrm{IR}^d}=\frac{1}{2\chi (\rho )}\underset{g}{inf}\mathrm{IE}^{\mu ^\rho }\left[\underset{i=1}{\overset{d}{}}c_{e_i}^0(\eta )\left(v_i(\eta (e_i)\eta (0))\underset{x\mathrm{ZZ}^d}{}\mathrm{\Delta }_{(0,e_i)}g\left(\tau _x\eta \right)\right)^2\right],$$
$`(2.12)`$
for every $`v\mathrm{IR}^d`$. In (2.12) $`e_i`$ denotes the unit vector in the $`i`$ direction, $`,_{\mathrm{IR}^d}`$ is the scalar product in $`\mathrm{IR}^d`$ and the infimum is taken over all local functions $`g`$. We will take the freedom of using both the notation $`v_i`$ and the notation $`v_{e_i}`$. Moreover for $`e`$ a unit vector, $`v_e=\pm v_{e_i}`$ if $`e=\pm e_i`$.
A number of facts are known about $`D`$: first of all it is a continuous function of $`\rho `$, cf. \[VY\], and there exists a constant $`c`$ such that in the sense of matrices
$$\frac{I}{c\chi (\rho )}D(\rho )cI,$$
$`(2.13)`$
where $`I`$ is the $`d\times d`$ identity matrix. While the upper bound is an immediate consequence of (2.12), the lower bound is much more subtle and it has been established in \[SY\]. In \[VY, Lemma 8.3\] it is shown moreover that in our isotropic set up, $`D(\rho )`$ is a multiple of $`I`$. We will keep the notation $`D(\rho )`$ also to denote the scalar proportionality factor between $`I`$ and the matrix $`D(\rho )`$.
We are going to consider initial particle configurations associated to a density profile $`\rho _0:\mathrm{IT}[0,1]`$, in the following sense: if we define the the empirical density field
$$\nu ^\gamma (t,x)=\gamma ^d\underset{y\mathrm{\Lambda }}{}\delta (x\gamma y)\eta _{\gamma ^2t}(y),$$
$`(2.14)`$
we require that for any smooth test function $`G`$ from $`\mathrm{IT}^d`$ to $`\mathrm{IR}`$ and $`\delta >0`$
$$\underset{\gamma 0}{lim}𝐏_{\mu _\gamma }^\gamma \left(\left|_{\mathrm{IT}^d}\nu ^\gamma (x,0)G(x)𝑑x_{\mathrm{IT}^d}\rho _0(x)G(x)dx\right|>\delta \right)=0.$$
$`(2.15)`$
Denote by $`𝐐^\gamma `$ the law of the process $`\{\nu ^\gamma (t)\}_{t[0,T]}`$ on the space $`D([0,T],)`$, induced by $`𝐏_{\mu _\gamma }^\gamma `$, where $`=(\mathrm{IT}^d)`$ is the space of nonnegative measures on the torus with total mass bounded by $`1`$: this space is endowed with the weak topology weakened by the continuous functions. We consider also the subspace $`_1=_1(\mathrm{IT}^d)`$ of $``$ consisting of absolutely continuous measures with density bounded above by one. Our notation for the gradient in $`d`$–dimensional Euclidean space is $``$.
Theorem 2.1 Consider an initial datum satisfying (2.15). Then, the sequence of probability measures $`𝐐^\gamma `$ is tight and all its limit points $`𝐐`$ are concentrated on absolutely continuous paths whose densities $`\rho C^0([0,T];_1(\mathrm{IT}^d))L^2([0,T],H_1(\mathrm{IT}^d))`$ are weak solutions of the equation
$$\{\begin{array}{cc}& _t\rho =\left\{D(\rho )\left[\rho \beta \chi (\rho )\left(J\rho \right)\right]\right\},\hfill \\ & \rho (0,)=\rho _0().\hfill \end{array}$$
$`(2.16)`$
Moreover if the diffusion coefficient $`D`$ is Lipschitz continuous, then $`𝐐^\gamma `$ converges and the limit point $`𝐐`$ is the unique weak solution of (2.16).
Throughout the text $`\mathrm{\Phi }`$, $`J`$, $`\kappa `$ and $`\beta `$ will be considered fixed. We will be often interested in getting estimates which are uniform on the configuration $`\eta `$ or on the history of the process $`\{\eta _t\}_{t0}`$. We therefore introduce the notation $`o_u()`$ and $`O_u()`$ in the standard sense of $`o()`$ and $`O()`$ but uniformly with respect to the configuration or to the history of the process.
3. The fluctuation–dissipation equation.
In this section we recall the fundamental result proven in \[VY\]: the approximate decomposition of the current in a gradient term and a fluctuation term.
Let us recall the definition of current in the general $`J`$–dependent case. The current is defined for every $`x`$, every unit vector $`e`$ and every $`\eta `$ as
$$w_{x,x+e}^\gamma (\eta )=c_{x,x+e}^\gamma (\eta )[\eta (x)\eta (x+e)].$$
$`(3.1)`$
It has the property that
$$_\gamma \eta (x)=\underset{j=1}{\overset{d}{}}\left[w_{x,x+e_j}^\gamma (\eta )w_{xe_j,x}^\gamma (\eta )\right].$$
$`(3.2)`$
The analogous definitions in the case of the unperturbed dynamics generated by $`^0`$ are just obtained by setting $`\gamma =0`$.
We follow very closely \[VY\]. For $`\zeta \mathrm{\Omega }_{𝐙^d}`$ or $`\zeta :𝐙^d𝐑`$ and $`\mathrm{}\mathrm{IR}^+`$ we set
$$\mathrm{Av}_{\mathrm{}}\zeta (x)=\frac{1}{(2\mathrm{}+1)^d}\underset{y:|yx|\mathrm{}}{}\zeta (y).$$
$`(3.3)`$
We also set $`\mathrm{\Lambda }_{\mathrm{}}^{}=\{x𝐙^d:|x|\mathrm{}\}`$. We define the $`\sigma `$–algebra $`_{x,s}`$ generated by $`\{\mathrm{Av}_s\eta (x)\}\{\eta (y):|yx|>s\}`$ and the space $`𝒢`$ of local functions $`h:\mathrm{\Omega }_{\mathrm{ZZ}^d}\mathrm{IR}`$ with the property $`\mathrm{IE}^\mu [h|_s]=0`$ for some $`s`$. To any $`h𝒢`$ and any $`\eta \mathrm{\Omega }_{\mathrm{ZZ}^d}`$ we associate an element (still denoted by $`h`$) of $`\mathrm{\Omega }_{\mathrm{ZZ}^d}`$ by setting $`h(x)=\tau _xh(\eta )`$. We call $`\mathrm{\Omega }_𝒢`$ the subset of $`\mathrm{\Omega }_{\mathrm{ZZ}^d}`$ obtained from $`𝒢`$ with this procedure. We then introduce the finite volume variance
$$V_{\mathrm{}}(h,\rho ,\xi )=\mathrm{}^d\mathrm{Av}_\mathrm{}_1h,(_\mathrm{\Lambda }_{\mathrm{}}^{}^0)^1\mathrm{Av}_\mathrm{}_1h_{\mu _\mathrm{\Lambda }_{\mathrm{}}^{},\rho ,\xi },$$
$`(3.4)`$
where $`\mathrm{}_1=\mathrm{}\sqrt{\mathrm{}}`$, $`_\mathrm{\Lambda }^0=_{b\mathrm{\Lambda }^{}}_b^0`$ for any finite set $`\mathrm{\Lambda }𝐙^d`$ and $`\mu _{\mathrm{\Lambda }_{\mathrm{}}^{},\rho ,\xi }`$ is the canonical Gibbs measure with interaction $`F`$ on $`\mathrm{\Lambda }_{\mathrm{}}^{}`$ with boundary condition $`\xi \mathrm{\Omega }_{\mathrm{ZZ}^d}`$ and density $`\rho [0,1]`$. Two observations are in order for canonical measures. The first is that we view $`\mu _{\mathrm{\Lambda }_{\mathrm{}}^{},\rho ,\xi }`$ as an element of the probability measures over $`\mathrm{\Omega }_{\mathrm{ZZ}^d}`$: the extension is made in the natural way obtaining a measure concentrated on $`\{\eta :\eta (y)=\xi (y)`$ for every $`y\mathrm{\Lambda }_{\mathrm{}}^{}{}_{}{}^{c}\}`$. Second, if $`\rho `$ is not an integer multiple of $`1/|\mathrm{\Lambda }_{\mathrm{}}^{}|`$, the density of the canonical Gibbs measure is taken to be $`\mathrm{max}\{k/|\mathrm{\Lambda }_{\mathrm{}}^{}|:k\mathrm{ZZ},k\rho |\mathrm{\Lambda }_{\mathrm{}}^{}|\}`$. We observe then that if $`x\mathrm{\Lambda }_\mathrm{}_1^{}`$, $`h(x)`$ depends only on $`\eta (y)`$, $`y\mathrm{\Lambda }_{\mathrm{}}^{}`$, for sufficiently large $`\mathrm{}`$. Moreover $`V_{\mathrm{}}(h,\rho ,\xi )`$ is an $`_{\mathrm{}}`$–measurable function of $`\xi \mathrm{\Omega }`$. Given $`\mu ^\rho `$, the infinite volume Gibbs measure with density $`\rho `$, we define $`V:\mathrm{\Omega }_𝒢\times [0,1]\mathrm{IR}^+`$ by
$$V(h,\rho )=\underset{\mathrm{}\mathrm{}}{lim\; sup}\left(V_{\mathrm{}}(h,\rho ,\xi )\mathrm{d}\mu ^\rho (\xi )\right).$$
$`(3.5)`$
Finally we extend this definition to every local function $`h`$ by setting $`h_k(x)=h(x)\mathrm{IE}^{\mu _{\mathrm{\Lambda }_{\mathrm{}}^{},\rho ,\xi }}[h(x)|_{x,k}]`$ and taking the limit
$$\underset{k\mathrm{}}{lim\; sup}V(h_k,\rho ).$$
$`(3.6)`$
Below we use the notation
$$_{e_j}\eta (x)=\eta (x+e_j)\eta (x).$$
$`(3.7)`$
We are now ready to state the Fluctuation–Dissipation Theorem \[VY, Theorem 3.4\].
Theorem 3.2 The symmetric, in fact diagonal, density dependent matrix $`D`$ defined in (2.12) coincides with the following Green–Kubo formula:
$$\chi (\rho )D_{i,j}(\rho )=\delta _{i,j}\frac{1}{2}\mathrm{IE}^{\mu ^\rho }\left[c_{(0,e_j)}^0\right]_0^{\mathrm{}}\underset{x}{}\mathrm{IE}^{\mu ^\rho }\left[w_{0,e_i}^0\mathrm{exp}\left(^0t\right)\tau _xw_{0,e_j}^0\right]\mathrm{d}t,$$
$`(3.8)`$
where $`\chi `$ is the compressibility defined in (2.11) and $`\delta _{i,j}`$ is the Kronecker delta. Then for any $`\alpha \mathrm{IR}^d`$
$$\underset{h𝒢^d}{inf}V(\underset{j=1}{\overset{d}{}}\alpha _j\left[w_{0,e_j}^0(\eta )+(D(\rho )\eta (0))_j+^0h_j(\eta )\right],\rho )=0.$$
$`(3.9)`$
Moreover, for any $`\delta >0`$ there exists $`g^\delta :[0,1]\times \mathrm{\Omega }_{\mathrm{ZZ}^d}\mathrm{IR}^d`$, $`g^\delta (\rho ,)𝒢^d`$ for every $`\rho `$ and $`g^\delta (,\eta )`$ is smooth for every $`\eta `$, such that
$$\underset{\rho [0,1]}{sup}V(\underset{j}{}\alpha _j\left[w_{0,e_j}^0(\eta )+(D(\rho )\eta (0))_j+^0g_j^\delta \right],\rho )\delta .$$
$`(3.10)`$
We observe that, by polarization, from the bilinear functional $`V(,\rho )`$ we can define a scalar product (covariance) that will be denoted by $`f,g(\rho )`$, defined for $`f`$ and $`g`$ local function on $`\{0,1\}^{\mathrm{ZZ}^d}`$. This covariance is carefully analyzed in Section 8 of \[VY\]. We collect here two properties that will be crucial for us. First, formula (8.7) in \[VY\] tells us that
$$w_{0,e_i}^0,w_{0,e_j}^0(\rho )=\frac{1}{2}\delta _{i,j}𝐄^{\mu ^\rho }\left[c_{0,e_j}(\eta )(_{e_j}\eta (0))^2\right].$$
$`(3.11)`$
Moreover formula (8.13) in \[VY\]) tells us that
$$w_{0,e_i}^0,_{e_j}\eta (0)(\rho )=\delta _{i,j}\chi (\rho ).$$
$`(3.12)`$
4. Proof of Theorem 2.1
4.1. Preliminary lemmas. We will repeatedly need the expansion (in powers of $`\gamma `$) of the jump rates: we take advantage of the fact that an exchange of two particles changes the long–range energy of $`O_u(\gamma )`$. We will use the following notation for discrete convolution
$$(f\eta )(x)=\gamma ^d\underset{z\mathrm{\Lambda }_\gamma }{}f\left(\gamma (xz)\right)\eta (z),$$
$`(4.1)`$
where $`f`$ is a function from $`\mathrm{IT}^d`$ to $`\mathrm{IR}`$ and $`x\mathrm{\Lambda }_\gamma `$.
Lemma 4.3 For every $`\mathrm{\Phi }`$ and $`J`$, there exists $`C`$ such that for every $`b\mathrm{\Lambda }_{\gamma }^{}{}_{}{}^{}`$, every $`\eta \mathrm{\Omega }_{\mathrm{\Lambda }_\gamma }`$ and every $`\gamma (0,1)`$
$$\left|\mathrm{\Delta }_b\left(\underset{x,y}{}J_\gamma (x,y)\eta (x)\eta (y)\right)\right|C\gamma ,$$
$`(4.2)`$
and
$$\left|c_b^\gamma (\eta )c_b^0(\eta )\gamma \beta \mathrm{\Phi }^{}\left(\kappa \mathrm{\Delta }_bH(\eta )\right)\left[\eta (x+e)\eta (x)\right]e\left(J\eta \right)(x)\right|C\gamma ^2,$$
$`(4.3)`$
where $`b=(x,x+e)`$.
Proof. The proof follows immediately from the expansion in Taylor series of $`\mathrm{\Phi }`$ (2.4).
A first application of Lemma 4.3 is in proving that the concept of superexponentially small event coincides for perturbed and unperturbed processes. As usual, $`\beta `$ and $`\kappa `$ are fixed, but recall that we denote by $`𝐏^{\gamma ,0}`$ the law of the process in the box $`\mathrm{\Lambda }_\gamma `$ with $`\beta =0`$, that is the reference process in the periodic box.
Lemma 4.4 Let $`T`$ be a fixed positive number, $`\mathrm{\Lambda }=\mathrm{\Lambda }_\gamma `$ and let $`_{\gamma ,𝐪}`$ be a bounded functional of the process $`\{\eta _s\}_{s[0,T\gamma ^2]}`$ which depends on $`\gamma >0`$ and on a vector parameter $`𝐪`$. Furthermore let $`\mu =\mu _\gamma ^{0,\kappa ,\lambda }`$ (recall (2.2)) and let $`\mu ^{}`$ be any probability measure on $`\mathrm{\Omega }_{\mathrm{\Lambda }_\gamma }`$. If for every $`p\mathrm{IR}`$
$$\underset{𝐪}{lim}\underset{\gamma 0}{lim\; sup}\gamma ^d\mathrm{log}𝐄_\mu ^{\gamma ,0}\left(\mathrm{exp}\left(p\gamma ^d_{\gamma ,𝐪}\right)\right)=0,$$
$`(4.4)`$
then the same is true with $`𝐄_\mu ^{\gamma ,0}`$ replaced by $`𝐄_\mu ^{}^\gamma `$. Analogously if $`\{E_{\gamma ,𝐪}\}_{\gamma ,𝐪}`$ is a family of $`_{T\gamma ^2}`$ measurable events which are superexponentially small under $`𝐏_\mu ^{\gamma ,0}`$, that is
$$\underset{𝐪}{lim}\underset{\gamma 0}{lim\; sup}\gamma ^d\mathrm{log}𝐏_\mu ^{\gamma ,0}\left(E_{\gamma ,𝐪}\right)=\mathrm{},$$
$`(4.5)`$
then they are superexponentially small also under $`𝐏_\mu ^{}^\gamma `$.
Proof. First we observe that, since for some $`c=c(F,\kappa ,\lambda )𝐑`$ we have that $`sup_{\mu ^{},\eta }(\mathrm{d}\mu ^{}/\mathrm{d}\mu )(\eta )e^{c|\mathrm{\Lambda }_\gamma |}`$, it is sufficient to prove the statements in the case $`\mu ^{}=\mu `$. We therefore omit the subscript $`\mu `$ in this proof.
For $`b=(x,y)`$ and $`t0`$, let $`N_b(t)`$ denote the number of jumps between $`x`$ and $`y`$ in the time span $`[0,t]`$. The Radon–Nikodym derivative of $`𝐏^\gamma `$ with respect to $`𝐏^{\gamma ,0}`$, both restricted to $`_{t\gamma ^2}`$ will be denoted by $`M_t`$ and it is given by
$$M_t=\mathrm{exp}\left(_0^{t\gamma ^2}\underset{b}{}\left[c_b^\gamma (\eta _s)c_b^0(\eta _s)\right]\mathrm{d}s+_0^{t\gamma ^2}\underset{b}{}\mathrm{log}\left(\frac{c_b^\gamma (\eta _s^{})}{c_b^0(\eta _s^{})}\right)\mathrm{d}N_b(s)\right),$$
$`(4.6)`$
c.f. \[Prop. 2.6, App. 1, KL\]. With respect to the measure $`𝐏^{\gamma ,0}`$ and the filtration $`\{_t\}_{t0}`$, the process $`\{M_t\}_{t0}`$ is a martingale. Therefore, for $`p>1`$, $`\{M_t^p\}_{t0}`$ is a submartingale. If we define
$$A_t=$$
$`(4.7)`$
$$_0^{t\gamma ^2}pM_s^p\underset{b}{}\left[c_b^\gamma (\eta _s)c_b^0(\eta _s)\right]\mathrm{d}s+_0^{t\gamma ^2}M_s^p\underset{b}{}c_b^0(\eta _s)\left[e^{p\mathrm{log}\left(\frac{c_b^\gamma (\eta _s)}{c_b^0(\eta _s)}\right)}1\right]\mathrm{d}s,$$
$`(4.8)`$
we have that $`\stackrel{~}{M}_t=M_t^pA_t`$ is a martingale with $`\stackrel{~}{M}_0=1`$. We now expand the expression in the right–hand side of (4.8) by taking advantage of the fact that $`(c_b^\gamma (\eta _s)/c_b^0(\eta _s))1=O_u(\gamma )`$. Precisely, by Lemma 4.3,
$$\mathrm{exp}\left(p\mathrm{log}\left(\frac{c_b^\gamma (\eta _s)}{c_b^0(\eta _s)}\right)\right)1=p\left(\frac{c_b^\gamma (\eta _s)}{c_b^0(\eta _s)}1\right)+\frac{p(p1)}{2}\left(\frac{c_b^\gamma (\eta _s)}{c_b^0(\eta _s)}1\right)^2+O_u(\gamma ^3),$$
$`(4.9)`$
where we have assumed $`p`$ bounded, say $`p2`$, for the last term. Therefore we obtain that there exists $`C`$ such that
$$\begin{array}{cc}\hfill A_t& =_0^{t\gamma ^2}M_s^p\underset{b}{}\left[c_b^0(\eta _s)\frac{p(p1)}{2}\left(\frac{c_b^\gamma (\eta _s)}{c_b^0(\eta _s)}1\right)^2+O_u(\gamma ^3)\right]\mathrm{d}s\hfill \\ & C(p(p1)\gamma ^{d+2}+\gamma ^{d+3})_0^tM_s^pds,\hfill \end{array}$$
$`(4.10)`$
where in the last step we have used the fact that $`c_b^0(\eta )`$ is uniformly bounded as well as the positivity of $`M_t`$. By taking the expectation of this last expression, recalling that $`𝐄^{\gamma ,0}[\stackrel{~}{M}_t]=1`$, by Gronwall’s Lemma we obtain
$$𝐄^{\gamma ,0}\left[M_t^p\right]\mathrm{exp}\left(Ct\left(p(p1)\gamma ^d+\gamma ^{d+1}\right)\right).$$
$`(4.11)`$
This suffices for our purposes since, by applying Hölder inequality, we obtain that for every $`p>1`$ and $`q=p/(p1)`$
$$\begin{array}{cc}\hfill \underset{𝐪}{lim\; sup}\underset{\gamma 0}{lim\; sup}\gamma ^d& \mathrm{log}𝐄^\gamma (\mathrm{exp}(p\gamma ^d_{\gamma ,𝐪})\hfill \\ & \underset{\gamma 0}{lim\; sup}\frac{\gamma ^d}{p}\mathrm{log}𝐄^{\gamma ,0}\left(M_T^p\right)+\underset{𝐪}{lim}\underset{\gamma 0}{lim\; sup}\frac{\gamma ^d}{q}\mathrm{log}𝐄^{\gamma ,0}(\mathrm{exp}(q\gamma ^d_{\gamma ,𝐪})\hfill \\ & Ct(p1).\hfill \end{array}$$
$`(4.12)`$
By letting $`p1`$ the first statement is proven. The proof of the second statement runs in the same way.
If $`h`$ is a local function, we define $`\stackrel{~}{h}(\rho )=\mathrm{IE}^{\mu ^\rho }[h]`$ for $`\rho [0,1]`$. Here is the first application of Lemma 4.4.
Lemma 4.5 (Replacement Lemma). Let $`h`$ be a local function and
$$_{b\gamma ^1}(\eta )=\left|\frac{1}{(2b\gamma ^1+1)^d}\underset{|y|b\gamma ^1}{}\left[h(\tau _y\eta )\stackrel{~}{h}\left(\mathrm{Av}_{[b\gamma ^1]}\eta (0)\right)\right]\right|,$$
$`(4.13)`$
for $`b>0`$. Then, for any $`\delta >0`$
$$\underset{b0}{lim\; sup}\underset{\epsilon 0}{lim}𝐏^\gamma \left[_0^T\gamma ^d\underset{x}{}_{b\gamma ^1}\left(\tau _x\eta _{\gamma ^2t}\right)\mathrm{d}t\delta \right]=0.$$
$`(4.14)`$
Proof. This goes through the by now classical one block and two blocks estimates. These can be found in \[VY\] (Lemma 5.2 and Theorem 6.2) for the unperturbed process and these estimates are superexponential. The extension is therefore just Lemma 4.4.
We conclude this subsection with a computation that is very relevant for us to identify the limit equation.
Lemma 4.6 . For any bounded local function $`f:\mathrm{\Omega }_{\mathrm{ZZ}^d}\mathrm{IR}`$ we have that
$$\mathrm{IE}^{\mu ^\rho }\left[\mathrm{\Phi }^{}(\kappa \mathrm{\Delta }_{0,e}H)_e\eta (0)\mathrm{\Delta }_{0,e}f(\eta )\right]=\frac{1}{2}\mathrm{IE}^{\mu ^\rho }\left[\mathrm{\Phi }(\kappa \mathrm{\Delta }_{0,e}H)_e\eta (0)\mathrm{\Delta }_{0,e}f(\eta )\right].$$
$`(4.15)`$
Proof. By differentiating both sides of (2.4), we reduce (4.15) to proving that
$$E^{\mu ^\rho }\left[\mathrm{exp}\left(\kappa \mathrm{\Delta }_{0,e}H/2\right)\varphi ^{}(\kappa \mathrm{\Delta }_{0,e}H)\left(\eta (e)\eta (0)\right)\mathrm{\Delta }_{0,e}f(\eta )\right]=0.$$
$`(4.16)`$
Let us now approximate $`\mu ^\rho `$ with a sequence of finite volume grand–canonical Gibbs measures on $`\mathrm{\Omega }_{\mathrm{\Lambda }_\gamma }`$. The result follows because
$$\begin{array}{cc}& =\underset{\eta \mathrm{\Omega }_{\mathrm{\Lambda }_\gamma }}{}e^{\kappa \mathrm{\Delta }_{0,e}H(\eta )/2}\varphi ^{}(\kappa \mathrm{\Delta }_{0,e}H(\eta ))\left(\eta (e)\eta (0)\right)\mathrm{\Delta }_{0,e}f(\eta )e^{\kappa H(\eta )}e^{\lambda _x\eta (x)}\hfill \\ & =\underset{\eta \mathrm{\Omega }_{\mathrm{\Lambda }_\gamma }}{}\left(\eta (e)\eta (0)\right)\mathrm{\Delta }_{0,e}f(\eta )e^{\kappa [H(\eta )+H(\eta ^{0,e})]/2}\varphi ^{}(\kappa \mathrm{\Delta }_{0,e}H(\eta ))e^{\lambda _x\eta (x)}\hfill \\ & =\underset{\eta \mathrm{\Omega }_{\mathrm{\Lambda }_\gamma }}{}\left(\eta (e)\eta (0)\right)\mathrm{\Delta }_{0,e}f(\eta )e^{\kappa [H(\eta ^{0,e})+H(\eta )]/2}\varphi ^{}(\kappa \mathrm{\Delta }_{0,e}H(\eta ))e^{\lambda _x\eta (x)},\hfill \end{array}$$
$`(4.17)`$
where in the last step we used the fact that $`(\eta (e)\eta (0))\mathrm{\Delta }_{0,e}f(\eta )`$ is invariant under the transformation $`\eta \eta ^{0,e}`$. Recall now that $`\varphi ^{}`$ is odd and the proof of (4.15) is complete.
4.2. Tightness and energy estimate.
We go now to the set up of Theorem 2.1. Recall that $`𝐐^\gamma `$ is the law of the empirical process $`\{\nu ^\gamma (t)\}_{t[0,T]}`$, cf. (2.14), on the Skorohod space $`D([0,T],)`$.
Lemma 4.7 The sequence $`\{𝐐^\gamma \}_{\gamma >0}`$ is tight and every limit point $`𝐐`$ is concentrated on absolutely continuous paths whose densities belong to
$`C^0([0,T];_1(\mathrm{IT}^d))L^2([0,T],H_1(\mathrm{IT}^d))`$.
Proof. This is standard. One way to prove tightness in $`D([0,T],)`$ is to repeat the proof in \[§4, VY\]: Lemma 4.1 in \[VY\] depends only on the uniform boundedness of the jump rates and Lemma 4.2, still in \[VY\], is easily upgraded to our situation via Lemma 4.4. The limit points actually lie in a smaller space: by the exclusion rule it is immediate to verify that the limit is in $`_1`$ for every $`t`$ and, since every jump produces a discontinuity $`O_u(\gamma ^d)`$, we can substitute $`D`$ with $`C^0`$. The energy estimate, that is the existence of a constant $`C`$ such that any limit point $`𝐐`$ satisfies
$$\mathrm{IE}^𝐐\left[_0^T_{\mathrm{IT}^d}(\rho (s,r))dr\mathrm{ds}\right]<C,$$
$`(4.18)`$
requires a more sophisticated argument. Once again most of the work has been done in \[VY\], Section 5: it is sufficient to upgrade (5.25) and (5.28)–(5.30) in \[VY\] to our situation. One way would be to get a volume order estimate on the relative entropy of $`𝐏^\gamma `$ with respect to $`𝐏^0`$, both measures restricted to $`[0,T\gamma ^2]`$, but we choose to stick to $`L^p`$ estimates on the Radon–Nicodym derivative $`M_T`$ of the process. We will be as close as possible to the notations of Lemma 4.4 and, like in its proof, we will omit the dependence on the initial condition: the only difference is that we will not need any extra parameter $`𝐪`$ and we will therefore omit the subscript $`𝐪`$ from the notation. We choose, with $`GC^1(𝐑^+,\mathrm{IT}^d;𝐑)`$ and $`C_1>0`$,
$$_{\gamma ,𝐪}=_\gamma =\gamma ^{d1}_0^T\underset{x}{}G(t,x\gamma )_e\eta _{t\gamma ^2}(x)\mathrm{d}t\frac{C_1}{2}\gamma ^d_0^T\underset{x}{}\left|G(t,x/N)\right|^2\mathrm{d}t.$$
$`(4.19)`$
In the proof of Lemma 3.2 in \[VY\] it is shown that for some $`C_1`$ (4.4) holds for $`p=1`$, that is
$$\underset{\gamma 0}{lim}\gamma ^d\mathrm{log}𝐄^{\gamma ,0}\left[\mathrm{exp}\left(\gamma ^d_\gamma \right)\right]=0.$$
$`(4.20)`$
Using (4.20) we will show that there exists $`C_2`$ such that
$$\underset{\gamma 0}{lim\; sup}𝐄^\gamma \left[_\gamma \right]C_2.$$
$`(4.21)`$
We refer the reader to \[Section 5,VY\] for a proof of the fact that (4.21), with the choice (4.19), implies (4.18) with $`C=\mathrm{max}(C_1,C_2)`$. Here we prove (4.21); it is just an application of Jensen’s inequality and Cauchy–Schwarz:
$$\begin{array}{cc}\hfill 𝐄^\gamma \left[_\gamma \right]& 2\gamma ^d\mathrm{log}𝐄^{\gamma ,0}\left[M_T\mathrm{exp}\left(\frac{\gamma ^d}{2}_\gamma \right)\right]\hfill \\ & \gamma ^d\mathrm{log}𝐄^{\gamma ,0}\left[M_T^2\right]+\gamma ^d\mathrm{log}𝐄^{\gamma ,0}\left[\mathrm{exp}\left(\gamma ^d_\gamma \right)\right],\hfill \end{array}$$
$`(4.22)`$
and, recalling (4.20), (4.21) follows from (4.12) with $`p=2`$.
4.3. The (partial) identification of the limit. By the definition of the process, for every $`GC^1(\mathrm{IT}^d;\mathrm{IR})`$ we can write
$$\begin{array}{cc}\hfill \gamma ^d\underset{x}{}G(\gamma x)\eta _{t\gamma ^2}(x)& \gamma ^d\underset{x}{}G(\gamma x)\eta _0(x)=\hfill \\ & \gamma ^d_0^{t\gamma ^2}\underset{x,e}{}\left[G(\gamma (x+e))G(\gamma x)\right]w_{x,x+e}^\gamma (\eta _s)\mathrm{d}s+M_\gamma ^G(t)\hfill \end{array}$$
$`(4.23)`$
in which $`M_\gamma ^G=\{M_\gamma ^G(t)\}_{r\mathrm{IR}^+}`$ is a $`𝐏^\gamma `$–martingale with respect to the filtration associated to $`\{\eta _t\}_{t\mathrm{IR}^+}`$. The quadratic variation of $`M_\gamma ^G`$ is easily computed and estimated: if we set $`X_\gamma (\eta )=\gamma ^d_xG(\gamma x)\eta (x)`$ it is immediate to see that there exists a constant $`c=c(d,G)`$ such that for every $`\gamma >0`$
$$M_\gamma ^G,M_\gamma ^G(t)=_0^{t\gamma ^2}\left(_\gamma X_\gamma ^22X_\gamma _\gamma X_\gamma \right)(\eta _s)dsc\gamma ^d,$$
$`(4.24)`$
and thus, by Doob’s inequality, for every $`T>0`$ and every $`\delta >0`$
$$\underset{\gamma 0}{lim}𝐏^\gamma \left(\underset{t[0,T]}{sup}\left|M_\gamma ^G(t)\right|>\delta \right)=0.$$
$`(4.25)`$
We now split the current:
$$w^\gamma =[w^0]+[w^\gamma w^0].$$
$`(4.26)`$
The non gradient difficulties come from the first term. Below we use the convention that $`_e`$ is the sum over the unit vectors $`\{e_j\}_{j=1,2,\mathrm{},d}`$.
We first have a fast look at the easy second term. By Lemma 4.3, for every $`x`$ and every unit vector $`e`$
$$[w^\gamma w^0]_{x,x+e}(\eta )=\beta \gamma \mathrm{\Phi }^{}\left(\mathrm{\Delta }_{x,x+e}H(\eta )\right)[_e\eta ]^2(x)e(J\eta )(x)+O_u(\gamma ^2).$$
$`(4.27)`$
If we define
$$\begin{array}{cc}\hfill _{b,\gamma }\left(\left\{\eta _s\right\}_{s0}\right)& =|_0^{\gamma ^2T}\gamma ^d\underset{x,e}{}[G(\gamma (x+e))G(\gamma x)][w_{x,x+e}^\gamma (\eta _t)w_{x,x+e}^0(\eta _t)]\mathrm{d}t\hfill \\ \hfill +\beta _0^{\gamma ^2T}& \gamma ^{d+2}\underset{x,e}{}\left(_eG\right)(\gamma x)(_eJ\eta _t)(x)\frac{1}{(2b\gamma ^1+1)^d}\underset{|y|b\gamma ^1}{}h_e\left(\tau _{x+y}\eta _t\right)\mathrm{d}t|,\hfill \end{array}$$
$`(4.28)`$
where $`h_e(\eta )=\mathrm{\Phi }^{}(\mathrm{\Delta }_{0,e}H(\eta ))(\eta (e)\eta (0))^2`$, and therefore
$$\stackrel{~}{h}_e(\rho )=\mathrm{IE}^{\mu ^\rho }\left[\mathrm{\Phi }^{}\left(\kappa \mathrm{\Delta }_{0,e}H(\eta )\right)(\eta (e)\eta (0))^2\right].$$
$`(4.29)`$
Note that, by applying Lemma 4.6, with $`f1`$, we can immediately rewrite
$$\stackrel{~}{h}_e(\rho )=\frac{1}{2}\mathrm{IE}^{\mu ^\rho }\left[c_{0,e}^0(\eta )(\eta (e)\eta (0))^2\right].$$
$`(4.30)`$
By the smoothness of $`G`$ and $`J`$ it is immediate to obtain that
$$\underset{b0}{lim}\underset{\gamma 0}{lim\; sup}\underset{\left\{\eta _s\right\}_{s0}}{sup}_{b,\gamma }\left(\left\{\eta _s\right\}_{s0}\right)=0.$$
$`(4.31)`$
This, together with Lemma 4.5 applied to the spatial average of $`h`$, immediately implies that
$$\begin{array}{cc}\hfill |_0^{\gamma ^2T}\gamma ^d\underset{x,e}{}[G(\gamma (x+e))G(\gamma x)]& \left[w_{x,x+e}^\gamma (\eta _t)w_{x,x+e}^0(\eta _t)\right]\mathrm{d}t\hfill \\ \hfill _0^T_{\mathrm{IT}^d}G(r)& (J\nu ^\gamma )(t,r)\stackrel{~}{h}(\mathrm{𝟏}_{(b,b)^d}\nu ^\gamma )(t,r)\left)\mathrm{d}r\mathrm{d}t\right|\hfill \end{array}$$
$`(4.32)`$
tends to zero in $`𝐏^\gamma `$–probability as $`\gamma 0`$ and then $`b0`$.
Now we turn to the non gradient term. Given $`a`$, $`b`$, $`\delta `$ and $`\delta _1>0`$, we define
$$\begin{array}{cc}\hfill _{a,b,\delta }=\{& \underset{t[0,T]}{sup}|\gamma ^{d+1}_0^{\gamma ^2t}\underset{x}{}\underset{e}{}_eG(\gamma x)w_{x,x+e}^0(\eta _t)\mathrm{d}t+\hfill \\ \hfill \gamma ^{d+1}_0^{\gamma ^2T}(2b)^1\underset{x,e,e^{}}{}& \left[\eta _s(x+b\gamma ^1e)\eta _s(xb\gamma ^1e)\right]D_{e,e^{}}\left(\mathrm{Av}_{[a\gamma ^1]}\eta _s(x)\right)_eG(\gamma x)+\hfill \\ & \gamma ^{d+1}_0^{\gamma ^2T}\underset{x,e}{}_eG(\gamma x)^0g_e^\delta \left(\tau _x\eta _t\right)|\delta _1\}.\hfill \end{array}$$
$`(4.33)`$
The following result follows from Theorem 3.3 in \[VY\] and Lemma 4.4, with the only observation that in the statement in \[VY\] the term $`^0g^\delta `$ does not appear: in their case it is irrelevant (see Section 3 of \[VY\]).
Lemma 4.8 For every $`\delta _1>0`$
$$\underset{a0}{lim}\underset{b0}{lim\; sup}\underset{\delta 0}{lim\; sup}\underset{\gamma 0}{lim\; sup}𝐏^\gamma \left(_{a,b,\delta }\right)=0.$$
$`(4.34)`$
We now observe that in (4.33) we can replace $`^0g^\delta `$ with $`(^0_\gamma )g^\delta `$ and Lemma 4.8 still holds since
$$\begin{array}{cc}& \gamma ^{d+1}_0^{\gamma ^2T}\underset{x,e}{}_eG(\gamma x)_\gamma g_e^\delta \left(\tau _x\eta _t\right)\hfill \\ \hfill =\gamma & \left[\gamma ^d\underset{x,e}{}_eG(\gamma x)\left(g_e^\delta \left(\tau _x\eta _{\gamma ^2T}\right)g_e^\delta \left(\tau _x\eta _0\right)\right)\right]+\gamma \stackrel{~}{M}_\gamma (T)+E_\gamma (T).\hfill \end{array}$$
$`(4.35)`$
where $`\{\stackrel{~}{M}^\gamma (t)\}_{t0}`$ with respect to the natural filtration is a $`𝐏^\gamma `$–martingale and $`E_\gamma (T)`$ is the error we made by considering $`g^\delta `$ independent of $`\{\mathrm{Av}_\mathrm{}_\delta \eta (x)\}_x`$, $`\mathrm{}_\delta `$ the range of $`g^\delta `$. The first term in the second line of (4.35) is $`O_u(\gamma )`$, the martingale $`\stackrel{~}{M}_\gamma `$ has a quadratic variation $`O(\gamma ^d)`$ and $`sup_{t[0,T]}|E_\gamma (t)|`$ tends to zero in probability as $`\gamma 0`$ and $`\delta 0`$. The last claim follows from the argument in \[VY\], formulas (3.26)–(3.27), and Lemma 4.4.
We are therefore left with evaluating
$$\gamma _0^{\gamma ^2T}\gamma ^d\underset{x,e}{}_eG(\gamma x)\left[^0_\gamma \right]g_e^\delta \left(\tau _x\eta _t\right)\mathrm{d}t.$$
$`(4.36)`$
By using Lemma 4.3 to express $`^0_\gamma `$, recalling that $`g_e^\delta `$ is a local function and repeating exactly the same steps, i.e. smearing and using the replacement lemma, as in (4.28)–(4.32), we obtain that, up to a negligible error term, this term can be replaced by
$$\beta \underset{e,e^{}}{}_0^T_{\mathrm{IT}^d}_eG(r)\theta _{e,e^{}}^\delta \left(\chi _{[\delta ,\delta ]^d}\nu ^\gamma (t,r)\right)_e^{}J\nu ^\gamma (t,r)\mathrm{d}r\mathrm{d}t,$$
$`(4.37)`$
where
$$\theta _{e,e^{}}^\delta (\rho )=𝐄^{\mu ^\rho }\left[\underset{z}{}\left(\eta (z+e^{})\eta (z)\right)\mathrm{\Phi }^{}\left(\kappa \mathrm{\Delta }_{z,z+e^{}}H_0(\eta )\right)\left[g_e^\delta (\eta ^{z,z+e^{}})g_e^\delta (\eta )\right]\right].$$
$`(4.38)`$
Let us define $`\theta (\rho )=lim\; sup_{\delta 0}\theta ^\delta (\rho )`$ for every $`\rho [0,1]`$. By Lemma 4.7 and collecting the results (4.23), (4.25), (4.31) and Lemma 4.8, under the hypothesis that $`\theta ^\delta `$ converges uniformly as $`\delta 0`$, we obtain that every limit of the sequence $`\{\nu ^\gamma \}_{\gamma >0}`$ concentrates on trajectories $`\rho `$ which solve weakly the PDE
$$\{\begin{array}{cc}& _t\rho =\underset{i,j}{}_i\left\{D_{i,j}(\rho )_j\rho \beta \left(\delta _{i,j}\stackrel{~}{h}_j(\rho )+\theta _{i,j}(\rho )\right)\left(_jJ\rho \right)\right\},\hfill \\ & \rho (0,)=\rho _0().\hfill \end{array}$$
$`(4.39).`$
We are therefore left with showing that
Lemma 4.9 The sequence $`\theta ^\delta `$ converges uniformly and for every $`\rho [0,1]`$, every $`i`$ and every $`j=1,2,\mathrm{},d`$ we have that
$$\delta _{i,j}\stackrel{~}{h}_j(\rho )+\theta _{i,j}(\rho )=\chi (\rho )D_{i,j}(\rho ).$$
$`(4.40)`$
Proof. We divide this algebraic computations in steps
Step 1. We start by rewriting $`\theta `$ in a more convenient form, using Lemma 4.6 with $`f=g_{e_j}^\delta `$. We obtain that
$$\theta _{i,j}^\delta (\rho )=\frac{1}{2}\mathrm{IE}^{\mu ^\rho }\left[\underset{z}{}c_{z,z+e_j}(\eta )\left(\eta (z+e_j)\eta (z)\right)\mathrm{\Delta }_{z,z+e_j}g_{e_i}^\delta (\eta )\right].$$
$`(4.41)`$
Step 2. Reversibility and summation by parts on (4.41) imply
$$\theta _{i,j}^\delta (\rho )=\mathrm{IE}^{\mu ^\rho }\left[\underset{z}{}c_{z,z+e_j}(\eta )\left(\eta (z+e_j)\eta (z)\right)g_{e_i}^\delta (\tau _z\eta )\right]=w_{0,e_j}^0,g_{e_i}^\delta _0(\rho ),$$
$`(4.42)`$
where in the last step we used the notation
$$f,g_0(\rho )=\underset{x}{}\mathrm{IE}^{\mu ^\rho }\left[f(\eta )g(\tau _x\eta )\right],$$
$`(4.43)`$
for $`f`$ and $`g`$ local functions. It is not too difficult to show \[Section 8, VY\] that
$$f,g_0=f,_0g,$$
$`(4.44)`$
and we recall that the covariance $`,`$ is defined right after Theorem 3.2. Therefore
$$\theta _{i,j}^\delta =w_{0,e_j}^0,_0g_{e_i}^\delta .$$
$`(4.45)`$
Step 3. We now use the decomposition induced by (3.9) to express $`_0g_{e_j}^\delta `$ obtaining that $`\theta _{i,j}^\delta `$ is equal to
$$w_{0,e_j}^0,_0g_{e_i}^\delta (\rho )=w_{0,e_i}^0,w_{0,e_i}^0(\rho )w_{0,e_j}^0,(D\eta )_{e_i}(0)(\rho )+R(\rho ,\delta ),$$
$`(4.46)`$
where $`lim_{\delta 0}sup_\rho |R(\rho ,\delta )|=0`$.
Completing the proof of (4.40) is now just a matter of applying (3.11), note the cancellation with the term $`\stackrel{~}{h}`$, c.f. (4.30), and (3.12) and the proof is complete.
Lemma 4.9 leads us to the weak formulation of the PDE (2.16), that is that every limit measure $`𝐐`$ is concentrated on trajectories $`\rho C^0([0,T];_1(\mathrm{IT}^d))L^2([0,T],H_1(\mathrm{IT}^d))`$ which solve
$$\begin{array}{cc}\hfill _{\mathrm{IT}^d}G(r)\rho (T,r)dr& _{\mathrm{IT}^d}G(r)\rho (0,r)dr=\hfill \\ & _0^T_{\mathrm{IT}^d}G(r)\left(D(\rho (t,r))\chi (\rho (t,r))J\rho D(\rho )\rho (t,r)\right)\mathrm{d}r\mathrm{d}t,\hfill \end{array}$$
$`(4.47)`$
for every $`GC^1(\mathrm{IT}^d)`$ and every $`T>0`$. Therefore the proof of Theorem 2.1 is complete modulo discussing the uniqueness issue. This is considered in the next subsection.
4.4. Uniqueness. A proof of uniqueness is available if $`D`$ is a uniformly Lipschitz continuous function. This follows from the control of $`H^1`$ norm. In \[GL2\] this result is proven under an additional condition on the time derivative, in the sense of distributions, of $`\rho `$. This condition can be removed if one replaces the kernel of the $`H^1`$ norm with a smoothened kernel, as it is done in \[Appendix A, KL\]. Note that the weak formulation (4.47) is in this set up equivalent to the formulation that we get by multiplying both sides of (2.16) by $`\stackrel{~}{G}C^1(\mathrm{IR}^+\times \mathrm{IT}^d;\mathrm{IR})`$ and formally integrating by parts.
5. Multispecies systems
The scheme of proof of Section 3 and Section 4 can be applied to several other systems. Of particular interest for applications is the case of systems with several types of particles. One (apparently) simple system in this class is the $`n`$–color exclusion process: take a simple exclusion process and distinguish the particles by painting them with $`n`$ colors ($`n`$ is kept fix). The hydrodynamics of this system has been done in \[Q\]: here we would like to consider the case in which the particles interact via long range potentials that distinguish between colors.
From the technical viewpoint this case has essentially been already considered in \[QRV\], where a Large Deviation principle for a $`n`$–color exclusion system is proven as a step to prove a process Large Deviation for the simple exclusion. The lower bound of the Large Deviations depends on the standard change of measure argument which entails studying the $`n`$–color system driven by a weak external force. This is also our case: the weak driving force is however configuration dependent, but since it depends on the configuration only via an empirical average, the changes with respect to \[QRV\] are minimal. We will therefore simply state the result and make some observations on the proof.
To simplify the notation and the statement of the result we restrict our attention to the case of two species (A and B) and to the case in which A and B interact with each other, but A does not interact with any A and the same for B particles.
5.1 The A–B model. As in the previous sections $`\mathrm{\Lambda }_\gamma `$ will denote the lattice torus with diameter $`[\gamma ^1]`$. We are looking at a $`d`$–dimensional system of A and B particles evolving via a Kawasaki dynamics with Kac Hamiltonian
$$H_\gamma (\eta )=\frac{1}{2}\underset{x,y\mathrm{\Lambda }_\gamma }{}J_\gamma (x,y)\eta ^A(x)\eta ^B(y)$$
$`(5.1)`$
where as usual $`\gamma >0`$ and $`\eta ^A`$ and $`\eta ^B`$ are elements of $`\{0,1\}^{\mathrm{\Lambda }_\gamma }`$ with the hard core restriction that there can be at most one particle per site
$$\eta ^A+\eta ^B1.$$
$`(5.2)`$
$`J_\gamma (x,y)=\gamma ^dJ(\gamma (xy))`$, where $`J`$ is a smooth function from the $`d`$–dimensional unit torus to $`\mathrm{IR}`$. The particle configuration can be alternatively described by
$$\eta \{0,A,B\}^{\mathrm{\Lambda }_\gamma },$$
$`(5.3)`$
and we will identify $`\eta `$ with $`(\eta ^A,\eta ^B)`$. The dynamics which conserves both $`A`$ and $`B`$ particles (or interchanging with) are specified by $`A`$ and $`B`$ particles hopping to nearest neighbor empty sites at a rate $`Dc_b^{\gamma ,AB}(\eta )`$ but no direct exchanges between $`A`$ and $`B`$ particles are permitted. This models the physical situation of polymers in a fluid considered in \[WD\]. The generator of the dynamics is
$$L_\gamma f(\eta )=$$
$`(5.4)`$
$$D\underset{b=(x,y)\mathrm{\Lambda }_\gamma ^{}}{}c_b^{\gamma ,AB}(\eta )\left[(\eta ^A(x)\eta ^B(y))^2+(\eta ^B(x)\eta ^A(y))^2\right]\left[f(\eta ^{x,y})f(\eta )\right],$$
where $`D>0`$, $`f`$ is a real valued function defined in $`\{0,A,B\}^{\mathrm{\Lambda }_\gamma }`$ and $`\eta ^{x,y}`$ is $`\eta `$ with the occupation of the sites $`x`$ and $`y`$ exchanged. As before, the rates $`c_b^{\gamma ,AB}`$ are such that the dynamics satisfies the detailed balance condition with respect to the Gibbs measure with Hamiltonian (5.1) at inverse temperature $`\beta >0`$:
$$c_{(x,y)}^{\gamma ,AB}(\eta )=\mathrm{\Phi }\left(\beta \left[H_\gamma (\eta ^{x,y})H_\gamma (\eta )\right]\right),$$
$`(5.5)`$
$`\mathrm{\Phi }`$ as in (2.4). To denote the process we keep the same notation of the previous sections.
5.2 The hydrodynamics of the A–B model. For the hydrodynamic limit we look at the empirical measure
$$\nu _\alpha ^\gamma (t,x)=\gamma ^d\underset{y\mathrm{\Lambda }_\gamma }{}\delta (x\gamma y)\eta _{\gamma ^2t}^\alpha (y),$$
$`(5.6)`$
for $`\alpha \{A,B\}`$, $`x\mathrm{IT}^d`$ and $`\eta ^\alpha (x,s)`$ specifies the presence or absence of an $`\alpha `$-particle on site $`x`$ at time $`s`$.
For the initial datum we assume $`\nu _\alpha ^\gamma (0,)`$ to be close to $`\rho ^\alpha ()`$ in the sense of (2.15). In this case we have to make further assumptions on $`\rho ^\alpha `$, precisely that there exists $`\delta >0`$ such that for every $`x`$ and every $`\alpha `$
$$\delta \rho ^\alpha (x)1\delta ,$$
$`(5.7)`$
and that $`\rho ^\alpha `$ is differentiable with bounded derivatives
$$\underset{x,\alpha }{sup}\left|\rho ^\alpha (x)\right|<\mathrm{}.$$
$`(5.8)`$
For the moment let us set $`J=0`$. In \[Q\] it is shown that $`\nu _\alpha ^\gamma `$ converges weakly in probability, that is in the sense of (2.15) for every $`t0`$, to $`\rho ^\alpha :\mathrm{IR}^+\times \mathrm{IT}^d[0,1]`$ and the couple $`(\rho ^A,\rho ^B)`$ solves
$$_t\left(\begin{array}{c}\rho ^A\\ \rho ^B\end{array}\right)=\left[\left(\begin{array}{cc}\frac{\rho ^B}{\rho }D_s+\frac{\rho ^A}{\rho }D& \frac{\rho ^A}{\rho }(DD_s)\\ \frac{\rho ^B}{\rho }(DD_s)& \frac{\rho ^A}{\rho }D_s+\frac{\rho ^B}{\rho }D\end{array}\right)\left(\begin{array}{c}\rho ^A\\ \rho ^B\end{array}\right)\right],$$
$`(5.9)`$
with initial condition $`(\rho ^A(0,),\rho ^B(0,))=(\rho ^A(),\rho ^B())`$. In (5.9) $`\rho =\rho ^A+\rho ^B`$ and $`D_s=D_s(\rho )`$ is the self diffusion coefficient. The expression for the diffusion matrix $`D`$ in (5.9) can be derived from elementary considerations on the microscopic system from which it is derived \[LS\]. We observe that, as expected, the evolution equation for $`\rho `$ is simply
$$_t\rho =D\mathrm{\Delta }\rho .$$
$`(5.10)`$
This follows from the observation that if we ignore the distinction between $`A`$ and $`B`$ particles then, in the absence of interactions, we are just dealing with the one component simple symmetric exclusion process \[Sp\].
We can rewrite the system (5.9) as
$$_t\underset{¯}{\rho }=\left[𝒟\underset{¯}{\rho }\right]=\left[𝐌\frac{\delta _0}{\delta \underset{¯}{\rho }}\right]$$
$`(5.11)`$
in which $`\underset{¯}{\rho }=(\rho ^A,\rho ^B)^t`$, and
$$_0(\rho ^A,\rho ^B)=\frac{1}{\beta }s(\rho ^A,\rho ^B)dx,$$
$`(5.12)`$
with
$$s(\rho ^A,\rho ^B)=\left[\rho ^A\mathrm{log}\rho ^A+\rho ^B\mathrm{log}\rho ^B+(1\rho ^A\rho ^B)\mathrm{log}(1\rho ^A\rho ^B)\right]$$
$`(5.13)`$
and
$$𝐌=\beta 𝒟(\mathrm{𝐇𝐞𝐬𝐬}(s))^1.$$
$`(5.14)`$
Explicitely
$$\mathrm{𝐇𝐞𝐬𝐬}(s)^1=\left(\begin{array}{cc}\rho ^A(1\rho ^A)& \rho ^A\rho ^B\\ \rho ^A\rho ^B& \rho ^B(1\rho ^B)\end{array}\right),$$
$`(5.15)`$
and
$$𝐌=\beta \left(\begin{array}{cc}D_s\frac{\rho ^A\rho ^B}{\rho }+D\frac{(\rho ^A)^2(1\rho )}{\rho }& D_s\frac{\rho ^A\rho ^B}{\rho }+D\frac{\rho ^A\rho ^B(1\rho )}{\rho }\\ D\frac{\rho ^B\rho ^A(1\rho )}{\rho }D_s\frac{\rho ^A\rho ^B}{\rho }& D\frac{(\rho ^B)^2(1\rho )}{\rho }+D_s\frac{\rho ^A\rho ^B}{\rho }\end{array}\right).$$
$`(5.16)`$
We now claim that, in the case in which $`J`$ is not zero, the limit equation (5.11) has to be changed by replacing $`_0`$ with
$$=_0+\frac{1}{2}J(xx^{})\rho ^A(x)\rho ^B(x^{})dxdx^{}.$$
$`(5.17)`$
Proposition 5.10 Set $`d3`$. For every $`t0`$ the empirical field $`(\nu _a^\gamma (t,),\nu _B^\gamma (t,))`$ converges weakly in probability, i.e. in the sense of (2.15), to the unique weak solution of
$$_t\underset{¯}{\rho }=\left[𝒟\underset{¯}{\rho }+𝐌J\underset{¯}{\rho }\right].$$
$`(5.18)`$
The restriction to $`d3`$ is due to the fact that only under this restriction $`D_s`$ is known to be a Lipschitz function and uniqueness of the weak solution holds.
Acknowledgments. R. M. would like to thank R. Esposito and E. Presutti for useful discussions. Research supported by NSF Grant DMR-9813268, AFOSR Grant F49620-98-1-0207, and DIMACS and its supporting agencies the NSF under contract STC–91–19999 and the NJ Commission on Science and Technology. G.G. acknowledges the support of MURST (cofin99) and of GNAFA and R. M. the support of MURST (cofin98).
References
\[A\] A. Asselah and L. Xu, private communication.
\[BL\] P. Buttà, J. L. Lebowitz, Hydrodynamic limit of Brownian particles interacting with short– and long–range forces, J. Statist. Phys. 94 (1999), no. 3–4, 653–694.
\[DG\] P. G. de Gennes, Dynamics of fluctuations and spinodal decomposition in polymer blends, J. Chem. Phys. 72 (1980), no. 9, 4756–4763.
\[EP\] W. E and P. Palffy-Muhoray, Phase separation in incompressible systems, Phys. Rev. E (3) 55 (1997), no. 4, R3844–R3846.
\[GL1\] G. Giacomin and J. L. Lebowitz, Phase segregation dynamics in particle systems with long range interaction I. Macroscopic limits, J. Stat. Phys. 87, 37-61. (1997).
\[GL2\] G. Giacomin and J. L. Lebowitz, Phase segregation dynamics in particle systems with long range interaction II. Interface motion, SIAM J. Appl. Math. 58, 1707–1729 (1998).
\[GLP\] G. Giacomin, J. L. Lebowitz and E. Presutti, Deterministic and Stochastic Hydrodynamic Equations Arising From Simple Microscopic Model Systems, Stochastic partial differential equations: six perspectives, 107–152, Math. Surveys Monogr., 64, Amer. Math. Soc., Providence, RI, 1999.
\[GPV\] M. Z. Guo, G. C. Papanicolaou and S. R. S. Varadhan, Nonlinear diffusion limit for a system with nearest neighbor interactions, Comm. Math. Phys., 118(1988), 31-59.
\[KL\] C. Kipnis and C. Landim, Scaling limits of interacting particle systems, Grundlehren der Mathematischen Wissenschaften \[Fundamental Principles of Mathematical Sciences\], 320, Springer-Verlag, Berlin, 1999.
\[Li\] T. Liggett, Interacting particle systems, Springer–Verlag (1985).
\[LOP\] J. L. Lebowitz, E. Orlandi, E. Presutti, A particle model for spinodal decomposition, J. Statist. Phys. 63 (1991), no. 5–6, 933–974.
\[LS\] J.L. Lebowitz and H. Spohn, Comment on: ”Onsager reciprocity relations without microscopic reversibility” by D. Gabrielli, G. Jona–Lasinio and C. Landim, Phys. Rev. Lett. 78 (1997), no. 2, 394–395.
\[MM\] R. Marra and M. Mourragui, Phase segregation dynamics for the Blume-Capel model with Kac interaction. to appear on Stoc. Processes and Appl. (2000).
\[Q\] J. Quastel, Diffusion of a color in the simple exclusion process, Comm. Pure Appl. Math. XLV, 623–679 (1992).
\[QRV\] J. Quastel, F. Rezakhanlou and S. R. S. Varadhan, Large Deviations for the symmetric simple exclusion process in dimension $`d3`$, Probab. Theory Related Fields 113 (1999), no. 1, 1–84.
\[R\] F. Rezakhanlou, Hydrodynamic limit for a system with finite range interactions, Comm. Math. Phys. 129 (1990), no. 3, 445–480.
\[Sp\] H. Spohn, Large scale dynamics of interacting particles, Springer, Berlin, (1991).
\[V\] S.R.S. Varadhan, Nonlinear diffusion limit for a system with nearest neighbor interactions II, in Asymptotic Problems in Probability Theory : Stochastic Models and Diffusion on Fractals, edited by K. Elworthy and N. Ikeda, Pitman Research Notes in Mathematics 283, Wiley, (1994).
\[VY\] S. R. S. Varadhan and H.–T. Yau, Diffusive Limit of Lattice Gas with Mixing Conditions. Asian J. Math. 1, 623–678 (1997).
|
warning/0003/cond-mat0003085.html
|
ar5iv
|
text
|
# Non-Fermi Liquid Behavior in U and Ce Alloys: Criticality, Disorder, Dissipation, and Griffiths-McCoy singularities.
## I Introduction
In this paper we are interested in the quantum critical behavior of alloys of actinides and rare earths with metallic atoms. There is a large number of these alloys and they can be classified into two main groups accordingly to their chemical composition: 1) Kondo hole systems which are made out of the substitution of the rare earth or actinide (R) by a non-magnetic metallic atom (M) with a chemical formula R<sub>1-x</sub>M<sub>x</sub> (a typical example is U<sub>1-x</sub>Th<sub>x</sub>Pd<sub>2</sub>Al<sub>3</sub>); 2)Ligand systems where one of the metallic atoms (M1) is replaced by another (M2) but the rare earths or actinides are not touched and thus have the formula R(M1)<sub>1-x</sub>(M2)<sub>x</sub> (as, for instance, UCu<sub>5-x</sub>Pd<sub>x</sub>). In most of these alloys one usually has an ordered magnetic phase at $`x=0`$ which is destroyed at some critical value $`x=x^{}`$ with a non-Fermi liquid (NFL) phase in the vicinity where strong deviations from the predictions of Landau’s theory of the Fermi liquid are observed (see Fig.1). In this paper we propose that the origin for the NFL behavior is the existence of Griffiths-McCoy singularities at low temperatures. As we explain below, these singularities have their origin in a quantum percolation problem driven by the number of magnetically compensated moments. In this percolation problem magnetic clusters can tunnel in the presence of a metallic environment which produces dissipation.
On the one hand, at $`x=1`$ a Kondo hole system becomes an ordinary Fermi liquid with a temperature independent specific heat coefficient, $`\gamma (T)=C_V/T`$, (by low temperature we mean $`TE_F`$ where $`E_F`$ is the Fermi energy of the metal), the magnetic susceptibility is paramagnetic and given by Pauli’s expression, $`\chi (T)=\chi _0`$, and the resistivity has the Fermi liquid form $`R(T)=R_0+AT^2`$. On the other hand, at large $`x`$ a ligand system is usually a heavy fermion , that is, it can also be described by the Fermi liquid expressions but with coefficients $`\gamma `$ and $`\chi _0`$ orders of magnitude larger than in ordinary metals. The nature of this heavy fermion behavior can be tracked down to the presence of the rare earths or actinides in the alloy. In the NFL phase the specific heat coefficient and the magnetic susceptibility show divergent behavior as the temperature is lowered.
The root for the understanding of the problem lies on the fact that in Landau liquids the thermodynamic and response functions are always well behaved functions of the temperature. This is clearly inconsistent with the behavior in the paramagnetic phase close to a quantum critical point (QCP). Exactly at the QCP divergences are expected since the system is ordering magnetically, thus, QCP can generate NFL behavior (albeit with fine tuning of the chemical composition). It turns out, however, that many times NFL behavior is observed away from the QCP and inside of the paramagnetic phase. One possibility is that the NFL behavior is due to single ion physics and therefore not at all related to the QCP physics. Another possibility is that there are still “traces” of the QCP inside of the paramagnetic phase. This is possible in the presence of disorder which can “pin” pieces of the ordered phase even in the absence of long range order. It is even conceivable that single ion physics and critical behavior co-exist close to a QCP. There is no clear consensus among researchers on the nature of this NFL behavior and the scientific debate is intense. The objective of this work is to shed light into this controversy.
We organize the paper in the following fashion: Section II briefly reviews the problem of NFL behavior in U and Ce intermetallics; in Section III we write an effective Hamiltonian where Kondo effect and RKKY interaction appear explicitly and we discuss the problem of magnetic ordering and dilution within this Hamiltonian; Section IV contains a detailed discussion of the one impurity and two impurity Kondo problem and the extension of the problem to $`N`$ spin clusters where we define a Kondo cluster effect; in Section V we propose the dissipative quantum droplet model which is the basis for the description of the problem of magnetic ordering in metallic magnetic alloys and we show that at low temperatures dissipation dominates the behavior of the system leading to universal logarithmic divergences, and that at higher temperatures non-universal power law behavior is expected; in Section VI we study the intermediate temperature range where quantum Griffiths singularities are expected; Section VII contains our conclusions. We also have included a few appendices with detailed calculations of the results used in the paper.
## II NFL behavior in U and Ce intermetallics
The basis for the study of metals in the last 50 years has been Landau’s theory of the interacting Fermi gas . As consequences of Landau’s theory the thermodynamic and response functions of the electron fluid are smooth functions of the temperature. The magnetic susceptibility has the form (we use units such that $`\mathrm{}=k_B=1`$)
$`\chi (T)={\displaystyle \frac{(g\mu _B)^2N(0)}{4(1+F_0^a)}},`$ (1)
where $`g2`$ is the Landé factor for the electron, $`\mu _B`$ is the Bohr magneton, $`F_0^a`$ is an antisymmetric Landau parameter,
$`N(0)={\displaystyle \frac{m^{}k_F}{\pi ^2}}={\displaystyle \frac{3\rho _e}{2E_F}},`$ (2)
is the density of states at the Fermi energy, $`m^{}`$ is the effective mass of the quasiparticles (the effective mass is related to the bare mass, $`m`$, by a symmetric Landau parameter: $`m^{}/m=1+F_1^s/3`$),
$`k_F=(3\pi ^2\rho _e)^{1/3},`$ (3)
is the Fermi momentum, $`\rho _e`$ is the number of electrons per unit of volume and $`E_F=k_F^2/(2m^{})`$. Moreover, the electronic specific heat, $`C_V`$, is given by the Fermi liquid expression
$`{\displaystyle \frac{C_V}{T}}=\gamma (T)={\displaystyle \frac{\pi ^2N(0)}{3}}.`$ (4)
Naturally these expressions resemble the ones obtained for the free electron gas where the bare mass $`m`$ is replaced by the effective mass $`m^{}`$ and Landé factor is renormalized by a factor of $`1/\sqrt{1+F_0^a}`$. Furthermore, at low temperatures one expects the electronic resistivity to behave like
$`R(T)=R_0+AT^2`$ (5)
where $`R_0`$ is the resistivity due to impurities and $`A`$ is a constant. In a magnetic alloy where the magnetic moments are decoupled from the conduction electrons and do not interact directly among each other we expect an ordinary paramagnetic behavior for the susceptibility in the low field limit ($`Tg\mu _BH`$ where $`H`$ is an applied magnetic field)
$`\chi (T)={\displaystyle \frac{𝒞}{T}}`$ (6)
where $`𝒞=\rho _fS(S+1)(g\mu _B)^2/3`$ is the Curie constant ($`\rho _f`$ is the number of magnetic atoms per unit of volume, $`S`$ is the magnetic angular momentum of the atom). At low enough temperatures ($`Tg\mu _BH`$) the susceptibility vanishes.
The effect of disorder in ordinary Fermi liquids has been studied in great detail and it was found that the results quoted above, especially the temperature dependences of the physical quantities, do not change appreciably (at least when disorder is weak enough to be treated in perturbation theory). Therefore, for a metallic system where Anderson localization does not play a role we still expect Fermi liquid behavior to be a robust feature. Thus, it is indeed very surprising that for such a broad class of U and Ce alloys (which clearly show three-dimensional behavior) deviations from Landau’s theory are so abundant.
In NFL systems it is usually observed that even in the paramagnetic phase the specific heat coefficient and the magnetic susceptibility do not saturate as expected from the Landau scenario. In all the cases studied so far the data for the susceptibility and specific heat have been fitted to weak power laws or logarithmic functions . The resistivity of the systems discussed here can be fitted with $`R(T)=R_0+AT^\alpha `$ where $`\alpha <2`$. Neutron scattering experiments in UCu<sub>5-x</sub>Pd<sub>x</sub> show that the imaginary part of the frequency dependent susceptibility, $`\mathrm{}(\chi (\omega ))`$, has power law behavior, that is, $`\mathrm{}(\chi (\omega ))\omega ^{1\lambda }`$ with $`\lambda 0.7`$, over a wide range of frequencies (for a Fermi liquid one expects $`\lambda =1`$). Moreover, consistent with this behavior the static magnetic susceptibility seems to diverge with $`T^{1+\lambda }`$ at low temperatures . What is interesting about UCu<sub>5-x</sub>Pd<sub>x</sub> is that it has been shown in recent EXAFS experiments that this compound has a large amount of disorder consistent with early NMR and $`\mu `$SR experiments .
From the magnetic point of view the U and Ce intermetallics show a rich variety of magnetic ground states: antiferromagnetic as in the case of UCu<sub>5-x</sub>Pd<sub>x</sub> for $`0x1`$ , ferromagnetic as in the case of U<sub>1-x</sub>Th<sub>x</sub>Cu<sub>2</sub>Si<sub>2</sub> for $`0.6x0.85`$ and spin glass as in the case of U<sub>1-x</sub>Y<sub>x</sub>Pd<sub>3</sub> for $`0.6x0.8`$ . There are many indications that the magnetism in these alloys is inhomogeneous with strong coupling to the lattice. This seems to be the case of CeAl<sub>3</sub> or CePd<sub>2</sub>Al<sub>3</sub> where $`\mu `$SR experiments have shown clear signs of microscopic inhomogeneity. Moreover, because of the strong spin-orbit coupling in these systems a myriad of magnetic states with different ordering configurations are found .
On the theoretical side we can classify the approaches to NFL as of single ion character or cooperative behavior. Single impurity approaches for the NFL problem are very important because the Kondo effect has been suggested as the source of heavy fermion behavior in undiluted ligand systems (say, CeAl<sub>3</sub>) . Nozières and Blandin proposed the multichannel Kondo effect as a possible source of NFL behavior. D. Cox proposed that such a multichannel effect of quadrupolar origin should exist in these systems . Another source of NFL behavior based on single impurity physics is the so-called Kondo disorder approach which postulates an a priori broad distribution of single ion Kondo temperatures . The cooperative approaches, on the other hand, stress the proximity to a critical point as a source of NFL behavior and was pioneered by Hertz and has been applied to many of the systems we discuss here . The problem of disorder close to the QCP has been discussed by many authors over the years. Hertz studied a classical XY model and showed that disorder is a relevant perturbation in dimensions smaller than $`4`$. Bhatt and Fisher and more recently Sachdev have argued that for Hubbard-like models an instability to a inhomogeneous phase should exist close to the QCP. A more recent work extending Hertz calculations to the case of disorder has reached similar conclusions, namely, that close to the QCP a new type of critical behavior, very different from the critical behavior of the clean system, should exist .
It is clear from the experimental and theoretical point of view that magnetic interactions and the Kondo effect should be relevant for the physics of U and Ce intermetallics. Moreover, it has been claimed for a long time that it is the competition between Kondo effect and magnetic order that determines the phase diagram of these systems. This idea was put forward by Doniach and we discuss it in more detail below. In trying to reconciliate the Doniach argument with the existence of disorder in these systems we proposed recently a new explanation for the NFL behavior in these systems which is based on the magnetic inhomogeneity due to this competition. In other words, if the Kondo effect is responsible for the destruction of long range order in the magnetically ordered system then at the QCP the competition between magnetic ordering and magnetic quenching is strongest . We have argued that close to the QCP, in the paramagnetic phase, finite clusters of magnetically ordered atoms can fluctuate quantum mechanically. In this case a so-called quantum Griffiths-McCoy singularities would be generated and zero temperature divergences of the physical quantities should be observed. In the next sections we explain exactly how this process occurs.
Classical Griffiths singularities appear in the context of classical Ising models as due to the response of rare and large clusters to an external magnetic field . A classical model with columnar disorder was solved exactly by McCoy and Wu in two dimensions and shows clearly the importance of Griffiths singularities close to the critical point for magnetic order. There are indeed experimental indications for the existence of such singularities in Fe<sub>0.47</sub>Zn<sub>0.53</sub>F<sub>2</sub> and other Ising magnets . Because of the relationship between the classical statistical mechanics in $`d`$ dimensions and the quantum statistical mechanics in $`d+1`$ dimensions the McCoy-Wu model can be mapped at $`T=0`$ into the transverse field Ising model which has been studied by D. Fisher . The transverse field Ising model is supposed to be applicable to insulating magnets since conduction electron degrees of freedom are not present. Senthil and Sachdev have studied the transverse field Ising model in higher dimensions on a percolating lattice and have found that quantum Griffiths singularities should be present in the paramagnetic phase in these systems close to the QCP . Moreover, many physical quantities diverge at zero temperature as a result of the strong response of the large and rare clusters. Such anomalous behavior close to the QCP has been confirmed numerically in dimensions higher than one and there is today strong experimental indication of such behavior in U and Ce intermetallics .
## III Effective Hamiltonian: RKKY $`\times `$ Kondo
In our discussion of the physics of U and Ce intermetallics we are going to use the so-called Kondo lattice Hamiltonian which describes the exchange interactions between itinerant electron spins and the localized magnetic moments:
$`H={\displaystyle \underset{𝐤,\kappa }{}}ϵ_\kappa (𝐤)c_{𝐤,\kappa }^{}c_{𝐤,\kappa }+{\displaystyle \underset{i}{}}{\displaystyle \underset{a,b,\kappa ,\kappa ^{}}{}}J_{a,b}(i)S^a(i)c_\kappa ^{}(i)\tau _{\kappa ,\kappa ^{}}^bc_\kappa ^{}(i)`$ (7)
where $`\kappa =1,2`$ labels the spin states, $`J_{a,b}(i)`$ are the effective exchange constants between the localized spins and conduction electrons at sites $`𝐫_i`$. $`S^a(i)`$ is the localized electron spin operator and $`_{\kappa ,\kappa ^{}}c_\kappa ^{}(i)\tau _{\kappa ,\kappa ^{}}^bc_\kappa ^{}(i)`$ the conduction electron spin operator where $`c_{𝐤,\kappa }`$ annihilates an electron with momentum $`𝐤`$ and spin projection $`\kappa `$ ($`\tau ^b`$ with $`a,b=x,y,z`$ are Pauli spin matrices). Observe that we are allowing for the possibility of the conduction electron dispersion $`ϵ_\kappa (𝐤)`$ to be dependent of the electron spin. Indeed, one can show that (7) can be obtained directly from the Anderson Hamiltonian when spin-orbit effects are included and the hybridization energy between localized and itinerant electrons is much smaller than the atomic energy scales . One of the striking features of (7) is the fact that the exchange interaction is not symmetric in spin space because of spin-orbit coupling. As we discuss below besides canted magnetism (7) gives rise to what we call a canted Kondo effect.
The main problem in studying the competition of Rudermann-Kittel-Kasuya-Yosida interaction (RKKY) and Kondo effect in the Hamiltonian (7) is related to the fact that both the RKKY and the Kondo effect have origin on the same magnetic coupling between spins and electrons. What allows us to treat this problem is the fact that the RKKY interaction is perturbative in $`J/E_F`$ while the Kondo effect is non-perturbative in this parameter and requires different techniques. Moreover, the RKKY interaction depends on electronic states deep inside of the Fermi sea, in addition to those at $`E_F`$, while the Kondo effect is purely a Fermi surface effect. Thus, it seems to be possible to use perturbation theory to treat the RKKY interaction.
We will consider, for simplicity the case where the Fermi surfaces for the two species of electrons are spherical. The local electron operator can be written in momentum space as
$`c_\kappa (i)={\displaystyle \frac{1}{\sqrt{N}}}{\displaystyle \underset{𝐤}{}}e^{i𝐤𝐫}c_{𝐤,\kappa }.`$ (8)
We now separate the states in momentum space into three different regions of energy as shown in Fig.2, namely, $`\mathrm{\Omega }_0`$: $`k_{F,\alpha }\mathrm{\Lambda }<k<k_{F,\alpha }+\mathrm{\Lambda }`$; $`\mathrm{\Omega }_1`$: $`k<k_{F,\alpha }\mathrm{\Lambda }`$; and $`\mathrm{\Omega }_2`$: $`k>k_{F,\alpha }+\mathrm{\Lambda }`$ where $`\mathrm{\Lambda }`$ is an arbitrary cut-off. As we discuss below the value of the cut-off is related to the number of compensated spins in the system. But for the moment being we consider it as some arbitrary quantity we can vary, like in a renormalization group calculation. Observe that the sum in (8) can also be split into these three different regions. The problem we want to address is how the states in region $`\mathrm{\Omega }_0`$ close to the Fermi surface renormalize as one traces out high energy degrees of freedom which are present in regions $`\mathrm{\Omega }_1`$ and $`\mathrm{\Omega }_2`$. We are going to do this calculation perturbatively in $`J/E_F`$ . For this purpose it is convenient to use a path integral representation for the problem and write the quantum partition function as
$`Z={\displaystyle D𝐒(n,t)D\overline{\psi }(𝐫,t)D\psi (𝐫,t)\mathrm{exp}\left\{i𝒮[𝐒,\overline{\psi },\psi ]\right\}}`$ (9)
in terms of Grassman variables $`\overline{\psi }`$ and $`\psi `$ and where the path integral over the localized spins also contains the constraint that $`𝐒^2(n,t)=S(S+1)`$. The quantum action in (9) can be separated into three different pieces, $`𝒮=𝒮_0[𝐒]+𝒮_0[\overline{\psi },\psi ]+𝒮_I[𝐒,\overline{\psi },\psi ]`$, where $`𝒮=𝒮_0[𝐒]`$ is the free actions of the spins (which can be written, for instance, in terms of spin coherent states ),
$`𝒮_0[\overline{\psi },\psi ]={\displaystyle \underset{\alpha ,\gamma }{}}{\displaystyle \frac{d\omega }{2\pi }\underset{𝐤}{}\overline{\psi }_\alpha (𝐤,\omega )\left(\omega +\mu ϵ_\alpha (𝐤)\right)\delta _{\alpha ,\gamma }\psi _\gamma (𝐤,\omega )}`$ (10)
is the free action for the conduction electrons and
$`𝒮_I[𝐒,\overline{\psi },\psi ]={\displaystyle 𝑑t\underset{\alpha ,\gamma }{}\underset{n}{}J_{a,b}(n)S_a(n,t)\tau _{\alpha ,\gamma }^b\overline{\psi }_\alpha (n,t)\psi _\gamma (n,t)}`$ (11)
is the exchange interaction between conduction electrons and localized moments. We now split the Grassman fields into the momentum shells defined above, that is, we rewrite the path integral as
$`Z={\displaystyle D𝐒(n,t)\underset{i=0}{\overset{2}{}}D\overline{\psi }_i(𝐫,t)D\psi _i(𝐫,t)\mathrm{exp}\left\{i𝒮[𝐒,\{\overline{\psi }_i,\psi _i\}]\right\}}`$ (12)
where the indices $`0,1,2`$ refer to the degrees of freedom which reside in the momentum regions $`\mathrm{\Omega }_0`$, $`\mathrm{\Omega }_1`$ and $`\mathrm{\Omega }_2`$, respectively. The action of the problem can be rewritten as $`𝒮=𝒮_0[𝐒]+_{i=0}^2𝒮_0[\overline{\psi }_i,\psi _i]+𝒮_I[𝐒,\overline{\psi },\psi ]`$. Notice the free part of the electron action is just a sum of three terms (essentially by definition since the non-interacting problem is diagonal in momentum space). Moreover, the exchange part mixes electrons in all three regions defined above:
$`𝒮_I={\displaystyle \underset{n}{}}{\displaystyle 𝑑t\underset{i,i^{}=0}{\overset{2}{}}J_{a,b}(𝐫_n)S_a(𝐫_n,t)\tau _{\alpha ,\gamma }^b\overline{\psi }_{\alpha ,i}(𝐫_n,t)\psi _{\gamma ,i^{}}(𝐫_n,t)}.`$ (13)
Since we are interested only on the physics close to the Fermi surface we trace out the fast electronic modes in the regions $`\mathrm{\Omega }_1`$ and $`\mathrm{\Omega }_2`$ assuming that $`J_{a,b}\mu `$. As we show in Appendix (A), besides the renormalization of the parameters in free action of the electrons in the region $`\mathrm{\Omega }_0`$, we get the RKKY interaction between localized moments. The effective action of the problem becomes:
$`𝒮_{eff}[𝐒,\overline{\psi }_0,\psi _0]`$ $`=`$ $`𝒮_0[\overline{\psi }_0,\psi _0]+{\displaystyle \underset{n}{}}{\displaystyle 𝑑t\underset{\alpha ,\gamma ,a,b}{}J_{a,b}^R(𝐫_n)S_a(𝐫_n,t)\tau _{\alpha ,\gamma }^b\overline{\psi }_{\alpha ,0}(𝐫_n,t)\psi _{\gamma ,0}(𝐫_n,t)}`$ (14)
$`+`$ $`{\displaystyle \underset{n,m,a,b}{}}{\displaystyle 𝑑t\mathrm{\Gamma }_{a,b}^R(𝐫_n𝐫_m,\mathrm{\Lambda })S_a(𝐫_n,t)S_b(𝐫_m,t)}`$ (15)
where $`\mathrm{\Gamma }_{a,b}^R(𝐫_n𝐫_m,\mathrm{\Lambda })`$ is the cut-off dependent RKKY interaction between the local moments and $`J_{a,b}^R(n)`$ is the Kondo electron-spin coupling renormalized by the high energy degrees of freedom. As we show in Appendix (A) this renormalization can be calculated exactly. For a spherical Fermi surface the exchange interaction between spins can be written as:
$`\mathrm{\Gamma }_{a,b}^R(r,\mathrm{\Lambda })={\displaystyle \frac{9c^2\overline{n}}{E_F}}{\displaystyle \underset{c}{}}J_{a,c}J_{b,c}(2k_Fr,\mathrm{\Lambda }/k_F)e^{r/\mathrm{}}`$ (16)
where $`c`$ is the number of magnetic moments per atom, $`\overline{n}`$ is the number of electrons per atom, $`\mathrm{}`$ is the electron mean-free path and $`(x,y)`$ is given in (A56). In Fig.3 we compare the usual RKKY with $`\mathrm{\Lambda }=0`$ and the RKKY with finite $`\mathrm{\Lambda }`$. At short distances, $`rk_F^1,\mathrm{\Lambda }^1`$, the interaction is ferromagnetic and is renormalized by a factor of $`(1\mathrm{\Lambda }/k_F)^3`$. Moreover, as shown in Fig.4 the first zero of the RKKY interaction is shifted from $`k_Fr2.2467`$ at $`\mathrm{\Lambda }=0`$ to smaller values. At intermediate distances, that is, $`\mathrm{\Lambda }^1r>k_F^1`$, the RKKY interaction decays like $`1/r^3`$ as in the case of $`\mathrm{\Lambda }=0`$ but the most striking result is that for large distances, $`r>>k_F^1,\mathrm{\Lambda }^1`$ the RKKY interactions decays like $`1/r^4`$ instead of the usual $`1/r^3`$. Thus, for finite $`\mathrm{\Lambda }`$ the RKKY interaction has a shorter range. Another interesting result of a finite $`\mathrm{\Lambda }`$ is that at short and intermediate distances the RKKY oscillations are mostly antiferromagnetic. As one can see directly from Fig.3 the ferromagnetic part of the RKKY is suppressed after the first zero. This could perhaps explain why most of the alloys of the type discussed here have antiferromagnetic ground states.
We observe further that the perturbation theory here is well behaved and there are no infrared singularities in the perturbative expansion. Thus, the limit of $`\mathrm{\Lambda }0`$ is well-defined. In this limit $`\mathrm{\Gamma }_{a,b}^R(𝐫_n𝐫_m,\mathrm{\Lambda }0)`$ becomes the usual RKKY interaction one would calculate by tracing all the energy shells of the problem. Observe that there are no retardation effects in tracing this high energy degrees of freedom since they are much faster than the electrons close to the Fermi surface and therefore adapt adiabatically to their motion. Hamiltonian (15) is the basic starting point to our approach and contains the basic elements for the discussion of magnetic order in the system. Observe that the RKKY interaction depends on the electronic states far away from the Fermi surface while the Kondo interaction is a pure Fermi surface effect. While the RKKY interaction leads to order of the magnetic moments the Kondo coupling induces magnetic quenching. It is the interplay of these two interaction which leads to the physics we discuss here. The action (15) has been used as a starting point for many theoretical discussions of rare earth alloys . We stress, however, that the action (15) describes only the low energy degrees of freedom of the problem and therefore the exchange constants that appear there can have strong renormalizations due to the high energy degrees of freedom. Moreover, as we explain below the cut-off $`\mathrm{\Lambda }`$ depends on the number of compensated moments.
### A Magnetically ordered phase: the role of RKKY
The existence of local moments is not a sufficient condition for the existence of long range magnetic order. It is exactly the interaction between the spins which determines the ordering temperature $`T_c`$ of the material. There are various ways localized moments can interact: dipolar-dipolar interactions, kinetic exchange interactions due to the overlap of the f orbitals and RKKY interaction. Dipolar interactions are too small to account for the ordering temperature in these systems (they range from $`100`$ K down to $`10`$ K in the pure compounds) and the direct exchange between f orbitals is very weak since the spatial extent of the f orbitals is small (with the possible exception of the 5f orbitals of U). The RKKY interaction is by far the most important interactions in metallic rare earth alloys and it will be the only interaction we will consider in detail.
As it is well-known and as shown in Appendix A in the ordered case ($`\mathrm{\Lambda }=0`$) the RKKY interaction decays like $`1/r^3`$ and oscillates in real space with wave-vector $`2k_F`$. The oscillatory terms have to do with the sharpness of the Fermi surface. Moreover, non-spherical Fermi surfaces will also lead to an angular dependence on the RKKY interaction with decaying rates which vary with the direction . The specific form of the RKKY interaction is not important in our discussion but the fact that the RKKY is an interaction which is perturbative in $`J/\mu `$ and scales like $`N(0)J^2`$. The exponential factor due to disorder was obtained originally by de Gennes .
Observe that the spin-orbit coupling generates an RKKY interaction which is anisotropic and therefore can give rise to canted magnetism. This effect is the analogue of the anisotropic spin exchange interaction, or Dzyaloshinsky-Moriya (DM) exchange interaction in insulating magnets which is obtained via the kinetic exchange between localized moments in the presence of spin-orbit coupling. The interaction (16) is an indirect exchange interaction in the presence of spin-orbit coupling. Thus, as in the case of DM interactions one expects parasitic ferromagnetism within antiferromagnetic phases. This effect has been observed long ago in R-Cr0<sub>3</sub> systems. The existence of ferromagnetic and antiferromagnetic coupling creates a very rich situation where many different magnetic phases are possible in the presence of a Lifshitz point . Recent theoretical approaches for the NFL problem in CeCu<sub>6-y</sub>Au<sub>y</sub> are based on the idea that the Lifshitz point in these systems is a QCP and therefore the quantum fluctuations associated with this point induce NFL behavior in the conduction band.
The critical temperature of the system can be estimated directly from the mean field theory for (7) and it is given by
$`T_c={\displaystyle \frac{2S(S+1)}{3}}{\displaystyle \underset{𝐑0}{}}\mathrm{\Gamma }_M(𝐑)\mathrm{cos}\left(𝐐𝐑\right)`$ (17)
where $`𝐐`$ is the ordering vector ($`Q=0`$ for ferromagnetism) and $`\mathrm{\Gamma }_M(𝐑)`$ is the largest eigenvalue of $`\mathrm{\Gamma }_{a,b}(𝐑)`$. Observe that $`T_c`$ scales with $`\mathrm{\Gamma }`$ and therefore it is proportional to $`N(0)J^2`$. This value of $`T_c`$ gives the order of magnitude of the transition temperature in Kondo hole or ligand systems, that is, the magnetically ordered Kondo lattice. In what follows we discuss the effect of disorder on the magnetic order in these systems.
### B Magnetic dilution
In Kondo hole systems the transition to the paramagnetic phase happens because the magnetic sublattice is diluted with non-magnetic atoms. In this case two main effects occur: (1) the magnetic system loses its magnetic atoms; (2) because the non-magnetic atoms do not have the same size of the magnetic ones there is a local lattice contraction or expansion. The first effect created by the dilution is to introduce disorder in the electronic environment and produce a finite scattering time $`\tau `$ for the electron. As shown by de Gennes long ago , the RKKY interaction decays exponentially with the electron mean-free path, $`\mathrm{}=v_F\tau `$. Notice that this is only true if there is true magnetic long range order in the problem. In the case of a spin glass order this is argument is not correct .
The problem of destruction of magnetic order in a ligand system is more complex since the magnetic atoms are not replaced. In an alloy like UCu<sub>4-x</sub>Pd<sub>x</sub> the Cu atoms are replaced by the somewhat smaller Pd atoms. This difference between the Pd and the Cu leads to a local lattice contraction which modifies the local hybridization matrix elements. Since these matrix elements are exponentially sensitive to the overlap between different angular momentum orbitals one can have large local effects in the system. This change of local matrix elements induces changes in the exchange constants between the conduction band and the localized moments, $`J(𝐫_i)`$, in (7).
Let us start with (7) and in the homogeneous ordered phase where we assume that $`J_{a,b}(𝐫_i)=J_{a,b}`$ for all sites. The transition temperature is given by (17) with $`\mathrm{\Gamma }`$ given by (16) with the electron mean free path determined by the extrinsic impurities in the system. As the Kondo lattice is doped, either by substitution of a magnetic atom by a non-magnetic one (as in the case of the Kondo hole systems) or just disordered (as in the case of ligand systems) the local coupling between the localized moments and electrons is changed from $`J_{a,b}`$ to a different value $`\stackrel{~}{J}_{a,b}`$ (for simplicity we will assume just a binary distribution but this assumption can be easily generalized). In this case can rewrite (7) as $`H=H_0+H_{KL}+H_{imp}`$ where:
$`H_0`$ $`=`$ $`{\displaystyle \underset{𝐤,\kappa }{}}ϵ_\kappa (𝐤)c_{𝐤,\kappa }^{}c_{𝐤,\kappa }`$ (18)
$`H_{KL}`$ $`=`$ $`{\displaystyle \underset{i}{}}{\displaystyle \underset{a,b,\kappa ,\kappa ^{}}{}}J_{a,b}S_a(𝐫_i)c_\kappa ^{}(𝐫_i)\tau _{\kappa ,\kappa ^{}}^bc_\kappa ^{}(i)`$ (19)
$`H_{imp}`$ $`=`$ $`{\displaystyle \underset{a,b,i}{}}p_i\delta J_{a,b}S_a(i)c_\kappa ^{}(i)\tau _{\kappa ,\kappa ^{}}^bc_\kappa ^{}(i)`$ (20)
where $`\delta J_{a,b}=\stackrel{~}{J}_{a,b}J_{a,b}`$ where the summation over $`i`$ includes all sites and $`p_i=1`$ if a particular site $`i`$ was changed by disorder and $`p_i=0`$ otherwise.
Let us consider first the case where the system the ordered state is just slightly doped. In first order in $`x`$ (the concentration) the problem reduces to a single ion problem. If $`|\delta J_{a,b}|>|J_{a,b}|`$ then one has to treat first $`H_0+H_{imp}`$ and then add $`H_{KL}`$. It is obvious that we have a Kondo effect on the sites for which $`p_i=1`$ with the original conduction band of the system. In the case of lattice contraction we would have $`\delta J_{a,b}>0`$ ($`\stackrel{~}{J}_{a,b}>J_{a,b}`$) and therefore an antiferromagnetic Kondo effect and a singlet is formed below a Kondo temperature $`T_KE_Fe^{1/(N(0)\delta J)}`$. All sites with $`p_i=1`$ are magnetically quenched. If $`\delta J_{a,b}<0`$ ($`\stackrel{~}{J}_{a,b}<J_{a,b}`$), which is the case of local lattice dilation, we would have a ferromagnetic Kondo effect and a local triplet state. Thus, there is an increase of the local magnetization of the system as a function of $`x`$. The effect here is very similar to the problem of enhancement of the magnetic moment by a highly polarizable metallic environment and creation of “giant moments” as it was discussed long ago by Jaccarino and Walker and observed experimentally in Pd<sub>1-x</sub>Ni<sub>x</sub> and other similar alloys .
As it is well-known in a disordered system, the electron acquires a life-time, $`\tau _K`$, due to the Kondo effect which, at zero temperature, reads :
$`{\displaystyle \frac{1}{\tau _K}}={\displaystyle \frac{3\pi (\delta J)^2S(S+1)x}{2E_F}}`$ (21)
where $`\delta J`$ is the largest eigenvalue of $`\delta J_{a,b}`$. This finite lifetime leads to a finite mean-free path, $`\mathrm{}_K=v_F\tau _K`$ for the conduction band motion. After the mean-free path is taken into account one proceeds as before (described in Appendix A) to calculate the RKKY interaction which will be given by (16) with $`1/\mathrm{}`$ substituted by $`1/\mathrm{}+1/\mathrm{}_K`$ as in Matthiessen’s rule.
In the opposite limit of $`|\delta J_{a,b}|<|J_{a,b}|`$ we have to consider $`H_0+H_{KL}`$ first and then add to it $`H_{imp}`$. This problem is more complicated because the ordered Kondo lattice problem has to be solved first. Here we just consider the simplest mean field theory in which the magnetic moments order along the $`z`$ axis. The mean field Hamiltonian can be written as $`H=H_{MF}+H_{imp}`$ where
$`H_{MF}`$ $`=`$ $`{\displaystyle \underset{𝐤,\kappa }{}}ϵ_\kappa (𝐤)c_{𝐤,\kappa }^{}c_{𝐤,\kappa }+{\displaystyle \underset{i}{}}\left[H_S\left(n_{i,}n_{i,}\right)+H_sS_z(𝐫_i)\right]`$ (22)
where $`H_S=J_zS_z(𝐫_i)`$ and $`H_s=J_z\left(n_{i,}n_{i,}\right)`$ are the molecular fields applied by the localized spins on the conduction electrons and by the conduction electrons on the localized spins, respectively.
The problem described by Hamiltonian (22) can be easily solved for the case of ferromagnetism as we show in Appendix (B). As a result the electronic degrees of freedom are renormalized in different ways depending on geometry of the Fermi surface and the type of ordering one has. For ferromagnetism the main change in the problem is the change in the density of states for different electron species. In the case of antiferromagnetism the situation can be more complicated because one generates a well defined momentum $`𝐐`$ and therefore Umklapp scattering is possible depending on the shape of the Fermi surface. A spin density wave state (SDW) can be generated and a gap can open on regions of the Fermi surface. Experimentally, the systems we are discussing here are metallic over the ordered phase which implies that the whole Fermi surface or perhaps large portions of it would remain gapless. This is easily understood by the fact that the conditions for commensurability are hard to obtain in these systems which have rather complicated Fermi surfaces. Therefore, the conclusions we reach for the ferromagnetic case can be easily generalized to more complicated magnetic structures.
Assuming that the system remains metallic in the magnetically ordered phase we see that $`H_K`$ describes the Kondo effect on this new metallic band in the presence of a magnetic field. We show in Appendix B that at $`T=0`$ the local magnetic energies are $`H_S=SJ_z`$ and $`H_s=SN(0)J_z^2/\rho _f`$. Observe that while the magnetic field applied on the electron, $`H_S`$, by the localized spin can be positive or negative depending if the local exchange is antiferromagnetic or ferromagnetic, the local field applied on the local spin, $`H_s`$, by the conduction electrons is always ferromagnetic. In the paramagnetic case ($`S_z=(n_{}n_{})=0`$) the local state of the system is degenerate. That is, one has a quartet, made out of $`|,`$, $`|,`$,$`|,`$ and $`|,`$ where $`,`$ represents the conduction band spin and $`,`$ the local moment spin. Since these are all eigenstates of $`S_z`$ and $`n_{}n_{}`$ their energies in term of the molecular fields are
$`|,`$ $``$ $`E_1=SJ_z{\displaystyle \frac{SN(0)J_z^2}{\rho _f}}`$ (23)
$`|,`$ $``$ $`E_2=SJ_z+{\displaystyle \frac{SN(0)J_z^2}{\rho _f}}`$ (24)
$`|,`$ $``$ $`E_3=SJ_z{\displaystyle \frac{SN(0)J_z^2}{\rho _f}}`$ (25)
$`|,`$ $``$ $`E_4=SJ_z+{\displaystyle \frac{SN(0)J_z^2}{\rho _f}}.`$ (26)
Observe that there is level crossing when $`J_z=\pm J_c=\pm \rho _f/N(0)`$. Furthermore, $`J_ccE_F`$ where $`c`$ is the number of local moments per atom. Since $`J_zE_F`$ we can only have $`J_zJ_c`$ when the density of local moments is very low (dilute limit). In general we expect $`J_zJ_c`$ in concentrated Kondo lattices.
When $`J_z>J_c`$ (antiferromagnetic coupling) the local state of the system is $`|,`$ which is separated from the state $`|,`$ by an energy amount $`\delta E=2SJ_z`$ and therefore the Kondo effect is suppressed (notice that $`\delta J<\delta E`$ and therefore the Kondo effect cannot bring these two states together). The problem is very similar to the usual ferromagnetic Kondo effect where quantum fluctuations are totally suppressed. Observe, however, that the local moment is compensated in the same way it would be if we just eliminate the local moment from the lattice. (Note that the electronic phase shift due to scattering by the impurity is zero in both cases.) Thus doping decreases the magnetization of the system as one would have in the usual dilution problem (the magnetic dilution of a ligand system is essentially identical to the dilution of the Kondo hole system). In the intermediate coupling regime of $`0<J_z<J_c`$ the two lowest energy states are $`|,`$ and $`|,`$ which are separated in energy $`\delta E=SN(0)J_z^2/\rho _f`$ and no Kondo effect happens. The ferromagnetic case is somewhat similar with the difference that the local state of the system is $`|,`$ but also in this case the Kondo effect does not take place because the electron spin states are quenched by the local molecular fields. Therefore, in a ferromagnetically ordered lattice the Kondo effect is absent independent of the sign of the Kondo coupling. This state of affairs is very similar to the one found by Larkin and Mel’nikov in the case of magnetic impurities in nearly ferromagnetic Fermi liquids . In summary, we conclude that the case of $`|\delta J_{a,b}|<|J_{a,b}|`$ there is no real Kondo effect in a magnetically ordered Kondo lattice. We note, however, that in the case of antiferromagnetic coupling the magnetization of the system drops because of a formation of $`|,`$ states.
We have seen that the ordered state of a Kondo lattice is destroyed via compensation of the magnetic moments either via the Kondo quenching or moment compensation. Thus, the percolation parameter in this problem is the density of quenched moments, $`\rho _Q`$. In percolation theory we assign a percolation parameter $`p`$ which in the case of the Kondo lattice is essentially $`\rho _Q`$. Let us now consider the situation of one of the alloys mentioned previously where we chemically substitute the atoms by an amount $`x`$. In a Kondo hole system the number of magnetic moments decreases with the alloying because the magnetic moments are replaced by non-magnetic atoms. At the same time the number of compensated moments grows because of the changes in the local structure of the lattice and the increase in the Kondo coupling. At some particular value of $`x`$, $`\rho _Q`$ reaches a maximum since it cannot grow beyond the actual number of magnetic moments which are left in the system. Moreover, at percolation threshold, $`p_c`$, which is determined by the dilution and lattice changes, the last infinite cluster disappears and long range order is lost. Beyond this point only finite clusters can exist. Eventually both the number of magnetic atoms and the number of compensated atoms drop to zero. This situation is depicted in Fig.5(a).
In a ligand system the density of magnetic moments is kept constant with chemical substitution because the magnetic atoms are not replaced. The number of compensated moments grows because of the local growth of the hybridization. At some critical value of $`\rho _Q`$ we reach the percolation threshold $`p_c`$ and long range order is lost because the last infinite cluster of uncompensated moments disappears. Moreover, when $`\rho _Q`$ grows beyond the threshold it will eventually reach the value $`\rho _f`$ and all the magnetic moments in the lattice are compensated. At this large value of doping one can find a heavy fermion ground state. We depict this situation in Fig.5(b).
We have to be careful in interpreting the heavy fermion ground state in light of the Kondo model we are studying. As we mentioned previously, the dilution in a Kondo hole system leads to a trivial Fermi liquid state while in a ligand system it can eventually lead to a Heavy Fermion (HF) state which is not straightforward to describe. Since the HF state is non-magnetic it is clear that dilution can drive the system to such a state by increasing the local hybridization of the conduction band with the localized f-electrons. If the hybridization becomes of the order of the local atomic energy scales the Kondo Hamiltonian (7) is not a good starting point for the description of the magnetic correlations. We should work directly with the Anderson Hamiltonian. This is usually the route taken by many approaches to the HF ground state . The f-electrons mix with the conduction band electrons and the Fermi surface is large since it counts all the electrons in the system. In what we have discussed we considered only the Kondo Hamiltonian which does not contain this kind of physics since the occupation of the f-atomic states is fixed to be $`1`$. Since we are interested mainly in the behavior of the system close to the magnetic ordered phase the Kondo Hamiltonian should give a good description of the problem but one has to be cautious about the transition from localized to itinerant behavior in these systems.
Assuming that the Kondo lattice is a good starting problem we immediately see that the Hamiltonian generated by (15) has the right properties associated with Doniach’s famous argument , namely, there are two main energy scales described in (15): the RKKY strength $`\mathrm{\Gamma }^R`$ and the Kondo temperature $`T_K`$ generated by $`J^R`$. These two energy scales are local properties of the alloying procedure and they scale in a very different way with the bare exchange between conduction electrons and magnetic moments. The RKKY coupling is proportional to $`(N(0)J(i))^2/E_F`$ in the weak coupling regime while $`J^RJ`$ and therefore $`T_K(i)E_Fe^{1/(N(0)J(i))}`$ (we are going to show below that this $`T_K`$ is slightly more complicated if we take anisotropy into account). If we plot these two quantities together as in Fig.6 we see that there is a critical value $`J_c`$ for the local exchange constant $`J(i)`$ above which the Kondo temperature is larger than the RKKY energy scale and therefore as the the system is cooled from high temperatures the moment is locally compensated and the RKKY interaction for that particular place becomes irrelevant in the limit of large $`J(i)`$. For $`J(i)`$ below $`J_c`$ the RKKY energy scale is larger and therefore as the system is cooled down the moment can locally order with its environment. (Note that for just two impurities, a partial Kondo effect occurs for finite $`J(i)`$, and the ground state is always a singlet, even for ferromagnetic RKKY coupling . Here, however, we are treating the case of a larger number of magnetic moments, tending toward a magnetic ground state for large RKKY.) We have to stress, however, that in the presence of disorder this effect is local and represents the a quantum percolation problem and has nothing to do with the homogeneous change in the exchange $`J`$! Indeed, if one interprets the chemical alloying of the system as a simple change of the exchange over the entire lattice then the Doniach argument would predict an ordering temperature $`T_N`$ as in Fig.6 which vanishes at a QCP where a transition from the ordered state to a fully Kondo compensated state happens. In this picture the QCP has nothing to do with a percolation problem where moments are compensated due to local effects. In our picture this is not possible since a local change in a coupling constant does not immediately imply a change of the “average coupling” constant. Thus, we believe that interpretations of the Doniach argument based on homogeneous changes in the couplings are actually erroneous.
We also would like to point out that the same discussion carried out in terms of Hamiltonian (7) can be done in terms of the effective action (15) by introducing a Hubbard-Stratonovich transformation and studying the saddle point equations by assuming a homogeneous solution for the spin field.
We have argued that by chemical substitution the order in a Kondo lattice can be taken away even when the substitution is not on the magnetic site because individual moments are compensated magnetically due to the distribution of exchange constants. At some finite concentration $`x^{}`$ long range order is lost and the system enters a paramagnetic phase. Since the problem at hand is a percolative one, the paramagnetic phase can still contain clusters of atoms in a relatively ordered state. In the rest of the paper we are going to discuss exactly the physics of these clusters and how they respond to external probes. We are going to show how finite clusters give rise to Griffiths-McCoy singularities in these systems.
### C The value of $`\mathrm{\Lambda }`$
In our previous discussion the cut-off $`\mathrm{\Lambda }`$ (cf. Fig. (2)) was considered as a free parameter but now we discuss this problem in more detail. It is clear from the above discussion that in the magnetically ordered phase we can set $`\mathrm{\Lambda }=0`$ because all the electrons on the Fermi sea are participating in the order of the magnetic moments. As one starts to dilute the Kondo lattice some particular sites with hybridization larger than average will be magnetically compensated by the Kondo effect. Therefore $`\mathrm{\Lambda }`$ has to increase in order to accommodate the electrons which participate in the Kondo coupling. Thus we expect $`\mathrm{\Lambda }`$ to increase with the number of compensated magnetic moments in the system.
Let $`\rho _Q`$ be the number of compensated moments per unit of volume at a given chemical concentration. As we have discussed previously this number is not given only by the Kondo effect alone but by the competition between the Kondo effect and the RKKY interaction in (15). If the electrons which participate in the Kondo effect are the ones in the region $`\mathrm{\Omega }_0`$ in Fig.2 then it is easy to see that
$`\rho _Q`$ $`=`$ $`\rho _e2{\displaystyle \frac{d^3k}{(2\pi )^3}\mathrm{\Theta }(k_F\mathrm{\Lambda }k)}`$ (27)
$`=`$ $`\rho _e{\displaystyle \frac{(k_F\mathrm{\Lambda })^3}{3\pi ^2}}`$ (28)
which can inverted to give (using (3))
$`{\displaystyle \frac{\mathrm{\Lambda }}{k_F}}=1\left(1{\displaystyle \frac{\rho _Q}{\rho _e}}\right)^{1/3}.`$ (29)
This equation gives a simple relationship between the number of compensated moments and the cut-off to be used in (15). When $`\rho _Q\rho _e`$ we have $`\mathrm{\Lambda }/k_F\rho _Q/(3\rho _e)1`$ and $`\mathrm{\Lambda }`$ is very small compared with the Fermi momentum. Naturally, since each magnetic atom in the unit cell gives at least one electron to the conduction band, we must have $`\rho _f\rho _e`$. This is true even when the electronic density changes as a function of the chemical substitution or percolation parameter $`p`$ as shown in Fig.5. Obviously, we have $`\rho _f\rho _Q`$ and therefore $`\rho _Q<\rho _e`$ from which follows that $`\mathrm{\Lambda }k_F`$. The extreme case of $`\rho _Q\rho _e`$ ($`\mathrm{\Lambda }k_F`$) indicates that Kondo process involves all electrons in the Fermi sea. Notice that the perturbative expansion in terms of $`J`$ in the effective Hamiltonian (15) is still valid but the form of the RKKY interaction will not be the usual one, that is, (16), since it will depend strongly on $`\mathrm{\Lambda }/k_F`$ (as shown in Appendix A).
Our discussion should be contrasted with the well-known exhaustion paradox proposed by Nozières : because the characteristic energy in the Kondo effect is $`T_K`$ the number of electrons participating in the Kondo effect is supposed to be
$`{\displaystyle \frac{\rho _K}{\rho _f}}{\displaystyle \frac{N(0)}{\rho _f}}T_K={\displaystyle \frac{3\rho _e}{2\rho _f}}{\displaystyle \frac{T_K}{E_F}}.`$ (30)
So, in order for all moments to be compensated one has to require this fraction to be $`1`$. But because $`T_KE_F`$ the condition in (30) is only observed at extremely small moment concentrations. As one increases $`\rho _f`$ there are not enough electrons at the Fermi surface to quench the magnetic moments and one should observe free magnetic moments. This paradox is based on an energetic argument which takes only the Kondo effect into account and is correct in the dilute limit. Indeed when $`\rho _F>\rho _Q0`$ we would have $`T_Kv_F\mathrm{\Lambda }`$. In the concentrated limit, where interactions start to play a role, it is misleading. The energy scales that produce $`\mathrm{\Lambda }`$ involve not only the Kondo coupling but also the RKKY interaction. Since the RKKY interaction involves states deep inside of the Fermi sea the energy scales involved in $`\mathrm{\Lambda }`$ can be rather large, of the order of the Fermi energy itself. Indeed, recent infrared conductivity measurements in YbInCu<sub>4</sub> indicate that a large fraction of the Fermi sea must be involved in the Kondo quenching in clear contradiction to the exhaustion paradox and in agreement with our discussion . The calculation of $`\rho _Q`$ thus involves a self-consistent calculation of the combined effect of Kondo and RKKY and goes beyond the scope of this paper.
## IV $`1,2,\mathrm{},N`$
We have argued in the previous section that the local changes in the hybridization will lead to a finite density $`\rho _Q`$ of compensated magnetic moments. The number of compensated moments depends strongly on lattice structure and the local changes in the electronic wavefunctions due to alloying. It is obvious, however, that alloying leads to a percolation problem and the number of magnetic moments drops as the ordered system moves towards the paramagnetic phase. Instead of working directly with $`\rho _Q`$ we define a percolation parameter $`p`$. For $`p<p_c`$ magnetic order exists and for $`p>p_c`$ long range order is not possible. $`p_c`$ is therefore the percolation threshold of the lattice. Thus, above $`p_c`$ there are only finite clusters of magnetic moments which are more coupled than the average. The probability of having $`N`$ moments together is given in percolation theory by
$`P(N,p)=N^{1\tau _d}f_p\left(N/N_\xi \right)`$ (31)
where $`N_\xi =(\xi (p)/a)^D`$ is the number of spins within a correlation length $`\xi `$, $`d`$ is the spatial dimension of the lattice, $`\tau _d`$ is a percolation exponent ($`\tau _2=187/912.05`$, $`\tau _32.18`$), $`D`$ is the fractal dimension of the cluster ($`\tau _d=1+d/D`$). Close to percolation threshold $`p_c`$ the system is critical and therefore
$`{\displaystyle \frac{\xi (p)}{a}}{\displaystyle \frac{1}{|pp_c|^\nu }}`$ (32)
where $`\nu `$ is the correlation length exponent. The scaling function $`f_p(x)`$ is such that $`f_p(x0)=1`$ and for $`x1`$ one has
$`f_p(x)x^{\theta _{d,p}+\tau _d}e^{c_px^{\zeta _p}}`$ (33)
where $`\theta _{2,p>p_c}=1`$, $`\theta _{3,p>p_c}=3/2`$, $`\theta _{2,p<p_c}=5/4`$ and $`\theta _{3,p<p_c}=1/9`$, $`c_p`$ is a constant of order of unit, $`\zeta _{p>p_c}=1`$ and $`\zeta _{p<p_c}=11/d`$. Observe that the exponential behavior given in (33) is the dominant part of the probability.
The above equations are easy to understand in the dilute limit, that is, when $`p>>p_c`$. If $`c`$ is the concentration of magnetic atoms in the systems then the probability of having $`N`$ atoms together is
$`P(N,p>>p_c)c^N=e^{N\mathrm{ln}(1/c)}`$ (34)
and the probability is exponentially small. In particular, the probability of finding $`N`$ nearest neighbor atoms in a lattice with coordination number $`Z`$ is
$`P(N,Z)={\displaystyle \frac{Z!}{N!(ZN)!}}c^N(1c)^{ZN}`$ (35)
which reduces to (34) when $`c0`$.
It is also important to understand what happens close to percolation threshold ($`pp_c0^+`$) where from (31) and (33) we have
$`P(N,pp_c)\mathrm{exp}\left\{{\displaystyle \frac{c_{p_c}N}{(\xi (p)/a)^D}}\right\}.`$ (36)
Therefore, from (34) and (36), the exponential part of the probability can be written in a simple form
$`P(N,p)e^{\kappa _pN^{\zeta _p}}`$ (37)
where
$`\kappa _{pp_c}`$ $``$ $`\mathrm{ln}(1/c)`$ (38)
$`\kappa _{pp_c}`$ $``$ $`(\xi (p)/a)^D|pp_c|^{\nu D}0.`$ (39)
It is obvious from the above equations that the mean size of the clusters diverges at $`p=p_c`$ and that for $`p>p_c`$ the cluster size is determined by $`\xi (p)`$. In this paper we are mainly interested in the physics of clusters in the paramagnetic case ($`p>p_c`$). For a given concentration $`c`$ there are going to be clusters of all sizes but with different probabilities. For instance, for a cubic system ($`Z=6`$), the number of isolated atoms ($`N=0`$) becomes smaller than the number of dimers ($`N=1`$) only when the concentration is larger than $`c>0.14`$. In understanding the response of such an inhomogeneous to external probes we have to understand how a particular cluster responds to these probes. We are going to assume that the clusters only couple weakly to each other and that in first approximation they can be thought of isolated and permeated by a paramagnetic matrix. The weak coupling among the clusters, as we discuss at the end of the paper, can lead to glassy states which do not show strong deviations from a Fermi liquid ground state. In order to build up intuition about the physics of the clusters we discuss the case of $`1`$ and $`2`$ magnetic atoms in a paramagnetic matrix. After the discussion it will become quite clear how a cluster of $`N`$ atoms should behave in the presence of a paramagnetic environment (this is the $`N`$ impurity cluster Kondo effect).
### A $`1`$ Magnetic Moment Kondo Effect
The one impurity Kondo effect should occur in the case of Kondo hole systems at very low magnetic atom concentration and according to the argument given in Subsection III A also in the ligand systems in the ordered phase. Our discussion now will follow very closely the approach to the single impurity problem via bosonization techniques . In order to do so we reduce (22) to an effective one dimensional problem via the bosonization technique.
The bosonization procedure follows three different steps. In the first step we trace out the electrons far away from the Fermi surface exactly as in Section III. Since there is just one magnetic moment in the problem no RKKY term is generated and the only effect of this trace is to renormalize the Kondo couplings $`J`$. Indeed, as shown by Anderson et al. for the one impurity Kondo problem the renormalization of the Kondo coupling is given by
$`J_z^R=8v_F\delta (J_z)`$ (40)
where $`\delta (J_z)`$ is the phase shift of the electrons due to the scattering of a static impurity (corresponding to the Ising component of (15)) which is given by
$`\delta (J_z)=\mathrm{arctan}\left({\displaystyle \frac{\pi N(0)J_z}{4}}\right).`$ (41)
On a second stage we linearize the electron dispersion close to the Fermi surface:
$`E_{𝐤,\sigma }=E_F+v_{F,\sigma }(|𝐤|k_{F,\sigma }),`$ (42)
where we are considering the generic case where the spin flavors can have different Fermi velocities $`v_{F,\sigma }`$. In order to linearize the problem we also have to introduce a momentum cut-off, $`\mathrm{\Lambda }`$, which is a non-universal constant (anisotropies in the shape of the Fermi surface can be absorbed in $`\mathrm{\Lambda }`$). The conduction band Hamiltonian is written as
$`H_C={\displaystyle \underset{p,\sigma }{}}v_{F,\sigma }pc_{p,\sigma }^{}c_{p,\sigma }`$ (43)
where $`c_{p,\sigma }`$ creates an electron with spin $`\sigma `$, momentum $`|𝐤|=p+k_F`$ and angular momentum $`l=0`$. Moreover, since we are treating the problem of a single impurity it involves electrons moving in a single direction associated with incoming or outgoing waves. Thus, in writing (43) we have reduced the problem to an effective one-dimensional problem. In order to do it we introduce right, $`R`$, and left, $`L`$, moving electron operators
$`\psi _{R(L),\sigma }(x)={\displaystyle \frac{1}{\sqrt{2\pi }}}{\displaystyle _\mathrm{\Lambda }^\mathrm{\Lambda }}𝑑kc_{k\pm k_{F,\sigma },\sigma }e^{ikx}`$ (44)
which are used to express the electron operator as
$`\psi _\sigma (x)=\psi _{R,\sigma }(x)e^{ik_{F,\sigma }x}+\psi _{L,\sigma }(x)e^{ik_{F,\sigma }x}.`$ (45)
In any impurity problem the right and left moving operators produce a redundant description of the problem since they are actually equivalent to incoming or outgoing waves out of the impurity. Therefore we have two options: either we work with right and left movers in half of the line or we work in the full line but impose the condition $`\psi _{R,\sigma }(x)=\psi _{L,\sigma }(x)`$. Following tradition we choose the latter approach. Thus, from now on we drop the symbol $`R`$ from the problem and work with left movers only. The left mover fermion can be bosonized as
$`\psi _\sigma (x)={\displaystyle \frac{K_\sigma }{\sqrt{2\pi a}}}e^{i\mathrm{\Phi }_\sigma (x)}`$ (46)
where
$`\mathrm{\Phi }_\sigma (x)={\displaystyle \underset{p>0}{}}\sqrt{{\displaystyle \frac{\pi }{pL}}}((b_p+\sigma a_p)e^{i\sigma px}h.c.)`$ (47)
and $`K_\sigma `$ is a factor which preserves the correct commutation relations between electrons, that is, $`\{\psi _\sigma (x),\psi _\sigma ^{}(y)\}=\delta (xy)\delta _{\sigma ,\sigma ^{}}`$ and the bosons obey canonical commutation relations $`[a_k,a_p^{}]=[b_k,b_p^{}]=\delta _{p,k}`$.
In what follows we are going to assume the simple case in which $`J_{a,b}=J_a\delta _{a,b}`$ where $`J_z<J_x<J_y`$. In terms of the boson operators, the Kondo Hamiltonian (15) becomes
$`H`$ $`=`$ $`\overline{v}_F{\displaystyle \underset{p>0}{}}p(a_p^{}a_p+b_p^{}b_p)+\delta v_F{\displaystyle \underset{p>0}{}}p\left(b_p^{}a_p+a_p^{}b_p\right)+J_z^RS^z{\displaystyle \underset{k>0}{}}\sqrt{{\displaystyle \frac{k}{\pi L}}}(a_k+a_k^{})`$ (48)
$`+`$ $`{\displaystyle \frac{J_+}{4\pi a}}(S^+K_{}^{}K_{}e^{_{k>0}\sqrt{\frac{4\pi }{kL}}(a_ka_k^{})}+h.c.)`$ (49)
$`+`$ $`{\displaystyle \frac{J_{}}{4\pi a}}(S^{}K_{}^{}K_{}e^{_{p>0}\sqrt{\frac{4\pi }{kL}}(a_ka_k^{})}+h.c.)`$ (50)
where $`\overline{v}_F=(v_{F,}+v_{F,})/2`$ is the average Fermi velocity, $`\delta v_F=(v_{F,}v_{F,})/2`$ is the mismatch between Fermi velocities in different spin branches, $`J_+=(J_x+J_y)/2`$ and $`J_{}=(J_xJ_y)/2`$. Notice that because we are allowing for different Fermi velocities the charge and spin degrees of freedom do not decouple from each other. Moreover, this Hamiltonian can be brought to a simpler form if one performs a unitary transformation
$`U=e^{S^z_{k>0}\sqrt{\frac{\pi }{kL}}(a_ka_k^{})}`$ (51)
which transforms the Hamiltonian to
$`H^{}`$ $`=`$ $`U^1HU=\overline{v}_F{\displaystyle \underset{p>0}{}}p(a_p^{}a_p+b_p^{}b_p)+\delta v_F{\displaystyle \underset{p>0}{}}p\left(b_p^{}a_p+a_p^{}b_p\right)`$ (52)
$`+`$ $`S^z\left[\left(J_z^R\pi v_F\right){\displaystyle \underset{k>0}{}}\sqrt{{\displaystyle \frac{k}{\pi L}}}(a_k+a_k^{})+\delta v_F{\displaystyle \underset{k>0}{}}\sqrt{{\displaystyle \frac{k}{\pi L}}}(b_k+b_k^{})\right]`$ (53)
$`+`$ $`{\displaystyle \frac{J_+}{4\pi a}}\left(S^+K_{}^{}K_{}+S^{}K_{}^{}K_{}\right)`$ (54)
$`+`$ $`{\displaystyle \frac{J_{}}{4\pi a}}(S^{}K_{}^{}K_{}e^{4_{p>0}\sqrt{\frac{\pi }{kL}}(a_ka_k^{})}+h.c.).`$ (55)
An important observation here is that the unitary transformation does not affect $`S^z`$. The anti-commutation factors can be rewritten in terms of spin operators in which case we can rewrite
$`H^{}`$ $`=`$ $`\overline{v}_F{\displaystyle \underset{p>0}{}}p(a_p^{}a_p+b_p^{}b_p)+\delta v_F{\displaystyle \underset{p>0}{}}p\left(b_p^{}a_p+a_p^{}b_p\right)+{\displaystyle \frac{J_+}{2\pi a}}\sigma _x`$ (56)
$`+`$ $`\sigma _z\left[\left(J_z^R\pi \overline{v}_F\right){\displaystyle \underset{k>0}{}}\sqrt{{\displaystyle \frac{k}{\pi L}}}(a_k+a_k^{})+\delta v_F{\displaystyle \underset{k>0}{}}\sqrt{{\displaystyle \frac{k}{\pi L}}}(b_k+b_k^{})\right]`$ (57)
$`+`$ $`{\displaystyle \frac{J_{}}{4\pi a}}(\sigma ^+e^{4_{p>0}\sqrt{\frac{\pi }{kL}}(a_ka_k^{})}+h.c.)`$ (58)
where $`\sigma _x=(\sigma ^++\sigma ^{})/2`$. Observe that (58) describes the physics of a two level system coupled to a bosonic environment .
When $`J^z2v_F`$ and $`J_{}=0`$ (the limit of large uniaxial anisotropy) we see from (40) that $`J_R^z\pi v_F`$ ($`\delta v_F=0`$) and the Hamiltonian reduces to
$`H={\displaystyle \frac{J_{}}{2\pi a}}\sigma _x+v_F{\displaystyle \underset{p>0}{}}pa_p^{}a_p`$ (59)
with the decoupling of the spin degrees of freedom to the bosonic modes ($`J_x=J_y=J_+=J_{}`$). This is the dissipationless limit of the problem. Observe that in this limit the eigenstates of the system are eigenstates of $`\sigma _x`$, that is, the transverse field.
The physical interpretation in terms of the Kondo problem is also very straightforward: in the limit of $`J^z2v_F`$ the electron forms a virtual bound state with the localized spin. There are two states which are degenerate for an antiferromagnetic coupling
$`|+`$ $`=`$ $`|,`$ (60)
$`|`$ $`=`$ $`|,`$ (61)
which are the eigenstates of $`\sigma ^z`$. This degeneracy is lifted by the transverse field $`\sigma ^x`$ and one ends up with two states
$`|s`$ $`=`$ $`{\displaystyle \frac{1}{\sqrt{2}}}(|+|)={\displaystyle \frac{1}{\sqrt{2}}}(|,|,)`$ (62)
$`|t`$ $`=`$ $`{\displaystyle \frac{1}{\sqrt{2}}}(|++|)={\displaystyle \frac{1}{\sqrt{2}}}(|,+|,)`$ (63)
which are the singlet and triplet states.
Thus, we can define a tunneling splitting, $`\mathrm{\Delta }_0`$, between the two magnetic states and the coupling constant to the bosonic bath, $`\alpha `$, which are given by
$`\mathrm{\Delta }_0`$ $`=`$ $`{\displaystyle \frac{J_{}}{2\pi a}}`$ (64)
$`\alpha `$ $`=`$ $`\left[1{\displaystyle \frac{J_R^z}{\pi v_F}}\right]^2=\left[1{\displaystyle \frac{2}{\pi }}\mathrm{arctan}\left({\displaystyle \frac{\pi N(0)J_z}{4}}\right)\right]^2.`$ (65)
For simplicity we will consider the case of uniaxial symmetry in which $`J_{}=0`$ and $`v_{F,}=v_{F,}`$ so we rewrite (58) as
$`H^{}=\mathrm{\Delta }_0\sigma _x\pi v_F\sqrt{\alpha }\sigma _z{\displaystyle \underset{k>0}{}}\sqrt{{\displaystyle \frac{k}{\pi L}}}(a_k+a_k^{})+{\displaystyle \underset{k>0}{}}v_Fka_k^{}a_k.`$ (66)
Observe that the magnetic moment flips at a rate given by $`1/\mathrm{\Delta }_0`$. The bosons with energy much larger than $`\mathrm{\Delta }_0`$ (that is, with momentum close to $`\mathrm{\Lambda }`$) will follow adiabatically the motion of the spin and their effect is to renormalize the fluctuation rate. The low energy bosons (that is, with $`k0`$) are too slow and cannot follow the motion of the spin. The spin can dissipate energy into this slow bosonic bath. In order to take into account the effect of the fast bosons into the motion of the spin we can consider the effects of bosons which live in a thin shell $`\mathrm{\Lambda }d\mathrm{\Lambda }<k<\mathrm{\Lambda }`$ and treat their coupling to the spin in perturbation theory. It is simple to show that in this case the bare tunneling splitting $`\mathrm{\Delta }_0`$ is renormalized to
$`\mathrm{\Delta }_R=E_c\left({\displaystyle \frac{\mathrm{\Delta }_0}{E_c}}\right)^{\frac{1}{1\alpha }}`$ (67)
where $`\mathrm{\Delta }_0=\mathrm{\Delta }(\mathrm{\Lambda }_0)`$ is the value of the tunneling at the bare value of the coupling constant and $`E_cE_F`$ is a high energy cut-off. The above discussion is valid at zero temperature. At finite temperatures the renormalization group flow has to stop at $`max(T,\mathrm{\Delta }_R)`$ because after this point there are no bosonic modes to renormalize anymore. Thus, there is a crossover in the problem when the temperature becomes of order $`TT_R=\mathrm{\Delta }_R`$ so that for temperatures larger than $`T_R`$ the tunneling is suppressed and for temperatures smaller than $`T_R`$ the tunneling is renormalized to $`\mathrm{\Delta }_R`$. This crossover temperature can be associated with the Kondo temperature as we discuss below.
It turns out that the physics of (66) is now understood and many physical quantities can be computed exactly . For instance, it was found that,
$`\sigma _z(t)=\sigma _z(0)e^{\mathrm{\Gamma }t}\mathrm{cos}(\mathrm{\Omega }t)`$ (68)
where
$`\mathrm{\Gamma }(\alpha ,\mathrm{\Delta }_0)`$ $`=`$ $`{\displaystyle \frac{2T_K}{\pi }}\mathrm{sin}^2\left({\displaystyle \frac{\pi \alpha }{2(1\alpha )}}\right)`$ (69)
$`\mathrm{\Omega }(\alpha ,\mathrm{\Delta }_0)`$ $`=`$ $`{\displaystyle \frac{T_K}{\pi }}\mathrm{sin}\left({\displaystyle \frac{\pi \alpha }{(1\alpha )}}\right)\mathrm{\Theta }(1/2\alpha )`$ (70)
where
$`K_BT_K={\displaystyle \frac{T_R}{\alpha }}={\displaystyle \frac{E_c}{\alpha }}\left({\displaystyle \frac{\mathrm{\Delta }_0}{E_c}}\right)^{\frac{1}{1\alpha }}`$ (71)
is the actual Kondo temperature of the system which depends on a non-universal cut-off energy scale $`E_c`$. Observe that the ratio
$`{\displaystyle \frac{\mathrm{\Omega }}{\mathrm{\Gamma }}}=\mathrm{cot}\left({\displaystyle \frac{\pi \alpha }{2(1\alpha )}}\right)`$ (72)
only depends on $`\alpha `$ and it is universal. Moreover, it was shown that the zero temperature contribution of the magnetic moment to the susceptibility is simply
$`\chi _{imp}(T=0)={\displaystyle \frac{1}{2\pi \mathrm{\Delta }_R}}={\displaystyle \frac{1}{2\pi \alpha T_K}}`$ (73)
while the contribution to the specific heat is
$`{\displaystyle \frac{C_{imp}(T)}{T}}={\displaystyle \frac{\pi }{3}}{\displaystyle \frac{\alpha }{\mathrm{\Delta }_R}}={\displaystyle \frac{\pi }{3}}{\displaystyle \frac{1}{T_K}}.`$ (74)
The imaginary part of the frequency dependent susceptibility is given by
$`\mathrm{}\left[\chi _{imp}(\omega )\right]={\displaystyle \frac{8\mathrm{\Omega }\mathrm{\Gamma }}{\alpha }}\mathrm{sin}\left({\displaystyle \frac{\pi \alpha }{2(1\alpha )}}\right){\displaystyle \frac{\omega }{\left(\omega ^2\mathrm{\Omega }^2+\mathrm{\Gamma }^2\right)^2+4\mathrm{\Gamma }^2\mathrm{\Omega }^2}}\mathrm{tanh}({\displaystyle \frac{\beta \omega }{2}})`$ (75)
where $`\beta =1/T`$. Notice that the Kramers-Kronig relation ($`\chi (0)=(1/\pi )𝑑\omega ^{}𝒫(1/\omega ^{})\mathrm{}[\chi (\omega ^{})]`$), leads to (73). At high temperatures, $`TT_K`$, we have, for instance,
$`\chi _{imp}(T)`$ $``$ $`{\displaystyle \frac{1}{4T}}`$ (76)
$`C_{imp}(T)`$ $``$ $`\left({\displaystyle \frac{\mathrm{\Delta }_R}{T}}\right)^{2(1\alpha )}.`$ (77)
The physics of the single impurity Kondo problem follows immediately from the exact solution (68): the electron spin acts as a transverse field on the local moment (see (66)) which flips (precesses) at a rate given by $`\mathrm{\Omega }`$. In the process of flipping the magnetic moment produces particle-hole excitations in the Fermi sea which lead to an effective damping of the flipping process which is given by $`\mathrm{\Gamma }`$. Observe that the flipping process is underdamped when $`\alpha <1/3`$ (since $`\mathrm{\Omega }(\alpha =1/3)=\mathrm{\Gamma }(\alpha =1/3)`$) and it is overdamped when $`1/2<\alpha <1/3`$. Oscillations cease to exist when $`\alpha >1/2`$. Going back to the Kondo problem we see that the oscillations can be classified as: underdamped if $`N(0)J_z>0.996`$, overdamped if $`0.63>N(0)J_z>0.996`$ and damped if $`N(0)J_z<0.63`$. Finally, just to make connection with the SU(2) Kondo problem let us observe that for $`J_z,J_{}E_c`$ the Kondo temperature looks very similar to the SU(2) expression $`T_KE_c\mathrm{exp}\{1/(N(0)J)\}`$. Indeed, from (71) we have
$`T_KE_c\mathrm{exp}\left\{{\displaystyle \frac{\mathrm{ln}[1/(N(0)J_{})]}{N(0)J_z}}\right\}`$ (78)
which is the form of the Kondo temperature for the anisotropic Kondo problem. Observe that the Kondo temperature of an anisotropic Kondo problem is not a single parameter quantity since it depends on the Ising component $`J_z`$ and the XY component given by $`J_{}`$.
### B $`2`$ Magnetic Moments Kondo Effect
The single impurity Kondo problem discussed in the last subsection gives us a hint on how a cluster of coupled moments should behave, that is, one would expect renormalizations of the tunneling energy due to dressing of high energy particle-hole excitations and dissipation due to the low energy particle-hole excitations. It is however too simple because it does not contain the RKKY coupling between spins. The next level of complexity is the two impurity Kondo model which has the first features of a cluster problem.
We consider the problem of two magnetic atoms at a distance $`𝐑`$ from each other interacting via an RKKY interaction $`\mathrm{\Gamma }_{a,b}=\mathrm{\Gamma }_a\delta _{a,b}`$ in the presence of a metallic host, that is, the so-called two impurity Kondo problem which is described by (15):
$`H_d`$ $`=`$ $`{\displaystyle \underset{𝐤,\alpha }{}}ϵ_kc_{𝐤,\alpha }^{}c_{𝐤,\alpha }+{\displaystyle \underset{a}{}}\mathrm{\Gamma }_a^R(R)S_a(𝐑/2)S_a(𝐑/2)`$ (79)
$`+`$ $`{\displaystyle \underset{𝐤,𝐤^{},a}{}}J_a^R\left[e^{i(𝐤𝐤^{})𝐑/2}c_{𝐤,\alpha }^{}\sigma _{\alpha ,\beta }^ac_{𝐤^{},\beta }S_a(𝐑/2)+e^{i(𝐤𝐤^{})𝐑/2}c_{𝐤,\alpha }^{}\sigma _{\alpha ,\beta }^ac_{𝐤^{},\beta }S_a(𝐑/2)\right]`$ (80)
where the couplings are renormalized by the trace over high energy degrees of freedom.
In order to study this problem we reduce it to an one-dimensional problem by rewriting the electron operators as
$`\psi _{j,\alpha }(k)={\displaystyle \frac{k}{\sqrt{2}}}{\displaystyle 𝑑\mathrm{\Omega }\left[\frac{1}{N_e(k)}\mathrm{cos}\left(\frac{𝐤𝐑}{2}\right)+i(1)^j\frac{1}{N_o(k)}\mathrm{sin}\left(\frac{𝐤𝐑}{2}\right)\right]c_{𝐤,\alpha }}`$ (81)
where $`j=1,2`$ and $`\mathrm{\Omega }`$ is the solid angle and
$`N_{e(o)}(k)=\sqrt{1\pm {\displaystyle \frac{\mathrm{sin}(kR)}{kR}}}`$ (82)
and $`\{\psi _{j,\alpha }^{}(k),\psi _{l,\beta }(k^{})\}=2\pi \delta (kk^{})\delta _{\alpha ,\beta }\delta _{j,l}`$. Furthermore, observe that all the momenta here are defined in a thin shell around the Fermi surface. Thus, we can linearize the band by writing $`ϵ_k=v_F(k_Fk)`$ and rewrite the whole Hamiltonian close to the Fermi surface as (see Appendix D):
$`H`$ $`=`$ $`iv_F{\displaystyle \underset{j,\alpha }{}}{\displaystyle 𝑑x\psi _{j,\alpha }^{}(x)\frac{}{x}\psi _{j,\alpha }(x)}+{\displaystyle \underset{a}{}}\mathrm{\Gamma }_a^RS_{1,a}S_{2,a}+V{\displaystyle \underset{j,l,\alpha }{}}\psi _{j,\alpha }^{}(0)\tau _{j,l}^x\psi _{l,\alpha }(0)`$ (83)
$`+`$ $`{\displaystyle \frac{v_F}{2}}{\displaystyle \underset{a,j,l\alpha ,\beta }{}}[J_{+,a}\psi _{j,\alpha }^{}(0)\delta _{j,l}\sigma _{\alpha ,\beta }^a\psi _{l,\beta }(0)S_a+J_{m,a}\psi _{j,\alpha }^{}(0)\tau _{j,l}^z\sigma _{\alpha ,\beta }^a\psi _{l,\beta }(0)\delta S_a`$ (84)
$`+`$ $`J_{,a}\psi _{j,\alpha }^{}(0)\tau _{j,l}^x\sigma _{\alpha ,\beta }^a\psi _{l,\beta }(0)S_a]`$ (85)
where $`\tau `$ are Pauli matrices which act in the states $`j=1,2`$, $`J_{s,a}`$ are exchange constants, $`\psi _{j,\alpha }(x)`$ is the Fourier transform of (81) and
$`S_a`$ $`=`$ $`S_{1,a}+S_{2,a}`$ (86)
$`\delta S_a`$ $`=`$ $`S_{1,a}S_{2,a}.`$ (87)
Moreover, we have explicitly introduced a new coupling constant $`V`$ associated with impurity scattering and which breaks the particle-hole symmetry. This kind of term is unavoidable in any realistic model of impurities and plays an important role in what follows.
Exactly as in the case of the single impurity given in (46) we bosonize the problem but take into account that besides $``$ and $``$ spin states we also have $`j=1,2`$. This leads to four types of bosonic fields, namely ,
$`\varphi _c(x)`$ $`=`$ $`(\varphi _{1,}(x)+\varphi _{1,}(x)+\varphi _{2,}(x)+\varphi _{2,}(x))/2`$ (88)
$`\varphi _s(x)`$ $`=`$ $`(\varphi _{1,}(x)\varphi _{1,}(x)+\varphi _{2,}(x)\varphi _{2,}(x))/2`$ (89)
$`\varphi _f(x)`$ $`=`$ $`(\varphi _{1,}(x)+\varphi _{1,}(x)\varphi _{2,}(x)\varphi _{2,}(x))/2`$ (90)
$`\varphi _{sf}(x)`$ $`=`$ $`(\varphi _{1,}(x)\varphi _{1,}(x)\varphi _{2,}(x)+\varphi _{2,}(x))/2`$ (91)
which are associated with the charge, spin, flavor and spin-flavor currents of the problem. And again, like in the one impurity Kondo problem, we perform a rotation in the spin space in order to eliminate the spin currents. This can be accomplished with the unitary transformation
$`U=e^{i(S_{1,z}+S_{2,z})\mathrm{\Phi }_s(0)}`$ (92)
in which case the Hamiltonian becomes
$`H`$ $`=`$ $`{\displaystyle \frac{v_F}{2}}{\displaystyle \underset{i=s,f,sf}{}}{\displaystyle 𝑑x\left[\mathrm{\Pi }_i^2(x)+\left(\frac{\varphi _i}{x}\right)^2\right]}+\stackrel{~}{\mathrm{\Gamma }}_zS_{1,z}S_{2,z}`$ (93)
$`+`$ $`\mathrm{\Gamma }_{}(S_{1,x}S_{2,x}+S_{2,y}S_{1,y})+{\displaystyle \frac{2V}{\pi a}}\mathrm{cos}(\mathrm{\Phi }_{sf}(0))\mathrm{cos}(\mathrm{\Phi }_f(0)\theta )`$ (94)
$`+`$ $`{\displaystyle \frac{v_F\stackrel{~}{J}_{+,z}}{2\pi }}{\displaystyle \frac{\mathrm{\Phi }_s(0)}{x}}S_z+{\displaystyle \frac{v_FJ_{+,}}{\pi a}}\mathrm{cos}(\mathrm{\Phi }_{sf}(0))S_x`$ (95)
$`+`$ $`{\displaystyle \frac{v_FJ_{m,z}}{2\pi }}{\displaystyle \frac{\mathrm{\Phi }_{sf}(0)}{x}}\delta S_z{\displaystyle \frac{v_FJ_{m,}}{\pi a}}\mathrm{sin}(\mathrm{\Phi }_{sf}(0))\delta S_y`$ (96)
$`+`$ $`{\displaystyle \frac{v_FJ_{,z}}{\pi a}}\mathrm{sin}(\mathrm{\Phi }_f(0)\theta )\mathrm{sin}(\mathrm{\Phi }_{sf}(0))S_z+{\displaystyle \frac{v_FJ_,}{\pi a}}\mathrm{cos}(\mathrm{\Phi }_f(0)\theta )S_x`$ (97)
where
$`\mathrm{\Phi }_i(x)`$ $`=`$ $`\sqrt{\pi }\left(\varphi _i(x){\displaystyle _{\mathrm{}}^x}𝑑y\mathrm{\Pi }_i(y)\right)`$ (98)
$`\theta `$ $`=`$ $`{\displaystyle \frac{1}{2}}{\displaystyle _{\mathrm{}}^+\mathrm{}}𝑑x{\displaystyle \frac{}{x}}(\mathrm{\Phi }_c(x)\mathrm{\Phi }_{sf}(x))`$ (99)
and
$`\stackrel{~}{\mathrm{\Gamma }}_z`$ $`=`$ $`\mathrm{\Gamma }_z^R{\displaystyle \frac{2v_F}{\pi a}}(J_{+,z}\pi )`$ (100)
$`\stackrel{~}{J}_{+,z}`$ $`=`$ $`J_{+,z}2\pi `$ (101)
are the new renormalized couplings. Observe that (97) has strong similarities with its one impurity counterpart (66) but the bosons have not decoupled from the transverse field terms because of the appearance of bosonic modes $`\mathrm{\Phi }_{sf}`$. The most studied case of this Hamiltonian is associated with the NFL fixed point which is not our main interest since it does not describe the cluster physics we are looking for. The main problem with the NFL fixed point is that it is unstable to the particle-hole symmetry breaking operator $`V`$ defined in (97). This operator is relevant under the renormalization group and drives the system away from the NFL fixed point. In the presence of scattering the system flows to a line of fixed points which represent the different phase shifts. This line of fixed points can be reached in different ways as we discuss below
Let us consider the case of $`V0`$. As shown in refs. $`V`$ grows under the RG and accordingly to (97) the value of the $`\mathrm{\Phi }_{sf}`$ and $`\mathrm{\Phi }_f`$ fields freeze at values such that
$`\mathrm{\Phi }_{sf}(0)=m\pi `$ (102)
$`\mathrm{\Phi }_f(0)=\pi n+\theta `$ (103)
where $`m`$ and $`n`$ are integers. By its definition in (99) we see that $`\theta `$ is purely topological and depends only on the value of the fields at infinity, that is, depends on the boundary conditions. Thus it is clear that $`\theta `$ is related to the phase shift the electrons acquire by scattering from the impurities. Therefore to each value of $`\theta `$ we have a different fixed point and (103) shows that in the absence of particle-hole symmetry the system flows to a line of fixed points as $`\theta `$ varies in between $`0`$ and $`\pi /2`$. Using (103) we find that (97) simplifies to:
$`H`$ $`=`$ $`\stackrel{~}{\mathrm{\Gamma }}_zS_{1,z}S_{2,z}+\mathrm{\Gamma }_{}\left(S_{1,x}S_{2,x}+S_{1,y}S_{2,y}\right)+{\displaystyle \frac{v_FJ_{}}{\pi a}}S_x`$ (104)
$`+`$ $`{\displaystyle \frac{v_F}{2}}{\displaystyle \underset{i=s,sf}{}}{\displaystyle 𝑑x\left[\mathrm{\Pi }_i^2(x)+\left(\frac{\varphi _i}{x}\right)^2\right]}+{\displaystyle \frac{v_F\stackrel{~}{J}_{+,z}}{2\pi }}{\displaystyle \frac{\mathrm{\Phi }_s(0)}{x}}S_z+{\displaystyle \frac{v_FJ_{m,z}}{2\pi }}{\displaystyle \frac{\mathrm{\Phi }_{sf}(0)}{x}}\delta S_z`$ (105)
where $`J_{}=J_{+,}+J_,`$. Notice that the flavor degree of freedom decouples in this limit. This Hamiltonian has a form which is very similar to the two level system (66) we have studied previously. The main difference is that the spin operators for each impurity appear in a linear combination and therefore are not naively related to the two level system problem.
The only trivial limit of the the Hamiltonian (105) is when $`J\mathrm{\Gamma }`$ in which case the Kondo coupling has to be treated first. The impurities decouple and we essentially have two independent Kondo effects . We are interested in the opposite limit since in the cluster the RKKY is stronger than the Kondo effect.
In order to understand the physics of Hamiltonian (105) one needs to consider the type of RKKY interaction we are dealing with. Let us first consider the case where $`\stackrel{~}{J}_{+,z}=J_{m,z}=0`$. The problem is purely magnetic since the bosons decouple from the spins. We can diagonalize the magnetic Hamiltonian since the it describes the simple problem of two interacting spins in a transverse field. When $`J_{}=0`$ the system decouples into two triplets, $`|,`$ and $`|,`$ with $`S_z=+1`$ and $`1`$, respectively, and with energy $`\stackrel{~}{\mathrm{\Gamma }}_z/4`$; a $`S_z=0`$ triplet, $`(|,+|,)/\sqrt{2}`$, with energy $`\stackrel{~}{\mathrm{\Gamma }}_z/4+\mathrm{\Gamma }_{}/4`$; and a singlet
$`|+,1={\displaystyle \frac{1}{\sqrt{2}}}(|,|,),`$ (106)
with energy $`\stackrel{~}{\mathrm{\Gamma }}_z/4\mathrm{\Gamma }_{}/4`$. On the one hand, the transverse field $`J_{}`$ splits the degeneracy of the $`|S_z|=1`$ triplets down to
$`|+,2={\displaystyle \frac{1}{\sqrt{2}}}(|,|,)`$ (107)
with energy $`\stackrel{~}{\mathrm{\Gamma }}_z/4`$ and (not normalized)
$`|,2={\displaystyle \frac{1}{\sqrt{2}}}(|,+|,)+ϵ(|,|,)`$ (108)
with energy $`\sqrt{(\stackrel{~}{\mathrm{\Gamma }}_z/4\mathrm{\Gamma }_{}/4)^2+\stackrel{~}{J}_{}^2}+\mathrm{\Gamma }_{}/4`$ where $`\stackrel{~}{J}_{}=v_FJ_{}/(2\pi a)`$. The coefficient $`ϵ`$ is a normalization coefficient which depends on the relation between energy scales. On the other hand, while the singlet $`|+,1`$ is still an eigenstate of the problem with energy $`\stackrel{~}{\mathrm{\Gamma }}_z/4\mathrm{\Gamma }_{}/4`$ and we get a new state
$`|,1={\displaystyle \frac{1}{\sqrt{2}}}(|,|,)+ϵ(|,+|,)`$ (109)
with energy $`\sqrt{(\stackrel{~}{\mathrm{\Gamma }}_z/4\mathrm{\Gamma }_{}/4)^2+\stackrel{~}{J}_{m,}^2}+\mathrm{\Gamma }_{}/4`$. The splitting of the energy levels is shown in the diagram Fig. 7.
Let us consider first the case of antiferromagnetic coupling where $`\stackrel{~}{\mathrm{\Gamma }}_z>0`$ and large ($`\stackrel{~}{\mathrm{\Gamma }}_z\mathrm{\Gamma }_{},\stackrel{~}{J}_{m,}`$). The lowest energy levels are $`|+,1`$ and $`|,1`$ which are separated in energy by $`\mathrm{\Gamma }_{}/2+2\stackrel{~}{J}_{}^2/\stackrel{~}{\mathrm{\Gamma }}_z\stackrel{~}{J}_{}^2/\stackrel{~}{\mathrm{\Gamma }}_z`$ (since $`\mathrm{\Gamma }_{}J_{}^2/E_F`$ and $`E_F\stackrel{~}{\mathrm{\Gamma }}_z`$). Moreover, $`ϵ\stackrel{~}{J}_{}/\stackrel{~}{\mathrm{\Gamma }}_z1`$ and therefore the main effect of the transverse field is to split the degeneracy of the $`|,`$ and $`|,`$ states. At low temperatures we just have to keep these two low lying states (we always assume $`T\stackrel{~}{\mathrm{\Gamma }}_z`$) and introduce back the couplings $`J_{+,z}=J_{m,z}`$. We see from (105) that $`\mathrm{\Phi }_s`$ decouples from the spins since
$`S_z|\sigma ,\sigma =(S_{1,z}+S_{2,z})|\sigma ,\sigma =0`$ (110)
while the coupling $`J_{m,z}`$ gets a renormalization of a factor of $`2`$ since
$`\delta S_z|\sigma ,\sigma =(S_{1,z}S_{2,z})|\sigma ,\sigma =2\sigma |\sigma ,\sigma `$ (111)
which just tells that the sub-Hilbert space is spanned by the states $`|,`$ and $`|,`$. The effective Hamiltonian can be written as
$`H_{AF}={\displaystyle \frac{v_F}{2}}{\displaystyle 𝑑x\left[\mathrm{\Pi }_{sf}^2(x)+\left(\frac{\varphi _{sf}}{x}\right)^2\right]}2{\displaystyle \frac{v_FJ_{m,z}}{2\pi }}{\displaystyle \frac{\mathrm{\Phi }_{sf}(0)}{x}}\sigma _z+{\displaystyle \frac{\stackrel{~}{J}_{}^2}{\stackrel{~}{\mathrm{\Gamma }}_z}}\sigma _x`$ (112)
where
$`\sigma _x|1,+`$ $`=`$ $`+|1,+`$ (113)
$`\sigma _x|1,`$ $`=`$ $`|1,`$ (114)
$`\sigma _z|1,+`$ $`=`$ $`+|1,`$ (115)
$`\sigma _z|1,`$ $`=`$ $`+|1,+.`$ (116)
Thus (112) has essentially the same physics as the initial Hamiltonian and its physical meaning is obvious, namely, it describes a dissipative two level system problem (66) where the two levels represent the two states $`|,`$ and $`|,`$ of the magnetic moments, in order words, it describes a Kondo effect of the conduction band with the antiferromagnetic cluster made out of two local moments. Observe that the heat bath for the antiferromagnetic cluster is made out of the spin-flavor bosons. From the above Hamiltonian we immediately conclude that the Kondo temperature of the antiferromagnetic cluster is given by
$`T_K(J_{m,},J_{m,z})`$ $`=`$ $`{\displaystyle \frac{E_c}{\alpha (J_{m,z})}}\left({\displaystyle \frac{v_F^2J_{}^2}{4\pi ^2a^2\stackrel{~}{\mathrm{\Gamma }}_zE_c}}\right)^{\frac{1}{1\alpha (J_{m,z})}}`$ (117)
$`\alpha (J_{m,z})`$ $`=`$ $`4{\displaystyle \frac{J_{m,z}^2}{(2\pi )^2}}=(N(0)J_z^R)^2\left[1\left({\displaystyle \frac{\mathrm{sin}(k_FR)}{k_FR}}\right)^2\right]`$ (118)
where we used (D10). Observe that we have extracted a factor of $`4`$ in front of $`\alpha `$ just to stress that $`\alpha `$ is coming from the coupling of the staggered moment of two impurities and therefore $`\alpha `$ scales like $`2^2\alpha `$ since $`\sqrt{\alpha }2\sqrt{\alpha }`$ in contrast with the single impurity case.
In the ferromagnetic case the situation is reversed since the two lying states due to the RKKY and transverse field are $`|2,+`$ and $`|2,`$ which are split from each other in energy by $`\sqrt{(\stackrel{~}{\mathrm{\Gamma }}_z/4\mathrm{\Gamma }_{}/4)^2+\stackrel{~}{J}_{}^2}+\mathrm{\Gamma }_{}/4\stackrel{~}{\mathrm{\Gamma }}_z/4\stackrel{~}{J}_{}^2/\stackrel{~}{\mathrm{\Gamma }}_z`$. Moreover, we have
$`S_z|\sigma ,\sigma =(S_{1,z}+S_{2,z})|\sigma ,\sigma =2|\sigma ,\sigma `$ (119)
$`\delta S_z|\sigma ,\sigma =(S_{1,z}S_{2,z})|\sigma ,\sigma =0`$ (120)
and therefore from (105) the $`\mathrm{\Phi }_{sf}`$ fields decouple in the ferromagnetic case and the effective Hamiltonian in the sub-Hilbert space spanned by the triplet states $`|,`$ and $`|,`$ reads,
$`H_F={\displaystyle \frac{v_F}{2}}{\displaystyle 𝑑x\left[\mathrm{\Pi }_s^2(x)+\left(\frac{\varphi _s}{x}\right)^2\right]}2{\displaystyle \frac{v_F\stackrel{~}{J}_{+,z}}{2\pi }}{\displaystyle \frac{\mathrm{\Phi }_s(0)}{x}}\sigma _z+{\displaystyle \frac{\stackrel{~}{J}_{}^2}{\stackrel{~}{\mathrm{\Gamma }}_z}}\sigma _x`$ (121)
where
$`\sigma _x|2,+`$ $`=`$ $`+|2,+`$ (122)
$`\sigma _x|2,`$ $`=`$ $`|2,`$ (123)
$`\sigma _z|2,+`$ $`=`$ $`+|2,`$ (124)
$`\sigma _z|2,`$ $`=`$ $`+|2,+.`$ (125)
Notice again that (121) describes a Kondo effect between the two low-lying states of a ferromagnetic cluster where the heat bath is provided $`\mathrm{\Phi }_s`$. The Kondo temperature of the ferromagnetic cluster is given by
$`T_K(J_{m,},J_{+,z})`$ $`=`$ $`{\displaystyle \frac{E_c}{\alpha (J_{+,z})}}\left({\displaystyle \frac{v_F^2J_{}^2}{4\pi ^2a^2\stackrel{~}{\mathrm{\Gamma }}_zE_c}}\right)^{\frac{1}{1\alpha (J_{+,z})}}`$ (126)
$`\alpha (J_{+,z})`$ $`=`$ $`4{\displaystyle \frac{(\stackrel{~}{J}_{+,z})^2}{(2\pi )^2}}=4\left(1{\displaystyle \frac{N(0)J_z^R}{2}}\right)^2`$ (127)
where we used (101) and (D10). Notice the close resemblance of (127) and the single impurity problem where $`\alpha `$ is given by (65). The only difference, like in the antiferromagnetic case, is a number in front which is associated with the number of spins involved. Indeed, the ferromagnetic problem is more directly related to the usual Kondo effect than the antiferromagnetic case. A direct comparison between the antiferromagnetic (118) and the ferromagnetic case (127) reveals a basic difference between the two problems: while dissipation scales like $`(N(0)J_z)^21`$ in the antiferromagnetic case it scales like $`1N(0)J_z1`$ in the ferromagnetic case. Thus, for the same coupling constants the ferromagnetic case has a stronger coupling to the bosonic bath. The antiferromagnetic, on the other hand, is weakly coupled and therefore dissipation is weaker. The same type of effect occurs in the problem of tunneling of magnetic grains where dissipation is more important for ferromagnetic than for antiferromagnetic granular systems .
The conclusion of the two-impurity Kondo problem is that the stable fixed points of the system represent either a ferromagnetic or an antiferromagnetic cluster undergoing a Kondo effect with the conduction band. The effective spin in this case is associated with the low lying doublet generated by the RKKY interaction which is split by the XY component of the Kondo coupling. Because the splitting of the doublet requires the flip of two spins the transverse field scales like the $`J_{}^2`$ in direct contrast with the single impurity case where it is a linear function of the XY coupling. Moreover, the coupling to the bath is also modified since the flip of two spins leads to the production of more particle-hole excitations close to the Fermi surface. Although there is no way to define long range order for the two impurity problem it is clear that when the RKKY is dominant it is the “order parameter”, magnetization or staggered magnetization depending on the sign of the RKKY interaction, which couples to the heat bath. This trend seems to be easily generalized to more than two impurities, that is, although NFL fixed points are probably possible due to specific symmetries of the problem, when symmetries are broken a simpler Kondo effect of many moments is possible where the moments flip coherently in the spin field generated by the Kondo coupling.
### C $`N`$ Magnetic Moments - XYZ Magnetism
Consider now the problem of $`N`$ magnetic moments forming a cluster close to the QCP. If $`N`$ is large we can envisage this cluster as a large magnetic grain. The ground state of the grain is just the classical one, that is, it is the fully ordered state for a ferromagnet ($`|,,\mathrm{},`$ or $`|,,\mathrm{},`$) or the Neél state for an antiferromagnet ($`|,,,\mathrm{}`$ or $`|,,,\mathrm{}`$). Because of time reversal symmetry the ground state of a magnetic system has to be at least double degenerate. Like in the examples of the single impurity or two impurity the cluster can fluctuate quantum mechanically between the two degenerate states in the absence of an applied magnetic field which breaks explicitly the symmetry and bias one of the configurations.
The tunneling between degenerate states can have many origins. In the preceding sections we have discussed the tunneling due to the XY coupling of the Kondo interaction to the magnetic moment. Another more “mundane” source of interaction is the magnetic anisotropy in XYZ magnets. This kind of anisotropy exists even in insulating magnets and has to do with the interplay between spin-orbit and crystal field effects. To understand the origin of this anisotropy we can just look at the RKKY interaction alone in (15). Observe that the RKKY interaction commutes with $`𝐒_T^2`$ where $`𝐒_T=_{i=1}^N𝐒_i`$ is the total spin of the cluster. Thus, the eigenstates of the problem can be classified accordingly to the eigenstates of $`𝐒_T^2`$, that is, $`S_T(S_T+1)`$ where $`S_T=0,1,2,\mathrm{},N/2`$ if $`N`$ is even or $`S_T=1/2,3/2,..,N/2`$ if $`N`$ is odd. Because the cluster is in the ordered state it will select $`S_T`$ accordingly to the interactions ($`S_T=N/2`$ for a ferromagnet and $`S_T=0`$ or $`1/2`$ for an antiferromagnet). If the cluster is large it will behave like a magnetic grain and the total spin operator $`𝐒_T`$ behaves like a classical variable. Let us consider the simplest case of a ferromagnetic cluster (the antiferromagnetic case is essentially analogous) with $`N`$ atoms. Since the atoms are all locked together we can describe their spins as classical variables:
$`S_{z,i}`$ $`=`$ $`S_T\mathrm{sin}(\theta )\mathrm{cos}(\varphi )`$ (128)
$`S_{x,i}`$ $`=`$ $`S_T\mathrm{sin}(\theta )\mathrm{sin}(\varphi )`$ (129)
$`S_{y,i}`$ $`=`$ $`S_T\mathrm{cos}(\theta )`$ (130)
where $`\theta `$ is the angle the spins make with the $`Y`$ axis and $`\varphi `$ is the angle in the $`XZ`$ plane. The RKKY interaction between the spins is given by (15) with $`\mathrm{\Gamma }_{a,b}=\mathrm{\Gamma }_z\delta _{a,b}`$ with $`|\mathrm{\Gamma }_z|>|\mathrm{\Gamma }_x|>|\mathrm{\Gamma }_y|`$ are the principal magnetic axis of the crystal. The energy due to the RKKY interaction can be rewritten in terms of the angles as
$`E(\theta ,\varphi )=N\left[(\overline{\mathrm{\Gamma }}_y\overline{\mathrm{\Gamma }_x})\mathrm{cos}^2(\theta )+(\overline{\mathrm{\Gamma }}_z\overline{\mathrm{\Gamma }}_x)\mathrm{sin}^2(\theta )\mathrm{cos}^2(\varphi )+\overline{\mathrm{\Gamma }}_x\right]`$ (131)
where $`\overline{\mathrm{\Gamma }}_a=_i\mathrm{\Gamma }_a(i)`$ is the average exchange within the cluster. The energy (131) describes a magnet with $`Z`$ easy axis and a $`XZ`$ easy plane if $`0>\overline{\mathrm{\Gamma }}_y>\overline{\mathrm{\Gamma }_x}>\overline{\mathrm{\Gamma }}_z`$. The energy of the cluster is minimized at $`\varphi =0,\pi `$ and $`\theta =\pi /2`$ (that is, all the spins point along the Z axis). It is usual to rewrite (131) as
$`E(\theta ,\varphi )=N\left[K_{}\mathrm{cos}^2(\theta )K_{||}\mathrm{sin}^2(\theta )\mathrm{cos}^2(\varphi )\right]`$ (132)
where $`K_{}=\overline{\mathrm{\Gamma }}_y\overline{\mathrm{\Gamma }_x}>0`$ and $`K_{||}=\overline{\mathrm{\Gamma }}_x\overline{\mathrm{\Gamma }}_z>0`$ are the anisotropy energies of the cluster (we have subtracted an unimportant constant from (132)).
The dynamics of a cluster described by the energy (132) is given by the well-known Landau-Lifshitz equations :
$`{\displaystyle \frac{d𝐒_T}{dt}}=𝐒_T\times {\displaystyle \frac{\delta E}{\delta 𝐒_T}}`$ (133)
which, in terms of the angle variables, are
$`{\displaystyle \frac{d\varphi }{dt}}`$ $`=`$ $`{\displaystyle \frac{1}{S_T\mathrm{sin}\theta }}{\displaystyle \frac{E}{\theta }}`$ (134)
$`{\displaystyle \frac{d\theta }{dt}}`$ $`=`$ $`{\displaystyle \frac{1}{S_T\mathrm{sin}\theta }}{\displaystyle \frac{E}{\varphi }}.`$ (135)
Of course, such equations do not allow for any tunneling. To study tunneling for such a magnetic grain we have to allow for solutions which are not classical in nature. The simplest way to do it is by the path integration method in imaginary time. The generating functional for the magnetic grain can be written as
$`Z={\displaystyle 𝑑\mu [\varphi ]𝑑\mu [\theta ]e^{𝒮_E}}`$ (136)
where $`S_E`$ is the Euclidean action
$`𝒮_E={\displaystyle 𝑑\tau \left\{iS_T[1\mathrm{cos}(\theta )]\frac{d\varphi }{d\tau }+E(\theta ,\varphi )\right\}}.`$ (137)
The tunneling process is now described as an instanton solution of the equations of motion for $`𝒮_E`$ which interpolate between the minima of the potential. It can be shown that the tunneling energy is given by
$`\mathrm{\Delta }_A=\omega _0|\mathrm{cos}(\pi S_T)|\mathrm{exp}\left\{2S_T\mathrm{ln}\left[\sqrt{1+K_{||}/K_{}}+\sqrt{K_{||}/K_{}}\right]\right\}`$ (138)
where
$`\omega _0=2\sqrt{K_{||}(K_{||}+K_{})}`$ (139)
is the attempt frequency of the cluster. Since $`S_TN/2`$ we see that the tunneling splitting $`\mathrm{\Delta }`$ can be written as
$`\mathrm{\Delta }=\omega _0e^{\gamma N}`$ (140)
where
$`\gamma `$ $`=`$ $`\mathrm{ln}\left[\sqrt{1+K_{||}/K_{}}+\sqrt{K_{||}/K_{}}\right]`$ (141)
$``$ $`{\displaystyle \frac{1}{2}}\mathrm{ln}\left({\displaystyle \frac{K_{||}}{K_{}}}\right).`$ (142)
Notice that the result (140) is what one would expect from a WKB calculation for the tunneling splitting of $`N`$ atoms. The factor of $`\mathrm{cos}(\pi S_T)`$ has to do with the Kramer’s theorem: if $`S_T`$ is an integer (even number of spins) the magnetization can tunnel and split the degeneracy of the ground state; if $`S_T`$ is a half-integer (odd number of electrons) tunneling is not allowed and the degeneracy of the ground state remains. This term has its origin in the first term in the Euclidean action (137) and it is topological in origin . Furthermore, the splitting is exponentially small in the number of spins in the cluster since $`S_TN`$ for a ferromagnet. Notice that tunneling is only possible if $`K_{}0`$ (that is, $`\overline{\mathrm{\Gamma }}_x\overline{\mathrm{\Gamma }}_y`$) which requires a magnetic cluster with very low spin isotropy, that is, XYZ magnetism. In an isotropic or uniaxial magnet (Heisenberg or XXZ, respectively) tunneling is suppressed. As we show below, only the Kondo effect can lead to tunneling. We also observe that at finite temperatures the system can be thermally activated from one minimum to another. The process has the usual Ahrenius factor $`\mathrm{exp}\{\beta NK_{||}\}`$ related to the jump of the system over an energy barrier of height $`NK_{||}`$. Comparing this exponent with the one found in (138) we see that there is a temperature $`T_A`$ above which thermal activation dominates and below which quantum tunneling dominates. This temperature is approximately given by
$`T_A{\displaystyle \frac{NK_{||}}{2S_T\mathrm{ln}\left[\sqrt{1+K_{||}/K_{}}+\sqrt{K_{||}/K_{}}\right]}}.`$ (143)
All the arguments presented here can be easily generalized to the case of antiferromagnetic grain .
In a metallic substrate dissipation due to particle-hole excitations in the conduction band plays an important role. Dissipation can be introduced in the problem via (15) . Since within the cluster the moments are locked together we can rewrite (15) in Hamiltonian form as
$`H=\mathrm{\Delta }_A\sigma _x+{\displaystyle \underset{𝐤,\alpha }{}}ϵ_kc_{𝐤,\alpha ,0}^{}c_{𝐤,\alpha ,0}+{\displaystyle \underset{a,b}{}}\overline{J}_{a,b}M_a\sigma _a\tau _{\alpha ,\gamma }^b{\displaystyle \underset{𝐤,𝐪}{}}F_𝐪c_{𝐤+𝐪,\alpha ,0}^{}c_{𝐤,\gamma ,0}`$ (144)
where $`\overline{J}_{a,b}`$ is the average exchange in the cluster, $`M_a`$ is the order parameter ($`S_a(𝐫)=M_a\sigma _a\mathrm{cos}(𝐐𝐫)`$ where $`𝐐`$ is the ordering vector in the cluster) and
$`F_𝐪={\displaystyle \underset{i=1}{\overset{N}{}}}\mathrm{cos}(𝐐𝐫_i)e^{i𝐪𝐫_i}`$ (145)
is the form factor of the cluster.
The Hamiltonian (144) describes the Kondo scattering of the electrons by the cluster in the presence of a transverse field. As shown in ref. the transverse field is a relevant perturbation in the problem and all the processes associated with the XY component of the Kondo effect are irrelevant. In other words, in the presence of tunneling due to the anisotropy in the RKKY interaction the cluster flipping due to the Kondo effect does not play an important role. We have seen, however, that the splitting due to the RKKY interaction vanishes for Heisenberg or XXZ magnets and therefore the Kondo scattering becomes relevant. Thus, in the presence of $`\mathrm{\Delta }_A`$ we just have to keep the component of $`M_a`$ in the ordering direction (say, $`z`$). Because the Kondo coupling does not play a role any longer it is possible to apply perturbation theory to Hamiltonian (144). It is easy to see that (144) maps again into the dissipative two level system (66) with $`\mathrm{\Delta }_0`$ replaced by $`\mathrm{\Delta }_A`$ and the dissipative constant $`\alpha `$ is given by a Fermi surface average :
$`\alpha `$ $`=`$ $`2(N(0)\overline{J}_z)^2M_z^2{\displaystyle \frac{d𝐫}{V_c}\frac{d𝐫^{}}{V_c}\mathrm{cos}(𝐐𝐫)\mathrm{cos}(𝐐𝐫^{})\left[\frac{\mathrm{sin}(k_F|𝐫𝐫^{}|)}{k_F|𝐫𝐫^{}|}\right]^2}`$ (146)
$`=`$ $`2(N(0)\overline{J}_z)^2M_z^2{\displaystyle \frac{1}{2V_c}}(\delta _{𝐐,0}+1)\mathrm{}\left\{{\displaystyle 𝑑𝐫e^{i𝐐𝐫}\frac{\mathrm{sin}^2(k_Fr)}{k_F^2r^2}}\right\}`$ (147)
where the integrals are performed in the volume $`V_c`$ of the cluster. Thus, we introduce a cut-off term $`e^{r/L}`$ (so that $`V_c=8\pi L^3`$) and perform the integration exactly:
$`\alpha `$ $`=`$ $`(N(0)\overline{J}_z)^2M_z^2[{\displaystyle \frac{\delta _{𝐐,0}}{1+(2k_FL)^2}}`$ (148)
$`+`$ $`{\displaystyle \frac{1}{4L^3k_F^2Q}}(\mathrm{arctan}(QL){\displaystyle \frac{\mathrm{arctan}[(Q+2k_F)L]}{2}}{\displaystyle \frac{\mathrm{arctan}[(Q2k_F)L]}{2}})].`$ (149)
The physical situation here is quite interesting. Although the XY coupling of the Kondo effect does not play a role, the anisotropy of the RKKY interaction introduces a transverse field. Thus, we can map back the two level system problem to a pure Kondo effect of the cluster! The only difference is that now in this “fake” Kondo effect the transverse coupling (what we called $`J_{}`$ in (65)) is related to $`\mathrm{\Delta }_A`$. The only difference from the original Kondo effect is that the transverse component is proportional to the original exchange as $`J^2`$ instead of $`J`$ as in the true Kondo effect. Thus, one has a cluster Kondo effect with the RKKY coupling with a characteristic Kondo temperature
$`T_K={\displaystyle \frac{E_F}{\alpha }}\left({\displaystyle \frac{\mathrm{\Delta }_A}{E_F}}\right)^{\frac{1}{1\alpha }}.`$ (150)
Observe that for a ferromagnet ($`Q=0`$) the magnetization is $`M_z=N/2`$ and in the large cluster size limit, $`k_FL1`$, we find
$`\alpha _F{\displaystyle \frac{1}{2(3\pi )^{2/3}}}\left({\displaystyle \frac{\rho _f}{\rho _e}}\right)^{2/3}(N(0)\overline{J}_z)^2N^{4/3}`$ (151)
where we assumed the cluster to be homogeneous with $`\rho _f=N/(8\pi L^3)`$ and $`\rho _e`$ is the electronic density given in (3). Observe that dissipation grows like a power law of the number of spins while the splitting decreases exponentially with the number of spins.
In the antiferromagnetic case $`Q0`$, the staggered magnetization is $`M_z=N/2`$, and from (149) we find that for large clusters ($`QL,2k_FL1`$) we have
$`\alpha _{AF}(2\pi N(0)\overline{J}_z)^2\left({\displaystyle \frac{\rho _f}{k_F^2Q}}\right)L^3\left[{\displaystyle \frac{\pi }{2}}\mathrm{\Theta }(2k_FQ)+{\displaystyle \frac{(2k_F)^2}{Q^2(2k_F)^2}}{\displaystyle \frac{1}{QL}}\right]+𝒪(1/L^3)`$ (152)
Thus, in the antiferromagnetic case $`\alpha `$ has a singularity at $`Q=2k_F`$ due to the Fermi surface effect. If $`Q2k_F`$ the leading order term is written
$`\alpha _{AF}{\displaystyle \frac{\pi ^2}{4}}(N(0)\overline{J}_z)^2\left({\displaystyle \frac{\rho _f}{k_F^2Q}}\right)N`$ (153)
and dissipation grows linearly with the number of atoms. Observe that the dissipation is substantially smaller than in the ferromagnetic case. For $`Q>2k_F`$ we have
$`\alpha _{AF}4\pi ^{4/3}(N(0)\overline{J}_z)^2\left({\displaystyle \frac{\rho _f^{1/3}}{Q}}\right)^4{\displaystyle \frac{N^{2/3}}{1(2k_F/Q)^2}}`$ (154)
which grows much slower with $`N`$ as the previous cases. From this simple calculations we can conclude that the dissipation is much weaker effect in antiferromagnetic clusters.
We can summarize these results as
$`\alpha =\left({\displaystyle \frac{N}{N_c}}\right)^\phi `$ (155)
where
$`\phi `$ $`=`$ $`4/3`$ (156)
$`N_c`$ $`=`$ $`\sqrt{6\pi }\left({\displaystyle \frac{\rho _e}{\rho _f}}\right)^{1/2}{\displaystyle \frac{1}{(N(0)J_z)^{3/2}}}`$ (157)
for a ferromagnetic cluster,
$`\phi `$ $`=`$ $`1`$ (158)
$`N_c`$ $`=`$ $`{\displaystyle \frac{4k_F^2Q}{\pi ^2\rho _f(N(0)J_z)^2}}`$ (159)
for an antiferromagnetic cluster with $`Q2k_F`$ and
$`\phi `$ $`=`$ $`2/3`$ (160)
$`N_c`$ $`=`$ $`{\displaystyle \frac{(1(2k_F/Q)^2)^{3/2}}{8\pi ^2(N(0)J_z)^3}}\left({\displaystyle \frac{Q}{\rho _f^{1/3}}}\right)^6`$ (161)
for an antiferromagnetic cluster with $`Q>2k_F`$. Notice that $`N_c`$ gives the critical number of spins in a given cluster above which the Kondo effect ceases to occur because for $`N>N_c`$ we find $`\alpha >1`$ which corresponds to the ferromagnetic Kondo coupling and therefore no Kondo effect. As we discussed before this case is related to the cessation of tunneling and the freezing of the cluster motion. Thus, the value of $`N_c`$ determines the largest size of a cluster which can still tunnel in the presence of a metallic environment when the anisotropy generated by the RKKY interaction is the source of quantum tunneling.
We can estimate the value of $`N_c`$ for typical values of the constants. We assume $`E_F10^4K`$, $`\overline{J}_z300K`$ ( which corresponds to an ordering temperature of the order $`\overline{\mathrm{\Gamma }}_z\overline{J}_z^2/E_F10K`$). Moreover, for U<sup>+3.5</sup>Cu$`{}_{4}{}^{}{}_{}{}^{+1}`$Pd<sup>0</sup> we have $`\rho _f1.2\times 10^2\AA ^3`$ and $`\rho _e8.7\times 10^2\AA ^3`$ (which corresponds to $`k_F1.4\AA ^1)`$. Moreover, from neutron scattering we have $`Q0.8\AA ^1`$ . In the ferromagnetic case we find $`N_c2\times 10^3`$ atoms and for the antiferromagnetic case $`N_c6\times 10^4`$ atoms! Thus, in the case of RKKY anisotropy a large number of atoms can quantum tunnel at low temperatures. We are going to see that the Kondo effect imposes much stronger restrictions on these numbers.
### D $`N`$ Magnetic Moments - XXZ and Heisenberg magnets
We describe in this section that when the system has XXZ or Heisenberg symmetry ($`\overline{\mathrm{\Gamma }}_x=\overline{\mathrm{\Gamma }}_y`$) the tunneling due to RKKY is suppressed. In this case the only source of tunneling, as we have discussed for the two impurity problem, is the Kondo effect itself. Again the XY component of the Kondo effect acts as a local magnetic field that flips the magnetic moment. This Kondo effect, in perfect analogy with the two-impurity Kondo problem, is due to the dissipative dynamics of states, say, $`|+=|,;,;\mathrm{}`$ and $`|=|,;,;\mathrm{}`$ in the case of ferromagnetic coupling. The existence of quasi-degenerate low lying states separated from higher energy states is guaranteed by the fact that the cluster is effectively within the ordered phase and therefore states of the spins can be related by a finite number of spin flips. Now we can just borrow the results from the previous sections for the response functions. In particular, the cluster Kondo temperature is given by
$`T_K(N){\displaystyle \frac{E_F}{\alpha (N)}}\left({\displaystyle \frac{\mathrm{\Delta }_0(N)}{E_F}}\right)^{1/(1\alpha (N))}`$ (162)
where $`\mathrm{\Delta }_0(N)`$ is the bare splitting between the two low lying states of the cluster and $`\alpha (N)`$ is the dissipative coupling to the bath. Since $`\mathrm{\Delta }_0(N)`$ is the splitting between two states of the cluster where all the spins are flipped it has to scale like $`(J_{}/\mathrm{\Gamma }_z^R)^N`$ which corresponds to the energy required to flip $`N`$ spins. Furthermore, for a cluster the dissipation comes from the fact that the “order parameter” flips and produces a wake of particle-hole excitations close to the Fermi surface. It is clear that in this case in (170) “order parameter” is an extensive quantity which implies that $`\alpha (N)N^2`$ in complete accordance with the discussion of the two-impurity Kondo problem. Thus, we conclude that for a cluster one has
$`\mathrm{\Delta }_0(N)\mathrm{\Gamma }_z^R\left({\displaystyle \frac{J_{}}{\mathrm{\Gamma }_z^R}}\right)^N=\mathrm{\Gamma }_z^Re^{N\mathrm{ln}(\mathrm{\Gamma }_z^R/J_{})}`$ (163)
which by direct comparison with (140) we find $`\gamma \mathrm{ln}(\mathrm{\Gamma }_z^R/J_{})`$ and $`\omega _0\mathrm{\Gamma }_z^R`$. Notice from (101) that the RKKY interaction is strongly renormalized and in general $`\mathrm{\Gamma }_z^RJ_{}`$. Furthermore, for the antiferromagnetic case the coupling of the cluster to the bath can be written as
$`\alpha (N)N^2\left({\displaystyle \frac{J_z}{E_F}}\right)^2=\left({\displaystyle \frac{N}{N_c}}\right)^2`$ (164)
while for a ferromagnetic cluster we expect (see (127))
$`\alpha (N)N^2\left(1{\displaystyle \frac{J_z}{E_F}}\right)^2=\left({\displaystyle \frac{N}{N_c}}\right)^2`$ (165)
where $`N_cE_F/J_z`$ ($`(1\frac{J_z}{E_F})^1`$) for an antiferromagnetic (ferromagnetic) cluster. Observe that in a typical antiferromagnetic system $`N_c1001000`$ spins while for a ferromagnetic cluster $`N_c1`$. We can immediately conclude that in the case of ferromagnetic clusters, like in the case where tunneling is generated by anisotropy of the RKKY interaction, dissipation will be so important that the cluster will freeze at low temperatures. We expect that ordinary superparamagnetism will occur - a situation much closer to a classical spin glass. Moreover, observe that the Kondo temperature drops exponentially with the number of spins in the cluster and therefore for practical purposes the Kondo temperature drops to vanishing small values before the cluster size reaches $`N_c`$. Thus, the region where dissipation occurs is rather narrow in the case of antiferromagnetic clusters. The Kondo effect, therefore, imposes much stronger requirements in the cluster size than the RKKY interaction. This is due to the fact that the Kondo effect requires the collective flipping of the cluster magnetization from electron scattering while the RKKY tunneling is a product of lattice anisotropy. Which effect is the dominant one in real systems depends on the lattice structure and the spin-orbit coupling. Furthermore, without dissipation the tunneling splitting vanishes only when the number of spins in the cluster diverges. In the presence of dissipation it vanishes for a finite number of spins given by $`N_c`$.
## V The Dissipative Quantum Droplet Model
We have seen in the previous section that the problem of a set of magnetic impurities interacting through a conduction band has two main features: tunneling and dissipation. In a simple mean field like picture there are three different energy scales in this problem: the ordering temperature which scales like $`T_cN(0)J_z^2`$, the tunneling energy which is given by $`\mathrm{\Omega }`$ and the damping energy which is given by $`\mathrm{\Gamma }`$ both given in (70).
In order to understand how these energy scales affect the Kondo lattice consider a ligand system in the compositionally ordered phase. In this case the exchange between atoms is small enough so that the Kondo effect does not take place. As the system is doped with a metallic atom with size smaller than the original one the lattice locally contracts. The local matrix elements have exponentially large values and thus also the exchange $`J_z`$. Then a particular region of the system can have an exchange parameter much larger than the average exchange in the lattice. This is the situation described by Hamiltonian (22), for instance. If the exchange is locally very large then in face of the discussion of the single impurity Kondo problem it will be given by a Hamiltonian like (66) where the tunnelling splitting is replaced by $`\mathrm{\Omega }`$. Thus in the limit of large magnetic anisotropy the effective magnetic Hamiltonian which is generated from (15) is given by
$`H_{eff}`$ $`=`$ $`{\displaystyle \underset{n,m}{}}\mathrm{\Gamma }_z(𝐫_n𝐫_m)S_z(𝐫_n)S_z(𝐫_m)+{\displaystyle \underset{n}{}}\mathrm{\Omega }(𝐫_n)S_x(𝐫_n)`$ (166)
$`+`$ $`{\displaystyle \underset{n,\alpha }{}}S_z(𝐫_n)\left(\lambda _\alpha (𝐫_n)b_\alpha +\lambda _\alpha ^{}(𝐫_n)b_\alpha ^{}\right)+{\displaystyle \underset{\alpha }{}}\omega _\alpha b_\alpha ^{}b_\alpha `$ (167)
where $`\alpha `$ labels the relevant quantum numbers for the particle-hole excitations and $`\lambda (𝐫_n)`$ is the local coupling constant of the spin to the electronic bath. When $`\lambda =0`$ the bath decouples from the spins and the problem is mapped into the transverse field Ising model. We should stress at this point that (167) can be only formally demonstrated for weak magnetic moment concentration. For larger concentrations it becomes an ansatz which has the correct features of the problems we are discussing.
It is clear from the above that (167) will drive the system through a quantum phase transition as $`\mathrm{\Omega }`$ increases. The simplest way to understand this effect is to rewrite the problem in path integral form and use the Suzuki-Troter trick to map the $`d`$-dimensional quantum problem (167) into a $`d+1`$ classical problem by breaking the imaginary time direction into $`N_\tau =\beta /ϵ`$ pieces. The effective Hamiltonian associated with (167), after tracing out the bosons degree of freedom, reads:
$`H_e={\displaystyle \underset{i=1}{\overset{N_\tau }{}}}{\displaystyle \underset{n,m}{}}ϵ\mathrm{\Gamma }_{n,m}S(i,n)S(i,n)+{\displaystyle \underset{i=1}{\overset{N_\tau }{}}}{\displaystyle \underset{n}{}}J_nS(i,n)S(i+1,n)+{\displaystyle \underset{ij=1}{\overset{N_\tau }{}}}{\displaystyle \underset{n}{}}{\displaystyle \frac{\alpha _n}{8}}{\displaystyle \frac{S(i,n)S(j,n)}{(ij)^2}}`$ (168)
where $`J_n=\mathrm{ln}[\mathrm{tanh}(ϵ\mathrm{\Omega }_n)]`$ and $`\alpha _n`$ is the dissipation constant of each spin (observe that in the limit of $`ϵ0`$ we have $`J_n\mathrm{}`$, thus, the mapping into the classical problem is valid when we take $`\beta \mathrm{}`$). The original problem can now be thought as an anisotropic $`d+1`$ classical Ising model with long range interactions in the imaginary time direction and short range interactions in the space direction. In the classical problem the sign of $`\mathrm{\Gamma }_n`$ is irrelevant and for small $`J_n`$ the system will order magnetically (ferro, antiferro or spin-glass, depending on $`\mathrm{\Gamma }_n`$) in $`d`$ dimensions. Because the interactions are long-range in the imaginary direction the system orders in this direction as well. As in the classical case one assumes that the lowest excitation energy above the classical configuration is a droplet involving $`N`$ spins which can be reversed at some energy cost $`E(N)`$ which scales with the size of the droplet as $`N^\theta `$. Moreover, we will assume that the droplets are very diluted and do not interact with each other in the spatial domain. It is clear that the problem is equivalent to independent droplets which are correlated in the imaginary time direction. The effective action for this problem is obtained directly from (168) by coarse-graining the spatial coordinate:
$`H_D={\displaystyle \underset{N}{}}\left\{{\displaystyle \underset{j=1}{\overset{N_\tau }{}}}\left[ϵE(N)S(ϵj)+K(N)S(ϵj)S(ϵ(j+1))\right]+{\displaystyle \frac{\alpha (N)}{8}}{\displaystyle \underset{jk}{}}{\displaystyle \frac{S(ϵj)S(ϵk)}{(kj)^2}}\right\}`$ (169)
where $`S=1`$ corresponds to the ground state and $`S=+1`$ corresponds to the droplet excitation present at some imaginary time. The problem of a droplet at low energies is totally equivalent to a two level system problem. This approach is an extension of the quantum droplet model proposed by Thill and Huse in the context of insulating magnets. We should point out that the quantum droplet model gives essentially the same results obtained numerically and analytically for the random transverse field Ising model.
Notice that the energy barriers involved in the imaginary time are given by $`\mathrm{\Delta }_0(N)`$ which is the surface free energy in $`d+1`$ dimensions. It is obvious from the above discussion that in real time the dynamics of the problem is described by the Hamiltonian
$`H={\displaystyle \underset{N}{}}\left(E(N)\sigma _z(N)+\mathrm{\Delta }_0(N)\sigma _x(N)\pi \sqrt{\alpha (N)}\sigma _z(N){\displaystyle \underset{k>0}{}}\sqrt{{\displaystyle \frac{k}{\pi L}}}(b_k+b_k^{})+{\displaystyle \underset{k>0}{}}kb_k^{}b_k\right)`$ (170)
which is the analogue of the two level system Hamiltonian (66). Observe that $`\sigma _z`$ is therefore related to magnetic order parameter of the system and $`\sigma _x`$ is associated with the coherent flip of $`N`$ spins within the droplet. In a ferro or antiferromagnetic state $`E(N)`$ is proportional to the mean magnetic field of the spins which keep the droplet in the ordered phase. In a spin glass this energy is distributed accordingly to some distribution $`P(E(N))`$, say,
$`P(E)={\displaystyle \frac{2}{\sqrt{\pi }\overline{\mathrm{\Gamma }}}}e^{E^2/\overline{\mathrm{\Gamma }}^2}`$ (171)
where $`\overline{\mathrm{\Gamma }}=\sqrt{\overline{\mathrm{\Gamma }_{n,m}^2}}`$ is the average interaction within the droplet. Moreover, as it is well-known in the context of the two-level system problem the bias $`E(N)`$ is a relevant perturbation and will lead to ordering of the moments. In the rest of the paper we will interested only in the paramagnetic phase where $`E(N)=0`$ and the droplets are just the clusters discussed in this paper.
The statistical problem is now very similar to the one discussed in Section IV. We have a distribution of energy scales (or cluster Kondo temperatures) which have their origin in a distribution of cluster sizes. Thus, contrary to the well-known distribution of Kondo temperatures approach to NFL behavior the effect is not of single impurity nature and the distribution is fixed by percolation theory. Using (67), (163) and (164) we see that the renormalized tunneling splitting is given by
$`\mathrm{\Delta }_R(N)=We^{\left(\frac{\gamma N+\mathrm{ln}(W/\omega _0)}{1(N/N_c)^\phi }\right)}`$ (172)
where $`WE_F`$. Usually it is not possible to invert the $`N`$ as a function of $`\mathrm{\Delta }_R`$ for a generic value of $`\phi `$. Thus, for practical purposes the averages over the distributions of clusters have to be done with (31) and (33):
$`P(N)={\displaystyle \frac{N^{1\theta }e^{N/N_\xi }}{\mathrm{\Gamma }(2\theta )N_\xi ^{2\theta }}}.`$ (173)
We will consider first the example of the cluster Kondo effect for $`\phi =2`$ which serves as a good illustration of the effect of dissipation in the cluster problem. When $`\phi =2`$, (172) can be inverted in order to give the size of the cluster in terms of the splitting or fluctuation time $`\tau =1/\mathrm{\Delta }_R`$ as
$`{\displaystyle \frac{N}{N_c}}={\displaystyle \frac{1}{2\mathrm{ln}(W/\mathrm{\Delta }_R)}}\left[\sqrt{(N_c\gamma )^2+4\mathrm{ln}(W/\mathrm{\Delta }_R)\mathrm{ln}(\omega _0/\mathrm{\Delta }_R)}N_c\gamma \right].`$ (174)
Notice that $`N`$ vanishes when $`\mathrm{\Delta }_R>\omega _0`$ which corresponds to the smallest cluster (that is, $`N=1`$ single Kondo impurity). Therefore the probability of finding a cluster with splitting $`\mathrm{\Delta }_R`$ is given by
$`P(\mathrm{\Delta }_R)`$ $`=`$ $`P(N(\mathrm{\Delta }_R))\left|{\displaystyle \frac{dN}{d\mathrm{\Delta }_R}}\right|`$ (175)
$``$ $`{\displaystyle \frac{\kappa _pN_c\left\{\mathrm{ln}(W/\mathrm{\Delta }_R)\mathrm{ln}(W/\omega _0)+\frac{\gamma N_c}{2}\left[\sqrt{(N_c\gamma )^2+4\mathrm{ln}(W/\mathrm{\Delta }_R)\mathrm{ln}(\omega _0/\mathrm{\Delta }_R)}N_c\gamma \right]\right\}}{(1e^{N_c\kappa _p})\mathrm{\Delta }_R\mathrm{ln}^2(W/\mathrm{\Delta }_R)\sqrt{(N_c\gamma )^2+4\mathrm{ln}(W/\mathrm{\Delta }_R)\mathrm{ln}(\omega _0/\mathrm{\Delta }_R)}}}`$ (176)
$`\times `$ $`\mathrm{exp}\left\{\kappa _p{\displaystyle \frac{N_c}{2\mathrm{ln}(W/\mathrm{\Delta }_R)}}\left[\sqrt{(N_c\gamma )^2+4\mathrm{ln}(W/\mathrm{\Delta }_R)\mathrm{ln}(\omega _0/\mathrm{\Delta }_R)}N_c\gamma \right]\right\}.`$ (177)
Observe that when $`N_c\mathrm{}`$ we have
$`P(\mathrm{\Delta })\mathrm{\Delta }^{\kappa _p/\gamma 1}`$ (178)
which gives a power law distribution for the energy levels. If $`\kappa _p/\gamma <1`$ the probability is divergent when $`\mathrm{\Delta }0`$. As we are going to see in the next section it is this power law that gives rise to quantum Griffiths singularities. However, for finite $`N_c`$ the power law is modified at very small splittings. Indeed, when $`\mathrm{\Delta }_R0`$ we find:
$`P(\mathrm{\Delta }_R0){\displaystyle \frac{N_c\kappa _p(\mathrm{ln}(W/\omega _0)+\gamma N_c)}{2(e^{N_c\kappa _p}1)}}{\displaystyle \frac{1}{\mathrm{\Delta }_R\mathrm{ln}^2(\omega _0/\mathrm{\Delta }_R)}}`$ (179)
which, unlike (178), diverges logarithmically. Thus, there is a crossover in the problem from the power law given by (178) for $`\mathrm{\Delta }_R>\mathrm{\Delta }^{}`$ to (179) for $`\mathrm{\Delta }_R<\mathrm{\Delta }^{}`$ where:
$`\mathrm{\Delta }^{}=\sqrt{W\omega _0}\mathrm{exp}\left\{{\displaystyle \frac{1}{2}}\sqrt{\mathrm{ln}^2(W/\omega _0)+(\gamma N_c)^2}\right\}.`$ (180)
Thus, when $`N_c\mathrm{}`$ we have $`\mathrm{\Delta }^{}0`$ and the power law behavior dominates the entire range of energy scales.
### A Magnetic properties: $`T<T^{}`$
Since we now have the distribution of energy scales it is possible to calculate the physical properties of the clusters. For that we need the response functions for individual clusters which are given in Section IV. For the Kondo effect there are no analytic expressions for the physical properties which have to be usually calculated numerically . Here we will study only the asymptotic behavior. The temperature crossovers, however, can only be obtained numerically and in some special points which are not of practical interest.
The magnetic susceptibility can be obtained directly from (73) and (77) and for $`T\mathrm{\Delta }_R(N)`$ is given by
$`\chi (T,\mathrm{\Delta }_R(N)){\displaystyle \frac{1}{2\pi \mathrm{\Delta }_R(N)}}`$ (181)
while for $`T\mathrm{\Delta }_R(N)`$ we have
$`\chi (T,\mathrm{\Delta }_R(N)){\displaystyle \frac{1}{4T}}.`$ (182)
The average susceptibility is given by for $`T<\omega _0`$
$`\overline{\chi }(T)`$ $`=`$ $`{\displaystyle _0^{N_c}}𝑑NP(N)\chi (T,\mathrm{\Delta }_R(N))`$ (183)
$``$ $`{\displaystyle _0^{N^{}(T)}}𝑑N{\displaystyle \frac{P(N)}{2\pi \mathrm{\Delta }_R(N)}}+{\displaystyle _{N^{}(T)}^{N_c}}𝑑N{\displaystyle \frac{P(N)}{4T}}`$ (184)
where $`\mathrm{\Delta }_R(N^{}(T))=T`$ and we have approximated the integral by its two asymptotic pieces. Since $`\mathrm{\Delta }_R\omega _0`$ in the high temperature limit ($`T>\omega _0`$) we have $`\overline{\chi (T)}1/(4T)`$ and therefore one has the usual Curie behavior.
In what follows we will always assume that $`T\omega _0`$ in which case $`N^{}(T)N_c`$ since $`\mathrm{\Delta }_R(N_c)=0`$. Using (172) we easily find
$`N^{}(T)N_c\left(1{\displaystyle \frac{\gamma N_c+\mathrm{ln}(W/\omega _0)}{\gamma N_c+\phi \mathrm{ln}(\beta W)}}\right).`$ (185)
In this limit the second integral in (184) can be computed immediately
$`{\displaystyle _{N^{}(T)}^{N_c}}𝑑N{\displaystyle \frac{P(N)}{4T}}{\displaystyle \frac{N_c^{2\theta }(\gamma N_c+\mathrm{ln}(W/\omega _0))e^{N_c/(\xi /a)^D}}{4\mathrm{\Gamma }(2\theta )(\xi /a)^{2\theta }T(\gamma N_c+\phi \mathrm{ln}(\beta W))}}`$ (186)
and therefore for $`TT^{}=We^{\gamma N_c/\phi }`$ we see that this term diverges like $`1/(T\mathrm{ln}(W/T)`$. The first integral in (184) is more complicated but can be written in terms of exponential integrals and we find that in the limit of $`TT^{}`$ it diverges like $`1/(T\mathrm{ln}^2(W/T)`$ and therefore is less singular than the second term. Thus, we conclude that when $`TT^{}`$ the susceptibility diverges at zero temperature as
$`\overline{\chi }(T){\displaystyle \frac{1}{T\mathrm{ln}(W/T)}}`$ (187)
which has stronger divergence a the power law singularity and it is independent on the value of $`\phi `$.
In the opposite limit, that is $`TT^{}`$ we see that $`N^{}(T)`$ is very small and given by
$`N^{}(T){\displaystyle \frac{1}{\gamma }}\mathrm{ln}(\beta \omega _0)`$ (188)
which is what we expect from the usual Griffiths-McCoy case. We conclude therefore that power law behavior has to be observed at temperatures larger than $`T^{}`$. In systems where RKKY is the source of tunneling as we saw in Subsection IV C the value of $`N_c`$ is very large and therefore $`T^{}`$ is effectively zero. Thus power law behavior will be predominant over all realistic low temperatures. In systems where the Kondo effect is responsible for tunneling, as described in Subsection IV D we can have $`N_c10`$ if $`J_z1,000K`$ and $`\gamma 2`$ (for ordering temperatures of the order of $`\mathrm{\Gamma }_z^R10K`$ and $`J_{}1K`$) and therefore $`T^{}0.5K`$ which is a reasonable energy scale. Notice, however, that $`T^{}`$ is an exponential function of $`N_c`$ and therefore a change to $`J_z500K`$ (which is probably a better estimative for the exchange in rare earths) leads to $`T^{}10^9K`$!
### B Specific Heat: $`T<T^{}`$
Once again analytic expressions for the specific heat of a Kondo system are not known. We will use the asymptotic forms given in (74) and (77). For $`T\mathrm{\Delta }_R(N)`$ we have
$`\gamma (T,N)={\displaystyle \frac{C_V(T,N)}{T}}{\displaystyle \frac{\pi }{3}}{\displaystyle \frac{\alpha (N)}{\mathrm{\Delta }_R(N)}}`$ (189)
while for $`T\mathrm{\Delta }_R(N)`$
$`\gamma (T,N)\left({\displaystyle \frac{\mathrm{\Delta }_0}{E_c}}\right)^2\left({\displaystyle \frac{E_c}{T}}\right)^{2(1\alpha (N))}`$ (190)
where we used (67). Exactly as in the case of the susceptibility we break the integral of the specific heat into two pieces:
$`\overline{\gamma }(T){\displaystyle \frac{\pi }{3}}{\displaystyle _0^{N^{}(T)}}𝑑N{\displaystyle \frac{P(N)\alpha (N)}{\mathrm{\Delta }_R(N)}}+\left({\displaystyle \frac{\mathrm{\Delta }_0}{E_c}}\right)^2{\displaystyle _{N^{}(T)}^{N_c}}𝑑NP(N)\left({\displaystyle \frac{E_c}{T}}\right)^{2(1\alpha (N))}.`$ (191)
For $`TT^{}`$ the integrals can be evaluated as in the case of the susceptibility. The main difference here is that the second integral in (191) which gave the most divergent contribution in the case of the susceptibility diverges only like $`1/\mathrm{ln}(\beta W)`$ because $`\alpha (N_c)=1`$. Thus in the case of the specific heat the dominating term is the first one which diverges like
$`\overline{\gamma }(T){\displaystyle \frac{1}{T\mathrm{ln}^2(W/T)}}`$ (192)
which is a weaker divergence than the susceptibility. In the limit of $`TT^{}`$ we recover the power law behavior characteristic of Griffiths-McCoy singularities. Other physical quantities can be directly calculated from the probability distribution (173) and the dependence of $`\mathrm{\Delta }_R(N)`$ on $`N`$ given in (172).
## VI $`T>T^{}`$: quantum Griffiths singularities
We have shown in the previous section that at temperatures smaller than $`T^{}`$ the cluster dynamics is essentially dissipative and the singularities in the thermodynamic functions have logarithmic character. A reasonable estimate of $`T^{}`$ gives a very low temperature which in many systems is experimentally not reachable. Thus, for most systems the physics in a wide temperature range from $`T^{}`$ to $`T\overline{\mathrm{\Gamma }}`$ where $`\overline{\mathrm{\Gamma }}`$ is the average RKKY interaction in the system the physics is non-dissipative. For temperatures above $`\overline{\mathrm{\Gamma }}`$ the cluster does not exist as a well-defined object since temperature fluctuations can excite isolated spins within the cluster. In the temperature range $`T^{}<T<\overline{\mathrm{\Gamma }}`$ we can effectively treat the system as non-dissipative, that is, set $`\alpha =0`$ (or equivalently, $`N_c\mathrm{}`$). In this crossover temperature scale the physics is highly non-universal and depends strongly from the distance from percolation threshold. Indeed, consider the probability of finding a cluster with an energy splitting $`\mathrm{\Delta }`$ which is given in (178). Since there is a broad distribution of energy scales in disordered systems it is often convenient to use a logarithmic scale so that
$`P(\mathrm{ln}(\mathrm{\Delta }))\mathrm{\Delta }^{\lambda _p}`$ (193)
where
$`\lambda _p={\displaystyle \frac{\kappa _p}{\gamma }}`$ (194)
is an important exponent of the theory. Equation (193) is quite revealing if one considers the probability of finding a spin in a cluster of size $`N`$ or excitation energy $`\mathrm{\Delta }`$. Since the probability is proportional to the size of the cluster one sees that this probability is
$`P_L(\mathrm{ln}(\mathrm{\Delta }))`$ $``$ $`N\mathrm{\Delta }^{\kappa _p/\gamma }L^D\mathrm{\Delta }^{\kappa _p/\gamma }`$ (195)
$``$ $`(L\mathrm{\Delta }^{1/Z_p})^D`$ (196)
where
$`Z_p={\displaystyle \frac{D}{\lambda _p}}`$ (197)
is the so-called dynamical exponent. Observe that $`Z_p`$ depends directly on $`\lambda _p`$ which is a non-universal quantity which depends on many microscopic details as one can easily see from the definition (194). Thus $`Z_p`$ varies over the phase diagram. Indeed, using (39) and (197)
$`Z_{pp_c}`$ $``$ $`{\displaystyle \frac{\gamma D}{\mathrm{ln}(1/c)}}`$ (198)
$`Z_{pp_c}`$ $``$ $`\gamma D(\xi (p)/a)^D{\displaystyle \frac{1}{|pp_c|^{\nu D}}}\mathrm{}`$ (199)
which diverges at percolation threshold. The physical meaning of the dynamical exponent is clear from (196): it defines the relationship between length and energy scales in a correlated volume $`L`$ of the system. The thermodynamic and response functions of the system are directly related to $`Z_p`$. Finally, we have to point out that the dynamical exponent in the Griffiths phase has a different meaning than the one usually used for clean systems (thus we use the capital letter $`Z_p`$ for the dynamical exponent in the Griffiths phase). We should stress once again that the behavior of the system for $`T<T^{}`$ also corresponds to $`Z\mathrm{}`$ but unlike the non-dissipative system $`Z`$ is divergent away from the critical point since it is independent of $`\kappa _p`$ which vanishes at $`p=p_c`$. Observe that even at this point (179) is finite.
All the thermodynamic functions can be now calculated from the knowledge of $`\lambda _p`$ or $`Z_p`$. Although we were able to find microscopic expressions for $`\gamma `$ and $`\omega _0`$ the relationship between $`p`$ and $`Z_p`$ depends on details of the percolation problem. Thus, another way to face the problem here is to assume $`\lambda _p`$ as a phenomenological parameter which can be varied as doping is varied and which vanishes when $`pp_c`$. Thus, from now on we drop the subscript $`p`$ on $`\lambda _p`$. In the rest of the section we give an explicit calculation for the response functions in the crossover temperature range. Using (178) and (140) we find that the normalized distribution function (178) is given by
$`P(\mathrm{\Delta })={\displaystyle \frac{[\mathrm{ln}(\omega _0/\mathrm{\Delta })]^{1\theta }}{\mathrm{\Gamma }[2\theta ](\gamma N_\xi )^{2\theta }\omega _0}}\left({\displaystyle \frac{\mathrm{\Delta }}{\omega _0}}\right)^{\lambda 1}\mathrm{\Theta }(\omega _0\mathrm{\Delta }).`$ (200)
Notice that the Heavyside step function appears because the largest value for the tunneling splitting is $`\omega _0`$.
The calculation of physical quantities is now rather simple because we have just a two level system Hamiltonian to deal with. For generality let us consider the problem of a magnetic field $`H`$ applied along the easy axis of the cluster so that the effective cluster Hamiltonian can be written as
$`H_C=E_H\sigma _z+\mathrm{\Delta }\sigma _x`$ (201)
where $`E_H`$ is the magnetic energy of the cluster. This magnetic energy can be written as $`E_H=m_cH`$ where $`m_c`$ is the mean magnetic moment of the clusters and it depends on the magnetic nature of the cluster and the number of spins in the cluster. For a ferromagnetic cluster we would have $`m_c=g\mu _BS_T`$. If the cluster is fully aligned then $`S_T=N/2`$. For an antiferromagnetic cluster in a percolating lattice we can have the following possibilities: (1) the cluster has random fluctuations of the spins (uncompensated) in its surface in which case the net magnetization is proportional to $`(N)^{1/D}`$ (this result can be obtained from the fact that the average size $`R`$ of cluster scales like $`RN^{1/D}`$ and therefore its area scales like $`AN^{2/D}`$ leading to fluctuations or the order $`\sqrt{A}N^{1/D}`$ as $`N1`$); (2) the cluster has uncompensated spins in its volume so that the net magnetization proportional to $`(N)^{1/2}`$. In general we write $`m_c=\mu _BqN^\varphi `$ where $`q`$ gives the magnitude of the average moment in the cluster (for a spin glass it is the Edwards-Anderson parameter) and $`\varphi =1(1/2)`$ for ferromagnets (antiferromagnets and spin glasses). Thus, using (140) we can write
$`E_H(\mathrm{\Delta })=q\mu _B\left[{\displaystyle \frac{1}{\gamma }}\mathrm{ln}\left({\displaystyle \frac{\omega _0}{\mathrm{\Delta }}}\right)\right]^\varphi H.`$ (202)
Notice, however, that the dependence of $`E_H`$ on $`\mathrm{\Delta }`$ is rather weak and since $`\mathrm{\Delta }`$ is cut-off from above by $`\omega _0`$, $`H`$ or $`T`$, we can safely replace the expression above by
$`E_H(H,T)q\mu _B\left[{\displaystyle \frac{1}{\gamma }}\mathrm{ln}\left({\displaystyle \frac{\omega _0}{max(H,T)}}\right)\right]^\varphi H.`$ (203)
In doing so we disregard higher logarithmic corrections in the expressions given above.
The cluster magnetization is obtained from (201):
$`M(\mathrm{\Delta },H,T)={\displaystyle \frac{\mu _B^2E_H}{\sqrt{E_H^2+\mathrm{\Delta }^2}}}\mathrm{tanh}\left(\beta \sqrt{E_H^2+\mathrm{\Delta }^2}\right),`$ (204)
the static magnetic susceptibility can obtained directly from (204):
$`\chi _0(\mathrm{\Delta },H,T)=\mu _B^3\gamma \left[{\displaystyle \frac{\beta E_H}{E_H^2+\mathrm{\Delta }^2}}sech^2\left(\beta \sqrt{E_H^2+\mathrm{\Delta }^2}\right)+{\displaystyle \frac{\mathrm{\Delta }^2}{(E_H^2+\mathrm{\Delta }^2)^{3/2}}}\mathrm{tanh}\left(\beta \sqrt{E_H^2+\mathrm{\Delta }^2}\right)\right],`$ (205)
the imaginary part of the frequency dependent susceptibility for $`\omega >0`$ is (see Appendix (C)):
$`\mathrm{}\left[\chi (\omega ,\mathrm{\Delta },H,T)\right]=2\pi {\displaystyle \frac{\mathrm{\Delta }^2}{(E_H^2+\mathrm{\Delta }^2)}}\mathrm{tanh}\left(\beta \sqrt{E_H^2+\mathrm{\Delta }^2}\right)\delta (\omega 2\sqrt{E_H^2+\mathrm{\Delta }^2}),`$ (206)
and the specific heat is given by:
$`C_V(\mathrm{\Delta },H,T)=\beta ^2(E_H^2+\mathrm{\Delta }^2)sech^2\left(\beta \sqrt{E_H^2+\mathrm{\Delta }^2}\right).`$ (207)
Any physical constant can now be calculated directly from the expressions above and the distribution of tunneling splittings given by (200).
#### 1 Magnetic Response
From (204) and (200) we see that the average magnetization is given by
$`\overline{M}(H,T)`$ $``$ $`H{\displaystyle _0^{\omega _0}}𝑑\mathrm{\Delta }{\displaystyle \frac{\mathrm{\Delta }^{\lambda 1}}{\sqrt{E_H^2+\mathrm{\Delta }^2}}}\mathrm{tanh}\left(\beta \sqrt{E_H^2+\mathrm{\Delta }^2}\right)[\mathrm{ln}{\displaystyle \frac{\omega _0}{\mathrm{\Delta }}}]^{1\theta +\varphi }`$ (208)
$``$ $`H\beta ^{1\lambda }{\displaystyle _{\beta E_H}^{\beta \sqrt{E_H^2+\omega _0^2}}}𝑑x{\displaystyle \frac{\mathrm{tanh}(x)}{\left(x^2(\beta E_H)^2\right)^{1\lambda /2}}}[\mathrm{ln}{\displaystyle \frac{\beta \omega _0}{\sqrt{x^2(\beta E_H)^2}}}]^{1\theta +\varphi }.`$ (209)
Observe that the magnetization has a scaling form
$`\overline{M}(H,T)={\displaystyle \frac{H}{T^{1\lambda }}}f_\lambda ({\displaystyle \frac{H}{T}},{\displaystyle \frac{T}{\omega _0}})`$ (210)
where $`f_\lambda (x,y)`$ is a simple scaling function.
Let us consider first the high temperature case ($`T\omega _0,E_H`$) in which case the integral simplifies to
$`\overline{M}{\displaystyle \frac{E_H}{T}}`$ (211)
which is just the usual Curie behavior. Observe that the high temperature results have to be understood with a little grain of salt since the cluster decomposes when the temperature becomes of the order of the characteristic coupling between spins (which is of order of ordering temperature of the pure system, $`T_c`$). At high temperatures one has indeed paramagnetic behavior like the one described by (211) but the Curie constant can be quite different from the one we would get from the cluster calculation since it comes from individual atoms. Thus, high temperature behavior in the cluster picture only makes sense if $`T_cT\omega _0`$ which puts a strong constraint on the temperature range. Moreover, at high temperatures the clusters can be thermally activated over the barrier and the two level system description is not complete. We therefore focus on the case where $`\omega _0T,E_H`$. Notice that in this case the scaling function in (210) depends only on the ratio of $`H/T`$.
Let us consider the case where $`\omega _0T,E_H`$. If $`TE_H`$ (low field limit) (209) can be safely replaced by
$`\overline{M}{\displaystyle \frac{(\mathrm{ln}\beta \omega _0)^{1\theta +\varphi }}{T^{1\lambda }}}H.`$ (212)
Observe that the magnetization is linear in the field and diverges at low temperatures as $`T^{\lambda 1}`$ for $`\lambda <1`$. Thus, the scaling function is $`f_\lambda (x,0)1`$ when $`x0`$.
On the other hand, if $`\omega _0E_HT`$ (high field limit) the integral has a different behavior, namely,
$`\overline{M}`$ $``$ $`H(\beta )^{1\lambda }{\displaystyle _{\beta E_H}^{\mathrm{}}}𝑑x{\displaystyle \frac{\left[\mathrm{ln}\frac{\beta \omega _0}{\sqrt{x^2(\beta E_H)^2}}\right]^{1\theta +\varphi }}{\left(x^2(\beta E_H)^2\right)^{1\lambda /2}}}`$ (213)
$``$ $`H^\lambda \left(\mathrm{ln}{\displaystyle \frac{\omega _0}{H}}\right)^{1\theta +\varphi }`$ (214)
which shows that the magnetization scales like $`H^\lambda `$ and therefore the susceptibility behaves like $`H^{\lambda 1}`$. Thus, the scaling function behaves like $`f_\lambda (x,0)x^{\lambda 1}`$ when $`x\mathrm{}`$. The crossover from (214) to (212) occurs when $`E_HT`$.
The imaginary part of the average frequency dependent susceptibility can be easily calculated from (200) and (206) since the only contribution comes from the Dirac delta function:
$`Im[\chi (\omega ,E_H,T)]\left(\omega ^2(2E_H)^2\right)^{\lambda /2}{\displaystyle \frac{\mathrm{tanh}\left(\frac{\beta \omega }{2}\right)}{\omega }}\left(\mathrm{ln}{\displaystyle \frac{2\omega _0}{\sqrt{\omega ^2(2E_H)^2}}}\right)^{1\theta +\varphi }\mathrm{\Theta }(\omega 2E_H).`$ (215)
This function can be measured in momentum integrated neutron scattering experiments since it is the local response function. Observe that for $`E_H=0`$ it reduces to
$`Im[\chi (\omega ,0,T)]{\displaystyle \frac{\mathrm{tanh}\left(\frac{\beta \omega }{2}\right)}{\omega ^{1\lambda }}}\left(\mathrm{ln}{\displaystyle \frac{2\omega _0}{\omega }}\right)^{1\theta +\varphi }`$ (216)
which is rather simple function. The fact that this function vanishes for $`\omega <2E_H`$ has to do with the blocking of the tunneling by a magnetic field which biases one specific magnetic configuration. Observe that the application of a magnetic field suppresses the quantum fluctuations and therefore destroys the NFL behavior caused by the clusters.
Another important quantity for the characterization of a Griffiths singularity is the nonlinear magnetic susceptibility which can be calculated from (210) as the third derivative of the magnetization. Due to the scaling behavior it is obvious that the leading behavior is
$`\chi _3(T){\displaystyle \frac{1}{T^{3\lambda }}}`$ (217)
which therefore diverges more strongly than the linear susceptibility.
In certain experiments, like the Knight shift $`K(T)`$ measurement in NMR, it is possible to observe the width of the distribution of energy scales by studying the distribution of local susceptibilities in the system. Indeed, we can write:
$`{\displaystyle \frac{\delta \chi (T)}{\chi }}=D{\displaystyle \frac{\delta K(T)}{K}}`$ (218)
where $`D`$ is a constant which depends on the range of the interactions. For long range interactions $`D=1`$ while for short range interactions one has $`D>1`$. $`\delta \chi (T)=\sqrt{\overline{\chi ^2}(T)\overline{\chi }(T)^2}`$ can be calculated directly from (205) and (200)
$`{\displaystyle \frac{\delta \chi (T)}{\overline{\chi }}}{\displaystyle \frac{1}{T^{\lambda /2}}}`$ (219)
for $`T0`$. This result shows that as the temperature is lowered the width of the distribution of susceptibilities grows and eventually diverges at $`T=0`$. In fact any moment of the susceptibility can be calculated from (205) and (200).
#### 2 Neutron Scattering
Consider now the problem of scattering of neutrons by the magnetic clusters discussed here. The differential cross-section for neutron scattering is given by the usual expression
$`{\displaystyle \frac{d^2\sigma }{d\mathrm{\Omega }d\omega }}=N\left({\displaystyle \frac{\gamma e^2}{m_Nc^2}}\right)^2{\displaystyle \frac{k^{}}{k}}S(𝐪,\omega )`$ (220)
where $`𝐪=𝐤𝐤^{}`$ is the momentum transfer and $`\omega =((k^{})^2k^2)/(2m_N)`$ the energy transfer in the scattering. $`S(𝐪,\omega )`$ is the dynamical form factor which can be written as
$`S(𝐪,\omega )={\displaystyle \frac{F^2(𝐪)}{N\pi }}{\displaystyle \underset{i,j}{}}{\displaystyle _{\mathrm{}}^+\mathrm{}}𝑑te^{i(\omega t𝐪(𝐫_i𝐫_j)}S_i(0)S_j(t),`$ (221)
where $`F(𝐪)`$ is the nuclear form factor. It is usual to rewrite the spin-spin correlation function as $`S_i(0)S_j(t)=(S_i(0)S_j(t)S_iS_j)+S_iS_j`$ so that we can divide the dynamical form factor into a static and a dynamic part: $`S(𝐪,\omega )=S_S(𝐪)\delta (\omega )+S_D(𝐪,\omega )`$ where
$`S_S(𝐪)={\displaystyle \frac{2F^2(𝐪)}{N}}{\displaystyle \underset{i,j}{}}e^{i𝐪(𝐫_i𝐫_j)}S_iS_j`$ (222)
gives the static response. We are only interested in the dynamic response which can be rewritten in terms of the imaginary part of the susceptibility as
$`S_D(𝐪,\omega )={\displaystyle \frac{2F^2(𝐪)}{N\pi }}\left[{\displaystyle \frac{1}{1e^{\beta \omega }}}\right]\mathrm{}\left\{\chi (𝐪,\omega )\right\}.`$ (223)
In the previous subsections we have calculated the $`q=0`$ component of $`\mathrm{}\left\{\chi (𝐪,\omega )\right\}`$ but for neutron scattering we need a more detailed description. In order to do that we again consider the clusters as independent scatters. Moreover, since the cluster scatters as a whole we write for a cluster of $`N`$ atoms that
$`\mathrm{}\left\{\chi _N(𝐪,\omega )\right\}=F_N^2(q)\mathrm{}\left\{\chi (\omega ,\mathrm{\Delta }(N),T)\right\}`$ (224)
where $`F_N^2(q)`$ is the form factor of the cluster and $`\mathrm{}\left\{\chi (\omega ,\mathrm{\Delta }(N),T)\right\}`$ is given in (206). The main theoretical problem is to calculate the cluster form factor since in a percolating lattice clusters have different sizes and shapes. We observe however, that if we are sufficiently close to the ordering vector $`q=Q`$ we can write
$`F_N(q)=e^{R^2(N)[(qQ)^2Q^2]}`$ (225)
where $`R(N)`$ is the gyration radius of the cluster and is roughly given by $`N^{1/D}a`$ (we normalized the form factor such that $`F_N(q=0)=1`$). Substituting (225) and (206) into (224) and averaging the result with (200) we easily find
$`\mathrm{}\left\{\chi (𝐪,\omega )\right\}{\displaystyle \frac{\left[\mathrm{ln}(2\omega _0/\omega )\right]^{1\theta }\mathrm{tanh}(\beta \omega /2)}{(\omega /\omega _0)^{1\lambda }}}\mathrm{exp}\left\{\left[\mathrm{ln}(2\omega _0/\omega )/\gamma \right]^{2/D}[(qQ)^2Q^2]a^2\right\}`$ (226)
where $`c`$ is a constant of order of unit. Notice that for $`q=0`$ we obtain (216). Also notice that $`𝑑𝐪\mathrm{}\left\{\chi (𝐪,\omega )\right\}\mathrm{}\left\{\chi (𝐪=0,\omega )\right\}`$ apart from logarithmic corrections.
A first direct consequence of the Griffiths phase is the weak interplay between $`q`$ and $`\omega `$ which are essentially decoupled. The scattering is peaked around the ordering wave-vector $`Q`$ and the width of the peak is given by
$`\delta q(\omega ){\displaystyle \frac{1}{a\left[\mathrm{ln}(2\omega _0/\omega )/\gamma \right]^{1/D}}}`$ (227)
which is independent of temperature and weakly dependent on the frequency. Actually only measurements over many decades of frequency can detect any significant changes in the width of the peak. Notice, moreover, that the width of the peak does not depend on the distance from the QCP which is given by $`\lambda \xi ^D(p)`$ and therefore we predict the width of the peak to be essentially independent on the concentration. All these results, however, have a limitation which is the assumption of the gaussian form factor given in (225).
#### 3 Specific Heat
An important function in the context of NFL physics is the low temperature specific heat which is given by
$`\overline{C}_V(H,T)`$ $``$ $`\beta ^2{\displaystyle _0^{\omega _0}}𝑑\mathrm{\Delta }\mathrm{\Delta }^{\lambda 1}(E_H^2+\mathrm{\Delta }^2)sech^2\left(\beta \sqrt{E_H^2+\mathrm{\Delta }^2}\right)\left[\mathrm{ln}{\displaystyle \frac{\omega _0}{\mathrm{\Delta }}}\right]^{1\theta }`$ (228)
$``$ $`\beta ^\lambda {\displaystyle _{\beta E_H}^{\beta \sqrt{E_H^2+\omega _0^2}}}𝑑x{\displaystyle \frac{x^3sech^2(x)}{\left(x^2(\beta E_H)^2\right)^{1\lambda /2}}}\left[\mathrm{ln}{\displaystyle \frac{\beta \omega _0}{\sqrt{x^2(\beta E_H)^2}}}\right]^{1\theta }`$ (229)
which has the scaling form:
$`\overline{C}_V(H,T)=T^\lambda g_\lambda ({\displaystyle \frac{H}{T}},{\displaystyle \frac{T}{\omega _0}}).`$ (230)
Again we will consider only the case of $`T,E_H\omega _0`$. At very low fields, $`E_HT`$, we see that (229) reduces to:
$`\overline{C}_V(H,T)T^\lambda (\mathrm{ln}\beta \omega _0)^{1\theta }.`$ (231)
Observe that in this case the specific heat coefficient diverges at low temperatures:
$`\gamma (T)={\displaystyle \frac{\overline{C}_V(H,T)}{T}}{\displaystyle \frac{1}{T^{1\lambda }}}(\mathrm{ln}\beta \omega _0)^{1\theta }`$ (232)
with the same leading exponent as the susceptibility. Notice that the scaling function is such that $`g_\lambda (x0)1`$.
On the other hand at high fields, $`E_HT`$, we have as leading behavior:
$`\overline{C}_V(H,T)`$ $``$ $`{\displaystyle \frac{(\beta H)^{2+\lambda /2}}{(\beta \omega _0)^\lambda }}e^{\beta H}`$ (233)
$``$ $`{\displaystyle \frac{H^{2+\lambda /2}}{T^{2\lambda }}}\mathrm{exp}\left\{{\displaystyle \frac{H}{T}}\right\}`$ (234)
which has a Sckotty anomaly due to the field. All these results are summarized in Fig.8. These results are all valid for temperatures above $`T^{}`$. Below $`T^{}`$ we expect the crossover discussed in the previous section and a deviation from power law behavior.
## VII Conclusions
The problem of the effect of disorder in magnetic phase transitions is probably one of the most important problems of modern condensed matter. The intrinsic complexity of this problem is related to the fact that at very low temperatures close to a QCP statistical fluctuations due to disorder and quantum fluctuations due to quantum mechanics are intrinsically linked to each other. This is especially true for the alloys of U and Ce because of the existence of the Kondo effect.
We have presented a unified picture of Kondo hole and ligand systems and the role played by alloying in the physical properties of these systems. Starting from the Kondo lattice model we have shown that it is possible to write down an effective quantum action (15) which contains both the RKKY and the Kondo effect by tracing energy degrees of freedom outside of a shell of momentum $`\mathrm{\Lambda }`$ around the Fermi surface. This treatment allows for a direct comparison between the effects of RKKY and Kondo and reproduces, at the Hamiltonian level, the famous Doniach argument. We have shown, however, that alloying leads to a broad distribution of exchange couplings and therefore to a very inhomogeneous environment. This should be contrasted with old approaches to this problem where the concept of chemical pressure has been introduced. Just briefly, this concept proposes that chemical substitution is equivalent to applying pressure to the system: if one substitutes a smaller atom by a larger one then the lattice expands and it is equivalent to negative pressure; if we substitute a larger atom by a smaller one the lattice contracts and therefore it is equivalent to positive pressure. The next step on the chemical pressure argument is to assert that the exchange $`J`$ over the entire lattice changes uniformly under pressure and the system can undergo a phase transition from an ordered state to a Kondo compensated state as shown by $`T_N`$ in Fig.6. From our point of view chemical pressure and applied pressure have different effects on the physics of these systems.
A simple example of the large difference between pressure and chemical substitution can be captured in specific heat measurements in CeCu<sub>6</sub> under pressure and with substitution of Ce by Th in Ce<sub>1-x</sub>Th<sub>x</sub>Cu<sub>6</sub> . From the experimental data on CeCu<sub>6</sub> one finds that the inverse of $`\gamma `$ can be fitted by a linear relation with the applied pressure, $`P`$:
$$\frac{\gamma _0}{\gamma (P)}=1+1.017\frac{P}{P_{max}}$$
(235)
where $`\gamma _0=1.67`$ J mole<sup>-1</sup> K<sup>-2</sup> is the zero pressure result and $`P_{max}=8.8`$ kbar is the maximum applied pressure in the experiment. Moreover, the unit cell volume in pure CeCu<sub>6</sub> at zero applied pressure is $`𝒱_0=420.6\AA ^3`$ while at maximum applied pressure, $`P_{max}`$, of $`8.8`$ kbar the volume is $`416.5\AA ^3`$. Assuming a linear relation between the volume and the pressure (a constant bulk modulus) one finds,
$`{\displaystyle \frac{𝒱(P)}{𝒱_0}}=1\beta {\displaystyle \frac{P}{P_{max}}}`$ (236)
where $`\beta =9.7\times 10^3`$. In the case of the doped samples it is experimentally established that pure ThCu<sub>6</sub> has a unit cell volume of $`417.1\AA ^3`$. Since the contraction of the lattice seems to be homogeneous and isotropic with doping one finds that the volume as a function of doping can be written as,
$`{\displaystyle \frac{𝒱(c)}{𝒱_0}}=1\alpha c`$ (237)
where $`\alpha =8.3\times 10^3`$. The main idea behind the “chemical pressure” theory is to assert that (236) and (237) are equivalent to each other, that is, there is a one-to-one correspondence between an applied pressure and doping the system. If one assumes that this is indeed the case we can find the correspondence between applied pressure and doping:
$$\frac{P}{P_{max}}=\frac{\alpha }{\beta }c.$$
(238)
Using (238) together with (235) one finds that the “chemical pressure” theory yields to the following relation,
$$\left[\frac{\gamma _0}{\gamma (c)}\right]_P=1+\zeta _Pc$$
(239)
where $`\zeta _P=0.858`$. Now one has to compare (239) to the actual dependence of $`\gamma `$ on doping. Since we are interested in the low doping regime one uses the experimental data up to $`c=0.4`$ which can be well fitted by a straight line of the form ,
$$\frac{\gamma _0}{\gamma (c)}=1+\zeta _Dc$$
(240)
where $`\zeta _D=0.249`$. We must stress that (240) is obtained from an extrapolation of the data to the zero temperature limit. A more careful look at the data, however, seems to indicate a divergence of the specific heat exponent in Ce<sub>1-x</sub>Th<sub>x</sub>Cu<sub>6</sub>. We will however, insist using the extrapolation just to check the identity between chemical pressure and real pressure. We see the clear difference between the actual doped sample (240) and the calculation based on the “chemical pressure” argument, (239), for the same value of doping. We find that $`\zeta _P/\zeta _D3.4`$ and therefore a quite large difference between pressure and doping. The conclusion is that there is more than a change in the lattice constants or the volume of the unit cell when a system is doped. Actually, from our arguments we actually expect that the ration $`\zeta _D/\zeta _P`$ to be temperature dependent!
In this paper we have proposed that the cut-off $`\mathrm{\Lambda }`$ which determines the energetic crossover between Kondo effect and RKKY coupling is directly related to the number of compensated moments in the system which is given by (29). Thus, $`\mathrm{\Lambda }`$ has to be determined in a self-consistent way from the disorder and characteristic energy scales in the problem. We argue that this scale is not determined by the Kondo temperature alone but also by the RKKY coupling between moments. Since the RKKY coupling involves states well deep into the Fermi surface $`\mathrm{\Lambda }`$ can be rather large and of the order of the Fermi momentum. As a secondary consequence of this discussion we show that the exhaustion paradox proposed by Nozières does not apply in the concentrated limit of the Kondo lattice because it fails to take into account the high energy degrees of freedom associated with the RKKY interaction.
Within the percolation theory of the Kondo lattice we show that close to the QCP strong response from quantum fluctuation magnetic clusters gives singular contributions at low temperatures which are known to be Griffiths-McCoy singularities. We have presented a phenomenological approach for this problem which allows the calculation of the various physical quantities of interest such as magnetic susceptibilities (as given in (187) and (212)), dynamical form factors for neutron scattering (given in (226)) and thermodynamic quantities (such as the specific heat in (192) and (231)). The integrals involved are rather complicated and sometimes one has to perform them numerically. Moreover, we have shown that when a magnetic field is applied to the system there is a recovery of the ordinary behavior due to the quenching of the clusters in the presence of the magnetic field (see (234)). We have also discussed the various scalings in the presence of a magnetic field which are given in (210) and (230).
We have also reviewed the single impurity Kondo problem and have shown that in the anisotropic case, which is relevant for the case of spin-orbit coupling, it can be interpreted as the dissipative dynamics of a two level system in a transverse field. We have shown that while the XY component of the Kondo coupling acts as a transverse field on the magnetic spin, the Ising component gives rise to dissipation due to the production of particle-hole excitations at the Fermi surface. Using exact results for the dissipative two level system we have shown that the physical quantities differ slightly from the non-dissipative two level system by the presence of the parameter $`\alpha `$ that controls the coupling of the two level system to the electronic heat bath. We have also revisited the two impurity Kondo problem and have shown that the stable fixed points are related to the coherent flipping motion of the spins as a single degree of freedom (as shown in (112) and (121)). This is a clear sign of cluster behavior, that is, coherent quantum mechanical motion. In a cluster of $`N`$ atoms we have identified two mechanisms for quantum tunneling: in XYZ magnets the anisotropy of the RKKY interaction generated coherent cluster tunneling (with a cluster Kondo temperature given in (150)) while in XXZ or Heisenberg systems the Kondo effect generates the tunneling (with a cluster Kondo temperature given in (162)). In both cases the tunneling splitting is an exponential function of the number of atoms in the cluster as shown in (140). This calculation allows us to relate the microscopic quantities in the Hamiltonian (15) and the phenomenological parameters in Section VI. We have shown that there is a cluster Kondo effect which is parameterized also by a dissipative constant $`\alpha (N)=(N/N_c)^\phi `$ where $`\phi `$ is an exponent which depends on the type of coupling with the particle-hole continuum (see (155)) and $`N_c`$ (given in (157)-(161) and (164)-(165)) is the largest size of the cluster for which the Kondo effect ceases to exist and the cluster are frozen. Thus, the Kondo temperature of the cluster vanishes at $`N_c`$. In this respect our picture is quite close to the Kondo disorder picture proposed for the explanation of NFL in these systems . We have to point out, however, that the microscopic origin of both pictures is quite different since in our scenario the system has to be close to the magnetic phase transition and the Kondo temperature is defined for an extensive object which is determined by the laws of percolation theory.
Based on these results we proposed a dissipative droplet model for the magnetic behavior in disordered metallic magnets which is described by the action (170). This picture is the extension of the quantum droplet model proposed by Thill and Huse for the case of insulating magnets. We have shown that in this model there is a crossover temperature $`T^{}E_Fe^{\gamma N_c/\phi }`$ above which the results of the Griffiths-McCoy phase are still valid with power law behavior in many physical quantities (see (180)). Below $`T^{}`$ the behavior of the physical quantities is modified by dissipation and a new behavior emerges where the magnetic susceptibility and specific heat diverge stronger than power law and do not scale together as shown in (187) and (192). Actually the magnetic susceptibility has a stronger divergence than the specific heat coefficient. This could perhaps explain why the exponents $`\lambda `$ found from magnetic susceptibility measurements are systematically smaller than the ones found in the specific heat data . We have estimated, however, that $`T^{}`$ is rather small. Thus, power laws as given in (212) and (231) should dominate the crossover behavior at low temperatures. We have calculated the thermodynamic properties of the clusters alone but it is clear that in a real system one has to add also the contributions coming from the paramagnetic environment (Fermi liquid or heavy Fermi liquid) which are given by (1) and (4), for instance. Since the response functions we have calculated are singular as $`T0`$ they dominate the response at low enough temperatures.
Finally, we should mention that our picture is based on two main features: the short range character of the disorder and the complete decoupling of the clusters. On the one hand, if the disorder is long range correlated then stronger divergences of the physical quantities are possible . On the other hand, if the cluster are not decoupled but there is a residual interaction between them it is possible that the system reaches a spin-glass state at lower temperatures. The nature of the spin glass state depends on the dissipation. If the freezing temperature is of order of $`T^{}`$ then the dissipation will kill quantum coherence and a classical spin glass state is formed. If $`T^{}`$ is smaller than the freezing temperature then a quantum spin glass ground state is possible . The existence of such phases can only be decided on experimental basis.
Thus, in this paper we put forward the theoretical basis for Griffiths-McCoy singularities in the paramagnetic region close to the QCP for magnetic order of U and Ce intermetallics. We show that strong temperature dependence in the physical quantities is expected at low temperatures due to the quantum mechanical response of magnetic clusters. We believe that our picture is in agreement with a large number of experiments in these systems . It would be interesting to investigate further the existence of a even lower temperature regime where dissipation plays a decisive role in the physics of these systems.
We would like to acknowledge E. Abrahams, I. Affleck, M. Aronson, G. Castilla, T. Costi, D. Cox, I. Dzyaloshinsky, R. Egger, F. Guinea, J. Kaufman, H. v. Löhneysen, D. MacLaughlin, B. Maple, N. Prokof’ev, R. Ramazashvili, H. Rieger, H. Saleur, S. Senthil, G. Stewart and A. Tsvelik for illuminating comments and discussions. A. H. C. N. acknowledges partial support from the Alfred P. Sloan foundation and a Los Alamos CULAR grant under the auspices of the U. S. Department of Energy.
## A Perturbation Theory
Our starting point is (13) where we integrate out the electrons in regions $`\mathrm{\Omega }_1`$ and $`\mathrm{\Omega }_2`$ and obtain an effective action for the electrons residing in region $`\mathrm{\Omega }_0`$:
$`Z={\displaystyle D𝐒(n,t)D\overline{\psi }_0(𝐫,t)D\psi _0(𝐫,t)e^{iS_{eff}(𝐒,\overline{\psi }_0,\psi _0)}}`$ (A1)
where
$`S_{eff}=S_0itr\mathrm{ln}(G^1)i{\displaystyle \underset{𝐫,𝐫^{}}{}}{\displaystyle 𝑑t𝑑t^{}\underset{\alpha ,\alpha ^{}}{}\underset{\iota ,\iota ^{}}{}\overline{\eta }_{\alpha ,\iota }(𝐫,t)G_{\alpha ,\alpha ^{}}^{\iota ,\iota ^{}}(𝐫,t;𝐫^{},t^{})\eta _{\alpha ^{},\iota ^{}}(𝐫^{},t^{})}`$ (A2)
where $`S_0`$ is (10) for the electrons in region $`\mathrm{\Omega }_0`$ only. Furthermore,
$`\eta _{\alpha ,\iota }(𝐫,t)={\displaystyle \underset{a,b}{}}J_{a,b}(𝐫)S^a(𝐫,t)\tau _{\alpha ,\beta }^b\psi _{\beta ,0}(𝐫,t),`$ (A3)
and
$`𝐫,t,\iota ,\alpha |G^1|𝐫^{},t^{},\iota ^{},\alpha )=\left(i{\displaystyle \frac{}{t}}+\mu +ϵ_\alpha (i)\right)\delta _{𝐫,𝐫^{}}\delta (tt^{})\delta _{\iota ,\iota ^{}}\delta _{\alpha ,\alpha ^{}}`$ (A4)
$`+\sigma _{\iota ,\iota ^{}}^x{\displaystyle \underset{a,b}{}}J_{a,b}(𝐫)S^a(𝐫,t)\tau _{\alpha ,\beta }^b\delta (tt^{})\delta _{𝐫,𝐫^{}}`$ (A5)
where $`\sigma _{\iota ,\iota ^{}}^x`$ is a Pauli matrix which works in the subspace of states $`\mathrm{\Omega }_1`$ and $`\mathrm{\Omega }_2`$. Moreover,
$`J_{a,b}(𝐫)={\displaystyle \underset{n}{}}J_{a,b}(n)\delta _{𝐫,𝐫_n}.`$ (A6)
It is convenient to define the bare Green’s function as the solution of the $`J=0`$ problem, that is,
$`\left(i{\displaystyle \frac{}{t}}+\mu +ϵ_\alpha (i)\right)G_{0,\alpha }(𝐫𝐫^{},tt^{})=\delta (tt^{})\delta _{𝐫,𝐫^{}}.`$ (A7)
Notice that the time ordered Green’s function can be trivially obtained from the above equation and reads,
$`G_{0,\alpha }(𝐤,\omega )={\displaystyle \frac{1}{\omega +\mu ϵ_\alpha (𝐤)\frac{i}{\tau }sgn(ϵ_\alpha \mu )}}`$ (A8)
where we have introduced the electron relaxation time $`\tau `$ due to weak disorder. Thus, in the regions of momentum space introduced in Section (III) we have
$`G_{0,\alpha ,1}(𝐤,\omega )={\displaystyle \frac{\mathrm{\Theta }(k_{F,\alpha }\mathrm{\Lambda }k)}{\omega +\mu ϵ_\alpha (𝐤)+\frac{i}{\tau }}}`$ (A9)
$`G_{0,\alpha ,2}(𝐤,\omega )={\displaystyle \frac{\mathrm{\Theta }(kk_{F,\alpha }\mathrm{\Lambda })}{\omega +\mu ϵ_\alpha (𝐤)\frac{i}{\tau }}}.`$ (A10)
Notice that the trace over the fast modes has generated two kinds of terms in (A2): new interactions between the $`\mathrm{\Omega }_0`$ electrons and the spins (through $`\eta `$) and spin-spin interactions through $`G`$. In what follows we will consider perturbation theory of $`J`$ for these fast modes. First let us consider the new interaction terms between electrons and spins. Since $`\eta `$ is already order $`J`$ we can consider $`GG_0`$ in (A2) and rewrite
$`I={\displaystyle \underset{𝐫,𝐫^{}}{}}{\displaystyle 𝑑t𝑑t^{}\underset{\alpha ,\alpha ^{}}{}\underset{\iota ,\iota ^{}}{}\overline{\eta }_{\alpha ,\iota }(𝐫,t)G_{\alpha ,\alpha ^{}}^{\iota ,\iota ^{}}(𝐫,t;𝐫^{},t^{})\eta _{\alpha ^{},\iota ^{}}(𝐫^{},t^{})}`$ (A11)
$`I={\displaystyle \underset{𝐫,𝐫^{}}{}}{\displaystyle }dt{\displaystyle }dt^{}{\displaystyle \underset{\gamma ,\delta }{}}\overline{\psi }_{\gamma ,0}(𝐫,t)[{\displaystyle \underset{a,b,c,d}{}}{\displaystyle \underset{\alpha ,\beta }{}}{\displaystyle \underset{\iota ,\iota ^{}}{}}J_{a,b}(𝐫)J_{c,d}(𝐫^{})S^a(𝐫,t)S^c(𝐫^{},t^{})`$ (A12)
$`\tau _{\alpha ,\gamma }^b\tau _{\delta ,\beta }^dG_{\alpha ,\alpha ^{}}^{\iota ,\iota ^{}}(𝐫,t;𝐫^{},t^{})]\psi _{\delta ,0}(𝐫^{},t^{}).`$ (A13)
Thus, in order to evaluate $`I`$ one needs an expression for $`G`$. From (A5) we have (symbolically),
$`G^1`$ $`=`$ $`G_0^1+𝒥=G_0^1\left(1+G_0𝒥\right)`$ (A14)
$`G`$ $`=`$ $`\left(1+G_0𝒥\right)^1G_0`$ (A15)
$``$ $`G_0+G_0𝒥G_0+h.o.t.`$ (A16)
where $`𝒥`$ is the second term in (A5).
To leading order one has,
$`I`$ $``$ $`{\displaystyle \underset{𝐫,𝐫^{}}{}}{\displaystyle 𝑑t𝑑t^{}\underset{\gamma ,\delta }{}\overline{\psi }_{\gamma ,0}(𝐫,t)\psi _{\delta ,0}(𝐫^{},t^{})\underset{a,b,c,d}{}J_{a,b}(𝐫)J_{c,d}(𝐫^{})S^a(𝐫,t)S^c(𝐫^{},t^{})}`$ (A17)
$`\times `$ $`{\displaystyle \underset{\alpha }{}}\tau _{\alpha ,\gamma }^b\tau _{\delta ,\alpha }^d\left({\displaystyle \underset{\iota }{}}G_{0,\alpha ,\iota }(𝐫𝐫^{},tt^{})\right)`$ (A18)
which is a electron spin-flip scattering process involving two different local moments. Observe that this interaction is mediated by the single particle Green’s function $`G_0`$ which can be evaluated from (A10). After a trivial frequency integral we find
$`{\displaystyle \underset{\iota }{}}G_{0,\alpha ,\iota }(𝐫,t)`$ $`=`$ $`{\displaystyle \frac{ie^{|t|/\tau }}{\rho _L}}{\displaystyle }{\displaystyle \frac{d𝐤}{(2\pi )^3}}e^{i(𝐤𝐫(ϵ_𝐤\mu )t)}[\mathrm{\Theta }(kk_{F,\alpha }\mathrm{\Lambda })\mathrm{\Theta }(t)`$ (A19)
$``$ $`\mathrm{\Theta }(k_{F,\alpha }\mathrm{\Lambda }k)\mathrm{\Theta }(t)]`$ (A20)
where $`\rho _L=N/V`$ is the lattice density. Since we are only interested in the asymptotic behavior of the system we observe that when $`t\mathrm{}`$ the integral is dominated by the saddle point at $`\mu =ϵ_\alpha (k_{F,\alpha })`$ but the $`\mathrm{\Theta }`$ functions are only finite outside this region, thus, the above integral has very fast oscillations given null contribution. In order to see this result explicitly let us expand around the saddle point, that is, we change variables:
$`𝐤=k_{F,\alpha }+q𝐮_{F,\alpha }`$ (A21)
where $`𝐮_{F,\alpha }`$ is a vector perpendicular to the Fermi surface (that is, in the direction of the Fermi velocity). Thus we can write
$`ϵ_𝐤`$ $``$ $`\mu +v_{F,\alpha }q`$ (A22)
$`k`$ $``$ $`k_F+q`$ (A23)
and the integral above becomes
$`{\displaystyle \underset{\iota }{}}G_{0,\alpha ,\iota }(𝐫,t)`$ $``$ $`{\displaystyle \frac{ie^{|t|/\tau }}{\rho _L}}{\displaystyle \underset{𝐤_{F,\alpha }}{}}{\displaystyle \frac{k_{F,\alpha }^2}{(2\pi )^3}}{\displaystyle _{\mathrm{}}^+\mathrm{}}dqe^{i[(𝐤_{F,\alpha }+q𝐮_{F,\alpha })𝐫qv_{F,\alpha }t]}[\mathrm{\Theta }(q\mathrm{\Lambda })\theta (t)`$ (A24)
$``$ $`\mathrm{\Theta }(\mathrm{\Lambda }q)\mathrm{\Theta }(t)]e^{\nu |q|}`$ (A25)
where we introduced an infinitesimal converging factor $`\nu 0`$. The integrals are now trivial and give
$`{\displaystyle \underset{\iota }{}}G_{0,\alpha ,\iota }(𝐫,t)`$ $``$ $`{\displaystyle \underset{𝐤_{F,\alpha }}{}}{\displaystyle \frac{k_{F,\alpha }^2}{8\pi ^3\rho _L}}e^{i𝐤_{F,\alpha }𝐫}{\displaystyle \frac{e^{i\mathrm{\Lambda }(𝐮_{F,\alpha }𝐫v_{F,\alpha }t)sgn(t)}}{𝐮_{F,\alpha }𝐫v_{F,\alpha }ti\nu sgn(t)}}`$ (A26)
which shows that at long times the Green’s function oscillates with frequency $`v_F\mathrm{\Lambda }`$ and therefore averages out to zero.
We now look at the the first term in (A2) and write
$`tr\mathrm{ln}(G_0^1+𝒥)`$ $`=`$ $`tr\mathrm{ln}[G_0^1(1+G_0𝒥)]=tr\mathrm{ln}G_0^1+tr\mathrm{ln}(1+G_0𝒥)`$ (A27)
$``$ $`tr\mathrm{ln}G_0^1+tr(G_0𝒥){\displaystyle \frac{1}{2}}tr(G_0𝒥G_0𝒥)+h.o.t.`$ (A28)
In order to simplify the notation we will use a simple dummy index $`i=(𝐫,t,\alpha ,\iota )`$ for the matrix elements. Observe that $`tr(G_0𝒥)=_{i,j}G_{0,i}\delta _{i,j}𝒥_{i,j}=_iG_{0,i}𝒥_{i,i}=0`$ because $`\sigma _{\iota ,\iota }^x=0`$ in (A5). The second order term in (A28) becomes
$`tr(G_0𝒥G_0𝒥)`$ $`=`$ $`{\displaystyle \underset{i,j,k,l}{}}G_{0,i}\delta _{i,j}𝒥_{j,k}G_{0,k}\delta _{k,l}𝒥_{k,i}`$ (A29)
$`=`$ $`{\displaystyle \underset{i,k}{}}{\displaystyle \underset{\alpha ,\gamma }{}}G_{0,i}𝒥_{i,k}G_{0,k}𝒥_{k,i}.`$ (A30)
Using (A5) we find explicitly
$`tr(G_0𝒥G_0𝒥)={\displaystyle 𝑑𝐫𝑑𝐫^{}𝑑t𝑑t^{}\underset{a,b,c,d}{}J_{a,b}(𝐫)J_{c,d}(𝐫^{})S^a(𝐫,t)S^c(𝐫^{},t^{})\mathrm{\Pi }_{b,d}(𝐫𝐫^{},tt^{})}`$ (A31)
where
$`\mathrm{\Pi }_{b,d}(𝐫,t)=2{\displaystyle \underset{\iota ,\iota ^{}}{}}{\displaystyle \underset{\alpha ,\gamma }{}}\tau _{\alpha ,\gamma }^b\tau _{\gamma ,\alpha }^dG_{0,\alpha ,i}(𝐫,t)\sigma _{\iota ,\iota ^{}}^xG_{0,\gamma ,\iota ^{}}(𝐫,t)`$ (A32)
is the polarization function of the electron band. It can be rewritten in momentum space as
$`\mathrm{\Pi }_{b,d}(𝐩,\mathrm{\Omega })={\displaystyle \frac{1}{2}}{\displaystyle \underset{\alpha ,\gamma }{}}{\displaystyle \underset{\iota ,\iota ^{}}{}}\tau _{\alpha ,\gamma }^b\tau _{\gamma ,\alpha }^d\sigma _{\iota ,\iota ^{}}^x{\displaystyle \frac{d\omega }{2\pi }\frac{d𝐤}{(2\pi )^d}G_{0,\alpha ,\iota }(𝐤,\omega )G_{0,\gamma ,\iota ^{}}(𝐤+𝐩,\omega +\mathrm{\Omega })}`$ (A33)
which can be evaluated once the conduction band is known. Thus, to second order the integration of the fast degree of freedom generates an interaction among the spins in different sites. One further approximation which is usually used is to take the static limit in (A31). This approximation is good if one looks at times scales which are much larger than the scattering time of the electrons between moments which is $`\mathrm{}/v_{F,\alpha }`$ where $`\mathrm{}`$ is the distance between moments and $`v_{F,\alpha }`$ is the Fermi velocity for the $`\alpha `$ branch. We replace $`\mathrm{\Pi }_{b,d}(𝐩,\omega )`$ by $`\mathrm{}[\mathrm{\Pi }_{b,d}(𝐩,0)]`$ (since $`\mathrm{}[\mathrm{\Pi }_{b,d}(𝐩,0)]=0`$). In this limit the second order contribution is the RKKY interaction between localized spins (which depends on the cut-off $`\mathrm{\Lambda }`$) which can be written as
$`H_{RKKY}(\mathrm{\Lambda })={\displaystyle \underset{m,n}{}}{\displaystyle \underset{a,b}{}}\mathrm{\Gamma }_{a,b}(𝐫_n𝐫_m,\mathrm{\Lambda })S^a(𝐫_n)S^b(𝐫_m)`$ (A34)
where
$`\mathrm{\Gamma }_{a,b}(𝐫,\mathrm{\Lambda })={\displaystyle \underset{c,d}{}}J_{a,c}J_{b,d}{\displaystyle }d𝐪e^{i𝐪𝐫}\mathrm{}[\mathrm{\Pi }_{c,d}(𝐪,\mathrm{\Omega }=0,\mathrm{\Lambda })].`$ (A35)
The integral in (A35) can be written directly from (A33) and reads
$`{\displaystyle \underset{\iota ,\iota ^{}}{}}`$ $`\sigma _{\iota ,\iota ^{}}^x{\displaystyle 𝑑𝐤^{}e^{i(𝐤𝐤^{})𝐫}\frac{d\omega }{2\pi }\frac{d𝐤}{(2\pi )^d}G_{0,\alpha ,\iota }(𝐤,\omega )G_{0,\gamma ,\iota ^{}}(𝐤^{},\omega )}=`$ (A36)
$`=`$ $`{\displaystyle }d𝐤^{}{\displaystyle }{\displaystyle \frac{d𝐤}{(2\pi )^d}}e^{i(𝐤𝐤^{})𝐫}{\displaystyle \frac{d\omega }{2\pi }}({\displaystyle \frac{\mathrm{\Theta }(k_{F,\alpha }\mathrm{\Lambda }k)\mathrm{\Theta }(k^{}k_{F,\gamma }\mathrm{\Lambda })}{(\omega +\mu ϵ_\alpha (k)i/\tau )(\omega +\mu ϵ_\gamma (k^{})+i/\tau )}}`$ (A37)
$`+`$ $`{\displaystyle \frac{\mathrm{\Theta }(kk_{F,\alpha }\mathrm{\Lambda })\mathrm{\Theta }(k_{F,\gamma }\mathrm{\Lambda }k^{})}{(\omega +\mu ϵ_\alpha (k)+i/\tau )(\omega +\mu ϵ_\gamma (k^{})i/\tau )}})`$ (A38)
$`=`$ $`i{\displaystyle }d𝐤^{}{\displaystyle }{\displaystyle \frac{d𝐤}{(2\pi )^d}}e^{i(𝐤𝐤^{})𝐫}({\displaystyle \frac{\mathrm{\Theta }(k_{F,\alpha }\mathrm{\Lambda }k)\mathrm{\Theta }(k^{}k_{F,\gamma }\mathrm{\Lambda })}{ϵ_\gamma (k^{})ϵ_\alpha (k)+2i/\tau }}`$ (A39)
$`+`$ $`{\displaystyle \frac{\mathrm{\Theta }(kk_{F,\alpha }\mathrm{\Lambda })\mathrm{\Theta }(k_{F,\gamma }\mathrm{\Lambda }k^{})}{ϵ_\alpha (k)ϵ_\gamma (k^{})+2i/\tau }})`$ (A40)
Let us consider the special case where the dispersion are the same in both spin branches and the Fermi surface is assumed to be spherical. In this case the bare Green’s function does not depend on the spin label and $`\mathrm{\Pi }_{b,d}(𝐫,t)=\delta _{b,d}\mathrm{\Pi }^0(𝐫,t)`$. We can write (A40) as
$`(r,\mathrm{\Lambda })`$ $`=`$ $`i{\displaystyle }d𝐤^{}{\displaystyle }{\displaystyle \frac{d𝐤}{(2\pi )^3}}e^{i(𝐤𝐤^{})𝐫}({\displaystyle \frac{\mathrm{\Theta }(k_F\mathrm{\Lambda }k)\mathrm{\Theta }(k^{}k_F\mathrm{\Lambda })}{ϵ(k^{})ϵ(k)+2i/\tau }}`$ (A41)
$`+`$ $`{\displaystyle \frac{\mathrm{\Theta }(kk_F\mathrm{\Lambda })\mathrm{\Theta }(k_F\mathrm{\Lambda }k^{})}{ϵ(k)ϵ(k^{})+2i/\tau }})=`$ (A42)
$`=`$ $`{\displaystyle 𝑑𝐤^{}\frac{d𝐤}{(2\pi )^3}\frac{\mathrm{cos}[(𝐤𝐤^{})𝐫]}{ϵ(k^{})ϵ(k)+2i/\tau }\mathrm{\Theta }(k_F\mathrm{\Lambda }k)\mathrm{\Theta }(k^{}k_F\mathrm{\Lambda })}`$ (A43)
$`=`$ $`{\displaystyle \frac{2}{\pi r^2}}{\displaystyle _0^{k_F\mathrm{\Lambda }}}𝑑k{\displaystyle _{k_F+\mathrm{\Lambda }}^{\mathrm{}}}𝑑k^{}{\displaystyle \frac{kk^{}\mathrm{sin}(kr)\mathrm{sin}(k^{}r)}{ϵ(k^{})ϵ(k)+2i/\tau }}`$ (A44)
When $`\mathrm{\Lambda }0`$ and $`E_F1/\tau `$ we find the usual result
$`(r,0)={\displaystyle \frac{4mk_F^2}{\pi r^2}}e^{r/\mathrm{}}_0(2k_Fr)`$ (A45)
where
$`\mathrm{}={\displaystyle \frac{k_F\tau }{m}}=v_F\tau `$ (A46)
is the electron mean-free path and
$`_0(x)={\displaystyle \frac{\pi }{4}}\left[{\displaystyle \frac{\mathrm{sin}(x)x\mathrm{cos}(x)}{x^2}}\right]`$ (A47)
which leads to the final expression for the RKKY interaction in $`d=3`$.
Let us now consider the corrections to the RKKY interaction due to a finite $`\mathrm{\Lambda }/k_F`$. In all the cases here we will consider the situation in which $`k_F\mathrm{\Lambda }1/\tau `$. From (A44) we see that the correction has the form
$`\delta (r,\mathrm{\Lambda })={\displaystyle \frac{4mk_F^2}{\pi r^2}}e^{r/\mathrm{}}(2k_Fr,\mathrm{\Lambda }/k_F)`$ (A48)
where, $`=_1+_2+_3`$, with
$`_1(2x,y)`$ $`=`$ $`{\displaystyle _0^1}𝑑z{\displaystyle _1^{1+y}}𝑑w{\displaystyle \frac{zw\mathrm{sin}(xz)\mathrm{sin}(xw)}{z^2w^2+i\eta }}`$ (A49)
$`_2(2x,y)`$ $`=`$ $`{\displaystyle _{1y}^1}𝑑z{\displaystyle _1^{\mathrm{}}}𝑑w{\displaystyle \frac{zw\mathrm{sin}(xz)\mathrm{sin}(xw)}{z^2w^2+i\eta }}`$ (A50)
$`_3(2x,y)`$ $`=`$ $`{\displaystyle _{1y}^1}𝑑z{\displaystyle _1^{1+y}}𝑑w{\displaystyle \frac{zw\mathrm{sin}(xz)\mathrm{sin}(xw)}{z^2w^2+i\eta }}`$ (A51)
where $`\eta 0`$. All the integrals can be evaluated after a very tedious algebra. The final result reads
$`(x,y)`$ $`=`$ $`{\displaystyle \frac{1}{4x^2}}\{\pi \mathrm{cos}(x)(\pi 2xy)\mathrm{sin}(x)2x\mathrm{sin}(xy)2x^2y(Ci(x)Ci(xy))`$ (A52)
$`+`$ $`\mathrm{cos}[(1+y)x]\left[(Ci(x)Ci(xy))(1+y)x(Si(x)Si(xy))\right]`$ (A53)
$`+`$ $`\mathrm{sin}[(1+y)x]\left[(Si(x)Si(xy))+(1+y)x(Ci(x)Ci(xy))\right]`$ (A54)
$``$ $`\mathrm{cos}[(1y)x]\left[(Ci(x)Ci(xy))(1y)x(Si(x)+Si(xy))\right]`$ (A55)
$``$ $`\mathrm{sin}[(1y)x][(Si(x)+Si(xy))+(1y)x(Ci(x)Ci(xy))]\}`$ (A56)
where
$`Ci(x)`$ $`=`$ $`{\displaystyle _x^{\mathrm{}}}𝑑t{\displaystyle \frac{\mathrm{cos}(t)}{t}}`$ (A57)
$`Si(x)`$ $`=`$ $`{\displaystyle _x^{\mathrm{}}}𝑑t{\displaystyle \frac{\mathrm{sin}(t)}{t}}`$ (A58)
are cosine and sine integrals. A plot of $`(x,y)`$ is shown in Fig.3. Thus, the actual RKKY interaction in the presence of a finite $`\mathrm{\Lambda }/k_F`$ is given by the sum of (A45) plus (A48). Observe that besides the usual $`2k_F`$ oscillations a finite $`\mathrm{\Lambda }/k_F`$ produces other oscillations at $`2\mathrm{\Lambda }`$. Thus, the structure of the RKKY interaction is more complicated.
At very short distances, $`2k_Fr1`$, we have
$`(r,\mathrm{\Lambda }){\displaystyle \frac{2mk_F^3}{3r}}\left(1{\displaystyle \frac{\mathrm{\Lambda }}{k_F}}\right)^3`$ (A59)
which, as in the case of the usual RKKY interaction, leads to ferromagnetic coupling. The main difference, however is that the slope of the RKKY is reduced by the factor $`(1\mathrm{\Lambda }/k_F)^3`$ and therefore the strength of the ferromagnetic coupling is reduced as $`\mathrm{\Lambda }`$ increases. Furthermore, while the first zero of the RKKY interaction for $`\mathrm{\Lambda }=0`$ occurs for $`k_Fr2.2467`$; when $`\mathrm{\Lambda }0`$ the first zero is reduced as shown in Fig.4 and saturates at $`k_Fr1`$ when $`\mathrm{\Lambda }k_F`$. Overall the effect of a finite $`\mathrm{\Lambda }`$ is to increase the range of the antiferromagnetic coupling for small to moderate $`k_Fr`$. The component near the Fermi energy contributes a largely ferromagnetic component at these $`k_Fr`$. Note that overall the contributions to RKKY from the energy region away from the Fermi surface is at least comparable to that for near the Fermi surface. At intermediate distances, that is, $`\mathrm{\Lambda }^1r>k_F^1`$, the RKKY interaction can be written as
$`(r,\mathrm{\Lambda })`$ $``$ $`(r,0)+{\displaystyle \frac{2mk_F\mathrm{\Lambda }}{r^2}}\mathrm{sin}(k_Fr)[\pi \mathrm{cos}(k_Fr)2(1𝒞+Ci(2k_Fr)\mathrm{ln}(2\mathrm{\Lambda }r))\mathrm{sin}(k_Fr)`$ (A60)
$`+`$ $`2\mathrm{cos}(k_Fr)(Si(2k_Fr)+\pi /2)]`$ (A61)
where $`𝒞0.577216`$ is the Euler constant. We see from these expression that the correction to the $`\mathrm{\Lambda }=0`$ case vanishes as $`(\mathrm{\Lambda }/k_F)\mathrm{ln}(\mathrm{\Lambda }/k_F)`$ as $`\mathrm{\Lambda }/k_F0`$. The most striking difference between the behavior of the RKKY interaction with $`\mathrm{\Lambda }/k_F0`$ is it behavior at large distances. It is very simple to show that for $`r>>k_F^1,\mathrm{\Lambda }^1`$ one has
$`(r,\mathrm{\Lambda }){\displaystyle \frac{mk_F^3}{2\mathrm{\Lambda }\pi r^4}}\left[1\left({\displaystyle \frac{\mathrm{\Lambda }}{k_F}}\right)^2\right]\left(\mathrm{cos}(2k_Fr)+\mathrm{cos}(2\mathrm{\Lambda }r)\right)`$ (A62)
which decays like $`1/r^4`$ instead of the usual $`1/r^3`$. Thus, for finite $`\mathrm{\Lambda }`$ the RKKY interaction has a shorter range.
## B Mean field theory of itinerant ferromagnetism in the Kondo lattice
In this appendix we will discuss the physics of $`H_{MF}`$ in (22). Assuming that the magnetic order is homogeneous we can write
$`H_{MF}={\displaystyle \underset{𝐤,\sigma }{}}ϵ_\sigma (𝐤)c_{𝐤,\sigma }^{}c_{𝐤,\sigma }+J_zM_z{\displaystyle \underset{i=1}{\overset{N_s}{}}}(n_{i,}n_{i,})+J_zm_z{\displaystyle \underset{i=1}{\overset{N_s}{}}}S_i^z`$ (B1)
where
$`M_z`$ $`=`$ $`{\displaystyle \frac{1}{N_s}}{\displaystyle \underset{i=1}{\overset{N_s}{}}}S_i^z`$ (B2)
$`m_z`$ $`=`$ $`{\displaystyle \frac{1}{N_s}}{\displaystyle \underset{i=1}{\overset{N_s}{}}}\left(n_{i,}n_{i,}\right)`$ (B3)
where $`N_s`$ is the number of magnetic atoms. As usual we can now perform a Fourier transform of the electron operator
$`c_{n,\sigma }={\displaystyle \frac{1}{\sqrt{N}}}{\displaystyle \underset{𝐤}{}}e^{i𝐤𝐫_n}c_{𝐤,\sigma }`$ (B4)
where $`N`$ is the total number of atoms. We rewrite (B1) as
$`H_{MF}`$ $`=`$ $`{\displaystyle \underset{𝐤}{}}\left[\left(ϵ_{}(𝐤)+J_zcM_z\right)c_{𝐤,}^{}c_{𝐤,}+\left(ϵ_{}(𝐤)J_zcM_z\right)c_{𝐤,}^{}c_{𝐤,}\right]`$ (B5)
$`+`$ $`J_zm_z{\displaystyle \underset{i}{}}S_i^z`$ (B6)
which brings the Hamiltonian to diagonal form ($`c=N_s/N`$). In the ground state we have
$`n_{𝐤,\sigma }=\mathrm{\Theta }(k_{F,\sigma }k)`$ (B7)
and therefore
$`n_\sigma `$ $`=`$ $`{\displaystyle \frac{1}{N_s}}{\displaystyle \underset{i=1}{\overset{N_s}{}}}n_{i,\sigma }={\displaystyle \frac{1}{N}}{\displaystyle \underset{𝐤}{}}n_{𝐤,\sigma }`$ (B8)
$`=`$ $`{\displaystyle \frac{1}{\rho _L}}{\displaystyle \frac{k_{F,\sigma }^3}{6\pi ^2}}`$ (B9)
where $`\rho _L=N/V`$ is the lattice density. On the other hand, the total number of electrons is
$`N_e`$ $`=`$ $`{\displaystyle \underset{i=1}{\overset{N}{}}}{\displaystyle \underset{\sigma }{}}n_{i,\sigma }={\displaystyle \underset{𝐤}{}}{\displaystyle \underset{\sigma }{}}n_{𝐤,\sigma }`$ (B10)
$`=`$ $`V{\displaystyle \underset{\sigma }{}}{\displaystyle \frac{k_{F,\sigma }^3}{6\pi ^2}}`$ (B11)
and thus the number of electrons per lattice site is
$`\overline{n}={\displaystyle \frac{N_e}{N}}={\displaystyle \underset{\sigma }{}}n_\sigma `$ (B12)
moreover
$`m_z=n_{}n_{}`$ (B13)
from which we conclude
$`n_{}`$ $`=`$ $`{\displaystyle \frac{\overline{n}+m_z}{2}}`$ (B14)
$`n_{}`$ $`=`$ $`{\displaystyle \frac{\overline{n}m_z}{2}}.`$ (B15)
Since the up and down spins are in thermal equilibrium they must have the same chemical potential, that is,
$`\mu ={\displaystyle \frac{k_{F,}^2}{2m^{}}}+J_zM_zc={\displaystyle \frac{k_{F,}^2}{2m^{}}}J_zM_zc`$ (B16)
which can be rewritten with the help of (B9) as
$`(n_{})^{2/3}(n_{})^{2/3}={\displaystyle \frac{4m^{}J_zM_zc}{(6\pi ^2\rho _L)^{2/3}}}.`$ (B17)
Using (B15) we find
$`(\overline{n}+m_z)^{2/3}(\overline{n}m_z)^{2/3}={\displaystyle \frac{4m^{}J_zM_zc}{(3\pi ^2\rho _L)^{2/3}}}`$ (B18)
which has to be solved for $`m_z`$. If one is interested in the critical temperature $`T_c`$ of the problem which is much smaller than $`\mu `$ it is sufficient to consider the case of $`m_z\overline{n}`$ in which case we find
$`m_z{\displaystyle \frac{3m^{}J_zM_zc(\overline{n})^{1/3}}{(3\pi ^2\rho _L)^{2/3}}}.`$ (B19)
On the other hand, from (B1) we have
$`M_z=SB_S\left({\displaystyle \frac{\beta SJ_zcm_z}{2}}\right).`$ (B20)
The solution of the problem is given by the set (B19) and (B20). In particular, substituting (B19) into (B20)
$`M_z=SB_S\left({\displaystyle \frac{3Sm^{}\beta J_z^2c^2(\overline{n})^{1/3}}{2(3\pi ^2\rho _L)^{2/3}}}M_z\right).`$ (B21)
At $`TT_c^{}`$ we have $`M_z0`$ and therefore we find
$`T_c`$ $`=`$ $`{\displaystyle \frac{S(S+1)m^{}J_z^2c^2(\overline{n})^{1/3}}{2(3\pi ^2\rho _L)^{2/3}}}`$ (B22)
$`=`$ $`{\displaystyle \frac{S(S+1)}{4}}{\displaystyle \frac{c^2\overline{n}J_z^2}{E_F}}={\displaystyle \frac{S(S+1)}{6\rho _L}}N(0)J_z^2`$ (B23)
which agrees with (16).
## C Frequency dependent susceptibility of a two level system
Let us consider the response of a system described by the Hamiltonian (201) to an oscillating magnetic field of frequency $`\omega `$. In linear response this is given by the retarded spin-spin correlation function averaged over the thermal ensemble
$`\chi (t)`$ $`=`$ $`i\mathrm{\Theta }(t)[\delta \sigma _z(0),\delta \sigma _z(t)]=i{\displaystyle \frac{\mathrm{\Theta }(t)}{Z}}tr\left(e^{\beta H}[\delta \sigma _z,e^{iHt}\delta \sigma _ze^{iHt}]\right)`$ (C1)
$`=`$ $`i{\displaystyle \frac{\mathrm{\Theta }(t)}{Z}}{\displaystyle \underset{n,m}{}}\left(e^{\beta E_n}e^{\beta E_m}\right)e^{i(E_nE_m)t}|n|\delta \sigma _z|m|^2`$ (C2)
where $`H|n=E_n|n`$ and $`Z=_ne^{\beta E_n}`$ and $`\delta \sigma _z=\sigma _z\sigma _z`$. A simple Fourier transform gives
$`\chi (\omega )`$ $`=`$ $`{\displaystyle \frac{1}{Z}}{\displaystyle \underset{n,m}{}}\left(e^{\beta E_n}e^{\beta E_m}\right){\displaystyle \frac{|n|\delta \sigma _z|m|^2}{\omega +(E_nE_m)iϵ}}`$ (C3)
$`=`$ $`{\displaystyle \frac{1}{Z}}{\displaystyle \underset{n,m}{}}\left(e^{\beta E_n}e^{\beta E_m}\right){\displaystyle \frac{|n|\delta \sigma _z|m|^2}{E_nE_m}}\left(1{\displaystyle \frac{\omega }{\omega +(E_nE_m)iϵ}}\right)`$ (C4)
$`=`$ $`\chi (0)+{\displaystyle \frac{\omega }{Z}}{\displaystyle \underset{n,m}{}}\left(e^{\beta E_n}e^{\beta E_m}\right){\displaystyle \frac{|n|\delta \sigma _z|m|^2}{(E_nE_m)(\omega +(E_nE_m)iϵ)}}`$ (C5)
where $`ϵ0`$ and
$`\chi (0)={\displaystyle \frac{1}{Z}}{\displaystyle \underset{n,m}{}}\left(e^{\beta E_n}e^{\beta E_m}\right){\displaystyle \frac{|n|\delta \sigma _z|m|^2}{E_nE_m}}`$ (C6)
is the static susceptibility.
The calculation is very simple because we are dealing with a two-level system problem. Indeed the Hamiltonian (201) can be diagonalized by a unitary transformation
$`U`$ $`=`$ $`e^{i\theta \sigma _y}`$ (C7)
$`\mathrm{tan}(2\theta )`$ $`=`$ $`{\displaystyle \frac{\mathrm{\Delta }}{E_H}}`$ (C8)
which transforms
$`UH_CU^1`$ $`=`$ $`\sqrt{E_H^2+\mathrm{\Delta }^2}\tau _z`$ (C9)
$`U\sigma _zU^1`$ $`=`$ $`{\displaystyle \frac{1}{\sqrt{E_H^2+\mathrm{\Delta }^2}}}\left[E_H\tau _z\mathrm{\Delta }\tau _x\right]`$ (C10)
$`U\sigma _xU^1`$ $`=`$ $`{\displaystyle \frac{1}{\sqrt{E_H^2+\mathrm{\Delta }^2}}}\left[\mathrm{\Delta }\tau _z+E_H\tau _x\right].`$ (C11)
Moreover, in the new basis of the Pauli matrices $`\tau `$ with eigenstates $`|\pm `$ and eigenenergies $`\pm \sqrt{E_H^2+\mathrm{\Delta }^2}`$, respectively, we can easily show that
$`\sigma _z`$ $`=`$ $`{\displaystyle \frac{E_H}{\sqrt{E_H^2+\mathrm{\Delta }^2}}}\mathrm{tanh}\left(\beta \sqrt{E_H^2+\mathrm{\Delta }^2}\right)`$ (C12)
$`\pm |\sigma _z|\pm `$ $`=`$ $`\pm {\displaystyle \frac{E_H}{\sqrt{E_H^2+\mathrm{\Delta }^2}}}`$ (C13)
$`\pm |\sigma _z|`$ $`=`$ $`{\displaystyle \frac{\mathrm{\Delta }}{\sqrt{E_H^2+\mathrm{\Delta }^2}}}.`$ (C14)
Substituting (C14) into (C6) we easily find
$`\chi (0)={\displaystyle \frac{\beta E_H}{E_H^2+\mathrm{\Delta }^2}}sech^2\left(\beta \sqrt{E_H^2+\mathrm{\Delta }^2}\right)+{\displaystyle \frac{\mathrm{\Delta }^2}{(E_H^2+\mathrm{\Delta }^2)^{3/2}}}\mathrm{tanh}\left(\beta \sqrt{E_H^2+\mathrm{\Delta }^2}\right)`$ (C15)
which is the result (205). The dynamic part can be obtained in analogous way and it gives
$`\chi (\omega )`$ $`=`$ $`\chi (0){\displaystyle \frac{\beta E_H^2}{E_H^2+\mathrm{\Delta }^2}}sech^2\left(\beta \sqrt{E_H^2+\mathrm{\Delta }^2}\right){\displaystyle \frac{\omega }{\omega iϵ}}`$ (C16)
$``$ $`{\displaystyle \frac{\mathrm{\Delta }^2}{(E_H^2+\mathrm{\Delta }^2)^{3/2}}}\mathrm{tanh}\left(\beta \sqrt{E_H^2+\mathrm{\Delta }^2}\right)\left[{\displaystyle \frac{\omega }{\omega 2\sqrt{E_H^2+\mathrm{\Delta }^2}iϵ}}+{\displaystyle \frac{\omega }{\omega +2\sqrt{E_H^2+\mathrm{\Delta }^2}iϵ}}\right]`$ (C17)
from which we can extract the real and imaginary parts:
$`\mathrm{}\left[\chi (\omega )\right]`$ $`=`$ $`\chi (0){\displaystyle \frac{\beta E_H^2}{E_H^2+\mathrm{\Delta }^2}}sech^2\left(\beta \sqrt{E_H^2+\mathrm{\Delta }^2}\right){\displaystyle \frac{\omega ^2}{\omega ^2+ϵ^2}}`$ (C18)
$``$ $`{\displaystyle \frac{\mathrm{\Delta }^2}{2(E_H^2+\mathrm{\Delta }^2)^{3/2}}}\mathrm{tanh}\left(\beta \sqrt{E_H^2+\mathrm{\Delta }^2}\right)[{\displaystyle \frac{\omega (\omega 2\sqrt{E_H^2+\mathrm{\Delta }^2})}{(\omega 2\sqrt{E_H^2+\mathrm{\Delta }^2})^2+ϵ^2}}`$ (C19)
$`+`$ $`{\displaystyle \frac{\omega (\omega +2\sqrt{E_H^2+\mathrm{\Delta }^2})}{(\omega +2\sqrt{E_H^2+\mathrm{\Delta }^2})^2+ϵ^2}}]`$ (C20)
$`\mathrm{}\left[\chi (\omega )\right]`$ $`=`$ $`2\pi {\displaystyle \frac{\mathrm{\Delta }^2}{(E_H^2+\mathrm{\Delta }^2)}}\mathrm{tanh}\left(\beta \sqrt{E_H^2+\mathrm{\Delta }^2}\right)\left[\delta (\omega 2\sqrt{E_H^2+\mathrm{\Delta }^2})\delta (\omega +2\sqrt{E_H^2+\mathrm{\Delta }^2})\right]`$ (C21)
which are used in the calculations.
## D Two impurity calculations
Starting from (80) and using the expansion (81)the Hamiltonian is written as
$`H`$ $`=`$ $`{\displaystyle \underset{j,\alpha }{}}{\displaystyle \frac{dk}{2\pi }ϵ_k\psi _{j,\alpha }^{}(k)\psi _{j,\alpha }(k)}+{\displaystyle \underset{a}{}}\mathrm{\Gamma }_a^RS_{1,a}S_{2,a}`$ (D1)
$`+`$ $`{\displaystyle \frac{v_F}{2}}{\displaystyle \underset{a,j,l,\alpha ,\beta }{}}{\displaystyle }{\displaystyle \frac{dk}{2\pi }}{\displaystyle }{\displaystyle \frac{dk^{}}{2\pi }}[J_{+,a}(k,k^{})\psi _{j,\alpha }^{}(k)\delta _{j,l}\sigma _{\alpha ,\beta }^a\psi _{l,\beta }(k^{})(S_{1,a}+S_{2,a})`$ (D2)
$``$ $`J_{m,a}(k,k^{})\psi _{j,\alpha }^{}(k)\tau _{j,l}^z\sigma _{\alpha ,\beta }^a\psi _{l,\beta }(k^{})(S_{1,a}S_{2,a})+J_{,a}(k,k^{})\psi _{j,\alpha }^{}(k)\tau _{j,l}^x\sigma _{\alpha ,\beta }^a\psi _{l,\beta }(k^{})(S_{1,a}+S_{2,a})`$ (D3)
$``$ $`iJ_{t,a}(k,k^{})\psi _{j,\alpha }^{}(k)\tau _{j,l}^y\sigma _{\alpha ,\beta }^a\psi _{l,\beta }(k^{})(S_{1,a}S_{2,a})]`$ (D4)
where $`\tau ^a`$ are Pauli matrices which act on the subspace of $`j=1,2`$, and
$`J_{\pm ,a}`$ $`=`$ $`{\displaystyle \frac{kk^{}J_a^R}{4\pi v_F}}\left(N_e(k)N_e(k^{})\pm N_o(k)N_o(k^{})\right)`$ (D5)
$`J_{m(t),a}`$ $`=`$ $`{\displaystyle \frac{kk^{}J_a^R}{4\pi v_F}}\left(N_e(k)N_o(k^{})\pm N_o(k)N_e(k^{})\right)`$ (D6)
are the exchange constants. Observe that all the momenta here are defined in a thin shell around the Fermi surface. Thus, we can linearize the band by writing $`ϵ_k=v_F(k_Fk)`$ and expand the above exchange constants to leading order in $`k`$ in which case we get
$`J_{+,a}(k,k^{})`$ $``$ $`\pi N(0)J_a^R`$ (D7)
$`J_{,a}(k,k^{})`$ $``$ $`\pi N(0)J_a^R{\displaystyle \frac{\mathrm{sin}(k_FR)}{k_FR}}`$ (D8)
$`J_{m,a}(k,k^{})`$ $``$ $`\pi N(0)J_a^R\sqrt{1\left({\displaystyle \frac{\mathrm{sin}(k_FR)}{k_FR}}\right)^2}`$ (D9)
$`J_{t,a}(k,k^{})`$ $``$ $`0.`$ (D10)
in which case there is no momentum dependence in the coupling constants of (D4) which transforms it to an impurity problem if we define the Fourier transform:
$`\psi _{j,\alpha }(x)={\displaystyle _{\mathrm{}}^+\mathrm{}}{\displaystyle \frac{dk}{2\pi }}e^{ikx}\psi _{j,\alpha }(k)`$ (D11)
in which case (D4) becomes turns into (85).
An important term we have not included is due to ordinary scattering of electrons at the impurity. This term is just the electrostatic coupling which is given by
$`H_V`$ $`=`$ $`{\displaystyle \underset{\alpha }{}}\left(V(𝐑/2)\psi _\alpha ^{}(𝐑/2)\psi _\alpha (𝐑/2)+V(𝐑/2)\psi _\alpha ^{}(𝐑/2)\psi _\alpha (𝐑/2)\right)`$ (D12)
$`=`$ $`{\displaystyle \underset{\alpha }{}}{\displaystyle \underset{𝐤,𝐤^{}}{}}\left(V(𝐑/2)e^{i(𝐤𝐤^{})𝐑/2}+V(𝐑/2)e^{i(𝐤𝐤^{})𝐑/2}\right)c_{𝐤,\alpha }^{}c_{𝐤^{},\alpha }`$ (D13)
$`=`$ $`{\displaystyle \underset{\alpha }{}}{\displaystyle }{\displaystyle \frac{dk}{2\pi }}{\displaystyle }{\displaystyle \frac{dk^{}}{2\pi }}[V_e(k,k^{})\psi _{1,\alpha }^{}(k)\psi _{1,\alpha }(k^{})+V_o(k,k^{})\psi _{2,\alpha }^{}(k)\psi _{2,\alpha }(k^{})`$ (D14)
$`+`$ $`V_{eo}(k,k^{})(\psi _{1,\alpha }^{}(k)\psi _{2,\alpha }(k^{})+\psi _{2,\alpha }^{}(k)\psi _{1,\alpha }(k^{}))]`$ (D15)
where we used (81). As previously, assuming that the electron momenta is close to the Fermi surface we can rewrite the above term as
$`H_V`$ $`=`$ $`{\displaystyle \underset{\alpha }{}}[V_1\psi _{1,\alpha }^{}(x=0)\psi _{1,\alpha }(x=0)+V_2\psi _{1,\alpha }^{}(x=0)\psi _{1,\alpha }(x=0)`$ (D16)
$`+`$ $`V(\psi _{1,\alpha }^{}(x=0)\psi _{2,\alpha }(x=0)+\psi _{2,\alpha }^{}(x=0)\psi _{1,\alpha }(x=0))].`$ (D17)
The bosonization procedure goes like the one described for the single impurity problem in Subsection IV A. The only difference is that we are going to construct the factors $`K_\sigma `$ explicitly in terms of the fermion fields. We define
$`K_{1,\sigma }`$ $`=`$ $`\mathrm{exp}\left\{i\pi {\displaystyle 𝑑x\psi _{1,}^{}\psi _{1,}}\right\}`$ (D18)
$`K_{2,\sigma }`$ $`=`$ $`\mathrm{exp}\left\{i\pi {\displaystyle 𝑑x\left[\underset{\sigma }{}\psi _{1,\sigma }^{}\psi _{1,\sigma }\right]}+\psi _{2,}^{}\psi _{2,}\right\}`$ (D19)
which can be rewritten in terms of the bosonic fields if we use that
$`\psi _{i,\sigma }^{}\psi _{i,\sigma }={\displaystyle \frac{1}{2\pi }}{\displaystyle \frac{\mathrm{\Phi }_{i,\sigma }}{x}}.`$ (D20)
It is easy to show that (85) reads (for simplicity we work again with the problem of uniaxial symmetry)
$`H`$ $`=`$ $`{\displaystyle \frac{v_F}{2}}{\displaystyle \underset{i=c,s,f,sf}{}}{\displaystyle 𝑑x\left[\mathrm{\Pi }_i^2(x)+\left(\frac{\varphi _i}{x}\right)^2\right]}+{\displaystyle \underset{a}{}}\mathrm{\Gamma }_aS_{1,a}S_{2,a}+{\displaystyle \frac{2V}{\pi a}}\mathrm{cos}(\mathrm{\Phi }_{sf}(0))\mathrm{cos}(\mathrm{\Phi }_f(0)\theta )`$ (D21)
$`+`$ $`{\displaystyle \frac{v_FJ_{+,z}}{2\pi }}{\displaystyle \frac{\mathrm{\Phi }_s(0)}{x}}(S_{1,z}+S_{2,z})`$ (D22)
$`+`$ $`{\displaystyle \frac{v_FJ_{+,}}{\pi a}}\mathrm{cos}(\mathrm{\Phi }_{sf}(0))\left(\mathrm{cos}(\mathrm{\Phi }_s(0))(S_{1,x}+S_{2,x})\mathrm{sin}(\mathrm{\Phi }_s(0))(S_{1,y}+S_{2,y})\right)`$ (D23)
$`+`$ $`{\displaystyle \frac{v_FJ_{m,z}}{2\pi }}{\displaystyle \frac{\mathrm{\Phi }_{sf}(0)}{x}}(S_{1,z}S_{2,z})`$ (D24)
$``$ $`{\displaystyle \frac{v_FJ_{m,}}{\pi a}}\mathrm{sin}(\mathrm{\Phi }_{sf}(0))\left(\mathrm{sin}(\mathrm{\Phi }_s(0))(S_{1,x}S_{2,x})+\mathrm{cos}(\mathrm{\Phi }_s(0))(S_{1,y}S_{2,y})\right)`$ (D25)
$`+`$ $`{\displaystyle \frac{v_FJ_{,z}}{\pi a}}\mathrm{sin}(\mathrm{\Phi }_{sf}(0))\mathrm{sin}(\mathrm{\Phi }_f(0)\theta )(S_{1,z}+S_{2,z})`$ (D26)
$`+`$ $`{\displaystyle \frac{v_FJ_,}{\pi a}}\mathrm{cos}(\mathrm{\Phi }_f(0)\theta )\left(\mathrm{cos}(\mathrm{\Phi }_s(0))(S_{1,x}+S_{2,x})\mathrm{sin}(\mathrm{\Phi }_s(0))(S_{1,y}+S_{2,y})\right)`$ (D27)
which can be rewritten in a more economical format:
$`H`$ $`=`$ $`{\displaystyle \frac{v_F}{2}}{\displaystyle \underset{i=c,s,f,sf}{}}{\displaystyle 𝑑x\left[\mathrm{\Pi }_i^2(x)+\left(\frac{\varphi _i}{x}\right)^2\right]}+{\displaystyle \underset{a}{}}\mathrm{\Gamma }_aS_{1,a}S_{2,a}`$ (D28)
$`+`$ $`{\displaystyle \frac{2V}{\pi a}}\mathrm{cos}(\mathrm{\Phi }_{sf}(0))\mathrm{cos}(\mathrm{\Phi }_f(0)\theta )`$ (D29)
$`+`$ $`{\displaystyle \frac{v_FJ_{+,z}}{2\pi }}{\displaystyle \frac{\mathrm{\Phi }_s(0)}{x}}(S_{1,z}+S_{2,z})`$ (D30)
$`+`$ $`{\displaystyle \frac{v_FJ_{+,}}{2\pi a}}\mathrm{cos}(\mathrm{\Phi }_{sf}(0))\left(e^{i\mathrm{\Phi }_s(0)}(S_{1,+}+S_{2,+})+e^{i\mathrm{\Phi }_s(0)}(S_{1,}+S_{2,})\right)`$ (D31)
$`+`$ $`{\displaystyle \frac{v_FJ_{m,z}}{2\pi }}{\displaystyle \frac{\mathrm{\Phi }_{sf}(0)}{x}}(S_{1,z}S_{2,z})`$ (D32)
$`+`$ $`i{\displaystyle \frac{v_FJ_{m,}}{2\pi a}}\mathrm{sin}(\mathrm{\Phi }_{sf}(0))\left(e^{i\mathrm{\Phi }_s(0)}(S_{1,+}+S_{2,+})e^{i\mathrm{\Phi }_s(0)}(S_{1,}+S_{2,})\right)`$ (D33)
$`+`$ $`{\displaystyle \frac{v_FJ_{,z}}{\pi a}}\mathrm{sin}(\mathrm{\Phi }_f(0)\theta )\mathrm{sin}(\mathrm{\Phi }_{sf}(0))(S_{1,z}+S_{2,z})`$ (D34)
$`+`$ $`{\displaystyle \frac{v_FJ_,}{2\pi a}}\mathrm{cos}(\mathrm{\Phi }_f(0)\theta )\left(e^{i\mathrm{\Phi }_s(0)}(S_{1,+}+S_{2,+})+e^{i\mathrm{\Phi }_s(0)}(S_{1,}+S_{2,})\right)`$ (D35)
where $`S_{i,\pm }=S_{i,x}\pm iS_{i,y}`$ and $`\theta `$ is defined in (99). Notice that the charge degrees of freedom have decoupled entirely and can be disregarded and that the diagonal terms in (D17) can be written as gradients of the boson operators and therefore can be trivially absorbed into a shift of the bosons and do not appear here. After the unitary transformation: $`U=e^{i(S_{1,z}+S_{2,z})\mathrm{\Phi }_s(0)}`$ the Hamiltonian becomes (97).
|
warning/0003/quant-ph0003111.html
|
ar5iv
|
text
|
# Quantum communication between atomic ensembles using coherent light
## Abstract
Protocols for quantum communication between massive particles, such as atoms, are usually based on transmitting nonclassical light, and/or super-high finesse optical cavities are normally needed to enhance interaction between atoms and photons. We demonstrate a surprising result: an unknown quantum state can be teleported from one free-space atomic ensemble to the other by transmitting only coherent light. No non-classical light and no cavities are needed in the scheme, which greatly simplifies its experimental implementation.
The goal of quantum communication is to transmit an unknown quantum state from one particle to another one at a distant location. This can be obtained either by direct transmission of the state , or by disembodied transport, i.e., quantum teleportation . Quantum teleportation of an unknown state from a photon to a photon , or from a single-mode light to another single-mode beam of light has been demonstrated experimentally. A desired goal is to obtain quantum teleportation of the state of massive particles, since the massive particles are ideal for storage of quantum information, and they play an important role in local quantum information processing, such as quantum computation. At the same time, the information should be transferred from one location to another via optical states, since light is the best long distance carrier of information. There have been several proposals for quantum teleportation of atomic motional or internal states . by transmitting single-photon or non-classical light . Most of these proposals are based on the assumption that atoms are trapped inside high-Q optical cavities, which is difficult to achieve experimentally . The recent proposal eliminates this requirement, however it still requires an external source of entanglement (non-classical light). Here, we propose and analyze a quantum communication scheme, which teleports an unknown collective internal state from one free-space atomic ensemble to another only using coherent light. This result is indeed surprising, since strong coherent light (light from an ordinary laser) is usually thought to be ‘purely classical’, but via it unknown quantum states of free-space atomic ensembles can nonetheless be teleported from one location to another!
The system we are considering is a cloud of identical atoms with the relevant level structure shown in Fig. 1. Each atom has two degenerate ground states and two degenerate excited states. The transitions $`|1|3`$ and $`|2|4`$ are coupled with a large detuning $`\mathrm{\Delta }`$ to propagating light fields with different circular polarizations according to the angular-momentum selection rules. This kind of interaction has been analyzed semiclassically in , and recently shown to be applicable for quantum non-demolition measurements and teleportation with non-classical light , with an adiabatic Hamiltonian and neglecting the noise due to spontaneous emission. Our goal here is twofold: first, we show based on this Hamiltonian, entanglement can be generated and furthermore quantum communication can be achieved between distant atomic ensembles using only coherent light; and second, we deduce this Hamiltonian through a full quantum description of the interaction of the atomic ensemble with free-space propagating light, taking into account the noise. The latter is an essential result since we make use of the quantum nature of both light and atoms in quantum communication, and it not clear from the outset that the noise can indeed be neglected during the interaction process.
Entanglement generation is basic to quantum communication. We create entanglement between two atomic ensembles through a nonlocal Bell measurement with the schematic setup shown by Fig. 2. The atomic ensemble is assumed to be of a pencil shape with Fresnel number $`F=A/\lambda _0L=1`$, where $`A`$ and $`L`$ are the cross section and the length of the ensemble, respectively, and $`\lambda _0`$ is the optical wave length. In this case, it is justified to use a one-dimensional theory to describe the propagating light field . The input laser pulse is linearly polarized and expressed as $`E^{(+)}(z,t)=\sqrt{\frac{\mathrm{}\omega _0}{4\pi ϵ_0A}}\underset{i=1,2}{}a_i(z,t)e^{i\left(k_0z\omega _0t\right)}`$, where $`\omega _0=k_0c=2\pi c/\lambda _0`$ is the carrier frequency, and $`i`$ denotes two orthogonal circular polarizations, with the standard commutation relations $`[a_i(z,t),a_j(z^{},t)]=\delta _{ij}\delta \left(zz^{}\right)`$. The light is weakly focused with cross section $`A`$ to match the atomic ensemble. For a strong coherent input with linear polarization, the initial condition is expressed as $`a_i(0,t)=\alpha _t,`$ with the total photon number over the pulse duration $`T`$ satisfies $`2N_p=2c_0^T\left|\alpha _t\right|^2𝑑t1`$. The Stokes operators are introduced for the free-space input and output light (light before entering or after leaving the atomic ensemble) by $`S_x^p=\frac{c}{2}_0^T\left(a_1^{}a_2+a_2^{}a_1\right)𝑑\tau ,`$ $`S_y^p=\frac{c}{2i}_0^L\left(a_1^{}a_2a_2^{}a_1\right)𝑑\tau ,`$ $`S_z^p=\frac{c}{2}_0^L\left(a_1^{}a_1a_2^{}a_2\right)𝑑\tau .`$ In free space, $`a_i(z,t)`$ only depends on $`\tau =tz/c`$, and then the Stokes operators satisfy the spin commutation relations $`[S_y^p,S_z^p]=iS_x^p`$. For our coherent input, we have $`S_x^p=N_p`$ and $`S_y^p=`$ $`S_z^p=0`$. With a very large $`N_p`$, the off-resonant interaction with atoms is only a small perturbation to $`S_x^p`$, and we can treat $`S_x^p`$ classically by replacing it with its mean value $`S_x^p`$. Then, we define two canonical observables for light by $`X^p=S_y^p/\sqrt{S_x^p},`$ $`P^p=S_z^p/\sqrt{S_x^p}`$ with a standard commutator $`[X^p,P^p]=i`$. These operators, initially in a vacuum state, are the quantum variables we are interested in. Similar operators can be introduced for atoms. For an atomic ensemble with many atoms, it is convenient to define the continuous atomic operators $`\sigma _{\mu \nu }(z,t)=lim_{\delta z0}\frac{1}{\rho A\delta z}_i^{zz_i<z+\delta z}|\mu _i\nu |`$ ($`\mu ,\nu =1,2,3,4)`$ with the commutation relations $`[\sigma _{\mu \nu }(z,t),\sigma _{\nu ^{}\mu ^{}}(z^{},t)]=\left(1/\rho A\right)\delta \left(zz^{}\right)\left(\delta _{\nu \nu ^{}}\sigma _{\mu \mu ^{}}\delta _{\mu \mu ^{}}\sigma _{\nu ^{}\nu }\right)`$. In the definition, $`z_i`$ is the position of the $`i`$ atom, and $`\rho `$ is the number density of the atomic ensemble with the total atom number $`2N_a=\rho AL1`$. The collective spin operators are introduced for the ground states of the atomic ensemble by $`S_x^a=\frac{\rho A}{2}_0^L\left(\sigma _{12}+\sigma _{12}^{}\right)𝑑z,`$ $`S_y^a=\frac{\rho A}{2i}_0^L\left(\sigma _{12}\sigma _{12}^{}\right)𝑑z,`$ $`S_z^a=\frac{\rho A}{2}_0^L\left(\sigma _{11}\sigma _{22}\right)𝑑z`$. All the atoms are initially prepared in the superposition of the two ground states $`\left(|1+|2\right)/\sqrt{2}`$ (this can be obtained with negligible noise by applying classical laser pulses with detuning $`\mathrm{\Delta }\gamma `$), which is an eigenstate of $`S_x^a`$ with a very large eigenvalue $`N_a`$. Similarly, we treat $`S_x^a`$ classically, and define the canonical operators for atoms by $`X^a=S_y^a/\sqrt{S_x^a},`$ $`P^a=S_z^a/\sqrt{S_x^a}`$ with $`[X^a,P^a]=i`$ and an initial vacuum state. As shown below, after the laser pulse passes through the atomic ensemble, the off-resonant interaction changes the canonical operators according to
$`X^p`$ $`=`$ $`\sqrt{1\epsilon _p}\left(X^p\kappa P^a\right)+\sqrt{\epsilon _p}X_s^p,`$ (1)
$`X^a`$ $`=`$ $`\sqrt{1\epsilon _a}\left(X^a\kappa P^p\right)+\sqrt{\epsilon _a}X_s^a,`$ ()
$`P^\beta `$ $`=`$ $`\sqrt{1\epsilon _\beta }P^\beta +\sqrt{\epsilon _\beta }P_s^\beta ,\text{ }\left(\beta =a,p\right),`$ (2)
where the symbols with (without) a prime denote the operators after (before) the interaction, and $`X_s^a,P_s^a`$ and $`X_s^p,P_s^p`$ are the standard vacuum noise operators with variance $`1/2.`$ The interaction and damping coefficients $`\kappa ,\epsilon _p,\epsilon _a`$ are given respectively by $`\kappa =\frac{2\sqrt{N_pN_a}\left|g\right|^2}{\mathrm{\Delta }c},`$ $`\epsilon _p=\frac{N_a\left|g\right|^2\gamma }{\mathrm{\Delta }^2c},`$ $`\epsilon _a=\frac{N_p\left|g\right|^2\gamma ^{}}{\mathrm{\Delta }^2c}`$, where $`g`$ is the coupling constant and $`\gamma ,\gamma ^{}`$ are spontaneous emission rates (see Fig. 1). Equation (1) is obtained under the conditions $`\epsilon _{p,a}1`$ and $`\kappa \sqrt{N_{p,a}}.`$ For our application, we would like to have $`\kappa 1`$. This is possible if we choose $`N_pN_a1`$ and $`\mathrm{\Delta }\gamma `$. The number matching condition $`N_pN_a`$ is an important requirement obtained here to minimize the noise effect, since we have $`\kappa =2\sqrt{\epsilon _p\epsilon _a}\mathrm{\Delta }/\sqrt{\gamma \gamma ^{}}`$ and the best choice is $`\epsilon _p\epsilon _a`$ to increase the signal-to-noise ratio.
Before we proceed to demonstrating Eq. (1), first we show that this transformation allows us to generate entanglement, and to achieve quantum communication between atomic ensembles using only coherent light. Entanglement is generated through a nonlocal Bell measurement of the EPR operators $`X_1^aX_2^a`$ and $`P_1^a+P_2^a`$ with the setup depicted by Fig. 2. This setup measures the Stokes operator $`X_2^p`$ of the output light. Using Eq. (1) and neglecting the small loss terms, we have $`X_2^p=X_1^p+\kappa \left(P_1^a+P_2^a\right)`$, so we get a collective measurement of $`P_1^a+P_2^a`$ with some inherent vacuum noise $`X_1^p`$. The efficiency $`1\eta `$ of this measurement is determined by the parameter $`\kappa `$ with $`\eta =1/\left(1+2\kappa ^2\right)`$. After this round of measurements, we rotate the collective atomic spins around the $`x`$ axis to get the transformations $`X_1^aP_1^a,`$ $`P_1^aX_2^a`$ and $`X_2^aP_2^a,`$ $`P_2^aX_2^a`$. The rotation of the atomic spin can be easily obtained with negligible noise by applying classical laser pulses with detuning $`\mathrm{\Delta }\gamma `$. After the rotation, the measured observable of the first round of measurement is changed to $`X_1^aX_2^a`$ in the new variables. We then make another round of collective measurement of the new variable $`P_1^a+P_2^a`$. In this way, both the EPR operators $`X_1^aX_2^a`$ and $`P_1^a+P_2^a`$ are measured, and the final state of the two atomic ensembles is collapsed into a two-mode squeezed state with variance $`\delta \left(X_1^aX_2^a\right)^2=\delta \left(P_1^a+P_2^a\right)^2=e^{2r}`$, where the squeezing parameter $`r`$ is given by
$$r=\frac{1}{2}\mathrm{ln}\left(1+2\kappa ^2\right).$$
(3)
Thus, using only coherent light, we generate continuous variable entanglement between two nonlocal atomic ensembles. With the interaction parameter $`\kappa 5`$, a high squeezing (and thus a large entanglement) $`r2.0`$ is obtainable. Note that entanglement generation is the key step for many quantum protocols, and is the basis of quantum communication, quantum cryptography, and tests of Bell inequality. In the following, we show as an example how to achieve indirect quantum communication, i.e., quantum teleportation, between distant atomic ensembles using only coherent light.
We consider unconditional quantum teleportation of continuous variables from one atomic ensemble to the other since we have continuous variable entanglement. To achieve quantum teleportation, first two distant atomic samples 1 and 2 are prepared in a continuously entangled state using the nonlocal Bell measurement described above. Then, a Bell measurement with the same setup as shown by Fig. 2 on the two local samples 1 and 3, together with a straightforward displacement of $`X_3^a,`$ $`P_3^a`$ on the sample 3, will teleport an unknown collective spin state from the atomic sample 3 to 2. The teleported state on the sample 2 has the same form as that in the original proposal of continuous variable teleportation using squeezing light , with the squeezing parameter $`r`$ replaced by Eq. (2) and with an inherent Bell detection inefficiency $`\eta =1/\left(1+2\kappa ^2\right)`$. The teleportation quality is best described by the fidelity, which, for a pure input state, is defined as the overlap of the teleported state and the input state. For any coherent input state of the sample 3, the teleportation fidelity is given by
$$F=1/\left(1+\frac{1}{1+2\kappa ^2}+\frac{1}{2\kappa ^2}\right).$$
(3)
Equation (3) shows, if there is no extra noise, a high fidelity $`F96\%`$ would be possible for the teleportation of the collective atomic spin state with the interaction parameter $`\kappa 5`$.
Next we will include noise and derive expressions for the squeezing and the fidelity under realistic experimental conditions. Before we analyze the effects of noise, let us first demonstrate Eq. (1) with a full quantum approach. The demonstration of Eq. (1) including the spontaneous emission noise is necessary in the following context: First, it is not clear that the spontaneous emission is indeed negligible through a simple estimation of the noise, since during the interaction approximately $`\frac{N_pN_a\left|g\right|^2\gamma }{\mathrm{\Delta }^2c}`$ atoms in the atomic ensemble (normally much large than $`1`$) will be subjected to quantum jumps caused by the spontaneous emission . We need to show that quantum jumps of individual atoms have negligible influence on the collective spin operators which are the quantities of interest. Second, the maximally allowable interaction parameter $`\kappa `$ is mainly limited by the noise. We need a balance between the desired interaction and the noise to maximize the squeezing and the teleportation fidelity. Third, some subtle experimental requirements, such as the number matching condition $`N_pN_a`$, is only obtainable by considering the noise.
With introduction of the continuous atomic operators, the interaction between atoms and the propagating light $`E^{(+)}(z,t)`$ is described by the following Hamiltonian (in the rotating frame),
$$H=\mathrm{}\underset{i=1,2}{}_0^L[\mathrm{\Delta }\sigma _{i+2,i+2}(z,t)+(ge^{ik_0z}a_i(z,t)\sigma _{i,i+2}(z,t)+h.c)]\rho Adz,$$
(4)
where the coupling constant $`g=\sqrt{\frac{\omega _0}{4\pi \mathrm{}ϵ_0A}}d`$ and $`d`$ is the dipole moment of the $`|i|i+2`$ transition. Corresponding to this Hamiltonian, the Maxwell-Bloch equations are written as as
$`\left({\displaystyle \frac{}{t}}+c{\displaystyle \frac{}{z}}\right)a_i(z,t)=ig^{}e^{ik_0z}\rho A\sigma _{i+2,i}(z,t),`$ (5)
$`{\displaystyle \frac{}{t}}\sigma _{\mu \nu }={\displaystyle \frac{i}{\mathrm{}}}[\sigma _{\mu \nu },H]{\displaystyle \frac{\gamma _{\mu \nu }}{2}}\sigma _{\mu \nu }+\sqrt{\gamma _{\mu \nu }}\left(\sigma _{\nu \nu }\sigma _{\mu \mu }\right)F_{\mu \nu }\text{ }\left(\mu <\nu \right),`$ ()
where the spontaneous emission rates (see Fig. 1) are $`\gamma _{13}=\gamma _{24}\gamma =\frac{\omega _0^3\left|d\right|^2}{3\pi ϵ_0\mathrm{}c^3},`$ $`\gamma _{14}=\gamma _{23}\gamma ^{},`$ and $`\gamma _{12}=0`$, respectively. Assuming that the spontaneous emission is independent for different atoms (because the distance between atoms is larger than optical wave length), the vacuum noise operators $`F_{\mu \nu }`$ satisfy the $`\delta `$-commutation relations $`[F_{\mu \nu }(z,t),F_{\mu ^{}\nu ^{}}^{}(z^{},t^{})]=\left(1/\rho A\right)\delta _{\mu \mu ^{}}\delta _{\nu \nu ^{}}\delta \left(zz^{}\right)\delta \left(tt^{}\right)`$. To simplify Eq. (5), first we change the variables by $`\tau =tz/c`$, and then adiabatically eliminate the excited states $`|3`$ and $`|4`$ of atoms in the case of a large detuning, i.e., $`\mathrm{\Delta }ga_i(z,t)g\sqrt{N_p/\left(cT\right)}`$. The resultant equations read
$`{\displaystyle \frac{}{z}}a_i(z,\tau )={\displaystyle \frac{i\left|g\right|^2\rho A\sigma _{ii}}{\mathrm{\Delta }c}}a_i(z,\tau ){\displaystyle \frac{\left|g\right|^2\rho A\gamma \sigma _{ii}}{2\mathrm{\Delta }^2c}}a_i(z,\tau )+{\displaystyle \frac{g^{}e^{ik_0z}\rho A\sqrt{\gamma }\sigma _{ii}}{\mathrm{\Delta }c}}F_{i,i+2}(z,\tau ),`$ (6)
$`{\displaystyle \frac{}{\tau }}\sigma _{12}={\displaystyle \frac{i\left|g\right|^2\left(a_2^{}a_2a_1^{}a_1\right)}{\mathrm{\Delta }}}\sigma _{12}{\displaystyle \frac{\left|g\right|^2\gamma ^{}\left(a_2^{}a_2+a_1^{}a_1\right)}{2\mathrm{\Delta }^2}}\sigma _{12}+{\displaystyle \frac{\sqrt{\gamma ^{}}}{\mathrm{\Delta }}}\left(g^{}e^{ik_0z}a_2^{}\sigma _{11}F_{14}+ge^{ik_0z}a_1\sigma _{22}F_{23}^{}\right).`$ ()
The physical meaning of the above equation is quite clear: The first term at the right hand side is the phase shift caused by the off-resonant interaction between light and atoms, and the second and the third terms represent the damping and the corresponding vacuum noise caused by the spontaneous emission, respectively. In Eq. (3), the $`\sigma _{ii}`$ and $`a_i^{}a_i`$ are approximately constant operators, only with a small damping caused by the spontaneous emission. To consider the spontaneous emission noise to the first order, it is reasonable to assume constant $`\sigma _{ii}`$ and $`a_i^{}a_i`$ for Eq. (6). Then, this equation can be easily solved by integrating over $`z,\tau `$ on both sides. In this way we obtain Eq. (1) with the introduced canonical operators. The vacuum noise operators in Eq. (1) are defined from the integration of $`F_{\mu \nu }(z,\tau )`$, $`X_s^p=\sqrt{\frac{c}{4N_pN_a\left|g\right|^2}}_0^T_0^L\rho A[ig^{}e^{ik_0z}(a_2^{}\sigma _{11}F_{13}a_1^{}\sigma _{22}F_{24})+h.c.]dzd\tau `$ for instance. It should be noted that the damping term cannot be directly neglected in Eq. (6) compared with the phase shift term, even when $`\mathrm{\Delta }\gamma `$, since $`a_2^{}a_2+a_1^{}a_1a_2^{}a_2a_1^{}a_1`$. What is remarkable is that due to the collective effect, the phase shift term obtains another large prefactor $`\sqrt{N_{p,a}}`$ when we perform the integration in Eq. (6), which makes this contribution well exceed the noise term.
In the derivation above, we have neglected motion of the atoms. The atomic motion introduces two effects: the Doppler broading, and decoherence of the ground states caused by the atomic collisions. Doppler broading is negligible here, since it is suppressed significantly for off-resonant interactions with the collinear input and output light. On the other hand, the ground state coherence time ($`1`$ms$`1`$s) is much larger than the interaction time scale considered here ($`1`$ns$`1\mu `$s) under realistic experimental conditions, both for a cold trapped atomic ensemble and for a room-temperature atomic cell with a buffer gas , so that this kind of decoherence can be safely neglected. It is helpful to give an estimation of the relevant parameters for typical experiments. The interaction parameter $`\kappa `$ can be rewritten as $`\kappa =\left(3\rho \lambda _0^2L\gamma \right)/\left(8\pi ^2\mathrm{\Delta }\right)`$ with $`N_p=N_a`$. For a atomic sample of density $`\rho 5\times 10^{12}`$cm<sup>-3</sup> and of length $`L2`$cm, $`\kappa 5`$ is obtainable with the choice $`\mathrm{\Delta }300\gamma `$, and at the same time the loss $`\epsilon _p\epsilon _a<1\%`$.
As our last point, let us return to the analysis of the influence of some important noise terms on the teleportation fidelity. The noise includes the spontaneous emission noise described by Eq. (1), the detector inefficiency, and the transmission loss of the light from the first sample to the second sample. The spontaneous emission noise can be included partly in the transmission loss and partly in the detector efficiency, so we do not analyze it separately. The effect of the detector inefficiency $`\eta _d`$ is to replace $`\kappa ^2`$ in Eqs. (2) and (3) with $`\kappa ^2\left(1\eta _d\right)`$, and the teleportation fidelity is decreased by a term $`\eta _d/\kappa ^2`$, which is very small and can be safely ignored. The most important noise comes from the transmission loss. The transmission loss is described by $`X_2^p=\sqrt{1\eta _t}X_1^p^{}+\sqrt{\eta _t}X_s^t`$ (see Fig. 2), where $`\eta _t`$ is the loss rate and $`X_s^t`$ is the standard vacuum noise. The transmission loss changes the measured observables to be $`\sqrt{1\eta _t}X_1^aX_2^a`$ and $`\sqrt{1\eta _t}P_1^a+P_2^a`$. These two observables do not commute, and the two rounds of measurements influence each other. To minimize the influence on the teleportation fidelity, we choose the following configuration (for simplicity, we assume we have the same loss rate $`\eta _t`$ from the sample 1 to 2 and from 1 to 3): In the nonlocal Bell measurements on the samples 1 and 2 (the entanglement generation process), we choose a suitable interaction coefficient $`\kappa _2`$ (where its optimal value will be determined below) for the second round measurement, whereas $`\kappa _1`$ for the first round of measurement is large with $`\kappa _1^2\kappa _2^2`$ (the interaction coefficient can be easily adjusted, for instance, by changing the detuning). In the local Bell measurement, we choose the same $`\kappa _2`$ for the first round of measurement and the large $`\kappa _1`$ for the second round of measurement. For a coherent input state of the sample 3, the teleported state on the sample 2 is still Gaussian, and the teleportation fidelity $`F^{}`$ is found to be
$$F^{}2/\left(2+\frac{1}{\kappa _2^2}+\kappa _2^2\eta _t\right)1/\left(1+\sqrt{\eta _t}\right),$$
(7)
which is still independent of the coherent input state with suitable gain for the displacements . The optimal value for $`\kappa _2`$ is thus given by $`\kappa _2=1/\sqrt[4]{\eta _t}`$. Even with a notable transmission loss rate $`\eta _t0.2`$, quantum teleportation with a remarkable high fidelity $`F0.7`$ is still achievable. It is known that for coherent inputs a fidelity exceeding $`1/2`$ has ensured quantum teleportation .
In summary, we have shown that quantum communication between free space atomic ensembles can be achieved using only coherent laser beams. Quantum teleportation of the atomic spin state is observable even in the presence of significant noise. This result, together with the much simplified experimental setup proposed here, suggests that efficient quantum communication between atomic samples is within reach of present experimental conditions. ESP acknowledges fruitful discussions with A. Kuzmich. LMD acknowledges discussions with A. Sorensen. This work was supported by the Austrian Science Foundation, by the European TMR network Quantum Information, and by the Institute for Quantum Information.
|
warning/0003/hep-th0003037.html
|
ar5iv
|
text
|
# Flux Stabilization of D-branes
## 1 Introduction
String compactifications on group manifolds have long been of interest. Their world-sheet theories are the exactly solvable Wess-Zumino-Witten (WZW) conformal field theories and thus stringy effects can be understood in detail. The case of primary interest in superstring theory is the $`SU(2)`$ group manifold, because the near-horizon geometry of $`k`$ coincident NS fivebranes is a direct product including $`S^3`$, and the corresponding CFT is a product of supersymmetric $`SU(2)`$ level $`k`$, a “Feigin-Fuchs superfield”, and six free superfields . Another exact supersymmetric string background is $`S^3\times AdS_3`$ – this corresponds to the CFT of two supersymmetric WZW models, one for the $`SU(2)`$ and one for the $`SL(2,R)`$ group manifold, and it describes the near-horizon geometry of intersecting branes .
D-branes in group manifolds have been studied in a series of papers and the basic story, at least for the compact case, is fairly well understood. The natural boundary conditions (those for which the gluing can be expressed in terms of an automorphism $`\omega `$ of the current algebra) can be classified purely in CFT terms. In the case of trivial gluing, $`\omega =1`$, Cardy’s general theory , puts them in one-to-one correspondence with primary fields. <sup>1</sup><sup>1</sup>1Cardy’s theory can be generalized to non-trivial $`\omega `$; in this case, one obtains a correspondence of boundary conditions with primary fields in twisted sectors of appropriate orbifold theories, . The results turn out to be geometrical : an allowed boundary condition corresponds to a D-brane wrapped on an allowed (twisted) conjugacy class of the group. The only sign of the underlying CFT is a quantization condition on the allowed (twisted) conjugacy classes. For example, in the $`SU(2)`$ level $`k`$ model D-branes can wrap on $`k1`$ distinct $`S^2`$’s around any point (these are subject to a $`_2`$ identification). There is also a D$`0`$-brane, to complete the spectrum (there is no D$`3`$ because $`H0`$ ).
One then can study the world-volume theories of these branes by classifying massless modes and computing interactions. The results show an intriguing parallel with the noncommutative torus in that the algebra of open string primary fields (for large $`k`$ but finite conjugacy class) is the algebra of the “fuzzy sphere,” the natural quantization of the sphere .
However, there is a more elementary question one might ask first. Why is it that branes wrapped about spheres which are not minimal volume surfaces are stable at all ?
It is easiest to check that the other boundary conditions lead to stable branes by considering the parallel with the boundary state describing a D$`2`$-brane ending on an NS fivebrane. This boundary state is a tensor product of “D$`0`$” in the WZW sector with “Neumann” in the linear dilaton sector and in two of the Minkowski dimensions. It preserves half of the supersymmetry of the fivebrane theory. If $`k>1`$, the WZW component of the boundary state can be replaced with a different WZW boundary condition, leaving everything else unchanged. In particular the supersymmetry of the object is unchanged, so it is stable. The geometrical interpretation of these objects is “conical” D$`4`$-branes again wrapping a non-minimal $`S^2`$. Thus our question is appropriate.
It seems clear that this stability is linked with the origin of the quantization condition, and there are two ways one might try to explain this. One might imagine that the integrated NS two-form potential $`\widehat{B}`$ takes quantized values on a D-brane world-sheet. This would imply an implausible non-local constraint on the allowed embeddings of the D-brane, with no known origin in string theory.
A better idea is that it follows from the usual quantization of $`U(1)`$ gauge field strength, relevant because the D-brane is wrapped on $`S^2`$. Indeed there is a well known mechanism of “flux stabilization” (which has been invoked in large extra dimension scenarios, for example) which could then explain the brane’s stability. It is simply that the energy $`F^2`$ of a constant flux will be inversely proportional to the volume, and thus the total energy with the brane tension will have a minimum at non-vanishing volume.
This argument is not really correct, because, unlike what happens in the Maxwell energy, the Born-Infeld energy of a constant flux stays finite as the brane shrinks to zero volume (this is why there are no stable spherical D-branes in flat spacetime). What enters in the D-brane energy, on the other hand, is the flux of the gauge-invariant combination $`=B+2\pi \alpha ^{}F`$. In a varying external $`B`$ field the total energy including the brane tension can indeed have a minimum at nonzero volume, as we will show.
Our main result is to show that this explanation works not only qualitatively but quantitatively: indeed, up to the well known one loop shift $`kk+2`$ which renormalizes the radius of $`S^3`$, computations starting from the Born-Infeld action precisely reproduce the masses, multiplicities, and even the spectrum of small fluctuations of these branes, as calculated in CFT. The exact agreement implies that higher-order corrections to the Born-Infeld theory must vanish – this could be related to the BPS property of the corresponding objects in the fivebrane or $`S^3\times AdS_3`$ geometries.
We consider the above results convincing evidence for the advocated explanation of stability. They do in particular confirm the fact that it is the $`U(1)`$ flux $`F`$ (rather than the flux of $``$) that must be quantized. This leads, however, to an apparent paradox: the RR charges of the branes are not quantized. We will explain why the $`F`$-flux quantization is correct, and discuss possible resolutions of the paradox in the final section.
## 2 Semiclassical brane solutions
Consider the WZW model on the group manifold of SU(2) – we will comment on the generalization to other groups later. A general group element can be parametrized as $`U=\mathrm{exp}(i\stackrel{}{\psi }\stackrel{}{\sigma })`$, where $`\stackrel{}{\psi }`$ is a 3-vector of length $`\psi `$ pointing in the direction ($`\theta ,\varphi `$), and $`\stackrel{}{\sigma }`$ are the usual Pauli matrices. The coordinate $`\psi `$ takes values in the interval $`[0,\pi ]`$ with the two extremes corresponding to the two elements of the center. In these coordinates the metric and Neveu-Schwarz three-form backgrounds read
$$ds^2=k\alpha ^{}\left[d\psi ^2+\mathrm{sin}^2\psi \left(d\theta ^2+\mathrm{sin}^2\theta d\varphi ^2\right)\right],$$
(1)
and
$$HdB=2k\alpha ^{}\mathrm{sin}^2\psi \mathrm{sin}\theta d\psi d\theta d\varphi ,$$
(2)
with $`k`$ the (integer) level of the associated current algebra. We can choose a gauge in which the NS two-form is proportional to the volume form of the two-sphere spanned by $`(\theta ,\varphi )`$,
$$B=k\alpha ^{}\left(\psi \frac{\mathrm{sin2}\psi }{2}\right)\mathrm{sin}\theta d\theta d\varphi .$$
(3)
This is a smooth choice everywhere except at the point $`\psi =\pi `$. The wavefunction of a fundamental string wrapping around this potential singularity picks up a Bohm-Aharonov phase equal to $`_{S^2}B/2\pi \alpha ^{}=2\pi k`$. The singularity is therefore unobservable for integer $`k`$ as it should be.
Let us next put this WZW model together with seven flat space-time coordinates, so that the full geometry is $`S^3\times R^7`$. This is a non-critical background for type-II string theory because the central charges don’t add up to ten, but the dilaton tadpole will not affect our discussion of D-branes at leading order in the string-loop expansion. Consider now a static D2-brane wrapping the $`(\theta ,\varphi )`$ two-sphere at fixed value of $`\psi `$. This configuration breaks the $`SU(2)_L\times SU(2)_R`$ symmetry of the background to a diagonal $`SU(2)`$. If the dominant brane energy were tensive our configuration would tend to shrink to a point at one of the two poles of $`S^3`$, either $`\psi =0`$ or $`\psi =\pi `$. The total brane energy, on the other hand, has contributions also from the induced NS-NS two-form $`\widehat{B}`$ and from the worldvolume gauge field $`F=dA`$, which enter through the invariant combination $`=\widehat{B}+2\pi \alpha ^{}F`$. Consistently with the symmetry we may turn on a uniform worldvolume flux,
$$F=dA=\frac{n}{2}\mathrm{sin}\theta d\theta d\varphi $$
(4)
where $`n`$ is the ‘magnetic monopole’ number. For $`0<n<k`$ one can check that $`||`$ is locally maximum at the poles, so this could prevent the D2-brane from collapsing.
A crucial point in the further considerations is that it is the flux of $`F`$, rather than that of $``$, which is quantized. This may seem counterintuitive in that $`F`$ is not invariant under the gauge transformations $`\delta \widehat{B}=2\pi \alpha ^{}d\mathrm{\Lambda }`$ and $`\delta A=\mathrm{\Lambda }`$. One might have expected the quantization condition to apply to the gauge invariant $``$.
A first comment one can make is that gauge transformations for which $`\mathrm{\Lambda }`$ is single-valued do not affect $`F`$, so claiming that $`F`$ is quantized is not evident nonsense. Although $`\mathrm{\Lambda }`$ need not be single-valued, in fact such large gauge transformations can only shift $`F`$ by an integer. This is how the usual quantization condition on $`H`$ arises in space-time language: one must define $`\widehat{B}`$ in patches on $`S^3`$, and the allowed transition functions between the patches are those respecting this quantization condition.
This shows that the claim that $`F`$ and not $``$ is quantized is sensible, but does not really prove it. Indeed, any argument for this claim which starts from a conventional world-volume gauge theory (such as the Born-Infeld action) would be circular, as the conventional gauge potential only makes sense if $`F`$ is quantized.
As is by now well-known, there do exist other gauge theories such as noncommutative gauge theory in which $`F`$ is not quantized in the usual way. However, the examples in which this is known to make sense at present are related to manifolds with non-trivial fundamental group, such as the torus. Indeed the case of $`S^2`$ has been much studied and the only noncommutative gauge theories which are known to make sense in this case are based on the “fuzzy sphere” and have finitely many degrees of freedom (the original algebra of functions on $`S^2`$ is truncated). There is even a “no-go” theorem to the effect that deformations of the algebra of functions which do not make this truncation, and which respect the natural $`SO(3)`$ symmetry, do not exist as bounded algebras.
Although this is a theorem, we have not proven that its assumptions are the physically appropriate ones, and so we are not claiming at this point to show that noncommutative gauge theory with a field theoretic number of degrees of freedom does not exist on $`S^2`$. The point of this discussion is to explain why there is no sensible candidate low energy theory (at this writing) to describe the alternate hypothesis that $`F`$ is not quantized.
In any case, our hypothesis that $`F`$ is quantized will be confirmed shortly by the beautiful agreement of our results with those of conformal field theory.
Let us now fill in the appropriate formulae. The energy of our D2-brane with $`n`$ units of worldvolume magnetic flux in the semiclassical (large-$`k`$) limit reads
$`E_n(\psi )`$ $`=`$ $`T_{(2)}{\displaystyle _0^\pi }𝑑\theta {\displaystyle _0^{2\pi }}𝑑\varphi \sqrt{\mathrm{det}(\widehat{G}+)}+\mathrm{}`$ (5)
$`=`$ $`4\pi k\alpha ^{}T_{(2)}\left(\mathrm{sin}^4\psi +(\psi {\displaystyle \frac{\mathrm{sin2}\psi }{2}}{\displaystyle \frac{\pi n}{k}})^2\right)^{1/2}+\mathrm{}`$
where $`T_{(2)}`$ is the D-brane tension and the dots stand for higher $`\alpha ^{}`$ corrections (see for instance ). For $`0<n<k`$ this expression has a unique minimum away from the poles, at
$$\psi _n=\frac{\pi n}{k},$$
(6)
where it takes the value
$$M_n=4\pi k\alpha ^{}T_{(2)}\mathrm{sin}\frac{\pi n}{k}.$$
(7)
For values of $`n`$ outside this range the minimum of the energy is at $`\psi =0`$ (if $`n<o`$) or $`\psi =\pi `$ (if $`n>k`$) and it corresponds to a singular configuration of the brane. This gives a total of $`k1`$ non-singular configurations. In order to take into account the well-known one loop shift of the sphere curvature, we should replace everywhere $`k`$ by $`k+2`$. There is however no reason at this point to trust our expressions beyond the large $`k`$ and $`n`$ (with $`n/k`$ held fixed) limit, since only in this limit are the wordvolume curvatures small in string units.
We can also evaluate the D-particle charge induced by the background flux on the above stable non-degenerate D2-branes. Using the standard formulae one finds
$$Q_n=T_{(2)}_{S^2}=2\pi k\alpha ^{}T_{(2)}\mathrm{sin}\frac{2\pi n}{k}.$$
(8)
These charges are not even rationally related to each other – here is the apparent paradox we have alluded to in the introduction. Note also that in the flat limit, $`k\mathrm{}`$ with $`n`$ held fixed, eqs. (7), (8) reduce to the mass and charge of $`n`$ free D-particles, $`E_nQ_nnT_{(0)}`$. The above stable configurations should have in fact a dual description as bound states of $`n`$ D-particles on the sphere, which can be analyzed along the lines of . From the exact properties of these bound states we can place restrictions on the non-abelian Born Infeld theory, similar to those of , but this is outside the scope of the present work.
## 3 Small fluctuations
The ‘mini-superspace’ analysis of the previous section took only into account the degree of freedom corresponding to rigid motions of the D2-brane in the $`\psi `$-direction. In this section we derive the complete spectrum of small quadratic fluctuations around the above D-brane solutions. This will allow us to confirm their stability, to find their classical moduli space, and to compare later on with the spectrum of boundary operators for the corresponding Cardy states.
We use static gauge in which the worldvolume is parametrised by $`(t,\theta ,\varphi )`$, and impose $`A_0=0`$ for the worldvolume gauge field. We ignore for simplicity brane fluctuations and gauge-field components in the extra spectator spatial directions, and concentrate on the three remaining degrees of freedom
$$\psi =\frac{\pi n}{k}+\delta ,A_\theta =\frac{k}{2\pi }\alpha _\theta ,\mathrm{and}A_\varphi =\frac{n}{2}(\mathrm{cos}\theta 1)+\frac{k}{2\pi }\alpha _\varphi .$$
(9)
Here the small fluctuations $`\delta ,\alpha _\theta ,\alpha _\varphi `$ are arbitrary functions on $`R\times S^2`$, and the $`k/2\pi `$ normalization is introduced for convenience. The Born-Infeld energy-density reads
$$_{BI}=T_{(2)}\sqrt{\mathrm{det}(\widehat{G}+)}$$
(10)
where
$$\widehat{G}+=k\alpha ^{}\left(\begin{array}{ccccc}\frac{1}{k\alpha ^{}}+(_t\delta )^2\hfill & & _t\delta _\theta \delta +_t\alpha _\theta \hfill & & _t\delta _\varphi \delta +_t\alpha _\varphi \hfill \\ _t\delta _\theta \delta _t\alpha _\theta \hfill & & \mathrm{sin}^2\psi +(_\theta \delta )^2\hfill & & _\theta \delta _\varphi \delta +_{\theta \varphi }\hfill \\ _t\delta _\varphi \delta _t\alpha _\varphi \hfill & & _\theta \delta _\varphi \delta _{\theta \varphi }\hfill & & \mathrm{sin}^2\psi \mathrm{sin}^2\theta +(_\varphi \delta )^2\hfill \end{array}\right)$$
(11)
with
$$_{\theta \varphi }\left(\delta \frac{\mathrm{sin}(2\psi )}{2}\right)\mathrm{sin}\theta +_\theta \alpha _\varphi _\varphi \alpha _\theta .$$
(12)
In expanding out the determinant to quadratic order, terms involving $`\delta `$ off the diagonal drop out. After some tedious but straightforward algebra the Born-Infeld lagrangian up to quadratic order takes the form
$$\begin{array}{c}_{BI}2f\mathrm{cot}(\frac{n\pi }{k})+k\alpha ^{}\mathrm{sin}\theta \left[(_t\delta )^2+(_t\alpha _\theta )^2+(_t\alpha _\varphi )^2/\mathrm{sin}^2\theta \right]\hfill \\ \\ (1/\mathrm{sin}\theta )\left[(_\varphi \delta )^2+\mathrm{sin}^2\theta (_\theta \delta )^2+2\mathrm{sin}^2\theta \delta ^2+4\mathrm{sin}\theta \delta f+f^2\right].\hfill \end{array}$$
(13)
We have here denoted $`f_\theta \alpha _\varphi _\varphi \alpha _\theta `$ for short, and have dropped the leading, fluctuation-independent term of the lagrangian as well as an irrelevant multiplicative constant.
The above expression starts out with a linear term, which seems to contradict our assertion that we are expanding around a classical solution. This is however not the case. The linear term is proportional to the fluctuation of the integrated flux, which must be set to zero because of the quantization condition. This demonstrates explicitly that it is the magnetic flux that stabilizes the D-brane. The quadratic terms in (13) are furthermore independent of the background solution we expand around. This means that the spectrum of fluctuations is independent of $`n`$, in agreement with the conformal field-theory result as we will see soon.
From the above lagrangian we derive the following linearized equations for the fluctuation fields,
$$\frac{d^2}{dt^2}\left(\begin{array}{c}\delta \hfill \\ \alpha _\theta \hfill \\ \alpha _\varphi \hfill \end{array}\right)=𝒪\left(\begin{array}{c}\delta \hfill \\ \alpha _\theta \hfill \\ \alpha _\varphi \hfill \end{array}\right),$$
(14)
where the operator-valued matrix is
$$𝒪=\frac{1}{k\alpha ^{}}\left(\begin{array}{ccccc}\text{ }\text{ }\text{ }\text{ }+2\hfill & & \frac{2}{\mathrm{sin}\theta }_\varphi \hfill & & \frac{2}{\mathrm{sin}\theta }_\theta \hfill \\ & & & & \\ \frac{2}{\mathrm{sin}\theta }_\varphi \hfill & & \frac{1}{\mathrm{sin}^2\theta }_\varphi ^2\hfill & & \frac{1}{\mathrm{sin}^2\theta }_\varphi _\theta \hfill \\ & & & & \\ \frac{2}{\mathrm{sin}\theta }_\theta \hfill & & \mathrm{sin}\theta _\theta \frac{1}{\mathrm{sin}\theta }_\varphi \hfill & & \mathrm{sin}\theta _\theta \frac{1}{\mathrm{sin}\theta }_\theta \hfill \end{array}\right)$$
(15)
and
$$\text{ }\text{ }\text{ }\text{ }\text{ }=\frac{1}{\mathrm{sin}^2\theta }_\varphi ^2\frac{1}{\mathrm{sin}\theta }_\theta \mathrm{sin}\theta _\theta $$
(16)
is the covariant Laplacian on $`S^2`$. The operator $`𝒪`$ has zero eigenvalues corresponding to the unphysical longitudinal polarization of the photon. We can extract the transverse polarization by combining the last two equations so as to express everything in terms of the physical fluctuation $`f`$. After some algebra the answer is
$$\frac{d^2}{dt^2}\left(\begin{array}{c}\delta \hfill \\ f/\mathrm{sin}\theta \hfill \end{array}\right)=\frac{1}{k\alpha ^{}}\left(\begin{array}{ccc}\text{ }\text{ }\text{ }\text{ }+2\hfill & & 2\hfill \\ 2\text{ }\text{ }\text{ }\text{ }\hfill & & \text{ }\text{ }\text{ }\text{ }\hfill \end{array}\right)\left(\begin{array}{c}\delta \hfill \\ f/\mathrm{sin}\theta \hfill \end{array}\right).$$
(17)
We can now readily diagonalize this operator by going to a basis of spherical harmonics,
$$\delta =\underset{l=0}{\overset{\mathrm{}}{}}\underset{m=l}{\overset{l}{}}\delta _{lm}Y_{lm}(\theta ,\varphi )\mathrm{and}f=\mathrm{sin}\theta \underset{l=1}{\overset{\mathrm{}}{}}\underset{m=l}{\overset{l}{}}f_{lm}Y_{lm}(\theta ,\varphi )$$
(18)
with the reality conditions $`\delta _{lm}=\delta _{lm}^{}`$ and similarly for $`f`$. Notice that the absence of the s-wave in the expansion of $`f`$ guarantees the flux quantization condition, as can be checked using the orthonormality of the spherical harmonics. For $`l=0`$ there is therefore only the $`\delta `$-fluctuation, and its frequency squared is $`2/k\alpha ^{}`$. For all other $`l`$ we need to consider the matrix
$$\frac{1}{k\alpha ^{}}\left(\begin{array}{ccc}l(l+1)+2\hfill & & 2\hfill \\ 2l(l+1)\hfill & & l(l+1)\hfill \end{array}\right)$$
(19)
whose eigenvalues are $`(l+1)(l+2)/k\alpha ^{}`$ and $`l(l1)/k\alpha ^{}`$. Putting it all together we thus have the following spectrum of quadratic fluctuations
$$m^2=j(j+1)/k\alpha ^{},\mathrm{in}\mathrm{reps}.(j1)(j+1),\mathrm{for}j=0,1,2,\mathrm{}$$
(20)
with the understanding that only the spin-one representation appears in the special case $`j=0`$. This corresponds precisely to a triplet of zero modes, corresponding to arbitrary rotations of the worldvolume two-sphere inside $`S^3`$. All other excitations have positive mass, confirming the stability of our solutions.
## 4 CFT analysis
Let us now compare the results of the last two sections with those of conformal field theory. In the diagonal-invariant bosonic theory there are ($`k+1`$) Cardy boundary states preserving a SU(2) symmetry,
$$|q_C\underset{p=0}{\overset{k}{}}\frac{S_{qp}}{\sqrt{S_{0p}}}|p_I$$
(21)
where $`|p_I`$ is the Ishibashi (character) state corresponding to the chiral primary of spin $`p/2`$, and
$$S_{ij}=\sqrt{\frac{2}{k+2}}\mathrm{sin}\left(\frac{(i+1)(j+1)\pi }{k+2}\right)$$
(22)
is the modular-transformation matrix.
We want to identify these states with the semiclassical configurations of section 2. We have seen that there are $`k1`$ stable non-degenerate D2-branes, when the $`S^3`$ radius is $`\sqrt{k\alpha ^{}}`$, and adding the two D-particles at the north and south poles gives the correct total number. Alternatively, if we take into account the quantum shift $`kk+2`$, we find exactly $`k+1`$ non-degenerate D2 branes. These two ways of counting are of course indistinguishable in the semiclassical large $`k`$ limit, since it is hard to tell the difference between a point-like D-particle and one with radius $`\sqrt{\alpha ^{}}`$ (this is the radius of the worldvolume for $`n=1`$ units of flux). Adopting the latter point of view makes, however, the Born-Infeld results exact – this may be related to supersymmetry in the fivebrane context, but we did not have any reason to expect it a priori. To exhibit this precise agreement of the formulae we will assume that the $`S^3`$ radius is $`\sqrt{(k+2)\alpha ^{}}`$, and identify
$$|n1_C(\mathrm{flux}n\mathrm{D2}\mathrm{brane})$$
(23)
where the flux $`n`$ takes the values $`1,2,..,k+1`$.
The mass of a Cardy state can be read off from the coefficient of its $`p=0`$ Ishibashi component which has non-vanishing overlap with the (seven-dimensional) graviton and dilaton, see for instance . To be more precise, the interaction energy between two D-branes a distance $`r`$ apart due to the exchange of the (seven-dimensional) graviton and dilaton can be calculated along the lines of with the result<sup>2</sup><sup>2</sup>2We can formally ignore the dilaton tadpole in this calculation.
$$(r)=2\kappa _{(7)}^2M_n^2\mathrm{\Delta }_{(6)}(r)$$
(24)
where $`\kappa _{(7)}`$ is the gravitational coupling in seven dimensions, $`M_n`$ the mass of the D-branes and $`\mathrm{\Delta }_{(6)}(r)`$ the six-dimensional Euclidean Green’s function. The string-theory calculation for this interaction energy on the other hand is
$$(r)=(2\pi )^4\frac{|S_{n0}|^2}{S_{00}}\mathrm{\Delta }_{(6)}(r)$$
(25)
where we have here projected onto the identity-operator in the closed-string channel of the amplitude, and we have set $`\alpha ^{}=1/2`$. Comparing the two expressions gives
$$M_n=\frac{2\sqrt{2}\pi ^2}{\kappa _{(7)}}\frac{|S_{n0}|}{\sqrt{S_{00}}}=\frac{(2\pi )^{3/2}(2k+4)^{1/4}}{\kappa _{(7)}}\mathrm{sin}\left(\frac{n\pi }{k+2}\right),$$
(26)
where we have here used the expressions for the S-matrix coefficients. Using now the relations
$$T_{(2)}=\frac{\sqrt{2\pi ^3}}{\kappa _{(10)}},\mathrm{and}\kappa _{(10)}^2=\kappa _{(7)}^2\left(\frac{k+2}{2}\right)^{3/2}2\pi ^2$$
(27)
one can verify that the above mass agrees precisely with our semiclassical result (7), including all the numerical prefactors.
Consider next the spectrum of quadratic fluctuations, to be compared with the open-string excitations in the $`_{(n1)(n1)}`$ Hilbert space. If we neglect transverse spatial dimensions, the light states in this Hilbert space are of the form
$$|\mathrm{open}>=J_1^a|j>$$
(28)
where $`J^a`$ are the SU(2) currents and $`|j>`$ is created by a primary field with $`j=0,\mathrm{}k/2`$. These transform in the $`(j1)j(j+1)`$ representations of $`SU(2)`$, but imposing the (super)Virasoro constraint will project the representation $`j`$ out of the spectrum. One way to see this is to note that there are as many constraints as number of primaries, namely $`2j+1`$, and since physical states must form $`SU(2)`$ representations it is necessarily the $`j`$ representation that is projected out. The conformal weight of the vertex operators corresponding to the states (28) is $`j(j+1)/(k+2)`$ , in complete agreement with the semiclassical mass formula of small fluctuations derived in section three.
Another qualitative confirmation of the results in section two follows from an analysis of the ‘wavefunctions’ of the Cardy states in position space . These are peaked around equally-spaced values of the polar angle $`\psi `$, in agreement again with the semiclassical result (6). Notice that the moduli space for rigid translations of the D2-branes on the group manifold corresponds to the freedom of obtaining equivalent Cardy states by group conjugation.
Finally, let us compare the induced D-particle charge (8) with the result of CFT. In the CFT this charge is given by the $`p=1`$ coefficient of the Cardy state, because the corresponding closed-string RR states transform in the $`(p/21/2,p/21/2)`$ representation of $`SU(2)_L\times SU(2)_R`$. The reason is that the zero modes of the supersymmetric WZW fermions are realized on a bispinor of $`SU(2)_L\times SU(2)_R`$. Now the D-particle charge of interest is a $`SO(4)`$ singlet – this can be verified explicitly by checking that it does not transform under rigid translations of the D2-brane on $`S^3`$. Thus only $`p=1`$ contributes to this coupling, and a calculation similar to the one for the mass gives
$$Q_n=\frac{(2\pi )^{3/2}(2k+4)^{1/4}}{2\kappa _{(7)}}\mathrm{sin}\left(\frac{2n\pi }{k+2}\right),$$
(29)
in perfect agreement again with the result of section two. That these charges are not rationally related to each other is thus confirmed by the CFT analysis – we will return to the point in the final section.
## 5 General group manifolds
In this section we discuss some aspects of the generalization of our results to compact Lie groups $`G`$ which we assume to be simple, connected and, for simplicity, to be simply connected. Our discussion will be entirely topological, the precise form of the metric and antisymmetric tensor will not play any role. The reader not interested in this generalization can go directly to the final section.
The D-brane world volumes for all boundary conditions for which the gluing of left movers and right movers at the boundary is given by an automorphism $`\omega `$ of $`G`$ have been described in . They are (regular) twined conjugacy classes, i.e. they are subspaces of the form
$$𝒞_\omega (g)=\{hg\omega (h)^1\text{ with }hG\}$$
(30)
where $`gG`$ is a regular element.
Let us describe the geometry of twined conjugacy classes in somewhat more detail: for any automorphism $`\omega `$, there is a maximal torus $`T`$ of $`G`$ that is invariant under $`\omega `$. The subgroup of elements of $`T`$ that are left pointwise fixed by $`\omega `$,
$$T^\omega =\{tT|\omega (t)=t\},$$
(31)
is not a torus, but a semi-direct product of a torus and a finite abelian group; its connected component $`T_0^\omega `$ of the identity is a torus. In case $`\omega `$ is an inner automorphism – and this is always true for $`G=SU(2)`$ – all subgroups of $`G`$ coincide:
$$T=T_0^\omega =T^\omega .$$
(32)
For inner automorphisms, this torus actually coincides with a maximal torus and the dimension of the torus equals the rank of $`G`$; so e.g. for inner automorphisms of $`SU(3)`$ we have a two-dimensional torus. For outer automorphisms, the dimension of $`T_0^\omega `$ is smaller than the rank; for outer automorphisms of $`SU(3)`$ it is equal to one.
Weyl’s classical theory of conjugacy classes has in fact a nice generalisation to twined conjugacy classes. We will sketch some statements of this theory. The central tool is the following: given a maximal torus $`T`$, we define a map from the coset space $`G/T`$ and the maximal torus $`T`$ to the group by using conjugation:
$$\begin{array}{c}q:G/T\times TG\hfill \\ q([g],t)=gtg^1\hfill \end{array}$$
(33)
Weyl could show that the mapping degree of $`q`$ equals the number of elements in the Weyl group $`W`$. Maps with positive degree are surjective, so in particular one sees that any element of $`G`$ is conjugated to some element in the maximal torus $`T`$. So any conjugacy class can be characterized by an element of the maximal torus. In fact, different elements of the maximal torus parametrize identical conjugacy classes if and only if they are related by the action of the Weyl group. In the case of $`SU(2)`$, a maximal torus is one-dimensional; examples are given by circles of constant values of $`\varphi `$ and $`\theta `$; they can be parametrized by $`\psi `$. The Weyl group is just $`Z_2`$, and its action has been taken into account by restricting $`\psi `$ to the range $`0\psi \pi `$. Finally, fixing $`t`$ we see that regular conjugacy classes are isomorphic to the homogeneous space $`G/T`$. In the case of $`SU(2)`$ this gives $`SU(2)/U(1)`$ which is isomorphic to the two-sphere.
The results nicely generalize to twisted conjugacy classes (for details see ). For any automorphism $`\omega `$, $`q`$ is replaced by $`q_\omega `$ that is defined via twisted conjugation:
$$\begin{array}{c}q_\omega :G/T_0^\omega \times T_0^\omega G\hfill \\ q_\omega ([g],t)=gt\omega (g^1)\hfill \end{array}$$
(34)
The mapping degree of $`q_\omega `$ can be shown to be positive. To state more precise results, we need the subgroup $`W_\omega `$ of the Weyl group $`W`$ that commutes with the action of $`W`$ on the weight space:
$$W_\omega :=\{wW|\omega ^{}w=w\omega ^{}\text{ for all }wW\}$$
(35)
The group $`W_\omega `$ has been shown in to be isomorphic to the Weyl group of the so-called orbit Lie algebra . For the outer automorphism of $`SU(3)`$ this group is $`Z_2`$, which is the Weyl group of the orbit Lie algebra $`SU(2)`$.
The mapping degree of $`q_\omega `$ is just $`n_{T^\omega }|W_\omega |`$, where $`n_{T^\omega }`$ is the number of connected components of $`T^\omega `$. Weyls classical results can now be generalized to twined conjugacy classes. All statements remain true, provided one replaces the maximal torus $`T`$ by $`T_0^\omega `$ and the Weyl group $`W`$ by $`W_\omega `$: Twined conjugacy classes are characterized by elements of $`T_0^\omega `$; different elements of $`T_0^\omega `$ describe identical twined conjugacy classes if and only if they are related by the action of $`W_\omega `$. Regular twined conjugacy classes are isomorphic to the homogeneous space $`G/T_0^\omega `$. Even a generalization of Weyl’s integration formula holds .
To give an explicit example, regular D-branes for inner automorphisms of $`SU(3)`$ are isomorphic to $`SU(3)/U(1)^2`$ and are thus six-dimensional. They are characterized by two parameters. For outer automorphisms, they are isomorphic to $`SU(3)/U(1)`$ and therefore seven-dimensional. For their characterization a single parameter suffices. Outer automorphisms therefore change the dimensionality of the worldvolume.
Extending the analysis of $`SU(2)`$ to other groups requires a detailed knowledge of the differential geometry of the group manifold, in particular a good choice of coordinates. The corresponding calculations become rather complicated and are beyond the scope of the present note. However, there is a simple and yet non-trivial check of the stabilization mechanism: we expect as many independent $`U(1)`$ fluxes as the number of transverse brane coordinates that must be stabilized.
The possible $`U(1)`$ fluxes on the worldvolume of the D-brane are given by
$$\mathrm{dim}H^2(G/T_0^\omega ,\mathrm{R})$$
(36)
Following our previous discussion, the description of a specific D-brane requires on the other hand $`dimT_0^\omega `$ parameters. We should thus expect the general relation
$$\mathrm{dim}H^2(G/T_0^\omega ,𝐑)=\mathrm{dim}T_0^\omega $$
(37)
Such a relation does indeed hold in full generality: for a simply connected compact Lie group $`G`$ also the second homotopy group $`\pi _2(G)`$ vanishes. The long exact sequence in homotopy
$$\mathrm{}\pi _k(G)\pi _k(G/T_0^\omega )\pi _{k1}(T)\pi _{k1}(G)\mathrm{}$$
(38)
implies for $`k=1`$ that the homogeneous space $`G/T_0^\omega `$ is simply connected and for $`k=2`$ that $`\pi _2(G/T_0^\omega )`$ is isomorphic to $`\pi _1(T_0^\omega )`$. The latter is a free abelian group whose rank is $`dimT_0^\omega `$. The homotopy group $`\pi _2(G/T_0^\omega )`$ therefore coincides with the homology we want to determine, and we find indeed that
$$H^2(G/T_0^\omega ,R)\pi _2(G/T_0^\omega )\pi _1(T_0^\omega )𝐙^{\mathrm{dim}T_0^\omega }$$
(39)
Notice in particular that the line bundles over $`G/T_0^\omega `$ do not have any continuous parameters. This generalizes the situation of $`SU(2)`$, where we consider bundles over $`S^2`$. This is reflected in the conformal field theory analysis by the fact that we find D-brane worldvolumes whose only continuous deformations are given by (inner) automorphisms of the group, but which do not have any other moduli.
The fact that the relation (37) always holds shows that the advocated mechanism could indeed be responsible for the stability of all known WZW D-branes.
## 6 Fivebrane and a paradox
To further discuss the physics of the $`SU(2)`$ branes, let us consider a configuration of $`N`$ coincident supersymmetric (NS) five-branes in type II theory.
The full fivebrane background is (in string frame)
$`ds^2`$ $`=`$ $`dx^2+f(r)dy^2`$ (40)
$`e^{2\varphi }`$ $`=`$ $`g_s^2f(r)`$ (41)
$`f(r)`$ $`=`$ $`1+{\displaystyle \frac{N\alpha ^{}}{r^2}}`$ (42)
$`H`$ $`=`$ $`N\alpha ^{}ϵ_3`$ (43)
where $`x`$ are the $`5+1`$ longitudinal coordinates, $`y`$ are $`4`$ transverse coordinates and $`r=|y|`$.
In the near-horizon limit $`r0`$, the background factorizes into a radial component and an $`S^3`$. The corresponding CFT has a Feigin-Fuchs (linear dilaton) field in addition to the supersymmetric SU(2) WZW model . The supersymmetric WZW model can be realized as a tensor product of three free fermions (with level $`2`$ current algebra) and a bosonic WZW model of level $`k`$ (the $`k`$ of the previous sections). The fivebrane number $`N`$ is identified with the total central charge $`k+2`$ of the $`SU(2)`$ current algebra. This may sound unsatisfactory, as there is apparently no candidate theory for a single five-brane, but it agrees with the standard lore that the center-of-mass degrees of freedom of the branes are not be visible in the dual holographic theory. A single fivebrane has no degrees of freedom other than center-of-mass, and hence no dual holographic description.
The D particles of the previous section are nonsupersymmetric and unstable in this background. They are momentum modes in the eleventh dimension, which tend to fall towards the core of the fivebrane where the eleventh dimension blows up and the D-particles become massless. This agrees with the expectation that they should complete SO(5) representations of the fivebrane fields .
D$`2`$-branes which end on the fivebrane are supersymmetric and correspond to the product of a Neumann boundary state in the linear dilaton theory, and a Dirichlet (D$`0`$) boundary state in the WZW model. In this context, the additional WZW boundary states will correspond to D$`4`$ branes extending along the radial direction (and so ending on the fivebranes), but now (in the near horizon regime) “wrapped” on an transverse $`S^2`$, forming a conical geometry.
To study the supersymmetry properties of these branes, we need to write down the space-time supersymmetry generators. The world-volume supersymmetry generators are given in and the D$`2`$ boundary conditions in the WZW model preserve an $`N=4`$ world-sheet supersymmetry. This is enough to guarantee that the world-volume operator corresponding to the space-time supercharge also exists with these boundary conditions.
The conjugate brane (in the sense of electric-magnetic duality) would be a D$`4`$-brane with three dimensions wrapped on $`S^3`$ and $`1+1`$ extending in Minkowski space. Although this of course does not exist because it is wrapping a surface with $`H0`$, there is a similar object which is believed to exist . The total flux $`H`$ on the brane can be made zero by allowing $`k`$ D$`2`$-branes to end on the brane (and extend outward), analogous to the “baryon” of .
The existence of the conjugate object would appear to require D$`2`$ charge quantization. So why is RR charge not quantized ?
We first note that there is a superficially similar effect, already visible in toroidal compactification, with a much simpler explanation. Since the integrals $`\widehat{B}`$ over two-cycles in the target space can take arbitrary non-zero values, the induced RR charges $`C\widehat{B}`$ are non-integral. However, this is not a violation of quantization but rather a rotation of the entire charge lattice.
A simple way to see that this is not what is happening here is to realize that charge quantization requires that there be an integral basis of the charge lattice of rank equal to the number of charges (one must then check that the DSZ form is integral of course). In the present case (and for $`SU(2)`$), we have found $`O(k/2)`$ distinct ‘charges’ satisfying no integer relations, but at most two independent RR charges (from the two-dimensional cohomology of $`S^3`$).
The way one avoids an immediate contradiction is by noting that the RR fields in question are massive in the near-horizon geometry. This can be seen in the CFT where the normalizable vertex operators have (six-dimensional) mass bounded below by the background charge of the Feigin-Fuchs coordinate . Alternatively, in the low energy effective theory, this follows from the Chern-Simons coupling $`GG\widehat{B}`$. We work in the usual string conventions in which the RR kinetic terms and sources are independent of the dilaton. This leads to
$$dG=HG+\delta ^{(7)}+B\delta ^{(5)}$$
(44)
and
$$dG=\delta ^{(5)}$$
(45)
where $`G=dC^{(3)}`$ is the four-form field strength, $`\delta ^{(9p)}`$ is the source associated to a $`p+1`$-brane (a $`9p`$ form normal to the world-volume). The conserved electric and magnetic RR charges are then
$$Q_M=G$$
and
$$Q_E=GBG.$$
In the near-horizon limit of the five-brane background, $`B`$ is independent of distance $`r`$. In this case the CS term makes the RR field effectively massive, so the quantization condition will not be visible.
In the true five-brane background, when we go to asymptotic infinity (the Minkowski region), the $`B`$ field does fall off with distance, and the charge quantization must become observable again. This regime is of course not described by our explicit CFT. Since the volume of the $`S^3`$ grows with radius in the normal way here, presumably the conical four-brane must asymptote to a cylindrical four-brane (or even $`n`$ two-branes again). This fits with the asymptotic BPS bound which implies that the tension here is just $`n`$ times that of the original two-brane.
In a general background with non-zero $`H`$, it is clear that the induced RR charge will depend on the embedding of the brane through $`\widehat{B}`$ and so this contribution cannot be quantized. However this variation could be cancelled by the bulk CS term: the total variation of the right hand side of (44) under a variation of the embedding of the D$`4`$-brane is zero, under a suitable interpretation of the boundary terms.
These points seem to us to resolve the paradox both in the context of the near-horizon geometry (because the RR fields are massive) and in the full geometry (in which the $`H`$ flux falls off fast enough to apply the second argument). However they leave open the interesting question of just what the CFT results (29) are measuring. One possibility is that they are related to the $`N1=k+1`$ independent charges of the holographically dual field theory on the fivebranes. For $`N`$ type IIB fivebranes the worldvolume theory has $`SU(N)`$ gauge symmetry, and hence $`N1`$ independent charges. These charges may indeed correspond to the different allowed couplings of the massive RR field in the bulk theory.
We feel there is more to say about this issue, but will leave it for future work.
Acknowledgements
The authors thank the organizers of the workshop on ‘Non-Commutative Gauge Theory’ at the Lorentz Center in Leiden, Holland, where this collaboration started. CB thanks the New Center for High Energy Physics at the University of Rutgers for hospitality during completion of this work. MRD would like to thank A. Rajaraman and M. Rozali for discussions on this problem and especially for emphasizing the importance of RR charge quantization. We also thank O. Aharony, J. Maldacena and J. Polchinski for useful comments.
|
warning/0003/nucl-th0003044.html
|
ar5iv
|
text
|
# Momentum conservation and local field corrections for the response of interacting Fermi gases
\[
## Abstract
We reanalyze the recently derived response function for interacting systems in relaxation time approximation respecting density, momentum and energy conservation. We find that momentum conservation leads exactly to the local field corrections for both cases respecting only density conservation and respecting density and energy conservation. This rewriting simplifies the former formulae dramatically. We discuss the small wave vector expansion and find that the response function shows a high frequency dependence of $`\omega ^5`$ which allows to fulfill higher order sum rules. The momentum conservation also resolves a puzzle about the conductivity which should only be finite in multicomponent systems.
\]
Recently the improvement of the response function in interacting quantum systems has regained much interest . This quantity is important in a variety of fields and describes the induced density variation if the system is externally perturbed: $`\delta n=\chi V^{\mathrm{ext}}`$. As an example for an interacting system with potential $`V`$ the conductivity can be calculated from the response function via
$`\mathrm{Re}\sigma ={\displaystyle \frac{V}{4\pi }}\omega \mathrm{Im}\chi .`$ (1)
One of the most fruitful concepts to improve the response functions including correlations are the local field corrections $`G`$
$`\chi ={\displaystyle \frac{\chi _0}{1+G\chi _0}},`$ (2)
see and references therein.
On the other hand there exists an extremely useful form of the response function when the interactions are abbreviated in the relaxation time approximation $`\tau `$ respecting density conservation. One of the advantages of the resulting Mermin formula (11) is that it leads to the Drude -like form of the dielectric function in the long wavelength limit
$`ϵ=1V\chi =1{\displaystyle \frac{\omega _p^2}{\omega (\omega +\frac{i}{\tau })}}`$ (3)
with the plasma frequency $`\omega _p`$ for the Coulomb potential $`V`$ from which follows the conductivity
$`\mathrm{Re}\sigma ={\displaystyle \frac{ne^2\tau }{m(1+\omega ^2\tau ^2)}}=\{\begin{array}{c}\frac{ne^2\tau }{m}+o(\omega )\\ \frac{ne^2}{m\omega ^2\tau }+o(\frac{1}{\omega })\end{array}.`$ (4)
However one should note that this formula is valid only for the extension to a multicomponent system (at least a two-component system) since it makes no sense to speak of conductivity in a single component system where the conductivity should be infinite. Clearly the Mermin formula does not distinguish these cases and cannot be sufficient to describe the response. Therefore we will show that the inclusion of additional momentum conservation will repair this defect (39) and will lead to a conductivity
$`\mathrm{Re}\sigma ={\displaystyle \frac{ne^2\tau }{m(1+\omega ^2\tau ^2)}}{\displaystyle \frac{nq^2}{m\omega ^2}}\left({\displaystyle \frac{1}{_\mu n}}{\displaystyle \frac{2E}{n^2}}\right)`$ (5)
which shows indeed for the static limit a diverging behavior in contrast to (4).
There are two distinguishable cases, the single component case where we have to include momentum conservation and obtain divergent conductivity and the multicomponent case where we should expect Mermin-like formulae in order to render the conductivity finite. In order to bring these two extreme cases together the response function for multicomponent systems should be considered .
In this letter we want to restrict to the one - component situation. In we have derived the density, current and energy response $`\chi ,\chi _J,\chi _E`$ of an interacting quantum system
$`\left(\begin{array}{c}\delta n\\ \delta 𝐉\\ \delta E\end{array}\right)=\left(\begin{array}{c}\chi \\ \chi _J\\ \chi _E\end{array}\right)V^{\mathrm{ext}}𝒳\left(\begin{array}{c}1\\ 0\\ 0\end{array}\right)V^{\mathrm{ext}}𝒳\nu ^{\mathrm{ext}}`$ (6)
(7)
to the external perturbation $`V^{\mathrm{ext}}`$ provided the density, momentum and energy are conserved. The interacting system has been described by the quantum kinetic equation for the density operator in relaxation time approximation where the relaxation is considered with respect to the local density operator or the corresponding local equilibrium distribution function. This local equilibrium is given by a local chemical potential $`\mu `$, a local temperature $`T`$ and a local momentum $`Q`$ of mass motion. These local quantities are specified by the requirement that the expectation values for density, momentum and energy are the same when calculated from local distribution function or performed with the density operator.
The density response functions have been expressed in for the inclusion of successively more conservation laws in terms of polarization functions $`𝒫=\{\mathrm{\Pi },\mathrm{\Pi }_n,\mathrm{\Pi }_E\}`$ and have the general form
$`𝒳=𝒫(1𝒱𝒫)^1`$ (8)
due to the induced mean fields which can have density - and momentum - dependent Skyrme form.
When we note the free response function or Lindhard polarization function without collisions as
$`\mathrm{\Pi }_0=s{\displaystyle \frac{d𝐩}{(2\pi )^3}\frac{f_0(𝐩+\frac{𝐪}{2})f_0(𝐩\frac{𝐪}{2})}{\frac{\mathrm{𝐩𝐪}}{m}\omega i0}}`$ (9)
with finite temperature Fermi functions $`f_0`$, the inclusion of only density conservation leads to the Mermin polarization
$`\mathrm{\Pi }^\mathrm{n}(𝐪,\omega )`$ $`=`$ $`{\displaystyle \frac{\mathrm{\Pi }_0(𝐪,\omega +i/\tau )}{1{\displaystyle \frac{1}{1i\omega \tau }}\left[1{\displaystyle \frac{\mathrm{\Pi }_0(𝐪,\omega +i/\tau )}{\mathrm{\Pi }_0(𝐪,0)}}\right]}}`$ (10)
$`=`$ $`(1i\omega \tau ){\displaystyle \frac{g_1(\omega +\frac{i}{\tau })g_1(0)}{h_1}}.`$ (11)
If we include also the energy conservation we obtain an additional term to (11)
$`\mathrm{\Pi }^{\mathrm{n},\mathrm{E}}(\omega )`$ $`=`$ $`(1i\omega \tau )({\displaystyle \frac{g_1(\omega +\frac{i}{\tau })g_1(0)}{h_1}}`$ (13)
$`\omega \tau i{\displaystyle \frac{(h_ϵg_1(0)h_1g_ϵ(0))^2}{h_1(h_ϵ^2h_{ϵϵ}h_1)}})`$
where we use the abbreviation
$`h_\varphi =g_\varphi (\omega +{\displaystyle \frac{i}{\tau }})\omega \tau ig_\varphi (0).`$ (14)
The different occurring correlation functions can be written in terms of moments of the usual Lindhard polarization function (9) as follows :
$`g_1`$ $`=`$ $`\mathrm{\Pi }_0,`$ (15)
$`g_ϵ`$ $`=`$ $`{\displaystyle \frac{n}{2}}+{\displaystyle \frac{m\omega ^2}{2q^2}}\mathrm{\Pi }_0+{\displaystyle \frac{1}{2m}}\stackrel{~}{\mathrm{\Pi }}_2,`$ (16)
$`g_{ϵϵ}`$ $`=`$ $`{\displaystyle \frac{7}{6}}E{\displaystyle \frac{nq^2}{16m}}(1+{\displaystyle \frac{4m^2\omega ^2}{q^4}}){\displaystyle \frac{m^2\omega ^4}{4q^4}}\stackrel{~}{\mathrm{\Pi }}_0`$ (18)
$`{\displaystyle \frac{\omega ^2}{2q^2}}\stackrel{~}{\mathrm{\Pi }}_2{\displaystyle \frac{1}{4m^2}}\stackrel{~}{\mathrm{\Pi }}_4.`$
Integration via the chemical potential yields the higher moments of the polarization function
$`\stackrel{~}{\mathrm{\Pi }}_2`$ $`=`$ $`2m{\displaystyle \underset{\mathrm{}}{\overset{\mu }{}}}𝑑\mu ^{}\mathrm{\Pi }_0,`$ (19)
$`\stackrel{~}{\mathrm{\Pi }}_4`$ $`=`$ $`2(2m)^2{\displaystyle \underset{\mathrm{}}{\overset{\mu }{}}}𝑑\mu ^{}{\displaystyle \underset{\mathrm{}}{\overset{\mu ^{}}{}}}𝑑\mu ^{\prime \prime }\mathrm{\Pi }_0`$ (20)
and the density and energy are given by
$`n`$ $`=`$ $`{\displaystyle \frac{dp}{(2\pi \mathrm{})^3}f_0(p)},`$ (21)
$`E`$ $`=`$ $`{\displaystyle \frac{dp}{(2\pi \mathrm{})^3}\frac{p^2}{2m}f_0(p)}.`$ (22)
For the inclusion of additional momentum conservation to formulae (11) or (13) we obtain now a tremendous simplification by observing that the formulae given in can be rewritten as
$`{\displaystyle \frac{1}{\mathrm{\Pi }^{\mathrm{n},\mathrm{j}}(\omega )}}{\displaystyle \frac{1}{\mathrm{\Pi }^\mathrm{n}(\omega )}}`$ $`=`$ $`{\displaystyle \frac{1}{\mathrm{\Pi }^{\mathrm{n},\mathrm{j},\mathrm{E}}(\omega )}}{\displaystyle \frac{1}{\mathrm{\Pi }^{\mathrm{n},\mathrm{E}}(\omega )}}`$ (23)
$`=`$ $`{\displaystyle \frac{i\omega }{\tau }}{\displaystyle \frac{m}{nq^2}}G.`$ (24)
This shows that the inclusion of momentum conservation leads to nothing but the local field correction with the same form $`G`$ for both cases, the inclusion of only density conservation and additional energy conservation. Formula (24) is the main result of this paper since it leads to a tremendous simplification. To see the advantages more clearly we discuss now limiting cases.
The long wave length expansion is particularly important for the classical limit and for the discussion of sum rules . Since the discussion above has shown the advantage of discussing the inverse polarization function instead of the polarization function itself we proceed and give the expansion for the inverse polarization functions (9), (11) and (13)
$`{\displaystyle \frac{1}{\mathrm{\Pi }_0}}`$ $`=`$ $`{\displaystyle \frac{m\omega ^2}{nq^2}}{\displaystyle \frac{2E}{n^2}}+o(q^2),`$ (25)
$`{\displaystyle \frac{1}{\mathrm{\Pi }^\mathrm{n}}}`$ $`=`$ $`{\displaystyle \frac{m\omega (\omega +\frac{i}{\tau })}{nq^2}}\left({\displaystyle \frac{i}{\omega \tau }}{\displaystyle \frac{1}{_\mu n}}+{\displaystyle \frac{2E}{n^2}}\right){\displaystyle \frac{\omega }{\omega +\frac{i}{\tau }}}+o(q^2),`$ (26)
$`{\displaystyle \frac{1}{\mathrm{\Pi }^{\mathrm{n},\mathrm{E}}}}`$ $`=`$ $`{\displaystyle \frac{1}{\mathrm{\Pi }_\mathrm{n}}}{\displaystyle \frac{nq^2}{18m}}\left({\displaystyle \frac{9}{_\mu n}}{\displaystyle \frac{10E}{n^2}}\right)^2{\displaystyle \frac{\omega }{(\omega +\frac{i}{\tau })^3}}+o(q^4).`$ (27)
From equations (24) it is straight forward to derive the expansions for $`\mathrm{\Pi }^{\mathrm{nj}}`$ and $`\mathrm{\Pi }^{\mathrm{njE}}`$.
The first observation is that up to zeroth order in $`q`$ the local field corrections (24) induced by momentum conservation lead to an exact cancellation of the effect of collisions in (11) since we have
$`{\displaystyle \frac{1}{\mathrm{\Pi }^\mathrm{n}}}{\displaystyle \frac{1}{\mathrm{\Pi }_0}}=G+o(q^0)`$ (29)
which shows that we have to go to the next order in $`q`$ as done in (LABEL:nn).
Also one recognizes that the inclusion of energy conservation leads only to corrections in next order of $`q^2`$ with respect to $`\mathrm{\Pi }^\mathrm{n}`$. Moreover, we observe that this correction even vanishes if we employ the zero temperature limit. For zero temperature we have $`E=3ϵ_fn/5`$ and $`_\mu n=3n/2ϵ_f`$ with the Fermi energy $`ϵ_f`$ such that
$`{\displaystyle \frac{1}{\mathrm{\Pi }^{\mathrm{njE}}}}`$ $`=`$ $`{\displaystyle \frac{1}{\mathrm{\Pi }^{\mathrm{nj}}}}+o(q^4)`$ (30)
$`=`$ $`{\displaystyle \frac{m\omega ^2}{nq^2}}{\displaystyle \frac{2ϵ_f}{15n}}{\displaystyle \frac{9\omega +\frac{5i}{\tau }}{\omega +\frac{i}{\tau }}}+o(q^4).`$ (31)
Using (LABEL:nn) one can write all the effects of correlation including conservation laws in one common local field factor
$`\stackrel{~}{G}`$ $`=`$ $`{\displaystyle \frac{1}{\mathrm{\Pi }^{\mathrm{n},\mathrm{j},\mathrm{E}}}}{\displaystyle \frac{1}{\mathrm{\Pi }_0}}`$ (32)
$`=`$ $`{\displaystyle \frac{1}{1i\omega \tau }}\left({\displaystyle \frac{1}{_\mu n}}{\displaystyle \frac{2E}{n^2}}\right)+o(q^2)`$ (33)
$`=`$ $`{\displaystyle \frac{1}{1i\omega \tau }}{\displaystyle \frac{8ϵ_f}{15n}}+o(q^4)`$ (34)
where the last line is valid for zero temperature.
This allows in turn to give the small wave vector expression of the polarization function itself in a Drude-like expression
$`\underset{𝐪0}{lim}\mathrm{\Pi }^{\mathrm{n},\mathrm{j},\mathrm{E}}(𝐪,\omega )={\displaystyle \frac{nq^2}{m\omega \left[\omega +\frac{nq^2}{m\omega }\mathrm{Re}\stackrel{~}{G}(\omega )+\frac{i}{\stackrel{~}{\tau }(\omega )}\right]}}`$ (35)
with the modified frequency–dependent relaxation rate
$`\stackrel{~}{\tau }^1={\displaystyle \frac{nq^2}{m\omega }}\mathrm{Im}\stackrel{~}{G}(\omega )`$ (37)
similar to . The advantage here is that we have simple explicit formulae for the dynamical local field factor $`\stackrel{~}{G}`$ and the modified relaxation rate while in this could only be given in static approximation and involving complicated integrals. If we had used simply the Mermin formula (11) we would have obtained $`\stackrel{~}{\tau }=\tau `$ and $`\mathrm{Re}\stackrel{~}{G}=0`$.
In particular we find for the imaginary part
$`\underset{\omega \mathrm{}}{lim}\mathrm{Im}\mathrm{\Pi }^{\mathrm{n},\mathrm{j},\mathrm{E}}(𝐪,\omega )`$ $`=`$ $`{\displaystyle \frac{n^2q^4}{\omega ^5\tau m^2}}\left({\displaystyle \frac{1}{_\mu n}}{\displaystyle \frac{2E}{n^2}}\right)`$ (38)
$`=`$ $`{\displaystyle \frac{8ϵ_fnq^4}{15\omega ^5\tau m^2}},`$ (39)
$`\underset{\omega \mathrm{}}{lim}\mathrm{Im}\mathrm{\Pi }^\mathrm{n}(𝐪,\omega )`$ $`=`$ $`{\displaystyle \frac{n^2q^4}{\omega ^3\tau m^2}}`$ (40)
showing a characteristic different high frequency behavior. While in we have checked the improved convergence of first energy weighted sum rule for the full expression (24) we want to point out that the $`\omega ^5`$ decrease for high frequencies allows to fulfill higher order sum rules. The analytical discussion and proof similar to will be devoted to a forthcoming work.
In Fig. 1 we compare the imaginary part of the polarization function for the density and momentum approximation (gray lines) with the Mermin (density) approximation (dark lines) as function of energy with the corresponding limiting cases.
First we want to discuss the corresponding complete expressions (solid lines) of the Mermin formula (11) and the formula (24) including momentum, density and energy conservation. One recognizes that the low frequency limit agrees between Mermin (density) formula and the complete formula while the high frequency limit shows the characteristic different behavior of $`\omega ^3`$ for Mermin (40) and a stronger decrease of $`\omega ^5`$ for the complete expression as have been seen in (39). The high frequency expressions according to (39) and (40) are given by corresponding dashed lines in the figure.
Let us now examine the long–wave length limits (35) of the Mermin formula (11) and the one including momentum, density and energy conservation (24) plotted in the figure as dotted lines. We see that the long wave limit of the Mermin formula approximates the high frequency behavior of the Mermin formula (11) nicely but fails for low frequencies. In contrast, the long wave length expansion of the expression including momentum conservations (35) shows an excellent agreement with the complete expression (24) for both the high and low frequency limit. Please remember that in the latter expression (35) the corrections of order $`q^2`$ drop out and it is effectively of the order $`q^4`$. The nice numerical agreement of the expression (35) with the full result (24) underlines also the force of local field corrections in constructing approximate formulae for the response functions.
The valuable comments by J. D. Frankland (GANIL) are gratefully acknowledged.
|
warning/0003/cond-mat0003224.html
|
ar5iv
|
text
|
# Electron - hole asymmetry and activation energy in quantum Hall ferromagnets
\[
## Abstract
We argue that the dissipative transport in ferromagnetic quantum Hall effect liquids at $`\nu =2N+1`$ is dominated by the thermal activation of pairs consisting of an electron and an antiskyrmion (topological texture which represents a hole with ’screened’ exchange interaction), thus manifesting the lack of electron-hole symmetry in quantum Hall ferromagnets. We find that the activation energy of such a pair is not the exchange energy, but is determined by the interplay between the excess Zeeman energy of a skyrmion and the charging energy of its topological texture.
\]
Due to the electron-electron exchange interaction, the system of two-dimensional (2D) electrons filling completely an odd-integer number of spin-polarized Landau levels (LL) form an incompressible liquid with ferromagnetic properties . The incompressibility of such a liquid is usually studied experimentally by measuring the activation energy, $``$, in the dissipative part of the conductivity tensor, $`\sigma _{xx}e^{/2T}`$. The activation energy (at, for instance, $`\nu =1`$) reflects the enhancement of a gap in the excitation spectrum of quantum Hall ferromagnets (QHF’s), as compared to the bare Zeeman energy of 2D electrons, $`\epsilon _\mathrm{Z}`$. For a long time, it was believed that the experimentally observed $``$ is nothing but the activation energy, $`\mathrm{}_0`$, of a pair consisting of a free electron and a hole at the filled spin-polarized Landau level, which is determined by the intra-LL exchange interaction ($`\mathrm{}_0=𝑑zV(\lambda \sqrt{2z})e^z\sqrt{\pi /2}(e^2/\lambda \chi )`$ for $`\nu =1`$, $`\lambda =\sqrt{\mathrm{}c/eB}`$).
However, there is growing experimental evidence that the energy required for activation of a dissipative current in QHF’s in 2D systems with a reduced value of bare Zeeman splitting is much less than $`\mathrm{}_0`$. The accumulation of this evidence is accompanied by extensive theoretical studies of skyrmions \- charge carriers whose exchange interaction with the ground state electrons is ’screened’ by a topologically stable spin polarization texture. For $`\epsilon _\mathrm{Z}\mathrm{}_0`$, the activation energy of a skyrmion - antiskyrmion pair is determined by a smaller inter-LL exchange energy ($`\mathrm{}_i=𝑑zV(\lambda \sqrt{2z})e^z\mathrm{L}_1(z)\frac{1}{2}\sqrt{\pi /2}(e^2/\lambda \chi )`$ for $`\nu =1`$), which is the result of a local twist of the spin-polarization field and requires the mixing of different Landau levels.
Nevertheless, even despite the factor of two reduction, the activation energy of a skyrmion-antiskirmion pair is too large to explain the experimentally observed tendency of a gap in quantum Hall ferromagnets with a small Zeeman energy to diminish, when $`\epsilon _\mathrm{Z}0`$. In the present paper, we propose an alternative scenario for the activation of charge carriers in QHF’s by exploiting the lack of electron-hole symmetry in such a system. Below, we argue that a dissipative current in QHF’s (in the thermodynamic limit) is provided by thermal activation of a couple consisting of one single-particle electron and an antiskyrmion (the latter represents a hole with an exchange interaction ’screened’ by a topological texture). Such an asymmetric pair has activation energy
$$=a\epsilon _\mathrm{Z}^{1/3}E_\mathrm{C}^{2/3}\mathrm{ln}^{1/3}\left(\frac{\mathrm{}_i}{E_\mathrm{C}^{2/3}\epsilon _\mathrm{Z}^{1/3}}\right),E_\mathrm{C}=\frac{e^2}{\chi \lambda },$$
(1)
(with$`a0.9`$), which is much smaller than even the inter-LL exchange energy. The same quantity, $``$ also determines the jump in the chemical potential of the 2D system across the unit filling factor. Together with the idea of the electron-’hole’ asymmetry, Eq. (1) for the transport activation energy represents the main result of this paper.
First, we would like to comment on the idea of the electron-hole asymmetry of thermodynamic properties of the QHF. When analyzing QHF’s, it is very tempting to restrict the microscopic analysis to only a couple of spin-split Landau levels (spin-up/down $`n=0`$ for $`\nu =1`$). After such a premature projection onto the lowest LL, the symmetry between a single-particle electron at the excited-spin LL and a hole at a completely filled one is artificially created, and it dominates in the further analysis making all thermodynamic properties of a liquid symmetric ad hoc with respect to the electron-hole transformation. In particular, this places the energies of an electron/hole and also of a pair of a skyrmion and antiskyrnion symmetric with respect to the chemical potential of a liquid, which has, then, to be placed half-way between the energies of equivalent spin-split carriers. However, a more cautious analysis of the skyrmionic texture energy requires the mixing of Landau levels by the gradients of the exchange field and cannot be obtained without consistently taking into account other LL’s. Taking account of the higher LL’s can be done following various roots, and it results in both spin rigidity of a polarized liquid and in a topological term in its energy which strongly distinguishes between skyrmionic excitations corresponding to the addition or subtraction of a physical electron.
The formation of a skyrmion aims to reduce the exchange energy of a charged carrier (an electron or a hole at the filled LL) by smoothly adjusting the spinor wave functions of neighboring electrons in such a way that in the locally rotated spin-coordinate system there would be one completely filled spin-polarized LL. We describet he transformation $`\left(\begin{array}{c}\phi (𝐫)\\ 0\end{array}\right)=U(𝐫)\psi (𝐫)`$ to the local spin-coordinate system, where the spin-density matrix is diagonal $`\mathrm{\Lambda }=(1+\sigma ^3)/2`$, using unitary rotations, $`U`$ from the group U(2) . After such a tranformation, the LL in the rotated spin-coordinate system have to be slightly modified by taking into account an additional non-Abelian gauge field,
$$\stackrel{}{\mathrm{\Omega }}(𝐫)=U(𝐫)\left(i\stackrel{}{}U^{}(𝐫)\right)\left(i\stackrel{}{}U(𝐫)\right)U^{}(𝐫),$$
(2)
which has all three independent spin-matrix components: $`\stackrel{}{\mathrm{\Omega }}(𝐫)=\stackrel{}{a}^\alpha \sigma ^\alpha `$ ($`\sigma ^{1,2,3}`$ stand for Pauli matrices). One of those three, defined as $`\stackrel{}{a}^3(𝐫)=\mathrm{tr}(\mathrm{\Lambda }\stackrel{}{\mathrm{\Omega }}(𝐫))`$, produces the same effect, as an additional ’magnetic field’ ,
$$b=\frac{c}{e}\left[_xa_y^3_ya_x^3\right]=i\frac{c}{e}\mathrm{tr}\left(\mathrm{\Lambda }[\mathrm{\Omega }_x,\mathrm{\Omega }_y]\right),$$
(3)
which changes the capacity to accommodate electrons in one ’gauge-transformed’ LL (as compared to a homogeneously polarized one) by $`q=\rho (𝐫)𝑑𝐫`$ real particles, where
$$\rho =\frac{\theta i}{2\pi }\mathrm{tr}\left(\mathrm{\Lambda }[\mathrm{\Omega }_x,\mathrm{\Omega }_y]\right),\mathrm{and}\theta =\frac{eB_z}{\left|eB_z\right|}.$$
(4)
The latter relation takes into account whether the additional field $`b`$ should be added, or subtracted from the external magnetic field value, which is the the first indication the ’electron’-’hole’ symmetry in the system is violated. Note that, for the reversed-spin LL in the rotated frame, the same $`\stackrel{}{\mathrm{\Omega }}(r)`$ of the same texture in $`U(𝐫)`$ produces the opposite effect, so that the total number of states available at one LL remains the same ($`\mathrm{tr}\stackrel{}{\mathrm{\Omega }}=0`$).
The texture with $`q=1`$ corresponds to one added particle and represents an electron with ’screened’ exchange interaction (skyrmion); $`q=1`$ is analogous to a ’hole’ (antiskyrmion). Both of them are topologically stable textures in the local spin-polarization field of a liquid, $`n^\alpha =\mathrm{tr}(U\sigma ^\alpha U^{}\mathrm{\Lambda })`$, which can be classified using the homotopy group $`\pi _2(\mathrm{U}(2)/[\mathrm{U}(1)\times \mathrm{U}(1)])`$, that is, using the index of mapping of a 2D plane onto a unit sphere,
$`{\displaystyle \frac{d𝐫}{8\pi }ϵ^{\beta \gamma \delta }ϵ_{zij}n^\beta _in^\gamma _jn^\delta }=\theta q.`$
The presence of a spurious gauge field $`\stackrel{}{a}^\alpha `$ also affects the energy of a liquid, due to the exchange interaction, and it results in two contributions to the energy of a twisted texture. One is generated by off-diagonal components $`\stackrel{}{a}^1`$ and $`\stackrel{}{a}^2`$ of the gauge field $`\stackrel{}{\mathrm{\Omega }}`$. Since, in the rotated coordinate system, the Landau level is full and the liquid is polarized along the axis $`𝐥_3`$, these two components enter only as perturbations in the second order, $`\mathrm{tr}\left(\mathrm{\Lambda }\stackrel{}{\mathrm{\Omega }}(1\mathrm{\Lambda })\stackrel{}{\mathrm{\Omega }}\right)`$, thus producing the spin-rigidity term in the energy. This term always increases the energy of a liquid when the texture is deformed. The other contribution comes from the diagonal (in the $`𝐥_3`$-basis) component of the non-Abelian gauge, and, as a diagonal term, it depends on the sign of the ’magnetic field’ $`b`$. In fact, what matters is whether the exchange field of a texture tends to twirl electrons in the same direction as the external field $`B_z`$, or opposite to it. Since the relative signs of $`b`$ and $`B_z`$ determine whether the texture carries extra electrons, this generates the ’electron’-’hole’ asymmetry for excitations in QHF’s and makes an additional difference between the energy shift of a polarized liquid upon adding or subtracting a real particle (electron).
In the following, we focus our attention on a single-carrier texture and study its energy using the standard root of the Hubbard-Stratonovich transformation applied to the partition function, $`𝒵=e^{\mathrm{\Phi }/T}`$ of the interacting electron gas,
$`𝒵`$ $`=`$ $`{\displaystyle 𝔇\psi 𝔇\psi ^{}\mathrm{exp}\left\{\frac{\mathrm{\Phi }_0}{T}_0^\beta 𝑑\tau E_{\mathrm{int}}\right\}}`$
$`{\displaystyle \frac{\mathrm{\Phi }_0}{T}}`$ $`=`$ $`{\displaystyle _0^\beta }𝑑\tau {\displaystyle 𝑑𝐫\psi ^{}\left(_\tau \mu +H_0\frac{\epsilon _\mathrm{Z}}{2}\sigma ^z\right)\psi }`$
$`H_0`$ $`=`$ $`{\displaystyle \frac{1}{2m}}(i\stackrel{}{}{\displaystyle \frac{e}{c}}\stackrel{}{A})^2,\beta =T^1,\mathrm{}=1,`$
$`E_{\mathrm{int}}`$ $`=`$ $`{\displaystyle \frac{1}{2}}{\displaystyle 𝑑𝐫𝑑𝐫^{}V(𝐫𝐫^{})\psi _\alpha ^{}(𝐫)\psi _\alpha (𝐫)\psi _\beta ^{}(𝐫^{})\psi _\beta (𝐫^{})},`$
where $`\mu `$ is chemical potential placed somewhere in the gap of the single-particle excitation spectrum of the system, fields $`\psi `$ obey anti-periodic boundary conditions, as a function of imaginary time $`\tau `$, and $`V>0`$. Then, we expand the thermodynamic potential, $`\mathrm{\Phi }=E\mu N`$ with respect to the non-Abelian gauge field $`\mathrm{\Omega }_i(𝐫)`$ and its derivatives. The whole procedure is equivalent to the self-consistent Hartree-Fock calculation, and the only, but essential difference of this work from what has been published earlier is that (i) we do not assume a premature projection of electron states onto the lowest Landau level, and (ii) after we find the presence of a topological term in the energy, we take it seriously.
The Hubbard-Stratonovich transformation reduces functional integration over fermionic fields to the functional integration over the exchange mean field $`Q(\tau ;𝐫,𝐫^{})`$, $`Q^{}(𝐫,𝐫^{})=Q(𝐫^{},𝐫)`$:
$`e^{{\scriptscriptstyle 𝑑\tau H_{\mathrm{int}}}}={\displaystyle DQe^{{\scriptscriptstyle 𝑑\tau 𝑑𝐫𝑑𝐫^{}F_{\mathrm{HS}}}}}`$
where
$`F_{\mathrm{HS}}`$ $`=`$ $`{\displaystyle \frac{\mathrm{}_0}{2}}\mathrm{tr}Q(\tau ;𝐫,𝐫^{})Q^{}(\tau ;𝐫,𝐫^{})`$
$`+\sqrt{\mathrm{}_0V(𝐫𝐫^{})}\psi _\alpha ^{}(𝐫)Q_{\alpha \beta }(\tau ;𝐫,𝐫^{})\psi _\beta (𝐫^{}).`$
After integrating out fermions, the thermodynamic potential $`\mathrm{\Phi }`$ can be represented in the form of
$`{\displaystyle \frac{\mathrm{\Phi }}{T}}`$ $`=`$ $`{\displaystyle _0^\beta }𝑑\tau {\displaystyle 𝑑𝐫𝑑𝐫^{}\frac{\mathrm{}_0}{2}Q(\tau ;𝐫,𝐫^{})Q^{}(\tau ;𝐫,𝐫^{})}`$
$`\mathrm{Tr}\mathrm{ln}\left\{H_0+\sqrt{\mathrm{}_0V(𝐫𝐫^{})}Q(\tau ;𝐫,𝐫^{})\right\},`$
where Tr stands for the complete trace, including spatial and imaginary time integrations of a properly ordered product of matrices . For comparison, in a homogeneously polarized liquid, without taking into account the possibility to form topological textures, thermodynamic potential is equal to
$$\mathrm{\Phi }=|\epsilon _Z|N_\mathrm{e}+\mathrm{}_0N_\mathrm{h}\mu \left[N_\mathrm{e}N_\mathrm{h}\right].$$
(5)
When studying the textured state, we use the saddle-point method to find the value of a matrix $`\stackrel{~}{Q}=U(𝐫)Q(\tau ;𝐫,𝐫^{})U^{}(𝐫^{})`$, which describes the exchange within one completely filled LL in the rotated spin-frame. The use of a saddle-point $`\stackrel{~}{Q}`$ enables us to calculate the exchange part of the thermodynamic potential of the sea of electrons,
$`{\displaystyle \frac{\mathrm{\Phi }_{\mathrm{ex}}}{T}}={\displaystyle _0^\beta }𝑑\tau {\displaystyle 𝑑𝐫𝑑𝐫^{}\frac{\mathrm{}_0}{2}\stackrel{~}{Q}(\tau ;𝐫,𝐫^{})\stackrel{~}{Q}^{}(\tau ;𝐫,𝐫^{})}`$
$`\mathrm{Tr}\mathrm{ln}\left\{_\tau \mu +H_0+H_\mathrm{\Omega }+\sqrt{\mathrm{}_0V(𝐫𝐫^{})}\stackrel{~}{Q}(\tau ;𝐫,𝐫^{})\right\}`$
for a given texture with a given number of added or missing particles, $`q`$. The non-Abelian gauge field $`\mathrm{\Omega }`$ determined by the texture is included into the perturbation term,
$`H_\mathrm{\Omega }=U(𝐫)H_0U^{}(𝐫)H_0=U(𝐫)[H_0,U^{}(𝐫)].`$
The saddle point equation for the exchange field of a full LL in the rotated spin-coordinate system can be obtained after applying the variational principle to $`\mathrm{\Phi }_{\mathrm{ex}}`$,
$`\stackrel{~}{Q}(\tau ;𝐫,𝐫^{})=\sqrt{{\displaystyle \frac{V(𝐫𝐫^{})}{\mathrm{}_0}}}G(\tau ,𝐱;\tau +0,𝐱^{}),`$
and, then, solved perturbatively by expanding the Green function, $`G\left[_\tau \mu +H_0+H_\mathrm{\Omega }+\sqrt{\mathrm{}_0V(𝐫𝐫^{})}\stackrel{~}{Q}\right]^1`$ over$`H_\mathrm{\Omega }`$. An intermediate step of such a calculation, which is completely equivalent to the Hartree-Fock procedure, can be schematically represented as
$`{\displaystyle \frac{\mathrm{\Phi }_{\mathrm{ex}}}{T}}=\mathrm{Tr}\left(H_\mathrm{\Omega }G_0\right)+{\displaystyle \frac{1}{2}}\mathrm{Tr}\left(H_\mathrm{\Omega }G_0H_\mathrm{\Omega }G_0\right)`$
$`{\displaystyle \frac{1}{2}}{\displaystyle 𝑑\tau 𝑑𝐫𝑑𝐫^{}\mathrm{tr}\left\{\left(G_0H_\mathrm{\Omega }G_0\right)_{\tau 𝐫,\tau 𝐫^{}}V(𝐫𝐫^{})\left(G_0H_\mathrm{\Omega }G_0\right)_{\tau 𝐫^{},\tau 𝐫}\right\}},`$
where $`G_0`$ is the electron Green function in a system with one completely filled homogeneously polarized LL and indices after brackets indicate the ’outer’ coordinates/time, whereas inside each product the full integration is assumed. For the sake of convenience, we shall use the basis of LL’s and inter-Landau level operators $`a_{}=(p_x\pm i\theta p_y)\lambda /\sqrt{2}`$ (where $`\stackrel{}{p}=i\stackrel{}{}e\stackrel{}{A}/c`$, and $`\theta =eB_z/\left|eB_z\right|`$ indicates the direction of cyclotron rotation of a carrier), so that
$`H_\mathrm{\Omega }=\omega _\mathrm{c}\left\{a_+\mathrm{\Omega }_{}+\mathrm{\Omega }_+a_{}+\mathrm{\Omega }_+\mathrm{\Omega }_{}\right\},`$
where $`\mathrm{\Omega }_{}=U(𝐫)[a_{},U^{}(𝐫)]=(\mathrm{\Omega }_x\pm i\theta \mathrm{\Omega }_y)\lambda /\sqrt{2}`$, and $`\omega _\mathrm{c}=|eB_z|/mc`$.
Using integral properties of Hermitian polynomials and recursion relations for Laguerre polynomials, $`\mathrm{L}_N^m`$, one finally arrives (for $`\omega _\mathrm{c}\mathrm{}`$) at
$`\mathrm{\Phi }_{\mathrm{ex}}={\displaystyle \frac{\mathrm{}_i}{2\pi \lambda ^2}}{\displaystyle 𝑑𝐫\mathrm{tr}\left(\mathrm{\Lambda }\mathrm{\Omega }_+\mathrm{\Omega }_{}\mathrm{\Lambda }\mathrm{\Omega }_+\mathrm{\Lambda }\mathrm{\Omega }_{}\right)}`$
$`={\displaystyle \frac{\mathrm{}_i}{2}}{\displaystyle \frac{d𝐫}{2\pi }\left\{\mathrm{tr}\left(\mathrm{\Lambda }\stackrel{}{\mathrm{\Omega }}(1\mathrm{\Lambda })\stackrel{}{\mathrm{\Omega }}\right)+\theta i\mathrm{tr}\left(\mathrm{\Lambda }[\mathrm{\Omega }_x,\mathrm{\Omega }_y]\right)\right\}}`$
where $`\mathrm{}_i`$ is the inter-LL exchange, which can be also calculated for larger odd-integer filling factors, $`\nu =2K+1`$, as
$`\mathrm{}_i`$ $`=`$ $`{\displaystyle _0^{\mathrm{}}}dzV(\lambda \sqrt{2z})e^z\times `$
$`\times \left\{{\displaystyle \underset{M=K,K+1}{}}M\left[\mathrm{L}_M(z)\mathrm{L}_K^1(z)\mathrm{L}_{M1}(z)\mathrm{L}_{K1}^1(z)\right]\right\}`$
For the sake of a single-charged texture analysis, we use the parametrization $`U=\left(1+|w|^2\right)^{1/2}\left(\begin{array}{cc}w& 1\\ 1& w^{}\end{array}\right)`$ of a unitary matrix ($`w`$ can be related to the polarization vector as $`w=(n^x+in^y)/(1n^z)`$), and $`\mathrm{\Phi }_{\mathrm{ex}}`$ takes the form of a conformally-invariant Belavin-Polyakov (BP) action ,
$`\mathrm{\Phi }_{\mathrm{ex}}={\displaystyle \frac{\mathrm{}_i}{2}}{\displaystyle \frac{d𝐫}{2\pi }\frac{|\stackrel{}{}w|^2+\theta i(_xw_yw^{}_yw_xw^{})}{\left(1+|w|^2\right)^2}}.`$
When the latter is taken alone, it is minimized by the classical BP skyrmion, $`w_0=\left(x+i\theta qy\right)/R`$, where $`R`$ is the texture radius, and for one texture carrying $`q=1`$, or $`q=1`$ real particles,
$$\mathrm{\Phi }_{\mathrm{ex}}=\frac{\mathrm{}_i}{2}\left[|q|+q\right].$$
(6)
We would like to emphasize that the above calculation results in only one part of the total thermodynamic potential of the electron liquid, $`\mathrm{\Phi }=E\mu N`$: namely, the exchange energy of a texture. Since the total number of particles in the liquid is changed by the texture, an extra term $`\mu q`$ has to be added to $`\mathrm{\Phi }_{\mathrm{ex}}`$ in order to obtain the actual $`\mathrm{\Phi }`$. Besides that, the ’excess’ charge density, $`\rho `$ carried by the texture with $`R\lambda `$ is formed by many electrons, so that the thermodynamic potential has to be completed by the direct Coulomb energy term ,
$`\mathrm{\Phi }_\mathrm{C}={\displaystyle \frac{e^2}{2\chi }}{\displaystyle 𝑑𝐫𝑑𝐫^{}\frac{\rho (𝐫)\rho (𝐫^{})}{|𝐫𝐫^{}|}}={\displaystyle \frac{\xi e^2}{\chi R}},\xi {\displaystyle \frac{3\pi ^2}{64}}.`$
One also has to bring forward the excess Zeeman energy of a texture,
$$\mathrm{\Phi }_\mathrm{Z}=\frac{\epsilon _\mathrm{Z}}{2}(1n^z)\frac{d𝐫}{2\pi \lambda ^2}=\frac{d𝐫}{2\pi }\frac{|\epsilon _\mathrm{Z}|/\lambda ^2}{\left(1+|w|^2\right)}.$$
(7)
Since, for the BP skyrmion solution, $`w_0`$, the integral in Eq. (7) logarithmically diverges at long distances, one has to study a modification of a topological texture by a conformally non-invariant $`\mathrm{\Phi }_\mathrm{Z}`$. To that end, we expand the space of functions $`w`$ onto $`w=(x+i\theta qy)/[Rf(r)]`$, which incorporates a smooth cut-off function $`f(r)<1`$, $`f(0)=1`$ and $`f(\mathrm{})=0`$, and has the same topological charge $`q=\pm 1`$ and nominal radius $`R`$, as $`w_0`$. We find the cut-off function $`f(r)`$ by minimizing the functional$`\mathrm{\Phi }_{\mathrm{ex}}+\mathrm{\Phi }_\mathrm{Z}`$ with respect to $`f(r)`$, which results in $`f(r>R)zK_1(z)`$, where $`z=(r/\lambda )\sqrt{2\epsilon _\mathrm{Z}/\mathrm{}_i}`$ and $`K_1`$ is the Bessel function, and gives us the minimal value for $`\mathrm{\Phi }_{\mathrm{ex}}+\mathrm{\Phi }_\mathrm{Z}`$:
$`{\displaystyle \frac{\mathrm{}_i}{2}}\left[|q|+q\right]+{\displaystyle \frac{\epsilon _\mathrm{Z}}{2}}\left({\displaystyle \frac{R}{\lambda }}\right)^2\mathrm{ln}\left({\displaystyle \frac{\eta \mathrm{}_i}{\epsilon _\mathrm{Z}R^2/\lambda ^2}}\right),`$
where $`\eta =2e^{2C}`$, $`C`$ is the Euler number.
As the last step, we find the radius, $`R`$ of a texture by optimising the sum $`\mathrm{\Phi }_{\mathrm{ex}}+\mathrm{\Phi }_\mathrm{Z}(R)+\mathrm{\Phi }_\mathrm{C}(R)`$ for a given $`q`$, so that
$`{\displaystyle \frac{R}{\lambda }}=\left({\displaystyle \frac{\xi E_\mathrm{C}}{\epsilon _\mathrm{Z}}}\right)^{1/3}\left(\mathrm{ln}{\displaystyle \frac{\mathrm{}_i\eta /\xi ^{2/3}}{E_\mathrm{C}^{2/3}\epsilon _\mathrm{Z}^{1/3}}}\right)^{1/3},`$
and we find that for a liquid with $`N_1`$ skyrmions and $`N_1`$ antiskyrmion the thermodynamic potential is equal to
$$\mathrm{\Phi }=\underset{q=\pm 1}{}\left(\frac{\mathrm{}_i}{2}\left[|q|+q\right]+|q|\mu q\right)N_q,=\frac{3\xi }{2}\frac{e^2}{\chi R}.$$
(8)
Now, we can study the energy, $`E`$ of the QHF with a small number of charged excitations of all types: $`N_\mathrm{e}`$ single-particle (un-screened) electrons, $`N_\mathrm{h}`$ ’plain’ holes, $`N_1`$ skyrmions and$`N_1`$ anti-skyrmions (screened holes) added to exactly $`\nu =2N+1`$ state under the electrical neutrality condition, $`N_\mathrm{e}+N_1=N_\mathrm{h}+N_1`$. Using expressions in Eqs. (5),(8) and the fact that $`N=N_\mathrm{e}+N_1N_\mathrm{h}N_1=0`$, we find that
$$E=\epsilon _\mathrm{Z}N_\mathrm{e}+N_1+(\mathrm{}_i+)N_1+\mathrm{}_0N_\mathrm{h}.$$
(9)
The latter equation shows that when, formally, $`<\mathrm{}_0`$ and $`R\lambda `$ (which is a stronger restriction), the activation energy of a neutral pair of charged carriers in the thermodynamic limit is equal to $``$ and is determined by the thermally activated pair consisting of an electron and an antiskyrmion. Under the approximation specified above, this activation energy depends on the inter-LL exchange energy only weakly, that is, it is almost insensitive to the extent of the electron wave function across the heterostructure or quantum well, and also to the number $`K`$ of the highest occupied Landau level in the QHF ($`\nu =2K+1`$). The same energy parameter also determines the jump of the chemical potential of a liquid upon the variation of the filling factor across the integer value. The result specified in Eq. (1) agrees with the activation transport gaps measured in heterostructures with a vanishing Zeeman splitting . It also explains a small value of $``$ observed in narrow quantum wells , where the limit $`\epsilon _\mathrm{Z}=0`$ is partly hindered by spin-orbit coupling effects . However, the energy factors which determine dynamical properties of QHF’s may differ very strongly and require a further analysis.
Authors thank R.Nicholas, A.MacDonald, N.Cooper and I.Kukushkin for discussions. This work was supported by EPSRC, NATO CLG and EU High Field ICN.
|
warning/0003/cond-mat0003143.html
|
ar5iv
|
text
|
# Shear-induced vortex decoupling in Bi2Sr2CaCu2O8 crystals
## Abstract
Simultaneous transport and magnetization studies in Bi<sub>2</sub>Sr<sub>2</sub>CaCu<sub>2</sub>O<sub>8</sub> crystals at elevated currents reveal large discrepancies, including finite resistivity at temperatures of 40K below the magnetic irreversibility line. This resistivity, measured at the top surface, is non-monotonic in temperature and extremely non-linear. The vortex velocity derived from magnetization is six orders of magnitude lower than the velocity derived from simultaneous transport measurements. The new findings are ascribed to a shear-induced decoupling, in which the pancake vortices flow only in the top few CuO<sub>2</sub> planes, and are decoupled from the pinned vortices in the rest of the crystal.
Transport measurements are one of the most common methods to study vortex dynamics in superconductors. The derived resistivity $`\rho `$ describes the vortex motion as a function of temperature, field, and the applied current. In high-temperature superconductors the situation is more complicated due to their high anisotropy and layered structure. The corresponding resistivity has two main components, the in-plane resistivity $`\rho _{ab}`$ and the out-of-plane $`\rho _c`$, with typical ratio of $`\rho _c/\rho _{ab}10^4`$ in the normal state of Bi<sub>2</sub>Sr<sub>2</sub>CaCu<sub>2</sub>O<sub>8</sub> (BSCCO) crystals . As a result, the measured resistance $`R`$ is a non-trivial function of sample geometry and contact configuration , which is further significantly complicated by the nonlinear current dependence of $`\rho _{ab}`$ and $`\rho _c`$. Yet it is generally assumed that the physical mechanism that governs the dissipation can be described in terms of current density, namely, the local values of $`\rho _{ab}`$ and $`\rho _c`$ are determined by the corresponding in-plane and out-of-plane current densities $`j_{ab}`$ and $`j_c`$ (ignoring possible nonlocal effects ). In this paper we demonstrate that in highly anisotropic materials like BSCCO this assumption may not be valid. We find that at elevated currents an additional new term, the c-axis gradient of the in-plane current $`dj_{ab}/dz`$, becomes the dominant parameter in the description of the local dissipation. This current gradient induces large velocity gradients $`dv_{ab}/dz`$ of the pancake vortices in the different CuO<sub>2</sub> planes, leading to their decoupling and to corresponding dramatic increase in $`\rho _c`$ and anisotropy $`\gamma `$. In the presence of inhomogeneous currents, this mechanism results in fundamental changes in the transport behavior, including large measurable reentrant resistance well below the magnetically determined irreversibility line (IL).
The studies were carried out on several high quality BSCCO crystals , with $`T_c90`$ K. Four wires were attached to the gold pads evaporated on freshly cleaved top ab surfaces, as shown schematically in the inset to Fig. 1a. The bottom surface of the crystals, free of electrical contacts, was attached to an array of 19 2DEG Hall sensors , $`30\times 30`$ $`\mu m^2`$ each, allowing simultaneous resistance $`R`$ and local magnetization measurements in the presence of transport current. In addition, several crystals were irradiated by very low doses of 5.8 GeV Pb ions to produce columnar defects with concentrations corresponding to matching fields of $`B_\varphi `$ = 5, 20, and 60 G. In order to focus on bulk vortex dynamics and to avoid complications due to surface barriers , most of the samples were prepared in a form of large square platelets with electrical contacts positioned far from the edges . Some of the crystals were cut into strip shape for comparison. Although detailed behavior varies from sample to sample, the qualitative features described below were observed in all the samples regardless of the surface barriers and the irradiation doses. Here we present data for as-grown crystal A of size $`1700\times 1300\times 10\mu m^3`$, which was subsequently cut into a strip, and for an irradiated crystal B $`1000\times 700\times 30\mu m^3`$ with $`B_\varphi `$ = 60 G.
Our main general observation is displayed in Fig. 1. Figure 1a shows the $`R(T)`$ of the unirradiated crystal A at various applied fields $`H_a`$ c-axis at elevated applied current of 30 mA. At lower currents, that are usually used in transport studies, $`R(T)`$ behaves monotonically with temperature as shown for comparison by the dashed line in Fig. 1a for 10 mA at 500 Oe. However, when $`I_a`$ is increased we find surprisingly a very pronounced reentrant behavior, in which $`R(T)`$ reaches a minimum at some characteristic temperature $`T_{min}`$, but then increases again, often by more than an order of magnitude, at lower $`T`$. At 100 Oe a sharp drop in $`R(T)`$ is observed at the first-order melting transition (FOT), $`T_m`$, followed
by a monotonic resistive tail in the quasi-ordered-lattice phase, as described previously . At 200 Oe the behavior is still monotonic similar to 100 Oe data. At 300 Oe, however, $`R(T)`$ decreases monotonically down to our noise level at about 55 K, but then reappears again between 46 and 34 K. This reentrant behavior is strongly pronounced at 400 Oe and up to fields above 1000 Oe. In addition, a sharp discontinuity in the derivative $`dR(T)/dT`$ is often observed at $`T_{min}`$ as shown in the inset to Fig. 1b. The irradiated samples display essentially similar properties, as shown in Fig. 1b, with a few consistent differences: (i) The resistive kink at the FOT is smeared by the low dose of columnar defects, consistent with previous studies , (ii) $`T_{min}`$ is generally shifted to higher temperatures, and (iii) the anomalous behavior appears at lower currents, as shown by the 15 mA data in Fig. 1b. The general form of $`R(T)`$ in Fig. 1 is somewhat reminiscent of the “peak effect” observed in NbSe<sub>2</sub> near $`H_{c2}`$, or at the melting transition in YBa<sub>2</sub>Cu<sub>3</sub>O<sub>7</sub>. The behavior reported here, however, is much broader and is of a fundamentally different nature.
In the unirradiated crystals the reentrant features usually appear above 25 mA, close to our upper bound of 30 mA, limited by the contact resistance. The enhanced pinning in the irradiated samples causes the anomalous behavior to begin at $`I_a`$ as low as 10 mA, allowing a more systematic study. Figures 2a and 2b thus present $`R(T)`$
of crystal B at various $`I_a`$ at 200 and 500 Oe respective- ly. At low currents $`R(T)`$ is monotonic and only weakly current dependent, and has a rather well-defined temperature $`T_{R=0}`$ at which $`R(T)`$ drops below experimental resolution. All these qualities change drastically at higher currents. The appearance of non-monotonic $`R(T)`$ is accompanied by extreme non-linearity. An increase of $`I_a`$ by 30% or less may result in enhancement of $`R`$ by orders of magnitude. Furthermore, the same current increase causes a decrease of $`T_{R=0}`$ by tens of degrees K.
In order to elucidate the origin of this anomalous behavior we have carried out simultaneous local magnetization measurements in presence of $`I_a`$ as shown in Fig. 3a for $`T=30`$ K. The results seem to be paradoxical. Finite resistivity should be present only above the magnetically measured IL, which in Fig. 3a occurs above 1600 Oe. Below the IL the resistivity should be immeasurable in standard transport measurements. This is indeed the case at low $`I_a`$. However, at elevated currents, substantial $`R`$ is measured concurrently with the hysteretic magnetization well below the IL (in Fig. 3a). Figure 3b shows the corresponding field profile $`B_z(x)`$ obtained by the Hall sensors at 400 Oe on increasing and decreasing $`H_a`$ in presence of $`I_a`$. A clear Bean profile is observed . Fitting this profile to the theoretical $`B_z(x)`$ in platelet sample results in critical current density $`J_c=2\times 10^4`$ $`A/cm^2`$, which translates into total critical current of $`I_c=4.2`$ A. Obviously, a transport current of 25 mA, which is less than 1% of $`I_c`$, cannot result in measurable resistance in the standard critical state models.
Figure 2 shows $`I_c(T)`$ determined from the Bean pro- files together with the $`R(T)`$ data. The reentrant resistance always occurs in the region where $`I_aI_c`$. We can also compare the electrical fields. In the magnetization measurements of Fig. 3 the slow sweep of $`H_a`$ of about 0.3 Oe/sec induces an average electrical field of $`4\times 10^{10}`$ V/cm in the sample, which corresponds to a very low vortex velocity of 10<sup>-4</sup> cm/sec at 400 Oe, for example. The resistively measured electrical field, induced by the simultaneous transport current, on the other hand, is $`1.4\times 10^3`$ V/cm. This translates into vortex velocity of 350 cm/sec, which is 6 orders of magnitude higher. Clearly existence of such a high vortex flow rate induced applied by concurrently $`I_a`$ should have resulted in completely reversible magnetization, in sharp contrast to the data. So paradoxically we have a situation in which a transport current, which is two orders of magnitude lower than $`I_c`$, results in vortex velocity which is 6 orders of magnitude higher than the flux creep vortex velocity at $`I_c`$.
We emphasize that we do not observe any heating effects as confirmed by monitoring the location of the resistive transitions at $`T_c`$ and $`T_m`$ at various currents. Furthermore, and more importantly, since the two measurements are carried out simultaneously, the transport and the magnetization results should be affected by heating to the same extent, and therefore heating effects can be ruled out as a possible explanation of the observed large discrepancy between the two measurements.
We describe the observed phenomena in terms of shear-induced decoupling and shear-enhanced anisotropy. At elevated temperatures in the vortex-liquid phase the in-plane current density $`j_{ab}(z)`$ decreases approximately exponentially from the top surface to the bottom . In samples of typical dimensions the current at the bottom is about four orders of magnitude lower than at the top, due to high anisotropy $`\gamma `$ of BSCCO. As $`T`$ is decreased the pinning becomes more effective. At some temperature the current density at the bottom becomes comparable or lower than $`J_c`$. As a result the vortices at the bottom stop moving or reduce their velocity substantially, whereas the vortices at the top maintain their high velocity since the current density there is significantly above $`J_c`$. As a result, the velocity gradient between the planes $`dv_{ab}/dz`$ is increased. The thickness of the vortex pinned layer grows as $`T`$ is decreased resulting in progressively larger velocity gradients within the vortex mobile part at the top. The enhanced $`dv_{ab}/dz`$ results, in turn, in shear-induced phase slippage between the adjacent CuO<sub>2</sub> planes reducing the Josephson coupling and leading to decoupling of the planes . As a result the $`\rho _c`$, and hence the measured $`R`$, are increased significantly causing the reentrant $`R(T)`$. This process is highly nonlinear, since larger $`I_a`$ results in larger $`dj_{ab}/dz`$, causing an increase in $`\rho _c`$ and $`\gamma `$, which lead in turn to even shallower current profile and thus extremely nonlinear $`R(I)`$. Since the onset of the enhanced shear is induced by pinning, low dose irradiation causes the anomalous behavior to set in at higher $`T`$ and lower $`I_a`$ as seen in Fig. 1. It is important to realize that the enhanced $`\rho _c`$ prevents the lower vortex pinned region from effectively shunting the current. In the case of full decoupling, for example, $`\rho _c`$ becomes comparable to the normal state $`\rho _c`$, which in BSCCO crystals of typical dimensions results in about $`10^4`$ $`\mathrm{\Omega }`$ resistance between two adjacent CuO<sub>2</sub> planes. Thus the serial c-axis resistance of several decoupled layers in the upper part of the sample can be of the order of $`10^3`$ $`\mathrm{\Omega }`$, thus effectively insulating the lower zero-resistance region. Any velocity gradient $`dv_{ab}/dz`$ should, in principle, completely decouple the layers, since the time-averaged Josephson phase difference reduces to zero. More detailed analysis , however, shows that the pancake shear occurs through a ‘stick - slip’ process, which retains a finite Josephson coupling at low $`dv_{ab}/dz`$. Thus the Josephson current between the layers can be sustained at low $`dv_{ab}/dz`$, and is lost at higher rates of pancake shear.
The described mechanism is conceptually different from conventional electrodynamics in which the dissipation is determined only by the local vortex velocity, and not by the velocity gradient. The underlying physics is also very different from the process of nonlocal resistivity , in which a force applied to a pancake vortex results in a nonlocal drag of pancakes in adjacent layers. This drag, which is proportional to $`d^2v_{ab}/dz^2`$, induces a more uniform vortex flow, in contrast to the described shear-induced decoupling which has an opposite effect of amplification of the velocity differences.
Since in most of the crystal thickness the vortices are
pinned, Bean profiles are obtained in the bulk upon sweeping $`H_a`$ (Fig. 3b). At the same time the transport current confined to the top planes is sufficient to create vortex flow in these planes, resulting in a measurable voltage drop, and explaining the apparent discrepancy between the transport and magnetization data. A similar mechanism probably occurs when the c-axis properties are probed by transverse ac susceptibility and transport in the zero-field-cooled state as reported previously . Our model also explains the hysteretic $`R(H)`$ in Fig. 3a. The field $`B`$ in the top planes is determined by the hysteretic Bean profile in the rest of the crystal. As a result, the measured $`R(H_a)`$ displays hysteresis, even though the vortices are mobile in the top planes. Consequently, this apparent hysteresis practically disappears when $`R`$ is plotted vs. the average field $`B`$ measured by the sensors, rather than vs. $`H_a`$.
Figure 4 summarizes the $`BT`$ diagram of sample A at 30 mA. A similar diagram is obtained for the irradiated crystal B. The resistance vanishes within our resolution at $`T_{R=0}`$, the position of which is highly current-dependent. The resistively determined FOT extends to fields of about 500 Oe in this sample. The anomalous reentrant $`R(T)`$ is observed in the broad region between $`T_{R=0}`$ and $`T_{min}`$, where vortices are pinned in the bulk. It is interesting to note that the decoupling process occurs first in the entangled vortex-solid phase above 500 Oe, where vortex cutting is apparently relatively easy. As $`I_a`$ is increased, the shear-induced decoupling expands also into the quasi-ordered-lattice phase below $`T_m`$ as is the case in Fig. 4. The non-linearities expand also partially into the liquid phase above $`T_{min}`$ and $`T_m`$, where the shear process can still modify $`\rho _c`$ at higher currents.
One can evaluate the number of the CuO<sub>2</sub> planes participating in the flow of vortex pancakes. In Fig. 2b at 30 K, for example, the total critical current determined from the Bean profiles is $`I_c=3.7A`$. This translates into critical current of about 0.18 mA per (double) CuO<sub>2</sub> plane of the sample. Thus, the top 140 planes would be sufficient to carry the entire $`I_a`$=25mA without observable dissipation. In order to obtain the large measured resistance the actual current in the planes has to be significantly above 0.18 mA. Furthermore, the lower part of the sample also carries a substantial portion of $`I_a`$. We therefore conclude that only a few tens of CuO<sub>2</sub> planes at the top of the crystal participate in the described process. Current-induced decoupling was recently analyzed theoretically in absence of Josephson coupling . It was found that above some critical value of the current applied to the top plane, slippage between magnetically coupled planes occurs, leading to an abrupt increase in the vortex velocity in the top plane. Since in this model $`\rho _c`$ is infinite, there is no corresponding current redistribution. Computer simulations of Josephson-coupled planes also show sharp current-induced decoupling. Vortex cutting was also observed in YBCO crystals at elevated currents in flux transformer measurements . The much lower anisotropy of YBCO, however, causes only a small change in the transport properties. In BSCCO in contrast the shear-induced decoupling leads to variations by orders of magnitude in the observed behavior.
In conclusion, we reveal large apparent discrepancies between simultaneously measured magnetization and transport properties, which reveal an important feature of vortex dynamics in the presence of inhomogeneous currents. In layered superconductors at elevated currents the c-axis resistivity becomes a strong function of the local current gradient $`dj_{ab}/dz`$, instead of being determined just by the current density. This mechanism results in a highly nonlinear reentrant resistance well below the irreversibility line.
This work was supported by Israel Ministry of Science, by US-Israel Binational Science Foundation (BSF), by CREST, and by the Grant-in-Aid for Scientific Research from the Ministry of Education, Science, Sports, and Culture, Japan.
|
warning/0003/hep-lat0003026.html
|
ar5iv
|
text
|
# One loop renormalisation of Lattice NRQCD currents for semileptonic 𝐵→𝐷^(∗) decays to order 𝑝⃗/𝑀
## I Introduction
Forthcoming $`B`$ physics experiments such as BaBar, Belle, CLEO III and LHC-b will greatly improve the accuracy with which we know the rates of different decay modes of the B meson, allowing improved determinations of the CKM matrix elements involving the $`b`$ quark. The exclusive decay $`\overline{B}Dl\overline{\nu }`$ gives access to $`V_{cb}`$. For example, at BaBar the decay amplitude at zero recoil will be determined with a statistical error of less than 2% and with similar level of systematic errors, a factor of three improvement over CLEO and LEP data. In order to extract $`V_{cb}`$ we must divide out the form factor at zero recoil, introducing associated theoretical uncertainties. The form factors for $`BD`$ decays are calculable within lattice QCD <sup>*</sup><sup>*</sup>* There are a number of recent reviews of the status of lattice calculations , and it will be highly challenging for the community to produce model independent predictions of the form factors with sufficient statistical and systematic precision that theoretical uncertainty matches the improved experimental data.
The decay amplitudes can be factorised into hadronic form factors and perturbative electroweak terms as follows:
$$M(BD^{()}l^{}\overline{\nu })=\frac{iG_F}{\sqrt{2}}V_{cb}\overline{u}_l\gamma _\mu (1\gamma _5)v_\nu H^{\mu ^{()}},$$
(1)
where
$$H^\mu =D(p^{})|V^\mu |B(p)=f_+(q^2)(p+p^{})^\mu +f_{}(q^2)(pp^{})^\mu ,$$
(2)
$$H^\mu ^{}=D^{}(p^{},ϵ)|V^\mu A^\mu |B(p)=\left\{\begin{array}{c}\frac{2iϵ^{\mu \nu \alpha \beta }}{M_B+M_D^{}}ϵ_\nu ^{}p_\alpha ^{}p_\beta V(q^2)\\ (M_B+M_D^{})ϵ^\mu A_1(q^2)\\ +\frac{ϵ^{}q}{M_B+M_D^{}}(p+p^{})^\mu A_2(q^2)\end{array}\right\}.$$
(3)
These are often reparametrised in terms of form factors in the heavy quark effective theory,
$$D(p^{})|V^\mu |B(p)=\sqrt{M_BM_D}\left\{h_+(vv^{})(v+v^{})^\mu +h_{}(vv^{})(vv^{})^\mu \right\}$$
(4)
$$D^{}(p^{},ϵ)|V^\mu A^\mu |B(p)=\sqrt{M_BM_D^{}}\left\{\begin{array}{c}iϵ^{\mu \nu \alpha \beta }ϵ_\nu ^{}v_\alpha ^{}v_\beta h_V(vv^{})\\ ϵ^\mu (vv^{}+1)h_{A_1}(vv^{})\\ +ϵ^{}vv^\mu h_{A_2}(vv^{})\\ +ϵ^{}vv^\mu h_{A_3}(vv^{})\end{array}\right\}$$
(5)
The weak (vector and axial) currents in a lattice-regulated theory differ from the continuum $`\overline{MS}`$ currents via their different ultra-violet cut-offs. A correction between the results in the two regularisation schemes is therefore required in order to use simulated lattice hadronic currents in the more commonly quoted continuum $`\overline{MS}`$ scheme. Lattice QCD and $`\overline{MS}`$ have the same infra-red behaviour, so that these correction factors should be dominated by the differing ultra-violet behaviour and one expects that these corrections are calculable within perturbation theory. Knowledge of these renormalisation constants enables the various elements of the Cabibbo-Kobayashi-Maskawa weak mixing matrix to be probed using experimental decay rates and lattice calculations of hadronic matrix elements in conjunction with standard electroweak perturbation theory.
This paper aims to make possible the accurate determination of $`BD`$ matrix elements by showing how to construct the $`\overline{MS}`$ currents when using the $`O(\frac{1}{M})`$ lattice NRQCD action to simulate both the $`b`$ and $`c`$ quarks. We perturbatively match the corrections to the weak currents between on-shell quark states in lattice NRQCD to the same current in continuum $`\overline{MS}`$ with fully relativistic quarks, working consistently to $`O(\frac{1}{M})`$ throughout. The paper will be constructed as follows: in Section II we will present the one loop correction to the continuum current evaluated between an incoming $`b`$ quark and outgoing $`c`$ quark, retaining all terms to order $`\frac{\stackrel{}{p}}{M}`$. This is matched to continuum Euclidean NRQCD to $`O(\frac{1}{M})`$, giving a basis of continuum NRQCD operators for each of the currents $`V_0`$, $`V_k`$, $`A_0`$, $`A_k`$. In Section III we select a basis of operators in the Lattice theory and define the one-loop mixing matrix, giving the projection of each element of a basis for a given Lattice current at one loop on the basis of continuum NRQCD operators. Finally, we use these results to construct relativistic $`\overline{MS}`$ currents to this order from lattice operators in Section IV.
## II Continuum Calculation
We require the continuum renormalisation of the Vector and Axial currents involving an incoming $`b`$ quark with momentum $`p`$ and outgoing $`c`$ quark with momentum $`p^{}`$, i.e. $`\overline{u}_c(p^{})\mathrm{\Gamma }u_b(p)`$, where either $`\mathrm{\Gamma }=\gamma _\mu `$ or $`\mathrm{\Gamma }=\gamma _\mu \gamma _5`$. At one loop the graphs in Figures 1 to 3 contribute. The calculation was performed in Minkowski space using dimensional regularisation in Feynman gauge, the on-shell $`\overline{MS}`$ scheme, and having introduced a fictitious gluon mass $`\lambda `$ to regulate infra-red divergences. The treatment allowed for non-zero external momenta, keeping terms linear in $`\frac{\stackrel{}{p}}{M}`$ and $`\frac{\stackrel{}{p}^{}}{M}`$.
In general, we write the perturbative expansion of some quantity, $`Q`$, as
$$Q=\underset{k=0}{\overset{\mathrm{}}{}}Q^{[k]}\alpha _s^k.$$
(6)
We write the Feynman graph in Figure 1 as $`\alpha _s\mathrm{\Lambda }^{\mathrm{\Gamma }^{[1]}}`$ and the leg corrections in Figures 2 and 3 as $`\frac{1}{2}\alpha _s\mathrm{\Gamma }Z_{\psi _b}^{[1]}`$ and $`\frac{1}{2}\alpha _s\mathrm{\Gamma }Z_{\psi _c}^{[1]}`$ respectively. Dropping the external spinors, the total current correction may be written as
$$\mathrm{\Delta }\mathrm{\Gamma }=\alpha _s\left[\mathrm{\Lambda }^{\mathrm{\Gamma }^{[1]}}+\frac{1}{2}\mathrm{\Gamma }Z_{\psi _c}^{[1]}+\frac{1}{2}Z_{\psi _b}^{[1]}\right],$$
(7)
Here both $`\mathrm{\Lambda }^{\mathrm{\Gamma }^{[1]}}`$, and $`Z_\psi `$ contain (cancelling) infra-red divergences of modulus $`\frac{\alpha _s}{3\pi }4\mathrm{log}\frac{\lambda }{M}`$. The result is IR-finite, and we obtain that the $`\overline{MS}`$ current to one loop and $`O(\frac{\stackrel{}{p}}{M})`$ is as follows It is worth noting here that at higher orders in the $`\frac{\stackrel{}{p}}{M}`$expansion the coefficients $`C_i`$ become functions of the recoil $`\omega =\frac{pp^{}}{m_cm_b}`$.
$$\mathrm{\Delta }\mathrm{\Gamma }\stackrel{~}{\mathrm{\Gamma }}=C_1\gamma ^\mu +C_2\frac{p^\mu }{m_b}+C_3\frac{p^\mu }{m_c}$$
(8)
where,
$`C_1`$ $`=`$ $`{\displaystyle \frac{\alpha _s}{3\pi }}\left[3{\displaystyle \frac{1+\xi }{1\xi }}\mathrm{log}\xi +2𝒞6\right]`$ (9)
$`C_2`$ $`=`$ $`{\displaystyle \frac{\alpha _s}{3\pi }}\left[{\displaystyle \frac{4}{(1\xi )^2}}+2\xi \mathrm{log}\xi {\displaystyle \frac{\xi 3}{(1\xi )^3}}+𝒞\left\{2{\displaystyle \frac{1+\xi }{(1\xi )^2}}+{\displaystyle \frac{4\xi \mathrm{log}\xi }{(1\xi )^3}}\right\}\right]`$ (10)
$`C_3`$ $`=`$ $`{\displaystyle \frac{\alpha _s}{3\pi }}\left[2\xi {\displaystyle \frac{1+\xi }{(1\xi )^2}}+4\xi ^2{\displaystyle \frac{\mathrm{log}\xi }{(1\xi )^3}}+𝒞\left\{{\displaystyle \frac{4\xi ^2}{(1\xi )^2}}+2\xi \mathrm{log}\xi {\displaystyle \frac{13\xi }{(1\xi )^3}}\right\}\right]`$ (11)
and
$$\begin{array}{ccccc}\xi =\frac{m_c}{m_b}& ,& 𝒞=\{\begin{array}{cc}1& ;\mathrm{\Gamma }=\gamma ^\mu \hfill \\ 1& ;\mathrm{\Gamma }=\gamma ^\mu \gamma _5\hfill \end{array}& ,& \stackrel{~}{\mathrm{\Gamma }}=\{\begin{array}{cc}1& ;\mathrm{\Gamma }=\gamma ^\mu \hfill \\ \gamma _5& ;\mathrm{\Gamma }=\gamma ^\mu \gamma _5\hfill \end{array}\end{array}$$
(12)
The Gordon decomposition of the spinor gives that, for $`m_c=m_b`$,
$$\overline{u}\gamma ^\mu u=\frac{(p+p^{})^\mu }{2m}\overline{u}u$$
(13)
and so the Vector current between degenerate quarks is, of course, not renormalised.
In order to match these currents with Euclidean lattice NRQCD currents we first analytically continue this result to Euclidean space, then apply the Foldy-Wouthuysen expansion of the external spinors to the appropriate order, taking care with the Euclideanised $`\gamma `$-matrices.
### A Analytic Continuation
We define the Euclideanisation as follows. Denoting the Minkowski gamma matrices and four vectors with a caret, $`\widehat{\gamma }`$, so that $`\{\widehat{\gamma }_\mu ,\widehat{\gamma }_\nu \}=2g_{\mu \nu }`$, we use the Dirac-Pauli representation:
$$\begin{array}{ccccc}\widehat{\gamma }^0& =& \widehat{\gamma }_0& =& \left(\begin{array}{cc}I& 0\\ 0& I\end{array}\right)\\ \widehat{\gamma }^j& =& \widehat{\gamma }_j& =& \left(\begin{array}{cc}0& \sigma _j\\ \sigma _j& 0\end{array}\right)\end{array}$$
(14)
We also define the Euclidean four vectors in terms of the contravariant and covariant Minkowski vectors as:
$$\begin{array}{ccccccc}x^0& =& x_0& =& i\widehat{x}^0& =& i\widehat{x}_0\\ x^j& =& x_j& =& \widehat{x}^j& =& \widehat{x}_j\end{array}$$
(15)
We introduce Euclidean $`\gamma `$ matrices such that $`\{\gamma _\mu ,\gamma _\nu \}=2\delta _{\mu \nu }`$
$$\begin{array}{ccccccc}\gamma ^0& =& \gamma _0& =& \widehat{\gamma }^0& =& \widehat{\gamma }_0\\ \gamma ^j& =& \gamma _j& =& i\widehat{\gamma }^j& =& i\widehat{\gamma }_j\\ \gamma _5& =& \widehat{\gamma }_5& & & & \end{array}$$
(16)
For consistent notation in Euclidean space, we define Euclidean bilinear currents in terms of the Euclidean $`\gamma `$ matrices, so that
$$\begin{array}{ccccccc}A^0& =& A_0& =& \widehat{A}^0& =& \widehat{A}_0\\ A^j& =& A_j& =& i\widehat{A}^j& =& i\widehat{A}_j\\ V^0& =& V_0& =& \widehat{V}^0& =& \widehat{V}_0\\ V^j& =& V_j& =& i\widehat{V}^j& =& i\widehat{V}_j\end{array}$$
(17)
Dropping terms in $`A_k`$ which are subsequently of higher order in $`\frac{\stackrel{}{p}}{M}`$, and suppressing the spinors, this gives us in Euclidean space:
$`\mathrm{\Delta }V^0`$ $`=`$ $`\gamma ^0{\displaystyle \frac{\alpha _s}{3\pi }}\left[3{\displaystyle \frac{1+\xi }{1\xi }}\mathrm{log}\xi 4\right]2{\displaystyle \frac{\alpha _s}{3\pi }}`$ (18)
$`\mathrm{\Delta }V^k`$ $`=`$ $`\gamma ^k{\displaystyle \frac{\alpha _s}{3\pi }}\left[3{\displaystyle \frac{1+\xi }{1\xi }}\mathrm{log}\xi 4\right]i{\displaystyle \frac{p^k}{m_b}}{\displaystyle \frac{\alpha _s}{3\pi }}\left[{\displaystyle \frac{2}{1\xi }}{\displaystyle \frac{2\xi \mathrm{log}\xi }{(1\xi )^2}}\right]i{\displaystyle \frac{p^k}{m_c}}{\displaystyle \frac{\alpha _s}{3\pi }}\left[{\displaystyle \frac{2\xi }{1\xi }}+{\displaystyle \frac{2\xi \mathrm{log}\xi }{(1\xi )^2}}\right]`$ (19)
$`\mathrm{\Delta }A^0`$ $`=`$ $`\gamma ^0\gamma _5{\displaystyle \frac{\alpha _s}{3\pi }}\left[3{\displaystyle \frac{1+\xi }{1\xi }}\mathrm{log}\xi 8\right]+\gamma _5{\displaystyle \frac{\alpha _s}{3\pi }}\left[6{\displaystyle \frac{1+\xi }{1\xi }}12{\displaystyle \frac{\xi \mathrm{log}\xi }{(1\xi )^2}}\right]`$ (20)
$`\mathrm{\Delta }A^k`$ $`=`$ $`\gamma ^k\gamma _5{\displaystyle \frac{\alpha _s}{3\pi }}\left[3{\displaystyle \frac{1+\xi }{1\xi }}\mathrm{log}\xi 8\right].`$ (21)
### B Non-Relativistic Expansion
We now perform the Foldy-Wouthuysen expansion of the external spinors $`u_h`$ to order $`\frac{1}{m}`$ in terms of the Euclidean $`\gamma `$ matrices to produce two component spinors $`U_Q`$:
$$u_h(p)=\left[1\frac{i}{2m}\stackrel{}{\gamma }\stackrel{}{p}\right]u_Q$$
(22)
where
$$u_Q=\left[\begin{array}{c}U_Q\\ 0\end{array}\right]$$
(23)
and $`U_Q`$ is a two component spinor, and $`m`$ is the pole mass of the quark arising from the on-shell condition (more formally, $`m`$ is the one loop pole mass whenever it appears at tree level since we are working consistently to order one loop). Inserting this in the results for the continuum renormalisation and keeping terms to order $`\frac{1}{M}`$ results in the following mixed continuum currents, with the relevant two component spinors suppressed.
$`V^0`$ $`=`$ $`1_{\mathrm{pauli}}\left[1+{\displaystyle \frac{\alpha _s}{3\pi }}\left(3{\displaystyle \frac{1+\xi }{1\xi }}\mathrm{log}\xi 6\right)\right]`$ (24)
$`V^k`$ $`=`$ $`\left[i\sigma _k{\displaystyle \frac{\sigma p}{2m_b}}{\displaystyle \frac{\sigma p^{}}{2m_c}}i\sigma _k\right]\left[1+{\displaystyle \frac{\alpha _s}{3\pi }}\left(3{\displaystyle \frac{1+\xi }{1\xi }}\mathrm{log}\xi 4\right)\right]`$ (27)
$`i{\displaystyle \frac{p_k}{m_b}}{\displaystyle \frac{\alpha _s}{3\pi }}\left[{\displaystyle \frac{2}{1\xi }}{\displaystyle \frac{2\xi \mathrm{log}\xi }{(1\xi )^2}}\right]`$
$`i{\displaystyle \frac{p_k^{}}{m_c}}{\displaystyle \frac{\alpha _s}{3\pi }}\left[{\displaystyle \frac{2\xi }{1\xi }}+{\displaystyle \frac{2\xi \mathrm{log}\xi }{(1\xi )^2}}\right]`$
$`A^0`$ $`=`$ $`\left[{\displaystyle \frac{\sigma p}{2m_b}}+{\displaystyle \frac{\sigma p^{}}{2m_c}}\right]\left[1+{\displaystyle \frac{\alpha _s}{3\pi }}\left(3{\displaystyle \frac{1+\xi }{1\xi }}\mathrm{log}\xi 8\right)\right]`$ (29)
$`+\left[{\displaystyle \frac{\sigma p}{2m_b}}{\displaystyle \frac{\sigma p^{}}{2m_c}}\right]{\displaystyle \frac{\alpha _s}{3\pi }}\left[6{\displaystyle \frac{1+\xi }{1\xi }}12{\displaystyle \frac{\xi \mathrm{log}\xi }{(1\xi )^2}}\right]`$
$`A_k`$ $`=`$ $`i\sigma _k\left[1+{\displaystyle \frac{\alpha _s}{3\pi }}\left(3{\displaystyle \frac{1+\xi }{1\xi }}\mathrm{log}\xi 8\right)\right]`$ (30)
We therefore define continuum Euclidean operators $`\mathrm{\Omega }_i^J`$ in Table I, to which we wish to match a corresponding set of lattice operators, and use the notation
$$c(p^{})|J^{\overline{MS}}|b(p)^{1\mathrm{loop}}=\underset{\alpha }{}Z_\alpha ^{\overline{MS}}U_c^{}\mathrm{\Omega }_\alpha ^JU_b.$$
(31)
The vector current is a conserved current for $`m_b=m_c`$ and, for this reason, the continuum renormalisation ( factor in square brackets in (24) ) is unity in the $`\xi 1`$ limit. The operator corresponding to the temporal component of the vector current, $`1_{\mathrm{pauli}}`$, is a conserved current of both the lattice NRQCD and HQET actions, and we shall see that the corresponding lattice renormalisation is also trivial in the $`m_b=m_c`$ limit.
Unlike the lattice correction, the pure HQET contribution to the vector current renormalisation at zero recoil remains trivial with non-identical quark masses at leading order in $`\frac{1}{m_b}`$ and $`\frac{1}{m_c}`$. As a result the above expressions (24) and (30) are the same as the renormalisation at zero recoil for the continuum heavy quark effective theory, commonly referred to as $`\eta _V`$ and $`\eta _A`$ respectively . This is somewhat fortuitous since it will allow us, in section V, to use the existing result for the BLM scale $`\mu `$ for the processes contributing to expressions (24) and (30), with $`\mu _V0.92\sqrt{m_bm_c}`$ and $`\mu _A0.51\sqrt{m_bm_c}`$.
## III Lattice Calculation
### A Lattice action, currents, and Feynman rules
We use the usual Wilson plaquette action defined in terms of the usual SU(3) gauge links $`U_\mu (x)`$, and the tadpole improved $`O(\frac{1}{M})`$ lattice NRQCD action ,
$$\begin{array}{c}a_{\mathrm{NRQCD}}=\psi ^{}(x)\psi (x)\psi ^{}(x+a\widehat{t})\left(1\frac{a\delta H}{2}\right)\left(1\frac{aH_0}{2n}\right)^n\frac{U_t^{}(x)}{u_0}\left(1\frac{aH_0}{2n}\right)^n\left(1\frac{a\delta H}{2}\right)\psi (x)\end{array}$$
(32)
where
$`H_0`$ $`=`$ $`{\displaystyle \frac{\mathrm{\Delta }^2}{2M_0}},`$ (33)
$`\delta H`$ $`=`$ $`c_B{\displaystyle \frac{g}{2M_0}}\sigma B,`$ (34)
$$\mathrm{\Delta }^2\chi (x)=\frac{1}{u_0}\underset{i=x,y,z}{}\left(U_i(x)\chi (x+\widehat{i})+U_i^{}(x\widehat{i})\chi (x\widehat{i})2u_0\chi (x)\right),$$
(35)
and
$`B_i(x)`$ $`=`$ $`{\displaystyle \frac{1}{2}}ϵ_{ijk}\left\{_{jk}(x){\displaystyle \frac{1}{3}}\mathrm{Tr}_{jk}(x)\right\},`$ (36)
$`_{\mu \nu }(x)`$ $`=`$ $`{\displaystyle \frac{1}{4ag^2}}{\displaystyle \underset{\alpha =\pm \mu ,\beta =\pm \nu }{}}{\displaystyle \frac{\left[P_{\alpha \beta }(x)P_{\alpha \beta }^{}(x)\right]}{2i}},`$ (37)
$`P_{\alpha \beta }(x)`$ $`=`$ $`{\displaystyle \frac{1}{u_0^4}}U_\alpha (x)U_\beta (x+\widehat{\alpha })U_\alpha ^{}(x+\widehat{\beta })U_\beta ^{}(x).`$ (38)
In the final results quoted in this paper, $`u_0`$ is taken from the plaquette and to one loop order in perturbation theory $`u_0^{[1]}=\frac{\pi }{3}`$. However, sufficient information will presented to allow a reader to change these results for different values of $`u_0`$. To this order in $`\alpha _s`$ any dynamical quarks present in the simulation do not appear in the perturbation theory. This calculation is therefore applicable to both quenched and dynamical simulations. It is well known how to perform perturbative calculations of the renormalisation of weak currents in NRQCD , and the Feynman rules in the Euclidean lattice theory used here are the same as those published by Morningstar and Shigemitsu . This paper completes a programme of work previously discussed at two lattice conferences .
Inspection of the continuum currents suggests the basis for the lattice current operators given in Table II. The momentum dependent operators are reproduced on the lattice with gauge covariant derivatives,
$$_\mu \chi (x)=\frac{1}{2u_0}\left[U_\mu (x)\chi (x+\widehat{\mu })U_\mu ^{}(x\widehat{\mu })\chi (x\widehat{\mu })\right]$$
(39)
which introduces additional quark gluon vertices that are not present in the continuum theory, as well as a tadpole improvement counter term. For convenience in simulations, we define the lattice operators in terms of the bare mass, $`M_q`$. Since we require that the one-loop renormalised kinetic mass in NRQCD matches the one-loop renormalised mass in the relativistic theory Lorentz invariance implies that the kinetic and pole masses are consistent in the relativistic theory, we must introduce additional $`Z_m`$ factors for lattice currents wherever the bare mass $`M_q`$ appears at tree level. The mass ratio $`\xi `$ appears in the continuum coefficients only at order $`\alpha _s`$, and hence the bare ratio $`\frac{M_c}{M_b}`$ can be used to this order.
We have additional Feynman rules associated with the current operators involving derivatives (and hence gauge links), giving both one and two gluon couplings. The conventions for momenta involved in current insertions of an operator $`𝒪`$ both at leading order, and couplings to both 1 and 2 gluons of polarisation $`j`$ are as given in figure 4.
The Feynman rules associated with the spatial vector and temporal axial currents at the vertex are simply the insertion of the appropriate Pauli structure. The Feynman rules associated with the spatial vector current insertions are as follows
$`𝒪_1^{V_k}(\mathrm{tree})`$ $`=`$ $`{\displaystyle \frac{i}{2M_b}}{\displaystyle \underset{j}{}}\mathrm{sin}k_j\sigma _k\sigma _j`$ (40)
$`𝒪_1^{V_k}(1\mathrm{gluon})`$ $`=`$ $`{\displaystyle \frac{ig}{2M_b}}\mathrm{cos}\left[k_j{\displaystyle \frac{1}{2}}q_j\right]\sigma _k\sigma _j`$ (41)
$`𝒪_1^{V_k}(2\mathrm{gluon})`$ $`=`$ $`{\displaystyle \frac{ig^2}{4M_b}}\mathrm{sin}\left[k_j{\displaystyle \frac{1}{2}}(q_{1_j}+q_{2_j})\right]\sigma _k\sigma _j`$ (42)
$`𝒪_3^{V_k}(\mathrm{tree})`$ $`=`$ $`{\displaystyle \frac{i}{M_b}}\mathrm{sin}k_k`$ (43)
$`𝒪_3^{V_k}(1\mathrm{gluon})`$ $`=`$ $`{\displaystyle \frac{ig}{M_b}}\mathrm{cos}\left[k_j{\displaystyle \frac{1}{2}}q_j\right]\delta _{jk}`$ (44)
$`𝒪_3^{V_k}(2\mathrm{gluon})`$ $`=`$ $`{\displaystyle \frac{ig^2}{2M_b}}\mathrm{sin}\left[k_j{\displaystyle \frac{1}{2}}(q_{1_j}+q_{2_j})\right]\delta _{jk}`$ (45)
where the 1 and 2 gluon couplings are to gluons of polarisation $`j`$.
The rules associated with the leftward derivative currents are both the obvious analogues of the above and not in fact required since the diagrams involving the second current may be switched for those of the first current with the masses swapped and momenta reversed, halving the required coding effort. The Feynman rules associated with the first temporal axial lattice current are as follows,
$`𝒪_1^{A_0}(\mathrm{tree})`$ $`=`$ $`{\displaystyle \frac{1}{2M_b}}{\displaystyle \underset{j}{}}\mathrm{sin}k_j\sigma _j`$ (46)
$`𝒪_1^{A_0}(1\mathrm{gluon})`$ $`=`$ $`{\displaystyle \frac{g}{2M_b}}\mathrm{cos}\left(k_j{\displaystyle \frac{1}{2}}q_j\right)\sigma _j`$ (47)
$`𝒪_1^{A_0}(2\mathrm{gluon})`$ $`=`$ $`{\displaystyle \frac{g^2}{4M_b}}\mathrm{sin}\left(k_j{\displaystyle \frac{1}{2}}(q_{1j}+q_{2j})\right)\sigma _j`$ (48)
while, as above, the rule for the second current is not required.
### B Mixing Matrix
We define the 1-loop mixing matrix of the lattice theory for each of the four currents $`J\{V_0,V_k,A_0,A_k\}`$ as follows:
$$c(p^{})|𝒪_\alpha ^J|b(p)^{1\mathrm{loop}}=\underset{\beta }{}_{\alpha \beta }^JU_c^{}\mathrm{\Omega }_\beta ^JU_b,$$
(49)
where the range of the summation over $`\beta `$ depends on the dimension of the basis for each continuum current. To avoid ambiguity, we shall denote the symbols for lattice Feynman diagrams in bold font to differentiate them from symbols used in the continuum calculation (although the different symbols do not in fact appear in the same equations). For both the temporal vector and spatial axial currents the mixing matrix is rank one, and we have (where $`J`$ is either $`V_0`$ or $`A_k`$):
$$^J=1+\alpha _s\left[\frac{1}{2}\{𝐙_{\psi _c}^{[1]}+𝐙_{\psi _b}^{[1]}\}+𝚲^{J[1]}\right],$$
(50)
$`𝐙_\psi ^{[1]}`$ gives the relevant leg correction, whose relation to the self energy diagram (Figure 5) is described in appendix A, where we have followed closely the conventions of Morningstar . $`𝚲^{J[1]}`$ corresponds to contributions from the Feynman diagram of Figure 6 where gluon couplings from each of the $`_t`$, $`\frac{p^2}{2m}`$, and $`\frac{\sigma 𝐁}{2m}`$ terms in the action are included at each vertex.
For the spatial vector current (and indeed the temporal axial current) the mixing matrix is non-trivial. We denote the vertex correction to each of the lattice currents $`𝒪_\alpha ^{V_k}`$ from the sum of Feynman diagrams represented in Figures 6 and 8 by $`𝚲_\alpha ^{V_k[1]}`$. At one loop the the vertex correction overlaps with each of the corresponding continuum operators, giving rise to the non-trivial mixing problem
$$𝚲_\beta ^{V_k[1]}=\underset{\beta }{}𝚲_{\alpha \beta }^{V_k[1]}\mathrm{\Omega }_\beta ^{V_k}.$$
(51)
We resolve the Pauli and momentum structure of $`𝚲_1^{V_k[1]}`$ using the following projections with $`jk`$, then setting the spatial momenta to zero:
$$\begin{array}{cc}𝚲_{11}^{V_k[1]}=2iM_b\sigma _j\sigma _k\frac{}{p_j}𝚲_1^{V_k[1]}& 𝚲_{12}^{V_k[1]}=2iM_c\sigma _k\sigma _j\frac{}{p_j^{}}𝚲_1^{V_k[1]}\\ 𝚲_{13}^{V_k[1]}=iM_b\frac{}{p_k}𝚲_1^{V_k[1]}\frac{1}{2}𝚲_{11}^{V_k[1]}& 𝚲_{14}^{V_k[1]}=iM_c\frac{}{p_k^{}}𝚲_1^{V_k[1]}\frac{1}{2}𝚲_{12}^{V_k[1]}.\end{array}$$
(52)
We may use the bare lattice mass for convenience in place of the the pole mass in these projections since we are evaluating one-loop coefficients, the difference being $`O(\alpha _s^2)`$. Mixing matrix elements $`𝚲_{2\beta }`$ are related to $`𝚲_{1\beta }`$ by $`M_bM_c`$, and it will be seen later that $`𝚲_{3\beta }`$ and $`𝚲_{4\beta }`$ are not required for our calculation. Hence we write the contributions to the mixing matrix for the spatial vector current as follows,
$$_{\alpha \beta }^{V_k}=\delta _{\alpha \beta }+\alpha _s\left[\begin{array}{c}\frac{1}{2}\{𝐙_{\psi _c}^{[1]}+𝐙_{\psi _b}^{[1]}\}\delta _{\alpha \beta }+\left\{𝐙_{m_b}^{[1]}+𝐙_{m_b}^{\mathrm{TI}[1]}\right\}\left\{\delta _{\alpha 1}+\delta _{\alpha 3}\right\}\delta _{\alpha \beta }\\ +\left\{𝐙_{m_c}^{[1]}+𝐙_{m_c}^{\mathrm{TI}[1]}\right\}\left\{\delta _{\alpha 2}+\delta _{\alpha 4}\right\}\delta _{\alpha \beta }+𝚲_{\alpha \beta }^{V_k[1]}+𝚲_{\alpha \beta }^{\mathrm{TI}^{V_k}[1]}\end{array}\right].$$
(53)
where the the $`𝐙_m`$ factors arise from matching the bare mass in the lattice current to the pole mass in the continuum current to order $`\alpha _s`$, and the tadpole improvement counter terms are
$`𝐙_m^{\mathrm{TI}[1]}`$ $`=`$ $`u_0^{[1]}\left(1{\displaystyle \frac{3}{2naM_0}}\right)`$ (54)
$`𝚲_{\alpha \beta }^{\mathrm{TI}^{V_k}[1]}`$ $`=`$ $`\delta _{\alpha \beta }u_0^{[1]}.`$ (55)
Similarly, for the temporal axial current we project the vertex corrections on to the continuum basis using,
$$\begin{array}{cc}𝚲_{11}^{A_0[1]}=2M_b\sigma _j\frac{}{p_j}𝚲_1^{A_0[1]},& 𝚲_{12}^{A_0[1]}=2M_c\sigma _j\frac{}{p_j^{}}𝚲_1^{A_0[1]}.\end{array}$$
(56)
There is also a tadpole improvement counter term for the temporal axial current,
$$𝚲_{\alpha \beta }^{\mathrm{TI}^{A_0}[1]}=\delta _{\alpha \beta }u_0^{[1]}.$$
(57)
We write the contributions to the mixing matrix for the temporal axial current as follows,
$$_{\alpha \beta }^{A_0}=\delta _{\alpha \beta }+\alpha _s\left[\begin{array}{c}\frac{1}{2}\{𝐙_{\psi _c}^{[1]}+𝐙_{\psi _b}^{[1]}\}\delta _{\alpha \beta }+\{𝐙_{m_b}^{[1]}+𝐙_{m_b}^{\mathrm{TI}[1]}\}\delta _{\alpha 1}\delta _{\alpha \beta }\\ +\{𝐙_{m_c}^{[1]}+𝐙_{m_c}^{\mathrm{TI}[1]}\}\delta _{\alpha 2}\delta _{\alpha \beta }+𝚲_{\alpha \beta }^{A_0[1]}+𝚲_{\alpha \beta }^{\mathrm{TI}^{A_0}[1]}\end{array}\right].$$
(58)
The major task in the renormalisation is to calculate the terms in the above mixing matrices numerically using the Feynman rules for the lattice action.
The integrands for each of the Feynman graphs (and their first and second derivatives with respect to both $`p_i`$ and $`p_j^{}`$ where appropriate) were coded in two independent ways in order to verify the correctness of the calculation.
A C++ class was written to encode the Pauli algebra, and wrapped in a Taylor series class giving the first few entries of the (double) expansion in $`p_j`$ and $`p_k^{}`$. The standard operators were overloaded and standard functions applied the chain rule to their arguments. This enabled the Feynman diagrams to be constructed easily from the Feynman rules, with the Pauli manipulations taken care of by the code. The leading entries in the double Taylor series gave the first few derivatives with respect to $`p_j`$ and $`p_k^{}`$. Additionally much faster Fortran codes were written with the Pauli manipulations carried out manually, and all derivatives with respect to the external momenta taken by hand. VEGAS was used to evaluate the one loop integrals.
The lattice effective theory is constructed to have the same infra-red physics as the relativistic continuum theory, so that we find the same (cancelling) logarithmic infra-red divergences in the lattice renormalisation in the lattice diagrams as were present in continuum theory. These arise from graphs where a temporally polarised gluon is exchanged via the $`_t`$ coupling in the action. To handle the infra-red divergent graphs we always perform the integral over $`k_0`$ analytically, and then integrate infra-red finite sums of graphs, e.g. $`\frac{1}{2}\{𝐙_{\psi _c}^{[1]}+𝐙_{\psi _b}^{[1]}\}+𝚲_{11}^{[1]}`$, over the remaining three dimensions corresponding to the spatial components of the loop momentum.
## IV Matching and Results
We wish to form a linear combination of the lattice currents such that they match the continuum currents at one loop. We reconstruct the $`\overline{MS}`$ continuum currents from the lattice currents as follows,
$$\begin{array}{ccc}c(p^{})|J^{\overline{MS}}|b(p)^{1\mathrm{loop}}& =& \underset{\alpha ,\beta }{}Z_\alpha ^{\overline{MS}}(^J)_{\alpha \beta }^1c(p^{})|𝒪_\beta ^J|b(p)^{1\mathrm{loop}}\\ & =& \underset{\beta }{}\left[Z_\beta ^{\mathrm{tree}}+\alpha _s\rho _\beta ^J\right]c(p^{})|𝒪_\beta ^J|b(p)^{1\mathrm{loop}}\end{array}$$
(59)
where $`Z_\beta ^{\mathrm{tree}}\{0,1\}`$ is the tree level piece of $`Z_\beta ^{\overline{MS}}`$
In what follows we shall present results for the $`\rho _\beta `$ for each of the currents in turn as a function of both the bare lattice mass $`\frac{1}{M_b}`$ and the mass ratio $`\xi =\frac{M_c}{M_b}`$. For reasons of practicality, we make the stabilisation parameters for each quark, $`n_c`$ and $`n_b`$, implicit functions of the mass as indicated in table III.
### A Temporal Vector Current
For the temporal vector current we have
$`V_0^{\overline{MS}}`$ $`=`$ $`\left\{1+\alpha _s\rho ^{V_0}\right\}𝒪^{V_0},`$ (60)
$`\rho ^{V_0}`$ $`=`$ $`B^{V_0}{\displaystyle \frac{1}{2}}\{𝐙_{\psi _c}^{[1]}+𝐙_{\psi _b}^{[1]}\}𝚲^{V_0[1]},`$ (61)
$`B^{V_0}`$ $`=`$ $`{\displaystyle \frac{1}{3\pi }}\left[3{\displaystyle \frac{1+\xi }{1\xi }}\mathrm{log}\xi 6\right],`$ (62)
We plot the renormalisation coefficient $`\rho ^{V_0}`$ in figure 10. In table IV we give the correction as a function of $`\frac{1}{M_b}`$ for $`\xi =0.1,0.2,0.3,0.4,0.5`$ and $`1.0`$, useful ratios for lattice simulations. Since the lattice current is conserved, the one-loop renormalisation coefficient vanishes, for all $`M_b`$ in the case $`\xi =1`$, as clearly shown by figure 10.
Luke’s theorem shows that there is no $`O(\frac{1}{M})`$ correction to form factors involving the vector current in HQET at zero recoil. Luke’s theorem is a non-perturbative observation based upon the symmetries of HQET. The required symmetries are satisfied by Lattice NRQCD, and the theorem should certainly be manifested in the one-loop coefficients at zero-recoil. Demonstrating this is less trivial than in HQET however, since $`M`$ dependence arises out of terms in the action as well as through current insertions, and we proceed by numerically calculating the $`M`$ dependence of the lattice NRQCD contribution to the renormalisation. In figure 11 we display the lattice one loop renormalisation $`\rho ^{V_0}B^{V_0}`$ plotted versus $`(\frac{1}{M_c}\frac{1}{M_b})^2`$, for various values of $`\xi =\frac{M_c}{M_b}`$. This plot shows that the leading correction is linear in $`(\frac{1}{M_c}\frac{1}{M_b})^2`$ in the $`\frac{1}{M}0`$ region, with a universal slope. This is consistent with the absence of the $`\frac{1}{M}`$ term, as expected by Luke’s theorem applied to lattice NRQCD.
Of course the analogue of Luke’s theorem should be held non-perturbatively by lattice NRQCD, and a fit of the dependence of the non-perturbative lattice form factor on $`\frac{1}{M_b}`$ near the static limit should be made on simulation data to check the consistency with Luke’s theorem.
### B Spatial Axial Current
For the spatial axial current,
$`A_k^{\overline{MS}}`$ $`=`$ $`\left\{1+\alpha _s\rho ^{A_k}\right\}𝒪^{A_k},`$ (63)
$`\rho ^{A_k}`$ $`=`$ $`B^{A_k}{\displaystyle \frac{1}{2}}\{𝐙_{\psi _c}^{[1]}+𝐙_{\psi _b}^{[1]}\}𝚲^{A_k[1]},`$ (64)
$`B^{A_k}`$ $`=`$ $`{\displaystyle \frac{1}{3\pi }}\left[3{\displaystyle \frac{1+\xi }{1\xi }}\mathrm{log}\xi 8\right].`$ (65)
We plot the renormalisation coefficient $`\rho ^{A_k}`$ in figure 12, and give the correction as a function of $`\frac{1}{M_b}`$ for $`\xi =0.1,0.2,0.3,0.4,0.5`$ and $`1.0`$ in table V.
### C Temporal Axial Current
For the temporal axial current we have
$$A_0^{\overline{MS}}=\left\{1+\alpha _s\rho _1^{A_0}\right\}𝒪_1^{A_0}+\left\{1+\alpha _s\rho _2^{A_0}\right\}𝒪_2^{A_0}$$
(66)
where
$`\rho _1^{A_0}`$ $`=`$ $`B_1^{A_0}{\displaystyle \frac{1}{2}}\{𝐙_{\psi _c}^{[1]}+𝐙_{\psi _b}^{[1]}\}𝚲_{11}^{A_0[1]}𝚲_{21}^{A_0[1]}𝐙_{m_b}^{[1]}𝚲_{11}^{\mathrm{TI}^{A_0}[1]}𝐙_{m_b}^{\mathrm{TI}[1]}`$ (67)
$`\rho _2^{A_0}`$ $`=`$ $`B_2^{A_0}{\displaystyle \frac{1}{2}}\{𝐙_{\psi _c}^{[1]}+𝐙_{\psi _b}^{[1]}\}𝚲_{12}^{A_0[1]}𝚲_{22}^{A_0[1]}𝐙_{m_c}^{[1]}𝚲_{22}^{\mathrm{TI}^{A_0}[1]}𝐙_{m_c}^{\mathrm{TI}[1]}`$ (68)
and
$`B_1^{A_0}`$ $`=`$ $`{\displaystyle \frac{1}{3\pi }}\left[3{\displaystyle \frac{1+\xi }{1\xi }}\mathrm{log}\xi 86{\displaystyle \frac{1+\xi }{1\xi }}12{\displaystyle \frac{\xi \mathrm{log}\xi }{(1\xi )^2}}\right]`$ (69)
$`B_2^{A_0}`$ $`=`$ $`{\displaystyle \frac{1}{3\pi }}\left[3{\displaystyle \frac{1+\xi }{1\xi }}\mathrm{log}\xi 8+6{\displaystyle \frac{1+\xi }{1\xi }}+12{\displaystyle \frac{\xi \mathrm{log}\xi }{(1\xi )^2}}\right]`$ (70)
We plot the renormalisation coefficients $`\rho _1^{A_0}`$ and $`\rho _2^{A_0}`$ in figures 13 and 14. The various contributions to $`\rho _1^{A_0}`$ for $`\xi =0.3`$ are given in table VI allowing the reader to see relatively easily the change that must be applied to obtain the coefficients with a different tadpole improvement prescription, such as the mean link in Landau gauge. The diagonal elements $`𝚲_{\beta \beta }`$ of the mixing matrix both contain infra-red divergences which exactly cancel those of the wavefunction renormalisation, and have associated tadpole improvement counter terms associated with the $`u_0`$ factor in the derivative operator. It is, in fact, interesting to observe the tadpole improvement programme in action; there are large cancellations between the $`𝐙_m^{\mathrm{TI}[1]}`$ and $`𝐙_m^{[1]}`$ terms and between the $`\mathrm{\Lambda }_{11}^{\mathrm{TI}^{A_0}[1]}`$ and the $`\frac{1}{2}\{𝐙_{\psi _c}^{[1]}+𝐙_{\psi _b}^{[1]}\}+𝚲_{11}^{A_0[1]}`$ terms. This cancellation is even more manifest when we isolate the tadpole graph contributions within $`𝚲_{11}^{A_0[1]}`$ and $`𝐙_m^{[1]}`$. The two overall correction coefficients are tabulated as a function of $`\frac{1}{M_b}`$ for $`\xi =0.1,0.2,0.3,0.4,0.5`$ and $`1.0`$ in tables VII and VIII.
### D Spatial Vector Current
For the spatial vector current we obtain
$$\begin{array}{ccc}V_k^{\overline{MS}}& =& \left\{1+\alpha _s\rho _1^{V_k}\right\}𝒪_1^{V_k}+\left\{1+\alpha _s\rho _2^{V_k}\right\}𝒪_2^{V_k}+\alpha _s\rho _3^{V_k}𝒪_3^{V_k}+\alpha _s\rho _4^{V_k}𝒪_4^{V_k}\end{array}$$
(71)
where
$`\rho _1^{V_k}`$ $`=`$ $`B_1^{V_k}{\displaystyle \frac{1}{2}}\{𝐙_{\psi _c}^{[1]}+𝐙_{\psi _b}^{[1]}\}𝚲_{11}^{V_k[1]}𝚲_{21}^{V_k[1]}𝐙_{m_b}^{[1]}𝚲_{11}^{\mathrm{TI}^{V_k}[1]}𝐙_{m_b}^{\mathrm{TI}[1]}`$ (72)
$`\rho _2^{V_k}`$ $`=`$ $`B_2^{V_k}{\displaystyle \frac{1}{2}}\{𝐙_{\psi _c}^{[1]}+𝐙_{\psi _b}^{[1]}\}𝚲_{12}^{V_k[1]}𝚲_{22}^{V_k[1]}𝐙_{m_c}^{[1]}𝚲_{22}^{\mathrm{TI}^{V_k}[1]}𝐙_{m_c}^{\mathrm{TI}[1]}`$ (73)
$`\rho _3^{V_k}`$ $`=`$ $`B_3^{V_k}𝚲_{13}^{V_k[1]}𝚲_{23}^{V_k[1]}`$ (74)
$`\rho _4^{V_k}`$ $`=`$ $`B_4^{V_k}𝚲_{14}^{V_k[1]}𝚲_{24}^{V_k[1]}`$ (75)
and
$`B_1^{V_k}`$ $`=`$ $`B_2^{V_k}={\displaystyle \frac{1}{3\pi }}\left[3{\displaystyle \frac{1+\xi }{1\xi }}\mathrm{log}\xi 4\right]`$ (77)
$`B_3^{V_k}`$ $`=`$ $`{\displaystyle \frac{1}{3\pi }}\left[{\displaystyle \frac{2}{1\xi }}{\displaystyle \frac{2\xi \mathrm{log}\xi }{(1\xi )^2}}\right]`$ (78)
$`B_4^{V_k}`$ $`=`$ $`{\displaystyle \frac{1}{3\pi }}\left[{\displaystyle \frac{2\xi }{1\xi }}+{\displaystyle \frac{2\xi \mathrm{log}\xi }{(1\xi )^2}}\right].`$ (79)
Since the $`𝒪_3`$ and $`𝒪_4`$ do not appear at tree level, we can see that it is in fact unnecessary to calculate $`𝚲_{3\beta }^{V_k}`$ and $`𝚲_{4\beta }^{V_k}`$. We plot the renormalisation coefficients $`\rho _1^{V_k}\mathrm{}\rho _4^{V_k}`$ in figures 15 to 18. The various contributions to $`\rho _1^{V_k}`$ and $`\rho _3^{V_k}`$ are given for $`\xi =0.3`$ in table IX and table X. The four overall correction coefficients are given as a function of $`\frac{1}{M_b}`$ for $`\xi =0.1,0.2,0.3,0.4,0.5`$ an $`1.0`$ in tables XI to XIV.
## V Relevant Momentum Scale for $`\alpha _s`$
In section I we argued that the regularisation scheme correction is dominated by ultra-violet physics and is therefore calculable within perturbation theory. This is addressed in this section by attempting to calculate the appropriate scale for $`\alpha _s`$ in the correction to $`V_0`$ and $`A_k`$.
The Lepage-Mackenzie $`q^{}`$ scale for some one-loop process whose Feynman amplitude $`A=\frac{d^4k}{(2\pi )^4}I(k)`$ is defined by
$$\mathrm{log}(aq^{})^2=\frac{\frac{d^4k}{(2\pi )^4}I(k)\mathrm{log}(ak)^2}{\frac{d^4k}{(2\pi )^4}I(k)}.$$
(80)
When a one-loop expression is made up of two sub-processes, $`A=\frac{d^4k}{(2\pi )^4}I_A(k)`$ and $`B=\frac{d^4k}{(2\pi )^4}I_B(k)`$ we may combine the $`q^{}`$’s as follows,
$$\mathrm{log}(aq^{})_{A+B}^2=\frac{\frac{d^4k}{(2\pi )^4}I_A(k)\mathrm{log}(ak)^2+\frac{d^4k}{(2\pi )^4}I_B(k)\mathrm{log}(ak)^2}{\frac{d^4k}{(2\pi )^4}I_A(k)+\frac{d^4k}{(2\pi )^4}(I_B(k)}$$
(81)
or more conveniently,
$$q_{A+B}^{}=\left[(q_A^{})^A\times (q_B^{})^B\right]^{\frac{1}{A+B}}.$$
(82)
Here we shall calculate the Lepage-Mackenzie $`q^{}`$ from lattice perturbation theory, and then use this rule to combine it with the continuum BLM scales from section II B to obtain an overall scale.
The combination of lattice graphs containing the cancelling IR divergences proved troublesome using 4 dimensional numerical integration, so we used the (non-rigorous) Hernandez and Hill prescription to deal with the $`k_0`$ integral, and integrated the remaining spatial momenta numerically.
We present the results for $`q_{V_0}^{}`$ in table XV. The overall scale is reassuringly large, and is dominated by the continuum scale due to the typical relative size of the corrections. We therefore believe that the use of Hernandez and Hill prescription to obtain numerical stability constitutes an irrelevant error in the overall scale. We note that $`\alpha _V(q^{})0.22`$ for $`q^{}3GeV`$, which is the case in a typical simulation regimes for $`V_0`$ .
We present the results for $`q_{A_k}^{}`$ in table XVI. There is a line (straddled by the bold highlighted data in table XVI ) in the $`\xi M_b`$ plane where $`q^{}`$ becomes poorly behaved. It can be seen in table XVII that the individual scales $`q_{A_k}^{\mathrm{lat}}`$ and $`q_{A_k}^{\mathrm{cont}}`$ are large at this point, but that the overall one-loop correction is near vanishing. The location of this uncontrolled behaviour is not dependent on our use of the Hernandez and Hill prescription, however the details of the behaviour will, of course, be somewhat dependent. The zero in the overall correction introduces a vertical asymptote for $`\mathrm{log}(aq^{})^2`$, due to the vanishing denominator in (81) and consequently a limit of infinity or zero for $`q^{}`$ depending on the direction of approach. The resultant low scale on one side of the zero in the correction is not considered problematic by the authors, since it is a breakdown of the mean scale interpretation of (80), that occurs whenever the correction is small.<sup>§</sup><sup>§</sup>§As noted by Morningstar , the integrand $`I_A(k)+I_B(k)`$ must be either strictly positive or strictly negative for the mean value theorem to guarantee $`0aq^{}2\pi `$. Exact cancellation of the integrals is likely to exacerbate the problem.
The prescription used to set the scale at which to evaluate $`\alpha _s`$ in a mixing calculation is considered an open question by the authors, however we remain encouraged by the reliability of the perturbation theory suggested by the scales determined thus far.
## VI Summary
We have shown how to construct the $`\overline{MS}`$ weak decay currents for $`BD^{()}`$ transitions from operators in the lattice NRQCD effective theory at and around the zero recoil point for a wide range of phenomenologically relevant lattice mass parameters. These constants are of use to both recent and future calculations of the semileptonic decay form factors. The one-loop coefficients typically have modulus $`0.3`$ in the usual simulation regime. The lattice temporal vector current is not renormalised in the degenerate quark case (this is not the case for Wilson fermions), with the result that the current correction for physical $`\frac{m_c}{m_b}`$ ratios is particularly small. For two of the currents we have estimated the Lepage-Mackenzie $`q^{}`$ scales, and find that it is large, giving overall one-loop corrections of order $`5`$%, and suggesting perturbation theory is converging adequately. With further suppression, the next (and higher) order corrections in the operator renormalisation could well be below the expected statistical error of the BaBar data. Certainly, the two loop renormalisation correction to HQET operators is known to be $`O(1\%)`$ .
## VII Acknowledgments
We wish to thank Junko Shigemitsu for many useful conversations, and cross checking the mass renormalisation. We made use of unpublished notes by Colin Morningstar for checking our continuum calculation. Peter Boyle is supported by PPARC grant PP/CBA/62 . Both Peter Boyle and Christine Davies fondly acknowledge the hospitality of the UCSB, Santa Barbara where some of this work was carried out.
## A Self Energy Calculation
To one loop the graph topologies shown in figures 5 and 9 contribute to the lattice self energy. The diagrams contain contributions from both the temporally polarised gluon vertex, and from spatially polarised gluons via the $`\frac{p^2}{2m}`$ term and via the $`\sigma B`$ term. The lattice self energy must satisfy cubic invariance in the spatial momenta, and we write the first terms of a double taylor series for the self-energy as
$$\mathrm{\Sigma }^{\mathrm{lat}}(ap_0,a\stackrel{}{p})=g^2\left\{\mathrm{\Omega }_0+\mathrm{\Omega }_1iap_0+\mathrm{\Omega }_2\frac{p^2a}{2M_0a}+\mathrm{}\right\}$$
(A1)
where clearly,
$`g^2\mathrm{\Omega }_0`$ $`=`$ $`\mathrm{\Sigma }(0,0)`$ (A2)
$`g^2\mathrm{\Omega }_1`$ $`=`$ $`i{\displaystyle \frac{}{ap_0}}\mathrm{\Sigma }(0,0)`$ (A3)
$`g^2\mathrm{\Omega }_2`$ $`=`$ $`aM_0{\displaystyle \frac{1}{a^2}}{\displaystyle \frac{^2}{p_j^2}}\mathrm{\Sigma }(0,0)`$ (A4)
to order $`\alpha `$ then the small $`\frac{v^2}{c^2}`$ expansion of the heavy quark propagator is
$`G_r^1(p)`$ $`=`$ $`G^1(p)\mathrm{\Sigma }(ap_0,a\stackrel{}{p})`$ (A6)
$`=`$ $`ip_0+{\displaystyle \frac{p_0^2}{2}}+{\displaystyle \frac{p^2}{2M_0}}ip_0{\displaystyle \frac{p^2}{2M_0}}g^2\mathrm{\Omega }_0iap_0g^2\mathrm{\Omega }_1{\displaystyle \frac{p^2}{2M_0}}g^2\mathrm{\Omega }_2`$ (A7)
$`=`$ $`Z_\psi (ia\overline{p}_0+{\displaystyle \frac{p^2a^2}{2Z_m^{\mathrm{kin}}Ma}}+{\displaystyle \frac{\overline{p}_0^2a^2}{2}})`$ (A8)
where to order $`g^2`$ and $`\frac{v^2}{c^2}`$ we have,
$`i\overline{p}_0`$ $`=`$ $`ip_0g^2\mathrm{\Omega }_0`$ (A9)
$`Z_\psi `$ $`=`$ $`1g^2(\mathrm{\Omega }_0+\mathrm{\Omega }_1)`$ (A10)
$`Z_m^{\mathrm{kin}}`$ $`=`$ $`1+g^2(\mathrm{\Omega }_2\mathrm{\Omega }_1)`$ (A11)
The contributions from the tadpole diagram figure 9 to $`\mathrm{\Omega }_0`$ and $`\mathrm{\Omega }_1`$ explicitly cancel in $`Z_\psi `$ due to the equation of motion of the external spinor. In the $`Z_m`$ calculation, both the tadpole and normal topologies contribute. In order to compute $`Z_m`$ it was necessary to code a routine to compute the second derivative of the self energy with respect to one of the external spatial momentum components. This was fairly involved, due to the nature of the lattice Feynman rules. There are cancelling infra-red divergences in both $`\mathrm{\Omega }_0`$ and in $`\mathrm{\Omega }_1`$, so that the integral over the $`k_0`$ was performed analytically and the remaining three dimensions integrated numerically. Our numbers for both the $`Z_\psi `$, $`\mathrm{\Delta }_E`$, and $`Z_m`$ agree within statistical error with those of Morningstar and Shigemitsu .
|
warning/0003/cond-mat0003219.html
|
ar5iv
|
text
|
# Spin-orbit coupling in interacting quasi-one-dimensional electron systems
## Abstract
We present a new model for the study of spin-orbit coupling in interacting quasi-one-dimensional systems and solve it exactly to find the spectral properties of such systems. We show that the combination of spin-orbit coupling and electron-electron interactions results in: the replacement of separate spin and charge excitations with two new kinds of bosonic mixed-spin-charge excitation, and a characteristic modification of the spectral function and single-particle density of states. Our results show how manipulation of the spin-orbit coupling, with external electric fields, can be used for the experimental determination of microscopic interaction parameters in quantum wires.
In contemporary condensed matter physics there exists a great variety of electron systems that can be considered as quasi-one-dimensional (Q1D). Among them are conducting polymers , carbon nanotubes , and semiconductor heterostructures . They combine the richness of observable physical properties with the possibility of finding exact solutions to non-trivial interacting problems. The major theoretical advance in this field was the formulation and solution of the Tomonaga-Luttinger model , which revealed features that are generic to many interacting Q1D electron systems, such as separation of charge and spin degrees of freedom and the anomalous scaling of correlation functions .
The very existence of Q1D systems (or quantum wires) arises essentially from confining electrical forces, which prevent the lateral motion of electrons. The accompanying electric fields necessarily give rise to a coupling between the spin and orbital degrees of freedom of propagating electrons — the spin-orbit (SO) interaction. Although this interaction is relativistic in nature, it nevertheless results in a significant modification to the band structure of Q1D systems (see, e.g., experimental and theoretical papers). In strictly 1D and 2D systems, the role of the SO interaction basically reduces to a “horizontal” (in momentum space) splitting of the energy branches corresponding to spin-up and spin-down electron states. In contrast to this, the existence of quantised transverse energy subbands in quantum wires brings about an additional important manifestation of the SO coupling: the absence of a vertical symmetry axis for each spin branch (see Fig. 1 and Ref. ). This absence results in a breakdown of chiral symmetry so that the electron Fermi velocities become dependent on the direction of motion. This effect monotonically increases as the SO coupling is enhanced. As applied to interacting systems, the absence of chiral symmetry raises two fundamental questions: how does the system respond to the asymmetry of the single-particle spectrum; and what is the fate of the spin-charge separation in the presence of the SO coupling? It is the aim of this Letter, to establish a realistic model for the SO coupling in interacting 1D systems which answers these questions, and to provide experiment with a set of quantitative predictions which will allow their verification.
In constructing a Hamiltonian for such a system, we consider the case where the Fermi energy $`E_F`$ is sufficiently small, such that only the lowest energy subband in a quantum wire is partly filled, while all the others are empty. This regime has proven to be the richest in non-trivial experimental results , and the SO effects are expected to be most pronounced here. As a natural way of capturing the essential physics in a quantum wire, we suggest the use of a modification to the Tomonaga-Luttinger model that takes into account the asymmetric single-particle spectrum in Fig. 1. Namely, we take the Hamiltonian to have the form $`H=H_0+H_{int}`$, where $`H_0`$ describes the linearized spectrum,
$`H_0`$ $`=`$ $`iv_1{\displaystyle 𝑑x\left(\psi _{R,}^{}_x\psi _{R,}\psi _{L,}^{}_x\psi _{L,}\right)}`$ (2)
$`iv_2{\displaystyle 𝑑x\left(\psi _{R,}^{}_x\psi _{R,}\psi _{L,}^{}_x\psi _{L,}\right)},`$
and $`H_{int}`$ is responsible for the EE interaction:
$`H_{int}`$ $`=`$ $`g_2{\displaystyle 𝑑x\psi _{R,}^{}\psi _{R,}\psi _{L,}^{}\psi _{L,}}`$ (3)
$`+`$ $`g_4{\displaystyle }dx\psi _{R,}^{}\psi _{R,}\psi _{R,}^{}\psi _{R,}+(RL).`$ (4)
The operators $`\psi _{r,s}(x)`$ ($`r=R,L`$; $`s=,`$) annihilate spin-up ($``$) and spin-down ($``$) electrons near the right ($`R`$) and left ($`L`$) Fermi points. In Eq. (2) the electron velocities $`v_{1,2}=E(p_{1,2})/p`$ are derived from the dispersion law in Fig. 1. In general, $`v_1v_2`$ as long as the spin-orbit interaction is finite .
In the interaction Hamiltonian (4) we include only forward scattering contributions, while the backward and Umklapp scattering are left out. Neglecting the Umklapp scattering is known to be safe if energy bands are far from being half-filled, which is exactly the case, e.g., in quantum wires patterned in semiconductor heterostructures . As regards the backscattering, it can be shown that, similar to the case of zero SO coupling , it renormalizes to zero for repulsive interactions and therefore can also be omitted. In this letter, we concentrate on the major tendencies arising from the SO coupling in interacting electron systems, therefore we assume momentum-independent interaction potentials $`g_2>0`$ and $`g_4>0`$ in Eq. (4), which corresponds to a well-screened repulsion between electrons.
The Hamiltonian (2), (4) is reminiscent of that for the multi-component Tomonaga-Luttinger model which consists of mutually interacting Luttinger liquids with different Fermi velocities. However, in contrast to our case, the model assumes that each liquid has a symmetric single-electron spectrum (as occurs, for example, with Zeeman splitting ) and therefore, as we will show, it describes qualitatively different physical behaviour.
We study the interacting model (2), (4) with the help of the bosonization technique . Within this formalism, fermionic operators $`\psi _{r,s}(x)`$ are represented as $`\psi _{r,s}(x)\mathrm{exp}(i\mathrm{\Phi }_{r,s})`$, where the operators $`\mathrm{\Phi }_{r,s}(x)`$ are linear combinations of so-called phase fields $`\phi _{\rho (\sigma )}(x)`$ and $`\mathrm{\Pi }_{\rho (\sigma )}(x)`$ that obey the bosonic canonical commutation relations (see Refs. for details). In terms of these phase fields, the Hamiltonian (2), (4) takes the bosonized form:
$`H={\displaystyle \frac{1}{2\pi }}{\displaystyle 𝑑x}`$ $`\{`$ $`v_\rho K_\rho (\pi \mathrm{\Pi }_\rho )^2+{\displaystyle \frac{v_\rho }{K_\rho }}(_x\phi _\rho )^2`$ (5)
$`+`$ $`v_\sigma K_\sigma (\pi \mathrm{\Pi }_\sigma )^2+{\displaystyle \frac{v_\sigma }{K_\sigma }}(_x\phi _\sigma )^2`$ (6)
$`+`$ $`\delta v[(\pi \mathrm{\Pi }_\rho )_x\phi _\sigma +(\pi \mathrm{\Pi }_\sigma )_x\phi _\rho ]\},`$ (7)
where
$$v_0=(v_1+v_2)/2,\delta v=v_2v_1$$
(8)
and $`v_{\rho (\sigma )}=\left[(v_0\pm g_4/2\pi )^2(g_2/2\pi )^2\right]^{1/2}`$ and $`K_{\rho (\sigma )}=\left[(2\pi v_0g_2\pm g_4)/(2\pi v_0\pm g_2\pm g_4)\right]^{1/2}`$. The first (second) line in Eq. (7) describes free propagation of charge (spin) density wave with the velocity $`v_{\rho (\sigma )}`$ and “stiffness” coefficient $`K_{\rho (\sigma )}`$. The third line in Eq. (7) is proportional to the velocity difference $`\delta v`$ and therefore represents the strength of the SO interaction. In the Tomonaga-Luttinger model, where $`v_1=v_2`$, this term is absent and this results in a decoupling of the charge and spin degrees of freedom: the so-called spin-charge separation . This decoupling inhibits the existence of quasi-particles with spin $`1/2`$ and charge $`e`$, the basic excitations of a Fermi liquid, and gives rise to a different state of matter, the Luttinger liquid, which has bosonic excitations in the form of independent spin and charge density waves.
As the SO interaction is switched on, i.e. at $`\delta v0`$, the third term in Eq. (7) starts to affect the dynamics of the system. It couples together the $`\rho `$\- and $`\sigma `$-fields and thereby destroys this spin-charge separation. A similar effect is also found in a spinful Luttinger model in a magnetic field . In our case, the asymmetry of the single-electron spectrum in Fig. 1 results in a different mechanism for the violation of spin-charge separation from Ref. . We stress however, that our model is exactly solvable since the Hamiltonian (2), (4) is quadratic in the phase fields. Here we choose to solve it within the functional integration formalism in imaginary (Matsubara) time .
The most profound effects of the SO coupling on interacting Q1D systems can be seen in the behaviour of single-particle characteristics, such as the spectral function $`\rho _r(q,\omega )`$,
$$\rho _r(q,\omega )=\frac{1}{\pi }\underset{s}{}\mathrm{Im}G_{r,s}^{(ret)}(q,\omega ),r=R,L,$$
(9)
and the density of states $`N(\omega )=(1/2\pi )_r𝑑q\rho _r(q,\omega )`$. Here $`q`$ and $`\omega `$ are the momentum and energy of elementary excitations and $`G_{r,s}^{(ret)}(q,\omega )`$ is the Fourier transform of the retarded Green function
$`G_{r,s}^{(ret)}(x,t)`$ $``$ $`i\theta (t)\{\psi _{r,s}(x,t),\psi _{r,s}^{}(0,0)\}`$ (10)
$`=`$ $`i\theta (t)\left[G_{r,s}(x,t)+G_{r,s}(x,t)\right],`$ (11)
where $`\theta (t)`$ is the Heaviside step function and $`G_{r,s}(x,t)=\psi _{r,s}(x,\tau )\psi _{r,s}^{}(0,0)|_{\tau it}`$ with $`\tau `$ being the Matsubara time. Leaving detailed calculations to a regular article, we give here an exact expression for $`G_{r,s}(x,t)`$ at $`T=0`$:
$$G_{r,s}(x,t)=\frac{\mathrm{exp}(ip_{r,s}x)}{2\pi \mathrm{\Lambda }}\left(\frac{\mathrm{\Lambda }}{\mathrm{\Lambda }+i(u_1tx)}\right)^{\theta _1^+}\left(\frac{\mathrm{\Lambda }}{\mathrm{\Lambda }+i(u_1t+x)}\right)^{\theta _1^{}}\left(\frac{\mathrm{\Lambda }}{\mathrm{\Lambda }+i(u_2tx)}\right)^{\theta _2^+}\left(\frac{\mathrm{\Lambda }}{\mathrm{\Lambda }+i(u_2t+x)}\right)^{\theta _2^{}}.$$
(12)
A small (of the order of the lattice parameter) length scale $`\mathrm{\Lambda }`$ defines ultraviolet cut-off . From Eq. (12) it follows that although the SO coupling destroys the spin-charge separation, it nevertheless preserves the anomalous scaling of correlation functions. At $`\delta v=0`$, Eq. (12) coincides with the well-known expression . The exponents $`\theta _i^\pm `$ ($`i=1,2`$) are determined by values of $`v_{\rho (\sigma )}`$, $`K_{\rho (\sigma )}`$, $`\delta v`$ and depend on the chiral $`r`$ and spin $`s`$ indices. For arbitrary interaction constants $`g_2`$ and $`g_4`$, the expressions for $`\theta _i^\pm `$ are rather lengthy and here we give them for the realistic case of equal interaction strength between all the spectral branches, i.e. for $`g_2=g_4(\pi v_0)\beta `$:
$$\theta _i^\pm (r,s)=(1)^i\xi _i^\pm (r,s)\eta _i/2(\eta _1^2\eta _2^2),$$
(13)
$`\xi _i^\pm (r,s)`$ $`=`$ $`\pm r\left[(1+rsϵ/2)^2/\eta _i\eta _i\right](1rsϵ/2)`$ (14)
$`+`$ $`\left[(1+rsϵ/2)(1ϵ^2/4)\beta ^2/2\right]/\eta _i^2.`$ (15)
Here $`ϵ=\delta v/v_0`$, and on the r.h.s. of the last equation $`r=\pm 1`$ for right (left) spectrum branches and $`s=\pm 1`$ for spin-up (spin-down) electrons.
The quantities $`\eta _{1,2}\eta _{1,2}(ϵ)`$ defined by
$$\eta _{1,2}^2(u_{1,2}/v_0)^2=1+ϵ^2/4\sqrt{\beta ^2+ϵ^2}$$
(16)
are dimensionless velocities of the new independent bosonic excitations that replace the spin and charge density waves, respectively, as the single-electron spectrum becomes asymmetric ($`ϵ0`$). Each of these new excitations are superpositions of the old ones, they carry both charge and spin, and have a sound-like spectrum $`\omega =u_{1,2}q`$. For $`ϵ=0`$, we have $`u_{1,2}=v_{\sigma (\rho )}`$ and return to the spin-charge separation. Eq. (16) demonstrates that increasing $`ϵ`$ at constant $`\beta `$ pushes $`u_1`$ and $`u_2`$ away from $`v_\sigma `$ and $`v_\rho `$ as well as from each other: one of the excitations monotonically accelerates ($`u_2`$ grows) while the other monotonically slows down ($`u_1`$ decreases). This effect becomes more pronounced in systems with stronger EE interaction. Since the parameter $`ϵ`$ is controlled by the SO interaction and therefore by the lateral electrical confinement, the strength of this confinement must affect drastically the dynamics of elementary excitations in quantum wires. Further increasing $`ϵ`$ eventually results in a vanishing of the velocity $`u_1`$ at the “critical” point $`ϵ_0=2\sqrt{1\beta }`$ and in a possible phase transition accompanied by phase separation . This phenomenon is beyond the scope of this Letter.
Eq. (12) allows one to calculate the spectral function $`\rho _r(q,\omega )`$ which can be obtained in magnetotunneling experiments . Fig. 2 shows the spectral function $`\rho _R(q,\omega )`$ for right-moving electrons in the range of energies $`\omega `$ where all singularities of $`\rho _R(q,\omega )`$ are located. At each value of $`ϵ<ϵ_0`$, the spectral function has two singular points corresponding to the bosonic excitations with velocities $`u_1`$ and $`u_2`$ \[see Eq. (16)\]. In full accordance with Eq. (16), the distance between the singular points grows with $`ϵ`$. The singularities of $`\rho _R(q,\omega )`$ are governed by power laws with exponents depending on $`v_{\rho (\sigma )}`$, $`K_{\rho (\sigma )}`$, and $`ϵ`$. The dependence on $`ϵ`$ is such that one of the singularities becomes sharper while the other becomes smoother as $`ϵ`$ grows. It is noteworthy that the dependence of $`\rho _R`$ on $`q`$ at fixed $`\omega `$ is very similar to that shown in Fig. 2 and exhibits the same tendencies as a function of $`ϵ`$.
From Eqs. (9), (11), and (12) one can also calculate the single-particle density of states $`N(\omega )`$. This quantity may be obtained in tunneling measurements since the tunneling current, e.g., between a wide metal and an interacting quantum wire is proportional to $`N(eV)`$, where $`V`$ is an applied voltage (see Ref. ). It can also be determined in angle-integrated photoemission experiments . For $`\omega 0`$ we obtain the following behaviour:
$$\frac{N(\omega )}{N_0}=\frac{1}{4}\underset{r,s}{}\frac{(\omega /\omega _0)^{\theta _1+\theta _21}}{\eta _1^{\theta _1}\eta _2^{\theta _2}\mathrm{\Gamma }(\theta _1+\theta _2)},\theta _i=\theta _i^++\theta _i^{}.$$
(17)
Here $`\omega _0=v_0/\mathrm{\Lambda }`$ is the natural energy unit of the order of the Fermi energy, $`N_0=2/\pi v_0`$ is the constant density of states at $`\beta =ϵ=0`$, and $`\mathrm{\Gamma }(x)`$ is the Gamma function.
Fig. 3a demonstrates the effect of the EE interaction on the density of states $`N(\omega )`$ at a fixed strength of the SO coupling. At zero EE interaction ($`\beta =0`$), the function $`N(\omega )`$ is a constant, $`N(\omega )/N_0=1/(1ϵ^2/4)`$. As the EE interaction is switched on ($`\beta 0`$), it starts to affect the formation of the lowest lying excitations and $`N(\omega )`$ takes on a power-law behaviour in the vicinity of $`\omega =0`$. The width of this power-law interval becomes progressively larger as $`\beta `$ grows and leads to a monotonic suppression of the density of states. This effect is also present for zero SO coupling .
Fig. 3b shows the evolution of the function $`N(\omega )`$ for fixed $`\beta `$ as the SO coupling is varied. For large values of $`\omega `$, where the role of EE interactions is not significant, the magnitude of $`N(\omega )`$ increases as $`ϵ`$ grows. However, at small values of $`\omega `$, where the nature of the elementary excitations is essentially dictated by EE interactions, the effect of the SO coupling on $`N(\omega )`$ is qualitatively the same as that of the EE interaction, that is, increasing $`ϵ`$ leads to a suppression of the density of states. We emphasize that in the standard multi-component Tomonaga-Luttinger model as well as in spin-polarized Luttinger liquids , there is no interval of $`\omega `$ where the density of states is suppressed by increasing the velocity difference between spectral branches. The existence of such an interval is a unique manifestation of the SO coupling in quantum wires.
The characteristic dependence of the spectral function $`\rho (q,\omega )`$ and the density of states $`N(\omega )`$ on the microscopic parameter $`ϵ`$, for fixed $`\beta `$, should enable one to extract both quantities from experiment and thus determine how strong the EE and SO interactions are. Such dependencies can be obtained experimentally by changing the SO coupling directly by varying an electric field perpendicular to a quantum wire (quantum-well field) as it was done, e.g., in Ref. .
Theoretical calculations of the electron bandstructure modified by the SO coupling indicate that the value of $`ϵ`$ in typical Q1D semiconductor systems should be $`0.1`$$`0.2`$ and appears sufficiently large to observe the principal tendencies in the behaviour of $`\rho (q,\omega )`$ and $`N(\omega )`$ caused by the SO coupling.
In conclusion, we have formulated and solved analytically the problem of the interplay between EE and SO interactions in Q1D electron systems and found that: (i) the spin-charge separation breaks down while the anomalous scaling of correlation functions is preserved; (ii) the single-particle characteristics, such as the spectral function and the density of states, are essentially modified and controlled by the strength of the SO coupling; (iii) varying the SO coupling with the external electric field can be used to extract the microscopic parameters of quantum wires.
AVM thanks the ORS, COT, and Corpus Christi College for financial support. KVS and CHWB thank the EPSRC for financial support.
|
warning/0003/hep-ph0003011.html
|
ar5iv
|
text
|
# New constraints for non-Newtonian gravity in nanometer range from the improved precision measurement of the Casimir force
## ACKNOWLEDGMENTS
The authors are grateful to U. Mohideen for several helpful discussions. G.L.K. and V.M.M. are indebted to Center of Theoretical Sciences and Institute of Theoretical Physics of Leipzig University, where this work was performed, for kind hospitality. G.L.K. was supported by Graduate College on Quantum Field Theory at Leipzig University. V.M.M. was supported by Saxonian Ministry for Science and Fine Arts.
|
warning/0003/math0003169.html
|
ar5iv
|
text
|
# On effective non-vanishing and base-point-freeness
## 1 semipositivity
The semipositivity theorem proved in was used for the study of algebraic fiber spaces whose fibers have nonnegative Kodaira dimension. Now its logarithmic generalization will be applied for those with negative Kodaira dimension as well.
We start with recalling the semipositivity theorem ( Theorem 5) with slightly different expression:
###### Theorem 1.1.
Let $`X`$ and $`S`$ be smooth projective varieties and let $`f:XS`$ be a surjective morphism. Let $`n=dimXdimY`$. Assume that there exists a normal crossing divisor $`\mathrm{\Gamma }`$ on $`S`$ such that $`f`$ is smooth over $`S_0=S\mathrm{\Gamma }`$. Then the following hold:
(1) $`=f_{}𝒪_X(K_{X/S})`$ is a locally free sheaf, where $`K_{X/S}=K_Xf^{}K_S`$.
(2) Let $`\pi :P=()S`$ be the associated projective space bundle, and let $`P_0=\pi ^1(S_0)`$. Then the tautological invertible sheaf $`𝒪_P(1)`$ on $`P`$ has a singular hermitian metric $`h`$ which is smooth over $`P_0`$ and such that the curvature current $`\mathrm{\Theta }`$ is semipositive and that the corresponding multiplier ideal sheaf coincides with $`𝒪_P`$.
(3) Let $`X_0=f^1(S_0)`$ and $`f_0=f|_{X_0}`$. If the local monodromies of $`R^nf_0_{X_0}`$ around the branches of $`\mathrm{\Gamma }`$ are unipotent, then the Lelong number of $`\mathrm{\Theta }`$ vanishes at any point of $`P`$. In particular, $``$ is numerically semipositive. If $`\mathrm{\Theta }`$ is strictly positive at a point on $`P_0`$, then $`𝒪_P(1)`$ is also big.
###### Proof.
The hermitian metric $`\stackrel{~}{h}`$ on $`|_{S_0}`$ is defined by the integration along the fiber: for $`sS_0`$ and $`u,v_s`$,
$$\stackrel{~}{h}_s(u,v)=\mathrm{const}._{f^1(x)}u\overline{v}.$$
By , the curvature form of $`\stackrel{~}{h}`$ is Griffiths semipositive. Hence the smooth metric $`h`$ on $`𝒪_P(1)|_{P_0}`$ induced from $`\stackrel{~}{h}`$ has semipositive curvature form as well. Moreover, it is extended to a singular hermitian metric $`h`$ over $`P`$. The multiplier ideal sheaf is trivial because the sections of $``$ are $`L^2`$. In the case (3), the growth of the metric is logarithmic. So the Lelong number vanishes, and the last statements follow from the regularization of positive currents (). ∎
The semipositivity theorem is generalized to the logarithmic case by the covering method:
###### Theorem 1.2.
Let $`X`$ and $`S`$ be smooth projective varieties, let $`f:XS`$ be a surjective morphism, and let $`B`$ be an effective $``$-divisor on $`X`$ whose support is a normal crossing divisor and whose coefficients are strictly less than $`1`$. Assume that there exists a normal crossing divisor $`\mathrm{\Gamma }`$ on $`S`$ such that $`f`$ is smooth and $`\mathrm{Supp}(B)`$ is relative normal crossing over $`S_0=S\mathrm{\Gamma }`$. Let $`D`$ be a Cartier divisor on $`X`$. Assume that $`D_{}K_{X/S}+B`$. Then the following hold:
(1) $`=f_{}𝒪_X(D)`$ is a locally free sheaf.
(2) Let $`\pi :P=()S`$ be the associated projective space bundle, and let $`P_0=\pi ^1(S_0)`$. Then the tautological invertible sheaf $`𝒪_P(1)`$ on $`P`$ has a singular hermitian metric $`h`$ which is smooth over $`P_0`$ and such that the curvature current $`\mathrm{\Theta }`$ is semipositive and that the corresponding multiplier ideal sheaf coincides with $`𝒪_P`$.
(3) There exists a finite surjective morphism $`\sigma :S^{}S`$ from a smooth projective variety $`S^{}`$ such that $`\mathrm{\Gamma }^{}=\sigma ^1(\mathrm{\Gamma })`$ is a normal crossing divisor and satisfies the following conditions: Let $`X^{}X\times _SS^{}`$ be a birational morphism from a smooth projective variety which is isomorphic over $`S_0`$ and such that the union of the pull-back of the support of $`B`$, the pull-back of the support of $`\mathrm{\Gamma }`$ and the exceptional locus is a normal crossing divisor. Let $`f^{}:X^{}S^{}`$ and $`\tau :X^{}X`$ be the induced morphisms. An effective $``$-divisor $`B^{}`$ on $`X^{}`$ is defined such that its coefficients are strictly less than $`1`$ and that $`R=\tau ^{}(K_{X/S}+B)(K_{X^{}/S^{}}+B^{})`$ is a divisor. Let $`D^{}=\tau ^{}DR`$. Then $`R`$ is effective, and the assumptions of the theorem are satisfied by $`f^{}:X^{}S^{}`$, $`B^{}`$ and $`D^{}`$. The locally free sheaf $`^{}=f_{}^{}𝒪_X^{}(D^{})`$ on $`S^{}`$ satisfies that $`\sigma ^{}^{}`$. The singular hermitian metric $`h`$ induces a singular hermitian metric $`h^{}`$ on the tautological invertible sheaf $`𝒪_P^{}(1)`$ on $`P^{}=(^{})`$, and the Lelong number of the curvature current $`\mathrm{\Theta }^{}`$ vanishes at any point of $`P^{}`$. In particular, $`^{}`$ is numerically semipositive. If $`\mathrm{\Theta }^{}`$ is strictly positive at a point on $`P_0^{}`$, then $`𝒪_P^{}(1)`$ is also big.
###### Proof.
Let $`m`$ be the minimal positive number such that $`mDm(K_{X/S}+B)`$. We take a rational function $`h`$ on $`X`$ such that $`\text{div}(h)=mD+m(K_{X/S}+B)`$. Let $`\pi :YX`$ be the normalization of $`X`$ in the field $`(X)(h^{1/m})`$, and let $`\mu :Y^{}Y`$ be a desingularization such that the composite morphism $`g:Y^{}S`$ is smooth over $`S_0`$. We have
$$\pi _{}𝒪_Y\underset{k=0}{\overset{m1}{}}𝒪_X(kD+kK_{X/S}+\mathrm{}kB\mathrm{}).$$
The Galois group $`G/m`$ acts on $`Y`$ such that the above direct summands of $`\pi _{}𝒪_Y`$ are eigenspaces with eigenvalues $`\text{exp}(2\pi \sqrt{1}k/m)`$.
Since $`\pi `$ is etale outside the support of $`B`$, $`Y`$ has only rational singularities, hence $`\mu _{}𝒪_Y^{}(K_Y^{})=𝒪_Y(K_Y)`$. We apply Theorem 1.1 to the sheaf $`g_{}𝒪_Y^{}(K_{Y^{}/S})=f_{}\pi _{}𝒪_Y(K_{Y/S})`$. By duality, we have
$$\pi _{}𝒪_Y(K_Y)\underset{k=0}{\overset{m1}{}}𝒪_X(K_X+kDkK_{X/S}\mathrm{}kB\mathrm{}).$$
By taking $`k=1`$ (we may assume that $`m2`$), we obtain our assertions (1) and (2) since $`f_{}𝒪_X(D\mathrm{}B\mathrm{})=f_{}𝒪_X(D)`$.
For (3), we use the unipotent reduction theorem for the local monodromies of $`g`$ (). ∎
If the base space is $`1`$-dimensional, we have a simpler expression:
###### Corollary 1.3.
Let $`X`$ be a complete normal variety, and $`B`$ an effective $``$-divisor on $`X`$ such that the pair $`(X,B)`$ is KLT. Let $`f:XC`$ be a surjective morphism to a smooth curve. Let $`D`$ be a Cartier divisor on $`X`$ such that $`D_{}K_{X/C}+B`$. Then $`f_{}𝒪_X(D)`$ is a numerically semipositive locally free sheaf on $`C`$.
###### Proof.
Let $`\mu :X^{}X`$ be a log resolution for the pair $`(X,B)`$, and set $`\mu ^{}(K_X+B)=K_X^{}+B^{}`$. The coefficients of $`B^{}`$ are less than $`1`$ and negative coefficients appear only for exceptional divisors of $`\mu `$. We set $`B^{}=B_I^{}+B_F^{}`$ where $`B_I^{}`$ is an effective integral divisor and $`B_F^{}`$ is a $``$-divisor whose coefficients belong to the interval $`(0,1)`$. Since the support of $`B_I^{}`$ is exceptional for $`\mu `$, we have $`\mu _{}𝒪_X^{}(B_I^{})=𝒪_X`$. By applying the theorem to the pair $`(X^{},B_F^{})`$, we deduce that the sheaf $`f_{}𝒪_X(D)=f_{}\mu _{}𝒪_X^{}(\mu ^{}D+B_I^{})`$ is numerically semipositive. ∎
###### Corollary 1.4.
Let $`X,B`$ and $`f:XC`$ be as in Corollary 1.3. Let $`D`$ be a Cartier divisor on $`X`$ such that $`H=D(K_{X/C}+B)`$ is nef and big. Then $`f_{}𝒪_X(D)`$ is a numerically semipositive locally free sheaf on $`C`$.
###### Proof.
There exists an effective $``$-divisor $`B^{}`$ such that $`(X,B+B^{})`$ is KLT and $`D_{}K_{X/C}+B+B^{}`$. ∎
In the case of rank one sheaf, we have a more precise result which is not used later:
###### Corollary 1.5.
In Theorem 1.2, assume that $`=𝒪_S(F)`$ is an invertible sheaf. Let $`E`$ be an effective divisor such that $`EDf^{}F`$. Let $`\mathrm{\Delta }`$ be the smallest $``$-divisor supported on $`\mathrm{\Gamma }`$ such that $`(X,BE+f^{}(\mathrm{\Gamma }\mathrm{\Delta }))`$ is subLC over the generic points of $`\mathrm{\Gamma }`$. Then $`F\mathrm{\Delta }`$ is nef.
###### Proof.
Let $`\mu :X^{}X`$ be the log resolution of the pair $`(X,B+E+f^{}\mathrm{\Gamma })`$. We have $`\mu ^{}f^{}F\mu ^{}(DE)_{}\mu ^{}(K_{X/S}+BE)_{}K_{X^{}/S}+B^{}`$ for some $``$-divisor $`B^{}`$. Then our assertion is proved in Theorem 2. ∎
## 2 Reduction
We consider the following problem:
###### Conjecture 2.1.
Let $`X`$ be a complete normal variety, $`B`$ an effective $``$-divisor on $`X`$ such that the pair $`(X,B)`$ is KLT, and $`D`$ a Cartier divisor on $`X`$. Assume that $`D`$ is nef, and that $`H=D(K_X+B)`$ is nef and big. Then $`H^0(X,D)0`$.
This problem was considered in in order to construct ladders on log Fano varieties. By the generalization of the Kodaira Vanishing Theorem ( Theorem 1.2.5), we have $`H^p(X,D)=0`$ for any positive integer $`p`$. Thus the condition $`H^0(X,D)0`$ is equivalent to saying that $`\chi (X,D)0`$. Our problem is a topological question, unlike the case of the Abundance Conjectures.
The base point free theorem says that there exists a positive integer $`m_1`$ such that the linear system $`|mD|`$ is free for $`mm_1`$. The following reduction theorem is obtained as an application of the base point free theorem and the semipositivity theorem with the help of the perturbation technique.
###### Theorem 2.2.
In Conjecture 2.1, one may assume that $`B`$ is a $``$-divisor and that $`H`$ is ample. Moreover, one may assume that $`D`$ is also ample if one replaces $`X`$ suitably.
###### Proof.
By the Kodaira lemma, there exists an effecive $``$-divisor $`E`$ such that $`B+E`$ is a $``$-divisor, the pair $`(X,B+E)`$ is KLT, and that $`HE`$ is ample. Therefore, we may assume that $`B`$ is a $``$-divisor and that $`H`$ is ample.
By the Base Point Free Theorem, there exists a proper surjective morphism $`\varphi :XX^{}`$ with connected fibers to a normal projective variety such that $`D\varphi ^{}D^{}`$ for an ample Cartier divisor $`D^{}`$ on $`X^{}`$. We have $`H^0(X,D)0`$ if and only if $`H^0(X^{},D^{})0`$. We shall show that there exists an effective $``$-divisor $`B^{}`$ on $`X^{}`$ such that $`(X^{},B^{})`$ is KLT and $`D^{}(K_X^{}+B^{})`$ is ample.
Since $`H`$ is already assumed to be ample, we can write $`H=H_0+2\varphi ^{}H^{}`$ with $`H_0`$ and $`H^{}`$ being ample $``$-divisors. Since $`H_0`$ is ample, there exists an effective $``$-divisor $`B_0`$ such that $`B+H_0_{}B_0`$ and that $`(X,B_0)`$ is KLT. We set $`D_0=K_X+B_0`$. Then $`D_0_{}\varphi ^{}D_0^{}`$ for $`D_0^{}=D^{}2H^{}`$.
We construct birational morphisms $`\mu :YX`$ and $`\mu ^{}:Y^{}X^{}`$ from smooth projective varieties such that $`\varphi \mu =\mu ^{}\psi `$ for a morphism $`\psi :YY^{}`$. We write $`\mu ^{}(K_X+B_0)K_Y+E_0`$. If $`\mu `$ and $`\mu ^{}`$ are chosen suitably, then we may assume that the conditions of Theorem 2 are satisfied for $`\psi `$ and $`E_0`$. Then there exist $``$-divisors $`E_0^{}`$ and $`M`$ on $`Y^{}`$ such that $`K_Y+E_0_{}K_Y^{}+E_0^{}+M`$, $`\mu _{}^{}E_0^{}`$ is effective, $`\mathrm{}E_0^{}\mathrm{}0`$ and $`M`$ is nef. Since $`H^{}`$ is ample, there exists a $``$-divisor $`E^{}_{}E_0^{}+M+\mu ^{}H^{}`$ on $`Y^{}`$ such that $`B^{}=\mu _{}^{}E^{}`$ is effective and $`\mathrm{}E^{}\mathrm{}0`$. Then we have $`D_0^{}+H^{}_{}K_X^{}+B^{}`$ and $`(X^{},B^{})`$ is KLT. Since $`D^{}_{}K_X^{}+B^{}+H^{}`$, we obtain our assertion. ∎
## 3 Surface case
We have a complete answer in dimension $`2`$.
###### Theorem 3.1.
Let $`X,B`$ and $`D`$ be as in Conjecture 2.1. Assume that the numerical Kodaira dimension $`\nu (X,D)`$ is at most $`2`$; namely, assume that $`D^30`$. Then the following hold.
(1) $`H^0(X,D)0`$.
(2) The linear system $`|mD|`$ is free for any integer $`m`$ such that $`m2`$.
###### Proof.
We may assume that $`dimX=\nu (X,D)2`$ by Theorem 2.2. Let $`\mu :X^{}X`$ be the minimal resolution of singularities. Since $`\mu ^{}K_XK_X^{}`$ is effective, we can write $`\mu ^{}(K_X+B)=K_X^{}+B^{}`$ with $`(X^{},B^{})`$ being KLT. Therefore, we may assume that $`X`$ is smooth. By Theorem 2.2 again, we may also assume that $`H`$ is ample and that $`D`$ is big.
Assume first that $`dimX=1`$. Then the assertions follow immediately from the Riemann-Roch theorem.
We assume that $`dimX=2`$ in the following. We prove (1). By the Riemann-Roch theorem, $`\chi (X,D)=\frac{1}{2}D(B+H)+\chi (X,𝒪_X)`$. Thus, if $`\chi (X,𝒪_X)0`$, then $`\chi (X,D)>0`$. Let us assume that $`\chi (X,𝒪_X)=1g<0`$. Then there exists a surjective morphism $`f:XC`$ to a curve of genus $`g`$ whose generic fiber is isomorphism to $`^1`$. By Corollary 1.4, the vector bundle $`f_{}𝒪_X(Df^{}K_C)`$ is numerically semipositive. Since $`𝒪_X(D)`$ is $`f`$-nef, it is $`f`$-generated, hence we have a surjective homomorphism $`f^{}f_{}𝒪_X(Df^{}K_C)𝒪_X(Df^{}K_C)`$, and the latter sheaf is nef. Thus $`(Df^{}K_C)(B+H)0`$. Since $`f^{}K_C(B+H)f^{}K_CK_X=4g4`$, we have $`\chi (X,D)g1>0`$.
In order to prove (2), we take a general member $`Y|D|`$ as a subscheme of $`X`$. We have an exact sequence $`0𝒪_X((m1)D)𝒪_X(mD)𝒪_Y(mD)0`$ and $`H^1(X,(m1)D)=0`$, hence it is sufficient to prove the freeness of $`|𝒪_Y(mD)|`$. Let $`𝔪`$ be any ideal sheaf of $`𝒪_Y`$ of colength $`1`$. We shall prove that $`H^1(Y,𝔪(mD))=0`$. By duality, it is equivalent to $`\text{Hom}(𝔪,\omega _Y(mD))=0`$. Since $`\text{deg }\omega _Y`$ is even and $`Y(B+H)>0`$, we have $`\text{deg }\omega _Y(mD)2`$, and we have the desired vanishing. ∎
Our bound for the freeness in Theorem 3.1 is better than the one given by the Fujita conjecture. But we cannot expect similar thing in higher dimensions:
###### Example 3.2.
(1) (Oguiso) Let $`X`$ be a general weighted hypersurface of degree $`10`$ in a weighted projective space $`(1,1,1,2,5)`$. Then $`X`$ is smooth, $`dimX=3`$, and $`K_X0`$. Let $`D=H=𝒪_X(1)`$. We have $`H^0(X,D)0`$, and $`|2D|`$ is free. But $`|3D|`$ is not free, and $`|4D|`$ is not very ample.
(2) Let $`d`$ be an odd integer such that $`d3`$, and let $`X`$ be a general weighted hypersurface of degree $`2d`$ in $`(1,\mathrm{},1,2,d)`$, where the number of $`1`$’s is equal to $`n=\text{dim X}`$. Then $`X`$ is smooth. Let $`D=𝒪_X(1)`$. We have $`K_X(dn2)D`$, $`D^n=1`$ and $`|mD|`$ is not free if $`m`$ is odd and $`m<d`$. For example, if $`n=d2`$, then $`K_X0`$ and $`|nD|`$ is not free.
(3) Let $`d`$ be an integer such that $`d0`$ ($`\text{mod }3`$) and $`d4`$. Let $`X`$ be a general weighted hypersurface of degree $`3d`$ in $`(1,\mathrm{},1,3,d)`$ as in (2). Then $`X`$ is smooth. Let $`D=𝒪_X(1)`$. We have $`K_X(2dn3)D`$, $`D^n=1`$, and $`|mD|`$ is not free if $`m0`$ ($`\text{mod }3`$) and $`m<d`$. For example, $`|2D|`$ is not free, and $`|(d1)D|`$ is not free if $`d2`$ ($`\text{mod }3`$).
## 4 Minimal $`3`$-fold
We have so far an affirmative answer only for minimal varieties in the case of dimension $`3`$.
###### Proposition 4.1.
Let $`X`$ be a $`3`$-dimensional projective variety with at most canonical singularities, and $`D`$ a Cartier divisor. Assume that $`K_X`$ is nef, and $`DK_X`$ is nef and big. Then $`H^0(X,D)0`$.
###### Proof.
By a crepant blowings-up, we may assume that $`X`$ has only terminal singularities. Then we have $`\chi (𝒪_X)\frac{1}{24}K_Xc_2`$ by , and $`3c_2K_X^2`$ is pseudo-effective by Miyaoka (see also ). By the Riemann-Roch theorem, we calculate
$$\begin{array}{cc}\hfill h^0(X,D)& =\frac{1}{6}D^3\frac{1}{4}D^2K_X+\frac{1}{12}DK_X^2+\frac{1}{12}Dc_2+\chi (𝒪_X)\hfill \\ & =\frac{1}{12}(2DK_X)\{\frac{1}{6}D^2+\frac{2}{3}D(DK_X)+\frac{1}{6}(DK_X)^2\}\hfill \\ & +\frac{1}{72}(2DK_X)(3c_2K_X^2)+\frac{1}{24}K_Xc_2+\chi (𝒪_X)\hfill \\ & >0.\hfill \end{array}$$
###### Proposition 4.2.
Let $`X`$ be a complete variety of dimension $`3`$ with at most Gorenstein canonical singularities, and $`D`$ a Cartier divisor. Assume that $`K_X0`$ and $`D`$ is ample. Let $`Y|D|`$ be a general member whose existence is guaranteed by Proposition 4.1. Then the pair $`(X,Y)`$ is LC. In particular, $`Y`$ is SLC.
###### Proof.
Assume that $`(X,Y)`$ is not LC. Let $`c`$ be the LC threshold for $`(X,0)`$ so that $`c<1`$ and $`(X,cY)`$ is properly LC. Let $`W`$ be a minimal center. By Theorem 1, for any positive rational number $`ϵ`$, there exists an effective $``$-divisor $`B^{}`$ on $`W`$ such that $`(K_X+cY+ϵD)|_W_{}K_W+B^{}`$ and $`(W,B^{})`$ is KLT. By the perturbation technique, we may assume that $`W`$ is the only LC center for $`(X,cY+ϵD)`$ and there exists only one LC place $`E`$ above $`W`$ if we replace $`c`$ and $`ϵ`$ suitably.
Therefore, there exists a birational morphism $`\mu :YX`$ from a smooth projective variety such that we can write $`\mu ^{}(K_X+cY+ϵD)=K_Y+E+F`$, where the support of $`E+F`$ is a normal crossing divisor and the coefficients of $`F`$ are strictly less than $`1`$.
We consider an exact sequence
$$0_W(D)𝒪_X(D)𝒪_W(D|_W)0,$$
where $`_W=\mu _{}𝒪_Y(E)`$ is the ideal sheaf for $`W`$. Since $`D(K_X+cY+ϵD)`$ is ample, we have $`H^p(Y,\mu ^{}DE+\mathrm{}F\mathrm{})=0`$ and $`R^p\mu _{}𝒪_Y(\mu ^{}DE+\mathrm{}F\mathrm{})=0`$ for $`p>0`$ by the generalization of the Kodaira vanishing theorem. Since $`\mu _{}𝒪_Y(\mathrm{}F\mathrm{})=𝒪_X`$, we obtain $`H^1(X,_W(D))=0`$. Hence the homomorphism $`H^0(X,D)H^0(W,D|_W)`$ is surjective. We have $`H^0(W,D|_W)0`$ by Theorem 3.1. It follows that $`W`$ is not contained in the base locus of $`|D|`$, a contradiction. ∎
## 5 Weak log Fano varieties
The following is proved by Ambro . We shall give a shorter proof of the second part as an application of Theorem 3.1.
###### Theorem 5.1.
Let $`X,B`$ and $`D`$ be as in Conjecture 2.1. Assume that there exists a positive rational number $`r`$ such that $`r>dimX30`$ and $`(K_X+B)_{}rD`$. Then the following hold.
(1) $`H^0(X,D)0`$.
(2) Let $`Y|D|`$ be a general member. Then the pair $`(X,B+Y)`$ is PLT.
###### Proof.
(1) is proved in Lemma 2. We recall the proof for the convenience of the reader. We set $`n=dimX`$, $`d=D^n1`$, $`\beta =BD^{n1}0`$, and $`p(t)=\chi (X,tD)`$ for $`t`$. Since $`p(0)=1`$ and $`p(1)=p(2)=\mathrm{}=p(n+3)=0`$, we can write
$$\begin{array}{cc}\hfill p(t)& =\frac{d}{n!}t^n+\frac{\beta +dr}{2(n1)!}t^{n1}+\mathrm{}\hfill \\ & =\frac{d}{n!}(t+1)(t+2)\mathrm{}(t+n3)(t^3+at^2+bt+\frac{n(n1)(n2)}{d}\hfill \end{array}$$
for some numbers $`a,b`$. Hence
$$a=\frac{n(rn+5)}{2}3+\frac{\beta n}{2d}.$$
On the other hand, we have
$$0(1)^np(n+2)=\frac{d}{n(n1)}((n2)^2a(n2)+b)1.$$
Therefore
$$h^0(X,D)=p(1)=n1+(1)^np(n+2)+\frac{\beta +d(rn+3)}{2}>0.$$
(2) We may assume that $`D`$ is ample by the base point free theorem. Assume that $`(X,B+Y)`$ is not PLT, and let $`c1`$ be the LC threshold so that $`(X,B+cY)`$ is properly LC. Let $`W`$ be a minimal center. By , for any positive rational number $`ϵ`$, there exists an effective $``$-divisor $`B^{}`$ on $`W`$ such that $`(K_X+B+cY+ϵD)|_W_{}K_W+B^{}`$ and $`(W,B^{})`$ is KLT. Then $`(K_W+B^{})_{}r^{}D|_W`$ for $`r^{}=rcϵ`$. Since $`ϵ`$ can be arbitrarily small, we have $`r^{}>\text{dim }W3`$ and $`r^{}>1`$.
If $`\text{dim }W3`$, then $`H^0(W,D|_W)0`$ by (1). Otherwise, we have also $`H^0(W,D|_W)0`$ by Theorem 3.1. By the vanishing theorem, we have $`H^1(X,_W(D))=0`$ as in the proof of Proposition 4.2. From an exact sequence
$$0_W(D)𝒪_X(D)𝒪_W(D|_W)0,$$
it follows that $`W`$ is not contained in the base locus of $`|D|`$, a contradiction. ∎
The following result deals with the case which is just beyond the scope of Theorem 5.1.
###### Theorem 5.2.
Let $`X`$ be a complete variety of dimension $`4`$ with at most Gorenstein canonical singularities. Assume that $`DK_X`$ is ample. Then the following hold.
(1) $`H^0(X,D)0`$.
(2) Let $`Y|D|`$ be a general member. Then $`(X,Y)`$ is PLT, hence $`K_Y0`$ and $`Y`$ has only Gorenstein canonical singularities.
###### Proof.
We shall prove (1) and (2) simultaneously. Let $`m`$ be the smallest positive integer such that $`H^0(X,mD)0`$. We shall derive a contradiction from $`m>1`$. We take a general member $`Y|mD|`$.
Assume first that $`(X,Y)`$ is PLT and $`m>1`$. Then $`Y`$ is Gorenstein canonical. We have an exact sequence
$$0𝒪_X(D)𝒪_X((m1)D)𝒪_Y(K_Y)0.$$
We have $`\chi (X,𝒪_X(D))=1`$ and $`\chi (X,𝒪_X((m1)D))=0`$. On the other hand, we have $`\chi (Y,K_Y)=\frac{1}{24}K_Yc_20`$ by , a contradiction.
Next assume that $`(X,Y)`$ is not PLT and $`m1`$. Let $`c`$ be the LC threshold so that $`c1`$ and $`(X,cY)`$ is properly LC. Let $`W`$ be a minimal center. If $`\text{dim }W=3`$, then we have $`c\frac{1}{2}`$. By , for any positive rational number $`ϵ`$, there exists an effective $``$-divisor $`B^{}`$ on $`W`$ such that $`(K_X+cY+ϵD)|_W_{}K_W+B^{}`$ and $`(W,B^{})`$ is KLT. By the perturbation technique, we may assume that $`W`$ is the only center if we replace $`c`$ and $`ϵ`$ suitably.
We consider an exact sequence
$$0_W(mD)𝒪_X(mD)𝒪_W(mD|_W)0.$$
Since $`mD(K_X+cY+ϵD)`$ is ample, we have $`H^1(X,_W(mD))=0`$ by the vanishing theorem. Hence the homomorphism $`H^0(X,mD)H^0(W,mD|_W)`$ is surjective. If $`\text{dim }W2`$, then we have $`H^0(W,mD|_W)0`$ by Theorem 3.1. We shall also prove that $`H^0(W,mD|_W)0`$ in the case $`\text{dim }W=3`$. Then it follows that $`W`$ is not contained in the base locus of $`|mD|`$, a contradiction, and (1) and (2) are proved.
Assume that $`\text{dim }W=3`$. We set $`r^{}=cm1+ϵ`$ so that $`K_W+B^{}_{}r^{}D|_W`$. Let $`p(t)=\chi (W,tD|_W)`$. If we set $`d=(D|_W)^3>0`$ and $`\delta =B^{}(D|_W)^20`$, then
$$p(t)=\frac{d}{6}t^3+\frac{r^{}d+\delta }{4}t^2+bt+c$$
for some numbers $`b`$ and $`c`$ by the Riemann-Roch theorem. By the vanishing theorem, we have $`p(1)0`$ and $`p(m1)0`$, because $`r^{}<m1`$. Then
$$\begin{array}{cc}\hfill p(m)=& \frac{(m1)(m+1)d}{3}+\frac{(m+1)(r^{}d+\delta )}{4}\hfill \\ & +\frac{p(1)+(m+1)p(m1)}{m}>0.\hfill \end{array}$$
Thus we have $`H^0(W,mD|_W)0`$. ∎
Department of Mathematical Sciences, University of Tokyo,
Komaba, Meguro, Tokyo, 153-8914, Japan
kawamata@ms.u-tokyo.ac.jp
|
warning/0003/quant-ph0003073.html
|
ar5iv
|
text
|
# 1 (a) A welcher-weg biprism experiment to demonstrate violation of the Complementarity Principle; (b) a welcher-weg Mach-Zender interferometry experiment that can test the Complementarity Principle more rigorously.
Welcher Weg Experiments, Duality and the Orthodox Bohr’s Complementarity Principle
S. Bandyopadhyay<sup>1</sup><sup>1</sup>1Corresponding author. E-mail: bandy@quantum1.unl.edu
Department of Electrical Engineering, University of Nebraska, Lincoln, Nebraska 68588-0511, USA
## Abstract
In its most orthodox form, Bohr’s Complementarity Principle states that a quanton (a quantum system consisting of a Boson or Fermion) can either behave as a particle or as wave, but never simultaneosuly as both. A less orthodox interpretation of this Principle is the “duality condition” embodied in a mathematical inequality due to Englert \[B-G Englert, Phys. Rev. Lett., 77, 2154 (1996)\] which allows wave and particle attributes to co-exist, but postulates that a stronger manifestation of the particle nature leads to a weaker manifestation of the wave nature and vice versa. In this Letter, we show that some recent welcher weg (”which path”) experiments in interferometers and similar set-ups, that claim to have validated, or invalidated, the Complementarity Principle, actually shed no light on the orthodox interpretation. They may have instead validated the weaker duality condition, but even that is not completely obvious. We propose simple modifications to these experiments which we believe can test the orthodox Complementarity Principle and also shed light on the nature of wavefunction collapse and quantum erasure.
Keywords: Complementarity Principle, Wave-Particle Duality, Welcher Weg, Wavefunction Collapse
The orthodox Bohr’s complementarity principle states that a quanton can behave either as a particle or as a wave, but never as both at the same time. Welcher weg experiments conducted with two-path interferometers (or analogous set-ups) are a suitable vehicle to test this strict complementarity between the wave- and particle-nature. If one can determine - even in principle - which of the two paths in the interferometer was traversed by the quanton, then the entity behaves as a “particle” since a “wave” would have sampled both paths simultaneously. In this case, there should be no interference (or any other wave-like behavior) if the orthodox Complementarity Principle holds. On the other hand, if there is interference, then the wave property is intact in which case it should have been impossible to discern the particle attribute, i.e. to tell which path was traversed. The orthodox Complementarity principle therefore allows only sharp wave or sharp particle attribute, but not both. A somewhat tempered version of the Complementarity Principle is the duality principle due to Englert which states that a quantum system can simultaneosuly exhibit wave and particle behavior, but sharperning of the wave character blurs the particle character and vice versa. In fact, Englert derives an inequality
$$P^2+V^21$$
(1)
where $`P`$ is a measure of the “which-path” information (particle attribute) and $`V`$ is a measure of the “(interference) fringe-visibility” (wave attribute). Equation (1) immediately shows that stronger wave or stronger particle behavior can be manifested only at the expense of each other.
Welcher Weg experiments that question the orthodox Complementarity Principle
Experiments purported to demonstrate violation of the orthodox version of the Complementarity Principle were proposed and carried out in the past. Ghose, Home and Agarwal had proposed a biprism experiment schematically depicted in Fig. 1(a). A single photon source emits a single photon which is split into orthogonal states $`\psi _r`$ and $`\psi _t`$ by a 50:50 beam splitter. They are detected by two photon detectors $`D_r`$ and $`D_t`$. If the photon behaves truly as a particle, then it should be detected at either $`D_r`$ or $`D_t`$ (but never at both) since a particle cannot traverse two paths simultaneously. That is, there should be perfect anti-coincidence between $`D_r`$ and $`D_t`$, or, in other words, either $`D_r`$ or $`D_r`$ will click but both will never click in between the arrival of two successive photons. The clever twist in this experiment, motivated by an experiment performed in the 19th century by Jagdish Chandra Bose , is the placement of the biprism with a small tunneling gap in the path of the transmitted photon. If $`D_t`$ clicks and $`D_r`$ does not, then we have made a “which path” determination (the particle took the path of transmission as opposed to reflection) and a sharp particle nature is demonstrated . Yet, to arrive at $`D_t`$, the particle must have tunneled through the biprism and tunneling is a sharp wave attribute. In this experiment, later conducted by Mizobuchi and Ohtake , perfect anti-coincidence was found between $`D_t`$ and $`D_r`$ demonstrating the particle nature. Yet, the very fact that $`D_t`$ ever clicked required tunneling and hence the existence of a wave nature. It was claimed that in this experiment, a photon was behaving both sharply as a particle and as a wave in violation of Bohr’s Complementarity Principle. A slight modification of this experiment has been proposed by Rangwala and Roy where interference is used instead of tunneling to showcase the wave-like behavior. They claimed that quantum mechanics does not prohibit the demonstration of simultaneous wave and particle behavior; rather, it prohibits their simultaneity only when the wave and particle attributes are “complementary” in the sense that projection operators associated with them do not commute. This is actually consistent with Englert’s work in that Englert takes pain to point out that Equation (1) does not rely on Heisneberg type uncertainty, i.e. the following relation need not hold:
$$\mathrm{\Delta }P\mathrm{\Delta }V\frac{1}{2}|<[P,V]>|.$$
(2)
In the experiments of refs. , the wave and particle behavior supposedly are not truly complementary and hence not subject to the restrictions of the orthodox Complementarity Principle.
A mathematical framework to determine true “complementarity” between wave and particle attributes was first addressed by Kar. et. al. . Complementary observables as those whose projection operators do not commute, i.e. have no common eigenvectors. In the experiments of refs. , the Hilbert spaces $`H_r`$ and $`H_t`$ associated with the reflected state $`\psi _r`$ and the transmitted state $`\psi _t`$ are orthogonal (since there is always anti-coincidence between reflection and transmission). Hence the projection operators $`P_r`$ and $`P_t`$, corresponding to reflection and transmission respectively, always commute. If we assume that the wave property (tunneling) is represented by some projection operator $`P_{wave}`$, its Hilbert space is contained within the Hilbert space of $`P_t`$. i.e. $`<\psi |P_{wave}|\psi >`$ $``$ $`<\psi |P_t|\psi >`$ and is hence orthogonal to $`H_r`$. Thus, $`P_{wave}`$ commutes with $`P_r`$. But $`P_{wave}`$ must also commute with $`P_t`$ since every state $`\psi _t`$ that tunnels through the biprism and reaches $`D_t`$ is a common eigenvector of these two operators. Thus, $`P_{wave}`$, $`P_t`$ and $`P_r`$ all commute with each other. Hence the sharp wave property (tunneling) and the sharp particle property (anti-coincidence between detectors $`D_t`$ and $`D_r`$) are not complementary and their simultaneous observation is not prohibited by the Complementarity Principle (equation (1) however, must still be obeyed, but the experiments did not test this inequality). In concluding their paper, Kar et. al. point out that the experiments of refs. do not test wave and particle properties that are complementary and hence can draw no conclusion about the validity or invalidity of the “orthodox” Complementarity Principle.
An alternate welcher weg experiment to test the orthodox Complementarity Principle
We propose an alternate welcher weg experiment where sharp particle behavior and sharp wave behavior would be complementary in the sense defined by Kar and co-workers. Consequently, this is an unambiguous experiment where simultaneous exhibition of sharp particle- and wave-character will give lie to the orthodox version of the Complementarity Principle. We point out that these experiments are worth conducting since their outcome is by no means a foregone conclusion. The orthodox Complementarity Principle is not sacrosanct (even though it is viewed by some as a cornerstone of the Copenhagen interpretation of quantum mechanics); viewpoints due to Einstein and DeBroglie do not subscribe to the Complementarity Principle .
Consider a Mach-Zender type interferometer as shown in Fig. 1(b). Proximity photon detectors $`D_1`$ and $`D_2`$ are placed near each limb as shown. A single photon source injects photons at terminal $`S`$ which reaches a screen $`D`$ after traveling along the two possible paths comprising the arms of the ring (actually there are a denumerably infinite number of paths possible if we take into account multiple reflections, but they are not important in this context).
To demonstrate an invalidation of the orthodox Complementarity Principle, we need to demonstrate two effects simultaneously:
1. Perfect anti-coincidence between $`D_1`$ and $`D_2`$ (sharp particle nature).
2. Existence of an interference pattern at $`D`$ (sharp wave nature)
In this example, the projection operators $`P_1`$ and $`P_2`$ corresponding to the traversal of the two paths of the interferometer are orthogonal, but $`P_{wave}`$ is manifestedly not orthogonal to either one of them. Note that the Hilbert space $`H_{wave}`$ is not contained within either $`H_1`$ or $`H_2`$ since neither of the two paths alone is sufficient to cause interference (both paths are needed). Also note that $`\psi _1`$ and $`\psi _2`$ (while eigenfunctions of $`P_1`$ and $`P_2`$ respectively), are not eigenfunctions of $`P_{wave}`$. Hence, unlike in the experiments in ref. , $`P_{wave}`$ does not commute with $`P_1`$ and $`P_2`$ and therefore the wave and particle properties are indeed complementary. Thus, their simultaneous manifestation will definitely give lie to the Complementarity Principle.
Welcher weg experiments that claim to have validated Complementarity or Duality
The duality principle embodied in Englert’s inequality \[Equation (1)\] was verified in atom interference experiments and perhaps even in recent experiments conducted with electrons traversing an Aharonov-Bohm (A-B) quantum interferometer whose one arm contained a quantum dot (QD) with a nearby quantum point contact (QPC) . The QD has a non-critical, peripheral role; it merely serves to trap an electron traveling that path long enough for the QPC to detect it. The trapping changes the transmission probability through the QPC (and hence its conductance) thus allowing “which path” detection.
The experiment showed that the A-B interference was diluted if one could even in principle detect which path was traversed, irrespective of whether the detection actually took place. When the QPC detector was turned on, the interference peaks were diluted regardless of whether one monitored or not any change in the QPC conductance caused by a fleeting electron in the nearby path. This result, remarkable as it is, does not shed any light on the orthodox version of Bohr’s complementarity principle. Validation of the orthodox version would require demonstrating complete vanishing of interference along with the demonstration of perfect anti-coincidence between two QPC detectors placed near the two paths of the interferometer. Unfortunately, this was not attempted in the experiment. Second, the experimental result may be consistent with the Duality Principle \[Equation (1)\] (like ref. ), but does not quite validate it either (validation is a stronger condition than consistence). We say this because turning the QPC on introduces an asymmetry into the interferometer (e.g. decrease the transmittivity of one path relative to another) and this alone can cause a dilution of the interference as we show in the appendix.
We suggest some modifications to this experiment to test the orthodox version of Bohr’s complementarity principle. The configuration of the interferometer used in ref. is shown in Fig. 2(a). It was defined on a high mobility two-dimensional electron gas using standard split-gate technology. The modifications are the following:
1. Introduce electrons one at a time into the interferometer using single charge tunneling (single electron pump or turnstile) . This can be done by delineating a small island at the mouth of the emitter and isolating it from the emitter with a tunnel barrier whose resistance is much higher than $`h/e^2`$ ($`h`$ = Planck’s constant and $`e`$ = single electron charge). Modulating the barrier with an external potential can cause electrons to be injected one at a time. There may be other, more complicated, schemes for pumping single electrons that may be more appropriate depending on the measurement approach.
2. Place two tunnel barriers in the two arms of the A-B interferometer near the detector $`C`$. They can also be defined by split gates and are shown as striped metal gates in Fig. 2(b). If an electron is detected and it collapses to a “particle”, it cannot tunnel through the barrier and reach the collector C since only waves tunnel and particles do not tunnel. Therefore, it must be reflected back into the emitter. Thereafter it cannot take the other path to the collector either because there is a tunnel barrier in that path as well. Consequently there should be no current. The evolution of the particle nature should not only destroy the interference, but actually reduce the current to zero.
3. Place a narrow slit defined by split gates in between the QPC detectors and the detector C. This connects the two paths. Normally this slit is pinched off by the split gates to isolate the two paths, but it can be opened (made conducting) by the split gate voltages, if necessary. This arrangement is relevant to the study of ”quantum erasure” as shown later.
4. Place two QD-QPC detectors alongside both arms. They serve two purposes. They can make the interferometer symmetric and they could also be used to demonstrate a perfect anti-coincidence between the detection events in the two paths (detection events correspond to a change in the QPC current when an electron passes by).
With the above modifications, the A-B interferometer can now be used to remove all the objections that we raised in relation to the experiment of ref. . The interferometer is nominally symmetric as long as both detectors are turned on. Perfect anti-coincidence can be demonstrated if it exists, and the tunnel barriers in the interferometer arms will conclusively demonstrate if the wave nature is present or absent. Thus sharp particle- and wave-behavior can be exhibited with no ambiguity.
Which path determination and quantum erasure
In addition to other differences, a major point of departure between the experiment of ref. and the above experiment is that the photon experiment purports to demonstrate wave behavior before the particle behavior is evidenced in the transmission path, whereas here the opposite is proposed. The tunneling gap in Fig. 1(a) precedes the detector $`D_t`$. In our experiment (Fig. 2(b)), the reverse is true; the detector precedes the tunnel barrier. Thus, simultaneous exhibition of wave- and particle- nature in our proposed experiment would require an electron to first “become” a particle and then be reincarnated as a wave. “Which path” determination does not prohibit this since entanglement of the electron’s wavefunction with that of the detector does not cause the entangled pure state to evolve into a mixed state irreversibly. We show this below.
Let $`\psi _1`$ and $`\psi _2`$ be the wavefunctions in the two arms of the interferometer and $`\theta `$ is the A-B phase difference between the two arms. Then, neglecting multiple reflection effects , the wavefunction $`\psi `$ at the detector D is
$$\psi =\psi _1+e^{i\theta }\psi _2$$
(3)
The current density at the detector is
$$J=\frac{ie\mathrm{}}{2m^{}}\left[\psi (\psi )^{}\psi ^{}\psi \right]$$
(4)
Thus, for propagating states (plane waves along the direction of propagation), Equations (1) and (2) yield
$$J|\psi |^2=|\psi _1|^2+|\psi _2|^2+e^{i\theta }\psi _1^{}\psi _2+e^{i\theta }\psi _2^{}\psi _1$$
(5)
The last two terms account for the A-B interference.
A fundamental result of quantum measurement theory is that if a detector tries to detect which arm was traversed by the interferometer, the wavefunction of the detector becomes entangled with that of the electron. The entangled wavefunction can be written as
$$\mathrm{\Phi }=\psi _1|1>+e^{i\theta }\psi _2|2>$$
(6)
where the wavefunctions $`|1>`$ and $`|2>`$ span the Hilbert space of the detector. The state $`|1>`$ corresponds to detecting the particle in path 1 and $`|2>`$ corresponds to detecting the particle in path 2.
The current density associated with this wave function is
$`J_{entangled}|\mathrm{\Phi }|^2=|\psi _1|^2<1|1>+|\psi _2|^2<2|2>`$
$`+<1|2>e^{i\theta }\psi _1^{}\psi _2+<2|1>e^{i\theta }\psi _2^{}\psi _1`$ (7)
If the detector is an “unambiguous” detector which unambiguously determines which path is traversed by the particle, then $`|1>`$ and $`|2>`$ are orthonormal and hence the interference terms vanish in Equation (5) and the wave behavior is lost. This is interpreted as dephasing or collapse.
$$J_{collapsed}\left[|\psi _1|^2+|\psi _2|^2\right]$$
(8)
In truth however, the collapse is not quite irreversible since if we design an experiment whose result is the probability of a particular outcome of the welcher weg determination and finding the detector in the symmetric state $`[|1>+|2>]`$, we find
$`J_{erase}`$ $``$ $`|[<1|+<2|]|\mathrm{\Phi }>|^2`$ (9)
$`=`$ $`|[<1|+<2|]|[\psi _1|1>+e^{i\theta }\psi _2|2>]|^2`$
$`=`$ $`[\psi _1|^2+|\psi _2|^2+e^{i\theta }\psi _1^{}\psi _2+e^{i\theta }\psi _2^{}\psi _1]`$
$`=`$ $`|\psi _1+e^{i\theta }\psi _2|^2=|\mathrm{\Phi }|^2`$
Thus, the original wavefunction of Equation (4) (along with the interference terms) is regenerated from the entangled wavefunction once we choose to erase the ”which path” information. This is termed “quantum erasure” and is possible because the entangled state of Equation (4) is still a pure state and not a mixed state. Evolution of a pure state into a mixed state (termed orthodox collapse) is irreversible, but that is not the case here.
We can test the quantum erasure by opening the slit which, in principle, allows conduction between the two paths and hence effectively erases the which path information. This can then regenerate the wave nature and allow tunneling through the tunnel barriers in the two paths. Consequently, the current (which is ideally zero) with the slit closed and the QPC detectors on, will rise to a non-zero value when the slit is opened. This will be a demonstration of quantum erasure.
In conclusion, we have proposed modifications to some welcher weg experiments that we believe can rigorously test the Complementarity Principle.
Appendix
We show that introducing asymmetry between the two paths of an interferometer degrades the visibility of the interferograms independent of any other effect.
Consider a two-path interferometer. The current between the source and the detector can be written as (neglecting multiple reflection of the electron wave between the emitter and collector (due to geometric discontinuities)
$$I=|i_1+i_2e^{i\theta }|^2$$
(10)
where $`i_1(i_2)`$ is the complex amplitude of the current in path 1(2) and we assume that $`|i_1|`$ = $`\alpha |i_2|`$ (0 $``$ $`\alpha `$ 1), i.e. the transmissivity of path 2 is equal to or larger than that of path 1. Thus, $`\alpha `$ is a measure of the asymmetry; the farther $`\alpha `$ is from unity, the more is the asymmetry. The angle $`\theta `$ is the phase difference between the two paths. We now find
$$I=|i|^2[1+\alpha ^2+2\alpha cos\theta ]$$
(11)
We can adopt a suitable metric for the visibility of the oscillation. This could be the relative modulation $`M`$ of the current
$$M=\frac{I_{max}I_{min}}{I_{max}}=\frac{4\alpha }{(1+\alpha )^2}$$
(12)
From the above equation, we can see that $`M`$ is 100% if $`\alpha `$ = 1; otherwise, $`M`$ decreases as $`\alpha `$ decreases. Thus, increasing the asymmetry in a two-path interferometer degrades the visibility of the interference oscillations independent of any other effect.
|
warning/0003/math0003114.html
|
ar5iv
|
text
|
# Theorem 1.1.
### Acknowledgements
For very large $`D`$, Theorem 1.1 was originally obtained by D. Rohrlich \[Ro4\]. We are indebted to him for sharing his method with us, as well as for allowing us to mention several of his results here. Also, we thank him for his inspiration and for suggesting this problem to us. We are grateful to M. Baker, B. Conrad, N. Elkies, B. Gross, K. Rubin, Z. Rudnick, D. Zagier, and S. Zhang for discussions, and to Harvard University for their hospitality. S.M. was partially supported by an NSF post-doctoral fellowship as well as a Yale Hellman fellowship. T.Y. was partially supported by an AMS Centennial fellowship and NSF grant DMS-9700777.
## 2 Notation and Strategy
We first recall some facts about canonical Hecke characters from \[Ro2\]. They exist if and only if $`D3(\text{mod}4)`$ or is a multiple of 8. Multiplying a canonical character by an ideal class character always yields another canonical character. This operation preserves the root number and defines natural families of canonical Hecke characters. When $`D3(\text{mod}4)`$, there is exactly one family and it has root number $`\left(\frac{2}{D}\right)`$; when $`D`$ is a multiple of 8, there are two families – one has root number 1 and the other has root number -1.
To avoid confusion, we will sometimes write $`\chi _{\text{can}}`$ for a canonical Hecke character of $`K`$. In this paper we consider Hecke characters $`\chi `$ of $`K`$ satisfying conditions (1.1) and (1.2), which are always of the form
$$\chi _{D,d}=\chi _{\text{can}}(ϵ_dN_{K/}).$$
(2.1)
Here $`d`$ is a fundamental discriminant and $`ϵ_d=(\frac{d}{})`$ is the quadratic dirichlet character with conductor $`d`$, prime to $`D`$. The root number $`W(\chi )`$ is explicitly computed in \[Ro2\]. In particular, when $`D`$ is odd
$$W(\chi _{D,d})=\left(\frac{2}{D}\right)\text{sign}(d).$$
(2.2)
From now on we will assume that $`W(\chi )=1`$. Set
$$B=\sqrt{DN𝔣}=\{\begin{array}{cc}D|d|,\hfill & D\text{ odd}\hfill \\ 2D|d|,\hfill & 8D.\hfill \end{array}$$
(2.3)
The Hecke L-function is defined as
$`L(s,\chi )`$ $`={\displaystyle \underset{𝔞\text{ integral}}{}}\chi (𝔞)(N𝔞)^s`$
$`={\displaystyle \underset{\text{ideal classes }C}{}}L(s,\chi ,C),`$
where $`L(s,\chi ,C)`$ is the partial L-series summed over integral ideals in $`C`$. Their completed L-functions are defined by
$$\mathrm{\Lambda }(s,\chi )=\left(\frac{B}{2\pi }\right)^s\mathrm{\Gamma }(s)L(s,\chi )$$
and
$$\mathrm{\Lambda }(s,\chi ,C)=\left(\frac{B}{2\pi }\right)^s\mathrm{\Gamma }(s)L(s,\chi ,C).$$
###### Lemma 2.1.
When $`W(\chi )=1`$, $`\mathrm{\Lambda }^{}(1,\chi )=0`$ if and only if $`\mathrm{\Lambda }^{}(1,\chi ,C)=0`$ for each ideal class $`C`$ of $`K`$.
Proof: Associated to $`\chi `$ is a cuspidal new form $`f`$ of weight 2 and level $`B^2`$ such that $`L(s,f)=L(s,\chi )`$. So Corollary 2 of \[GZ\] implies that $`\mathrm{\Lambda }^{}(1,\chi )=0`$ if and only if $`\mathrm{\Lambda }^{}(1,\chi ^\sigma )=0`$ for every $`\sigma \text{Gal}(\overline{}/)`$. On the other hand, by Theorem 1 of \[Ro3\],
$$\{\chi ^\sigma :\sigma \text{Gal}(\overline{}/K)\}=\{\chi \varphi :\varphi \text{ is an ideal class character of }K\},$$
and $`L(s,\chi )=L(s,\overline{\chi })`$ by (1.1). Thus $`\mathrm{\Lambda }^{}(1,\chi )=0`$ if and only if $`\mathrm{\Lambda }^{}(1,\chi \varphi )=0`$ for all ideal class characters $`\varphi `$ of $`K`$. The ideal class characters are linearly independent and
$$L(s,\chi \varphi )=\underset{C}{}\varphi (C)L(s,\chi ,C),$$
so the lemma follows. $`\mathrm{}`$
To prove $`\mathrm{\Lambda }^{}(1,\chi )0`$, it now suffices to show $`\mathrm{\Lambda }^{}(1,\chi ,c_1)0`$ for the trivial class $`c_1`$ (i.e. the class of principal ideals). Since the root number $`W(\chi )=1`$, we have the functional equation
$$\mathrm{\Lambda }(s,\chi ,c_1)=\mathrm{\Lambda }(2s,\chi ,c_1).$$
(2.4)
By Cauchy’s theorem
$$\mathrm{\Lambda }^{}(1,\chi ,c_1)=\frac{1}{2\pi i}\left(_{2i\mathrm{}}^{2+i\mathrm{}}\mathrm{\Lambda }(s,\chi ,c_1)\frac{ds}{(s1)^2}_i\mathrm{}^{+i\mathrm{}}\mathrm{\Lambda }(s,\chi ,c_1)\frac{ds}{(s1)^2}\right).$$
Applying (2.4) we arrive at the formula
$$\frac{1}{2}\mathrm{\Lambda }^{}(1,\chi ,c_1)=\frac{1}{2\pi i}_{2i\mathrm{}}^{2+i\mathrm{}}\mathrm{\Lambda }(s,\chi ,c_1)\frac{ds}{(s1)^2}.$$
(2.5)
It is clear from property (1.2) of $`\chi `$ that there is a quadratic character $`ϵ`$ of $`(𝒪/𝔣)^{}`$ such that
$$\chi (\alpha 𝒪)=ϵ(\alpha )\alpha .$$
(2.6)
We can express $`L(s,\chi ,c_1)`$ as a sum over real and complex ideals:
$$L(s,\chi ,c_1)=\underset{n=1}{\overset{\mathrm{}}{}}ϵ(n)n^{12s}+\underset{n=1}{\overset{\mathrm{}}{}}a_nn^s,$$
(2.7)
where
$$a_n=\underset{\stackrel{N𝔞=n,𝔞\overline{𝔞}}{\stackrel{\text{principal,}}{\text{integral}}}}{}\chi (𝔞)=\underset{\stackrel{u^2+Dv^2=4n}{u,v>0}}{}ϵ\left(\frac{u+\sqrt{D}v}{2}\right)u.$$
(2.8)
Let
$$f(x)=\frac{\mathrm{\Gamma }(0,x)}{x}=\frac{1}{x}_x^{\mathrm{}}e^t\frac{dt}{t}$$
(2.9)
be the inverse Mellin transform of $`\frac{\mathrm{\Gamma }(s)}{(s1)^2}`$. Indeed
$`{\displaystyle _0^{\mathrm{}}}f(x)x^s{\displaystyle \frac{dx}{x}}`$ $`={\displaystyle _0^{\mathrm{}}}{\displaystyle _x^{\mathrm{}}}x^{s1}e^t{\displaystyle \frac{dt}{t}}{\displaystyle \frac{dx}{x}}`$
$`={\displaystyle _0^{\mathrm{}}}{\displaystyle _0^t}{\displaystyle \frac{dx}{x}}x^{s1}e^t{\displaystyle \frac{dt}{t}}`$
$`={\displaystyle \frac{1}{s1}}{\displaystyle _0^{\mathrm{}}}t^{s1}e^t{\displaystyle \frac{dt}{t}}=`$ $`{\displaystyle \frac{\mathrm{\Gamma }(s1)}{(s1)}}={\displaystyle \frac{\mathrm{\Gamma }(s)}{(s1)^2}},`$
so
$$f(y)=\frac{1}{2\pi i}_{Re(s)=2}\frac{\mathrm{\Gamma }(s)}{(s1)^2}y^s𝑑s.$$
(2.10)
Combining (2.5), (2.7), and (2.10) we obtain
$$\frac{1}{2}\mathrm{\Lambda }^{}(1,\chi ,c_1)=\stackrel{R}{\stackrel{}{\underset{n=1}{\overset{\mathrm{}}{}}ϵ(n)nf(2\pi n^2/B)}}+\stackrel{C}{\stackrel{}{\underset{n=1}{\overset{\mathrm{}}{}}a_nf\left(\frac{2\pi n}{B}\right)}}.$$
(2.11)
Formula (2.11) is essentially due to Rohrlich (\[Ro4\]), except that he expressed $`R`$ in terms of dirichlet L-functions.
### Examples: $`D=8`$ and $`11`$
We will now illustrate (2.11) with the first two discriminants which occur. Since both $`(\sqrt{8})`$ and $`(\sqrt{11})`$ have class number 1,
$$\mathrm{\Lambda }^{}(1,\chi ,c_1)=\mathrm{\Lambda }^{}(1,\chi ).$$
In order to compute it using (2.11), we must first describe the character $`ϵ:\left(𝒪/𝔣\right)^{}\{\pm 1\}`$. When $`D=8`$, $`𝔣=2\sqrt{D}𝒪=8\sqrt{32}`$, and $`\left(𝒪/𝔣\right)^{}`$ is generated by $`\left(/8\right)^{}`$ and $`1+\sqrt{2}`$. The character $`ϵ(n)`$ must restrict to $`\left(\frac{8}{n}\right)`$ for $`n\left(/8\right)^{}`$, and is thus determined by its value on $`1+\sqrt{2}`$. In fact, $`W(\chi )=ϵ(1+\sqrt{2})`$, so in our case the values of $`ϵ`$ on the relatively-prime residue classes are given in the following chart:
| $`ϵ(u+v\sqrt{2})`$ | $`u`$ | 1 | 3 | 5 | 7 |
| --- | --- | --- | --- | --- | --- |
| $`v`$ | | | | | |
| 0 | | 1 | 1 | -1 | -1 |
| 1 | | -1 | -1 | 1 | 1 |
| 2 | | -1 | -1 | 1 | 1 |
| 3 | | -1 | -1 | 1 | 1 |
When $`D=11`$, $`ϵ\left(\frac{u+\sqrt{11}v}{2}\right)=\left(\frac{2u}{11}\right)`$. We now compute $`L^{}(1,\chi )`$ for these canonical characters, and check (2.11) by comparing the known values of $`L^{}(1,E)`$ for the associated elliptic curves $`E`$.
| | $`D=8`$ | $`D=11`$ |
| --- | --- | --- |
| Term $`R`$ with $`n^250`$ | 1.82582357875147 | 0.81497705252487 |
| Term $`C`$ with $`n50`$ | -0.28596530872740 | -0.0600975766040368 |
| $`L^{}(1,\chi )=\left(\frac{2\pi }{B}\right)\mathrm{\Gamma }(1)\mathrm{\Lambda }^{}(1,\chi )`$ | 1.209401857169272 | 0.862372296690396 |
| $`=\frac{4\pi }{B}(R+C)`$ | | |
| Associated curve $`E`$ | $`y^2=x^3+4x^2+2x`$ | $`y^2+y=x^3x^27x+10`$ |
| | (\[Cr\], curve 256A) | (\[Cr\], curve 121B) |
| $`L^{}(1,E)`$ from \[Cr\] | 1.2094018572 | .8623722967 |
### Proof of Theorem 1.1
By Lemma 2.1 and (2.11), it suffices to prove $`R>|C|`$. In the next section, we will prove that $`R`$ is bounded below by
$`R`$ $`>`$ $`{\displaystyle \underset{n=1}{\overset{\mathrm{}}{}}}\lambda (n)nf\left({\displaystyle \frac{2\pi n^2}{B}}\right)`$ (2.12)
$`>`$ $`.5235B.8458B^{3/4}.3951B^{1/2}.`$
Here $`\lambda (n)`$ is Liouville’s function – the completely multiplicative function which is -1 at each prime.
In Section 4 we consider the special case $`d=1`$, and bound term $`C`$ by Proposition 4.1:
$$|C|<\{\begin{array}{cc}.2369D,\hfill & D\text{ even}\hfill \\ .0269D,\hfill & D\text{ odd}.\hfill \end{array}$$
(2.13)
Having collected these estimates, proving $`R>|C|`$ is a simple calculation. Indeed, if $`D24`$ is even, then $`B=2D`$, and
$$R>.5235(2D).8458(2D)(48)^{1/4}.3951(2D)(48)^{1/2}=.2902D,$$
so
$$R>.2369D>|C|.$$
If $`D19`$ is odd then
$$R>.5235D.8458D19^{1/4}.3951D19^{1/2}>.0277D,$$
$$R>.0269D>|C|.$$
There are only two values of $`D`$ not covered by this argument: $`D=8`$ and $`11`$, which were dealt with in the examples.
$`\mathrm{}`$
### Quadratic twists
To prove non-vanishing for quadratic twists of canonical characters, the bound (2.13) is not useful. In Section 5 we apply Rohrlich’s method to obtain the following bound on $`C`$ for $`\chi _{D,d}`$ (Proposition 5.1):<sup>1</sup><sup>1</sup>1The notation $`AB`$ means $`A=O(B)`$, i.e. there exists a positive constant $`C`$ such that $`|A|CB`$.
$$|C|D^{15/16+\delta }|d|^{51/16+\delta },$$
(2.14)
where $`\delta >0`$ is arbitrary and the implied constant depends only on it. Combining (2.11), (2.12), and (2.14) we conclude
###### Theorem 2.2.
For any fixed $`\delta >0`$,
$$\mathrm{\Lambda }^{}(1,\chi _{D,d})0$$
for $`|d|D^{1/35\delta }`$ and $`W(\chi _{D,d})=1`$.
Remarks. When the root number $`W(\chi )=1`$, similar non-vanishing results for twists were obtained in \[Ro1\], \[RVY\], and \[Ya\] for the central L-value.
For canonical Hecke characters, Rohrlich (\[Ro4\]) computed $`R`$ as a contour integral of dirichlet L-functions. By shifting contours, $`R`$ can be expressed as the sum of a residue and a remainder integral. He showed the residue is of size $`D`$, and used Burgess’ sub-convexity estimate to bound the remainder integral by $`D^\alpha `$, $`\alpha <1`$. Also, he used the method in Section 5 to show $`CD^\alpha `$. The power of the main term is larger than that of the other two terms, and positivity follows for large $`D`$. However, the implied constants one gets for these estimates are quite unfavorable. In our proof of Theorem 1.1 we sacrifice the gain in the powers of $`D`$ for a tie – in favor of better constants.
## 3 The Main Term $`R`$
The purpose of this section is to prove (2.12). We will show term $`R`$ is large and positive by eventually bounding it from below by the following sum.
###### Proposition 3.1.
$$\underset{n=1}{\overset{\mathrm{}}{}}\lambda (n)nf\left(\frac{2\pi n^2}{x}\right)$$
is always positive for $`x>0`$ and in fact
$$\underset{n=1}{\overset{\mathrm{}}{}}\lambda (n)nf\left(\frac{2\pi n^2}{x}\right)>.5235x.8458x^{3/4}.3951x^{1/2}$$
(3.1)
for $`x>1`$.
Proof: Using (2.10) and the identity
$$\underset{n=1}{\overset{\mathrm{}}{}}\lambda (n)n^s=\frac{\zeta (2s)}{\zeta (s)},$$
we write
$`{\displaystyle \underset{n=1}{\overset{\mathrm{}}{}}}\lambda (n)nf\left({\displaystyle \frac{2\pi n^2}{x}}\right)`$ $`={\displaystyle \frac{1}{2\pi i}}{\displaystyle _{\text{Re }s=2}}\left({\displaystyle \frac{x}{2\pi }}\right)^s{\displaystyle \frac{\mathrm{\Gamma }(s)}{(s1)^2}}{\displaystyle \frac{\zeta (4s2)}{\zeta (2s1)}}𝑑s`$
$`={\displaystyle \frac{1}{2\pi i}}{\displaystyle _\gamma }\left({\displaystyle \frac{x}{2\pi }}\right)^s{\displaystyle \frac{\mathrm{\Gamma }(s)}{(s1)^2}}{\displaystyle \frac{\zeta (4s2)}{\zeta (2s1)}}𝑑s`$
$`+Res_{s=1}\left({\displaystyle \frac{x}{2\pi }}\right)^s{\displaystyle \frac{\mathrm{\Gamma }(s)}{(s1)^2}}{\displaystyle \frac{\zeta (4s2)}{\zeta (2s1)}}`$
$`+Res_{s=3/4}\left({\displaystyle \frac{x}{2\pi }}\right)^s{\displaystyle \frac{\mathrm{\Gamma }(s)}{(s1)^2}}{\displaystyle \frac{\zeta (4s2)}{\zeta (2s1)}}.`$
Here $`\gamma =C_1C_2C_3C_4C_5`$ is the contour consisting of the union of the following five line segments:
$`C_1`$ $`\text{from }1i\mathrm{}`$ $`\text{to }17i,`$
$`C_2`$ $`\text{from }17i`$ $`\text{to }{\displaystyle \frac{1}{2}}7i,`$
$`C_3`$ $`\text{from }{\displaystyle \frac{1}{2}}7i`$ $`\text{to }{\displaystyle \frac{1}{2}}+7i,`$
$`C_4`$ $`\text{from }{\displaystyle \frac{1}{2}}+7i`$ $`\text{to }1+7i,`$
$`C_5`$ $`\text{from }1+7i`$ $`\text{ to }1+i\mathrm{}`$
(7 is chosen because the first critical zeroes of $`\zeta (s)`$ are approximately $`\frac{1}{2}\pm 14.13472i`$). The residue at $`s=1`$ is $`\frac{\pi x}{6}.523599x`$ and the residue at $`s=3/4`$ is
$$Res_{s=3/4}\left(\frac{x}{2\pi }\right)^s\frac{\mathrm{\Gamma }(s)}{(s1)^2}\frac{\zeta (4s2)}{\zeta (2s1)}=\frac{2^{5/4}x^{3/4}\mathrm{\Gamma }(3/4)}{\pi ^{3/4}\zeta (1/2)}.845767x^{3/4}.$$
One can easily estimate the integrals over $`\gamma `$ as follows.<sup>2</sup><sup>2</sup>2All computations were done using Mathematica v4.0 on an Intel Celeron processor under Windows 98. First,
$`\left|{\displaystyle _{C_1}}\left({\displaystyle \frac{x}{2\pi }}\right)^s{\displaystyle \frac{\mathrm{\Gamma }(s)}{(s1)^2}}{\displaystyle \frac{\zeta (4s2)}{\zeta (2s1)}}𝑑s\right|`$
$`=\left|{\displaystyle _{C_5}}\left({\displaystyle \frac{x}{2\pi }}\right)^s{\displaystyle \frac{\mathrm{\Gamma }(s)}{(s1)^2}}{\displaystyle \frac{\zeta (4s2)}{\zeta (2s1)}}𝑑s\right|`$
$`{\displaystyle \frac{x}{2\pi }}{\displaystyle _{t=7}^{\mathrm{}}}{\displaystyle \frac{|\mathrm{\Gamma }(1+it)|}{t^2}}{\displaystyle \frac{|\zeta (2+4it)|}{|\zeta (1+2it)|}}𝑑t`$
$`x(510^7).`$
Next
$`\left|{\displaystyle _{C_2}}\left({\displaystyle \frac{x}{2\pi }}\right)^s{\displaystyle \frac{\mathrm{\Gamma }(s)}{(s1)^2}}{\displaystyle \frac{\zeta (4s2)}{\zeta (2s1)}}𝑑s\right|`$
$`=\left|{\displaystyle _{C_4}}\left({\displaystyle \frac{x}{2\pi }}\right)^s{\displaystyle \frac{\mathrm{\Gamma }(s)}{(s1)^2}}{\displaystyle \frac{\zeta (4s2)}{\zeta (2s1)}}𝑑s\right|`$
$`x{\displaystyle _{\sigma =1/2}^1}(2\pi )^\sigma {\displaystyle \frac{|\mathrm{\Gamma }(\sigma +7i)|}{(\sigma 1)^2+49}}{\displaystyle \frac{|\zeta (4\sigma 2+28i)|}{|\zeta (2\sigma 1+14i)|}}𝑑\sigma `$
$`x(210^6).`$
Finally,
$`\left|{\displaystyle _{C_3}}\left({\displaystyle \frac{x}{2\pi }}\right)^s{\displaystyle \frac{\mathrm{\Gamma }(s)}{(s1)^2}}{\displaystyle \frac{\zeta (4s2)}{\zeta (2s1)}}𝑑s\right|`$
$`\sqrt{{\displaystyle \frac{x}{2\pi }}}{\displaystyle _{t=7}^7}{\displaystyle \frac{|\mathrm{\Gamma }(1/2+it)|}{1/4+t^2}}{\displaystyle \frac{|\zeta (4it)|}{|\zeta (2it)|}}𝑑t`$
$`2.48218\sqrt{x}.`$
Combining these estimates proves (3.1). For $`x20,`$
$`.5235x.8458x^{3/4}.3951x^{1/2}`$
$`(.5235.845820^{1/4}.395120^{1/2})x`$
$`.0351x>0.`$
The positivity for $`x<20`$ is handled by the next lemma. $`\mathrm{}`$
###### Lemma 3.2.
For $`0<x<20`$, one has
$$f(\frac{2\pi }{x})>\underset{n=2}{\overset{\mathrm{}}{}}nf\left(\frac{2\pi n^2}{x}\right).$$
Proof: It is easy to see that for any $`0<a<1`$ and $`t>\frac{a}{1a}`$
$$f(t)>ae^t/t^2.$$
Take $`a=\frac{\pi }{10+\pi }>.23`$. Then for $`0<x<20`$, one has
$$f(\frac{2\pi }{x})>.23\frac{x^2}{4\pi ^2}e^{\frac{2\pi }{x}}.$$
One the other hand, clearly, $`f(x)<e^x/x^2`$, and so
$`{\displaystyle \underset{n=2}{\overset{\mathrm{}}{}}}nf\left({\displaystyle \frac{2\pi n^2}{x}}\right)`$ $`{\displaystyle \underset{n=2}{\overset{\mathrm{}}{}}}n{\displaystyle \frac{x^2}{4\pi ^2n^4}}e^{2\pi n^2/x}`$
$``$ $`{\displaystyle \frac{x^2}{4\pi ^2}}{\displaystyle \underset{n=2}{\overset{\mathrm{}}{}}}n^3e^{2\pi n^2/x}.`$
Since $`n^2n+2`$ for $`n2`$, this is
$$\frac{x^2}{4\pi ^2}\frac{e^{8\pi /x}}{8}\underset{n=0}{\overset{\mathrm{}}{}}e^{2\pi n/x}=\frac{x^2e^{2\pi /x}}{4\pi ^2}\frac{1}{8}\frac{e^{6\pi /x}}{(1e^{2\pi /x})}.$$
(3.3)
Since $`\frac{1}{8}\frac{e^{6\pi /x}}{(1e^{2\pi /x})}`$ is clearly increasing, it is thus bounded above in $`0<x<20`$ by its value $`.181`$ at $`x=20`$. Therefore (3.3) is bounded above by
$$.19\frac{x^2}{4\pi ^2}e^{2\pi /x}<f(2\pi /x).$$
$`\mathrm{}`$
###### Proposition 3.3.
(1) If $`m`$ is any completely multiplicative function with values $`1,0,`$ or $`1`$, then
$$\underset{n=1}{\overset{\mathrm{}}{}}m(n)nf\left(\frac{2\pi n^2}{x}\right)>0,x>0.$$
(3.4)
(2) If $`m_1`$ and $`m_2`$ are two distinct such functions with $`m_1(p)m_2(p)`$ for every prime p, then
$$\underset{n=1}{\overset{\mathrm{}}{}}m_1(n)nf\left(\frac{2\pi n^2}{x}\right)>\underset{n=1}{\overset{\mathrm{}}{}}m_2(n)nf\left(\frac{2\pi n^2}{x}\right)$$
(3.5)
for all $`x>0`$.
Proof: We first assume that the functions $`m`$, $`m_1`$, and $`m_2`$ differ from $`\lambda `$ at only finitely many primes. Therefore, by Proposition 3.1 and induction, it suffices to prove (3.5) under the following conditions:
(a) $`m_2`$ satisfies (3.4) for all $`x>0`$.
(b) $`m_1`$ and $`m_2`$ differ at exactly one prime, say $`p`$, and $`m_1(p)=m_2(p)+1`$. Under assumption (b), the difference is
$$\underset{n=1}{\overset{\mathrm{}}{}}m_1(n)nf\left(\frac{2\pi n^2}{x}\right)\underset{n=1}{\overset{\mathrm{}}{}}m_2(n)nf\left(\frac{2\pi n^2}{x}\right)$$
$$=\underset{pn}{}(m_1(n)m_2(n))nf\left(\frac{2\pi n^2}{x}\right).$$
When $`m_2(p)=1`$, $`m_1(p)=0`$, the difference is then
$$\underset{pn}{}m_2(n)nf\left(\frac{2\pi n^2}{x}\right)=p\underset{n=1}{\overset{\mathrm{}}{}}m_2(n)nf\left(\frac{2\pi n^2}{x/p^2}\right)>0$$
by assumption (a). When $`m_2(p)=0`$, $`m_1(p)=1`$ and $`m_1(p^kn)=m_2(n)`$ for $`pn`$. So the difference is
$$\underset{pn}{}m_1(n)nf\left(\frac{2\pi n^2}{x}\right)=\underset{k=1}{\overset{\mathrm{}}{}}p^k\underset{n=1}{\overset{\mathrm{}}{}}m_2(n)nf\left(\frac{2\pi n^2}{x/p^{2k}}\right)>0$$
by assumption (a) again.
In the general case, define $`m^N(n)`$ to be the completely multiplicative function derived from $`m`$ by
$$m^N(p)=\{\begin{array}{cc}m(p),\hfill & pN\hfill \\ \lambda (p),\hfill & p>N.\hfill \end{array}$$
(3.6)
Then $`m^N`$ differs from $`\lambda `$ at only a finite number of primes, and thus
$$\underset{n=1}{\overset{\mathrm{}}{}}m^N(n)nf\left(\frac{2\pi n^2}{x}\right)\underset{n=1}{\overset{\mathrm{}}{}}\lambda (n)nf\left(\frac{2\pi n^2}{x}\right)>0.$$
Taking the limit as $`N\mathrm{}`$,
$$\underset{n=1}{\overset{\mathrm{}}{}}m(n)nf\left(\frac{2\pi n^2}{x}\right)\underset{n=1}{\overset{\mathrm{}}{}}\lambda (n)nf\left(\frac{2\pi n^2}{x}\right)>0.$$
This completes the proof of part (1); part (2) can be handled similarly. $`\mathrm{}`$
###### Corollary 3.4.
One has
$$R\underset{n=1}{\overset{\mathrm{}}{}}\lambda (n)nf\left(\frac{2\pi n^2}{B}\right).$$
Combining Proposition 3.1 with Corollary 3.4, one obtains (2.12).
## 4 The Trivial Bound on Remainder Term $`C`$
In this section, we will only treat canonical characters, and prove (2.13):
###### Proposition 4.1.
When $`d=1`$ and $`D7`$, term $`C`$ is bounded by
$$|C|<\{\begin{array}{cc}.0269D,\hfill & D\text{ odd}\hfill \\ .2369D,\hfill & D\text{ even}.\hfill \end{array}$$
(4.1)
Proof: We first assume $`D`$ is odd, so $`B=D`$. From (2.8) we can bound $`C`$ term-wise, without appealing to cancellation from the character. To wit,
$$|C|\underset{\stackrel{u,v>0}{uv(\text{mod}2)}}{}uf\left(\frac{\pi }{2}(v^2+u^2/D)\right).$$
Since $`f(x)<e^x/x^2`$,
$$|C|<\underset{\stackrel{u,v>0}{vu(\text{mod}2)}}{}ue^{\frac{\pi u^2}{2D}}\frac{e^{\pi v^2/2}}{\left(\frac{\pi }{2}(v^2+u^2/D)\right)^2}$$
$$<\underset{u=1}{\overset{\mathrm{}}{}}ue^{\frac{\pi u^2}{2D}}\frac{4}{\pi ^2}\underset{\stackrel{v=1}{vu(\text{mod}2)}}{\overset{\mathrm{}}{}}v^4e^{\pi v^2/2}.$$
The inside sum is bounded by
$$\underset{v=1,odd}{\overset{\mathrm{}}{}}v^4e^{\pi v^2/2}.20788,$$
so
$$|C|.0843\underset{u=1}{\overset{\mathrm{}}{}}ue^{\frac{\pi u^2}{2D}}.$$
If $`D`$ is even, then $`B=2D`$,and the same argument shows that
$`|C|`$ $`<`$ $`2\left({\displaystyle \frac{16}{\pi ^2}}{\displaystyle \underset{v=1}{\overset{\mathrm{}}{}}}v^4e^{\pi v^2/4}\right)\left({\displaystyle \underset{u=1}{\overset{\mathrm{}}{}}}ue^{\pi u^2/D}\right)`$
$``$ $`1.488{\displaystyle \underset{u=1}{\overset{\mathrm{}}{}}}ue^{\pi u^2/D}.`$
The proof of (4.1) now follows from bound in the Lemma below. $`\mathrm{}`$
###### Lemma 4.2.
For $`a1`$,
$$\underset{n=1}{\overset{\mathrm{}}{}}ne^{n^2/a}<a/2.$$
(4.2)
This conclusion actually holds for all $`a>0`$.
Proof: Using the Poisson summation formula applied to $`|n|e^{n^2/a}`$,
$$\underset{n=1}{\overset{\mathrm{}}{}}ne^{n^2/a}=_0^{\mathrm{}}ne^{n^2/a}𝑑n+2\underset{r=1}{\overset{\mathrm{}}{}}_0^{\mathrm{}}ne^{n^2/a}\mathrm{cos}(2\pi rn)𝑑n.$$
The first integral is $`a/2`$ and the others are actually negative. This is because
$$_0^{\mathrm{}}ne^{n^2/a}\mathrm{cos}(2\pi rn)𝑑n=\frac{a}{2}\left(1e^{a\pi ^2r^2}2\pi r\sqrt{a}_0^{\pi r\sqrt{a}}e^{t^2}𝑑t\right)$$
(cf. \[GR\], 17.13.27) and
$$_0^{\pi r\sqrt{a}}e^{t^2}𝑑t>1+_1^{a\pi ^2r^2}\frac{e^t}{2\sqrt{t}}𝑑t=1+\frac{1}{2\pi r\sqrt{a}}\left(e^{a\pi ^2r^2}e\right)>\frac{e^{a\pi ^2r^2}}{2\pi r\sqrt{a}}.$$
$`\mathrm{}`$
## 5 Rohrlich’s Bound on Remainder Term $`C`$
We now return to the general case of a canonical character twisted by $`ϵ_d=\left(\frac{d}{}\right)`$. The method here is adapted from \[Ro1\] and \[Ro4\].
###### Proposition 5.1.
For any $`\delta >0`$, term $`C`$ is bounded by
$$|C|D^{15/16+\delta }|d|^{51/16+\delta },$$
where the implied constant depends only on $`\delta `$.
Proof: Set $`A(t)=_{n<t}a_n`$. Integration by parts gives
$$C=_{D/4}^{\mathrm{}}f\left(\frac{2\pi t}{B}\right)\frac{d}{dt}A(t)𝑑t=_{D/4}^{\mathrm{}}A(t)\frac{d}{dt}f\left(\frac{2\pi t}{B}\right)𝑑t,$$
(5.1)
because there are no complex ideals of norm $`<D/4`$. By \[Ro1\], p. 553(27), $`A(t)`$ is bounded above by
$$|A(t)|t^{5/4}D^{5/16+\delta }|d|^{19/16+\delta }$$
for $`D>8`$ and $`t>0`$ (the implied constant again depends only on $`\delta `$). Along with the inequalities
$$0<\frac{d}{dt}f\left(\frac{2\pi t}{B}\right)<\frac{B}{2\pi t^2}e^{2\pi t/B}\left(1+\frac{B}{2\pi t}\right),$$
(5.1) implies
$`|C|`$ $``$ $`D^{5/16+\delta }|d|^{19/16+\delta }{\displaystyle _{D/4}^{\mathrm{}}}\left[t^{5/4}{\displaystyle \frac{B}{2\pi t^2}}e^{2\pi t/B}\left({\displaystyle \frac{4}{3}}+{\displaystyle \frac{B}{2\pi t}}\right)\right]𝑑t`$
$``$ $`D^{5/16+\delta }|d|^{19/16+\delta }{\displaystyle _{D/4}^{\mathrm{}}}{\displaystyle \frac{d}{dt}}\left[{\displaystyle \frac{B^2}{2\pi ^2}}t^{3/4}e^{2\pi t/B}\right]𝑑t`$
$``$ $`D^{17/16+\delta }|d|^{19/16+\delta }B^2.`$
Since either $`B=D|d|`$ or $`2D|d|`$, this completes the proof. $`\mathrm{}`$
## 6 Arithmetic Applications
Having completed their proofs, we will now give some arithmetic applications of Theorems 1.1 and 2.2, including Corollary 1.2.
Let $`j`$ be the $`j`$-invariant of a fixed isomorphism class of elliptic curves with complex multiplication (CM) by $`𝒪`$. Then $`H=K(j)`$ is the Hilbert class field of $`K`$. We can extend any Hecke character $`\chi `$ of $`K`$ to one on $`H`$ by
$$\psi =\chi N_{H/K}.$$
When $`\chi `$ satisfies (1.1) and (1.2),
$$\psi (𝔄^\sigma )=\psi (𝔄)^\sigma ,\sigma \text{Gal}(H/)$$
for every ideal $`𝔄`$ of $`H`$ relatively prime to the conductor of $`\psi `$. By Theorem 9.1.3 and Lemma 11.1.1 of \[Gr\], there is a unique elliptic $``$-curve $`A`$ over $`H`$ with
$$j(A)=j\text{ and }L(s,A/H)=L(s,\psi )L(s,\overline{\psi }).$$
(Here we recall that a “$``$-curve” is an elliptic curve over a number field which is isogenous to all of its Galois conjugates.) Furthermore, $`A`$ descends to two isogenous elliptic curves over the subfield $`F=(j)`$ (\[Gr\], Theorem 10.2.1). By abuse of notation we will also refer to these curves as $`A`$. Let $`B=\text{Res}_{F/}A`$ be the abelian variety over $``$ obtained from $`A`$ by restriction of scalars. When $`D=p`$ is prime, Gross proved (\[Gr\], Theorem 15.2.5) that $`T=\text{End}_KB`$ is a CM number field of degree $`2h`$; thus $`B`$ is also a CM abelian variety. This result actually extends to composite $`D`$ via a different argument:
###### Lemma 6.1.
(a) Let $`T`$ be the subfield of $``$ generated by $`\chi (𝔞)`$, where $`𝔞`$ runs over all ideals of K prime to $`\chi `$’s conductor. Then $`T`$ is a CM number field of degree $`2h`$, and $`\mathrm{\Phi }=\{\sigma :T|\sigma \text{ trivial on }K\}`$ is a CM type of T.
(b) B is a CM abelian variety of type $`(T,\mathrm{\Phi })`$.
Proof: For each embedding $`\sigma :T`$ fixing $`K`$, $`\sigma \chi `$ is another canonical Hecke character of $`K`$, and thus it is of the form $`\chi \varphi `$, where $`\varphi `$ is an ideal class character of $`K`$. By Theorem 1 of \[Ro3\], $`\sigma \varphi `$ actually gives a one-to-one correspondence between the complex embeddings of $`T`$ into $``$ fixing $`K`$, and the ideal class characters of $`K`$. Thus $`[T:K]=h`$ and $`[T:]=2h`$. It is a general fact that $`T`$ is a CM number field; in this case it can easily be verified using property (1.1) of $`\chi `$.
By \[Sh\], Theorem 10, there is a CM abelian variety $`B^{}/`$ of type $`(T,\mathrm{\Phi })`$ associated to $`\chi `$, and it is unique up to isogeny. In particular
$$L(s,B^{})=\underset{\sigma :T^+}{}L(s,\chi ^\sigma )=\underset{\varphi }{}L(s,\chi \varphi ),$$
(6.1)
where $`T^+`$ is the maximal totally-real subfield of $`T`$. On the other hand,
$$L(s,B)=L(s,A/F)=L(s,\psi )=\underset{\varphi }{}L(s,\chi \varphi ).$$
This shows $`L(s,B)=L(s,B^{})`$, so a theorem of Faltings \[Fa\] guarantees $`B`$ and $`B^{}`$ are isogenous, proving (b). $`\mathrm{}`$
###### Lemma 6.2.
Let $`\chi `$ be a Hecke character of $`K`$ of the form (2.1). Let $`A`$ be an associated $``$-curve over $`F=(j)`$ with $`j`$-invariant $`j`$, and let $`B=\text{Res}_{F/}A`$. If $`\text{ord}_{s=1}L(s,\chi )1`$ then
(a) The Mordell-Weil ranks of $`A`$ and $`B`$ are given by
$$\text{rank}_{}A(F)=\text{rank}_{}B()=h\text{ord}_{s=1}L(s,\chi ).$$
(b) The Shafarevich-Tate groups $`\text{X}(A/F)`$ and $`\text{X}(B/)`$ are finite.
Proof: Since the Mordell-Weil and Shafarevich-Tate groups of $`A`$ over $`F`$ are identical to those of $`B`$ over $``$, it is sufficient to prove the Lemma for $`B`$. Let $`f`$ be the normalized weight 2 new-form associated to $`\chi `$ as in the proof of Lemma 2.1. The field generated by $`f`$’s Fourier coefficients is generated by $`\chi (𝔞)+\overline{\chi (𝔞)}`$, and is thus $`T^+`$. Equation (6.1) implies
$$L(s,B)=\underset{\sigma :T^+}{}L(s,f^\sigma ).$$
Now the Lemma follows from a result of Kolyvagin and Logachev (\[KL\]). $`\mathrm{}`$
Combining Lemma 6.2 with the non-vanishing theorems above (Theorems 1.1 and 2.2) and in \[MR\], one gets the following two corollaries.
###### Corollary 6.3.
Let $`\chi =\chi _{\text{can}}`$ be a canonical Hecke character of $`K`$, and let $`A`$ and $`B=\text{Res}_{F/}A`$ respectively be associated $``$-curves and CM abelian varieties as above. Then
(a) The Mordell-Weil ranks of $`A`$ and $`B`$ are given in terms of the root number $`W(\chi )`$ by
$$\text{rank}_{}A(F)=\text{rank}_{}B()=\{\begin{array}{cc}h,\hfill & W(\chi )=1\hfill \\ 0,\hfill & W(\chi )=1.\hfill \end{array}$$
In particular, when $`D`$ is odd, these ranks are $`h`$ or zero depending on whether $`D3`$ or 7$`mod8`$.
(b) The Shafarevich-Tate groups $`\text{X}(A/F)`$ and $`\text{X}(B/)`$ are finite.
Proof of Corollary 1.2: Take $`D=p`$, $`j=j\left(\frac{1+\sqrt{p}}{2}\right)`$, $`A=A(p)`$, and apply Corollary 6.3. $`\mathrm{}`$
###### Corollary 6.4.
Let $`\chi =\chi _{D,d}`$ be a Hecke character of $`K`$ of the form (2.1). Let $`A`$ and $`B=\text{Res}_{F/}A`$ be as above, and fix any $`\delta >0`$. If $`|d|D^{1/35\delta }`$ (the implied constant depending on $`\delta `$) and $`W(\chi _{D,d})=1`$, then
(a) The Mordell-Weil ranks of $`A`$ and $`B`$ are
$$\text{rank}_{}A(F)=\text{rank}_{}B()=h.$$
(b) The Shafarevich-Tate groups $`\text{X}(A/F)`$ and $`\text{X}(B/)`$ are finite.
Finally, we wish to point out that when $`D`$ is prime, all $``$-curves over $`F`$ are associated to Hecke characters of the form (2.1), though this is not true for every composite $`D`$. See \[Na\] for a more-precise description.
| Stephen D. Miller | | Tonghai Yang |
| --- | --- | --- |
| Department of Mathematics | | Department of Mathematics |
| Yale University | | State University of New York |
| P.O. Box 208283 | | Stony Brook, NY 11794-3651 |
| New Haven, CT 06520 | | thyang@math.sunysb.edu |
| stephen.miller@yale.edu | | |
|
warning/0003/hep-th0003282.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
The component Lagrangian of matter-coupled supergravity can be derived from a superspace formulation or a tensor calculus . Both approaches inevitably lead to a component theory in which the gravitational action is of a generalized Brans-Dicke form,
$$e^1=\frac{1}{2}e^{K/3}+\mathrm{}.$$
(1.1)
In this expression, $`e`$ is the vielbein determinant, $``$ is the curvature scalar and $`K`$ is the Kähler potential, a function of the scalar fields $`A^{},A`$.
The Lagrangian (1.1) leads to a kinetic mixing between the graviton and the scalar fields. In addition, the kinetic terms of the scalar fields appear in non-Kähler form. It is therefore convenient and customary to carry out a field-dependent Weyl rescaling of the metric to bring the Lagrangian into canonical Einstein form. In supergravity, this Weyl rescaling must also be accompanied by a chiral rotation of the fermions. As we shall see, this rotation gives rise to an anomalous Jacobian in the supergravity action.
The field redefinitions needed to go to this “Einstein frame” are usually performed in terms of component fields . This obscures the symmetries of the theory and complicates the study of anomalies and their consequences. Therefore, in this paper we will use the superspace approach of to study anomalies in supergravity theories. We will show that:
1. The Einstein-frame field redefinitions can be carried out directly in superspace through a super-Weyl transformation of the vielbein. The corresponding component-field Lagrangian gives rise to ordinary Einstein gravity.
2. Local supersymmetry transformations in the Einstein frame involve chiral rotations of the fermions. They are anomalous at one loop.
3. The local supersymmetry anomaly is cancelled by a Jacobian that arises from the transition to the Einstein frame. This Jacobian is necessary to ensure the quantum consistency of Einstein-frame supergravity.
4. The anomalous Jacobian can have important physical consequences. For example, it is necessary to ensure the quantum equivalence of scattering amplitudes computed in different frames.
This paper is organized as follows. In Section 2 we define super-Weyl transformations and derive the corresponding Jacobians. In Section 3 we study the transition to the Einstein frame. We show that one-loop supergravity invariance of the Einstein-frame Lagrangian requires that a certain superspace Jacobian be added to the bare Lagrangian. In Section 4 we present an explicit example which illustrates the physical importance of this Jacobian. We summarize our results in Section 5.
## 2 Super-Weyl Transformations
### 2.1 Classical Level
In this section we study super-Weyl transformations in classical and quantum supergravity. These transformations will play an important role throughout this paper.
In what follows we use the notation and conventions of . We take the matter-coupled supergravity Lagrangian to be of the form
$`(X)`$ $`=`$ $`{\displaystyle }d^2\mathrm{\Theta }2[{\displaystyle \frac{3}{8}}(\overline{𝒟}^28R)\mathrm{exp}\{{\displaystyle \frac{1}{3}}(K(\mathrm{\Phi }^{},\mathrm{\Phi })+\mathrm{\Gamma }(\mathrm{\Phi }^{},\mathrm{\Phi },V))\}`$ (2.1)
$`+{\displaystyle \frac{1}{4}}H_{ab}(\mathrm{\Phi })W^{(a)}W^{(b)}+P(\mathrm{\Phi })]+\mathrm{h}.\mathrm{c}.,`$
where $`X=(\mathrm{\Phi },V,E_{M}^{}{}_{}{}^{A})`$ denotes the set of fields in the supergravity Lagrangian, $`\mathrm{\Phi }`$ and $`V`$ are the chiral and vector superfields, and $`E_M^A`$ is the supervielbein. In this expression, $`K`$ is the Kähler potential, $`P`$ the superpotential, and $`W^{(a)}`$ the field strength superfield, where $`(a)`$ is the index for the adjoint representation.<sup>#1</sup><sup>#1</sup>#1In this paper, we sometimes omit the spin index $`\alpha `$. Therefore $`W^{(a)}W^{(b)}`$ should be understood as $`W^{(a)\alpha }W_\alpha ^{(b)}`$. In addition, $`H_{ab}`$ is the gauge kinetic function, and $`\mathrm{\Gamma }`$ is the gauge counterterm which renders the Lagrangian gauge invariant.
In the superspace formalism, supergravity transformations are given by translations in superspace. Chiral and vector superfields transform as follows,
$`\delta _{\mathrm{SUSY}}\mathrm{\Phi }=\xi ^A𝒟_A\mathrm{\Phi },\delta _{\mathrm{SUSY}}V=\xi ^A𝒟_AV,`$ (2.2)
while the vielbein transforms as
$`\delta _{\mathrm{SUSY}}E_M{}_{}{}^{A}=𝒟_M\xi ^A\xi ^BT_{BM}{}_{}{}^{A}.`$ (2.3)
In these expressions, the $`𝒟_A`$ are covariant derivatives and the $`T_{BM}^A`$ are the superspace torsion. The superspace formalism ensures that the Lagrangian (2.1) is invariant,
$`(X+\delta _{\mathrm{SUSY}}X)=(X),`$ (2.4)
up to a total derivative, under the supersymmetry transformations (2.2) and (2.3).
Super-Weyl transformations are defined as rescalings of the superspace vielbein that preserve the torsion constraints .<sup>#2</sup><sup>#2</sup>#2It is important to distinguish between super-Weyl and super-Weyl-Kähler transformations. The latter are symmetries of the classical supergravity Lagrangian. A super-Weyl-Kähler transformation is a super-Weyl transformation, with chiral superfield parameter $`\mathrm{\Sigma }`$, combined with a redefinition of the Kähler potential and superpotential, $`KK+6\mathrm{\Sigma }+6\mathrm{\Sigma }^{}`$, $`P\mathrm{exp}(6\mathrm{\Sigma })P`$. They are parameterized by a chiral superfield $`\mathrm{\Sigma }`$. If we denote the super-Weyl transformed $`X`$ field as $`\widehat{X}`$, we can write
$`X=\widehat{X}+\delta _{\mathrm{SW}}X,`$ (2.5)
where $`\delta _{\mathrm{SW}}X`$ is the super-Weyl variation of $`X`$.
To linear order in $`\mathrm{\Sigma }`$, the super-Weyl transformations are given by
$`\delta _{\mathrm{SW}}`$ $`=`$ $`6\mathrm{\Sigma }+{\displaystyle \frac{}{\mathrm{\Theta }^\alpha }}\left(S^\alpha \right)`$
$`\delta _{\mathrm{SW}}\mathrm{\Phi }`$ $`=`$ $`S^\alpha {\displaystyle \frac{}{\mathrm{\Theta }^\alpha }}\mathrm{\Phi }`$
$`\delta _{\mathrm{SW}}\left(\overline{𝒟}^28R\right)U`$ $`=`$ $`\left(\overline{𝒟}^28R\right)\left(4\mathrm{\Sigma }2\mathrm{\Sigma }^{}\right)US^\alpha {\displaystyle \frac{}{\mathrm{\Theta }^\alpha }}\left(\overline{𝒟}^28R\right)U`$
$`\delta _{\mathrm{SW}}W_\alpha `$ $`=`$ $`3\mathrm{\Sigma }W_\alpha S^\beta {\displaystyle \frac{}{\mathrm{\Theta }^\beta }}W_\alpha ,`$ (2.6)
where $`U`$ is any real vector superfield of Weyl weight zero, and $`S^\alpha `$ is defined by
$`S^\alpha =\mathrm{\Theta }^\alpha (2\mathrm{\Sigma }^{}\mathrm{\Sigma })|+\mathrm{\Theta }^2(𝒟^\alpha \mathrm{\Sigma })|.`$ (2.7)
The bar $`|`$ denotes the $`\theta =\overline{\theta }=0`$ component of the superfield.
Note that super-Weyl transformations also induce chiral rotations of the component fermions. Taking the appropriate components of Eq. (2.6) and re-exponentiating, we find
$`\chi =\mathrm{exp}(\mathrm{\Sigma }|2\mathrm{\Sigma }^{}|)\widehat{\chi },\lambda =\mathrm{exp}(3\mathrm{\Sigma }|)\widehat{\lambda },`$ (2.8)
where $`\chi `$ and $`\lambda `$ are the fermions in the chiral and gaugino multiplets, respectively.
In general, super-Weyl transformations are not symmetries of the classical supergravity Lagrangian. Indeed, substituting the transformed variables (2.6) into the Lagrangian (2.1), all derivative terms cancel. Nevertheless, a nontrivial $`\mathrm{\Sigma }`$ dependence remains. At the classical level, the Lagrangian (2.1) becomes
$`(\widehat{X}+\delta _{\mathrm{SW}}X)`$ $`=`$ $`{\displaystyle }d^2\mathrm{\Theta }2\widehat{}[{\displaystyle \frac{3}{8}}(\widehat{\overline{𝒟}}^28\widehat{R})\mathrm{exp}\{{\displaystyle \frac{1}{3}}(\widehat{K}6\mathrm{\Sigma }6\mathrm{\Sigma }^{}+\widehat{\mathrm{\Gamma }})\}`$ (2.9)
$`+{\displaystyle \frac{1}{4}}\widehat{H}_{ab}\widehat{W}^{(a)}\widehat{W}^{(b)}+\mathrm{exp}(6\mathrm{\Sigma })\widehat{P}]+\mathrm{h}.\mathrm{c}.,`$
where the hatted objects are evaluated using the Weyl-transformed fields. The Kähler and superpotential are different, so the Lagrangian (2.1) is not invariant.
### 2.2 Quantum Level
We are now ready to discuss the anomalous Jacobian associated with a given super-Weyl transformation. A proper framework is provided by the one-particle-irreducible (1PI) effective Lagrangian $`_{1\mathrm{P}\mathrm{I}}`$, defined by
$`{\displaystyle d^4x_{1\mathrm{P}\mathrm{I}}(X_\mathrm{C})}=i\mathrm{log}\left[{\displaystyle [dX_\mathrm{Q}]\mathrm{exp}\left(id^4x_{\mathrm{bare}}(X_\mathrm{C}+X_\mathrm{Q})\right)}\right].`$ (2.10)
Here $`_{\mathrm{bare}}`$ is the bare Lagrangian, and $`X_\mathrm{C}`$ and $`X_\mathrm{Q}`$ are classical and quantum parts of the $`X`$ field, respectively.
In general, super-Weyl transformations are anomalous; they have a mixed super-Weyl-gauge anomaly.<sup>#3</sup><sup>#3</sup>#3It also has a mixed super-Weyl-gravity anomaly. We ignore the gravity anomaly here.<sup>#4</sup><sup>#4</sup>#4For a discussion of supergravity anomalies in the compensator formalism, see . Anomalies generate a set of non-local terms in the 1PI effective action . For the case at hand, the one-loop anomaly-induced terms are ,
$$\mathrm{\Delta }=\frac{1}{256\pi ^2}d^2\mathrm{\Theta }2W^{(a)}W^{(a)}\frac{1}{\mathrm{}}\left(\overline{𝒟}^28R\right)\left[4(T_R3T_G)R^{}\frac{1}{3}T_R𝒟^2K\right]+\mathrm{h}.\mathrm{c}.,$$
(2.11)
where we have omitted a term from the sigma-model anomaly that is irrelevant for our discussion. In Eq. (2.11), $`T_G`$ is the Dynkin index of the adjoint representation, normalized to $`N`$ for $`SU(N)`$, and $`T_R`$ is the Dynkin index associated with the matter fields. A sum over all matter representations is understood. The first term, which contains the $`R^{}`$ superfield, arises from the superconformal anomaly. It is proportional to the beta function coefficient, $`b_0=3T_GT_R`$. The second term expresses the Kähler anomaly .
The variation of $`\mathrm{\Delta }`$ can be computed by considering a super-Weyl transformation with superfield parameter $`\mathrm{\Sigma }`$. Under such a transformation, the superfield $`R`$ changes as follows,
$`\delta _{\mathrm{SW}}R=2(2\mathrm{\Sigma }\mathrm{\Sigma }^{})R{\displaystyle \frac{1}{4}}\overline{𝒟}^2\mathrm{\Sigma }^{}.`$ (2.12)
This induces a shift by $`\mathrm{\Sigma }`$ in the $`R^{}`$ term in Eq. (2.11). A second shift comes from replacing $`K`$ by $`\widehat{K}6\mathrm{\Sigma }6\mathrm{\Sigma }^{}`$. These two shifts induce the following change in $`\mathrm{\Delta }`$,
$$\mathrm{\Delta }\mathrm{\Delta }+_\mathrm{J},$$
(2.13)
where
$`_\mathrm{J}={\displaystyle \frac{1}{16\pi ^2}}(3T_R3T_G){\displaystyle d^2\mathrm{\Theta }2\mathrm{\Sigma }W^{(a)}W^{(a)}}+\mathrm{h}.\mathrm{c}.`$ (2.14)
The Lagrangian $`_\mathrm{J}`$ can be interpreted as the superspace Jacobian that arises from the super-Weyl transformation (2.5). (Note that the imaginary part of $`\mathrm{\Sigma }|`$ corresponds to the Jacobian from the anomalous $`U(1)_R`$ transformation.)
The nonvanishing Jacobian implies that the functional measure is not invariant. It transforms as follows under an arbitrary super-Weyl transformation:
$`[d\mathrm{\Phi }][dV]=[d(\widehat{\mathrm{\Phi }}+\delta _{\mathrm{SW}}\mathrm{\Phi })][d(\widehat{V}+\delta _{\mathrm{SW}}V)]=[d\widehat{\mathrm{\Phi }}][d\widehat{V}]\mathrm{exp}\left[i{\displaystyle d^4x_\mathrm{J}}\right],`$ (2.15)
The super-Weyl-rescaled 1PI Lagrangian is then
$`\mathrm{exp}\left[i{\displaystyle d^4x_{1\mathrm{P}\mathrm{I}}}\right]`$ $`=`$ $`{\displaystyle [d(\widehat{X}+\delta _{\mathrm{SW}}X)]\mathrm{exp}\left[id^4x_{\mathrm{bare}}(\widehat{X}+\delta _{\mathrm{SW}}X)\right]}`$ (2.16)
$`=`$ $`{\displaystyle [d\widehat{X}]\mathrm{exp}\left[id^4x\left\{_{\mathrm{bare}}(\widehat{X})|_{\widehat{K}\widehat{K}6\mathrm{\Sigma }6\mathrm{\Sigma }^{},\widehat{P}e^{6\mathrm{\Sigma }}\widehat{P}}+_\mathrm{J}\right\}\right]}`$
$`=`$ $`{\displaystyle [d\widehat{X}]\mathrm{exp}\left[id^4x\widehat{}_{\mathrm{bare}}(\widehat{X})\right]},`$
where
$`\widehat{}_{\mathrm{bare}}(\widehat{X})_{\mathrm{bare}}(\widehat{X}+\delta _{\mathrm{SW}}X)+_\mathrm{J}=_{\mathrm{bare}}(\widehat{X})|_{\widehat{K}\widehat{K}6\mathrm{\Sigma }6\mathrm{\Sigma }^{},\widehat{P}e^{6\mathrm{\Sigma }}\widehat{P}}+_\mathrm{J}.`$ (2.17)
In these expressions, $`\widehat{}_{\mathrm{bare}}`$ is the bare Lagrangian for the quantum theory with super-Weyl-rescaled variable $`\widehat{X}`$. The bare Lagrangian does not contain the anomaly term $`\mathrm{\Delta }`$, which arises from integrating out the massless quantum fields. It does, however, contain the Jacobian $`_\mathrm{J}`$. As we will see, $`_\mathrm{J}`$ is important for ensuring the quantum consistency of supergravity in the Einstein frame.
## 3 Einstein Supergravity
### 3.1 The Einstein Frame
In this section, we find the field-dependent Weyl rescaling that takes the “supergravity frame” Lagrangian of Eq. (2.1)
$$e^1=\frac{1}{2}e^{K/3}+\mathrm{}$$
(3.1)
into the Einstein frame. In the literature, this rescaling has traditionally been done in terms of component fields . Here we perform the transformation in superspace. This allows us to keep better track of the symmetries of the theory.
The relevant superfield rescaling is, as we will see below, a super-Weyl transformation with transformation parameter $`\mathrm{\Sigma }_\mathrm{E}`$,
$`_\mathrm{E}(X)`$ $`=`$ $`{\displaystyle }d^2\mathrm{\Theta }2[{\displaystyle \frac{3}{8}}(\overline{𝒟}^28R)\mathrm{exp}\{{\displaystyle \frac{1}{3}}(K6\mathrm{\Sigma }_\mathrm{E}6\mathrm{\Sigma }_\mathrm{E}^{}+\mathrm{\Gamma })\}`$ (3.2)
$`+{\displaystyle \frac{1}{4}}H_{ab}W^{(a)}W^{(b)}+\mathrm{exp}(6\mathrm{\Sigma }_\mathrm{E})P]+\mathrm{h}.\mathrm{c}.,`$
where we have omitted all “hats” in the above equation. (Here and hereafter, all quantities should be understood as being defined in the frame obtained after the super-Weyl transformation, unless specified otherwise.)
The parameter $`\mathrm{\Sigma }_\mathrm{E}`$ can be found by demanding that $`K6\mathrm{\Sigma }_\mathrm{E}6\mathrm{\Sigma }_\mathrm{E}^{}`$ have no lowest, $`\mathrm{\Theta }`$ and $`\mathrm{\Theta }^2`$ components since this combination appears in the exponent of the first term in Eq. (3.2). To see, for example, why $`(K6\mathrm{\Sigma }_\mathrm{E}6\mathrm{\Sigma }_\mathrm{E}^{})|`$ must vanish, note that Eq. (3.1) involves the lowest component of $`e^{K/3}`$. If the lowest component of this term is scaled to 1, the factor $`e^{K/3}`$ is absent and gravity is canonically normalized. The other two conditions lead to a canonical kinetic term for the gravitino, and to canonical Kähler kinetic terms for the matter multiplets.
The conditions on $`\mathrm{\Sigma }_\mathrm{E}`$ are, therefore,
$`K|=6\mathrm{\Sigma }_\mathrm{E}|+6\mathrm{\Sigma }_\mathrm{E}^{}|,(𝒟_\alpha K)|=6(𝒟_\alpha \mathrm{\Sigma }_\mathrm{E})|,(𝒟^2K)|=6(𝒟^2\mathrm{\Sigma }_\mathrm{E})|,`$ (3.3)
or the vanishing of the lowest, $`\mathrm{\Theta }`$ and $`\mathrm{\Theta }^2`$ components of $`K6\mathrm{\Sigma }_\mathrm{E}6\mathrm{\Sigma }_\mathrm{E}^{}`$. The Einstein frame conditions (3.3) almost completely determine the parameter $`\mathrm{\Sigma }_\mathrm{E}`$,
$`\mathrm{\Sigma }_\mathrm{E}=A_\mathrm{\Sigma }+\sqrt{2}\mathrm{\Theta }\chi _\mathrm{\Sigma }+\mathrm{\Theta }^2F_\mathrm{\Sigma },`$ (3.4)
with
$`A_\mathrm{\Sigma }={\displaystyle \frac{1}{12}}K+i\varphi ,\chi _\mathrm{\Sigma }={\displaystyle \frac{1}{6}}K_i\chi ^i,F_\mathrm{\Sigma }={\displaystyle \frac{1}{6}}K_iF^i{\displaystyle \frac{1}{12}}K_{ij}\chi ^i\chi ^j,`$ (3.5)
where the subscript $`i`$ on $`K`$ denotes the derivative with respect to $`A^i`$ ($`K_iK/A^i`$), and $`F^i`$ is the highest component of $`\mathrm{\Phi }^i`$. Note that the conditions (3.3) do not completely fix $`\mathrm{\Sigma }_\mathrm{E}`$; the imaginary part $`\varphi `$ of its lowest component is left undetermined. Also note that $`\mathrm{\Gamma }`$ does not contribute to the conditions (3.3): we assume that a field redefinition is performed so that the matter fields are in the Wess-Zumino gauge, where $`\mathrm{\Gamma }`$ has no lowest, $`\mathrm{\Theta }`$ and $`\mathrm{\Theta }^2`$ components.
With the above $`\mathrm{\Sigma }_\mathrm{E}`$, it is not hard to find the component Lagrangian. If we substitute Eq. (3.4) into (3.2), all terms leading to non-Einstein gravity vanish. The rest of the component Lagrangian can be readily evaluated and gives the well-known Einstein supergravity Lagrangian with canonical kinetic terms. The complete expression for the Lagrangian is given in . We have seen that the component Lagrangian can be directly obtained from superspace, without any extra Weyl rescalings of the component fields.
Let us now show that we can safely set the field $`\varphi `$ to zero. This field appears in the kinetic terms of the matter and gauge fermions,
$`e^1_{\mathrm{kin}}`$ $`=`$ $`iK_{ij^{}}\overline{\chi }^j^{}\overline{\sigma }^a\left[_a+{\displaystyle \frac{i}{6}}b_a{\displaystyle \frac{1}{6}}\left(K_k_aA^kK_j^{}_aA^j^{}12i_a\varphi \right)\right]\chi ^i`$ (3.6)
$`i\overline{\lambda }\overline{\sigma }^a\left(_a{\displaystyle \frac{i}{2}}b_a\right)\lambda .`$
(We omit terms with spin, sigma-model, and gauge connections because they are not relevant for our discussion.) The field $`\varphi `$ also appears in the solution to the equations of motion for the auxiliary field $`b_a`$,
$`b_a={\displaystyle \frac{i}{2}}\left(K_j_aA^jK_j^{}_aA^j^{}12i_a\varphi \right)+{\displaystyle \frac{1}{4}}K_{ij^{}}\chi ^i\sigma _a\overline{\chi }^j^{}+\mathrm{}.`$ (3.7)
(For the complete expression for $`b_a`$, see Appendix B.) In addition, it appears in the superpotential terms in the Einstein frame,
$`e^1_{\mathrm{Yukawa}}={\displaystyle \frac{1}{2}}\mathrm{exp}\left({\displaystyle \frac{1}{2}}K+6i\varphi \right)P_{ij}\chi ^i\chi ^j+\mathrm{h}.\mathrm{c}.+\mathrm{}.`$ (3.8)
(The complete expression for the Yukawa terms can be found in .)
Upon inspection of Eqs. (3.6), (3.7) and (3.8), one can check that the classical component Lagrangian can be made independent of $`\varphi `$ by redefining $`\chi e^{3i\varphi }\chi `$ and $`\lambda e^{3i\varphi }\lambda `$. In fact, the $`\varphi `$ dependence also cancels at the quantum level. The field redefinitions used to go to the Einstein frame are $`\chi e^{\mathrm{Re}\mathrm{\Sigma }_\mathrm{E}|+3i\varphi }\chi `$ and $`\lambda e^{3\mathrm{R}\mathrm{e}\mathrm{\Sigma }_\mathrm{E}|3i\varphi }\lambda `$. The Jacobian from these transformations exactly cancels the Jacobian from the redefinitions used to eliminate $`\varphi `$ from the component Lagrangian. Therefore the field $`\varphi `$ is unphysical, and we can safely set it to zero.
### 3.2 Supersymmetry Transformations in the Einstein Frame
In this subsection, we discuss the invariance of the classical supergravity Lagrangian in the Einstein frame. Then, in the next subsection, we consider quantum effects, and in particular, anomalies.
It is simple to see that the Einstein-frame Lagrangian is not invariant under the supersymmetry transformations (2.2). Under a supersymmetry transformation, the Kähler potential transforms as $`\delta _{\mathrm{SUSY}}K=\xi ^A𝒟_AK`$, in which case $`\mathrm{\Sigma }_\mathrm{E}`$ transforms to $`\mathrm{\Sigma }_\mathrm{E}^{}`$,
$`\mathrm{\Sigma }_\mathrm{E}^{}\mathrm{\Sigma }_\mathrm{E}|_{KK+\delta _{\mathrm{SUSY}}K}=\mathrm{\Sigma }_\mathrm{E}\xi ^A𝒟_A\mathrm{\Sigma }_\mathrm{E}\mathrm{\Sigma }_\xi ,`$ (3.9)
where
$`\mathrm{\Sigma }_\xi ={\displaystyle \frac{1}{6\sqrt{2}}}\left(\xi \chi ^iK_i\overline{\xi }\overline{\chi }^i^{}K_i^{}\right)+𝒪(\mathrm{\Theta })+𝒪(\mathrm{\Theta }^2)i\varphi _\xi +𝒪(\mathrm{\Theta })+𝒪(\mathrm{\Theta }^2).`$ (3.10)
The second term on the RHS of Eq. (3.9) is the supersymmetry transformation of a normal chiral superfield. The $`\mathrm{\Sigma }_\xi `$ term is an additional super-Weyl transformation under which the action is not invariant.
To see this explicitly, consider the supersymmetry transformation of the Einstein-frame Lagrangian,
$`_\mathrm{E}(X+\delta _{\mathrm{SUSY}}X)`$ $`=`$ $`{\displaystyle }d^2\mathrm{\Theta }2[{\displaystyle \frac{3}{8}}(\overline{𝒟}^28R)\mathrm{exp}\{{\displaystyle \frac{1}{3}}(K6(\mathrm{\Sigma }_\mathrm{E}\mathrm{\Sigma }_\xi )6(\mathrm{\Sigma }_\mathrm{E}^{}\mathrm{\Sigma }_\xi ^{})+\mathrm{\Gamma })\}`$ (3.11)
$`+{\displaystyle \frac{1}{4}}H_{ab}W^{(a)}W^{(b)}+\mathrm{exp}\left\{6(\mathrm{\Sigma }_\mathrm{E}\mathrm{\Sigma }_\xi )\right\}P]+\mathrm{h}.\mathrm{c}.`$
Since $`\mathrm{\Sigma }_\xi `$ is nonvanishing, the Lagrangian is not invariant under the supersymmetry transformation (2.2).
This lack of invariance stems from the fact that $`\mathrm{\Sigma }_\xi `$ takes the Lagrangian out of the Einstein frame. Invariance can be restored by returning to the Einstein frame through a compensating super-Weyl transformation. This is similar to gauge invariance in globally supersymmetric gauge theories. There, a supersymmetry transformation in the Wess-Zumino gauge must be supplemented by a superfield gauge transformation to restore the Wess-Zumino gauge condition. It is instructive to consider this case in some detail because of the close analogy to supergravity . To that end, we review the supersymmetry transformations of globally supersymmetric gauge theories in Appendix A.
For the case at hand, the compensating super-Weyl transformation has parameter $`\mathrm{\Sigma }_\xi `$. In the Einstein frame, therefore, we define a supersymmetry transformation to include a frame-restoring super-Weyl transformation with parameter $`\mathrm{\Sigma }_\xi `$:
$`\delta _\xi \delta _{\mathrm{SUSY}}+\delta _{\mathrm{SW}}.`$ (3.12)
Under such a transformation, chiral and vector superfields transform as follows,
$`\delta _\xi \mathrm{\Phi }=\xi ^A𝒟_A\mathrm{\Phi }+\delta _{\mathrm{SW}}\mathrm{\Phi },\delta _\xi V=\xi ^A𝒟_AV+\delta _{\mathrm{SW}}V,`$ (3.13)
and analogously for the vielbein. In these expressions, the first terms on the RHS are the original supersymmetry transformations; the second are the compensating super-Weyl transformations with parameter $`\mathrm{\Sigma }_\xi `$. These transformations eliminate the $`\mathrm{\Sigma }_\xi `$ in Eq. (3.11) and restore the classical invariance of the action,
$`_\mathrm{E}(X+\delta _\xi X)=_\mathrm{E}(X).`$ (3.14)
The transformation properties of the individual component fields can be derived by expanding Eq. (3.13). We have checked that they agree with the transformations given in after eliminating the auxiliary fields.
### 3.3 Quantum Consistency in the Einstein Frame
We are now ready to discuss anomalies in the supersymmetry transformations (3.12). As we have seen, these transformations include frame-restoring super-Weyl field rescalings that induce chiral rotations on the matter fermions,
$`\delta _\xi \chi =\mathrm{}+3i\varphi _\xi \chi ,\delta _\xi \lambda =\mathrm{}3i\varphi _\xi \lambda ,`$ (3.15)
where $`i\varphi _\xi =\mathrm{\Sigma }_\xi |`$. At the quantum level, these transformations are anomalous, so they should give rise to an anomalous variation of the 1PI Lagrangian,
$`\delta _\xi _{1\mathrm{P}\mathrm{I}}=e\left[{\displaystyle \frac{1}{16\pi ^2}}(3T_R3T_G)\varphi _\xi F_{mn}^{(a)}\stackrel{~}{F}^{mn(a)}+\mathrm{}\right],`$ (3.16)
If nothing were to cancel this variation, local supersymmetry in the Einstein frame would be anomalous. In what follows, we will show that the full 1PI effective action is, in fact, invariant. The variation (3.16) is cancelled by the variation of the Jacobian (2.14) that arises in passing to the Einstein frame.
In the Einstein frame, the complete 1PI effective Lagrangian is of the following form,
$`_{1\mathrm{P}\mathrm{I}}=_\mathrm{E}+\mathrm{\Delta }+_\mathrm{J}.`$ (3.17)
The first term is the classical part of the Einstein-frame Lagrangian, the second is the non-local term induced by anomalies, and the third is the Jacobian (2.14). The first term is invariant under the local supersymmetry transformation (3.12), as discussed in the previous subsection. The second and third terms are not. Under the supersymmetry transformation (3.12), the nonvanishing variation of $`\mathrm{\Delta }`$ expresses the anomaly associated with the frame-restoring super-Weyl transformation.<sup>#5</sup><sup>#5</sup>#5If we make the argument before integrating out the light fields, $`\mathrm{\Delta }`$ does not exist. In this case the anomaly arises as a change of the functional measure of the path integral. If this variation were the only change of the Lagrangian, supersymmetry would be explicitly broken by anomalies at the quantum level.
Fortunately, however, it is not. There is also $`_\mathrm{J}`$, the Jacobian that arises in the Einstein frame. This term is not invariant under (3.12). From Eq. (3.9), we have
$`\delta _\xi \mathrm{\Sigma }_\mathrm{E}=\xi ^A𝒟_A\mathrm{\Sigma }_\mathrm{E}\mathrm{\Sigma }_\xi .`$ (3.18)
The first term on the RHS is the supersymmetry transformation of a normal chiral superfield. The second is a super-Weyl transformation of $`\mathrm{\Sigma }`$. This gives
$`\delta _\xi _\mathrm{J}`$ $`=`$ $`{\displaystyle \frac{1}{16\pi ^2}}(3T_R3T_G){\displaystyle d^2\mathrm{\Theta }2\mathrm{\Sigma }_\xi W^{(a)}W^{(a)}}+\mathrm{h}.\mathrm{c}.`$ (3.19)
$`=`$ $`e\left[{\displaystyle \frac{1}{16\pi ^2}}(3T_R3T_G)\varphi _\xi F_{mn}^{(a)}\stackrel{~}{F}^{mn(a)}+\mathrm{}\right].`$
Equation (3.19) exactly cancels the variation (3.16) and restores supersymmetry invariance in the Einstein frame.
Thus we have seen that the Einstein-frame 1PI effective Lagrangian is invariant under local supersymmetry transformations – provided the Jacobian (2.14) is added to the bare Lagrangian. Otherwise, local supersymmetry is explicitly broken at the quantum level because of the anomaly associated with the frame-restoring super-Weyl transformations. The Jacobian (2.14) is essential for the consistency of the quantum theory. The component expression for the Jacobian is given in Appendix B.
## 4 Physical Implications of the Jacobian: An Example
In this section, we show that the anomalous Jacobian $`_\mathrm{J}`$ has physical consequences. We illustrate this with an example of a scattering amplitude which requires the Jacobian to give a frame-independent result.
In what follows we consider a model with a no-scale Kähler potential of the form
$`K=3\mathrm{log}\left(1{\displaystyle \frac{1}{3}}\mathrm{\Phi }^{}\mathrm{\Phi }\right).`$ (4.1)
This Kähler potential (4.1) is chosen for simplicity; upon substitution into the superspace Lagrangian (2.1), it gives rise to canonically normalized scalars with a conformal coupling to gravity,
$`_{\mathrm{SUGRA}}=\sqrt{g}\left[{\displaystyle \frac{1}{2}}\left(1{\displaystyle \frac{1}{3}}A^{}A\right)g^{mn}_mA^{}_nA+\mathrm{}\right],`$ (4.2)
where we used the metric $`g_{mn}`$ instead of a vielbein. In Eq. (4.2), the subscript SUGRA indicates that this is the supergravity-frame Lagrangian. As discussed in the previous sections, we can use a super-Weyl transformation to pass to the Einstein frame.
In this model, we study the scattering process $`A^{}Av_mv_m`$ via a graviton exchange, where $`v_m`$ is a gauge boson and $`A`$ is a massless neutral scalar field. For simplicity, we assume that the gauge theory is pure supersymmetric Yang Mills.<sup>#6</sup><sup>#6</sup>#6If the matter fields had gauge quantum numbers, the following discussion would still hold, provided we replace $`3T_G`$ by $`3T_GT_R`$.
The amplitude of this process can be written as follows (see Fig. 1)
$`i(A^{}Av_mv_m)`$ $`=`$ $`{\displaystyle \frac{i}{2}}v_m(ϵ_1,p_1^{})v_m(ϵ_2,p_2^{})|T^{kl}|0\times i\mathrm{\Delta }_{kl,mn}`$ (4.3)
$`\times {\displaystyle \frac{i}{2}}0|T^{mn}|A^{}(p_1)A(p_2)`$
$`+v_m(ϵ_1,p_1^{})v_m(ϵ_2,p_2^{})|i_\mathrm{J}|A^{}(p_1)A(p_2),`$
where the $`p_i`$ denote the momenta of the scalars, while $`ϵ_i`$ and $`p_i^{}`$ denote the polarizations and momenta of the gauge bosons. Here, $`T_{mn}`$ is the energy momentum tensor,
$`T_{mn}={\displaystyle \frac{2}{\sqrt{g}}}{\displaystyle \frac{\delta }{\delta g^{mn}}},`$ (4.4)
while $`\mathrm{\Delta }_{kl,mn}`$ is the graviton propagator, given by
$`\mathrm{\Delta }_{mn,kl}(q)={\displaystyle \frac{2}{q^2}}(\eta _{km}\eta _{ln}+\eta _{kn}\eta _{lm}\eta _{kl}\eta _{mn}),`$ (4.5)
where $`\eta _{mn}`$ is the flat-space metric. We omit the part that depends on the gauge parameter because it does not contribute to the amplitude of interest.
For the scalars, the matrix element of the energy-momentum tensor is
$`0|T^{mn}|A^{}(p_1)A(p_2)=p_1^mp_2^n+p_1^np_2^m(p_1p_2)\eta ^{mn}+{\displaystyle \frac{\zeta }{3}}(q^2\eta ^{mn}q^mq^n),`$ (4.6)
where $`qp_1+p_2`$, and $`\zeta `$ is a coefficient proportional to the coupling of the scalars to the scalar curvature $``$; $`\zeta =1`$ in the supergravity frame and $`\zeta =0`$ in the Einstein frame. Note that $`\zeta =1`$ corresponds to conformally coupled scalars; this is easy to see upon computing the trace of Eq. (4.6) with $`q^2=2p_1p_2`$.
Substituting Eqs. (4.5) and (4.6) into Eq. (4.3), we obtain the amplitude
$``$ $`=`$ $`{\displaystyle \frac{2}{q^2}}p_{1k}p_{2l}v_mv_m|T^{kl}|0+{\displaystyle \frac{\zeta }{6}}v_mv_m|T_k^k|0+v_mv_m|_\mathrm{J}|A^{}A`$ (4.7)
$``$ $`_0+{\displaystyle \frac{\zeta }{6}}v_mv_m|T_k^k|0+v_mv_m|_\mathrm{J}|A^{}A,`$
where $`_0`$ is frame independent. At the classical level, with $`_\mathrm{J}`$ absent, the amplitudes in both frames are identical because the gauge boson energy momentum tensor is traceless.
At the quantum level, the frame independence is more subtle. First, the gauge boson energy momentum tensor is no longer traceless. Second, there is a Jacobian in the Einstein frame, but not the supergravity frame. To see what happens, let us consider the supergravity frame. We take $`\zeta =1`$ and use the trace anomaly relation
$`T_k^k={\displaystyle \frac{3}{32\pi ^2}}T_GF_{mn}^{(a)}F^{mn(a)},`$ (4.8)
to find
$`_{\mathrm{SUGRA}}=_0+{\displaystyle \frac{3}{192\pi ^2}}T_Gv_mv_m|F_{mn}^{(a)}F^{mn(a)}|0.`$ (4.9)
In the Einstein frame, where $`\zeta =0`$, there is no contribution from the $`T_k^k`$ term. However, in this frame there is the super-Weyl Jacobian,
$`_\mathrm{J}`$ $`=`$ $`{\displaystyle \frac{1}{16\pi ^2}}(3T_G){\displaystyle d^2\mathrm{\Theta }2\mathrm{\Sigma }_\mathrm{E}W^{(a)}W^{(a)}}+\mathrm{h}.\mathrm{c}.`$ (4.10)
$`=`$ $`{\displaystyle \frac{3}{192\pi ^2}}T_GA^{}AF_{mn}^{(a)}F^{mn(a)}+\mathrm{},`$
where we have used $`\mathrm{\Sigma }_\mathrm{E}|=\frac{1}{12}A^{}A+\mathrm{}`$. With this Jacobian, the Einstein-frame matrix element is
$`_\mathrm{E}=_0+{\displaystyle \frac{3}{192\pi ^2}}T_Gv_mv_m|F_{mn}^{(a)}F^{mn(a)}|0,`$ (4.11)
which is in complete agreement with the amplitude in the supergravity frame. The results in the two frames are identical because of the super-Weyl Jacobian.
## 5 Summary
In this paper, we studied the quantum consistency of the supergravity Lagrangian. We used a superspace approach in which the supergravity Lagrangian does not automatically give canonically normalized Einstein gravity. In the literature, Einstein gravity is recovered after a redefinition of the component fields. In this paper, we showed that the field redefinition is, in fact, a super-Weyl transformation, and we demonstrated a systematic way to do the field redefinition in superspace. This approach provides us with a clear understanding of supersymmetry transformations in the Einstein frame.
Supersymmetry transformations in the Einstein frame must preserve the Einstein frame condition, so they differ from the original supersymmetry transformations defined in the supergravity frame. We showed that the Einstein-frame supersymmetry transformations are ordinary supersymmetry transformations, $`\delta _{\mathrm{SUSY}}`$, combined with compensating super-Weyl transformations, $`\delta _{\mathrm{SW}}`$, which are necessary to maintain the Einstein frame condition.
The compensating super-Weyl transformations are, at the quantum level, anomalous. Because of this fact, one must be careful when studying the supersymmetry invariance of the quantum effective action. In this paper we emphasized that the super-Weyl transformation used to pass to the Einstein frame is anomalous. It gives rise to an anomalous Jacobian that must be included in the bare Einstein-frame Lagrangian. The 1PI Einstein-frame Lagrangian is supersymmetric because the variation of this Jacobian precisely cancels the anomaly arising from the frame-restoring super-Weyl transformation. If the Jacobian were omitted, the 1PI Lagrangian would not be invariant under Einstein-frame supersymmetry transformations. Consistency demands that the Jacobian be included in the bare Einstein-frame Lagrangian.
## Acknowledgments
The work of J.B. is supported by the U.S. National Science Foundation, grant NSF-PHY-9970781. T.M. is supported by U.S. National Science Foundation under grant NSF-PHY-9513835, and by the Marvin L. Goldberger Membership. E.P. is supported by the Department of Energy, contract DOE DE-FG0292ER-40704.
## Appendix A Globally Supersymmetric Gauge Theories
In this Appendix, we show how the supersymmetry transformations are defined in globally supersymmetric gauge theories. In particular, we demonstrate how the Wess-Zumino gauge condition is maintained after supersymmetry transformations. We will see that there is a close analogy between supersymmetry transformations in globally supersymmetric gauge theories and supergravity transformations in the Einstein frame.
Consider a globally supersymmetric gauge theory, with the Lagrangian
$`=_{\mathrm{gauge}}+{\displaystyle d^4\theta \mathrm{\Phi }^{}e^V\mathrm{\Phi }},`$ (A.1)
where $`_{\mathrm{gauge}}`$ is the kinetic term for the gauge multiplet. This Lagrangian is invariant under the superspace supersymmetry transformations
$`\delta _{\mathrm{SUSY}}V=\xi ^A_AV,\delta _{\mathrm{SUSY}}\mathrm{\Phi }=\xi ^A_A\mathrm{\Phi },`$ (A.2)
where $`\xi ^a=i\xi \sigma ^a\overline{\theta }+i\theta \sigma ^a\overline{\xi }`$ and $`\xi ^\alpha `$ is the supersymmetry transformation parameter. The component transformations can be determined by expanding (A.2) in powers of $`\theta `$.
The Lagrangian (A.1) contains the lower components of $`V`$, which are gauge degrees of freedom. Usually, it is convenient to use a Lagrangian in which these lower components are eliminated by a gauge transformation. This “frame” is often called “the Wess-Zumino gauge;” it is obtained by the following field redefinitions:
$`V_{\mathrm{WZ}}=V\mathrm{\Lambda }\mathrm{\Lambda }^{},\mathrm{\Phi }_{\mathrm{WZ}}=e^\mathrm{\Lambda }\mathrm{\Phi },`$ (A.3)
where $`\mathrm{\Lambda }`$ is a chiral superfield. The gauge parameter $`\mathrm{\Lambda }`$ is chosen so that all the unphysical fields are eliminated from the Lagrangian. The conditions are
$`V|=\mathrm{\Lambda }|+\mathrm{\Lambda }^{}|,(D_\alpha V)|=(D_\alpha \mathrm{\Lambda })|,(D^2V)|=(D^2\mathrm{\Lambda })|,`$ (A.4)
so $`\mathrm{\Lambda }`$ in the chiral basis is simply
$`\mathrm{\Lambda }={\displaystyle \frac{1}{2}}C+i\varphi +i\theta \chi +{\displaystyle \frac{i}{2}}\theta ^2(M+iN),`$ (A.5)
where $`C`$, $`\chi `$, $`M+iN`$ are the lowest, $`\theta `$ and $`\theta ^2`$ components of the vector superfield $`V`$, respectively. Note that the field $`\varphi `$ is not determined by the Wess-Zumino gauge conditions; it is the parameter of an ordinary gauge transformation. In terms of the new variables, the Lagrangian becomes
$`_{\mathrm{WZ}}`$ $`=`$ $`_{\mathrm{gauge}}+{\displaystyle d^4\theta \mathrm{\Phi }_{\mathrm{WZ}}^{}e^{V\mathrm{\Lambda }\mathrm{\Lambda }^{}}\mathrm{\Phi }_{\mathrm{WZ}}}`$ (A.6)
$`=`$ $`_{\mathrm{gauge}}+{\displaystyle d^4\theta \mathrm{\Phi }_{\mathrm{WZ}}^{}e^{V_{\mathrm{WZ}}}\mathrm{\Phi }_{\mathrm{WZ}}}.`$
The Wess-Zumino gauge Lagrangian $`_{\mathrm{WZ}}`$ contains only the physical fields.
The Wess-Zumino gauge conditions, however, are not preserved by the supersymmetry transformations (A.2). They must be supplemented by compensating superfield gauge transformations,
$`\delta _\xi V_{\mathrm{WZ}}=\xi ^A_AV_{\mathrm{WZ}}+\mathrm{\Lambda }_\xi ,\delta _\xi \mathrm{\Phi }_{\mathrm{WZ}}=\xi ^A_A\mathrm{\Phi }_{\mathrm{WZ}}+\mathrm{\Lambda }_\xi \mathrm{\Phi }_{\mathrm{WZ}},`$ (A.7)
where, in the chiral basis,
$`\mathrm{\Lambda }_\xi =\theta \sigma ^m\overline{\xi }(2v_m+2_m\varphi )+\theta ^2\overline{\xi }\overline{\lambda }.`$ (A.8)
Here $`v_m`$ and $`\lambda `$ are the gauge boson and gaugino fields, respectively.
The transformations (A.7) are combinations of the original supersymmetry transformations (A.2) and the frame-restoring gauge transformations. They leave invariant the Wess-Zumino-gauge Lagrangian. Furthermore, if we expand the LHS of (A.7) in powers of $`\theta `$, we obtain the Wess-Zumino-gauge supersymmetry transformations given, for example, in . Note that in Wess-Zumino gauge, a theory with a gauge anomaly would also have a supersymmetry anomaly because the transformations (A.2) contain ordinary gauge transformations.
Finally, we comment on the difference between (A.8) and (3.10), that is, on the different way we treat the imaginary part of the lowest component of the compensating transformation parameters. In each case, the term is not determined by the Wess-Zumino gauge/Einstein frame conditions. In globally supersymmetric gauge theory, we choose not to fix $`\varphi _\xi `$ in $`\mathrm{\Lambda }_\xi |`$; it is the degree for freedom associated with an ordinary gauge transformation. By contrast, in supergravity, we completely fix it and demand that the imaginary part of $`\mathrm{\Sigma }_\mathrm{E}|`$ vanish. The first term in (3.10) ensures that the imaginary part of $`\mathrm{\Sigma }_\mathrm{E}|`$ does not reappear in the Lagrangian after a supersymmetry transformation.
## Appendix B Component Expression for the Jacobian
In this Appendix, we present the complete component expression for the Jacobian that arises from the super-Weyl transformation required to pass to the Einstein frame.
As we have seen in this paper, the bare Lagrangian in Einstein-frame supergravity is given by
$`\widehat{}_{\mathrm{bare}}=_\mathrm{E}+_\mathrm{J}`$ (B.1)
where $`_\mathrm{E}`$ is the classical supergravity Lagrangian whose component expression is given, for example, in . $`_\mathrm{J}`$ is the Jacobian. At one-loop level, $`_\mathrm{J}`$ is given by
$`_\mathrm{J}={\displaystyle \frac{1}{16\pi ^2}}(3T_R3T_G){\displaystyle d^2\mathrm{\Theta }2\mathrm{\Sigma }_\mathrm{E}W^{(a)}W^{(a)}}+\mathrm{h}.\mathrm{c}.,`$ (B.2)
where the chiral superfield $`\mathrm{\Sigma }_\mathrm{E}`$ is given in Eqs. (3.4) and (3.5) with $`\varphi =0`$,
$`\mathrm{\Sigma }_\mathrm{E}=A_\mathrm{\Sigma }+\sqrt{2}\mathrm{\Theta }\chi _\mathrm{\Sigma }+\mathrm{\Theta }^2F_\mathrm{\Sigma },`$ (B.3)
with
$`A_\mathrm{\Sigma }={\displaystyle \frac{1}{12}}K,\chi _\mathrm{\Sigma }={\displaystyle \frac{1}{6}}K_i\chi ^i,F_\mathrm{\Sigma }={\displaystyle \frac{1}{6}}K_iF^i{\displaystyle \frac{1}{12}}K_{ij}\chi ^i\chi ^j.`$ (B.4)
Expanding Eq. (B.2), we obtain
$`e^1_\mathrm{J}`$ $`=`$ $`{\displaystyle \frac{1}{16\pi ^2}}(3T_R3T_G)[A_\mathrm{\Sigma }F_{mn}^{(a)}F_{}^{mn}{}_{}{}^{(a)}`$ (B.5)
$`2iA_\mathrm{\Sigma }\lambda ^{(a)}\sigma ^m\left(𝒟_m\overline{\lambda }^{(a)}f^{abc}v_m^{(b)}\overline{\lambda }^{(c)}+{\displaystyle \frac{i}{2}}b_m\overline{\lambda }^{(a)}\right)`$
$`2iA_\mathrm{\Sigma }\overline{\lambda }^{(a)}\overline{\sigma }^m\left(𝒟_m\lambda ^{(a)}f^{abc}v_m^{(b)}\lambda ^{(c)}{\displaystyle \frac{i}{2}}b_m\lambda ^{(a)}\right)`$
$`+2A_\mathrm{\Sigma }D_{\mathrm{aux}}^{}{}_{}{}^{(a)}D_{\mathrm{aux}}^{}{}_{}{}^{(a)}`$
$`+iA_\mathrm{\Sigma }\left(\psi _m\sigma ^{kl}\sigma ^m\overline{\lambda }^{(a)}+\overline{\psi }_m\overline{\sigma }^{kl}\overline{\sigma }^m\lambda ^{(a)}\right)\left(F_{kl}^{(a)}+\widehat{F}_{kl}^{}{}_{}{}^{(a)}\right)`$
$`\sqrt{2}\left(\chi _\mathrm{\Sigma }\sigma ^{mn}\lambda ^{(a)}+\overline{\chi }_\mathrm{\Sigma }\overline{\sigma }^{mn}\overline{\lambda }^{(a)}\right)F_{mn}^{(a)}`$
$`+\sqrt{2}i\left(\chi _\mathrm{\Sigma }\sigma ^{mn}\lambda ^{(a)}\psi _m\sigma _n\overline{\lambda }^{(a)}+{\displaystyle \frac{1}{4}}\overline{\psi }_m\overline{\sigma }^m\chi _\mathrm{\Sigma }\lambda ^{(a)}\lambda ^{(a)}\right)`$
$`+\sqrt{2}i\left(\overline{\chi }_\mathrm{\Sigma }\overline{\sigma }^{mn}\overline{\lambda }^{(a)}\overline{\psi }_m\overline{\sigma }_n\lambda ^{(a)}+{\displaystyle \frac{1}{4}}\psi _m\sigma ^m\overline{\chi }_\mathrm{\Sigma }\overline{\lambda }^{(a)}\overline{\lambda }^{(a)}\right)`$
$`+\sqrt{2}i\left(\chi _\mathrm{\Sigma }\lambda ^{(a)}\overline{\chi }_\mathrm{\Sigma }\overline{\lambda }^{(a)}\right)D_{\mathrm{aux}}^{}{}_{}{}^{(a)}`$
$`F_\mathrm{\Sigma }\lambda ^{(a)}\lambda ^{(a)}F_\mathrm{\Sigma }^{}\overline{\lambda }^{(a)}\overline{\lambda }^{(a)}]`$
where $`\widehat{F}_{mn}^{}{}_{}{}^{(a)}`$ is the supercovariant field strength,
$`\widehat{F}_{mn}^{}{}_{}{}^{(a)}=F_{mn}^{(a)}{\displaystyle \frac{i}{2}}\left(\psi _m\sigma _n\overline{\lambda }^{(a)}+\overline{\psi }_m\overline{\sigma }_n\lambda ^{(a)}\psi _n\sigma _m\overline{\lambda }^{(a)}\overline{\psi }_n\overline{\sigma }_m\lambda ^{(a)}\right).`$ (B.6)
For a detailed explanation of the notation, see .
The full component expression is given by substituting the solutions to the Einstein-frame auxiliary field equations of motion,
$`F^i`$ $`=`$ $`(K^1)^{ij^{}}\left(e^{K/2}D_j^{}P^{}+{\displaystyle \frac{1}{2}}K_{j^{}kl}\chi ^k\chi ^l+{\displaystyle \frac{1}{4}}_j^{}h_{(ab)}\overline{\lambda }^{(a)}\overline{\lambda }^{(b)}\right)`$
$`D_{\mathrm{aux}}^{}{}_{}{}^{(a)}`$ $`=`$ $`h_{}^{\mathrm{R}(ab)}{}_{}{}^{1}\left[D^{(b)}+{\displaystyle \frac{i}{2\sqrt{2}}}\left(_ih_{(bc)}\chi ^i\lambda ^{(c)}_i^{}h_{(bc)}\overline{\chi }^i\overline{\lambda }^{(c)}\right)\right]`$
$`b_m`$ $`=`$ $`{\displaystyle \frac{i}{2}}\left(K_i\stackrel{~}{𝒟}_mA^iK_i^{}\stackrel{~}{𝒟}_mA^i\right)+{\displaystyle \frac{1}{4}}K_{ij^{}}\chi ^i\sigma _m\overline{\chi }^j`$ (B.7)
$`{\displaystyle \frac{3}{4}}h_{(ab)}^\mathrm{R}\lambda ^{(a)}\sigma _m\overline{\lambda }^{(b)}+i\left[{\displaystyle \frac{1}{2}}\left(K_iX^{i(a)}K_i^{}X^{i(a)}\right)+iD^{(a)}\right]v_m^{(a)},`$
where $`D_iPP_i+K_iP`$, $`X^{(a)}`$ is the Killing vector, $`D^{(a)}`$ is the Killing potential associated with $`X^{(a)}`$, and $`\stackrel{~}{𝒟}_mA^i𝒟_mA^iv_m^{(a)}X_{(a)}^i`$ .
|
warning/0003/quant-ph0003060.html
|
ar5iv
|
text
|
# Partial Teleportation of Entanglement in the Noisy Environment
## I introduction
Quantum teleportation of a single-particle state has been extensively studied both theoretically and experimentally . Quantum teleportation reproduces an unknown quantum state at a remote place while the original state is destroyed . The key of the quantum teleportation is the quantum channel composed of the quantum entangled pair. If the quantum channel is maximally entangled, for example, by using the singlet state, the quantum state is perfectly reproduced at the remote place and the fidelity of the teleportation is unity. However, in the real world, the quantum channel lies in the noisy environment, which degrades the entanglement of the channel. The less is the quantum channel entangled, the smaller is the fidelity . It has also been found that the fidelity of the quantum teleportation is always larger than that of any classical communication protocol even in the noisy environment.
In this paper, we are interested in partial teleportation of an entangled state of a two spin-1/2 system. An entangled pair of particles are prepared by Alice who wants to teleport one of the entangled pair to Bob as shown in Fig. 1. If the quantum channel is maximally entangled, the partial teleportation is nothing more than entanglement swapping . Bennett et al. argued that teleportation is a linear operation for the perfect quantum channel and could be extended to what is now called entanglement swapping , which has been experimentally realised . Entanglement swapping was considered for a more generalised multi-particle system and for concentration of partially entangled states . In this paper, we analyse the environmental effects on the partial teleportation of the entangled state, considering the entanglement transfer and the fidelity. We define a measure of entanglement using the partial transposition. Although this measure does not completely agree with the entropy of entanglement for a pure state, it is useful and gives qualitative information. As it is easily calculated and satisfies important conditions of the measure of entanglement, we use it to analyse the partial teleportation in this paper. The concept of correlation information is also introduced. The correlation information we define is, in general, dependent on classical and quantum correlation but we show that it bears a simple and useful linear relation for the partial teleportation.
We assume that the initial entangled state is pure and the mixed quantum channel is represented by the Werner state. We show that the calculation is much simpler as we employ the Werner-state channel while we do not lose the generality of the treatment. Entanglement transfer was extensively studied by Schumacher and teleportation through general channels was considered by Horodecki et al. .
## II measures of entanglement
For a pure state $`\widehat{\rho }`$ of a bipartite system, one can choose the entropy $`S`$ of entanglement as a measure of entanglement, where
$$S=\mathrm{Tr}\widehat{\rho }_a\mathrm{log}_2\widehat{\rho }_a=\mathrm{Tr}\widehat{\rho }_b\mathrm{log}_2\widehat{\rho }_b$$
(1)
where $`\widehat{\rho }_{a,b}=\mathrm{Tr}_{b,a}\widehat{\rho }`$ is the reduced density matrix for the subsystem $`a`$ or $`b`$. For a mixed state, there have been many definitions for the measure of entanglement such as entanglement of formation , quantum relative entropy, and Bures metric . Every measure $`E(\widehat{\rho })`$ should satisfy the following necessary conditions for a given density matrix $`\widehat{\rho }`$ ,
* $`E(\widehat{\rho })=0`$ if and only if $`\widehat{\rho }`$ is separable.
* A local unitary transformation leaves $`E(\widehat{\rho })`$ invariant;
$$E(\widehat{U}_1\widehat{U}_2\widehat{\rho }\widehat{U}_1^{}\widehat{U}_2^{})=E(\widehat{\rho })$$
(2)
for all unitary operators $`\widehat{U}_1`$ and $`\widehat{U}_2`$.
* $`E(\widehat{\rho })`$ cannot increase under local general measurements (LGM), classical communications (CC), and post selection of subensemble (PSS),
$$p_iE(\widehat{\rho }_i)E(\widehat{\rho }),$$
(3)
where $`p_i\widehat{\rho }_i=\widehat{A}_i\widehat{B}_i\widehat{\rho }\widehat{A}_i^{}\widehat{B}_i^{}`$ with $`p_i=\text{Tr}\widehat{A}_i\widehat{B}_i\widehat{\rho }\widehat{A}_i^{}\widehat{B}_i^{}`$; two set of LGM operators $`\{\widehat{A}_i\}`$ and $`\{\widehat{B}_i\}`$ are classically correlated by CC.
The requirement of the condition (C.1) is clear since a separable state is just classically correlated and should be independent of the entanglement. Since a local unitary operation is performed only locally, it cannot affect any entanglement, required by the condition (C.2). The condition (C.3) is related with the purification procedure which selects a subensemble of maximally-quantum-correlated pairs among an impure ensemble . The purification procedure can distill maximally entangled states such that $`E(\widehat{\rho }_i)E(\widehat{\rho })`$ for a certain route $`i`$, but the average entanglement cannot increase over the whole ensemble since quantum nonlocal operations are not introduced, represented in the condition (C.3).
For a two spin-1/2 system, we define the measure of entanglement in terms of the negative eigenvalues of the partial transposition of the state. Consider a density matrix $`\widehat{\rho }`$ for a two spin-1/2 system and its partial transposition $`\widehat{\sigma }=\widehat{\rho }^{T_2}`$. The density matrix $`\widehat{\rho }`$ is inseparable if and only if $`\widehat{\sigma }`$ has any negative eigenvalues. The measure of entanglement $`E(\widehat{\rho })`$ is then defined as
$$E(\widehat{\rho })=2\underset{i}{}\lambda _i^{}$$
(4)
where $`\lambda _i^{}`$ are the negative eigenvalues of $`\widehat{\sigma }`$ and the factor 2 is introduced to have $`0E(\widehat{\rho })1`$. In Appendix, we prove that the entanglement measure (4) satisfies the above necessary conditions.
In fact, there is the fourth condition which a measure of entanglement has to satisfy:
* For pure states, the measure of entanglement reduces to the entropy of entanglement.
We note that for a pure entangled state the entanglement measure (4) is not reduced to the entropy of entanglement $`S`$ but is a monotonously increasing function of $`S`$ as shown in Fig. 2. Vedral and Plenio showed that the Bures metric satisfies the condition (C.1)-(C.3) but is smaller than the entropy of entanglement for pure states. They then wrote that measures which do not satisfy condition (C.4) can nevertheless contain useful information on entanglement. The Schmidt norm is another example of the measure of entanglement which does not satisfy condition (C.4) . The entanglement measure (4) can qualitatively give information on the entanglement of a given state as it satisfies conditions (C.1)-(C.3). Because of convenience in calculation, we use $`E(\widehat{\rho })`$ in Eq. (4) as the measure of entanglement in this paper.
## III partial teleportation of entanglement
We consider the partial teleportation of entanglement as shown in Fig. 1, where Alice teleports one particle of her entangled pair to Bob. Alice and Bob share an ancillary pair of an entangled state. Alice performs the Bell-state measurement on one of her original entangled pair and her part of the ancillary pair. Upon receiving Alice’s measurement result through a classical channel, Bob unitary rotates his part of the ancillary pair based on it. If the ancillary pair is perfectly entangled, Bob’s particle and Alice’s unmeasured particle become entangled as the Alice’s original entangled pair. The basic idea of the partial teleportation is similar to the single-particle teleportation or entanglement swapping but our interest here does not stop at a simple result of the teleportation of a particle. In this paper, the quantum channel is represented by a mixed entangled state due to the influence from the environment. We are interested in how the entanglement of the teleported state is affected by such the imperfect quantum channel by studying the channel-dependent fidelity, information transfer and entanglement transfer. We assume for the simplicity that Alice’s initial state is pure and the quantum channel is represented by a Werner state .
An initial pure state for entangled two particles 1 and 2 is given in the Hilbert-Schmidt space by
$$\widehat{\rho }^{12}=\frac{1}{4}\left(\widehat{1}\widehat{1}+\stackrel{}{a}_0\stackrel{}{\sigma }\widehat{1}+\widehat{1}\stackrel{}{b}_0\stackrel{}{\sigma }+\underset{nm}{}c_0^{nm}\widehat{\sigma }_n\widehat{\sigma }_m\right)$$
(5)
where $`\stackrel{}{a}_0`$ and $`\stackrel{}{b}_0`$ are real vectors and $`c_0^{nm}`$ is an element of the real matrix $`C_0`$. The initial pure state $`\widehat{\rho }_0`$ in Eq. (5) satisfies the pure state condition $`\widehat{\rho }^2=\widehat{\rho }`$. We can also consider a general representation of the initial pure state (5), with help of the seed state $`\widehat{\rho }_s^{12}`$ defined as follows
$$\widehat{\rho }^{12}=(\widehat{U}^1\widehat{U}^2)\widehat{\rho }_s^{12}(\widehat{U}^1\widehat{U}^2)^{}$$
(6)
where $`\widehat{U}^1`$ and $`\widehat{U}^2`$ are local unitary operators acting respectively on the particles 1 and 2. The density operator for the seed state is
$$\widehat{\rho }_s^{12}=\frac{1}{4}\left(\widehat{1}\widehat{1}+a_0\widehat{\sigma }_z\widehat{1}+\widehat{1}a_0\widehat{\sigma }_z+\underset{n}{}c_n\widehat{\sigma }_n\widehat{\sigma }_n\right)$$
(7)
where $`a_0`$ is a positive real number and $`\stackrel{}{c}=(c_0,c_0,1)`$ a real vector, constrained by $`a_0^2+c_0^2=1`$. The vector $`\stackrel{}{c}`$ describes the quantum correlation of the pure state $`\widehat{\rho }_s^{12}`$, yielding the relation $`E_0=|c_0|`$, where $`E_0`$ is the measure of entanglement for $`\widehat{\rho }_s^{12}`$. The state has no entanglement, i.e., $`E_0=0`$, if and only if $`c_0=0`$. Now, the state $`\widehat{\rho }_s^{12}`$ is characterised by one parameter $`c_0`$ or equivalently by its entanglement $`E_0`$. By the definition of the measure of entanglement, the initial state of the density operator $`\widehat{\rho }^{12}`$ and the state of the seed density operator $`\widehat{\rho }_s^{12}`$ have the same measure of entanglement, $`E_0`$. It is thus clear that the initial state $`\widehat{\rho }^{12}`$ is fully determined by its measure of entanglement $`E_0`$ and the local unitary operations $`\widehat{U}^1`$ and $`\widehat{U}^2`$.
We take the Werner state for the quantum channel. In fact, any mixed state can be made a Werner state by random local $`SU(2)SU(2)`$ operations so that we do not lose the generality by taking the Werner state in describing the mixed channel. The Werner state $`\widehat{w}^{34}`$ of the ancillary particles 3 and 4 is
$$\widehat{w}^{34}=\frac{1}{4}\left(\widehat{1}\widehat{1}+\underset{nm}{}c_w^{nm}\widehat{\sigma }_n\widehat{\sigma }_n\right).$$
(8)
where $`c_w^{nm}`$ is an element of the real matrix $`C_w=(2\mathrm{\Phi }+1)/3\mathrm{diag}(1,1,1)`$. The Werner state becomes the singlet state when $`\mathrm{\Phi }=1`$. The parameter $`\mathrm{\Phi }`$ is related to the entanglement $`E_w`$ of the Werner state $`\widehat{w}^{34}`$. It is straightforward to show that $`E_w=\text{max}(0,\mathrm{\Phi })`$.
The Bell-state measurement by Alice is represented by a family of projectors
$$\widehat{P}_\alpha ^{23}=|\mathrm{\Psi }_\alpha ^{23}\mathrm{\Psi }_\alpha ^{23}|=\frac{1}{4}\left(II+\underset{nm}{}p_\alpha ^{nm}\sigma _n\sigma _m\right)$$
(9)
where $`|\mathrm{\Psi }_\alpha ^{23}`$ are the four possible Bell states and $`p_\alpha ^{nm}`$ is an element of the real matrix $`P_\alpha `$; $`P_0=\text{diag}(1,1,1)`$, $`P_1=\text{diag}(1,1,1)`$, $`P_2=\text{diag}(1,1,1)`$, $`P_3=\text{diag}(1,1,1)`$. The Bell measurement is performed on the particle 2 of the initial entangled pair and the particle 3 of the Werner state (see Fig. 1).
The quantum teleportation utilises both the classical and quantum channels. Upon receiving the two-bit classical message on the Bell-state measurement through the classical channel, Bob performs the unitary transformation on the particle 4 accordingly. If the quantum channel is in the spin singlet state, the teleportation can be perfectly completed by one of the following four possible unitary operators: $`\widehat{1}`$, $`\widehat{\sigma }_x`$, $`\widehat{\sigma }_y`$, and $`\widehat{\sigma }_z`$. However, for the mixed channel, it is difficult for Bob to decide which unitary transformation to get the final state $`\widehat{\rho }^{14}`$ maximally close to the initial state $`\widehat{\rho }^{12}`$. Let us consider how to determine the right unitary transformation when the channel is in the Werner state.
Suppose Bob receives a two-bit message through the classical channel saying that Alice’s measurement was $`|\mathrm{\Psi }_\alpha ^{23}`$. Bob then applies the unitary transformation $`\widehat{U}_\alpha ^4`$ on the particle 4, then the state $`\widehat{\rho }_\alpha ^{14}`$ of the two particles 1 and 4 becomes
$$\widehat{\rho }_\alpha ^{14}=\frac{1}{p_\alpha }\mathrm{Tr}_{2,3}\left[\widehat{P}_\alpha ^{23}\widehat{U}_\alpha ^4\left(\widehat{\rho }_0^{12}\widehat{w}^{34}\right)\widehat{P}_\alpha ^{23}\widehat{U}_{\alpha }^{4}{}_{}{}^{}\right]$$
(10)
where $`p_\alpha `$ is the probability of $`|\mathrm{\Psi }_\alpha ^{23}`$ to be the result of Alice’s measurement. Eq. (10) can be written in the Hilbert-Schmidt space as
$$\widehat{\rho }_\alpha ^{14}=\frac{1}{4}\left(\widehat{1}\widehat{1}+\stackrel{}{a}_\alpha \stackrel{}{\sigma }\widehat{1}+\widehat{1}\stackrel{}{b}_\alpha \stackrel{}{\sigma }+\underset{nm}{}\stackrel{~}{c}_\alpha ^{nm}\widehat{\sigma }_n\widehat{\sigma }_m\right).$$
(11)
with the parameters
$`\stackrel{}{a}_\alpha `$ $`=`$ $`\stackrel{}{a}_0,`$ (12)
$`\stackrel{}{b}_\alpha `$ $`=`$ $`{\displaystyle \frac{2\mathrm{\Phi }+1}{3}}O_\alpha ^TP_\alpha \stackrel{}{b}_0,`$ (13)
$`C_\alpha `$ $`=`$ $`{\displaystyle \frac{2\mathrm{\Phi }+1}{3}}O_\alpha ^TP_\alpha C_0.`$ (14)
Here we have the rotation matrix $`O_\alpha `$ in the Bloch space for a single particle state, obtained from the unitary operator $`\widehat{U}_\alpha `$ :
$$\widehat{U}_\alpha \stackrel{}{a}\stackrel{}{\sigma }\widehat{U}_\alpha ^{}=\left(O_\alpha ^T\stackrel{}{a}\right)\stackrel{}{\sigma }.$$
(15)
The fidelity $`F`$ measures how close the final state $`\widehat{\rho }^{14}`$ is to the initial state $`\widehat{\rho }^{12}`$; $`F=_\alpha p_\alpha \mathrm{Tr}\widehat{\rho }^{12}\widehat{\rho }_\alpha ^{14}`$. If the teleportation is perfect, the final state is the same as the initial state so that the fidelity is 1. By substituting $`\widehat{\rho }^{12}`$ in (5) and $`\widehat{\rho }_\alpha ^{14}`$ in (12) into the definition of the fidelity, we find the fidelity for the Werner-state channel
$`F`$ $`=`$ $`{\displaystyle \frac{1}{4}}[1+|\stackrel{}{a}_0|^2+{\displaystyle \frac{2\mathrm{\Phi }+1}{3}}\stackrel{}{b}_0({\displaystyle \frac{1}{4}}{\displaystyle \underset{\alpha }{}}O_\alpha ^TP_\alpha \stackrel{}{b}_0)`$ (17)
$`+{\displaystyle \frac{2\mathrm{\Phi }+1}{3}}\text{Tr}({\displaystyle \frac{1}{4}}{\displaystyle \underset{\alpha }{}}O_\alpha ^TP_\alpha C_0^TC_0)].`$
The task is now to find Bob’s unitary operations $`\widehat{U}_\alpha `$ to maximise fidelity (17). For a general mixed channel, the fidelity is a function of the initial and channel states as well as Bob’s unitary operation. However, the basic assumption of the quantum teleportation is that the initial state is unknown. In order to examine the faithfulness of quantum teleportation, we need to average the fidelity over the Hilbert space where the initial state lies in. The unitary operations should be determined to maximise the average fidelity .
For the Werner-state channel the fidelity has been calculated as in Eq. (17), where the measurement dependence is found in the terms including $`_\alpha O_\alpha ^TP_\alpha `$. It is clear that $`P_\alpha `$ in Eq. (9) is a rotation matrix thus $`|O_\alpha ^TP_\alpha |1`$. Choosing $`O_\alpha =P_\alpha `$ to maximise the fidelity (17), we find that the corresponding unitary operators are the same as in the singlet-state channel discussed above. This choice enables Bob to produce the measurement-independent final state $`\widehat{\rho }_\alpha ^{14}=\widehat{\rho }^{14}`$.
The fidelity for the Werner-state channel is then given by
$$F=\frac{1}{4}\left(1+|\stackrel{}{a}_0|^2+\frac{2\mathrm{\Phi }+1}{3}|\stackrel{}{b}_0|^2+\frac{2\mathrm{\Phi }+1}{3}\text{Tr}C_0^TC_0\right).$$
(18)
It is seen that the fidelity is invariant for the local unitary transformations on the initial pure state. Noting $`|\stackrel{}{a}_0|`$, $`|\stackrel{}{b}_0|`$ and $`\text{Tr}C_0^TC_0`$ are uniquely determined by the entanglement $`E_0`$ for the initial state, the fidelity can be finally written as
$$F=\frac{E_w+2}{3}+\frac{E_w1}{6}E_0^2.$$
(19)
It is clear that the fidelity (19) depends on the initial-state entanglement $`E_0`$ and channel entanglement $`E_w`$.
The large entanglement in the channel enhances the fidelity and the maximally entangled channel gives the unit fidelity independent from the initial-state entanglement $`E_0`$. When Alice’s initial state is disentangled, i.e. $`E_0=0`$, it can be written as a direct product of two individual states and the partial entanglement teleportation becomes equivalent to a single-particle case. In this case, the fidelity is $`F=(E_w+2)/3`$, equal to that for the single-particle teleportation . It has an upper bound of $`2/3`$ producible by classical communication of $`E_w=0`$ . For the entangled initial state with $`E_00`$, the large initial-state entanglement $`E_0`$ reduces monotonously the fidelity (19) for a given channel entanglement $`E_w<1`$ because the sign of the second term is negative in Eq. (19). This implies that the initial entanglement has fragile nature to teleport. In other words, the entanglement is destroyed easily by the environment.
To examine the loss of the initial entanglement, we consider the entanglement transfer by teleportation. We are interested in how much the initial-state entanglement is transferred to the final state. The measure of entanglement for the final state is calculated as
$$E(\widehat{\rho }^{14})=\frac{1}{3}\left[\sqrt{(1E_w)^2+3E_w(2+E_w)E_0^2}(1E_w)\right].$$
(20)
As $`E/E_w0`$, large entanglement of the channel enhances the entanglement transfer to the final state from the initial one. For a given entanglement of the channel, the entanglement of the final state increases as the initial-state entanglement gets larger. The entanglement of the final state is nonzero as far as $`E_w0`$ and $`E_00`$, which shows that the entangled channel transfers at least some of the initial entanglement to the final state.
## IV correlation information
Brukner and Zeilinger have recently derived a measure of information for a quantum state . The information measure is invariant for a choice of a complete set $`\{\widehat{A}_1,\mathrm{},\widehat{A}_m\}`$ of complementary observations and is conserved as far as there is no information exchange between the system and the environment . We employ the measure of information to study the quantum information transfer in the partial teleportation of entanglement.
Suppose an experimental arrangement for a measurement by observable $`\widehat{A}_j`$ which has $`n`$ possible outcomes with $`n`$ dimensional probability vector $`\stackrel{}{p}=(p_1,\mathrm{},p_i,\mathrm{},p_n)`$ for a given system. The system is supposed to have maximum $`k`$-bits of information such that $`n=2^k`$. The measure of information $`I_j`$ for the observable $`\widehat{A}_j`$ is defined as
$$I_j=𝒩\underset{i=1}{\overset{n}{}}\left(p_i\frac{1}{n}\right)^2.$$
(21)
where the normalisation constant $`𝒩=2^kk/(2^k1)`$. $`I_j`$ results in $`k`$ bits of information if one $`p_i=1`$ and 0 bits of information if all $`p_i`$ are equal. For a complete set of $`m`$ mutually complementary observables the measure of information is defined as the sum of the measures of information over the complete set
$$I(\widehat{\rho })=\underset{j=1}{\overset{m}{}}I_j.$$
(22)
for the quantum state $`\widehat{\rho }`$. A single spin-1/2 system, for example, is represented by the measure of information $`I(\widehat{\rho })=2\mathrm{T}\mathrm{r}\widehat{\rho }^21`$.
In the following, we define the measure of correlation information based on the measure of information introduced by Brukner and Zeilinger. Let us consider the measure of information for a composite system of two particles which can be decomposed into three parts. Each particle has its own information corresponding to its reduced density matrix, which we call the individual information. The two particles can also have the correlation information which depends on how much they are correlated.
The measure of total information for a density matrix $`\widehat{\rho }`$ of the two spin-1/2 particles is
$$I(\widehat{\rho })=\frac{2}{3}\left(4\mathrm{T}\mathrm{r}\widehat{\rho }^21\right).$$
(23)
The measures of individual information $`I_a(\widehat{\rho })`$ and $`I_b(\widehat{\rho })`$ for the particles $`a`$ and $`b`$ are
$`I_a(\widehat{\rho })`$ $`=`$ $`2\mathrm{T}\mathrm{r}_a\left(\widehat{\rho }_a\right)^21`$ (24)
$`I_b(\widehat{\rho })`$ $`=`$ $`2\mathrm{T}\mathrm{r}_b\left(\widehat{\rho }_b\right)^21`$ (25)
where $`\widehat{\rho }_{a,b}=\mathrm{Tr}_{b,a}\widehat{\rho }`$ are reduced density matrices for the particles $`a`$ and $`b`$. If the total density matrix $`\widehat{\rho }`$ is represented by $`\widehat{\rho }=\widehat{\rho }_a\widehat{\rho }_b`$, the total system is completely separable and we know that there is no correlation information. We thus define the measure of correlation information as
$`I_c(\widehat{\rho })=I(\widehat{\rho })I(\widehat{\rho }_a\widehat{\rho }_b)=I(\widehat{\rho }){\displaystyle \frac{2}{3}}[I_a(\widehat{\rho })+I_b(\widehat{\rho })`$ (26)
$`+I_a(\widehat{\rho })I_b(\widehat{\rho })].`$ (27)
The measures of individual and correlation information are invariant for any particular choice of the complete set of complementary observables.
If there is no correlation between the two particles, the measure of total information is a mere sum of the measures of individual information. On the other hand, the total information is imposed only on the correlation information, $`I=I_c`$, if there is no individual information, $`I_a=I_b=0`$. For a two spin-1/2 system, the maximally entangled states have only the correlation information.
Note that the correlation information is not the same as the measure of entanglement. Only when a pure state is considered the measure of the correlation is directly related to the measure of entanglement. For a mixed state, the correlation information also includes the information due to classical correlation. For example, the state of the density operator $`\widehat{\rho }_{cc}=1/4(\widehat{1}\widehat{1}\sigma _z\sigma _z)`$ is not quantum-mechanically entangled but classically correlated with the correlated information $`I_c0`$ .
The total information $`I(\widehat{\rho }^{14})`$ in the final state (11) is obtained using its definition (23):
$$I(\widehat{\rho }^{14})=\frac{2}{3}\left[1+2\left(\frac{2E_w+1}{3}\right)^2+\left\{\left(\frac{2E_w+1}{3}\right)^21\right\}E_0^2\right]$$
(28)
which depends on the initial-state and the channel entanglement. We can easily find that $`0I(\widehat{\rho }^{14})2`$ from the range of the entanglement measure $`0E_w,E_01`$. The measure of information for the initial state is 2 as it is a pure two spin-1/2 system. The final state can have at best the same amount of information as the initial state because the noisy environment acts only to dissipate the information. The total information is better preserved for the larger channel entanglement $`E_w`$. The total information is lost more easily for the larger initial entanglement as the sign of the coefficient for $`E_0^2`$ is negative. This is consistent with the discussions for the fidelity.
We evaluate the measures of individual $`I_1`$, $`I_4`$ and correlation $`I_c`$ information for the final state given by
$`I_1(\widehat{\rho }^{14})`$ $`=`$ $`I_1^0`$ (29)
$`I_4(\widehat{\rho }^{14})`$ $`=`$ $`\left({\displaystyle \frac{2E_w+1}{3}}\right)^2I_2^0`$ (30)
$`I_c(\widehat{\rho }^{14})`$ $`=`$ $`\left({\displaystyle \frac{2E_w+1}{3}}\right)^2I_c^0`$ (31)
where $`I_1^0=1E_0^2`$, $`I_2^0=1E_0^2`$, and $`I_c^0=2(4E_0^2)E_0^2/3`$ are the measures of individual and correlation information for the initial state $`\widehat{\rho }^{12}`$. Because there has been no action on Alice’s particle 1, its individual information remains unchanged with $`I_1=I_1^0`$. On the other hand, the individual information for the particle 2 is not fully transferred to the particle 4 and the correlation information is decreased. The coefficient for the decrease of the information is the same for $`I_4`$ and $`I_c`$ but we must remember that the maximum measure of correlation information is 2 while that of individual information is 1, which shows that the correlation information can be lost more easily.
Because the final state is a mixed state, its correlation information $`I_c`$ describes in general both the quantum and classical correlations. For an entangled initial state of $`E_00`$ we consider two cases: $`E_w=0`$ and $`E_w0`$. If $`E_w=0`$, the final state is classically correlated because $`I_c0`$ in Eq. (31) whereas $`E=E(\widehat{\rho }^{14})=0`$ in Eq. (20). On the other hand, if $`E_w0`$, the correlation information $`I_c`$ can be written in terms of the final entanglement $`E`$:
$`I_c=2\left({\displaystyle \frac{2E_w+1}{3}}\right)^2\left(43{\displaystyle \frac{E+(1E_w)}{E_w(2+E_w)}}E\right)`$ (32)
$`\times {\displaystyle \frac{E+(1E_w)}{E_w(2+E_w)}}E.`$ (33)
The correlation information $`I_c=0`$ if and only if $`E=0`$ for partial teleportation via the Werner channel. This shows that for $`E_w0`$ the correlation information of the final state is only due to the quantum correlation.
## V Remarks
The partial teleportation of entanglement has been considered in the noisy environment. The measures of individual and correlation information have been extensively studied for the two spin-1/2 system. As the Werner-state is employed for the quantum channel, the calculation of the fidelity, the information transfer and the entanglement transfer became extremely simple while we do not lose the generality to consider the noisy environment. The larger the initial-state entanglement is, the worse the fidelity becomes for any imperfect quantum channel. We, however, find more entanglement in the final state with the larger initial-state entanglement. The entangled channel transfers at least some of the entanglement to the final state.
###### Acknowledgements.
JL thanks Mr.Č. Bruckner for discussions at the Entanglement and Decoherence Workshop in Garda. This work was supported by the Brain Korea 21 grant (D-0055) of the Korean Ministry of Education.
## proof of the measure of entanglement Eq. (4)
Consider a density matrix $`\widehat{\rho }`$ for a two spin-1/2 system and the partial transposition $`\widehat{\sigma }=\widehat{\rho }^{T_2}`$ . The density matrix $`\widehat{\rho }`$ is inseparable if and only if $`\widehat{\sigma }`$ has any negative eigenvalues . The measure of entanglement $`(\widehat{\rho })`$ is defined as $`2_i(\lambda _i^{})`$ with the negative eigenvalues $`\lambda _i^{}`$ of $`\widehat{\sigma }`$. We will show that $`(\widehat{\rho })`$ satisfies the three conditions (C.1)-(C.3).
Let $`\widehat{d}(\widehat{\rho })`$ be a diagonal matrix of the partial transposition $`\widehat{\sigma }`$ such that, for some unitary operator $`U`$,
$`\widehat{d}(\widehat{\rho })`$ $`=`$ $`\left(U\widehat{\sigma }U^{}\right)`$ (34)
$`=`$ $`\text{diag}(\lambda _1,\lambda _2,\lambda _3,\lambda _4)`$ (35)
where $`\text{diag}(\{\lambda _i\})`$ represents a diagonal matrix with its diagonal elements $`\lambda _i`$ and thus $`\lambda _i`$ are eigenvalues of $`\widehat{\sigma }`$. For the given $`\widehat{d}(\widehat{\rho })`$, the density matrix space is decomposed into two subspaces; one is expanded by eigenvectors of the semi-positive (positive or zero) eigenvalues and the other of the negative eigenvalues of $`\widehat{d}`$. The identity operator $`\widehat{1}`$ is then the sum of two projectors $`\widehat{P}_+`$ and $`\widehat{P}_{}`$ such that
$$\widehat{1}=\widehat{P}_++\widehat{P}_{}$$
(36)
where $`\widehat{P}_+`$ ($`\widehat{P}_{}`$) projects the density matrix space onto the semi-positive (negative) eigenvalue subspace. Any hermitian matrix $`\widehat{H}`$ is decomposed into
$$\widehat{H}=\widehat{P}_+\widehat{H}\widehat{P}_++\widehat{P}_+\widehat{H}\widehat{P}_{}+\widehat{P}_{}\widehat{H}\widehat{P}_++\widehat{P}_{}\widehat{H}\widehat{P}_{}$$
(37)
and thus
$$\widehat{d}(\widehat{\rho })=\widehat{d}_+(\widehat{\rho })+\widehat{d}_{}(\widehat{\rho })$$
(38)
where $`\widehat{d}_+\widehat{P}_+\widehat{d}\widehat{P}_+`$ and $`\widehat{d}_{}\widehat{P}_{}\widehat{d}\widehat{P}_{}`$. Note that $`\widehat{P}_+\widehat{d}\widehat{P}_{}=\widehat{P}_{}\widehat{d}\widehat{P}_+=0`$ since $`\widehat{d}`$ is a diagonal matrix. Now, the measure of entanglement $`(\widehat{\rho })`$ is defined as twice the absolute value of the trace on the negative diagonal matrix $`\widehat{d}_{}(\widehat{\rho })`$, given by
$$(\widehat{\rho })2\text{Tr}\left[\widehat{d}_{}(\widehat{\rho })\right]=2\underset{\beta }{}(\lambda _\beta ^{})$$
(39)
where $`\lambda _\beta ^{}`$ is negative eigenvalue of $`\widehat{\sigma }`$ and the factor 2 is introduced to be $`0(\widehat{\rho })1`$.
Now we consider that $`E(\widehat{\rho })`$ in Eq. (4) satisfies the necessary conditions (C.1)-(C.3). If $`\widehat{\rho }`$ is separable, $`\widehat{\sigma }`$ has no negative eigenvalues and the converse statement also holds, satisfying the condition (C.1). A local unitary transformation leads to new density matrix $`\widehat{\rho }^{}=\widehat{U}_1\widehat{U}_2\widehat{\rho }\widehat{U}_1^{}\widehat{U}_2^{}`$ and its partial transposition $`\widehat{\sigma }^{}=\widehat{U}_1\widehat{U}_2^{}\widehat{\sigma }\widehat{U}_1^{}(\widehat{U}_2^{})^{}`$. Note that $`\widehat{U}_1\widehat{U}_2^{}`$ is a unitary operator such that $`\widehat{U}_1\widehat{U}_2^{}\widehat{U}_1^{}(\widehat{U}_2^{})^{}=\widehat{1}`$. Since the eigenvalues are independent of the unitary transformation, the condition (C.2) is satisfied with $`E(\widehat{\rho }^{})=E(\widehat{\rho })`$.
To consider the final condition (C.3), we introduce the LGM+CC(+PSS) which maps the density matrix $`\widehat{\rho }`$ into $`\widehat{\rho }^{}`$, defined by
$$\widehat{\rho }^{}=\underset{i}{}\widehat{V}_i\widehat{\rho }\widehat{V}_i^{}$$
(40)
where the classically correlated operator $`\widehat{V}_i=\widehat{A}_i\widehat{B}_i`$ satisfies the complete relation as $`_i(\widehat{V}_i)^{}\widehat{V}_i=\widehat{1}`$. Let $`p_i\widehat{\rho }_i=\widehat{V}_i\widehat{\rho }\widehat{V}_i^{}`$ with $`p_i=\text{Tr}\widehat{V}_i\widehat{\rho }\widehat{V}_i^{}`$ and $`\widehat{\sigma }_i=\widehat{\rho }_i^{T_2}`$. Since $`\widehat{V}_i`$ is a local operator, $`\widehat{\sigma }_i`$ is represented in terms of $`\widehat{\sigma }`$, namely,
$$p_i\widehat{\sigma }_i=\widehat{V}_i^{}\widehat{\sigma }\widehat{V}_{i}^{}{}_{}{}^{}$$
(41)
where $`\widehat{V}_i^{}=\widehat{A}_i\widehat{B}_i^{}`$ is an LGM+CC operator with a completeness relation $`_i\widehat{V}_{i}^{}{}_{}{}^{}\widehat{V}_i^{}=\widehat{1}`$.
Suppose $`\widehat{d}^i`$ are diagonal matrices of $`\widehat{\sigma }_i`$ with some unitary operator $`\widehat{U}_i`$ and $`\widehat{d}`$ of $`\widehat{\sigma }`$ with $`\widehat{U}`$. The diagonal matrix $`\widehat{d}^i`$ can be written as
$`p_i\widehat{d}^i`$ $`=`$ $`\left(\widehat{U}_i\widehat{V}_i^{}\widehat{U}^{}\right)\left(\widehat{U}\widehat{\sigma }\widehat{U}^{}\right)\left(\widehat{U}\widehat{V}_{i}^{}{}_{}{}^{}\widehat{U}_i^{}\right)`$ (42)
$`=`$ $`\widehat{W}_i\widehat{d}\widehat{W}_i^{}`$ (43)
where $`\widehat{W}_i=\widehat{U}_i\widehat{V}_i^{}\widehat{U}_0^{}`$ is also a LGM+CC operator. For the given $`\widehat{d}^i`$, two projectors $`\widehat{P}_{}^i`$ and $`\widehat{P}_+^i`$ are defined to project the density matrix space onto semi-positive and negative eigenvalue subspaces of $`\widehat{d}^i`$. The measure of entanglement $`E(\widehat{\rho }_i)`$ on the subensemble $`\widehat{\rho }_i`$ is given by
$`p_iE(\widehat{\rho }_i)`$ $`=`$ $`2p_i\text{Tr}\left[\widehat{d}_{}^i(\widehat{\rho }_i)\right]`$ (44)
$`=`$ $`2\text{Tr}\left(\widehat{P}_{}^i\widehat{W}_i\widehat{d}\widehat{W}_i^{}\widehat{P}_{}^i\right).`$ (45)
We represent $`\widehat{d}=_j\lambda _j|\psi _j\psi _j|=_\alpha \lambda _\alpha ^+|\psi _\alpha \psi _\alpha |+_\beta \lambda _\beta ^{}|\psi _\beta \psi _\beta |`$ where the sum is decomposed into two sums of semi-positive and negative eigenvalues. The whole-ensemble average of the measures of entanglement is given by
$`{\displaystyle \underset{i}{}}p_iE(\widehat{\rho }_i)`$ $`=`$ $`2{\displaystyle \underset{ij}{}}(\lambda _j)\psi _j|\widehat{W}_i^{}\widehat{P}_{}^i\widehat{P}_{}^i\widehat{W}_i|\psi _j`$ (46)
$`=`$ $`2{\displaystyle \underset{i\alpha }{}}(\lambda _\alpha ^+)\psi _\alpha |\widehat{W}_i^{}\widehat{P}_{}^i\widehat{P}_{}^i\widehat{W}_i|\psi _\alpha `$ (48)
$`+2{\displaystyle \underset{i\beta }{}}(\lambda _\beta ^{})\psi _\beta |\widehat{W}_i^{}\widehat{P}_{}^i\widehat{P}_{}^i\widehat{W}_i|\psi _\beta `$
where we separate the sum into the sums of semi-positive eigenvalues $`\lambda _\alpha ^+`$ and negative eigenvalues $`\lambda _\beta ^{}`$ of $`\widehat{d}`$. The inequality, $`\psi _j|\widehat{W}_i^{}\widehat{P}_{}^i\widehat{P}_{}^i\widehat{W}_i|\psi _j0`$, and the sign of eigenvalues make the first term negative and the second term positive in Eq. (46). Eq. (46) results in the following inequality
$$\underset{i}{}p_iE(\widehat{\rho }_i)2\underset{i\beta }{}(\lambda _\beta ^{})\psi _\beta |\widehat{W}_i^{}\widehat{P}_{}^i\widehat{P}_{}^i\widehat{W}_i|\psi _\beta .$$
(50)
Since $`0_i\psi |\widehat{W}_i^{}\widehat{P}_{}^i\widehat{P}_{}^i\widehat{W}_i|\psi 1`$ for arbitrary wave function $`|\psi `$ , we finally arrive at the inequality
$`{\displaystyle \underset{i}{}}p_iE(\widehat{\rho }_i)`$ $``$ $`2{\displaystyle \underset{\beta }{}}(\lambda _\beta ^{}){\displaystyle \underset{i}{}}\psi _\beta |\widehat{W}_i^{}\widehat{P}_{}^i\widehat{P}_{}^i\widehat{W}_i|\psi _\beta `$ (51)
$``$ $`2{\displaystyle \underset{\beta }{}}(\lambda _\beta ^{})=E(\widehat{\rho }),`$ (52)
which satisfies the condition (C.3) for an arbitrary set of LGM+CC operators.
As an example how to calculate the measure of entanglement, consider the Werner state $`\widehat{w}`$ in Eq. (8). When we select the following representations,
$$\begin{array}{cc}\widehat{1}\widehat{1}=\left(\begin{array}{cccc}1& 0& 0& 0\\ 0& 1& 0& 0\\ 0& 0& 1& 0\\ 0& 0& 0& 1\end{array}\right),\hfill & \widehat{\sigma }_x\widehat{\sigma }_x=\left(\begin{array}{cccc}0& 0& 0& 1\\ 0& 0& 1& 0\\ 0& 1& 0& 0\\ 1& 0& 0& 0\end{array}\right),\hfill \\ \widehat{\sigma }_y\widehat{\sigma }_y=\left(\begin{array}{cccc}0& 0& 0& 1\\ 0& 0& 1& 0\\ 0& 1& 0& 0\\ 1& 0& 0& 0\end{array}\right),\hfill & \widehat{\sigma }_z\widehat{\sigma }_z=\left(\begin{array}{cccc}1& 0& 0& 0\\ 0& 1& 0& 0\\ 0& 0& 1& 0\\ 0& 0& 0& 1\end{array}\right),\hfill \end{array}$$
(53)
the Werner state is written as
$$\widehat{w}=\left(\begin{array}{cccc}\frac{1f}{4}& 0& 0& 0\\ 0& \frac{1+f}{4}& \frac{f}{2}& 0\\ 0& \frac{f}{2}& \frac{1+f}{4}& 0\\ 0& 0& 0& \frac{1f}{4}\end{array}\right)$$
(54)
where $`f=(2\mathrm{\Phi }+1)/3`$, and the set of eigenvalues is $`\{(1f)/4,(1f)/4,(1f)/4,(1+3f)/4\}`$. The positivity of density matrix requires that $`1/3f1`$ and thus $`1\mathrm{\Phi }1`$. The partial transposition is now given by
$$\widehat{\sigma }=\widehat{w}^{T_2}=\left(\begin{array}{cccc}\frac{1f}{4}& 0& 0& \frac{f}{2}\\ 0& \frac{1+f}{4}& 0& 0\\ 0& 0& \frac{1+f}{4}& 0\\ \frac{f}{2}& 0& 0& \frac{1f}{4}\end{array}\right)$$
(55)
which has its eigenvalues $`\{(1+f)/4,(1+f)/4,(1+f)/4,(13f)/4\}`$. It is clear that three eigenvalues are positive since $`(1+f)/40`$ under the constraint of $`1/3f1`$. The other eigenvalue can be negative only if $`3f>1`$ or equivalently $`\mathrm{\Phi }>0`$. For $`3f>1`$, $`E(\widehat{\rho })2_\beta (\lambda _\beta ^{})=(3f1)/2`$ while $`E(\widehat{\rho })=0`$ for $`3f1`$. The Werner state with $`f=1`$ becomes a maximally entangled singlet state and then $`E(\widehat{w})=1`$.
|
warning/0003/quant-ph0003006.html
|
ar5iv
|
text
|
# Space Searches with a Quantum Robot
## 1. Introduction
Quantum computers have been the subject of much study, mainly because of computations that can be done more efficiently than on classical computers. Well known examples include Shor’s and Grover’s algorithms, . Quantum robots have also been recently described as mobile systems, with an on board quantum computer and ancillary systems, that move in and interact with environments of quantum systems . Dynamics of quantum robots are described as tasks consisting of alternating computation and action phases.
Quantum robots can be used to carry out many types of tasks. These range from simple ones such as searching a region of space to determine the unknown location of a system to complex tasks such as carrying out physical experiments. The fact that the spatial searches are similar to the data base searches which are efficiently implemented using Grover’s Algorithm suggests that similar results might hold for use of Grover’s Algorithm to process results of a quantum search of a spatial region. This is an example of the possible applicability of Grover’s algorithm to various physical measurements . If this can be done for the search task, then one would have an example of a task that can be carried out more efficiently by a quantum robot than by any classical robot.
This possibility is analyzed here for a search of a 2 dimensional space region by a quantum robot to locate a system. It is seen that for reversible searches with time independent unitary dynamics, there are two problems preventing the efficient use of Grover’s Algorithm. One is that it appears impossible to remove entanglements generated during the search process. The other problem is that the action in the search task, which is the equivalent of the ”oracle function”, which is assumed in Grover’s Algorithm to be evaluated on any argument in one step, takes many steps to evaluate. The result is that even if the entanglement problem is ignored, quantum searches of 2 dimensional space regions are no more efficient than classical searches. However quantum searches of higher dimensional space regions are more efficient than classical searches.
The plan of this paper is to give, in the next section, a brief description of quantum robots and a summary of how they are different from quantum computers. This is followed, in Section 2, by a description of the dynamics of tasks as sequences of alternating computation and action phases. An explicit description of the dynamics is given as a Feynman sum over computation-action phase paths. An example of a quantum robot searching a space area to determine the unknown location of a systems is then described, Section 3. The next section is concerned with the use of Grover’s algorithm to process the search results. A very brief summary of Grover’s Algorithm, in Subsection 4.1, is followed by a description of the problems encountered, Subsection 4.2. The paper finishes with a discussion of why quantum robots are interesting independent of these results.
## 2. Quantum Robots
### 2.1. Comparison with Quantum Computers
Quantum robots are similar to quantum computers in that an important component is an on board quantum computer. Other systems such as a memory system $`m`$, an output system $`o`$, and a control qubit $`c`$ are also present in quantum robots . A relatively minor difference is that quantum robots are mobile whereas quantum computers are stationary relative to the environment. For quantum Turing machine models of quantum computers the head moves but the quantum registers are stationary. For networks of quantum gates the qubit systems move but the gates are stationary. This is shown in physical models of interacting qubits. Examples include ion trap models or nuclear magnetic resonance models . In these cases the ion traps and the liquid of active molecules are stationary.
As is the case for quantum computers the effects of the environment on the component systems of the quantum robot need to be minimized. Methods to achieve this include the possible use of shielding or quantum error correction codes . However other than this the dynamical properties of the environment are completely arbitrary.
This is quite different than the case for quantum computers. To see this assume for quantum Turing machines that the quantum registers are the environment of the head. Similarly for gate networks the moving qubits may be considered as the environment of the network of quantum gates. Here the dynamics of these systems is quite restricted in that the states of the registers or moving qubits can change only during interaction with the head or the gate systems. Also the types of changes that can occur are limited to those appropriate for the specific computation being carried out.
This shows the main difference between quantum robots and quantum computers, namely, that for quantum computers the states of the qubits must not change spontaneously in the absence of interactions related to the computation. No such dynamical restrictions apply to the environment of quantum robots. The states of environmental systems may change spontaneously whether the quantum robot is or is not interacting with them.
Another aspect relates to the requirement that the quantum robot cannot be a multistate system with one or a very few number of degrees of freedom, but must include a quantum computer. One reason is the need for computations as part of the implementation of any task (see below). Another is that the quantum robot may need to be able to respond to a large number of different environmental states. Also a large repertoire of different (task dependent) responses to the same environmental state must be available. If the total number $`N`$ of needed responses is large then the only physically reasonable approach is to make the number of degrees of freedom of the quantum robot proportional to $`\mathrm{log}N`$. This is satisfied by including a quantum computer on board.
### 2.2. Task Dynamics
The dynamics of quantum robots are described as tasks consisting of alternating computation and action phases . The purpose of each computation phase is to determine what is to be done in the next action phase. The computation may depend on the states of $`o,m`$, and the local environment as input. The goal is to put $`o`$ in one of a set $``$ of basis states each of which specifies an action. During the computation the quantum robot does not move. Interactions with the environment, if any, are limited to local entanglement interactions of the type occurring in the measurement process (premeasurements in the sense of Peres ).
The purpose of an action phase is to carry out the action determined by the previous computation phase. The action is determined by the state of $`o`$ and may include local premeasurements of the environment state. Activities during this phase include motion of the quantum robot and local changes in the state of the environment. The action is independent of the state of the on board quantum computer and the $`m`$ system. If $`o`$ is in a state in $``$, then the state does not change during the action.
The purpose of the control qubit $`c`$ is to regulate which type of phase is active. The computation \[action\] phase is active only if $`c`$ is in state $`|1,[|0]`$. Thus the last step of the computation \[action\] phase is the change $`|1|0,[|0|1]`$.
The overall system evolution is described here using a discrete space time lattice. In this case a unitary elementary step operator $`\mathrm{\Gamma }`$ gives the overall system evolution during an elementary time step $`\mathrm{\Delta }`$. Since $`\mathrm{\Gamma }`$ has, in general, nonzero matrix elements between environmental degrees of freedom and quantum robot degrees of freedom, it describes the evolution of the environment and the quantum robot as well as interactions between the environment and the quantum robot.
It is useful to decompose $`\mathrm{\Gamma }`$ into two terms based on the states of the control qubit. If $`P_0^c`$ and $`P_1^c`$ are projection operators on the respective control qubit states $`|0`$ and $`|1`$, then
(2.1)
$$\mathrm{\Gamma }=\mathrm{\Gamma }(P_0^c+P_1^c)=\mathrm{\Gamma }_a+\mathrm{\Gamma }_c.$$
Here $`\mathrm{\Gamma }_a`$ and $`\mathrm{\Gamma }_c`$ are step operators for the action and computation phases. Interactions among environmental degrees of freedom as well as degrees of freedom of the quantum robot other than those taking part in the task dynamics, if any, are also included in both operators.
Some of the conditions described for the computation and action phases are reflected in properties that the operators $`\mathrm{\Gamma }_a`$ and $`\mathrm{\Gamma }_c`$ must satisfy. In particular, if $`P_x^{QR},P_d^o`$ are projection operators for finding the quantum robot at each lattice position $`x`$ ($`x=x_1,x_2,\mathrm{},x_d`$ in d-dimensional space) and the output system in any state $`|d`$ in $``$, then one has
$`\mathrm{\Gamma }_cP_x^{QR}`$ $`=`$ $`P_x^{QR}\mathrm{\Gamma }_c`$
(2.2) $`\mathrm{\Gamma }_aP_d^o`$ $`=`$ $`P_d^o\mathrm{\Gamma }_a`$
These commutation relations express the requirements that the position of the quantum robot does not change during the computation phase and, except for possible entanglements with environmental states, the state of the output system is not changed during the action phase. These entanglements would occur if $`o`$ was in a linear superposition of $``$ states each of which resulted in different environment states during the action. Another property of $`\mathrm{\Gamma }_a`$ is that it is the identity operator on the space of states for the on board quantum computer and memory system degrees of freedom. Note also that $`\mathrm{\Gamma }_a`$ and $`\mathrm{\Gamma }_c`$ do not commute.
If $`\mathrm{\Psi }_0`$ is the overall system state at time $`0`$ then the state at time $`n\mathrm{\Delta }`$ is given by $`\mathrm{\Psi }_n=(\mathrm{\Gamma }_a+\mathrm{\Gamma }_c)^n\mathrm{\Psi }_0`$. The amplitude for finding the quantum robot and environment in a state $`|w^{},j`$ is given by
(2.3)
$$\mathrm{\Psi }_n(w^{},j)=\underset{w,i}{}w^{},j|(\mathrm{\Gamma }_a+\mathrm{\Gamma }_c)^n|w,i\mathrm{\Psi }_0(w,i).$$
Here $`|w,|w^{}`$ denote the states of all environmental and quantum robot systems except the control qubit in some suitable basis, and $`i,j=0,1`$ refer to the states of $`c`$.
All the information about the dynamics of the system is given in the matrix elements $`w^{},j|(\mathrm{\Gamma }_a+\mathrm{\Gamma }_c)^n|w,i`$. For each $`w,w^{},n,i,j`$ the matrix element can be expanded in a Feynman sum over phase paths . One first expands $`(\mathrm{\Gamma }_a+\mathrm{\Gamma }_c)^n`$ as a sum of products of $`\mathrm{\Gamma }_a`$ and $`\mathrm{\Gamma }_c`$:
(2.4) $`(\mathrm{\Gamma }_a+\mathrm{\Gamma }_c)^n`$ $`=`$ $`{\displaystyle \underset{v_1=a,c}{}}{\displaystyle \underset{t=1}{\overset{n}{}}}{\displaystyle \underset{h_1,h_2,\mathrm{},h_t=1}{\overset{\delta ({\scriptscriptstyle },n)}{}}}(P_0^c+P_1^c)(\mathrm{\Gamma }_{v_t})^{h_t}(\mathrm{\Gamma }_{v_{t1}})^{h_{t1}},`$
$`\mathrm{},(\mathrm{\Gamma }_{v_2})^{h_2}(\mathrm{\Gamma }_{v_1})^{h_1}.`$
In this expansion the number of phases is given by the value of $`t`$ which ranges from $`t=1`$ corresponding to one phase with $`n`$ steps to $`t=n`$ corresponding to n alternating phases each of $`1`$ step. The duration of the $`\mathrm{}th`$ phase is given by the value of $`h_{\mathrm{}}`$ for $`\mathrm{}=1,2,\mathrm{},t`$. The requirement that the total number of steps equals $`n`$, or $`h_1+h_2+,\mathrm{},+h_t=n`$, is indicated by the upper limit $`\delta (,n)`$ on the $`h`$ sum. The alternation of phases is shown by $`v`$ where $`v_{m+1}=a`$ (or $`c`$) if $`v_m=c`$ (or $`a`$). The factor $`P_0^c+P_1^c`$ expresses the fact that the $`t`$th phase may not be completed.
Expansion in a complete set of states between each of the phase operators $`(\mathrm{\Gamma }_v_{\mathrm{}})^h_{\mathrm{}}`$ gives the desired path sum:
(2.5) $`w^{},j|(\mathrm{\Gamma }_a+\mathrm{\Gamma }_c)^n|w,i`$ $`=`$ $`{\displaystyle \underset{t=1}{\overset{n}{}}}{\displaystyle \underset{p_2,\mathrm{},p_t}{}}{\displaystyle \underset{h_1,h_2,\mathrm{},h_t=1}{\overset{\delta ({\scriptscriptstyle },n)}{}}}w^{},j|(\mathrm{\Gamma }_{v_t})^{h_t}|p_t`$
$`\mathrm{},p_3|(\mathrm{\Gamma }_{v(2)})^{h_2}|p_2p_2|(\mathrm{\Gamma }_i)^{h_1}|w,i`$
Here the sum is over all paths $`p`$ of states of length $`t+1`$ with beginning and endpoints given by the states $`|w`$ and $`|w^{}`$. That is $`|p_1=|w,|p_t=|w^{}`$. The states of the control qubit have been suppressed as they correspond to the values of $`v`$. Note that $`v(1)=i`$.
Each term in this large sum gives the amplitude for finding $`t`$ alternating phases in the first $`n`$ steps where the $`\mathrm{}th`$ phase begins with all systems (except for c) in state $`|p_{\mathrm{}}`$ and ends after $`h_{\mathrm{}}`$ steps in state $`|w_{\mathrm{}+1}`$. The sums express the dispersion in the duration or number of steps in each phase ($`h`$ sums), in the number of phases ($`t`$ sum), and in the initial and terminal states for each phase ($`p`$ sums).
## 3. An Example of Quantum Searching
Quantum robots are well suited for carrying out search tasks. As a simple example consider a search task where a quantum robot searches a large square area $`R`$ of $`N\times N`$ sites to locate a system $`s`$. To keep things simple $`s`$ is assumed to be motionless and located at just one unknown site. The goal of the search is to determine the location of $`s`$ in $`R`$.
The quantum robot consists of an on board quantum computer, memory and output systems, and a control qubit. The on board computer is assumed here to be a quantum Turing machine consisting of a head moving on a cyclic lattice of $`O(\mathrm{log}N)`$ qubits. $`O()`$ denotes of the order of. The memory system also is a cyclic lattice which is taken here to have about the same number of qubits and to lie adjacent to the computation lattice. A schematic representation of the quantum robot located at a corner (the origin) of $`R`$ is shown in the figure.
One method of carrying out the search is to let the coordinates $`X,Y`$ with $`0X,YN1`$ of each point of $`R`$ define a search path. If the memory is initially in state $`|X,Y_m`$ the quantum robot, starting from the location $`0,0`$ moves $`X`$ sites in the $`x`$ direction, then $`Y`$ sites in the $`y`$ direction and looks for $`s`$ at its location. After recording the presence or absence of $`s`$ at the site and further processing, if any, the quantum robot returns along the path to the origin.
A more detailed description starts with the qubits in the memory, $`m`$, and computation lattice, $``$, in the state $`|X,Y_m|0_{}`$ the output system $`o`$ in state $`|dn_o`$ and a computation phase active ($`c`$ in state $`|1_c`$). After copying the m state onto $``$ to give the state $`|X,Y_m|X,Y_{}`$, the computation phase checks to see if $`X=0`$ or $`X>0`$. If $`X>0`$ the computation phase continues by subtracting $`1`$ from $`|X,Y_{}`$ to give $`|X1,Y_{}`$. It ends by changing the $`o`$ state to $`|+x_o`$ and the $`c`$ state to $`|0_c`$.
The action phase consists of one step (one iteration of $`\mathrm{\Gamma }_a`$) in which the quantum robot moves one lattice site in the $`+x`$ direction and the $`c`$ state is converted back to $`|1_c`$. The process is repeated until the state with $`X=0`$ is reached on $``$. Then the above process is repeated for $`Y`$ (the $`o`$ state now becomes $`|+y_o`$ to denote one step motion in the $`+y`$ direction) until $`Y=0`$ is reached in the state of $``$.
At this point the presence or absence of $`s`$ at the location $`X,Y`$ of the quantum robot is recorded during a computation phase and, after further processing, if any, the quantum robot returns along the same path. This is done by interleaving motion of the quantum robot in the $`y`$ and $`x`$ directions, with corresponding $`o`$ states $`|y_o,|x_o`$, with adding $`1`$ to the $`y`$, and then $`x`$ components of the $``$ state with checking if the values $`Y`$ and then $`X`$ are reached. This is done by stepwise comparison with the state of $`m`$ which remains unchanged.
When the state of $``$ is the same as the state of $`m`$ the quantum robot has returned to the starting point at the origin of $`R`$. A computation phase changes the $`o`$ state to $`|dn_o`$ and transfers motion to some ballast system. As has been noted this is necessary to preserve reversibility and the corresponding unitarity of the dynamics .
Examples of ballast motion consist of repetitions of adding $`1`$ to a large lattice of $`M`$ qubits or emitting a particle which moves away from $`R`$. In the first case with a finite number $`2^M`$ of ballast states, the quantum robot remains in the final state of the search degrees of freedom for a finite time only before the search process is undone. This does not occur for the second case with an infinite number of ballast states.
## 4. Grover’s Algorithm and the Quantum Search
Before applying Grover’s Algorithm to process the results of the search,it is useful to understand what it does and how it works. A very brief summary, that follows Grover and Chen et al , is given next.
### 4.1. Grover’s Algorithm
Suppose one has a data base $`B`$ of $`N`$ elements and a function $`f`$ that takes the value $`0`$ on all elements except one, $`\omega `$, on which $`f`$ has value $`1`$. It is assumed that $`\omega `$ is completely unknown and that a procedure is available for obtaining the value of $`f`$ on any element of the data base in 1 step. Let each $`x`$ in $`B`$ correspond to a unique length $`n`$ binary string and $`|x_B`$ be the corresponding n-qubit state.
Let the initial state for the search be given by
(4.1)
$$\varphi =\frac{1}{\sqrt{N}}\underset{xϵB}{}|x$$
where the sum is over all $`N`$ elements $`x`$ in $`B`$. This corresponds to a coherent sum over all product $`|0,|1`$ states of $`n`$ qubits in a quantum computer if $`N=2^n`$. This state is easily constructed from the constant $`0`$ state $`|\underset{¯}{0}=_{j=1}^n|0_j`$ by applying the operator $`(1/\sqrt{2})(\sigma _z+\sigma _x)`$ to each qubit. Here $`\sigma _x,\sigma _z`$ are the Pauli matrices. This is referred to as the Walsh-Hadamard transformation $`W`$. Thus $`\varphi =W|\underset{¯}{0}`$.
Define the unitary operator $`Q`$ by $`Q=I_\varphi I_\omega `$ where $`I_\varphi =12P_\varphi `$ and $`I_\omega =12P_\omega `$. Both $`P_\varphi `$ and $`P_\omega `$ are projection operators on the states $`\varphi `$ and $`|\omega `$. Let $`|\alpha =(1/\sqrt{N1})_{x\omega }|x`$ be the coherent sum over all states $`|x`$ with $`x`$ in $`B`$ and different from $`\omega `$. Since $`|\alpha `$ and $`|\omega `$ are orthonormal they form a binary basis for a 2 dimensional Hilbert space.
One can expand $`\varphi `$ in this basis:
(4.2)
$$\varphi =\sqrt{\frac{N1}{N}}|\alpha +\frac{1}{\sqrt{N}}|\omega .$$
In the same basis $`Q`$ has the representation
(4.3)
$$Q=\left(\begin{array}{cc}\mathrm{cos}\theta & \mathrm{sin}\theta \\ \mathrm{sin}\theta & \mathrm{cos}\theta \end{array}\right)$$
where $`\mathrm{cos}\theta =12/N`$ and $`\mathrm{sin}\theta =2\sqrt{N1}/N`$.
This shows that $`Q`$ acting on $`\varphi `$ corresponds to a rotation by $`\theta `$, and $`m`$ iterations of $`Q`$ correspond to a rotation by $`m\theta `$. So, carrying out $`m`$ iterations of $`Q`$ on $`\varphi `$ where $`m\theta \pi /2`$, rotates $`\varphi `$ from a state that is almost orthogonal to $`|\omega `$ to a state that is almost parallel to $`|\omega `$. Measurement of this final state gives with high probability, the value of $`\omega `$.
Grover’s algorithm derives its efficiency from the fact that this rotation is achieved with $`m\sqrt{N}`$ whereas classically $`N`$ steps are needed to find $`\omega `$ with high probability. The iteration of $`Q`$ must be stopped at the right value of $`m`$ because additional iterations will continue to rotate $`\varphi `$.
Efficient implementation of this algorithm on a quantum computer, corresponds to iteration of $`Q`$ on each component of $`\varphi `$. This requires that it is possible to determine, in a small number of steps, if $`x=\omega `$ or $`x\omega `$. This is often described in terms of an unknown or ”oracle” function $`f`$ on $`B`$, where $`f(x)=0[=1]`$ if $`x\omega [x=\omega ]`$, that can be evaluated in one step on any $`x`$ in $`B`$. $`I_\varphi `$ is implementable in $`O(n)`$ steps as $`\varphi =W|\underset{¯}{0}`$ and $`W`$ is the product of $`n`$ single qubit operators.
Grover first introduced the algorithm for searching an unstructured data base of $`N=2^n`$ elements for a single element (f has value 1 on just one element). Since then the algorithm has been much studied under various generalizations. These include searches for several elements (f has value 1 on several elements) , searches in which $`N`$ is arbitrary , and searches in which the initial amplitude distribution of the component states is arbitrary . It has also been shown that the algorithm is optimally efficient . Further development is described in other work . However, as has been recently emphasized , all these searches depend on the fact that the evolving state is and remains a coherent superposition of components corresponding to elements of the data base.
### 4.2. Problems with the Use of Grover’s Algorithm
The description in Section 3 of the quantum search was for the initial memory state $`|X,Y_m`$. However if the memory system lattice is in the initial state
(4.4)
$$\psi _m=(1/N)\underset{X,Y=0}{\overset{N1}{}}|X,Y_m,$$
then the description also applies to each component state $`|X,Y`$.
As was the case for the initial state $`\varphi `$, the state $`\psi _m`$ can be efficiently prepared from the state $`|0_m`$ using the Walsh-Hadamard transformation. $`(1/\sqrt{2})(\sigma _z+\sigma _x)`$ on each qubit of $`m`$. For this initial memory state all $`N^2`$ searches are carried out coherently. Since the path lengths range from $`0`$ to $`2N`$, the quantum robot can search all sites of $`R`$ and return to the origin in $`O(N\mathrm{log}N)`$ steps. Since this is less than the number of steps, $`O(N^2\mathrm{log}N)`$, required by a classical robot, the question arises if Grover’s algorithm can be used to process the final memory state to determine the location of $`s`$. If this is possible, the overall search and processing should require $`O(N\mathrm{log}N)`$ steps which is less than that required by a classical robot.
It is worth a digression at this point to see that Grover’s Algorithm is not applicable to the usual method of recording the presence or absence of $`s`$ at a site. To see this assume that $`s`$ is at site $`X_0,Y_0`$ and that an extra qubit, $`r`$ of the memory is set aside to record the presence or absence of $`s`$. If $`r`$ is initially in state $`|0_r`$ and is changed to state $`|1_r`$ only in the presence of $`s`$, then the initial memory state $`\varphi _I=(1/N)_{X,Y=0}^{N1}|X,Y_m|0_r`$ is changed to the final memory state
(4.5)
$$\varphi _f=(1/N)(\underset{X,YX_0Y_0}{}|X,Y_m|0_r+|X_0,Y_0_m|1_r).$$
after the quantum robot has returned to the origin of $`R`$.
The idea then would be to use Grover’s algorithm by carrying out $`N`$ iterations of a unitary operator $`U`$ to amplify the component state $`|X_0,Y_0_m|1_r`$ at the expense of the other components. Following Grover and others , define $`U`$ by $`U=I_{\varphi _i}I_{|1_r}`$ where $`I_{\varphi _i}=12P_{\varphi _i}`$ and $`I_{|1_r}=12P_{|1_r}`$. Here $`P_{\varphi _i}`$ and $`P_{|1_r}`$ are projection operators on the memory state $`\varphi _i`$ and the record state $`|1_r`$. $`P_{|1_r}`$ is the identity on other memory degrees of freedom.
It is clear that for this case iterations of $`U`$ can be carried out efficiently. However, the problem is that the initial state $`\varphi _i`$ contains a component, $`|X_0,Y_0_m|0_r`$, not present in the final state $`\varphi _f`$, Eq. 4.5. Also the state $`\varphi _i`$ does not contain the state $`|X_0,Y_0_m|1_r`$. In this case $`U`$ does not have a two dimensional representation in the basis pair
(4.6)
$$\frac{1}{\sqrt{N^21}}\underset{X,YX_0Y_0}{}|X,Y_m|0_r;|X_0Y_0_m|1_r$$
obtained from Eq. 4.5. As a result $`U`$ cannot be represented as a rotation that, under iteration, rotates the desired component to be almost parallel to the initial state.
This problem can be avoided by changing the method of recording the presence or absence of $`s`$ and not using an extra qubit $`r`$. Here, following Grover , the sign of the component corresponding to the location of $`s`$ is changed. In this case the initial memory state, Eq. 4.4, becomes the final memory state
(4.7)
$$\psi _f=(1/N)(\underset{X,YX_0Y_0}{}|X,Y_m|X_0,Y_0_m).$$
after return of the quantum robot. In this case $`U`$ is defined as $`U=I_{\psi _m}I_{X_0Y_0}`$ where $`I_{\psi _m}=12P_m`$ and $`I_{X_0Y_0}=12P_{X_0Y_0}`$. Here $`P_m`$ and $`P_{X_0Y_0}`$ are projection operators on the memory states $`\psi _m`$ and $`|X_0Y_0_m`$.
The problem here is that the only way to determine the action of $`I_{X_0Y_0}`$ is by repeating the search part of the process. This is not efficient as it requires $`O(N\mathrm{log}N)`$ steps. In the language of much of the work on Grover’s algorithm this corresponds to the fact that it requires $`O(N\mathrm{log}N)`$ steps to determine the value of the oracle function instead of just one step as is usually assumed. In this case the advantage of quantum over classical searching is lost for 2 dimensional regions as use of Grover’s Algorithm would require $`O(N)`$ searches each requiring $`O(N\mathrm{log}N)`$ steps.
This suggests that a method be considered in which the Grover iterations are done prior to return when the quantum robot is at the path endpoint. At this point the component memory states are entangled with the quantum robot position states as the overall state has the form $`(1/N)_{X,Y}|X,Y_m|X,Y_{QR}`$ where $`|X,Y_{QR}`$ is the quantum robot position state for the site $`X,Y`$.
In this case $`I_{X_0Y_0}`$ can be efficiently carried out on each initial component memory state $`|X,Y`$ by a local observation to see if $`s`$ is or is not at the site $`X,Y`$. Also the action of $`U=I_{\psi _m}I_{X_0Y_0}`$ on each component memory state is given by
(4.8)
$$U|X,Y_m=I_m|X,Y_m=\frac{2}{\sqrt{N}}\psi _m|X,Y_m$$
if $`|X,Y_m|X_0,Y_0_m`$ and
(4.9)
$$U|X_0,Y_0_m=I_m|X_0,Y_0_m=\frac{2}{\sqrt{N}}\psi _m+|X_0,Y_0_m$$
if $`|X,Y_m=|X_0,Y_0_m`$. Here, as before, $`\psi _m=(1/N)_{X,Y=0}^{N1}|X,Y_m`$.
Here the problem is that there is no efficient way to carry out more than one iteration of $`U`$. As noted above the first iteration can be done efficiently. However additional iterations require that the action of $`I_{X_0Y_0}`$ on memory component states $`|X^{},Y^{}`$ be evaluated for arbitrary values of $`X^{},Y^{}`$ while the quantum robot remains at site $`X,Y`$. This cannot be done efficiently as the quantum robot has no way of knowing whether $`s`$ is or is not at these different locations. To know this the quantum robot must go to the site $`X^{},Y^{}`$ to see if $`s`$ is there. This is inefficient as such a trip requires $`O(N\mathrm{log}N)`$ steps.(Actions are efficient if they require $`O(\mathrm{log}N)`$ steps or less. Low powers of $`\mathrm{log}N`$ are also acceptable.)
One sees from this that implementation of Grover’s Algorithm using either of these methods requires $`O(N)`$ iterations of $`U`$ (as $`R`$ has $`N^2`$ sites) where each iteration requires $`O(N\mathrm{log}N)`$ steps. The resulting number of steps required, $`O(N^2\mathrm{log}N)`$, is the same as that needed by a classical robot. So quantum searches of 2 dimensional space regions combined with Grover’s Algorithm are no more efficient than classical searches.
It is of interest to note that quantum searches of higher dimensional space regions combined with Grover’s Algorithm are more efficient than classical searches. To see this assume a search of a d dimensional cube of $`N^d`$ sites with the memory in the initial state
(4.10)
$$\mathrm{\Psi }=\frac{1}{N^{d/2}}\underset{X_1,\mathrm{}X_d=0}{\overset{N1}{}}|X_1,X_2,\mathrm{}X_d_m$$
Carrying out Grover’s Algorithm requires $`O(N^{d/2})`$ iterations of $`U`$ where, as before, each iteration of $`U`$ requires $`O(N\mathrm{log}N)`$ steps. This follows from the fact that the number of dimensions appears as a multiplicative factor for the number of steps. Also $`O(dN\mathrm{log}N)=O(N\mathrm{log}N)`$. So the overall process requires $`O(N^{(d/2)+1}\mathrm{log}N)`$ steps. For $`d>2`$ this is more efficient than a classical search requiring $`O(N^d\mathrm{log}N)`$ steps.
The discussion so far has ignored the entanglement problems. These problems, which apply to all the above cases, result from the fact that the task evolution, starting from the initial unentangled product state $`\psi _m|0_{}|0,0_{QR}\mathrm{}`$, generates entanglements between the position states $`|X^{},Y^{}_{QR}`$ of the quantum robot and the components $`|X,Y_m`$ of the memory state $`\psi _m`$. In order for Grover’s Algorithm to work it is necessary to remove this entanglement at the end of each search cycle or iteration of $`U`$ so that the final memory state is $`\psi _f`$<sup>1</sup><sup>1</sup>1The entanglement referred to here is different from that of the memory state qubits. The latter is generated during iteration of the Grover operator and is necessary for successful operation of Grover’s Algorithm on multiqubit states ..
This entanglement occurs because the unitary dynamics is reversible and the number of steps needed to complete the search task is different for different component states of $`\psi _m`$. Here the number ranges from $`O(1)`$ for the path $`|0,0_m`$ to $`O(2N)`$ for the path $`|N1,N1_m`$. This means that the various components of the quantum robot complete an iteration of $`U`$ at different times. This is independent of whether the Algorithm is completed after or prior to return.
Because of the reversibility each component cannot simply stop and wait until the longest search component is complete. It must instead embark on motion of irrelevant or ballistic degrees of freedom. This means that the memory components $`|X,Y_m`$ exchange entanglement with the quantum robot position states for entanglement with states of ballistic degrees of freedom. Since use of Grover’s Algorithm requires the removal of this entanglement , the question arises whether it is possible to insert delays into each of the memory components that are computed, for example, after the quantum robot returns to the origin of $`R`$ at the end of each cycle. If this works then there would be some time or step number at which the entanglement is removed and the original product structure of the initial state recovered, with $`\psi _f`$ replacing $`\psi _m`$.
This use of delays to remove the entanglements reversibly requires that no memory of the magnitude of the delay be left in the delay degrees of freedom. Otherwise one ends up with entanglement with the delay degrees of freedom. Also determining the magnitude of each delay is not trivial as it depends not only on the lengths of each of the paths but on the number of steps in the computation phases used to determine motion along the paths. This includes the dependence of the number of steps required to subtract $`1`$ from a number $`M`$ on the value of $`M`$ (through the number of ”carry $`1`$” operations needed ).
Based on these considerations it is seems doubtful that one can use Grover’s Algorithm to efficiently process the results of a quantum search of a space region $`R`$. Even if the entanglement problem were solvable, the above results show that, for 2 dimensional space regions, use of Grover’s Algorithm is no more efficient than a classical search. For higher dimensional searches the Algorithm is more efficient. Note that this conclusion is independent of the details of the quantum robot. It applies to any quantum system such as a mobile head that contains sufficient information on board to tell it where to go, what to do on arrival at the endpoint, and how to return to the origin.
## 5. Discussion
In spite of these pessimistic results, quantum robots are interesting objects of study. For instance they may be useful test beds for study of control of quantum systems as the dependence of the task dynamics on the local environmental state is, for some tasks, similar to a feedback loop.
Quantum robots and the associated task dynamics also make clear what is and is not being done in any task. This is shown by the quantum search task in that the quantum robot does no monitoring or control of its behavior. It (or the on board quantum computer) has no knowledge of where it is in $`R`$ at any point or even if it is in $`R`$. For each component memory state $`|X,Y_m`$ there are $`X`$ computation phases with the output system o in state $`|+x_o`$ and $`Y`$ phases with o in state $`|+y_o`$. These phases are interspersed with $`X`$ and $`Y`$ action phases during which anything can happen. For example the quantum robot might move outside $`R`$ or it might not move at all. Of course for these cases it is unlikely that the quantum robot would return to the origin at the end of the task.
This illustrates a valuable aspect of the description of the task dynamics of quantum robots as sequences of alternating computation and action phases. This is that, for the search task examples described here, it makes very clear the lack of awareness and control the quantum robot has over what has happened in the action phases and what it is doing. This argument applies to the computation phases also. For example the ”subtract $`1`$” steps could carry out an arbitrary change to the memory state and the task would continue. In this case the task would no longer be a search task but would be something else.
These considerations are also part of foundational reasons why quantum robots and quantum computers are interesting. If quantum mechanics (or some extension such as quantum field theory) is assumed to be universally applicable, then all systems involved in the validation of quantum mechanics are quantum systems. This includes the systems that make theoretical computations (which includes quantum computers) and the systems that carry out experiments (which includes quantum robots). Thus, in some sense quantum mechanics must describe its own validation, to the maximum extent possible. Exploration of this and the questions of self consistency and possible incompleteness that may occur make this an interesting path of inquiry.
In addition quantum robots, and to some extent quantum computers, are natural systems for investigating several questions. In particular what physical properties must a quantum system have such that
* It is aware of its environment?
* It has significant characteristics of intelligence?
* It changes states of some quantum systems so that the new states can be interpreted as text having meaning to the system generating the text ?
In addition there is a sense in which the existence problem for quantum systems having all these properties is already solved. That is, these systems include the readers, and hopefully the author, of this paper.
PHYSICS DIVISION, ARGONNE NATIONAL LABORATORY, ARGONNE, IL 60439
Current address: Physics Division, Argonne National Laboratory Argonne, IL 60439
E-mail address: pbenioff@anl.gov
|
warning/0003/hep-ph0003201.html
|
ar5iv
|
text
|
# Quantum fluctuations of the quark condensate.
## 1 Introduction.
We study the effect of quantum fluctuations of the quark condensate on the physical vacuum. We use an $`SU\left(2\right)_f`$ Nambu Jona-Lasinio model in the chiral limit. Two ultra-violet regularizations are considered:
* regularization using a sharp cut-off, in which the quark propagators are set to zero when $`k_\mu ^2>\mathrm{\Lambda }^2`$, where $`k_\mu `$ is the euclidean 4-momentum of the quark propagator;
* regularization using a smooth gaussian regulator which will be described below.
We introduce a current quark mass $`m`$ as a source term $`m\overline{\psi }\psi `$ to calculate the quark condensate $`\overline{\psi }\psi `$. We introduce another source term $`\frac{1}{2}j\left(\overline{\psi }\mathrm{\Gamma }_a\psi \right)^2`$ which allows us to calculate the expectation value $`\left(\overline{\psi }\mathrm{\Gamma }_a\psi \right)^2`$ of the *squared* quark condensate. We shall however maintain $`j`$ at a finite value in order to study the response of the system to a deformation induced by this source term. We include the $`1/N_c`$ effects due to the quantum fluctuations of the meson fields. In the quark language, this means that we include the exchange (Fock) term as well as the (RPA) ring diagrams.
Although the quantum fluctuations of the meson fields do not restore chiral symmetry, we do find surprisingly large fluctuations of the quark condensate. We also find that the effective potential is very sensitive to the shape of the regulator. Apparent instabilities appear which we show to be artefacts sharp cut-of used in conjunction with the relatively low values of the cut-offs, used in chiral quark models.
The quantum fluctuations of the meson fields in the Nambu Jona-Lasinio model have been studied for several years , , , . In these studies, the quark and meson loops were regularized with different cut-offs. Since both the quark and meson loops diverge, their relative contribution could be adjusted at will by a proper choice of the cut-offs thereby making it impossible to estimate the importance of the quantum fluctuations of the meson fields.
In this study we consider the Nambu Jona-Lasinio model to be a quark model in the sense that all physical processes can be expressed in term of Feynman graphs involving only quark propagators. The meson fields which are introduced in the process of bosonization are mere intermediate quantities introduced to calculate the partition function. When the quark propagators are regularized, a single cut-off regularizes both the quark and meson loops. No further regularization is required for higher order loops. This approach has also been adopted in Ref. for example, with the exception that a 3-momentum cut-off was used. We shall show that results obtained with 3 and 4-momentum cut-offs can differ even qualitatively. We show that regularization with a 4-momentum cut-off introduces a non-locality which makes the ground state energy unbounded from below, or, equivalently, which makes the model acausal. We relate this instability to the unphysical poles of the quark propagator which are introduced by the regulator.
We consider the regularization of a model to be a physical phenomenon and not simply a way to be rid of infinities wherever they turn up. Because of this we regularize the model action from the outset before calculating the loop integrals and we shall see that this makes a significant difference with the common practice of regularizing the infinities of the loop integrals which are deduced from an unregularized action. We derive all quantities in terms of a regulator which is a function of the squared euclidean 4-momentum $`k^2`$.
The paper is composed of three parts. In the first part we set the notation, we explain how the calculation was performed and we define the relevant range of the model parameters. The second part is devoted to a discussion of the quantum fluctuations of the quark condensate. The last part discusses the instabilities which are displayed by the response of the system, subject to constraint proportional to the squared quark condensate.
## 2 The model euclidean action.
The model is defined by the euclidean action:
$$I_m(q,\overline{q})=\overline{q}\left|i_\mu \gamma _\mu +rmr\right|q\frac{g^2}{2N_c}d_4x\left(\overline{q}\left|r\right|x\mathrm{\Gamma }_ax\left|r\right|q\right)^2$$
(1)
It involves a quark field $`q\left(x\right)x|q`$. The euclidean Dirac matrices are $`\gamma _\mu =\gamma ^\mu =(i\beta ,\stackrel{}{\gamma })`$. The model includes a regulator $`r`$ which is assumed to be diagonal in $`k`$-space: $`k\left|r\right|k^{}=\delta _{kk^{}}r\left(k\right)`$. We consider both a gaussian regulator:
$$r\left(k\right)=e^{\frac{k^2}{2\mathrm{\Lambda }^2}}$$
(2)
and a sharp cut-off:
$$r\left(k\right)=1ifk^2<\mathrm{\Lambda }^2r\left(k\right)=0ifk^2>\mathrm{\Lambda }^2$$
(3)
The use of the the sharp cut-off is tantamount to the calculation of Feynman graphs in which the quark propagators are set to zero when their euclidean 4-momentum $`k_\mu ^2`$ exceeds the cut-off $`\mathrm{\Lambda }^2`$. The bra-ket notation for the Dirac fields is:
$$\overline{q}\left|i_\mu \gamma _\mu +m\right|qd_4x\overline{q}\left(x\right)\left(i_\mu \gamma _\mu +m\right)q\left(x\right)$$
$$x\left|r\right|q=d_4yx\left|r\right|yq\left(y\right)\overline{q}\left|r\right|x=d_4y\overline{q}\left(y\right)y\left|r\right|x$$
(4)
We use the current quark mass $`m`$ to evaluate a regularized quark condensate and that is why the current quark mass $`m`$ is multiplied by the regulator. The $`\mathrm{\Gamma }_a=(1,i\gamma _5\stackrel{}{\tau })`$ are the generators of chiral rotations. The number of flavors is $`N_f`$ and there are $`N_f^2`$ generators $`\mathrm{\Gamma }_a`$. We assumed that the coupling constant $`\frac{g^2}{N_c}`$ is inversely proportional to $`N_c`$.
The partition function $`W`$ is given by the euclidean path integral<sup>2</sup><sup>2</sup>2Pedantically, the partition function is $`Z=e^W`$. In the following we shall call $`W`$ the partition function without fear of confusion.:
$$e^{W\left(m\right)}=D\left(q\right)D\left(\overline{q}\right)e^{I(q,\overline{q})}$$
(5)
At zero temperature and for an infinite translationally invariant system, $`W=\mathrm{\Omega }\epsilon `$ where $`\mathrm{\Omega }`$ is the space-time volume $`\mathrm{\Omega }=d_4x\mathrm{\hspace{0.17em}1}`$ and where $`\epsilon `$ is the energy density $`\epsilon =\frac{E}{V}`$, which is the energy per unit volume of the system in its ground state.
The model is regularized in the same way as the effective quark model which has been derived from a study of the propagation of quarks in an instanton liquid , in which case both the shape of the regulator and the value of the cut-off are derived. However, the model action (1) we are using is not exactly the same as that derived from the instanton liquid . The model described by the action (1) has been actively investigated in both the soliton and meson sectors .
We now adopt a notation which adds considerable transparency to the manipulations made below. We define an interaction $`V`$ by its matrix element:
$$xa\left|V\right|yb=yb\left|V\right|xq=\delta _{ab}\delta \left(xy\right)\frac{g^2}{N_c}xa\left|V^1\right|yb=\delta _{ab}\delta \left(xy\right)\frac{N_c}{g^2}$$
(6)
We define a *delocalized* quark field $`\psi \left(x\right)`$ and the corresponding bilinear forms $`\overline{\psi }\mathrm{\Gamma }_a\psi `$:
$$\psi \left(x\right)=d_4xx\left|r\right|yq\left(y\right)\overline{\psi }\left(x\right)\mathrm{\Gamma }_a\psi \left(x\right)=d_4yd_4z\overline{q}\left(y\right)y\left|r\right|x\mathrm{\Gamma }_ax\left|r\right|zq\left(z\right)$$
(7)
In this notation, the action (1) takes the form:
$$I_m(q,\overline{q})=\overline{q}\left|i_\mu \gamma _\mu +rmr\right|q+\frac{1}{2}\left(\overline{\psi }\mathrm{\Gamma }\psi \right)V\left(\overline{\psi }\mathrm{\Gamma }\psi \right)$$
(8)
The chiral limit is defined to be $`m0`$.
## 3 The constrained system and the effective potential.
The current quark mass $`m`$ serves as a source term to calculate the normalized quark condensate $`\overline{\psi }\psi `$. In order to calculate the quantum fluctuations of the quark condensate, we introduce an extra source term:
$$\frac{1}{2}j\left(\overline{\psi }\mathrm{\Gamma }_a\psi \right)^2=\frac{1}{2}jd_4x\left[\left(\overline{\psi }\psi \right)^2+\left(\overline{\psi }i\gamma _5\tau _a\psi \right)^2\right]$$
(9)
which acts as a constraint on the system and which allows us to calculate the expectation value $`\left(\overline{\psi }\mathrm{\Gamma }_a\psi \right)^2`$ of the *squared* condensate. We prefer to work with the squared condensate $`\left(\overline{\psi }\mathrm{\Gamma }_a\psi \right)^2`$ rather than with $`\left(\overline{\psi }\psi \right)^2`$ because $`\overline{\psi }\psi `$ is only one of the four components of the chiral 4-vector $`\overline{\psi }\mathrm{\Gamma }_a\psi `$. If the system chooses to vibrate in the direction $`\overline{\psi }\psi `$ defined by the ground state, it will do so because the amplitudes of the vibrations in the four directions are independent. But there is no reason to prevent the system to vibrate in the other three directions.
The constraint is introduced into the action (8) and we define *the constrained system* by the partition function:
$$e^{W(j,m)}=D\left(q\right)D\left(\overline{q}\right)e^{I_m(q,\overline{q})+\frac{1}{2}\left(\overline{\psi }\mathrm{\Gamma }\psi \right)j\left(\overline{\psi }\mathrm{\Gamma }\psi \right)}=D\left(q\right)D\left(\overline{q}\right)e^{I_{j,m}(q,\overline{q})}$$
(10)
where:
$$I_{j,m}(q,\overline{q})=\overline{q}\left|i_\mu \gamma _\mu +rmr\right|q+\frac{1}{2}\left(\overline{\psi }\mathrm{\Gamma }\psi \right)\left(Vj\right)\left(\overline{\psi }\mathrm{\Gamma }\psi \right)$$
(11)
is the action of the constrained system.
The ground state expectation value $`\left(\overline{\psi }\mathrm{\Gamma }_a\psi \right)^2`$ of the squared condensate in the constrained system is given by:
$$\frac{1}{2}\mathrm{\Omega }\left(\overline{\psi }\mathrm{\Gamma }_a\psi \right)^2=\frac{W(j,m)}{j}|_{m=0}$$
(12)
and the quark condensate is given by the expression:
$$\mathrm{\Omega }\overline{\psi }\psi =\frac{W(j,m)}{m}|_{m=0}$$
(13)
where $`\mathrm{\Omega }`$ is the space-time volume in euclidean space.
The constrained system, described by the partition function $`W(j,m)`$ is not the same as the system described by the partition function $`W\left(m\right)`$ because it has an additional potential energy equal to $`\frac{1}{2}j\left(\overline{\psi }\mathrm{\Gamma }_a\psi \right)^2=j\frac{W(j,m)}{j}`$. This is why, in the presence of the constraint, the energy density of the system is defined in terms of the *effective potential*:
$$\mathrm{\Gamma }=W(j,m)+\frac{1}{2}j\left(\overline{\psi }\mathrm{\Gamma }_a\psi \right)^2=W(j,m)j\frac{W(j,m)}{j}|_{m=0}$$
(14)
Viewed as a function of $`j`$ (for a fixed $`m`$), a stationary point of the action occurs when:
$$\frac{W(j,m)}{j}=j\frac{^2W(j,m)}{j^2}$$
(15)
so that the effective action is stationary when $`j=0`$, that is, in the absence of the constraint. This is true whatever approximation we use to calculate $`W(j,m)`$.
The effective potential allows us to map out the energy of the system as it deforms under the effect of the constraint $`\frac{1}{2}j\left(\overline{\psi }\mathrm{\Gamma }_a\psi \right)^2`$. The choice of the constraint used to probe the energy surface of the system is, of course, arbitrary. Our choice is justified by the fact that, as we shall see in section 10, the system offers a relatively soft response to the constraint and in some cases it actually displays an instability of the ground state.
## 4 Bosonization in terms of local fields.
Bosonization is simply a convenient way to calculate the partition function $`W(j,m)`$. We introduce *local* auxiliary fields $`\phi _a\left(x\right)`$ and we consider the new euclidean action:
$$I_{j,m}(q,\overline{q},\phi )=\overline{q}\left|i_\mu \gamma _\mu +rmr+r\phi _a\mathrm{\Gamma }_ar\right|q\frac{1}{2}\phi \left(Vj\right)^1\phi $$
(16)
Consider the partition function $`W^{}(j,m)`$ defined by the path integral:
$$e^{W^{}(j,m)}=D\left(\phi \right)D\left(q\right)D\left(\overline{q}\right)e^{I_{j,m}(q,\overline{q},\phi )}$$
(17)
The integral over $`\phi `$ yields:
$$D\left(\phi \right)e^{\frac{1}{2}\left(\left(\overline{\psi }\mathrm{\Gamma }\psi \right)\phi \left(Vj\right)^1\right)\left(Vj\right)\left(\left(\overline{\psi }\mathrm{\Gamma }\psi \right)\left(Vj\right)^1\phi \right)}=e^{\frac{1}{2}tr\mathrm{ln}\left(Vj\right)}$$
(18)
where $`tr`$ denotes a trace in $`(x,a)`$ space:
$$tr\mathrm{\hspace{0.17em}1}=\mathrm{\Omega }N_f^2$$
(19)
(because there are $`N_f^2`$ generators $`\mathrm{\Gamma }_a`$). From (17) and (18) we deduce a relation between the partition functions $`W^{}(j,m)`$ and $`W(j,m)`$:
$$W(j,m)=W^{}(j,m)+\frac{1}{2}tr\mathrm{ln}\left(Vj\right)$$
(20)
If we integrate out the quarks in the path integral (17), we get the partition function $`W(j,m)`$ in the form:
$$e^{W(j,m)}=D\left(\phi \right)e^{I_{j,m}\left(\phi \right)\frac{1}{2}tr\mathrm{ln}\left(Vj\right)}$$
(21)
where $`I_{j,m}\left(\phi \right)`$ is the so-called ”bosonized action”:
$$I_{j,m}\left(\phi \right)=Tr\mathrm{ln}\left(i_\mu \gamma _\mu +rmr+r\phi _a\mathrm{\Gamma }_ar\right)\frac{1}{2}\phi \left(Vj\right)^1\phi $$
(22)
The trace $`Tr`$ is over the variables (space-time $`x`$, Dirac indices, flavor and color) which define the quark field:
$$Tr\mathrm{\hspace{0.17em}\hspace{0.17em}1}=4N_cN_f\mathrm{\Omega }$$
(23)
It is convenient to shift the constituent quark mass to the interaction term by writing $`m+\phi _a\mathrm{\Gamma }_a\phi _a^{}\mathrm{\Gamma }_a`$. The bosonized action (22) becomes:
$$I_{j,m}\left(\phi \right)=Tr\mathrm{ln}\left(i_\mu \gamma _\mu +r\phi _a\mathrm{\Gamma }_ar\right)\frac{1}{2}\left(\phi m\right)\left(Vj\right)^1\left(\phi m\right)$$
(24)
where dropped the prime on $`\phi `$. The partition function is then given by the expression (21) in which $`I_{j,m}\left(\phi \right)`$ is the action (24). In the expression (24), $`m`$ stands for the vector $`m_a`$ with components $`m_a=(m,0,0,0)`$.
It is usual to calculate the effective potential from the texbook formula :
$$\mathrm{\Gamma }\left(\phi \right)=I_m\left(\phi \right)+\frac{1}{2}tr\mathrm{ln}\frac{^2I_m\left(\phi \right)}{\phi \phi }$$
(25)
obtained from a Legendre transform which relates a source term to the expectation value of the field. We do not use this formalism because, in our case, the operator $`\frac{^2I_m\left(\phi \right)}{\phi \phi }`$ has negative eigenvalues. This is easily seen by considering cuts in the Mexican hat shaped $`I_m\left(\phi \right)`$.
## 5 The classical approximation to the constrained system.
The classical approximation consists in approximating the path integral (21) as:
$$W(j,m)=I_{j,m}\left(\phi _j\right)$$
(26)
where $`\phi _j`$ is a stationary point of the action $`I_{j,m}\left(\phi \right)`$, defined in (24). In the classical approximation, we also neglect the constant $`\frac{1}{2}tr\mathrm{ln}\left(Vj\right)`$ which we will include in section 6 together with the contribution of the field fluctuations. The term ”classical” will be used throughout although, at the quark level, it includes the quark loop. As it turns out, it is the classical approximation *to the constrained system* determines the response of the system to the constraint and the quantum fluctuations of the fields are only small corrections. The classical approximation is the leading order contribution in $`N_c`$ and, in the quark representation, it corresponds to the Hartree approximation.
### 5.1 The gap equation and the relation between $`j`$ and $`M`$.
A stationary point of the action (24) occurs for $`\phi _a=(M,0,0,0)`$ where $`M`$ is the solution of the so-called gap equation:
$$\left(Vj\right)^1=\frac{M}{Mm}\frac{1}{2\mathrm{\Omega }}Tr\frac{r^4}{_\mu ^2+r^4M^2}4N_cN_f\frac{M}{Mm}g_M$$
(27)
where $`g_M=g_M\left(q=0\right)`$ and $`g_M\left(q\right)`$ is the function (B.6), defined in appendix B. We shall denote by $`I_{j,m}\left(M\right)`$ the action at the stationary point. An explicit expression is given in appendix A.
Let $`M_0`$ be the solution of the gap equation in the absence of a constraint $`\left(j=0\right)`$:
$$V^1=4N_cN_f\frac{M_0}{M_0m}g_{M_0}$$
(28)
This equation relates $`M_0`$ to the interaction strength $`V`$. The minus sign indicates that the interaction $`V`$ is attractive. Otherwise chiral symmetry would not be spontaneously broken and $`\phi _a=0`$ would be the only stationary point of the action.
We can calculate $`j`$ from the equation:
$$j=\frac{1}{4N_cN_f}\left(\frac{1}{g_M}\frac{1}{g_{M_0}}\right)\frac{m}{4N_cN_f}\left(\frac{1}{Mg_M}\frac{1}{M_0g_{M_0}}\right)$$
(29)
In practice, we start by choosing a value of $`M_0`$, which determines $`V`$. We then choose a value of $`M`$ which determines $`j`$. We can easily derive the relation:
$$\frac{\delta M^2}{\delta j}=4N_fN_cg_M^2\left(\frac{dg_M}{dM^2}\right)^1\left(m=0\right)$$
(30)
which shows that $`M`$ is a monotonically increasing function of $`j`$ so that it makes little difference if we plot the effective potential as a function of $`j`$ or $`M`$. The choice of $`M`$ is more transparent.
### 5.2 The classical estimates of $`\overline{\psi }\psi `$ and $`\left(\overline{\psi }\mathrm{\Gamma }_a\psi \right)^2`$ in the chiral limit.
In the classical approximation, the quark condensate (13) is given by the expression:
$$\mathrm{\Omega }\overline{\psi }\psi _{class}=\frac{I_{j,m}\left(M\right)}{m}|_{m=0}=\mathrm{\Omega }\left(Vj\right)^1M=4\mathrm{\Omega }N_cN_fMg_M$$
(31)
The classical approximation to the squared quark condensate (12) is:
$$\frac{1}{2}\mathrm{\Omega }\left(\overline{\psi }\mathrm{\Gamma }_a\psi \right)^2_{class}=\frac{dI_{j,m}\left(M^2\right)}{dj}=\frac{1}{2}\mathrm{\Omega }M^2\left(4N_cN_fg_M\right)^2$$
(32)
where we used the fact that, for a given value of $`j`$, $`M`$ is a stationary point of the action $`I_{j,m}\left(M\right)`$ .
It follows that, in the classical approximation, the *fluctuation* of the quark condensate is zero:
$$\left(\overline{\psi }\mathrm{\Gamma }_a\psi \right)^2_{class}=\overline{\psi }\psi _{class}^2$$
(33)
as expected. The quantum fluctuations of the quark condensate are due to the quantum fluctuations of the $`\phi `$ fields and they are introduced in section 6.
The gap equation (27) relates $`M`$ to the classical quark condensate and $`M`$ is simply a Hartree insertion:
| $`M=\left(Vj\right)\overline{\psi }\psi _{class}`$ | |
| --- | --- |
(34)
### 5.3 The classical effective potential in the chiral limit.
Because $`M`$ is a stationary point of the action $`I_{j,m}\left(M\right)`$, the classical approximation to the effective potential (14) is:
$$\mathrm{\Gamma }_{class}=I_{j,m}\left(M\right)j\frac{I_{j,m}\left(M\right)}{j}=I_{j,m}\left(M\right)+j\frac{1}{2}\mathrm{\Omega }\left(Vj\right)^2M^2$$
(35)
Using (27) and (29), the classical effective potential can also be expressed as the following function of $`M`$:
$$\mathrm{\Gamma }_{class}\left(M\right)=I_{j,m}\left(M\right)+\frac{1}{2}\mathrm{\Omega }\left(4N_cN_f\right)M^2g_M\left(1\frac{g_M}{g_{M_0}}\right)\left(m=0\right)$$
(36)
An explicit expression for $`I_{j,m}\left(M\right)`$ is given in appendix A.
## 6 Inclusion of the quantum fluctuations of the fields $`\phi `$.
A saddle point evaluation of the partition function (21) yields:
$$W(j,m)=I_{j,m}\left(\phi _j\right)+\frac{1}{2}tr\mathrm{ln}\frac{\delta ^2I_{j,m}\left(\phi \right)}{\delta \phi \delta \phi }|_{\phi =\phi _j}+\frac{1}{2}tr\mathrm{ln}\left(Vj\right)$$
(37)
where $`\phi _j`$ is the stationary point of $`I_{j,m}\left(\phi \right)`$, which is determined by the gap equation (27).
### 6.1 Calculation of $`W(j,m)`$.
To calculate the second term of (37), we need to evaluate the inverse meson propagator matrix:
$$K_{ab}^1(x,y)=\frac{\delta ^2I_{j,m}\left(\phi \right)}{\delta \phi _a\left(x\right)\delta \phi _b\left(y\right)}$$
(38)
at the stationary point of $`I_{j,m}\left(\phi \right)`$. From the action (24), we see that:
$$K^1=\mathrm{\Pi }\left(Vj\right)^1$$
(39)
where $`\mathrm{\Pi }`$ is the polarization function (Lindhardt function):
$$\mathrm{\Pi }_{ab}(x,y)=\frac{\delta ^2}{\delta \phi _a\left(x\right)\delta \phi _b\left(y\right)}Tr\mathrm{ln}\left(i_\mu \gamma _\mu +rmr+r\phi _a\mathrm{\Gamma }_ar\right)$$
$$=Tr\frac{1}{i_\mu \gamma _\mu +rmr+r\phi _a\mathrm{\Gamma }_ar}|x\mathrm{\Gamma }_ax\left|\frac{1}{i_\mu \gamma _\mu +rmr+r\phi _a\mathrm{\Gamma }_ar}\right|y\mathrm{\Gamma }_by|$$
(40)
The partition function (37) becomes:
$$W(j,m)=I_{j,m}\left(\phi _j\right)+\frac{1}{2}tr\mathrm{ln}\left(1\mathrm{\Pi }\left(Vj\right)\right)|_{\phi =\phi _j}$$
(41)
The matrices $`\mathrm{\Pi }`$ and $`V`$, and therefore $`K`$, are diagonal in momentum space and in the flavor indices:
$$qa\left|K^1\right|q^{}b=\delta _{ab}\delta _{qq^{}}K_a^1\left(q\right)qa\left|\mathrm{\Pi }\right|q^{}b=\delta _{ab}\delta _{qq^{}}\mathrm{\Pi }_a\left(q\right)$$
In particular:
$$\mathrm{\Pi }_{a=0}\left(q\right)\mathrm{\Pi }_S\left(q\right)=4N_cN_f\left(\frac{1}{2}q^2f_M^{22}\left(q\right)+M^2\left(f_M^{26}\left(q\right)+f_M^{44}\left(q\right)\right)g_M\left(q\right)\right)$$
$$\mathrm{\Pi }_{a=1,2,3}\left(q\right)\mathrm{\Pi }_P\left(q\right)=4N_cN_f\left(\frac{1}{2}q^2f_M^{22}\left(q\right)+M^2\left(f_M^{26}\left(q\right)f_M^{44}\left(q\right)\right)g_M\left(q\right)\right)$$
(42)
where the functions $`f_M^{np}\left(q\right)`$ and $`g_M\left(q\right)`$ are defined by the expressions (B.5) and (B.6) of appendix B. The regulator $`r`$ ensures that the polarization functions $`\mathrm{\Pi }\left(q\right)`$ vanish when $`q>>2\mathrm{\Lambda }`$.
If we use the gap equation (27) to express $`\left(Vj\right)^1`$ in terms of $`g_Mg_M\left(q=0\right)`$, we obtain analogous expressions for the inverse meson propagators:
$$K_{a=0}^1\left(q\right)K_S^1\left(q\right)=4N_cN_f\left(\frac{1}{2}q^2f_M^{22}\left(q\right)+M^2\left(f_M^{26}\left(q\right)+f_M^{44}\left(q\right)\right)g_M\left(q\right)+\frac{M}{Mm}g_M\right)$$
$$K_{a=1,2,3}^1\left(q\right)K_P^1\left(q\right)=4N_cN_f\left(\frac{1}{2}q^2f_M^{22}\left(q\right)+M^2\left(f_M^{26}\left(q\right)f_M^{44}\left(q\right)\right)g_M\left(q\right)+\frac{M}{Mm}g_M\right)$$
(43)
The pion remains a Goldstone boson even in the constrained system because, in the chiral limit, $`K_{a=1,2,3}^1\left(q\right)\underset{q0}{}0`$. This is an important feature of the choice we have made of the constraint which does not break the chiral symmetry of the lagrangian.
The partition function (41) is then:
$$W(j,m)=I_{j,m}\left(M^2\right)$$
$$+\frac{1}{2}\underset{q}{}\mathrm{ln}\left(1+\frac{Mm}{M}\frac{1}{g_M}\left(\frac{1}{2}q^2f_M^{22}\left(q\right)+M^2\left(f_M^{26}\left(q\right)+f_M^{44}\left(q\right)\right)g_M\left(q\right)\right)\right)$$
$$+\frac{N_f^21}{2}\underset{q}{}\mathrm{ln}\left(1+\frac{Mm}{M}\frac{1}{g_M}\left(\frac{1}{2}q^2f_M^{22}\left(q\right)+M^2\left(f_M^{26}\left(q\right)f_M^{44}\left(q\right)\right)g_M\left(q\right)\right)\right)$$
(44)
It is because we took the trouble to keep track of the term $`\frac{1}{2}tr\mathrm{ln}\left(Vj\right)`$ that the sums over $`q`$ converge. Another way to write $`W(j,m)`$ is:
$$W(j,m)=I_{j,m}\left(M\right)+\frac{1}{2}\underset{qa}{}\mathrm{ln}\frac{K_a^1\left(q\right)}{4N_cN_fg_M}\frac{Mm}{M}$$
(45)
### 6.2 The Fock term (exchange energy) and the ring diagrams.
Consider the expression (41) of $`W(j,m)`$. Using the explicit form (24) of the action $`I_{j,m}\left(M\right)`$ as well as the relation (34), we can write the partition function $`W(j,m)`$, in the chiral limit, as follows:
$$W(j,m)=Tr\mathrm{ln}\left(i_\mu \gamma _\mu +r^2M\right)\frac{1}{2}\overline{\psi }\psi \left(Vj\right)\overline{\psi }\psi +\frac{1}{2}tr\mathrm{ln}\left(1\mathrm{\Pi }\left(Vj\right)\right)|_{\phi =\phi _j}$$
(46)
We can express the second term, together with the expansion of the logarithm, in terms of Feynman graphs. The second term of (46) is the direct (Hartree) energy:
| $`\frac{1}{2}\overline{\psi }\psi \left(Vj\right)\overline{\psi }\psi =`$ | |
| --- | --- |
(47)
The expansion of the logarithm in (46) generates the exchange (Fock) energy as well as the ring diagrams:
$$\frac{1}{2}tr\mathrm{ln}\left(1\mathrm{\Pi }\left(Vj\right)\right)$$
$$=\frac{1}{2}tr\mathrm{\Pi }\left(Vj\right)+\frac{1}{4}tr\mathrm{\Pi }\left(Vj\right)\mathrm{\Pi }\left(Vj\right)\frac{1}{6}tr\mathrm{\Pi }\left(Vj\right)\mathrm{\Pi }\left(Vj\right)\mathrm{\Pi }\left(Vj\right)+\mathrm{}$$
(48)
The terms of (48) are the next to leading order in $`N_c`$. The first term, which is the exchange (Fock) term, is special in that it does not induce any correlations in the quark wavefunction. It is easy to see that the exchange term is more sensitive than the ring diagrams to the high momenta running in the meson loop. As a result, when a sharp cut-off is used, it is the exchange term and not the ring diagrams which dominates the next to leading order contribution (48).
### 6.3 Expressions for the quark condensates and the quark condensates in the chiral limit.
The ground state expectation value of the squared condensate (12) can be calculated from the expression:
$$\frac{1}{2}\mathrm{\Omega }\left(\overline{\psi }\mathrm{\Gamma }_a\psi \right)^2=\frac{I_{j,m}\left(M\right)}{j}|_{m=0}\frac{\delta M^2}{\delta j}\frac{\delta }{\delta M^2}\left(\frac{1}{2}\underset{qa}{}\mathrm{ln}\frac{K_a^1\left(q\right)}{4N_cN_fg_M}\right)$$
(49)
where $`\frac{\delta M^2}{\delta j}`$ is given by (30). The contribution of the exchange term to the squared condensate is:
$$\frac{1}{2}\mathrm{\Omega }\left(\overline{\psi }\mathrm{\Gamma }_a\psi \right)^2_{exch}=\frac{\delta M^2}{\delta j}\frac{\delta }{\delta M^2}\left(\frac{1}{2}\left(Vj\right)\underset{qa}{}\mathrm{\Pi }_a\left(q\right)\right)$$
(50)
The quark condensate (13) can be calculated with the help of (41):
$$\mathrm{\Omega }\overline{\psi }\psi =\frac{W(j,m)}{m}=\frac{}{m}\left(I_{j,m}\left(M\right)+\frac{1}{2}tr\mathrm{ln}\left(1\mathrm{\Pi }\left(Vj\right)\right)|_M\right)$$
(51)
We calculate the derivative $`\frac{W(j,m)}{m}`$ keeping $`j`$ constant. However, we must remember that $`M`$ depends on both $`j`$ and $`m`$ and therefore a contribution arises from the change in $`M`$ when $`m`$ is varied because, in contrast to $`I_{j,m}`$, $`W(j,m)`$ is not stationary with respect to variations of $`M`$. When $`mm+\delta m`$, $`\left(Vj\right)`$ remains constant and:
$$\delta \frac{1}{2}tr\mathrm{ln}\left(1\left(Vj\right)\mathrm{\Pi }\right)=\frac{1}{2}tr\frac{Vj}{1\left(Vj\right)\mathrm{\Pi }}\delta \mathrm{\Pi }=\frac{1}{2}trK\delta \mathrm{\Pi }=\frac{1}{2}\frac{\delta M^2}{\delta m}trK\frac{\delta \mathrm{\Pi }}{\delta M^2}$$
(52)
The quark condensate is thus:
$$\overline{\psi }\psi =\overline{\psi }\psi _{class}+\frac{\delta M^2}{\delta m}\frac{1}{2}\underset{qa}{}K_a\left(q\right)\frac{\delta \mathrm{\Pi }_a\left(q\right)}{\delta M^2}$$
(53)
From the gap equation (27) we see that $`\frac{\delta M^2}{\delta m}=\frac{g_M}{M}\left(\frac{dg_M}{dM^2}\right)^1`$.
The contribution of the exchange term to the quark condensate is:
$$\mathrm{\Omega }\overline{\psi }\psi _{exch}=\left(Vj\right)\frac{\delta M^2}{\delta m}\frac{1}{2\mathrm{\Omega }}\underset{qa}{}\frac{\delta \mathrm{\Pi }_a\left(q\right)}{\delta M^2}$$
(54)
The effective potential (14) can be calculated from the expression
$$\mathrm{\Gamma }=W(j,m)+j\frac{1}{2}\left(\overline{\psi }\mathrm{\Gamma }_a\psi \right)^2$$
(55)
where $`j`$ is given by (29) and $`\frac{1}{2}\left(\overline{\psi }\mathrm{\Gamma }_a\psi \right)^2`$ by (49).
## 7 The relevant range of parameters.
We consider now the parameters of the model. We calculate all quantities in units of the cut-off $`\mathrm{\Lambda }`$ which appears in the regulators (2) and (3). This is convenient because, for example, the effective potential is proportional to $`\mathrm{\Lambda }^4`$ but otherwise it depends only on the ratios $`M_0/\mathrm{\Lambda }`$ and $`M/\mathrm{\Lambda }`$. Thus, the stability of the system depends on the single parameter $`M_0/\mathrm{\Lambda }`$ and, of course on the shape of the regulator. The ratio $`M_0/\mathrm{\Lambda }`$ is therefore the key parameter to consider and we must determine the physically meaningful range of values for $`M_0/\mathrm{\Lambda }`$.
In the vicinity of $`q=0`$, the inverse pion propagator determines the value of $`f_\pi `$. In the chiral limit, we have:
$$K_P^1\left(q\right)Z_Pq^2f_\pi =M_0\sqrt{Z_P}$$
(56)
and $`Z_P`$ is the residue of the pion propagator at the pole $`q=0`$. For a given value of $`M_0/\mathrm{\Lambda }`$, the value of $`f_\pi `$ depends on the value of $`M_0`$. Tables 1 and 2 give the values of $`Z_P`$ and $`f_\pi `$ for various values of $`\frac{M_0}{\mathrm{\Lambda }}`$ and for two values of $`M_0`$, namely $`M_0=300MeV`$ and $`M_0=400MeV`$. For $`M_0=300MeV`$, the observed value $`f_\pi =93MeV`$ is fitted with $`\frac{M_0}{\mathrm{\Lambda }}0.4`$. For $`M_0=\mathrm{\hspace{0.17em}400}MeV`$, higher values $`\frac{M_0}{\mathrm{\Lambda }}0.6`$ and $`\frac{M_0}{\mathrm{\Lambda }}0.7`$ are required respectively, when a sharp cut-off and a gaussian regulator are used. Soliton calculations require $`M_0`$ to lie between $`300`$ and $`400MeV`$ . Higher values of $`M_0`$ and therefore of $`M_0/\mathrm{\Lambda }`$ have also been considered in order to push the unphysical $`q\overline{q}`$ continuum well above the $`\rho `$ mass of $`770MeV`$. In Ref. for example, the value $`M_0/\mathrm{\Lambda }=0.74`$ is used. In order to cover the full range of physically meaningful parameters, we perform our calculations from $`M_0/\mathrm{\Lambda }=0.2`$, which means a relatively high value of the cut-off, up to $`M_0/\mathrm{\Lambda }=0.8`$, which means a low value of the cut-off.
One crucial point here is that the relevant range of parameters (typically $`0.4<M_0/\mathrm{\Lambda }<0.8`$) corresponds to uncomfortably *low* values of the cut-off. Most field theoretic methods applied to statistical mechanics and particle physics have been developed to systems in which $`M_0<<\mathrm{\Lambda }`$. Considerable errors can be (and have been) made by applying methods and concepts, borrowed from the study of systems in which $`M_0<<\mathrm{\Lambda }`$, and applied to systems in which the cut-off $`\mathrm{\Lambda }`$ is of the same order of magnitude as $`M_0`$. Our calculations focus on several problems which one encounters when the cut-off is not much larger than the calculated observables. This is the regime applicable to low-energy hadronic physics and it is not an artefact of the Nambu Jona-Lasinio model. For example, low energy effective theories derived from an instanton liquid yield values of $`M_0/\mathrm{\Lambda }`$ of the order of $`0.4`$.
As an example, consider the inverse $`\sigma `$-meson propagator at $`q=0`$ in the chiral limit. It is given by (43):
$$K_S^1\left(q=0\right)=8N_cN_fM^2f_M4M^2Z_S$$
(57)
where $`f_M`$ is the function $`f_M^{np}\left(q\right)`$, defined in (B.5), and taken at $`q=0`$, where it is independent of $`n`$ and $`p`$. This definition of $`Z_S`$ is quite arbitrary except for the fact that, for large values of the cut-off, that is, when $`\frac{M_0}{\mathrm{\Lambda }}0`$, we have $`Z_S=Z_P`$ and the Nambu Jona-Lasinio model reduces to a linear sigma model. The values of $`Z_S`$ are also listed in the tables 1 and 2. We see that, with a sharp cut-off, $`Z_S`$ and $`Z_P`$ differ by about 30% in the relevant parameter range $`\frac{M_0}{\mathrm{\Lambda }}=0.4\mathrm{\hspace{0.17em}0.8}`$. With a gaussian cut-off the equality $`Z_S=Z_P`$ is not even approximately obtained. The fact that $`Z_SZ_P`$ contradicts most, if not all previously reported calculations of the meson propagators, when they are derived from an unregularized action with loop integrals subsequently regularized. If we had proceeded this way, the loop integrals $`f_M^{np}\left(q\right)`$ would have been independent of $`n`$ and $`p`$, and the function $`g_M\left(q\right)`$ would have become independent of $`q`$. Instead of the expression (43), the meson propagators would have been equal to the usually quoted expressions (in the chiral limit):
$$K_S^1\left(q\right)=4N_cN_f\frac{1}{2}\left(q^2+4M^2\right)f_M\left(q\right)K_P^1\left(q\right)=4N_cN_f\frac{1}{2}q^2f_M\left(q\right)$$
(58)
and $`Z_S`$ would equal $`Z_P`$. The tables 1 and 2 show that considerable errors can be introduced if the regularization is not specified in the action from the outset and adhered to. Of course, these errors would be small if the cut-off were large, that is, if $`M_0/\mathrm{\Lambda }`$ were very small. However, in the applications of the Nambu Jona-Lasinio model to low energy hadronic physics, it is not.
The recently claimed instability of the Nambu Jona-Lasinio model, heralded by Kleinert et al., is based on a reduction of the Nambu Jona-Lasinio model to a non-linear $`\sigma `$-model. Although the arguments presented above raise doubts as to the validity of this reduction, and such doubts have also been voiced elsewhere , we shall show in section 10 that instabilities do indeed arise but that they are artefacts of the sharp cut-off regularization when the cut-off is close to $`M_0`$.
## 8 The quark condensate of the unconstrained system.
In this section we consider the quark condensate (53) of the unconstrained system $`\left(j=0\right)`$. Various contributions to the quark condensate are listed in table 3 for various values of $`M_0/\mathrm{\Lambda }`$. They are expressed in units of $`\mathrm{\Lambda }^3`$. We see that, throughout the range of relevant parameters, the quadratic fluctuations of the fields do not alter significantly the quark condensate. They show no sign of restoring chiral symmetry. These results agree with those found in Ref.. A finer analysis would show that the negative contributions of the pion field are due to the exchange term. Although the field fluctuations do not alter significantly the ground state expectation value of the quark condensate, we shall see in the next section that they do cause an appreciable *quantum fluctuation* of the condensate.
## 9 The quantum fluctuations of the quark condensate in the unconstrained system.
The expressions (53) and (49), which give the quark condensate $`\overline{\psi }\psi `$ and the squared condensate $`\left(\overline{\psi }\mathrm{\Gamma }_a\psi \right)^2`$, allow us to calculate the *quantum* *fluctuation* of the quark condensate $`\mathrm{\Delta }_{\overline{\psi }\psi }`$:
$$\mathrm{\Delta }_{\overline{\psi }\psi }=\sqrt{\left(\overline{\psi }\mathrm{\Gamma }_a\psi \right)^2\overline{\psi }\psi ^2}$$
(59)
The values are listed in table 4 for various values of $`M_0/\mathrm{\Lambda }`$. The fluctuations are due to the exchange and ring diagrams and they vanish in the classical approximation. We see that, relative to the quark condensate, they are quite large: about 50% when a sharp cut-off is used and between 70% and 80% when a gaussian regulator is used. We also see that the quark condensates are more sensitive to the shape of the regulator than $`f_\pi `$. This is because $`f_\pi `$ would diverge logarithmically with a large cut-off, whereas the quark condensates would have a quadratic dependence on the cut-off. This is also the reason why, once $`f_\pi `$ is fixed, larger condensates are obtained with a sharp cut-off than with a gaussian regulator.
## 10 The effective potential in the chiral limit.
Figure 1 shows the classical effective potential (36) in the chiral limit, as a function of $`\frac{M}{\mathrm{\Lambda }}`$, calculated with a gaussian regulator, for various values of $`\frac{M_0}{\mathrm{\Lambda }}`$. The minimum occurs at $`M=M_0`$ and in all this work, we define the zero of energy to be equal to the minimum of the classical action at the point $`M=M_0`$. These curves map out the energy surface of the system while it is being deformed by the constraint $`\frac{1}{2}j\left(\overline{\psi }\mathrm{\Gamma }_a\psi \right)^2`$. As stressed in section 5.1, it makes no difference whether we plot the effective potential as a function of $`j`$, or of $`M`$. Plotted as a function of $`M`$, the curves are easier to understand.
Figure 2 shows shows the classical effective potential (36) calculated with a sharp cut-off. The code can be checked against analytic expressions in this case. We see that for increasing values of $`\frac{M_0}{\mathrm{\Lambda }}`$, that is, for decreasing values of the cut-off, the minimum of the effective potential at $`M=M_0`$ becomes increasingly shallower and that it disappears altogether at the critical value $`\frac{M_0}{\mathrm{\Lambda }}\mathrm{0.\hspace{0.17em}742}`$which can be evaluated analytically. When a gaussian regulator is used, the onset of the instability occurs at the much higher value $`\frac{M_0}{\mathrm{\Lambda }}2.93`$ which was evaluated numerically.
The system appears to display an instability with respect to perturbations caused by the constraint $`\frac{1}{2}j\left(\overline{\psi }\mathrm{\Gamma }_a\psi \right)^2`$. Furthermore, the energy of the system does not seem bounded from below. It goes without saying that the classical action (as opposed to the classical effective potential) displays a minimum at $`M=M_0`$ for all values of $`M_0/\mathrm{\Lambda }`$.
Let us take a closer look at this apparent instability. It is not an artefact of the classical approximation. Figure 3 shows the various contributions to the effective potential when the $`1/N_c`$ field fluctuations are included, using a sharp cut-off with $`M_0/\mathrm{\Lambda }=0.8`$. We see that the field fluctuations lower the energy but that they do not significantly change the shape of the effective potential, so that the instability remains. We also see that the exchange (Fock) term dominates the $`1/N_c`$ corrections because it is more sensitive than the ring diagrams to the high momenta running in the meson loop. These conclusions remain valid for smaller values of $`M_0/\mathrm{\Lambda }`$.
By way of comparison, figure 4 shows the contributions of the $`1/N_c`$ corrections when a gaussian regulator is used. Although the effect of the meson fluctuations is somewhat larger, the stability of the system is not modified. We see however that the ring diagrams make a somewhat larger contribution than the exchange (Fock) term, because the gaussian regulator reduces the effect of the high momenta running in the meson loop. The same conclusions can be reached for different values of $`\frac{M_0}{\mathrm{\Lambda }}`$. As mentioned above, the instability also occurs when a gaussian cut-off is used, but at much higher values of $`M_0/\mathrm{\Lambda }2.93`$. For such high values, the cut-off is too small to be physically meaningful. With a gaussian regulator and in the relevant range of parameters $`0.4<M_0/\mathrm{\Lambda }<0.8`$, one needs to probe the system with values as high as $`M/\mathrm{\Lambda }>4`$ before it becomes apparent that the energy is not bounded from below.
A clue concerning the nature of the instability can be obtained by considering the effective potential obtained with a sharp 3-momentum regularisation. In this regularization, the trace of the quark loop is calculated by integrating the energy variable from $`\mathrm{}`$ to $`+\mathrm{}`$, and by limiting the 3-momentum by the condition $`\left|\stackrel{}{k}\right|<\mathrm{\Lambda }`$. This regularization is tantamount to a limitation of the quantum mechanical Hilbert space available to the quarks. Figure 5 shows that the effective potential, calculated with a sharp 3-momentum cut-off, behaves as expected and that it does not display the instability.
## 11 Unphysical poles of the quark propagator.
The fact that the effective potential, calculated with a sharp 4-momentum cut-off, displays an instability that does not occur when a sharp 3-momentum cut-off is used, can be understood as an effect of the unphysical poles of the quark propagator which are introduced by the 4-momentum regulator. For constant fields, the quark propagator can be written in the form:
$$\frac{1}{k_\mu \gamma _\mu +r^2M}=\frac{k_\mu \gamma _\mu +r_k^2M}{\omega ^2+\stackrel{}{k}^2+r_k^4M^2}\left(k_\mu =(\omega ,\stackrel{}{k})\right)$$
(60)
When a 3-momentum regularisation is used, we have $`r_k^2=1`$ when $`\left|\stackrel{}{k}\right|<\mathrm{\Lambda }`$ and $`r_k^2=0`$ otherwise. In the complex $`\omega `$-plane, the quark propagator has only on-shell poles at $`\omega =\pm i\sqrt{\stackrel{}{k}^2+M^2}`$.
However, when a 4-momentum regulator is used, the quark propagator acquires extra poles, which also occur when proper-time regularization is used . Such poles are unphysical in the sense that, taken seriously, they lead to instabilities of the vacuum. Equivalently one can say that the system behaves as if it was governed by a non-hermitian hamiltonian. This is sometimes also expressed by saying that the theory becomes acausal. We assign the cause of the instability discussed above to the existence of such unphysical poles. When the 4-momentum cut-off is high (and $`M_0/\mathrm{\Lambda }`$ correspondingly low) the effect of these unphysical poles is not felt. But this is not the case in low energy hadronic physics where the parameter $`M_0/\mathrm{\Lambda }`$ is in the range $`0.4<M_0/\mathrm{\Lambda }<0.8`$. Furthermore, the position (and therefore the effect) of the unphysical poles depends very much on the shape of the regulator.
A qualitative understanding of the difference between a sharp 4-momentum cut-off and a soft gaussian regulator can be understood by comparing the location of the poles of the quark propagator in the complex $`k^2`$ plane. When a gaussian regulator is used, the poles of the quark propagator occur when:
$$k^2+M^2e^{\frac{2k^2}{\mathrm{\Lambda }^2}}=0$$
(61)
Poles on the real axis of the complex $`k^2`$ plane occur only if $`M^2<\frac{\mathrm{\Lambda }^2}{2}`$. Otherwise and in addition, poles occur in the complex plane. Figure 6 shows the poles of the quark propagator in the two cases $`M/\mathrm{\Lambda }=0.2`$ (large cut-off) and $`M/\mathrm{\Lambda }=0.8`$ (small cut-off). The poles all lie to the left of the imaginary axis and they move closer to it when the cut-off gets small. Table 5 gives the residues of the poles.
Consider next regularisation using a sharp 4-momentum cut-off. The corresponding regulator does not have an analytic form, but we have checked that very similar results are obtained with a Wood-Saxon shaped regulator $`r_k^4=\frac{1}{1+e^{\frac{k^2\mathrm{\Lambda }^2}{c}}}`$ which becomes equivalent to a sharp cut-off when $`c0`$. The poles of the quark propagator are then the solution of the the equation:
$$k^2+\frac{M^2}{1+e^{\frac{k^2\mathrm{\Lambda }^2}{c}}}=0$$
(62)
Setting $`k^2=x+iy`$, this equation decomposes into the two equations:
$`x+M^2{\displaystyle \frac{1+e^{\frac{x\mathrm{\Lambda }^2}{c}}\mathrm{cos}\frac{y}{c}}{\left(1+e^{\frac{x\mathrm{\Lambda }^2}{c}}\mathrm{cos}\frac{y}{c}\right)^2+\left(e^{\frac{x\mathrm{\Lambda }^2}{c}}\mathrm{sin}\frac{y}{c}\right)^2}}`$ $`=`$ $`0`$ (63)
$`yM^2{\displaystyle \frac{e^{\frac{x\mathrm{\Lambda }^2}{c}}\mathrm{sin}\frac{y}{c}}{\left(1+e^{\frac{x\mathrm{\Lambda }^2}{c}}\mathrm{cos}\frac{y}{c}\right)^2+\left(e^{\frac{x\mathrm{\Lambda }^2}{c}}\mathrm{sin}\frac{y}{c}\right)^2}}`$ $`=`$ $`0`$ (64)
One pole occurs on the negative real axis $`\left(y=0xM^2\right)`$ in the vicinity of $`M^2`$. It is simple to see that the complex poles all occur in the vicinity of the line $`x=\mathrm{\Lambda }`$. The imaginary part $`y`$ is then the solution of the equation:
$$\frac{y}{2c}=\frac{M^2}{4c}\mathrm{tan}\frac{y}{2c}$$
(65)
The spacing between the poles is thus close to $`2\pi c`$ so that they get denser in number as $`c0`$. In that limit, the continuum of poles forms a cut which expresses the discontinuity of the regulator when a sharp cut-off is used. A more exact numerical calculation of the position and residues of the poles is given in the table 6 for the case where $`c=0.05`$$`\mathrm{\Lambda }^2`$ and $`M/\mathrm{\Lambda }=0.8`$.
The unphysical poles produced by a soft gaussian regulator lie to the left of the imaginary axis of the complex $`k^2`$ plane. They are therefore mostly felt at low values of $`k^2`$ where phase space factors reduce their effect. The unphysical poles produced by a sharp cut-off lie close to the boundary $`k^2=\mathrm{\Lambda }^2`$ of high values of $`k^2`$ from which diverging quantities derive most of their contribution. This explains qualitatively the difference between the effect of the two regularizations on the effective potential. It would be worth analyzing whether the instabilities, recently heralded by Kleinert et al.\[Kleinert\], are not also artefacts of the use of a sharp 4-momentum in conjunction with a low cut-off.
## 12 Conclusions.
Although the meson loop contributions do not modify appreciably the value of the quark condensate, they do cause large quantum fluctuations of the quark condensate. The Lorentz invariant regularization of the quark propagator, used in conjunction with the relatively low cut-off values required in low energy hadronic physics, makes the physical vacuum unstable against distortions caused by the squared quark condensate. The instability depends strongly on the shape of the regulator. It is only weakly felt when a soft gaussian regulator is used but its effects are greatly enhanced in calculations which use a sharp 4-momentum cut-off. Large errors can be made if loop integrals are regularized after being derived from an unregularized action, instead of including the regulator in the model action from the outset. The ground state instability can be traced to unphysical poles of the quark propagator which are introduced by the regulator.
## Appendix A Expressions for the classical action and the classical effective action.
In the chiral limit, the value of the classical action at the stationaty point is:
$$I_{j,m=0}\left(M^2\right)I\left(M_0^2\right)$$
(A.1)
$$=\frac{1}{2}4N_cN_f\frac{\mathrm{\Omega }}{\left(2\pi \right)^4}d_4k\left(\mathrm{ln}\frac{k^2+r_k^4M^2}{k^2+r_k^4M_0^2}+\frac{r_k^4M^2}{k^2+r_k^4M^2}\frac{r_k^4M_0^2}{k^2+r_k^4M_0^2}\right)$$
(A.2)
We measure all energies relative to the minimum $`I\left(M_0^2\right)`$ of the classical action in the unconstrained system where $`j=0`$. In units of $`\mathrm{\Omega }\mathrm{\Lambda }^4`$, the classical action depends only on the two variables $`\frac{M}{\mathrm{\Lambda }}`$ and $`\frac{M_0}{\mathrm{\Lambda }}`$.
## Appendix B Expressions for the meson propagators.
### B.1 The meson propagators $`K_{ab}^1`$ and the polarization function $`\mathrm{\Pi }_a`$ at a point $`\phi _a=(M,0,0,0)`$.
From the second order expansion of $`Tr\mathrm{ln}\left(i_\mu \gamma _\mu +r\phi _a\mathrm{\Gamma }_ar\right)`$ we obtain the following expression for the polarization function:
$$\mathrm{\Pi }^{\left(2\right)}\left(\delta \phi \right)=\frac{1}{2}Tr\frac{1}{i_\mu \gamma _\mu +r^2M}r\delta \phi _a\mathrm{\Gamma }_ar\frac{1}{i_\mu \gamma _\mu +r^2M}r\delta \phi _a\mathrm{\Gamma }_ar$$
(B.1)
The calculation is standard. Taking traces over the Dirac and flavor indices, and keeping track of the regulator, we obtain:
$$\mathrm{\Pi }^{\left(2\right)}=4N_cN_f\frac{1}{2\mathrm{\Omega }}\underset{q}{}\delta S\left(q\right)\delta S\left(q\right)\left(\frac{1}{2}q^2f_M^{22}\left(q\right)+M^2\left(f_M^{26}\left(q\right)+f_M^{44}\left(q\right)\right)g_M\left(q\right)\right)$$
$$+4N_cN_f\frac{1}{2\mathrm{\Omega }}\underset{q}{}\delta P_i\left(q\right)\delta P_i\left(q\right)\left(\frac{1}{2}q^2f_M^{22}\left(q\right)+M^2\left(f_M^{26}\left(q\right)f_M^{44}\left(q\right)\right)g_M\left(q\right)\right)$$
(B.2)
where we wrote the fields $`\phi _a`$ in terms of scalar and pseudoscalar fields $`S`$ and $`P_i:`$
$$\mathrm{\Gamma }_a\phi _a=S+i\gamma _5\tau _iP_i$$
(B.3)
The Fourier transforms are defined to be:
$$S\left(q\right)=d_4xe^{iq_\mu x_\mu }S\left(x\right)P_i\left(q\right)=d_4xe^{iq_\mu x_\mu }P_i\left(x\right)$$
(B.4)
The function $`f_M^{np}\left(q\right)`$ is:
$$f_M^{np}\left(q\right)=\frac{4\pi }{\left(2\pi \right)^4}_0^{\mathrm{}}k^3𝑑k_0^\pi 𝑑\alpha \mathrm{sin}{}_{}{}^{2}\alpha \frac{r_{k_1}^nr_{k_2}^p}{\left(k_1^2+r_{k_1}^4M^2\right)\left(k_2^2+r_{k_2}^4M^2\right)}$$
(B.5)
with $`k_1=k\frac{q}{2}`$ and $`k_2=k+\frac{q}{2}`$. The function $`g_M\left(q\right)`$ is:
$$g_M\left(q\right)=\frac{4\pi }{\left(2\pi \right)^4}_0^{\mathrm{}}k^3𝑑k_0^\pi 𝑑\alpha \mathrm{sin}{}_{}{}^{2}\alpha \frac{r_{k_1}^2}{k_1^2+r_{k_1}^4M^2}r_{k_2}^2$$
(B.6)
and we denote by $`g_M`$ the function $`g_M\left(q=0\right)`$.
It follows that the polarization function is diagonal in momentum space:
$$qa\left|\mathrm{\Pi }\right|q^{}b=\delta _{ab}\delta _{qq^{}}\mathrm{\Pi }_a\left(q\right)$$
(B.7)
where:
$$\mathrm{\Pi }_{a=0}\left(q\right)\mathrm{\Pi }_S\left(q\right)=4N_cN_f\left(\frac{1}{2}q^2f_M^{22}\left(q\right)+M^2\left(f_M^{26}\left(q\right)+f_M^{44}\left(q\right)\right)g_M\left(q\right)\right)$$
$$\mathrm{\Pi }_{a=1,2,3}\left(q\right)\mathrm{\Pi }_P\left(q\right)=4N_cN_f\left(\frac{1}{2}q^2f_M^{22}\left(q\right)+M^2\left(f_M^{26}\left(q\right)f_M^{44}\left(q\right)\right)g_M\left(q\right)\right)$$
(B.8)
The inverse propagator matrix $`K^1`$ is obtained by adding $`xa\left|\left(Vj\right)^1\right|yb`$. We use the gap equation (27) to write:
$$\left(Vj\right)^1=4N_cN_f\frac{M}{Mm}g_M$$
(B.9)
so that:
$$K_S^1\left(q\right)=4N_cN_f\left(\frac{1}{2}q^2f_M^{22}\left(q\right)+M^2\left(f_M^{26}\left(q\right)+f_M^{44}\left(q\right)\right)g_M\left(q\right)+\frac{M}{Mm}g_M\left(0\right)\right)$$
$$K_P^1\left(q\right)=4N_cN_f\left(\frac{1}{2}q^2f_M^{22}\left(q\right)+M^2\left(f_M^{26}\left(q\right)f_M^{44}\left(q\right)\right)g_M\left(q\right)+\frac{M}{Mm}g_M\left(0\right)\right)$$
(B.10)
Acknowledgments.
The author wishes to thank W.Broniowski and B.Golli for numerous discussions, and J.Zinn-Justin and L.Pitaevskii for help in undestanding reference .
|
warning/0003/hep-ph0003073.html
|
ar5iv
|
text
|
# Charm multiplicity and the branching ratios of inclusive charmless b quark decays in the general two-Higgs-doublet models
## I. Introduction
In the forthcoming years, experiments at SLAC and KEK B-factories, HERA-B and other high energy colliders will measure various branching ratios and CP-violating asymmetries of B decays . The expected large number of B decay events ( say $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{10}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{10}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{9}$}`$) may allow us to explore the physics of CP violation, to determine the flavor parameters of the electroweak theory, and to probe for signals or evidences of new physics beyond the Standard Model (SM) \[1 - 6\].
Among various B meson decay modes, the decay $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}`$ have been, for example, the hot subject of many investigations , since these decay modes may be affected by loop contributions from various new physics models. Great progress in both the theoretical calculation and the experimental measurement enable one to constrain the new physics models, such as the two-Higgs-doublet model (2HDM) , the minimal supersymmetric standard model and the Technicolor models .
For many years, it appeared that the SM prediction for the semileptonic branching ratio $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ is much larger than the values measured at $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{{\rm Y}}}$}`$ resonance and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Z}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$-peak . More recently, the theoretical predictions have been refined by including full $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{O}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ QCD corrections . These progress, consequently, have lowered the predicted $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ and now adequately reproduce the experimental results . However, the measurements of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ obtained at the $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{{\rm Y}}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Z}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$ resonance are still disagree slightly . Besides the $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ problem, there is another so-called ”missing charm puzzle” : the charm multiplicity $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ measured at CLEO and LEP ( especially at CLEO, the $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{{\rm Y}}}$}`$ resonance ) is smaller than the theoretical prediction. Among various possible explanations for the missing charm/$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ problem, the most intriguing one would be an enhanced $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{X}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{o}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{h}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{a}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}}`$ rate due to new physics beyond the SM . An enhanced $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}`$ can decrease the values of both $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ and the $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ simultaneously . The large branching ratio $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{X}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ reported recently by CLEO provided a new hint for enhanced $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}`$. Besides those explanations based on the SM , new physics interpretation for this large ratio is also plausible .
In a previous paper , we calculated, from the first principle, the new contributions to inclusive charmless b quark decays $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}}`$ from the gluonic charged-Higgs penguin diagrams in the so-called Model III: the two-Higgs-doublet model with flavor changing couplings . In the considered parameter space, we found that the branching ratio $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ ($`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$) can be increased by roughly an order of magnitude, which is much larger than that in the ordinary 2HDM’s . In , however, we used the language of form factors $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{F}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{F}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}`$ and took into account the QCD corrections partially by using the $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ directly to calculate the branching ratios.
In this paper, in the framework of general 2HDM’s, we will calculate the branching ratios of various inclusive charmless b decays by using the low energy effective Hamiltonian including next-to-leading order (NLO) QCD corrections , and investigate the new physics effects on the theoretical predictions for both $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$.
This paper is organized as follows. In Sec.II, we describe the basic structures of the model III, extract out the Wilson coefficients, draw the constraint on parameter space of the model III from currently available data. In Sec.III, we calculate the branching ratios $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}`$ in the model III and II with the inclusion of NLO QCD corrections. In Sec.IV, we examine the current status and new physics effects on the determination of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$. The conclusions and discussions are included in the final section.
## II. The general 2HDM’s and experimental constraint
The simplest extension of the SM is the so-called two-Higgs-doublet models. In such models, the tree level flavor changing neutral currents(FCNC’s)are absent if one introduces an ad hoc discrete symmetry to constrain the 2HDM scalar potential and Yukawa Lagrangian. Lets consider a Yukawa Lagrangian of the form
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Y}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{U}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}\stackrel{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{~}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\varphi }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{U}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{D}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\varphi }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{D}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\xi }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{U}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}\stackrel{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{~}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\varphi }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{U}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\xi }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{D}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\varphi }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{D}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{h}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{.}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{.}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (1)
where $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\varphi }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}`$ ($`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}`$) are the two Higgs doublets of a two-Higgs-doublet model, $`\stackrel{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{~}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\varphi }$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\tau }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\varphi }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ ($`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{U}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}}`$) with $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ are the left-handed isodoublet quarks (right-handed up-type quarks), $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{D}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}}`$ are the right-handed isosinglet down-type quarks, while $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{U}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{D}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\xi }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{U}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{D}$}}`$ ($`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}`$ are family index ) are generally the nondiagonal matrices of the Yukawa coupling. By imposing the discrete symmetry
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\varphi }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\varphi }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\varphi }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\varphi }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{D}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{D}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{U}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{U}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}`$ (2)
one obtains the so called Model I and Model II. In Model I the third and fourth term in eq.(1) will be dropped by the discrete symmetry, therefore, both the up- and down-type quarks get mass from Yukawa couplings to the same Higgs doublet $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\varphi }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}`$, while the $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\varphi }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}`$ has no Yukawa couplings to the quarks. For Model II, on the other hand, the first and fourth term in Eq.(1) will be dropped by imposing the discrete symmetry. Model II has, consequently the up- and down-type quarks getting mass from Yukawa couplings to two different scalar doublets $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\varphi }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\varphi }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}`$.
During past years, the models I and II have been studied extensively in literature and tested experimentally, and the model II has been very popular since it is the building block of the minimal supersymmetric standard model. In this paper, we focus on the third type of 2HDM , usually known as the model III . In the model III, no discrete symmetry is imposed and both up- and down-type quarks then may have diagonal and/or flavor changing couplings with $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\varphi }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\varphi }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}`$. As described in , one can choose a suitable basis $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ to express two Higgs doublets
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\varphi }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}}{\sqrt{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\left(}$}\begin{array}{c}\sqrt{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\chi }$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\\ \colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{v}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\chi }$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\end{array}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\right)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\varphi }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}}{\sqrt{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\left(}$}\begin{array}{c}\sqrt{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\\ \colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\end{array}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\right)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (7)
and take their vacuum expectation values as the form
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{<}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\varphi }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{>}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\left(}$}\begin{array}{c}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\\ \colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{v}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\sqrt{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}\end{array}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\right)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{<}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\varphi }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{>}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (10)
where $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{v}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\sqrt{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{F}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{246}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}`$. The transformation relation between $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ and the mass eigenstates $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{h}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{A}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ can be found in . The $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}`$ are the physical charged Higgs boson, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{h}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$ are the physical CP-even neutral Higgs boson and the $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{A}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$ is the physical CP-odd neutral Higgs boson. After the rotation of quark fields, the Yukawa Lagrangian of quarks are of the form ,
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Y}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{U}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}\stackrel{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{~}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\varphi }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{U}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{D}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\varphi }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{D}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\widehat{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\xi }$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{U}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}\stackrel{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{~}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\varphi }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{U}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\widehat{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\xi }$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{D}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\varphi }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{D}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{.}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{.}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (11)
where $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{U}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{D}$}}`$ correspond to the diagonal mass matrices of up- and down-type quarks, while the neutral and charged flavor changing couplings will be <sup>1</sup><sup>1</sup>1We make the same ansatz on the $`\colorbox[rgb]{1,1,1}{$\xi $}_{\colorbox[rgb]{1,1,1}{$i$}\colorbox[rgb]{1,1,1}{$j$}}^{\colorbox[rgb]{1,1,1}{$U$}\colorbox[rgb]{1,1,1}{$,$}\colorbox[rgb]{1,1,1}{$D$}}`$ couplings as the Ref.. For more details about the definition of $`\widehat{\colorbox[rgb]{1,1,1}{$\xi $}}^{\colorbox[rgb]{1,1,1}{$U$}\colorbox[rgb]{1,1,1}{$,$}\colorbox[rgb]{1,1,1}{$D$}}`$ one can see Ref..
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\xi }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{U}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{D}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}{\displaystyle \frac{\sqrt{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{v}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\widehat{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\xi }$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{a}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{l}$}}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{U}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{D}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\xi }$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{U}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{D}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\widehat{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\xi }$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{h}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{a}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{U}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\xi }$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{U}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{K}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\widehat{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\xi }$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{h}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{a}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{D}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{K}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\xi }$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{D}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (12)
where $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{K}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}}`$ is the Cabibbo-Kabayashi-Maskawa mixing matrix , $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ are the generation index. The coupling constants $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}`$ are free parameters to be determined by experiments, and they may also be complex.
In the model II and assuming $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{tan}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\beta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}`$, the constraint on the mass of charged Higgs boson due to CLEO data of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}`$ is $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{350}$}`$ ( 200 ) GeV at the LO (NLO) level . For the model I, however, the limit can be much weaker due to the possible destructive interference with the SM amplitude.
For the model III, the situation is not as clear as the model II because there are more free parameters here. As pointed in , the data of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{K}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{K}$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$ mixing processes put severe constraint on the FC couplings involving the first generation of quarks. One therefore assume that,
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{f}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{o}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}`$ (13)
Imposing the limit in Eq.(13) and assuming all other $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}`$ parameters are of order 1, Atwood et al. found a very strong constraint of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{>}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{600}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}`$ by using the CLEO data of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}`$ decay available in 1995. In Ref., Aliev et al. studied the $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}`$ decay in the model III by extending the NLO results of the model II to the case of model III, and found some constraints on the FC couplings.
In a recent paper , Chao et al., studied the decay $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}`$ by assuming that only the couplings $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}`$ are non-zero. They found that the constraint on $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}`$ imposed by the CLEO data of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}`$ can be greatly relaxed by considering the phase effects of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}`$. The constraints by $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}}`$ mixing, the neutron electric dipole moment(NEDM), the $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Z}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$-pole parameter $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\rho }$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}`$ give the following preferred scenario :
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.3}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{50}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{A}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{h}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{80}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{120}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{;}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{80}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{200}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}`$ (14)
In the following sections, we will calculate the new physics contributions to the inclusive charmless decays of b quark in the Chao-Cheung-Keung (CCK) scenario of model III . Such model III has following advantages:
1. Since we keep only the couplings $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}`$ none zero, the neutral Higgs bosons do not contribute at tree level or one-loop level. The new contributions therefore come only from the charged Higgs penguin diagrams with the heavy internal top quark.
2. The new operators $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{O}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{9}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{10}$}}`$ and all flipped chirality partners of operators $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{O}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{10}$}}`$ as defined in do not contribute to the decay $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}`$ and other inclusive charmless decays under study in this paper.
3. The free parameters in this model III are greatly reduced to $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}`$.
In order to find more details about the correlations between $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}`$ and couplings $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}`$ by imposing the new CLEO data of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}`$, we recalculate the decay $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}`$ in the model III. For the sake of simplicity, we do not consider the less interesting model I further in this paper.
The effective Hamiltonian for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{X}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}`$ at the scale $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{O}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ is given by
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{f}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{f}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{F}$}}{\sqrt{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\left[}$}{\displaystyle \underset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}}{\overset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{6}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{7}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{7}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\right]}$}`$ (15)
The explicit expressions of operators $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{6}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{7}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}`$, as well as the corresponding Wilson coefficients $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ in the SM can be found for example in .
In the model III, the left-handed QED magnetic-penguin operator $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{7}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}`$ and the left-handed QCD magnetic-penguin operator $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}`$ may also play an important role,
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{7}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pi }$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\sigma }$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\nu }$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{5}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{F}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\nu }$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (16)
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pi }$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\sigma }$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\nu }$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{5}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{T}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\beta }$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{a}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\beta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\nu }$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{a}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (17)
In the SM and ordinary 2HDM’s, both operators $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{7}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}`$ are absent because one usually assume that $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$. In the model III, however, these two left-handed operators may contribute effectively because the Wilson coefficients $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{7}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}`$ may be rather large to compensate for the suppression of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}`$.
In Ref., we calculated the $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}`$ decay in the model III from the first principle and obtained the corresponding form factors $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{F}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{F}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}`$. Following the standard procedure and using the Feynman rules in the model III , we evaluate the Feynman diagrams for both $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}`$ decay as shown in Fig.1, extract out the Wilson coefficients $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ at the energy scale $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}`$ by matching the full theory onto the effective theory,
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{5}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{6}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (18)
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (19)
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{7}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{18}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{D}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{y}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (20)
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{7}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{7}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{12}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{A}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{y}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{y}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\theta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (21)
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{12}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{D}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{y}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (22)
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{12}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{D}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{y}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{E}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{y}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\theta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (23)
with
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{7}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{A}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (24)
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{D}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (25)
where $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{y}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}`$, the phase angle $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\theta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\theta }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\theta }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}`$, while $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\theta }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}`$ ( $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\theta }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}`$) is the phase angle of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}`$ ($`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}}`$). When compared with the Eqs.(18,19) of Ref., the second and third terms in Eqs.(21) and (23) have an additional factor of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}`$, since $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\xi }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{U}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{D}$}}`$ used here has as additional factor $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\sqrt{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}`$. The Inami-Lim functions $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{A}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{D}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{E}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ are of the form,
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{A}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{7}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{5}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{12}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{log}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{[}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{]}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (26)
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{D}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{5}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{log}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{[}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{]}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (27)
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{y}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{y}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{5}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{y}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{12}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{y}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{y}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{y}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{6}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{y}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{log}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{[}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{y}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{]}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (28)
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{E}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{y}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{y}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{y}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{y}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{y}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{y}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{log}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{[}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{y}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{]}$}`$ (29)
The Wilson coefficients given in Eqs.(18-23) contained the contributions from both the $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}`$-penguin and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}`$-penguin diagrams.
It is easy to see that both $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{7}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ in Eqs.(20) and (22) will be doubly suppressed by the ratio $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}`$ when $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}`$ is small as preferred by the data of NEDM . For typical values of relevant parameters, say $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.3}$}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{40}$}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\theta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{200}$}`$ GeV, One finds numerically that $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{7}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{10}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{5}$}`$, while $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{7}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.8}$}`$. Consequently, the left-handed Wilson coefficients are much smaller than their right-handed counterparts and therefore will be neglected in the following calculations.
At the lower energy scale $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{O}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$, the Wilson coefficients $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ for the decay $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}`$ at the leading order are of the form
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`{\displaystyle \underset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}}{\overset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{6}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{k}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{a}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{6}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (30)
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{7}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}^{\frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{16}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{23}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{7}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\left(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}^{\frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{14}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{23}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}^{\frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{16}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{23}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\right)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}{\displaystyle \underset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}}{\overset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{h}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{a}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (31)
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{7}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}^{\frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{16}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{23}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{7}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\left(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}^{\frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{14}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{23}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}^{\frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{16}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{23}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\right)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}{\displaystyle \underset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}}{\overset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{h}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{a}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (32)
where $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$, and the scheme-independent numbers $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{a}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{k}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{h}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}`$ can be found in .
Using the effective Hamiltonian, the branching ratio of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}`$ at the leading order can be written as,
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{6}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pi }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{f}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{7}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\nu }$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (33)
where $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{O}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\nu }$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{10.7}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.4}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$ is the measured semileptonic branching ratio of b decay, and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{f}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ is the phase space factor,
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{f}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{6}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{24}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{log}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{[}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{]}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (34)
where $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{p}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{o}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{l}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{p}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{o}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{l}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}}`$. It is straightforward to write down the branching ratios $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ for the SM and model II.
In the numerical calculations, the following input parameters will be used implicitly:
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{80.41}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{91.187}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{137}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.118}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{F}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1.16639}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\times }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{10}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{5}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.13}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1.4}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4.8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{168}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Lambda }}$}_{\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}}}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{5}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.225}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{A}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.84}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.22}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\rho }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.20}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.34}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (35)
where $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{A}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\rho }$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}`$ are the Wolfenstein parameters of the CKM mixing matrix. $`\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ here refers to the running current top quark mass normalized at $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}`$ and is obtained from the pole mass $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{p}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{o}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{l}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{176}$}`$ GeV. For the running of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}`$, the two-loop formulae will be used.
Fig.2 shows the branching ratios $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ in the SM and models II and III, assuming $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.3}$}`$,$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{35}$}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\theta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{30}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{tan}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\beta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}`$. The horizontal band between two dotted lines corresponds to the CLEO data : $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\times }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{10}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4.5}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\times }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{10}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4}$}`$. The short-dashed line is the SM prediction, and the long-dashed and solid curve show the ratio in the model III for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\theta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{30}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$, respectively. The dot-dashed curve shows the same ratio at the leading order in the model II. From the Fig.2, the lower and upper limit on $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}`$ in the model III can be read out:
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{185}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{238}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{f}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{o}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\theta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{215}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{287}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{f}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{o}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\theta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{30}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$ (36)
These limits are consistent with those given in Eq.(14). If we take into account the errors of theoretical predictions in model III, the corresponding mass limit will be relaxed by about 20 GeV.
From above analysis, we get to know that for the model III the parameter space
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{f}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{o}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{o}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.3}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{35}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\theta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{30}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{200}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{100}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (37)
are allowed by the available data. For the mass $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}`$, searches for pair production at LEP have excluded masses $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{77}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}`$ . Combining the direct and indirect limits together, we here conservatively consider a larger range of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{100}$}`$ GeV $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{300}$}`$ GeV, while take $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{200}$}`$ GeV as the typical value.
## III. Inclusive charmless b quark decays
In this section, we will calculate the new physics contributions to the two-body and three-body inclusive charmless decays of b quark induced by the charged Higgs gluonic penguin diagrams in the models II and III.
### A. $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{l}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{o}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}`$ decay
The branching ratio of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}`$ at the leading order can be written as,
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pi }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{f}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\kappa }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\nu }$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (38)
with
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}^{\frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{14}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{23}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}{\displaystyle \underset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}}{\overset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{h}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{a}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (39)
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}^{\frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{14}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{23}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}{\displaystyle \underset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}}{\overset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{h}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{a}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (40)
where $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ with $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{O}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$, and the numbers $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{a}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}`$ and $`\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{h}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}`$ can be found in . The factor $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\kappa }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ contains the QCD correction to the semileptonic decay rate $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\nu }$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ . To a good approximation the $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\kappa }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ is given by
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\kappa }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pi }$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\left[}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\left(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pi }$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{31}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\right)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\right]}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{.}$}`$ (41)
And an exact analytic formula for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\kappa }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ can be found in ref..
For $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}`$ decay, one simply substitutes $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}`$ by $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}`$ in Eq.(38). For the model II, one simply replaces $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ in Eq.(38) with $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}}`$ as given in .
Fig.3 shows the branching ratios of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ in the SM and the models II and III, assuming $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.3}$}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{35}$}`$, and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\theta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{30}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$. The dots line in Fig.3 is the SM prediction $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.27}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$, while the short-dashed curve shows the branching ratio $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.81}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$ in the model II assuming $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{tan}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\beta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{200}$}`$ GeV. In the model III, the enhancement to the ratio $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ can be as large as an order of magnitude: $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2.34}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4.84}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$ for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{200}$}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{100}$}`$ GeV respectively, as illustrated by the long-dashed and solid curves in Fig.3. The model III is clearly more promising than the model II to provide a large enhancement to the decay $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}`$. Although the current enhancement is still smaller than $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{10}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$ as expected, for example in Refs., such a significant increase is obviously very helpful for us to provide a reasonable solution for the problems such as the “ missing charm puzzle” or the deficit $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$, as being discussed below.
### B. Three-body charmless b quark decays
Within the SM, the three-body inclusive charmless b quark decays have been calculated at LO and NLO level for example in refs.. In Ref., Lenz et al. took into account the NLO QCD corrections from the gluonic penguin diagrams with insertions of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}`$ and the diagrams involving the interference of the $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}`$ with $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{6}$}}`$ .
The standard theoretical frame to calculate the decays $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}}`$ for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\}}$}`$ is based on the effective Hamiltonian,
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{f}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{f}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{F}$}}{\sqrt{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\left\{}$}{\displaystyle \underset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}}{\overset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\left(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{v}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{v}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\right)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{v}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\left[}$}{\displaystyle \underset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}}{\overset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{6}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\right]}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\right\}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{h}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{.}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{.}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (42)
where $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{v}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}`$ and the corresponding operator basis reads:
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\beta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{A}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\beta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{A}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (43)
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{A}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\beta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\beta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{A}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (44)
with $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$, and
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{A}$}}{\displaystyle \underset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\beta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\beta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{A}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (45)
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\beta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{A}$}}{\displaystyle \underset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\beta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{A}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (46)
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{5}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{A}$}}{\displaystyle \underset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\beta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\beta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{A}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (47)
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{6}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\beta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{A}$}}{\displaystyle \underset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\beta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{A}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (48)
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pi }$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\sigma }$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\nu }$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{5}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{T}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\beta }$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{a}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\beta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\nu }$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{a}$}`$ (49)
where the $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}`$ are current-current operators, the $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{6}$}`$ are QCD penguin operators, while the $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}`$ is the chromo-magnetic dipole operator.
For the SM part, we will use the formulae presented in directly. For the new physics part in the models II and III under study here, we take into account the new contributions from charged-Higgs gluonic penguins by using the Wilson coefficient $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}}`$ as given in Eq.(40) in the calculation, this coefficient comprises both the SM and the new physics contributions. All other Wilson coefficients remain unmodified.
When the NLO QCD corrections are included, one usually expand the decay width to order $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}`$,
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pi }$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\left(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{p}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\right)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{O}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (50)
where $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}`$ denotes the decay rate at the LO level, while the second part represents the NLO QCD corrections. We here use the renormalization-scheme(RS) independent terms $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{p}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}`$. For the convenience of the reader, the explicit expressions of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{p}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}`$ will be given in Appendix. The term $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}`$ in Eq.(50) ( which will be defined below in Eq.(61) ) is already RS independent. For the three-body decays $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}}`$ one simply substitutes $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}`$ by $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}`$ in Eqs.(42-50).
At the NLO, the RS dependent Wilson coefficients $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ are given by
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pi }$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{6}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{.}$}`$ (51)
where $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}`$ are the RS independent LO Wilson coefficients, and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}`$ are the RS dependent NLO corrections ,
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`{\displaystyle \underset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}}{\overset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{k}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{a}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (52)
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`{\displaystyle \underset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}}{\overset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\left[}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{E}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{f}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\right]}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{a}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (53)
where $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\eta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}`$, the function $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{E}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ and all the numbers $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{a}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{k}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{f}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}}`$, and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}}`$ can be found in . The NLO QCD correction $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}`$ is RS dependent and can be split into two parts:
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}{\displaystyle \underset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{k}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}}{\overset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{6}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{J}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{k}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{k}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{6}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{.}$}`$ (54)
where parameters $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{J}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{k}$}}`$ are usually RS dependent, $`\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}`$ is RS independent, and the precise definitions of the terms in Eq.(54) can be found for example in . The terms involving $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{J}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{k}$}}`$ will be absorbed into $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{p}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}}`$ to make the latter scheme independent.
In the leading order the decays $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}}`$ are penguin-induced processes proceeding via $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{6}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}`$, while $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}}`$ also receive contributions from $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}`$. Combining both cases, the decay width at the LO level can be written as
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{F}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{5}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{64}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pi }$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}{\displaystyle \underset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}}{\overset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{v}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}{\displaystyle \underset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}}{\overset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{6}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{v}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}`$ (55)
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}{\displaystyle \underset{\genfrac{}{}{0pt}{}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{6}$}}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{Re}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\left(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{v}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{v}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\right)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\}}$}`$
with $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}`$ for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$ for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}`$. The coefficients $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}`$ read
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{16}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pi }$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{6}$}}}{\displaystyle \colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{𝑑}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Phi }}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pi }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}}`$ (56)
with $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}`$ here. Setting the final state quark masses to zero one finds
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\{}$}\begin{array}{cc}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}\hfill & \colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\text{ for }}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\text{ and }}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\text{ even }}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\hfill \\ \colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}\hfill & \colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\text{ for }}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\text{ and }}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\text{ odd }}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\hfill \end{array}`$ (59)
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{55}$}}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{66}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{56}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{65}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{.}$}`$ (60)
Here $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}`$ for the decays $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}`$, in which the final state contains two identical particles, and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$ otherwise. The remaining $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}`$’s are zero.
Now we turn to study the contributions from the interference of the tree diagram with $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}`$ with operators $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{6}$}}`$, as shown in Fig.3 of Ref.. The tree-level correction $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}`$ is already at the order of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}`$ and is given by
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{F}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{5}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{32}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pi }$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{Re}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\left[}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{v}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{v}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}}{\displaystyle \underset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}}{\overset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{v}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}}{\displaystyle \underset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}}{\overset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{6}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\right]}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{.}$}`$ (61)
in the model III, where $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}}`$ has been given in Eq.(40) with $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{O}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$. For the case of the SM and model II, simply replace $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}}`$ with the appropriate $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$. The definitions and numerical values of coefficients $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}}`$ can be found in . As mentioned previously, the Wilson coefficient $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}}`$ now comprises the contributions from both the W-penguin and the charged-Higgs penguin diagrams. In this way, the new physics contributions are taken into account.
For the b quark decay rates one usually normalize them to the semileptonic decay rate of b quark,
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{l}$}}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{l}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\nu }$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{l}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\nu }$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\nu }$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\nu }$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\nu }$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (62)
for the sake of eliminating the factor of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{5}$}`$ common to all b decay rates. One also define the charmless decay rate of b quark as
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`{\displaystyle \underset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\tau }$}}`$ (63)
where rare radiative decays, for example $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}`$, have been neglected. To order $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}`$, the semileptonic decay rate takes the form
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\nu }$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{F}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{5}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{192}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pi }$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{f}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\kappa }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ (64)
where the factors $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{f}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\kappa }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ have been given in Eqs.(34) and (41).
To calculate $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}}`$ we also need explicit expressions of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}}`$. For $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}}`$ one finds ,
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{f}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\left\{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\kappa }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\kappa }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{6}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\left[}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{f}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\right]}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\right\}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (65)
where $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.12}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}`$ encodes the chromomagnetic interaction of the b quark with light degrees of freedom, and the factors of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{f}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\kappa }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ have been given in Eqs.(34) and (41).
From Eq.(38), we get
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pi }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{f}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\kappa }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (66)
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pi }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{f}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\kappa }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{.}$}`$ (67)
For $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}}`$, we use the formulae as given in ,
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{16}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pi }$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{f}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\kappa }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{z}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}{\displaystyle \underset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{f}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (68)
where $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}`$, the functions $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{f}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ can be found for example in . In the numerical calculation, we assume that $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}`$. Since the new contribution to the decay $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}`$ due to the charged Higgs penguin is negligibly small , we do not consider the new physics corrections to this decay here. In Ref., the authors did not include $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}}`$ in the estimation of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}}`$. We here will include this mode, since its branching ratio is rather large , as shown in the Table 1.
The corresponding branching ratios for two-body and three-body charmless b decays are defined as
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{X}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{X}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\nu }$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{p}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (69)
where ratios $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{X}$}`$ have been defined previously. In the numerical calculations, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\nu }$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{p}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{10.70}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$ will be used <sup>2</sup><sup>2</sup>2For more details, one can see the discussions about the semileptonic branching ratios of b decay in next section..
By using the input parameters as given in Eq.(35) and assuming $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.3}$}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{35}$}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{200}$}`$GeV and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\theta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$ or $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{30}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$, we find the numerical results of the decay rates and the branching ratios for various charmless b quark decays and collect them in Table 1. We also show the corresponding results in the model II assuming $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{200}$}`$ GeV and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{tan}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\beta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}`$. For larger $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{tan}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\beta }$}`$ the new physics contributions in model II will become smaller. $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}`$ in Table 1 is defined as
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{X}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\left[}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{X}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{X}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\right]}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}}`$ (70)
Fig.4 shows the mass dependence of the branching ratios $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ with $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\}}$}`$ in the SM and model III, using the input parameters in Eq.(35) and assuming $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.3}$}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{35}$}`$, and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\theta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{30}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$. In Fig.4, the three curves ( horizontal lines) are the theoretical predictions in the model III ( SM ) for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}`$, respectively. For $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{200}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{V}$}`$, as listed in Table 1, the enhancement to the decay mode $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}}`$ is only $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4.7}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$, but the enhancements to other five three-body b quark decay modes are rather large: from $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{30}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$ to $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{70}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$. In the model II, however, the new contributions are negative and will decrease the branching ratios slightly, from $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.3}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$ to $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{13.5}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$ for different decay modes.
Fig.5 shows the branching ratio $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{o}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{h}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{a}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ in the SM and models II and III, using the input parameters in Eq.(35) and assuming $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.3}$}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\lambda }$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{35}$}`$, and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\theta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{30}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$. The dots line in Fig.5 is the SM prediction $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{o}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{h}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{a}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2.49}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$. The short-dashed curve shows the the ratio in the model II, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{o}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{h}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{a}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2.98}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$ ($`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3.23}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$) for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{200}$}`$ (100) GeV and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{tan}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\beta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}`$. The long-dashed and solid curve show the theoretical predictions in the model III: $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{o}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{h}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{a}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4.67}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$ ($`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4.91}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$) for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{200}$}`$ GeV and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\theta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{30}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$, respectively. For the model III with $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{100}$}`$ GeV, one finds that $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{o}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{h}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{a}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{7.27}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$ ($`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{7.60}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$) for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\theta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{30}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$, respectively.
It is easy to see from Fig.5 and Table 1 that the new physics enhancement to the branching ratios of three-body charmless b quark decays in the model III is much larger than that in model II within the parameter space considered.
## IV. $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$
The ratio $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ is the average over weakly-decaying hadrons containing one b quark. For the CLEO experiments running on the $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{{\rm Y}}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ resonance, the average is over $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$ and their charge conjugate hadrons. For the experiments running on $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Z}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$ resonance, however, the average is over $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{N}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}`$<sup>3</sup><sup>3</sup>3$`\colorbox[rgb]{1,1,1}{$N$}_\colorbox[rgb]{1,1,1}{$b$}`$ is in turn the mixture of $`\colorbox[rgb]{1,1,1}{$\mathrm{\Lambda }$}_\colorbox[rgb]{1,1,1}{$b$}\colorbox[rgb]{1,1,1}{$($}\colorbox[rgb]{1,1,1}{$u$}\colorbox[rgb]{1,1,1}{$d$}\colorbox[rgb]{1,1,1}{$b$}\colorbox[rgb]{1,1,1}{$)$}`$, $`\colorbox[rgb]{1,1,1}{$\mathrm{\Sigma }$}_\colorbox[rgb]{1,1,1}{$b$}\colorbox[rgb]{1,1,1}{$($}\colorbox[rgb]{1,1,1}{$u$}\colorbox[rgb]{1,1,1}{$s$}\colorbox[rgb]{1,1,1}{$b$}\colorbox[rgb]{1,1,1}{$)$}`$, $`\colorbox[rgb]{1,1,1}{$\mathrm{\Xi }$}_\colorbox[rgb]{1,1,1}{$b$}\colorbox[rgb]{1,1,1}{$($}\colorbox[rgb]{1,1,1}{$d$}\colorbox[rgb]{1,1,1}{$s$}\colorbox[rgb]{1,1,1}{$b$}\colorbox[rgb]{1,1,1}{$)$}`$ and $`\colorbox[rgb]{1,1,1}{$\mathrm{\Omega }$}_\colorbox[rgb]{1,1,1}{$b$}\colorbox[rgb]{1,1,1}{$($}\colorbox[rgb]{1,1,1}{$s$}\colorbox[rgb]{1,1,1}{$s$}\colorbox[rgb]{1,1,1}{$b$}\colorbox[rgb]{1,1,1}{$)$}`$..
The charm multiplicity $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ is the average over the b-hadrons produced in the given environment. CLEO and LEP collaborations presented new measurements of inclusive $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ transitions that can be used to extract $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$. One naively expect $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1.15}$}`$ with the additional $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{15}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$ coming from the tree-level decay chain $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}`$. This expectation can be verified experimentally by adding all inclusive $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ branching ratios, and counting twice for the decay modes with 2 charm quarks in the final state.
In this section, we will investigate the new physics contributions, induced by the charged Higgs penguins in the models II and III, to the ratio $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ and the charm multiplicity $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$.
### A. $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$: experimental measurements
The $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ deficit was first point out in around 1994 when the theoretical prediction was considered to be difficult to produce $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{12}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$ while the 1995 CLEO data on $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{{\rm Y}}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ resonance was $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{10.49}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.46}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$ . In the following, we use the 1998 Particle Data Group value
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{10.45}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.21}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$ (71)
as the measured $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ on $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{{\rm Y}}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$.
For the experiments on the $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Z}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$-peak, all the four LEP collaborations reported their measured values of the ratio $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ as listed in Table 2. The seventh row shows the averaged result of the ratio $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ on $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Z}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$-peak <sup>4</sup><sup>4</sup>4 We here made an arithmetic average over four results as done in , but the newest L3 data has been used here in the average. : $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{10.66}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.17}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$. This $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}`$ on the $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Z}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$-peak can be converted to $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{{\rm Y}}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ value by multiplying a factor of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\tau }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\tau }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1.026}$}`$: $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{10.94}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.19}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$ ($`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Z}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$ corrected). In fact, there is still a $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\sigma }$}`$ discrepancy in ratio $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ between the high energy $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Z}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$ value and the low energy $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{{\rm Y}}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ value. The average of the $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Z}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{{\rm Y}}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ values of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ is
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{10.70}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.21}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{Overall}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{average}}$}`$ (72)
where we conservatively chose $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.21}$}`$ as the overall error of the measured $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$.
As for the charm counting, the value of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ measured at the $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{{\rm Y}}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ is still smaller than that measured at $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Z}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$-peak :
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\{}$}\begin{array}{cc}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1.10}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.05}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\hfill & \colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{{\rm Y}}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\hfill \\ \colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1.20}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.07}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\hfill & \colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Z}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{peak}}$}\hfill \end{array}`$ (75)
The average of the $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{{\rm Y}}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Z}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$ result leads to
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1.14}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.04}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Z}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{{\rm Y}}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ (76)
### B. $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$: theoretical predictions
Within the SM, the basis of the prediction for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ is the assumption of quark-hadron duality. The estimation for various inclusive decay rates is usually performed by using the heavy-quark expansion(HQE) and the perturbative QCD in the framework of operator product expansion. The HQE allows to relate the inclusive decay rate of B meson to that of the underlying $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}`$ quark decay process: $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{X}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{O}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$.
The theoretical prediction for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ with the inclusion of the $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{O}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\alpha }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ QCD corrections and the hadronic corrections to the free quark decay of order $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}`$ is currently available . The $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ can be defined as
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{l}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{l}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (77)
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{l}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{l}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (78)
where $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\tau }$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.25}$}`$, and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}}`$ ($`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}`$) is the rate of the decay mode $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}`$ ($`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}`$) where $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}`$ ($`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}`$) is the appropriate Cabibbo mixture of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}`$ quarks.
The $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}}`$ has been defined and calculated in last section. In the SM, we have
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.23}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.08}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (79)
where the error mainly comes from the uncertainties of the scale $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}`$ and the mass ratio $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}`$ .
As is well known, the main difficulty in calculating $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ is in the non-leptonic branching ratios $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}`$. For $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}}`$, a complete NLO calculation has been performed which gives
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4.0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.4}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (80)
where the error mainly comes from the uncertainties of the scale $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}`$, the quark masses $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}`$ and the assumption of quark-hadron duality . Furthermore, the error of the estimation for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}}`$ is generally considered to be larger than that for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}}`$. The enhancement of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}`$ due to large QCD corrections is about $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{30}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$ . Such enhancement will decrease the value of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$, but increase the size of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$.
Using the on-mass-shell sheme, the SM theoretical predictions for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ at the NLO level are
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{12.0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.7}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.5}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.2}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1.2}$}}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.9}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (81)
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1.24}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.05}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.01}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (82)
as given in Ref.<sup>5</sup><sup>5</sup>5The last and largest error of $`\colorbox[rgb]{1,1,1}{$B$}_{\colorbox[rgb]{1,1,1}{$S$}\colorbox[rgb]{1,1,1}{$L$}}`$ comes from the uncertainty of the renormalization scale $`\colorbox[rgb]{1,1,1}{$\mu $}`$; while the main error of $`\colorbox[rgb]{1,1,1}{$n$}_\colorbox[rgb]{1,1,1}{$c$}`$ is the the uncertainty in $`\colorbox[rgb]{1,1,1}{$m$}_\colorbox[rgb]{1,1,1}{$b$}`$ .; and
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\{}$}\begin{array}{cc}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{12.0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1.0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}\hfill & \colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\hfill \\ \colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{10.9}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1.0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}\hfill & \colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\hfill \end{array}`$ (85)
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\{}$}\begin{array}{cc}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1.20}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.06}$}\hfill & \colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\hfill \\ \colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1.21}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.06}$}\hfill & \colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\hfill \end{array}`$ (88)
as given in Ref. with the error mainly result from the variation of the scale $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}`$.
Comparing the observed and predicted values of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$, one can see that: (a) after considering all the corrections, the theoretical values of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ now come down and more or less consistent with the measurement, but unfortunately at the expense of boosting $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$; (b) the central value of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ in Ref. is higher than that in Ref., although two predictions are agree within errors; (c) there is still $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2.8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\sigma }$}`$ discrepancy between the $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ measured by CLEO and the theoretical prediction : $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1.10}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.05}$}`$ against $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1.24}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.05}$}`$
If we’d like to drop down the large uncertainty in the calculation for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}`$ decay mode, we can eliminate the ratio $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}`$ from the expression of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ and find,
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\left(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2.25}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\right)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ (89)
which is a linear correlation between $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$. Using the values for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ (72), $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}}`$ (80), and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}}`$ (79), one finds
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1.28}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.05}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}`$ (90)
for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{10.70}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.21}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$. The overall uncertainty of this prediction of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ should be smaller than that as given in Eqs.(82) and (88). The $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2.6}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\sigma }$}`$ discrepancy between the $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ in Eq.(90) and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ measured at $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{{\rm Y}}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ motivated proposals of new physics which will enhance $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}}`$ and in turn decrease $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$. That is what we try to do here.
As shown in Table 1, the ratio $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}}`$ will be increased significantly after taking the new physics effects into account, which will in turn decrease both $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ accordingly. From Eq.(89) and using the values for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ (72), $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}}`$ (80), and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}}`$ (63), one finds
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\{}$}\begin{array}{cc}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1.23}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.05}$}\hfill & \colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{for}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{M}}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{H}}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{200}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{G}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{e}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{V}}$}\hfill \\ \colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1.18}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.05}$}\hfill & \colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{for}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{M}}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{H}}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{100}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{G}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{e}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{V}}$}\hfill \end{array}`$ (93)
for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\theta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{30}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}`$. The $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}`$\- and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\theta }$}`$-dependence of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ is rather weak: the central value of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ will go down (up) by only $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.01}$}`$ for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mu }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{/}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}`$ ($`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\theta }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$). For $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ in the model III, the agreement between the prediction and the data will be improved slightly by a decrease $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.003}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.005}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{200}$}`$ ($`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{100}$}`$) GeV due to the inclusion of new physics contributions. In the model II, the resulted decrease for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ ($`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$) is only $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.01}$}`$ ($`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.001}$}`$) and plays no real role. Most importantly, one can see from Eqs.(75,76,90,93) that the predicted $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ and the measured $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ now agree within roughly one standard deviation after taking into account the effects of gluonic charged Higgs penguins in the model III with a relatively light charged Higgs boson, as illustrated in Fig.6.
## V. Summary and discussions
In the framework of the general two-Higgs doublet models, we calculated the charged-Higgs penguin contributions to (a) the rare radiative decay $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}`$; (b) the inclusive charmless decays $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}}`$ with $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\}}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\}}$}`$; (c) the charm multiplicity $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ and semileptonic branching ratio $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$.
In section II, we studied the experimental constraint on the model III from the CLEO data of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\gamma }$}`$ decay. With the help of previous works , we found the parameter space of the model III allowed by the available data, as shown in Eq.(37).
In section III, we firstly calculated the new physics contributions to the decay $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}`$ and found that the branching ratio $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ can be greatly enhanced from the SM prediction of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.27}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$ to $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2.34}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$ ($`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4.84}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$) in the model III for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{200}$}`$ ($`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{100}$}`$) GeV, as illustrated in Fig.3. Such a significant enhancement is clearly very helpful to resolve the missing charm/$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ problem appeared in B experiments.
Following the method of Ref., we then calculated the new physics contributions to three-body inclusive charmless decays of b quark due to the interference between the operators $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{6}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}`$. The Wilson coefficient $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{8}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{I}$}}`$ in Eq.(40) now describe the contributions from both the $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}`$ QCD penguins, the latter is the new physics part we focus in here. From numerical calculations. we found that: (a) the new physics enhancement to the decay $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}}`$ is only $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1.6}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$ since this mode is dominated by the tree diagrams; (b) the branching ratios of other five three-body b decay modes are strongly enhanced by the new charged Higgs penguins: $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{30}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$ to $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{70}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$ increase can be achieved within the considered parameter space. The new contributions to the corresponding branching ratios in the model II is, however, small in size and negative in sign against the theoretical predictions in the SM. As shown in Table 1 and Fig.5, the ratio $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{o}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{h}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{a}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$ can be increased from the SM prediction $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{o}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{h}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{a}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2.49}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$ to $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{o}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{h}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{a}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{4.91}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$ ($`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{7.60}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\%}$}`$) in the model III for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{200}$}`$ ($`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{100}$}`$) GeV.
In section IV, we studied the current status about the theoretical predictions and experimental measurements for the semileptonic branching ratio of B meson decay $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ and the charm multiplicity $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$, and calculated the new physics contributions, induced by the charged Higgs penguins in the model III (II), to both $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$. With an enhanced ratio $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{o}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{h}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{a}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{r}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}`$, both the $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ will be decreased accordingly: (a) the central value of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ can be decreased slightly by $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.003}$}`$ ($`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.005}$}`$) for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{200}$}`$ ($`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{100}$}`$) GeV; (b) the value of $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ can be lowered significantly from the prediction $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1.28}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.05}$}`$ in the SM to $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1.23}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.05}$}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1.18}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pm }$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0.05}$}`$ for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{M}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{H}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{200}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{100}$}`$ GeV, respectively.
In short, the predicted $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ and the measured $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}`$ now agree within roughly one standard deviation after taking into account the effects of gluonic charged Higgs penguins in the model III with a relatively light charged Higgs boson, while the agreement between the theoretical prediction and the data for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{B}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{S}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{L}$}}`$ can also be improved by inclusion of new physics effects.
## ACKNOWLEDGMENTS
C.S. Li and K.T. Chao acknowledge the support by the National Natural Science Foundation of China, the State Commission of Science and technology of China and the Doctoral Program Foundation of Institution of Higher Education. Z.J. Xiao acknowledges the support by the National Natural Science Foundation of China under the Grant No.19575015 and the Outstanding Young Teacher Foundation of the Education Ministry of China.
## Appendix: RS independent $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{p}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}`$
For the convenience of the reader, we here present the explicit expressions of the RS independent NLO corrections $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}}`$, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{p}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}`$. For more details one can see the original paper .
The term $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}}`$ in Eq.(50) describes the current-current type corrections proportional to $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}`$
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{F}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{5}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{32}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pi }$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{v}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}{\displaystyle \underset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}}{\overset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\left[}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{h}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}{\displaystyle \underset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{k}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}}{\overset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{J}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{k}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{k}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\right]}$}`$ (94)
with $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}`$ for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$ for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}`$, and the coefficients $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{h}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{J}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{k}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}}`$ can be found in .
The term $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{p}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}}`$ in Eq.(50) describes the effect of penguin diagrams involving $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{Q}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}`$ ,
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{p}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{n}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{F}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{5}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{32}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pi }$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{Re}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{[}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}{\displaystyle \underset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{v}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{[}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{v}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{v}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{]}$}`$ (95)
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}{\displaystyle \underset{\genfrac{}{}{0pt}{}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{6}$}}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{v}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{[}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{v}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{x}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{c}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{v}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{]}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{]}$}`$
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{Re}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\left[}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{v}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{v}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}{\displaystyle \underset{\genfrac{}{}{0pt}{}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{k}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{6}$}}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{J}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{k}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{k}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\xi }$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}{\displaystyle \underset{\genfrac{}{}{0pt}{}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{k}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{6}$}}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{J}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{k}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{k}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\right]}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{.}$}`$
with $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}`$ for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$ for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}`$. The explicit expressions of coefficients $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{g}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{J}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{k}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}}`$ can be found in .
Finally, $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}`$ is given by
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Delta }}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{\Gamma }}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{W}$}`$ $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}`$ $`{\displaystyle \frac{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{G}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{F}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{m}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{5}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{32}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\pi }$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{[}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}{\displaystyle \underset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}}{\overset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{v}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\left[}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\right]}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}{\displaystyle \underset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}}{\overset{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{6}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{v}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{|}$}^\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\left[}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\right]}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}`$ (96)
$`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}{\displaystyle \underset{\genfrac{}{}{0pt}{}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{2}$}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{3}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\mathrm{}}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{6}$}}}{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{[}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{+}$}\overline{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{C}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{)}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{]}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{R}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{e}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\left(}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{v}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}^{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{v}$}_\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{\right)}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{]}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{.}$}`$
where $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{1}$}`$ for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{u}$}`$ and $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{t}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{0}$}`$ for $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{q}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{=}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{d}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{,}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{s}$}`$, the $`\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{b}$}_{\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{i}$}\colorbox[rgb]{1,1,1}{$\textcolor[rgb]{0,0,0}{j}$}}`$ have been given in Eq.(60).
|
warning/0003/hep-ph0003301.html
|
ar5iv
|
text
|
# Day - Night Effect Predictions for the SNO Detector
preprint: Ref. SISSA 30/2000/EP March, 2000
Detailed predictions for the day-night (D-N) asymmetry in the energy-integrated one year signals in the SNO detector in the case of the MSW $`\nu _e\nu _{\mu (\tau )}`$ and/or $`\nu _e\nu _s`$ transition solutions of the solar neutrino problem are presented. The asymmetries in the charged current (CC) and $`\nu e^{}`$ elastic scattering (ES) event rates are calculated for both MSW solutions; in the case of the $`\nu _e\nu _s`$ transition solution the D-N asymmetry in the neutral current (NC) event rate are derived as well. The asymmetries are calculated for three night samples of events which are produced by the solar neutrinos crossing i) the Earth mantle only (Mantle), ii) the Earth core (Core) and iii) the Earth core and/or the mantle (Night). The effects of the uncertainties i) in the values of the cross-sections of the CC and NC neutrino-induced reactions on deuterium, and ii) in the value of the bulk matter density and/or the chemical composition of the Earth core, on the corresponding D-N asymmetry predictions are analyzed. It is shown, in particular, that that due to the strong enhancement of the transitions of the solar neutrinos crossing the Earth core, at $`\mathrm{sin}^22\theta _V0.01`$ the corresponding one year average D-N asymmetry in the Core sample of CC events in the case of the $`\nu _e\nu _{\mu (\tau )}`$ solution can be larger by a factor of up to $`8`$ than the asymmetry in the Night sample. In certain subregions of the MSW solution regions at small $`\mathrm{sin}^22\theta _V`$, the predicted magnitude of the Core D-N asymmetry in the CC sample is very sensitive to the value of the electron fraction number in the Earth core. Iso - (D-N) asymmetry contours in the $`\mathrm{\Delta }m^2\mathrm{sin}^22\theta _V`$ plane for the SNO detector are derived in the region $`\mathrm{sin}^22\theta _V\stackrel{_>}{_{}}10^4`$ for the Core and Night samples of the CC, ES and NC events. The dependence of the D-N asymmetries considered on the final state e<sup>-</sup> threshold energy in the CC and ES reactions is also investigated. Our results show, in particular, that the SNO experiment will be able to probe substantial parts of the SMA and LMA MSW $`\nu _e\nu _{\mu (\tau )}`$ solution regions be performing Night and Core D-N asymmetry measurements.
I. Introduction
In spite of the remarkable progress made in the studies of solar neutrinos the true cause of the solar neutrino deficit observed in the $`ClAr`$, Kamiokande, $`GaGe`$, and Super-Kamiokande experiments is still not identified. The existing helioseismological data and its interpretation make very unlikely the possibility of an astrophysical origin of the discrepancy between the solar neutrino observations and the standard solar model predictions . The hypotheses of vacuum oscillations or MSW transitions of solar neutrinos continue to provide viable solutions of the solar neutrino problem (see, e.g., ).
The current (mean event rate) solar neutrino data admit three types of MSW solutions if the solar $`\nu _e`$ undergo two-neutrino transitions into active neutrinos, $`\nu _e\nu _{\mu (\tau )}`$: the well-known small mixing angle (SMA) non-adiabatic and large mixing angle (LMA) adiabatic (see, e.g., ) and the so-called “LOW” solution (very recent analyses can be found in, e.g., ). While the SMA and LMA solutions have been shown to be rather stable with respect to variations in the values of the various physical quantities which enter into the calculations (the fluxes of <sup>8</sup>B and <sup>7</sup>Be neutrinos, nuclear reaction cross-sections, etc.), and of the data utilized in the analyses, the LOW solution is of the “borderline” type: its existence even at 99% C.L. is not stable with respect to relatively small changes in the data and/or in the relevant theoretical predictions (see, e.g., and the references quoted therein.). To the three solutions there correspond three distinct regions in the plane of values of the two parameters, $`\mathrm{\Delta }m^2`$ and $`\mathrm{sin}^22\theta `$, characterizing the transitions. One finds using the standard solar model predictions for the solar neutrino fluxes (<sup>8</sup>B, <sup>7</sup>Be, $`pp`$, etc.) that at 99% C.L. the SMA MSW solution requires values in the intervals $`4.0\times 10^6\mathrm{eV}^2<\mathrm{\Delta }m^2<10.0\times 10^6\mathrm{eV}^2`$, $`1.3\times 10^3<\mathrm{sin}^22\theta _V<1.0\times 10^2`$, the LMA solutions is realized for $`\mathrm{\Delta }m^2`$ and $`\mathrm{sin}^22\theta _V`$ from the region $`7.0\times 10^6\mathrm{eV}^2<\mathrm{\Delta }m^2<2.0\times 10^4\mathrm{eV}^2`$, $`0.50<\mathrm{sin}^22\theta _V<1.0`$, and the LOW solution lies approximately in the region $`0.4\times 10^7\mathrm{eV}^2<\mathrm{\Delta }m^2<1.5\times 10^7\mathrm{eV}^2`$, $`0.80<\mathrm{sin}^22\theta _V<1.0`$. The SMA and LMA solution regions expand in the direction of smaller values of $`\mathrm{sin}^22\theta _V`$ up to $`0.6\times 10^3`$ and to $`0.3`$, respectively, if one adopts more conservative approach in analyzing the data in terms of the MSW effect and treats the <sup>8</sup>B neutrino flux as a free parameter in the analysis (see, e.g., ). This enlargement of the two solution regions is not uniform in $`\mathrm{\Delta }m^2`$ (see further). We will use the term “conservative” for the MSW solution regions derived by treating the <sup>8</sup>B neutrino flux as a free parameter.
The solar neutrino data can also be explained assuming that the solar neutrinos undergo MSW transitions into sterile neutrino in the Sun: $`\nu _e\nu _s`$ (see, e.g, ). In this case only a SMA solution is compatible with the data. The corresponding solution region obtained (at a given C.L.) using the mean event rate solar neutrino data and the predictions of ref. for the different solar neutrino flux components practically coincides in magnitude and shape with the SMA $`\nu _e\nu _{\mu (\tau )}`$ solution region, but is shifted by a factor of $`1.2`$ along the $`\mathrm{\Delta }m^2`$ axis to smaller values of $`\mathrm{\Delta }m^2`$. The “conservative” $`\nu _e\nu _s`$ transition solution region extends both in the direction of smaller and larger values of $`\mathrm{sin}^22\theta _V`$ down to $`0.7\times 10^3`$ and up to 0.4 .
The results on the spectrum of the final state $`e^{}`$ from the $`\nu e^{}`$ elastic scattering reaction, obtained in the range of the recoil$`e^{}`$ energy $`E_e(520)`$MeV by the Super-Kamiokande collaboration do not allow to rule out definitely any of the solutions of the solar neutrino problem indicated above . The observed rise of the spectrum at $`E_e>12`$MeV can be due to a contribution from the $`hep`$ solar neutrinos , which complicates the interpretation of the data. The spectrum measured below 12 MeV is compatible with an energy-independent suppression of the <sup>8</sup>B neutrino flux, or with a mild energy dependence of the suppression.
A unique testable prediction of the MSW solutions of the solar neutrino problem is the day-night (D-N) effect - a difference between the solar neutrino event rates during the day and during the night, caused by the additional transitions of the solar neutrinos taking place at night while the neutrinos cross the Earth on the way to the detector (see, e.g., and the references quoted therein). The experimental observation of a non-zero D-N asymmetry
$$A_{DN}^N\frac{R_NR_D}{(R_N+R_D)/2},$$
(1)
where $`R_N`$ and $`R_D`$ are, e.g., the one year averaged event rates in a given detector respectively during the night and the day, would be a very strong evidence in favor (if not a proof) of the MSW solution of the solar neutrino problem (see further).
Extensive predictions for the magnitude of the D-N effect for the Super-Kamiokande detector have been obtained in . High precision calculations of the D-N asymmetry in the one year averaged recoil-e<sup>-</sup> spectrum and in the energy-integrated event rates were performed for three event samples, Night, Core and Mantle, in . The night fractions of these event samples are due to neutrinos which respectively cross the Earth along any trajectory, cross the Earth core, and cross only the Earth mantle (but not the core), on the way to the detector. The measurement of the D-N asymmetry in the Core sample was found to be of particular importance because of the strong enhancement of the asymmetry, caused by a constructive interference between the amplitudes of the neutrino transitions in the Earth mantle and in the Earth core . The effect differs from the MSW one . The mantle-core enhancement effect is caused by the existence (for a given neutrino trajectory through the Earth core) of points of resonance-like total neutrino conversion in the corresponding space of neutrino oscillation parameters . The location of these points determines the regions where the relevant probability of transitions in the Earth of the Earth-core-crossing solar neutrinos is large Being a constructive interference effect between the amplitudes of neutrino transitions in the mantle and in the core, this is not just “core enhancement” effect, but rather mantle-core enhancement effect. . At small mixing angles and in the case of $`\nu _e\nu _{\mu (\tau )}`$ transitions the predicted D-N asymmetry in the Core sample of the Super-Kamiokande data was shown to be much bigger due to the mantle-core enhancement effect The term “neutrino oscillation length resonance” (NOLR) was used in , in particular, to denote the mantle-core enhancement effect in this case. \- by a factor of up to $`6`$, than the asymmetry in the Night sample. The asymmetry in the Mantle sample was found to be smaller than the asymmetry in the Night sample. On the basis of these results it was concluded in that it can be possible to test a substantial part of the MSW $`\nu _e\nu _{\mu (\tau )}`$ SMA solution region in the $`\mathrm{\Delta }m^2\mathrm{sin}^22\theta _V`$ plane by performing selective, i.e., Core D-N asymmetry measurements.
The current Super-Kamiokande data shows a D-N asymmetry in the Night sample, which is different from zero at 1.9 s.d. level:
$$A_{DN}^N=0.065\pm 0.031(stat.)\pm 0.013(syst.).$$
(2)
These data do not allow to test the SMA solution region: the predicted asymmetry is too small (see, e.g., ). However, the Super-Kamiokande night data is given in 5 bins and 80% of the events in the bin N5 are due to Earth-core-crossing solar neutrinos , while the remaining 20% are produced by neutrinos which cross only the Earth mantle. Since the predicted D-N asymmetry in the Mantle sample is practically negligible in the case of the MSW SMA solution of interest , we have for the D-N asymmetry measured using the night N5 bin data: $`A_{DN}^{N5}0.8A_{DN}^C`$, $`A_{DN}^C`$ being the asymmetry in the Core sample. The data on $`A_{DN}^{N5}`$ permitted to exclude a part of the MSW SMA solution region located in the area $`\mathrm{sin}^22\theta _V(0.0070.01)`$, $`\mathrm{\Delta }m^2(0.51.0)\times 10^5\mathrm{eV}^2`$. Let us note that it was possible to probe the indicated MSW SMA solution region only due to the mantle-core enhancement of the the Core part of the asymmetry $`A_{DN}^{N5}`$. It should be obvious from the above discussion that the measurement of the Core asymmetry $`A_{DN}^C`$, as suggested in , will provide a more effective test of the the MSW SMA solution <sup>§</sup><sup>§</sup>§We were happy to learn that recently the Super-Kamiokande collaboration has decided to present experimental results on the Core D-N asymmetry as well. than the measurement of $`A_{DN}^{N5}`$.
As was shown in , the mantle-core enhancement is not so effective in the <sup>8</sup>B neutrino energy interval $`E(512)`$MeV when the solar neutrino transitions are of the type $`\nu _e\nu _s`$: it does not produce such a dramatic enhancement of the Core D-N asymmetry as in the case of $`\nu _e\nu _{\mu (\tau )}`$ transitions. More generally, the measurements of the D-N effect related observables in the Super-Kamiokande experiment appears at present unlikely to provide an effective test of the MSW $`\nu _e\nu _s`$ (SMA) solution of the solar neutrino problem .
The studies of the D-N effect will be continued by the SNO experiment which is already operational . As is well known, the solar neutrinos can be detected in SNO via the charged current (CC) reaction on deuterium,
$$\nu _e+De^{}+p+p,(CC)$$
(3)
deuterium break up by neutrinos via the neutral current (NC) weak interaction,
$$\nu +D\nu +n+p,(NC)$$
(4)
and via the elastic scattering (ES) on electrons,
$$\nu +e^{}\nu +e^{}.(ES)$$
(5)
In the present article we derive detailed predictions for the D-N asymmetries in the CC, NC and ES event rates for the SNO detector. For the <sup>8</sup>B neutrino flux, $`\mathrm{\Phi }(B)`$, which at the Earth surface represents 47.5% of the flux predicted by the standard solar model , $`\mathrm{\Phi }_{SSM}(B)`$, the expected number of events per year due to the CC reaction $`\nu _e+De^{}+p+p`$ in the SNO detector is about 3300 for $`T_{e,th}=5`$ MeV, $`T_{e,th}`$ being the threshold kinetic energy of the final state $`e^{}`$. This is comparable to the event rate in the Super - Kamiokande experiment. The expected event rate due to the NC reaction $`\nu _l+D\nu _l+n+p`$ for a full standard solar model <sup>8</sup>B neutrino flux (and 45% detection efficiency) is approximately (5.5 - 6.0) events per day, while the event rate due to the $`\nu e^{}`$ elastic scattering (ES) reaction is predicted to be $`1.4`$ events per day for $`\mathrm{\Phi }(B)0.47\mathrm{\Phi }_{SSM}(B)`$ and $`T_{e,th}=5`$ MeV. The indicated statistics permits to perform a high precision search for the D-N effect and for the Earth mantle-core enhancement of the effect in the sample of events due to the CC reaction. Therefore the main emphasis of our study will be on the predictions for the magnitude of the D-N effect in the CC sample of events in the SNO detector, although rather detailed results for the expected magnitude of the effect in the ES and NC samples will also be presented. Let us remind the reader that the D-N asymmetry in the NC event rate can be nonzero in the case of the MSW solution of the solar neutrino problem with transitions into sterile neutrino .
Predictions for the D-N effect for the SNO detector were derived in refs. and recently in . However, our study overlaps little with those performed in .
2. Calculating the D-N Effect for the SNO Detector
2.1. The CC and NC Reaction Cross Section Uncertainties
The high precision calculations of the D-N effect for the SNO detector performed in the present article are based on the methods developed for our studies of the D-N effect for the Super-Kamiokande detector, which are described in detail in . The cross sections of the CC and NC reactions on deuterium were taken from . They were tabulated using the 1996 version of the E. Lisi $`\nu \mathrm{d}`$ FORTRAN library . The cross section of the $`\nu _ee^{}`$ elastic scattering reaction was taken from . We used in our calculations the <sup>8</sup>B neutrino spectrum derived in . The probability of survival of the solar $`\nu _e`$ when they travel in the Sun and further to the surface of the Earth, $`\overline{P}_{}(\nu _e\nu _e)`$, was computed on the basis of the analytic expression obtained in and using the method developed in . In the calculation of $`\overline{P}_{}(\nu _e\nu _e)`$ we have utilized (as in ) the density profile of the Sun and the <sup>8</sup>B neutrino production distribution in the Sun, predicted in . These predictions were updated in , but a detailed test study showed (see also ) that using the results of instead of those in leads to a change of the survival probability by less than 1% for any set of values of the parameters $`\mathrm{\Delta }m^2`$ and $`\mathrm{sin}^22\theta `$, relevant for the calculation of the D-N effect. As in , the Earth matter effects were calculated using the Stacey model from 1977 for the Earth density distribution. The latter practically coincides with that predicted by the more recent Earth model . Most of the results are obtained for two sets of values of the electron fraction number in the Earth mantle and the Earth core, $`Y_e(man)`$ and $`Y_e(core)Y_e^c`$, which reflect the chemical composition of the two major Earth structures: the standard ones $`Y_e(man)=0.49`$, $`Y_e^c=0.467`$, and for the possibly conservative upper bound of $`Y_e^c=0.500`$ (see, e.g, ). The latter effectively accounts for the uncertainties both in $`Y_e^c`$ and in the average matter density of the Earth core.
Three different calculations of the cross section of the $`\nu _eD`$ CC reaction are available in the literature . In two of these articles the cross section of the $`\nu _lD`$ NC reaction was also computed. Extensive numerical study has shown that i) the maximal differences between the various one year averaged D-N asymmetries in the CC event rate, calculated using the three different predictions for the CC cross section, do not exceed 6%, ii) the differences in the range of (1 - 6)% occur only in a set of disconnected regions in the $`\mathrm{\Delta }m^2\mathrm{sin}^22\theta _V`$ plane, which are essentially point-like and whose total area is exceedingly small, and iii) these small point-like regions of maximal spread in the D-N asymmetry predictions are generally located outside and sufficiently far from the MSW solution regions of interest (including the “conservative” ones). There is only one case when such a difference occurs on the borderline of an MSW “conservative” solution region: we found a spread in the prediction for D-N asymmetry in the Core sample of events of $`2.5\%`$ in the case of $`\nu _e\nu _{\mu (\tau )}`$ transitions for $`\mathrm{sin}^22\theta 0.6`$ and $`\mathrm{\Delta }m^23\times 10^5`$ eV<sup>2</sup>, which is at the border of the LMA “conservative” solution region.
The cross sections of the $`\nu _eD`$ CC reaction calculated in differ by $``$ (5 - 15)% at the <sup>8</sup>B neutrino energies of interest. The question of why the predicted D-N asymmetries in the CC and NC event rates are so insensitive to the choice of the cross section used in the calculations naturally arises. The product of the normalized to one <sup>8</sup>B neutrino spectrum, $`𝒮(E)`$, and of the CC reaction cross section, $`\sigma _{CC}(E)`$, which enters into the expression for the CC event rate, can be approximated by a resonance-like function defined by its full width at half maximum, the position of the maximum in $`E`$ and by its value at the maximum. These parameters were calculated for the three different cross sections and for different values of the threshold kinetic energy of the final state electron, $`T_{e,th}`$. The results are given in Table I. The largest differences between the three results for $`𝒮(E)\sigma _{CC}(E)`$ happen in the fastly decreasing part of the <sup>8</sup>B neutrino spectrum ($`E10`$ MeV). As Table I shows, the spread in the values at the maximum is between 3.5% and 10%, while the position of the maximum in $`E`$ and the F.W.H.M. change at most by $`0.8\%`$. Consequently, the differences between the three CC cross sections lead essentially only to a non-negligible change in the prediction for the overall normalization factor for the total CC event rate, which practically does not affect the D-N asymmetry. The same conclusions can be shown to be valid for the implications of the differences between the two NC cross sections for the uncertainties in the predicted values of the D-N asymmetries in the NC signal in SNO in the case of MSW $`\nu _e\nu _s`$ transition solution of the solar neutrino problem. These results are illustrated in Figs. 1a - 1f, where the difference between the values of the Core (Night) D-N asymmetry calculated utilizing the predictions for the relevant CC and NC reaction cross-sections derived in ref. and in ref. are shown as functions of $`\mathrm{\Delta }m^2`$ and $`\mathrm{sin}^22\theta _V`$ in the intervals ($`10^710^4`$) eV<sup>2</sup> and ($`10^41.0`$), respectively However, all the plots produced in the framework of this study are at disposition of the interested reader at the site http://www.pv.infn.it/$``$maris/nue/sno/index.html..
This analysis permits one to conclude that even neglecting the electron energy resolution function and the expected statistical and systematical errors in the measurements, the calculation of the D-N asymmetries in the CC and NC event rates in the SNO detector utilizing the different cross sections from should give the same results within 1% in the MSW “conservative” solution regions. Accordingly, we will present predictions for the D-N asymmetries in the CC and NC signals computed using the cross sections of ref. .
2.2 Comments on the Probability of Solar $`\nu _e`$ Survival
The SNO detector is located at $`\lambda _{\mathrm{SNO}}=46{}_{}{}^{}\mathrm{\hspace{0.17em}20}_{}^{}`$ North, $`\lambda _{SNO}^{}=81{}_{}{}^{}\mathrm{\hspace{0.17em}12}_{}^{}`$ West. At the indicated latitude we have: $`T_{res}^C/T_y=0.0396`$, where $`T_{res}^C`$ is the Core residence time, i.e., the total time in one year during which the solar neutrinos cross the Earth core at night on the way to the SNO detector, $`T_y=365.24`$ days. This implies that at the location of the SNO experiment only for $`7.9\%`$ of the one year night time the Sun is behind the Earth core with respect to SNO, to be compared with $`14.85\%`$ for the Super - Kamiokande detector (see, e.g., ). Thus, the statistical error in the measurement of the Core D-N asymmetry with SNO detector will be approximately by a factor of 3.55 bigger than in the measurement of the Night D-N asymmetry. In the case of the Super - Kamiokande detector this factor is 2.66. Let us note also that for the Mantle residence time, $`T_{res}^M`$, i.e., the total time in one year during which the solar neutrinos cross at night the Earth mantle but do not cross the Earth core on the way to the SNO detector, we have $`T_{res}^M/T_y=0.4619`$, and that the Night residence time is $`T_{res}^N=0.5015T_y`$.
As is well-known, one can use the $`\nu _2\nu _e`$ transition probability, $`P_{e2}`$, to account for the Earth matter effects in the solar $`\nu _e`$ survival probability, $`\overline{P}(\nu _e\nu _e)`$, in the case of the MSW solutions of the solar neutrino problem. During the day one has $`\overline{P}(\nu _e\nu _e)=\overline{P}_{}(\nu _e\nu _e)`$, where $`\overline{P}_{}(\nu _e\nu _e)`$ was defined earlier as the probability of survival of the solar $`\nu _e`$ when they travel to the Earth surface without traversing the Earth. A nonzero difference between $`\overline{P}(\nu _e\nu _e)`$ for the Earth-crossing neutrinos and $`\overline{P}_{}(\nu _e\nu _e)`$ produces a nonzero D-N asymmetry. It is possible to have $`\overline{P}(\nu _e\nu _e)\overline{P}_{}(\nu _e\nu _e)0`$ when i) $`\overline{P}_{}(\nu _e\nu _e)0.5`$ and ii) $`P_{e2}`$ differs from its value in vacuum, $`P_{e2}\mathrm{sin}^2\theta _V`$ (for further details see, e.g., ).
We have calculated the one year averaged $`\nu _2\nu _e`$ transition probability for the location of the SNO detector, $`<P_{e2}>`$, using the methods described in . As in , this was done for each of the three different event samples Core, Night and Mantle we are considering, $`<P_{e2}>^{C,N,M}`$. We comment below only on few specific features of the probabilities $`<P_{e2}>^{C,N,M}`$ and of the corresponding solar $`\nu _e`$ survival probabilities at night, $`\overline{P}^{C,N,M}(\nu _e\nu _e)`$.
2.2.1 $`\nu _e\nu _{\mu (\tau )}`$ Transitions
At small mixing angles, $`\mathrm{sin}^22\theta _V<0.01`$, $`<P_{e2}>^C`$ is considerably larger than $`<P_{e2}>^{N,M}`$ in most of the interval of values of $`E_\nu /\mathrm{\Delta }m^2`$ of interest: the absolute maximum of $`<P_{e2}>^C`$, for instance, exceeds the absolute maxima of $`<P_{e2}>^{N,M}`$ by a factor of $`(45)`$. This is illustrated in Figs. 2a - 2b. The strong resonance-like enhancement of $`<P_{e2}>^C`$ with respect to $`<P_{e2}>^{N,M}`$, seen, in particular, in Figs. 2a - 2b, is The plots of $`<P_{e2}>^{C,N,M}`$ as functions of $`\mathrm{\Delta }m^2/E`$ at fixed $`\mathrm{sin}^22\theta _V<0.01`$ are very similar to those for the Super-Kamiokande detector, published in ref. which includes a large set of such plots. due to a constructive interference between the amplitudes of the neutrino transitions in the Earth mantle and in the core . The interference produces points of total neutrino conversion , $`\mathrm{max}(P_{e2})=1`$. For each fixed $`\widehat{h}`$ from the interval $`\widehat{h}=(0^{}30^{})`$, there exists one such point in the region $`\mathrm{sin}^22\theta _V0.10`$, located in the interval $`\mathrm{sin}^22\theta _V(0.040.08)`$ (see Table 4 in ). At $`\widehat{h}=23^{}`$, for instance, we have $`\mathrm{max}(P_{e2})=1`$ at $`\mathrm{sin}^22\theta _V=0.06`$ and $`\mathrm{\Delta }m^2/E=6.5\times 10^7\mathrm{eV}^2/\mathrm{MeV}`$ <sup>\**</sup><sup>\**</sup>\** The points under discussion change somewhat their position with the change of $`\widehat{h}`$ (see Table 4 in ).. The probability $`P_{e2}`$ is enhanced at each given $`\widehat{h}30^{}`$ in a sufficiently wide region in the neighborhood of the corresponding point of total neutrino conversion, which causes the strong enhancement of $`<P_{e2}>^C`$ under discussion at $`\mathrm{sin}^22\theta _V<0.01`$.
For $`\mathrm{sin}^22\theta _V<0.004`$, $`<P_{e2}>^{C,N,M}`$ are nonnegligible in an interval of values of $`E/\mathrm{\Delta }m^2`$ where $`\overline{P}_{}(\nu _e\nu _e)>0.5`$ and the D-N asymmetries (defined as in eq. (1)) for all the three samples of events Core, Night and Mantle, $`A_{DN}^{C,N,M}`$, are negative. For the values of $`\mathrm{\Delta }m^2`$ from the SMA solution region and the neutrino energies of interest, $`E(5.014.4)`$MeV, the core asymmetry $`A_{DN}^C`$ goes through zero in the interval of $`\mathrm{sin}^22\theta _V(0.0040.006)`$. For each given value of $`\mathrm{sin}^22\theta _V`$ from this interval the zero D-N effect line $`\overline{P}_{}(\nu _e\nu _e)=0.5`$ taking place at a specific value of $`E/\mathrm{\Delta }m^2`$, $`(E/\mathrm{\Delta }m^2)_0`$, splits $`<P_{2e}>^C`$ into two parts: the part at $`E/\mathrm{\Delta }m^2>(E/\mathrm{\Delta }m^2)_0`$ generates a negative D-N asymmetry, while the part at $`E/\mathrm{\Delta }m^2<(E/\mathrm{\Delta }m^2)_0`$ creates a positive D-N asymmetry. When integrating over the neutrino energy the positive and the negative contributions to $`A_{DN}^C`$ tend to mutually cancel. For each given $`\mathrm{sin}^22\theta _V(0.0040.006)`$ there exists a $`\mathrm{\Delta }m^2`$ from the SMA solution region for which this cancellation is exact giving $`A_{DN}^C=0`$. In the case of the Night asymmetry, the interval in $`\mathrm{sin}^22\theta _V`$ in which one has $`A_{DN}^N=0`$ for a certain value of $`\mathrm{\Delta }m^2`$ from the SMA solution region is larger: $`\mathrm{sin}^22\theta _V(0.0040.008)`$. For $`\mathrm{sin}^22\theta _V>0.006(0.008)`$ we have $`A_{DN}^C>0`$ ($`A_{DN}^N>0`$).
As in , one can introduce the probability D-N asymmetries,
$$A_P^{C,N,M}=\frac{\overline{P}^{C,N,M}(\nu _e\nu _e)\overline{P}_{}(\nu _e\nu _e)}{0.5(\overline{P}^{C,N,M}(\nu _e\nu _e)+\overline{P}_{}(\nu _e\nu _e))},$$
(6)
which give an indication about the possible magnitude of the corresponding energy-integrated event rate asymmetries. For $`\mathrm{sin}^22\theta _V<0.004`$, $`A_P^{C,N}`$ are essentially negative and we have $`|A_P^C|<3.0\%;4.7\%`$, while $`|A_P^N|<1\%`$. At $`\mathrm{sin}^22\theta _V=0.004`$, $`A_P^C`$ and $`A_P^N`$ as functions of $`E/\mathrm{\Delta }m^2`$ have substantial positive and negative parts and $`4\%<A_P^C<5.0\%`$, $`1.5\%<A_P^N<3.5\%`$. As $`\mathrm{sin}^22\theta _V`$ increases beyond the zero asymmetry region, $`A_P^C`$ increases steadily while the increase of $`A_P^N`$ is rather slow. At $`\mathrm{sin}^22\theta _V=0.008;0.010`$, for instance, $`A_P^C`$ reaches the values of $`26\%;50\%`$ and $`A_P^N<2.7\%;8.0\%`$. This corresponds to an enhancement factors of the Core asymmetry at the maxima of $`9.6;6.3`$. The above results suggests that for values of $`\mathrm{sin}^22\theta _V`$ from the SMA solution region, the Core event rate asymmetry in SNO can be bigger than the Night asymmetry by factors which can exceed the similar factors for the predicted Core and Night asymmetries in the Super-Kamiokande detector: in the latter case the Core asymmetry enhancement factor does not exceed $`6`$ .
It is interesting to note that for $`\mathrm{sin}^22\theta _V(410)\times 10^3`$, $`A_P^{C,N}`$ have relatively large positive values at $`E/\mathrm{\Delta }m^2few\times 10^5\mathrm{MeV}/\mathrm{eV}^2`$. In the same region $`A_P^{C,N}`$ exhibit (as functions of $`E/\mathrm{\Delta }m^2`$) fast small amplitude oscillations as well. These maxima of $`A_P^{C,N}`$ are responsible for the noticeable “horn” at $`\mathrm{sin}^22\theta _V5\times 10^3`$ and $`\mathrm{\Delta }m^22\times 10^5`$ eV<sup>2</sup>, shown by the CC and NC D-N asymmetry contour plots in the case of the $`\nu _e\nu _{\mu (\tau )}`$ and $`\nu _e\nu _s`$ transitions, as well as by the ES D-N asymmetry contour plots for $`\nu _e\nu _s`$ transitions (see further). The lack of a similar “horn” in the ES D-N asymmetry contour plots in the $`\nu _e\nu _{\mu (\tau )}`$ case is due to the neutral current contribution of $`\nu _{\mu (\tau )}`$ to the ES event rate. The indicated feature of $`A_P^{C,N}`$ affects the D-N asymmetries of interest only for a set of neutrino parameters which lie outside the regions of the MSW solutions and we are not going to discuss it further.
The mantle-core enhancement of $`A_P^C`$ is much less dramatic in the region of the LMA solution: actually, in this region we typically have $`A_P^C(1.01.25)A_P^N`$, with $`A_P^C`$ exhibiting as a function of $`E/\mathrm{\Delta }m^2`$ rather large amplitude oscillations which are absent in $`A_P^N`$.
2.2.2 $`\nu _e\nu _s`$ Transitions
The Earth mantle-core enhancement of $`P_{e2}`$ at small mixing angles is operative in the case of $`\nu _e\nu _s`$ transitions as well . For $`\mathrm{sin}^22\theta _V(0.11.0)\times 10^2`$ the corresponding absolute maximum of the one-year averaged Core probability $`<P_{e2}>^C`$ exceeds the absolute maximum of $`<P_{e2}>^N`$ by a factor of $`4`$. However, at $`\mathrm{sin}^22\theta _V0.01`$, $`<P_{e2}>^C`$ is noticeably smaller in the case of $`\nu _e\nu _s`$ transitions of the solar neutrinos than in the case of the $`\nu _e\nu _{\mu (\tau )}`$ transitions considered above (for illustration see the corresponding plots for the Super-Kamiokande detector in ). For $`\mathrm{sin}^22\theta _V=0.01`$, for instance, the maximal values reached by $`<P_{e2}>^C`$ in the two cases are $`0.1`$ and $`0.3`$, respectively. Similar result was shown to be valid for $`<P_{e2}>^C`$ calculated for the location of the Super-Kamiokande detector. The explanation of this difference can be found in : it is related to the fact that the total neutrino conversion point at $`\mathrm{sin}^22\theta _V0.25`$, which determines the enhancement under discussion, takes place for the different $`\widehat{h}`$ at $`\mathrm{sin}^22\theta _V(0.120.25)`$, which is relatively “far” in $`\mathrm{sin}^22\theta _V`$ from the region of interest (see Table 4 in ). The primary source is, of course, the difference between the neutrino effective potentials in matter in the two cases (see also, e.g., ).
For $`\mathrm{sin}^22\theta _V(0.0090.010)`$, the probability D-N asymmetries for the Core and Night event samples have (as functions of $`E/\mathrm{\Delta }m^2`$) substantial negative parts in addition to the positive ones. For most of the indicated values of $`\mathrm{sin}^22\theta _V`$ and the values of $`\mathrm{\Delta }m^2`$ from the MSW $`\nu _e\nu _s`$ solution region and of the neutrino energy of interest, $`E6.44`$ MeV, the energy-integrated D-N asymmetries $`A^C`$ and $`A^N`$ are determined essentially by the negative parts of $`A^C`$ and $`A^N`$ and therefore are negative. The zero asymmetry regions take place for $`A^C`$ and $`A^N`$ roughly at $`\mathrm{sin}^22\theta _V0.009`$ and $`\mathrm{sin}^22\theta _V0.010`$, respectively. At small mixing angles, $`\mathrm{sin}^22\theta _V0.010`$, the CC asymmetries are relatively small. For $`\mathrm{sin}^22\theta _V=(0.0030.010)`$, for instance, one finds $`3.0\%A_P^C8.0\%`$ and $`0.8\%A_P^N6.0\%`$.
The solar neutrino data strongly disfavors large mixing angle MSW solution in the case of $`\nu _e\nu _s`$ transitions. Nevertheless, we would like to note that due to the existence of points of absolute minima of the probability $`P_{e2}`$ for the solar neutrinos crossing the Earth core , $`\mathrm{min}(P_{e2})=0`$, one finds at $`\mathrm{sin}^22\theta (0.40.5)`$ a large negative Core probability asymmetry, $`\mathrm{min}(A_P^C)70\%`$, at $`E/\mathrm{\Delta }m^2(23)\times 10^6\mathrm{MeV}/\mathrm{eV}^2`$.
2.2.3 On the Latitude Dependence of the Core Asymmetry Enhancement
The comparison of $`<P_{e2}>^C=<P_{e2}>^C(\rho _r)`$ plots at fixed $`\mathrm{sin}^22\theta _V0.01`$ for the Super - Kamiokande detector , where
$$\rho _r\frac{\mathrm{\Delta }m^2\mathrm{cos}2\theta _V}{\sqrt{2}G_FE}m_N,$$
(7)
$`m_N`$ being the nucleon mass, with those for the SNO detector shows that the main effect on $`<P_{2e}>^C`$ of a moderated increase of the SNO detector latitude is a shift of the position of the dominating absolute maximum of $`<P_{2e}>^C`$ and not a reduction of this maximum. Thus, the magnitude of the mantle-core enhancement for moderate increase of the detector latitude, and therefore the sensitivity of the SNO detector to the D-N effect, is primarily affected by the reduction in the Core residence time and by changes in the position of the $`<P_{e2}>^C`$ dominating maximum along the $`\rho _r`$ axis with respect to the $`\overline{P}_{}(\nu _e\nu _e)=1/2`$ line. This observation suggests that for SNO, ICARUS and other solar neutrino detectors situated at latitudes which are somewhat larger than that of the Super-Kamiokande detector, the sensitivity to the D-N effect may be better than is usually expected.
To understand this result it is helpful to consider the dependence of $`P_{e2}`$ on $`\rho _r`$ and the Nadir angle $`\widehat{h}`$. The behavior of $`<P_{e2}>^C`$ follows from this dependence. At small $`\mathrm{sin}^22\theta _V`$ and for any fixed value of $`\rho _r`$ (or $`E/\mathrm{\Delta }m^2`$), $`P_{e2}`$ is a decreasing function of $`\widehat{h}`$ for the Earth-core-crossing neutrinos <sup>††</sup><sup>††</sup>††This is related to the fact that with the increase of $`\widehat{h}`$ the position of the total neutrino conversion point where $`P_{e2}=1`$, which determines the magnitude of the mantle-core enhancement of $`P_{e2}`$ at $`\mathrm{sin}^22\theta _V0.01`$, shifts to larger values of $`\mathrm{sin}^22\theta `$ , e.g., from $`\mathrm{sin}^22\theta 0.04`$ at $`\widehat{h}=13^{}`$ to $`\mathrm{sin}^22\theta 0.08`$ at $`\widehat{h}=30^{}`$ in the $`\nu _e\nu _{\mu (\tau )}`$ case. The indicated shift is caused, in particular, by the decreasing of the neutrino path length through the Earth core and of the average density of the core along the neutrino path when $`\widehat{h}`$ increases.. Since the time averaged $`\widehat{h}`$ for the core-crossing neutrinos increases with the increasing of the detector latitude and given the fact that SNO is at a higher latitude than the Super-Kamiokande detector, a smaller $`<P_{e2}>^C`$ probability is expected for SNO than for the Super-Kamiokande detector, which seems to be in contradiction with the suggestion made above. A more detailed study of the $`\rho _r`$ and $`\widehat{h}`$ dependence of $`P_{e2}`$ at fixed $`\mathrm{sin}^22\theta _V0.01`$ reveals, however, that in the region of the absolute maximum of $`P_{2e}`$ (at a given $`\mathrm{sin}^22\theta _V`$) this dependence can be described as a part of a positive saddle with fastly decreasing “shoulders”, but whose main axis (i.e., the “ridge” or minimal gradient line) is curved towards smaller values of $`\rho _r`$ when $`\widehat{h}`$ increases (Fig. 3). Moreover, along its main axis and on a relatively large part of it, the saddle is nearly flat and therefore $`P_{2e}`$ changes very little. The plots of $`P_{2e}`$ versus $`\widehat{h}`$ at fixed $`\rho _r`$ are sections of the saddle at high inclination with respect of the saddle axis, so what they illustrate is the fast decrease of the saddle shoulders. The plots of the $`P_{2e}`$ dependence on $`\rho _r`$ at fixed but different $`\widehat{h}`$ show the shift of the position of the maximum along the $`\rho _r`$axis for changing $`\widehat{h}`$, thus exhibiting the curvature of the saddle axis. The behavior of $`P_{2e}`$ discussed above is illustrated in Fig. 3.
When translating these considerations into total event rate predictions, one obviously has to take into account the effects of the shape of the $`{}_{}{}^{8}B`$ neutrino spectrum and of the integration over the neutrino energy. For $`\mathrm{sin}^22\theta _V`$ in the range of $`(10^310^2)`$, the D-N effect is suppressed by the simultaneous presence of a negative and positive parts of the D-N probability asymmetry in the interval of values of $`E/\mathrm{\Delta }m^2`$ of interest . As already discussed above, one of the important parameters which determine for given $`\mathrm{sin}^22\theta _V`$ the magnitude of the D-N effect is the position of the absolute maximum of $`<P_{2e}>^C`$ along the $`E/\mathrm{\Delta }m^2`$ axis with respect to the zero asymmetry line $`E/\mathrm{\Delta }m^2=(E/\mathrm{\Delta }m^2)_0`$ corresponding to $`\overline{P}_{}=1/2`$. The shift towards smaller values of $`\rho _r`$ (or larger values of $`E/\mathrm{\Delta }m^2`$) of the position of the dominating maximum of $`<P_{2e}>^C`$ as the detector latitude increases makes the crossing of the $`\overline{P}_{}=1/2`$ line by the maximum occur at larger values of $`\mathrm{sin}^22\theta _V`$. Thus, for a given $`\mathrm{sin}^22\theta _V0.009`$, the region of negative values of $`A_P^C`$ in $`E/\mathrm{\Delta }m^2`$ increases somewhat with the increase of $`\widehat{h}`$ at the expense of the region where $`A_P^C>0`$. This results in a certain decreasing of the CC total event rate Core D-N asymmetry $`A_{DN}^C`$. Similar effect takes place for the Night asymmetry $`A_{DN}^N`$. The decreasing is roughly the same in magnitude as the one caused by the presence of the NC contribution due to the $`\nu _{\mu (\tau )}`$ in the Super-Kamiokande solar neutrino signal in the case of the MSW $`\nu _e\nu _{\mu (\tau )}`$ transitions of solar neutrinos. Actually, the ratio $`A_{DN}^C/A_{DN}^N`$ of the discussed asymmetries for SNO is typically bigger, as we shall see, than that for the Super-Kamiokande detector. This analysis shows also that in the case of the MSW SMA $`\nu _e\nu _{\mu (\tau )}`$ solution the enhancement of the Core D-N asymmetry in the CC event rates for detectors located at different latitudes not exceeding, say, the SNO latitude ($`\lambda _{\mathrm{SNO}}=46{}_{}{}^{}\mathrm{\hspace{0.17em}20}_{}^{}`$ North), is approximately the same in magnitude and that the sensitivity to the Core D-N effect, apart from the systematic error, is limited essentially by the statistics of the corresponding data samples.
3. Predictions for the Energy-Integrated D-N Asymmetries
In this Section we will discuss the predictions we have obtained for the values of the various energy-integrated, one year averaged D-N asymmetries for the SNO detector. The asymmetries are defined as in eq. (2) and in . We consider the following set of D-N asymmetries: i) Night, Core and Mantle asymmetries in the CC signal in SNO, $`A_{DN}^{N,C,M}(CC)=A_{DN}^{N,C,M}(CC;T_{e,th},Y_e^c)`$, in the case of the $`\nu _e\nu _{\mu (\tau )}`$ and $`\nu _e\nu _s`$ transitions of solar neutrinos, ii) Night, Core and Mantle asymmetries in the ES signal in SNO, $`A_{DN}^{N,C,M}(ES)=A_{DN}^{N,C,M}(ES;T_{e,th},Y_e^c)`$, in the case of the $`\nu _e\nu _{\mu (\tau )}`$ and $`\nu _e\nu _s`$ transitions, and iii) Night, Core and Mantle asymmetries in the NC signal in SNO, $`A_{DN}^{N,C,M}(NC)=A_{DN}^{N,C,M}(NC;Y_e^c)`$, in the case of the $`\nu _e\nu _s`$ transitions <sup>‡‡</sup><sup>‡‡</sup>‡‡Note that with the definition of the D-N asymmetry we use , the one year average energy-integrated D-N asymmetries we consider are zero in the case of massless standard model neutrinos.. The asymmetries $`A_{DN}^{N,C,M}(CC;T_{e,th},Y_e^c)`$ and $`A_{DN}^{N,C,M}(ES;T_{e,th},Y_e^c)`$ have been calculated for two values of $`T_{e,th}`$: 5.0 MeV and 7.5 MeV. Results for all asymmetries for $`Y_e(core)=0.467and0.500`$ are given <sup>\**</sup><sup>\**</sup>\**Predictions for the Night CC asymmetry $`A_{DN}^N(CC;T_{e,th},Y_e^c)`$ for $`T_{e,th}=5`$ and $`T_{e,th}=5.0;7.5`$ MeV for the standard values of $`Y_e(man)`$ and $`Y_e(core)`$ were given respectively in and . Results for the Night asymmetry in the NC data $`A_{DN}^N(NC)`$ for the standard values of $`Y_e(man)`$ and $`Y_e(core)`$ were obtained also in ..
Our results are presented in Tables II - X and in Figs. 4a - 11b. The tables contain values of the three types of asymmetries and of the ratio of the Core and Night asymmetries, calculated for 36 pairs of values of $`\mathrm{\Delta }m^2`$ and $`\mathrm{sin}^22\theta _V`$ distributed evenly in the “conservative” MSW solution regions (older and most recent). Some of these pairs of values, as like those with $`\mathrm{sin}^22\theta _V0.03`$ in the case of MSW $`\nu _e\nu _s`$ solution in Tables VI - X, have been excluded by the recent Super-Kamiokande data, but we kept them for giving an idea about the magnitude of the asymmetries corresponding to them. In Figs. 4a - 11b contours of constant (D-N) asymmetries in the $`\mathrm{\Delta }m^2\mathrm{sin}^22\theta _V`$ plane are shown (iso-(D-N) asymmetry contour plots) together with the “conservative” MSW solution regions from . We did not include in the present work the iso-(D-N) asymmetry contour plots for the different (CC, ES, NC) Mantle asymmetries because the latter are relatively small and thus difficult to observe <sup>\*†</sup><sup>\*†</sup>\*†These plots can be found at the site: http://www.pv.infn.it/$``$maris/nue/sno/index.html.. Below we comment briefly our results for the indicated D-N asymmetries.
3.1. Magnitude of the Asymmetries and their $`Y_e(core)`$ and $`T_{e,th}`$ Dependence
3.1.1. $`\nu _e\nu _{\mu (\tau )}`$ Transitions
A. Asymmetries in the CC Event Sample
The most remarkable feature of the D-N asymmetries under discussion in this case is the relatively large absolute values of the Core CC asymmetry $`A_{DN}^C(CC)`$ in most of the region MSW SMA solution, as well as the strong mantle-core enhancement of $`A_{DN}^C(CC)`$ with respect to the Night and Mantle asymmetries in this region (Tables II and III). Due to the indicated enhancement we typically have $`|A_{DN}^C(CC)|>1\%`$ at $`\mathrm{sin}^22\theta _V(0.0010.006)`$ and $`\mathrm{\Delta }m^2(3.08.0)\times 10^6\mathrm{eV}^2`$, while the Night and Mantle asymmetries $`|A_{DN}^{N,M}(CC)|<1\%`$ and are hardly observable. The asymmetry $`A_{DN}^C(CC)`$ is negative and greater than 1% in absolute value for $`\mathrm{sin}^22\theta _V(0.74.0)\times 10^3`$ and $`\mathrm{\Delta }m^2`$ in the interval above, reaching the minimal value of $`(3.2)\%`$ at $`\mathrm{sin}^22\theta _V0.002`$. The ratio $`|A_{DN}^C(CC)/A_{DN}^N(CC)|`$ varies with $`\mathrm{sin}^22\theta _V`$ and $`\mathrm{\Delta }m^2`$ in the MSW SMA solution region typically in the interval of $`(58)`$. For $`\mathrm{sin}^22\theta _V(0.0080.010)`$, we have $`A_{DN}^C(CC)(12.939.7)\%`$, while $`A_{DN}^N(CC)(1.86.6)\%`$, the ration between the two asymmetries varying in the range of $`(4.37.9)`$. Actually, the CC Night asymmetry $`A_{DN}^N(CC)`$ is hardly observable at $`\mathrm{sin}^22\theta _V<0.008`$. The same conclusion is valid for the Mantle asymmetry $`A_{DN}^M(CC)`$ at $`\mathrm{sin}^22\theta _V<0.010`$.
The mantle-core enhancement of the CC Core asymmetry is rather modest in the region of the MSW LMA solution, where we typically have $`A_{DN}^N(CC)/A_{DN}^N(CC)(1.01.3)`$. Nevertheless, both asymmetries $`A_{DN}^C(CC)`$ and $`A_{DN}^N(CC)`$ have values exceeding 1% in most of the solution’s region (Tables II and III). More specifically, for $`\mathrm{\Delta }m^29.0\times 10^5\mathrm{eV}^2`$, we have $`A_{DN}^N(CC;7.5\mathrm{MeV})2.5\%`$ and $`A_{DN}^C(CC;5.0\mathrm{MeV})5.0\%`$. These predictions suggest that the SNO experiment will be able to probe most of the LMA solution region by measuring the Night and/or Core D-N asymmetries in the CC event sample. The conclusion is robust against the change of $`Y_e(core)`$ in the interval 0.467 - 0.500.
Changing $`Y_e(core)`$ from 0.467 to 0.500 affects little the Night CC asymmetry, but can change noticeably, typically by $`(1030)\%`$, the Core asymmetry in the SMA solution region (see, e.g., cases 13 - 19 in Table II and 12-14 and 16 in Table III). For $`\mathrm{sin}^22\theta _V(0.0040.010)`$ and $`\mathrm{\Delta }m^2(6.510.0)\times 10^6\mathrm{eV}^2`$, the change of $`A_{DN}^C(CC)`$ can be quite dramatic - the asymmetry increases approximately by a factor of $`1.5`$. Thus, for values of $`\mathrm{sin}^22\theta _V`$ and $`\mathrm{\Delta }m^2`$ from a specific subregion of the MSW SMA solution region the magnitude of the predicted Core asymmetry $`A_{DN}^C(CC)`$ is very sensitive to the value of the electron fraction number $`Y_e(core)`$ and the average matter density of the Earth core. The former is determined by the chemical composition of the core (see also ). If the MSW SMA solution turned out to be the solution of the solar neutrino problem and the parameters $`\mathrm{sin}^22\theta `$ and $`\mathrm{\Delta }m^2`$ lied in the region specified above and are known with a sufficiently high precision, the measurement of the CC Core asymmetry can give unique information, e.g., about the chemical composition of the Earth core.
Increasing the value of the threshold kinetic energy of the final state electron $`T_{e,th}`$ in the CC reaction, eq. (3), from 5.0 MeV to 7.5 MeV can have a dramatic effect on the Core asymmetry $`A_{DN}^C(CC)`$ changing it by a factor of $`2`$ (compare cases 11, 18 and 20 in Tables II and III). For the values of $`\mathrm{sin}^22\theta _V`$ and $`\mathrm{\Delta }m^2`$ for which the Night asymmetry $`A_{DN}^N(CC;5\mathrm{MeV})>1\%`$, the asymmetry for $`T_{e,th}=7.5`$ MeV, $`A_{DN}^N(CC;7.5\mathrm{MeV})`$, can be both larger or smaller than $`A_{DN}^N(CC;5\mathrm{MeV})`$ by at most $`20\%`$. For given $`\mathrm{sin}^22\theta _V`$ and $`\mathrm{\Delta }m^2`$ from the MSW solution regions there should exist optimal value of $`T_{e,th}5`$ MeV for which $`|A_{DN}^C(CC;T_{e,th})|`$ and/or $`|A_{DN}^N(CC;T_{e,th})|`$ have maximal values.
All the above properties of the Night and Core CC asymmetries under discussion are exhibited clearly in the corresponding iso-(D-N) asymmetry contour plots for $`A_{DN}^N(CC)`$ and $`A_{DN}^C(CC)`$, shown in Figs. 4a-4b and 5a-5b, respectively.
B. ES Event Sample Asymmetries
The D-N asymmetries in the ES sample are typically smaller by a factor of $`2`$ than the asymmetries in the CC sample (compare Tables II, III with IV, V). There are exceptions when $`A_{DN}^{C(N)}(ES)`$ is of the order of or even larger than $`A_{DN}^{C(N)}(CC)`$ but not by factors which can compensate for the much lower statistics expected in the ES sample of SNO. The mantle-core enhancement of the Core asymmetry at $`\mathrm{sin}^22\theta 0.01`$ is operative in this case as well: one typically has $`|A_{DN}^C(ES)/A_{DN}^N(ES)|(58)`$ (see Tables IV and V).
The above features of the ES sample asymmetries are exhibited in Figs. 6a - 6b and 7a - 7b, representing the iso-(D-N) asymmetry contour plots for $`A_{DN}^N(ES)`$ and $`A_{DN}^C(ES)`$, respectively, derived for $`Y_e(core)=0.467;0.500`$ and for $`T_{e,th}=5\mathrm{MeV}(a)`$ and $`T_{e,th}=7.5\mathrm{MeV}(b)`$.
3.1.2. $`\nu _e\nu _s`$ Transitions
The predicted D-N asymmetries of the same type (Core, Night or Mantle) in the CC, ES and NC event samples do not differ much in magnitude in the case of the MSW $`\nu _e\nu _s`$ transition solution of the solar neutrino problem (Tables VI - X). Let us remind the reader that the observation of a D-N asymmetry in the NC sample of events would be a proof that solar neutrinos undergo transitions into sterile neutrinos : this asymmetry is zero in the case of the $`\nu _e\nu _{\mu (\tau )}`$ transitions. The main contribution to the energy-integrated D-N asymmetries in the three event samples comes, as in the case of the Super-Kamiokande detector , from the “high” energy tail of the <sup>8</sup>B neutrino spectrum. This is related to the fact the relevant neutrino effective potential difference in the Earth matter in the case of the $`\nu _e\nu _s`$ transitions is approximately by a factor of 2 smaller than in the case of $`\nu _e\nu _{\mu (\tau )}`$ transitions . Only the Core asymmetries can exceed $`1\%`$ in absolute value at $`\mathrm{sin}^22\theta _V0.01`$ in specific limited subregions of the “conservative” MSW solution region. The Night asymmetries in the “conservative” solution region are for all samples smaller than 1% for $`\mathrm{sin}^22\theta _V0.01`$. In the subregion where $`\mathrm{sin}^22\theta _V>0.01`$ both the Core and Night asymmetries for the three event samples, $`A_{DN}^{C,N}(CC)`$, $`A_{DN}^{C,N}(ES)`$ and $`A_{DN}^{C,N}(NC)`$, are greater than 1%, the Core asymmetries being larger than the corresponding Night asymmetries by a factor of $`(25)`$.
The above features can be seen in the corresponding iso-(D-N) asymmetry contour plots shown in Figs. 8a - 8b (CC Night asymmetry), 9a - 9c (CC Core asymmetry), 10a - 10b (ES Night and Core asymmetries) and 11a - 11b (NC Night and Core asymmetries).
Let us note finally that in the subregion $`\mathrm{\Delta }m^2(4.07.0)\times 10^6\mathrm{eV}^2`$, $`\mathrm{sin}^22\theta (0.0090.020)`$, of the “conservative”solution region, the CC Core asymmetry $`A_{DN}^C(CC)`$ exhibits a very strong dependence on the value of $`Y_e(core)`$: the change of $`Y_e(core)`$ from the value of 0.467 to 0.500 leads to an increase of $`A_{DN}^C(CC)`$ by a factor of $`(24)`$ (see cases 26 - 33 in Tables VI and VII). This becomes evident also by comparing Figs. 9a and 9b.
4. Conclusions
In the present article we have derived detailed predictions for the day-night (D-N) asymmetry in the energy-integrated one year signals in the SNO detector in the case of the MSW $`\nu _e\nu _{\mu (\tau )}`$ and/or $`\nu _e\nu _s`$ transition solutions of the solar neutrino problem. The asymmetries in the charged current (CC), eq. (3), and $`\nu e^{}`$ elastic scattering (ES), eq. (4), event rates were calculated for both MSW solutions; in the case of the $`\nu _e\nu _s`$ transition solution the D-N asymmetry in the neutral current (NC) event rate, eq. (5), was derived as well. The asymmetries were calculated for the three night samples of events which are produced by the solar neutrinos crossing i) the Earth mantle only (Mantle), ii) the Earth core (Core) and iii) the Earth core and/or the mantle (Night) on the way to the SNO dtector: $`A_{DN}^{C,N,M}(CC)`$, $`A_{DN}^{C,N,M}(ES)`$ and $`A_{DN}^{C,N,M}(NC)`$.
We have studied the effects of the uncertainties in the values of the cross-sections of the CC and NC neutrino-induced reactions on deuterium by comparing the predictions for the asymmetries $`A_{DN}^{C,N,M}(CC)`$ and $`A_{DN}^{C,N,M}(NC)`$, obtained using the CC cross-sections from refs. , and the NC cross-sections from refs. . Our analysis showed that even neglecting the electron energy resolution function and the expected errors in the measurements, the calculation of the D-N asymmetries in the CC and NC event samples in the SNO detector utilizing the different cross-sections from and , respectively, should give the same results within 1% in the MSW “conservative” solution regions.
As in the case of the same type of D-N asymmetries for the Super-Kamiokande detector , at small mixing angles the Core asymmetries were shown to be considerably larger than the corresponding Night and Mantle asymmetries due to the mantle-core enhancement effect . Due to this enhancement, e.g., the one year average Core CC asymmetry in the case of the SMA MSW $`\nu _e\nu _{\mu (\tau )}`$ solution is bigger than 1% in absolute value, $`|A_{DN}^C(CC)|>1\%`$, in large subregions of the “conservative” solution region which extends in $`\mathrm{sin}^22\theta _V`$ from $`0.7\times 10^3`$ to $`10^2`$ (Figs. 5a - 5b). In contrast, in the same solution region we have $`|A_{DN}^{N(M)}(CC)|<1\%`$ for the Night (Mantle) asymmetry at $`\mathrm{sin}^22\theta _V<0.008(0.010)`$ (Figs. 4a - 4b). The asymmetry $`A_{DN}^C(CC)`$ is negative and greater than 1% in absolute value for $`\mathrm{sin}^22\theta _V(0.74.0)\times 10^3`$ and $`\mathrm{\Delta }m^2(3.08.0)\times 10^6\mathrm{eV}^2`$, reaching the minimal value of $`(3.2)\%`$. The ratio $`|A_{DN}^C(CC)/A_{DN}^N(CC)|`$ varies with $`\mathrm{sin}^22\theta _V`$ and $`\mathrm{\Delta }m^2`$ in the MSW SMA solution region typically in the interval of $`(58)`$ (Tables II and III). For $`\mathrm{sin}^22\theta _V(0.0080.010)`$, we have in this region $`A_{DN}^C(CC)(12.939.7)\%`$, while $`A_{DN}^N(CC)(1.86.6)\%`$.
In the region of the MSW LMA $`\nu _e\nu _{\mu (\tau )}`$ solution we typically have $`A_{DN}^N(CC)/A_{DN}^N(CC)(1.01.3)`$. Both asymmetries $`A_{DN}^C(CC)`$ and $`A_{DN}^N(CC)`$ have values exceeding 1% in most of the solution region (Tables II, III and Figs. 4a - 5b).
The D-N asymmetries in the ES sample in the case of the MSW $`\nu _e\nu _{\mu (\tau )}`$ solution are in most of the “conservative” solution regions smaller by a factor of $`2`$ than the asymmetries in the CC sample (Tables II, III and IV, V, Figs. 6a - 7b). There are exceptions when $`A_{DN}^{C(N)}(ES)/A_{DN}^{C(N)}(CC)1`$, but the ratio factors involved cannot compensate for the much lower statistics expected in the ES sample of SNO.
The predicted D-N asymmetries of the same type (Core, Night or Mantle) in the CC, ES and NC event samples do not differ much in magnitude in the case of the MSW $`\nu _e\nu _s`$ transition solution of the solar neutrino problem (Tables VI - X, Figs. 8a - 11b). Only the Core asymmetries can exceed $`1\%`$ in absolute value at $`\mathrm{sin}^22\theta _V0.01`$ in specific limited subregions of the “conservative” solution region. In the subregion where $`\mathrm{sin}^22\theta _V>0.01`$ both the Core and Night asymmetries for the three event samples, are greater than 1%, the Core asymmetries exceeding the corresponding Night asymmetries by a factor of $`(25)`$.
We have found also that in certain subregions of the MSW solution regions at small $`\mathrm{sin}^22\theta _V`$, the predicted magnitude of the Core D-N asymmetry in the CC and ES samples is very sensitive to the value of the electron fraction number $`Y_e^c`$ (i.e., to the chemical composition) and to the average matter density of the Earth core: the asymmetry $`A_{DN}^C(CC)`$, for instance, can increase in the case of the $`\nu _e\nu _{\mu (\tau )}`$ ($`\nu _e\nu _s`$) solution by a factor of $`1.5(24)`$ when $`Y_e^c`$ is changed from the standard value of 0.467 to its possibly conservative upper limit of 0.500 (Tables II, III, VI, VII, Figs. 5a, 5b, 7a, 7b, 9a-9b).
We have studied the dependence of the D-N asymmetries considered on the final state e<sup>-</sup> threshold energy $`T_{e,th}`$ in the CC and ES reactions. Increasing the value of $`T_{e,th}`$ in the CC and ES reactions from 5.0 MeV to 7.5 MeV can have a dramatic effect on the Core asymmetries $`A_{DN}^C(CC)`$ and $`A_{DN}^C(ES)`$, changing, e.g., $`A_{DN}^C(CC)`$ by a factor of $`2`$ in the case of the $`\nu _e\nu _{\mu (\tau )}`$ solution (Tables II, III, VI, VII, Figs. 5a, 5b, 9a - 9c).
Our analysis showed also that for the MSW SMA $`\nu _e\nu _{\mu (\tau )}`$ and $`\nu _e\nu _s`$ solutions, the enhancement of the Core D-N asymmetry in the CC event rates for detectors located at different latitudes not exceeding, say, the SNO latitude ($`\lambda _{\mathrm{SNO}}=46{}_{}{}^{}\mathrm{\hspace{0.17em}20}_{}^{}`$ North), is approximately the same in magnitude. This result suggests that for SNO, ICARUS and other solar neutrino detectors operating at latitudes which are somewhat larger than that of the Super-Kamiokande detector, the sensitivity to the Core D-N effect, apart from the systematic error, is limited essentially by the statistics of the corresponding data samples.
The results of the present study allow us to conclude, in particular, that the SNO experiment has the potential of probing substantial parts of the SMA and LMA MSW $`\nu _e\nu _{\mu (\tau )}`$ solution regions by performing Night and Core D-N asymmetry measurements in the CC event sample. The measurements of the D-N effect related observables with the SNO detector appears unlikely to provide by itself a critical test of the MSW $`\nu _e\nu _s`$ (SMA) solution of the solar neutrino problem. It is expected that such a test will be provided by the comparison of the CC and NC SNO data. Nevertheless, certain subregions of the “conservative” MSW $`\nu _e\nu _s`$ solution region can also be probed through the measurements of the D-N effect, e.g., in the CC and NC event samples.
Acknowledgements.
We would like to thank H. Robertson for clarifying discussions concerning the SNO experiment. The work of (S.T.P.) was supported in part by the EC grant ERBFMRX CT96 0090.
FIGURE CAPTIONS
Figures 1a - 1f. The variation with $`\mathrm{\Delta }m^2`$ and $`\mathrm{sin}^22\theta _V`$ of the difference between the values of the Core (Night) D-N asymmetry (a) - (c) ((d) - (f)), calculated utilizing the predictions for the relevant CC and NC reaction cross-sections derived in ref. and in ref. : figures (a), (b), (e) ((c), (f)) show the asymmetry difference in the one year averaged CC (NC) event rate; figures (a), (d) ((b), (c), (e) and (f)) correspond to $`\nu _e\nu _{\mu (\tau )}`$ ($`\nu _e\nu _s`$) transitions. The light-grey spot-like regions in the upper left-hand panels correspond to asymmetry differences exceeding 1%, while the dark-grey areas are the regions of the MSW solutions.
Figures 2a - 2b. The probabilities $`<P_{e2}>^C`$ (solid line), $`<P_{e2}>^N`$ (dashed line) and $`<P_{e2}>^M`$ (dash-dotted line) as functions of $`E/\mathrm{\Delta }m^2`$ for $`\mathrm{sin}^22\theta _V=0.005(a);0.010(b)`$ in the case of $`\nu _e\nu _{\mu (\tau )}`$ transitions of solar neutrinos.
Figure 3. The dependence of $`P_{e2}`$ on $`\rho _r`$ and the Nadir angle $`\widehat{h}`$ for $`\mathrm{sin}^22\theta _V=0.01`$ in the case of MSW $`\nu _e\nu _{\mu (\tau )}`$ (upper left cannel) and $`\nu _e\nu _s`$ (upper right cannel) transitions of solar neutrinos. The grey-scales correspond to different values of $`P_{e2}`$, as is indicated in the two vertical columns between the two upper panels. The thin solid lines in both upper panels represent contours of constant $`P_{e2}`$ values: $`P_{e2}=0.01`$, $`0.025`$, $`0.05`$, $`0.1`$, $`0.15`$, $`0.2`$, $`0.25`$, $`0.30`$, $`0.40`$. The dotted and the thick black lines are minimum gradient axes, connecting the points of local maxima of $`P_{e2}`$ in the variable $`\rho _r`$ at fixed $`\widehat{h}`$ (“ridges”), while the dash-dotted lines are minimum gradient axes connecting points of local minima of $`P_{e2}`$ (“valleys”). The solid lines in the lower panels show $`P_{e2}`$ as a function of $`\widehat{h}`$, computed along the “ridge” leading to the absolute maximum of $`P_{e2}`$ at $`\widehat{h}=0^{}`$, while the dashed lines show the dependence of $`P_{e2}`$ on $`\widehat{h}`$ along the $`\rho _r=const.`$ line (dashed lines in the upper panels) starting from the point of the absolute maximum at $`\widehat{h}=0^{}`$.
Figures 4a - 4b, 5a - 5b. Iso - (D-N) asymmetry contour plots for the one year average CC Night (4a,4b), and Core (5a,5b) asymmetries for the SNO detector in the case of the $`\nu _e\nu _{\mu (\tau )}`$ transitions and $`T_{e,th}=5.0(a);7.5(b)`$ MeV. The solid (dashed) lines correspond to $`Y_e(core)=0.467(0.500)`$. The MSW SMA and LMA “conservative” solution regions from ref. are also shown.
Figures 6a - 6b, 7a - 7b. Iso - (D-N) asymmetry contour plots for the one year average ES Night (6a,6b), and Core (7a,7b) asymmetries for the SNO detector in the case of the $`\nu _e\nu _{\mu (\tau )}`$ transitions and $`T_{e,th}=5.0(a);7.5(b)`$ MeV. The solid (dashed) lines correspond to $`Y_e(core)=0.467(0.500)`$. The MSW SMA and LMA “conservative” solution regions are also shown.
Figures 8a - 8b, 9a - 9c. The same as in figures 4a - 4b and 5a - 5b in the case of the $`\nu _e\nu _s`$ transitions of solar neutrinos. The CC Core asymmetry in figures 9a - 9c correspond to $`T_{e,th}=5.0(a,b);7.5(c)`$ and $`Y_e(core)=0.467(a,c);0.500(b)`$. The MSW $`\nu _e\nu _s`$ transition solution regions from ref. are also shown (the “conservative” solution region is marked with thin dashed lines).
Figures 10a -10b. Iso - (D-N) asymmetry contour plots for the one year average ES Night (a) and Core (b) asymmetries for the SNO detector in the case of the $`\nu _e\nu _s`$ transitions of solar neutrinos for $`T_{e,th}=5.0`$ and $`Y_e(core)=0.467`$. The MSW $`\nu _e\nu _s`$ transition solution regions are also shown (the “conservative” solution region is marked with thin dashed lines).
Figures 11a - 11b. The same as in Fig. 10 for the one year average NC Night (a) and Core (b) asymmetries in the case of the $`\nu _e\nu _s`$ transitions of solar neutrinos. The solid (dashed) lines correspond to $`Y_e(core)=0.467(0.500)`$.
|
warning/0003/astro-ph0003423.html
|
ar5iv
|
text
|
# The Ratio of Total to Selective Extinction Toward Baade’s Window
## 1 Introduction
Paczyński & Stanek (1998) found that the $`VIK`$ colors of clump giants in Baade’s Window were anomalous in the sense that at fixed $`(VK)_0`$, they were redder in $`(VI)_0`$ than clump giants in the solar neighborhood by 0.2 mag. Stutz, Popowski, & Gould (1999) found a similar offset (0.17 mag) for RR Lyrae stars in Baade’s Window compared to those in the solar neighborhood. Popowski (2000) showed that part of these offsets was simply due to errors in the original photometry used by both groups. When he incorporated the revised OGLE photometry of Paczyński et al. (1999), he found that the offset in $`(VI)`$ shrank to $`0.11`$ mag in both cases.
Popowski (2000) then reviewed the various attempts to explain such an offset in terms of a difference between the intrinsic properties of stars in the two populations, an idea advanced by Paczyński (1998) and by Stutz et al. (1999). He argued that such an explanation was not impossible, but unlikely, and that a more plausible explanation is that the ratio of total to selective extinction $`R_{VI}A_V/E(VI)=2.5`$ adopted by both Paczyński & Stanek (1998) and Stutz et al. (1999) from Stanek (1996) was incorrect. Both color anomalies could be solved, he noted, by adopting $`R_{VI}=2.1`$. Popowski (2000) used this re-evaluation to draw various conclusions about the extragalactic distance scale.
In the course of calibrating a $`K`$-band clump-giant distance indicator and applying it to measure the distance modulus of the Baade’s Window bulge field, Alves (2000) measured $`R_{VI}=2.26`$. He did so by comparing the $`VIK`$ colors of his 20-star sample of Baade’s Window clump giants taken from Tiede, Frogel, & Terndrup (1995) with the $`VIK`$ colors of local clump stars.
Here we improve on the Alves (2000) measurement by incorporating a factor $`7`$ more stars into the analysis, i.e., all 138 G0-K4 giants with $`VIK`$ photometry from Tiede et al. (1995). We determine $`R_{VI}`$ by comparing these $`VIK`$ colors to the $`VIK`$ colors of nearby stars as determined by Bessell & Brett (1988). We find,
$$R_{VI}\frac{A_V}{E(VI)}=2.283\pm 0.016.$$
(1)
This reduces the color anomalies to $`0.05`$mag and so qualitatively confirms Popowski’s (2000) explanation of them.
## 2 Measurement of $`R_{VI}`$
To measure $`R_{VI}`$, we assume that the $`VIK`$ colors of giant stars in Baade’s Window as measured by Tiede et al. (1995) are intrinsically the same as those in solar neighborhood as determined by Bessell & Brett (1988). (We will partially test this assumption below.) Bessel & Brett (1988) give $`VIK`$ colors from G0 to M5, that is, over the color range $`1.75(VK)_05.96`$. However, there is considerable evidence that the spectral energy distributions of M giants (at fixed spectral type) differ significantly between Baade’s Window and the solar neighborhood (Frogel & Whitford 1982, 1987; Tiede et al. 1995). This does not necessarily mean that the $`(VI),(VK)`$ color-color relation is different, and in fact we will present evidence below that it is not. However, to be conservative, we restrict consideration to $`(VK)_03.5`$, which eliminates all M giants and K5 giants as well. That is,
$$1.75(VK)_03.50.$$
(2)
We begin with the sample of 509 stars with optical and infrared photometry from the BW4b field of Tiede et al. (1995). Note that the columns headed “$`V_0`$” and “$`(VI)_0`$” in that paper actually give $`V`$ and $`(VI)`$. We recover the original $`K`$ magnitudes by adding $`K=K_{0,\mathrm{Tiede}}+0.14`$, as indicated by Table 1 of Tiede et al. (1995). We then obtain $`(VK)_0`$ using visual extinctions, $`A_V`$, from Stanek’s (1996) extinction map together with the relation,
$$A_K=0.11A_V.$$
(3)
Note that the Stanek (1996) map (available by anonymous ftp at astro.princeton.edu, stanek/Extinction) has been corrected to the zero point found by Gould, Popowski, & Terndrup (1998) and Alcock et al. (1998). These authors made their zero-point determinations by comparing the $`(VK)`$ colors of local K giants and RR Lyrae stars, respectively, with the $`(VK)`$ colors of similar stars in Baade’s Window, making use of the extinction ratio given by equation (3). Hence, for consistency, we use the same ratio here.
Of the original 509 Tiede et al. (1995) stars, 185 lack $`VI`$ photometry. Of the remainder, 86 lack $`A_V`$ measurements because they fall within $`2^{}`$ of NGC 6522 where Stanek (1996) found the $`A_V`$ determinations to be unreliable. A further 96 stars are bluer than the color interval (2), and an additional 4 stars are redder. This leaves a total of 138 stars.
For each of these 138 stars, we use linear interpolation to estimate the $`(VI)_{0,\mathrm{Bessel}}`$ predicted from their measured $`(VK)_0`$ and the Bessell & Brett (1988) $`VIK`$ color-color relation. We then fit the data to a two-parameter model
$$(VI)_{\mathrm{Tiede}}(VI)_{0,\mathrm{Bessell}}=\alpha A_V+\beta [(VK)_02.348],$$
(4)
where the offset 2.348 is chosen to eliminate the correlation between $`\alpha `$ and $`\beta `$.
We remove outliers as follows. We do the fit using all the data, and determine the “errors” by forcing $`\chi ^2`$ per degree of freedom to unity. We find the largest $`\sigma `$ outlier, eliminate it, and repeat the process. We stop when the largest outlier is less than $`3\sigma `$. This eliminates six outliers. We find,
$$\alpha =0.4380\pm 0.0031,\beta =0.0023\pm 0.0127[1.75(VK)_03.50].$$
(5)
Since $`R_{VI}=\alpha ^1`$, we obtain equation (1). From the fact that $`\beta `$ is consistent with zero at the $`1\sigma `$ level, we conclude that the shift from $`(VI)`$ to $`(VI)_0`$ depends only on $`A_V`$ and not on the color of the star. We therefore set $`\beta =0`$ and show our resulting fit in Figure 1. Note that since $`R_{VI}`$ will in general be a function of color or spectral type, our measurement should be taken as applying to stars at the mean of our sample, $`(VK)_0=2.35`$, i.e., K0 giants. Our value, $`R_{VI}=2.28`$, is in good agreement with the one found by Alves (2000), $`R_{VI}=2.26`$, and the one adopted by Tiede et al. (1995), $`R_{VI}=2.25`$.
Also plotted on Figure 1 are the four stars (open circles) that were excluded from the fit because they were too red. Note that they lie very close to the local $`VIK`$ curve of Bessell & Brett (1988). If we repeat the entire procedure including these four stars, we obtain
$$\alpha =0.4384\pm 0.0033,\beta =0.0073\pm 0.0092[1.75(VK)_05.96].$$
(6)
with four outliers excluded. That is $`R_{VI}=2.281\pm 0.017`$, essentially identical to equation (1). In either case, the slope $`\beta `$ is consistent with zero, indicating that our determination is not a signficant function of spectral type.
This non-dependence on $`(VK)_0`$ color appears to be in strong conflict with Figure 14 of Tiede et al. (1995) which shows a large number of points in the range $`4<(VK)_0<6`$ that lie $`0.3`$ mag below the Bessell & Brett (1988) relation. In fact, all the points with CCD $`VI`$ photometry from Tiede et al. (1995) lie close to the line (as they do in our Fig. 1). The remaining points are from Frogel, Whitford & Rich (1984) and Frogel & Whitford (1987) who obtained their own single-channel infrared data, but relied on the earlier photographic data of Whitford & Blanco (1979) and Arp (1965) for the $`VI`$ photometry. From the small overlap between these older photographic data and the CCD photometry reported in Tiede et al. (1995), we estimate that the earlier photographic photometry may have a zero-point error of $`0.3`$ mag in $`(VI)`$ in the sense of being too blue. If this zero-point error is confirmed by future observations, it would mean that the $`(VI),(VK)`$ color-color relation is the same for Baade’s Window and the solar neighborhood, but not necessarily the correspondance between colors anad spectral type. Nevertheless, to be conservative, we base our results only on the G0–K4 sample.
## 3 Discussion
As with essentially all methods for determining total and selective extinction, our measurement relies on the assumption that the colors of stars in Baade’s Window are the same as those in the solar neighborhood. If the mean $`(VI)_0`$ color at fixed $`(VK)_0`$ differed between the two populations by $`\mathrm{\Delta }(VI)`$, then our estimate of $`\alpha `$ would likewise be in error by $`\mathrm{\Delta }\alpha =\mathrm{\Delta }(VI)/A_V`$ where $`A_V=1.496`$ is the mean value of $`A_V`$ over our final sample of 132 stars. However, from the fact that $`\beta `$ is consistent with zero (eq. ), such an offset would have to be independent of spectal type from G0 to K4 (and arguably to M5). This seems quite implausible. In addition, approximately the same offset would have to apply to RR Lyrae stars (Stutz et al. 1999). Most probably, the fault lies not in the stars, but in the dust.
Popowski (2000) reexamined the Woźniak & Stanek (1996) method by which Stanek (1996) determined $`R_{VI}=2.5`$. Making use of the original tests done by Woźniak & Stanek (1996), he found that this determination depends in part on an initial assumption about $`R_{VI}`$ so that values as low as $`R_{VI}2.3`$ would be consistent with the data.
Our measurement of $`R_{VI}`$ removes most, but not all, of the anomalous color problems found by Paczyński & Stanek (1998) and by Stutz et al. (1999). In the latter case, Popowski’s (2000) revised offset $`0.11\pm 0.02`$ mag is now reduced to $`0.04\pm 0.02`$, and so is only a $`2\sigma `$ discrepancy. However, for the clump giant anomaly found by Paczyński & Stanek (1998), the formal uncertainty is only $`0.003`$ mag, so statistical fluctuations do not provide a plausible explanation. Nevertheless, the remaining discrepancy is small and may be due to a combination of small offsets between the photometric zero points of the various measurement systems that are used to make the comparison.
Paczyński et al. (1999) used $`R_{VI}^{\mathrm{Stanek}}=2.50`$ from Stanek (1996) to deredden the observed centroid of the clump in their $`VI`$ color-magnitude diagram of Baade’s Window. The zero point of their $`V`$-band extinctions is based on Gould et al. (1998) and Alcock et al. (1998) and is not affected by the present paper. However, the $`I`$-band photometry should be adjusted fainter by $`\mathrm{\Delta }I_0=(1/R_{VI}^{\mathrm{Stanek}}1/R_{VI}^{\mathrm{GSF}})A_V=0.056`$, where $`R_{VI}^{\mathrm{GSF}}=2.283`$ is the value we determine here and $`A_V=1.48`$ is the mean extinction of clump stars measured by Paczyński et al. (1999). This adjustment yields a clump centroid of $`I_{0,\mathrm{RC}}=14.43`$ and $`(VI)_{0,\mathrm{RC}}=1.058`$. The color is only 0.05 mag redward of the centroid of the Hipparcos clump (Paczyński et al. 1999). Inserting the magnitude into Udalski’s (2000) $`I`$-band calibration of the clump standard candle, $`M_I=(0.26\pm 0.02)+(0.13\pm 0.07)([\mathrm{Fe}/\mathrm{H}]+0.25)`$, we obtain a distance modulus to Baade’s Window
$$\mu _{BW}=14.43M_I=14.68\pm 0.04,$$
(7)
where we have adopted \[Fe/H\]$`{}_{BW}{}^{}=0.15\pm 0.10`$. The error bar takes account of all statistical uncertainties, 0.025 mag for $`M_I`$ (Udalski 2000), $`0.02`$ mag for the observed brightness of the Baade’s Window clump (Paczyński et al. 1999), $`0.04\times (1R_{VI})=0.022`$ mag for the zero-point uncertainty of the Stanek (1996) map (Alcock et al. 1998), and $`0.0033\times A_V=0.005`$ mag for the uncertainty in $`R_{VI}`$ (this paper), but does not take account of systematic errors. Equation (7) is in good agreement with Alves’ (2000) determination using the $`K`$-band clump distance indicator, $`\mu _{BW}=14.58\pm 0.11`$.
Acknowledgements: We thank Piotr Popowski for his useful comments on the manuscript and David Alves for pointing out the error in the column headings of Tiede et al. (1995). This work was supported in part by grant AST 97-27520 from the NSF.
|
warning/0003/physics0003065.html
|
ar5iv
|
text
|
# 1 Problem
## 1 Problem
In optics, a reflective grating is a conducting surface with a ripple. For example, consider the surface defined by
$$z=a\mathrm{sin}\frac{2\pi x}{d}.$$
(1)
The typical use of such a grating involves an incident electromagnetic wave with wave vector k in the $`x`$-$`z`$ plane, and interference effects lead to a discrete set of reflected waves also with wave vectors in the $`x`$-$`z`$ plane.
Consider, instead, an incident plane electromagnetic wave with wave vector in the $`y`$-$`z`$ plane and polarization in the $`x`$ direction:
$$𝐄_{\mathrm{in}}=E_0\widehat{𝐱}e^{i(k_yyk_zz\omega t)},$$
(2)
where $`k_y>0`$ and $`k_z>0`$. Show that for small ripples ($`ad`$), this leads to a reflected wave as if $`a=0`$, plus two surface waves that are attenuated exponentially with $`z`$. What is the relation between the grating wavelength $`d`$ and the optical wavelength $`\lambda `$ such that the $`x`$ component of the phase velocity of the surface waves is the speed of light, $`c`$?
In this case, a charged particle moving with $`v_xc`$ could extract energy from the wave, which is the principle of the proposed “grating accelerator” .
## 2 Solution
The interaction between particle beams and diffraction gratings was first considered by Smith and Purcell , who emphasized energy transfer from the particle to free electromagnetic waves. The excitation of surface waves by particles near conducting structures was first discussed by Pierce , which led to the extensive topic of wakefields in particle accelerators. The presence of surface waves in the Smith-Purcell effect was noted by di Francia . A detailed treatment of surface waves near a diffraction grating was given by van den Berg . Here, we construct a solution containing surface waves by starting with only free waves, then adding surface waves to satisfy the boundary condition at the grating surface.
If the (perfectly) conducting surface were flat, the reflected wave would be
$$𝐄_\mathrm{r}=E_0\widehat{𝐱}e^{i(k_yy+k_zz\omega t)}.$$
(3)
However, the sum $`𝐄_{\mathrm{in}}+𝐄_\mathrm{r}`$ does not satisfy the boundary condition that $`𝐄_{\mathrm{total}}`$ must be perpendicular to the wavy surface (1). Indeed,
$$\left[𝐄_{\mathrm{in}}+𝐄_\mathrm{r}\right]_{\mathrm{surface}}=2iE_0\widehat{𝐱}e^{i(k_yy\omega t)}\mathrm{sin}k_zz2iak_zE_0\widehat{𝐱}e^{i(k_yy\omega t)}\mathrm{sin}k_xx,$$
(4)
where the approximation holds for $`ad`$, and we have defined $`k_x=2\pi /d`$.
Hence, we require additional fields near the surface to cancel that given by (4). For $`z0`$, these fields therefore have the form
$$𝐄=ak_zE_0\widehat{𝐱}e^{i(k_yy\omega t)}\left(e^{ik_xx}e^{ik_xx}\right).$$
(5)
This can be decomposed into two waves $`𝐄_\pm `$ given by
$$𝐄_\pm =ak_zE_0\widehat{𝐱}e^{i(\pm k_xx+k_yy\omega t)}.$$
(6)
Away from the surface, we suppose that the $`z`$ dependence of the additional waves can be described by including a factor $`e^{ik_z^{}z}`$. Then, the full form of the additional waves is
$$𝐄_\pm =ak_zE_0\widehat{𝐱}e^{i(\pm k_xx+k_yy+k_z^{}z\omega t)}.$$
(7)
The constant $`k_z^{}`$ is determined on requiring that each of the additional waves satisfy the wave equation,
$$^2𝐄_\pm =\frac{1}{c^2}\frac{^2𝐄_\pm }{t^2}.$$
(8)
This leads to the dispersion relation
$$k_x^2+k_y^2+k_z^{}_{}{}^{}2=\frac{\omega ^2}{c^2}.$$
(9)
The component $`k_y`$ of the incident wave vector can be written in terms of the angle of incidence $`\theta _{\mathrm{in}}`$ and the wavelength $`\lambda `$ as
$$k_y=\frac{2\pi }{\lambda }\mathrm{sin}\theta _{\mathrm{in}}.$$
(10)
Combining (9) and (10), we have
$$k_z^{}=2\pi i\sqrt{\frac{1}{d^2}\left(\frac{\mathrm{cos}\theta _{\mathrm{in}}}{\lambda }\right)^2}.$$
(11)
For short wavelengths, $`k_z^{}`$ is real and positive, so the reflected wave (3) is accompanied by two additional plane waves with direction cosines $`(k_x,k_y,k_z^{})`$. But for long enough wavelengths, $`k_z^{}`$ is imaginary, and the additional waves are exponentially attenuated in $`z`$.
When surface waves are present, consider the fields along the line $`y=0`$, $`z=\pi /2k_z`$. Here, the incident plus reflected fields vanish (see the first form of (4)), and the surface waves are
$$𝐄_\pm =ak_ze^{\pi \left|k_z^{}\right|/2k_z}E_0\widehat{𝐱}e^{i(\pm k_xx\omega t)}.$$
(12)
The phase velocity of these waves is
$$v_p=\frac{\omega }{k_x}=\frac{d}{\lambda }c.$$
(13)
When $`d=\lambda `$, the phase velocity is $`c`$, and $`k_z^{}=ik_y`$ according to (11). The surface waves are then,
$$𝐄_\pm =\frac{2\pi a\mathrm{cos}\theta _{\mathrm{in}}}{d}e^{(\pi /2)\mathrm{tan}\theta _{\mathrm{in}}}E_0\widehat{𝐱}e^{i(\pm k_xx\omega t)}.$$
(14)
A relativistic charged particle that moves in, say, the $`+x`$ direction remains in phase with the wave $`𝐄_+`$, and can extract energy from that wave for phases near $`\pi `$. On average, the particle’s energy is not affected by the counterpropagating wave $`𝐄_{}`$. In principle, significant particle acceleration can be achieved via this technique. For a small angle of incidence, and with $`a/d=1/2\pi `$, the accelerating field strength is equal to that of the incident wave.
|
warning/0003/hep-ph0003183.html
|
ar5iv
|
text
|
# QCD In Extreme Conditions Lectures given at CRM Summer School, June 27-July 10, Banff (Alberta), Canada.
## 1 Introduction
In some ways, QCD is a mature subject. Its principles are precisely defined, and they have been extensively confirmed by experiment. QCD specifies unambiguous algorithms, capable of transmission to a Turing machine, that supply the answer to any physically meaningful question within its domain – any question, that is, about the strong interaction. I believe there is very little chance that the foundational equations of QCD will require significant revision in the foreseeable future. Indeed, as we shall soon review, these equations are deeply rooted in profound concepts of symmetry and local quantum field theory, which lead to them uniquely. So one cannot revise the equations without undermining these concepts.
Granting that the foundations are secure, we have the task – which is actually a wonderful opportunity – of building upon them. Due to the peculiarities of QCD, this is a particularly interesting and important challenge.
Interesting, because while the foundational equations are conceptually simple and mathematically beautiful, they seem at first sight to have nothing to do with reality. Notoriously, they refer exclusively to particles (quarks and gluons) that are not directly observed. Less spectacular, but more profound, is the fact that equations exhibit a host of exact or approximate symmetries that are not apparent in the world. One finds that these symmetries are variously hidden: confined in the case of color, anomalous in the cases of scale invariance and axial baryon number, spontaneously broken in the case of chirality. It is fascinating to understand how a theory that superficially appears to be “too good for this world” actually manages to describe it accurately. Conversely it is pleasing to realize how the world is, in this profound and very specific way, simpler and more beautiful than it at first appears.
Important, because there are potential applications to which the microscopic theory has not yet rendered justice. An outstanding, historic, challenge is to derive the principles of nuclear physics. At present meaningful connections between the microscopic theory of QCD and the successful practical theory of atomic nuclei are few and tenuous, though in principle the former should comprehend the latter. This seems to be an intrinsically difficult problem, probably at least as difficult as computing the structure of complex molecules directly from QED. In both cases, the questions of interest revolve around small energy differences induced among valence structures, after saturation of much larger core forces. If one starts calculating from the basic equations, unfocussed, then small inaccuracies in the calculation of core parameters will blur the distinctions of interest.
A related but simpler class of problems is to calculate the spectrum of hadrons, their static properties, and a variety of operator matrix elements that are vital to the planning and interpretation of experiments. This is the QCD analogue of atomic physics. Steady progress has been made through numerical work, exploiting the full power of modern computing machines.
Other significant applications appear more accessible to analytic work, or to a combination of analytics and numerics.
The behavior of QCD at high temperature and low baryon number density is central to cosmology. Indeed, during the first few seconds of the Big Bang the matter content of the Universe was almost surely dominated by quark-gluon plasma. There are also ambitious, extensive programs planned to probe this regime experimentally.
The behavior of QCD at high baryon number density and (relatively) low temperature is central to extreme astrophysics – the description of neutron star interiors, neutron star collisions, and conditions near the core of collapsing stars (supernovae, hypernovae). Also, we might hope to find – and will find – insight into nuclear physics, coming down from the high-density side.
The special peculiarity of QCD, that its fundamental entities and abundant symmetries are well hidden in ordinary matter, lends elegance and focus to the discussion of its behavior in extreme conditions. Quarks, gluons, and the various symmetries will, in the right circumstances, come into their own. By tracing symmetries lost and found we will be able to distinguish sharply among different phases of hadronic matter, and to make some remarkably precise predictions about the transitions between them.
In these 5 lectures I hope to provide a reasonably self-contained introduction to the study of QCD in extreme conditions. The first lecture is a rapid tour of QCD itself. I’ve organized it as a cluster of related stories narrating how apparent symmetries of the fundamental equations are hidden by characteristic dynamical mechanisms. Lectures 2-3 are mainly devoted to QCD at high temperature, and lectures 4-5 to QCD at high baryon number density. I have structured the lectures so that they head toward two climaxes: the prediction of a true critical point in real QCD, that ought to be accessible to numerical and laboratory experiments, in Lecture 3, and the prediction that at asymptotic densities QCD goes over into a color-flavor locked phase with remarkable properties including fully calculable realizations of confinement and chiral symmetry breaking, in Lecture 5. These are, I believe, remarkable results, and they bring us to frontiers of current research.
A wide variety of tools will be brought to bear, including three different renormalization groups (the usual ‘asymptotic freedom’ version toward bare quarks and gluons at high virtuality, the usual ‘Kadanoff-Wilson-Fischer’ version toward critical modes at a second-order phase transition, and the ‘Landau-Anderson’ version toward quasiparticles near the Fermi surface), perturbative quantum field theory, effective field theory, instantons, lattice gauge theory, and BCS pairing theory. Of course I won’t be able to give full-blown introductions to all these topics here; but I’ll try to give coherent accounts of the concepts and results I actually need, and to supply appropriate standard references. In a few months, I hope, a comprehensive reference will be appearing.
## 2 Lecture 1: Symmetry and the Phenomena of QCD
### 2.1 Apparent and Actual Symmetries
Let me start with a slightly generalized and slightly idealized Lagrangian for QCD:
$$=\frac{1}{4g^2}\mathrm{tr}G^{\mu \nu }G_{\mu \nu }+\underset{j=1}{\overset{f}{}}\overline{\psi }_j(i\overline{)}D)\psi _j$$
(1)
where
$$G_{\mu \nu }=_\mu A_\nu _\nu A_\mu +i[A_\mu ,A_\nu ]$$
(2)
and
$$D=_\mu +iA_\mu $$
(3)
The $`A_\mu `$ are 3$`\times `$3 traceless Hermitean matrices. Each spin-1/2 fermion quark field $`\psi _j`$ carries a corresponding 3-component color index.
Eqn. 1 is both slightly generalized from real-world QCD, in that I’ve allowed for a variable number $`f`$ of quarks; and slightly idealized, in that I’ve set all their masses to zero. Also, I’ve set the $`\theta `$ parameter to zero, once and for all. We will have much occasion to focus on particular numbers and masses of quarks in our considerations below, but Eqn. 1 is a good simple starting point.
Our definition of the gauge potential differs from the more traditional one by
$$A_\mu ^{\mathrm{here}}=gA_\mu ^{\mathrm{usual}}$$
(4)
which is why the coupling constant does not appear in the covariant derivative. By writing the Lagrangian in this form we make it clear that $`1/g`$ is neither more nor less than a stiffness parameter. It informs us how big is the energetic cost to produce curvature in the gauge field.
The form of Eqn. 1 is uniquely fixed by a few abstract postulates of a very general character. These are $`SU(3)`$ gauge symmetry, together with the general principles of quantum field theory – special relativity, quantum mechanics, locality – and the criterion of renormalizability. It is renormalizability that forbids more complicated terms, such as an anomalous gluomagnetic moment term $`\overline{q}\sigma ^{\mu \nu }G_{\mu \nu }q`$.
Later I shall argue that it is very difficult to rigorously insure the existence of a quantum field theory that is not asymptotically free. Asymptotic freedom is a stronger requirement than renormalizability, and can only be achieved in theories containing nonabelian gauge fields. Thus one can say, without absurdity, that even our postulates of gauge symmetry and renormalizability are gratuitous: both are required for the existence of a relativistic local quantum field theory.
The apparent symmetry of Eqn. 1 is:
$$𝒢_{\mathrm{apparent}}=SU(3)^c\times SU(f)_L\times SU(f)_R\times U(1)_B\times U(1)_A\times _{\mathrm{scale}}^+,$$
(5)
together of course with Poincare invariance, P, C, and T. The factors, in turn, are local color symmetry, the freedom to freely rotate left-handed quarks among one another, the freedom to freely rotate right-handed quarks among one another, baryon number (= a common phase for all quark fields), axial baryon number (= equal and opposite phases for all left-handed and right-handed quark fields), and scale invariance.
The chiral $`SU(f)_L\times SU(f)_R`$ symmetries arise because the only interactions of the quarks, their minimal gauge couplings to gluons, are universal and helicity conserving. These chiral symmetries will be spoiled by non-zero quark mass terms, since such terms connect the two helicities. With quarks of non-zero but equal mass one would have only the diagonal (vector) $`SU(f)_{L+R}`$ symmetry, while if the quarks have unequal non-zero masses this breaks up into a product of $`U(1)`$s. Thus the choice $`m=0`$ can be stated as a postulate of enhanced symmetry. If the quarks masses are all non-zero, then P and T would be violated by a non-zero $`\theta `$ term, unless $`\theta =\pi `$. For massless quarks, all values of $`\theta `$ are physically equivalent, so we lose nothing by fixing $`\theta =0`$.
One could of course have written these chiral symmetries together with the two $`U(1)`$ factors as $`U(f)_L\times U(f)_R`$, but the unwisdom of so doing will become apparent momentarily.
Finally the $`_{\mathrm{scale}}^+`$ factor reflects that the only parameter in the theory, $`g`$, is dimensionless (in units with $`\mathrm{}=c=1`$, as usual). Thus the classical theory is invariant under a change in the unit used to measure length, or equivalently (inverse) mass. Indeed, the action $`d^4x`$ is invariant under the rescaling
$$x^\mu \lambda x^\mu ;A\lambda ^1A;\psi \lambda ^1\psi .$$
(6)
The actual symmetry of QCD, and of the real world, is quite different from the apparent one. It is
$`𝒢_{\mathrm{actual}}`$ $`=`$ $`SU(3)^c\times SU(f)_L\times SU(f)_R\times Z_A^f\times U(1)_B`$ (7)
$`(\mathrm{asymptotic}\mathrm{freedom},\mathrm{chiral}\mathrm{anomaly})`$
$``$ $`SU(3)^c\times SU(f)_{L+R}\times U(1)_B`$
$`(\mathrm{chiral}\mathrm{condensation})`$
$`=`$ $`SU(f)_{L+R}\times U(1)_B`$
$`(\mathrm{confinement}).`$
Let me explain this cascade of symmetry reductions.
In the first line of Eqn. 7, I’ve specified the subgroup of $`𝒢_{\mathrm{apparent}}`$ which survives quantization. The $`_{\mathrm{scale}}^+`$ of classical scale invariance is entirely lost, and the $`U(1)_A`$ of axial baryon number is reduced to its discrete subgroup $`Z_A^f`$. Both fall victims to the need to regulate quantum fluctuations of highly virtual degrees of freedom, as I shall elaborate below. The breaking of scale invariance is associated with the running of the effective coupling, asymptotic freedom, and dimensional transmutation. The breaking of axial baryon number is associated with the triangle anomaly and instantons. Thus these symmetry removals arise from dynamical features of QCD that reflect its deep structure as a quantum field theory.
In the second line of Eqn. 7, I’ve specified the subgroup of the symmetries of the quantized theory which are also symmetries of the ground state. This group is properly smaller, due to spontaneous symmetry breaking. That is, the stable solutions of the equations exhibit less symmetry than the equations themselves. Specifically, the ground state contains a condensate of quark-antiquark pairs of opposite handedness, which fills space-time uniformly. One cannot rotate the different helicities components independently while leaving the condensate invariant. The lightness of $`\pi `$ mesons, and much of the detailed phenomenology of their interactions at low energies, can be understood as direct consequences of this spontaneous symmetry breaking.
In the third line of Eqn. 7, I’ve acknowledged that local color gauge symmetry, which is so vital in formulating the theory, is actually not directly a property of any physical observable. Indeed, in constructing the Hilbert space of QCD, one must restrict oneself to gauge-invariant states. The auxiliary, extended Hilbert space that we use in perturbation theory does not have a positive-definite inner product. It’s haunted by ghosts. Moreover, unlike the situation for QED, one discerns in the low-energy physics of QCD no obvious traces of gauge symmetry. Specifically, there are no long-range forces, nor do particles come in color multiplets. This is the essence of confinement, a tremendously important but amazingly elusive concept, as we shall discover repeatedly in these lectures. (If you look only at the world, not at a postulated micro-theory, what exactly does confinement mean? Don’t fall into the trap of saying confinement means the unobservability of quarks – the quarks are a theoretical construct, not something you can observe (that’s what you said!).)
Clearly, a major part of understanding QCD must involve understanding how its many apparent symmetries are lost, or realized in peculiar ways. The study of QCD in extreme conditions gives us fruitful new perspectives on these matters, for we can ask new, very sharp questions: Are symmetries restored, or lost, in phase transitions? Are they restored asymptotically?
Now I shall discuss each of the key dynamical phenomena: asymptotic freedom, confinement, chiral condensation, and chiral anomalies, in more detail.
### 2.2 Asymptotic Freedom
#### 2.2.1 Running Coupling
Running of couplings is a general phenomenon in quantum field theory. Nominally empty space is full of virtual particle-antiparticle pairs of all types, and these have dynamical effects. Put another way, nominally empty space is a dynamical medium, and we can expect it to exhibit medium effects including dielectric and paramagnetic behavior, which amount (in a relativistic theory) to charge screening. In other words, the strength of the fields produced by a test charge will be modified by vacuum polarization, so that the effective value of its charge depends on the distance at which it is measured.
Asymptotic freedom is the special case of running couplings, in which the effective value of a charge measured to be finite at a given finite distance, decreases to zero when measured at very short distances. Thus asymptotic freedom involves antiscreening. Antiscreening is somewhat anti-intuitive, since it is the opposite of what we are accustomed to in elementary electrodynamics. Yet there is a fairly simple way to understand how it might be possible in a relativistic, nonabelian, gauge theory:
* Because the theory is relativistic, magnetic forces are just as important as electric ones.
* Because it is a gauge theory, it contains vector mesons, with spin.
* Because it is nonabelian these vector mesons carry charge, and their spins carry magnetic moments.
Now in electrodynamics we learn that spin response is paramagnetic – a spin (elementary magnetic dipole) tends to align with an imposed magnetic field, in such a way as to enhance the field. This, you’ll realize if you think about it a moment, is antiscreening behavior. Thus there is a competition between normal electric screening (together with orbital diamagnetism) and antiscreening through spin paramagnetism. For virtual gluons, it turns out that spin paramagnetism is the dominant effect numerically.
When the effective coupling becomes weak, one can calculate the screening (or antiscreening) behavior in perturbation theory. In QCD one finds for the change in effective coupling
$$\frac{dg(ϵ)}{d\mathrm{ln}ϵ}=\beta _0g^3+\beta _1g^5+\mathrm{}$$
(8)
with
$`\beta _0`$ $`=`$ $`{\displaystyle \frac{1}{16\pi ^2}}(11{\displaystyle \frac{2}{3}}f)`$ (9)
$`\beta _1`$ $`=`$ $`({\displaystyle \frac{1}{16\pi ^2}})^2(102{\displaystyle \frac{38}{3}}f),`$ (10)
where $`ϵ`$ is the energy, or equivalently inverse length, scale at which the effective charge is defined. Thus if the $`f16`$ the effective coupling decreases to zero as the energy at which it is measured increases to infinity, which is asymptotic freedom. Taking the first term only, we see that the asymptotic behavior is approximately
$$\frac{1}{g(ϵ)^2}=\frac{1}{g(1)^2}\beta _0\mathrm{ln}ϵ.$$
(11)
#### 2.2.2 Clouds, Jets, and Experiments
The asymptotic freedom of QCD can be exploited to simplify the calculation of many physical processes. To understand why that is so, consider the picture of quarks (or gluons) that it suggests. This is shown in Figure 1. The effective color charge of a quark, responsible for its strong interaction, accumulates over distance. We should think of it as a distributed cloud of induced color charge, all of the same sign, with no singular core. Now consider how this cloud is seen when probed at various wavelengths. If the wavelength is small, the effective color charge will be small, since only a small portion of the cloud is sampled. If the wavelength is large, the effective charge will be large.
One can also consider the complementary particle picture. If one suddenly imparts a large impulse to a quark (or gluon), liberating it from its cloud, the resulting object will have a small effective color charge, and will propagate almost freely, until it builds up a new cloud.
Of course the electric and weak charges of a quark are not shared by its color polarization cloud. These charges remain concentrated. Thus short-wavelength electromagnetic or weak probes can resolve pointlike quarks, whose ‘strong’ (color) interactions make only small corrections to free-particle propagation. More precisely, it is corrections that substantially change the energy-momentum of the color current which are guaranteed to be small. Big changes in the energy-momentum can only arise from radiation of gluons having short wavelength in the frame of the current, but such short-wavelength gluons see only a small effective color charge, as we’ve discussed.
So in physical process originating with the pointlike quarks directly produced by hard electroweak currents, as for instance in $`e^+e^{}`$ annihilation at high energy, one develops jets of rapidly-moving hadrons following the nominal paths of the original quarks. Hard gluon radiation processes do occasionally occur, of course, but at a small rate, calculable in perturbation theory. These hard gluons can (with small probability) induce jets of their own, which themselves occasionally radiate, and so forth. The “antenna pattern” of jets – including the relative probabilities for different topologies (numbers of jets), the energy and angular distributions for a given topology, and how all these quantities depend upon the total energy, can be calculated in exquisite detail. These predictions, which reflect in 1-1 fashion the structure of the fundamental interactions in the theory and the running of the coupling, have been extensively, and successfully, tested experimentally.
The science of exploiting asymptotic freedom to predict rates for a wide variety of hard processes is highly developed. Figure 2 displays results of comparisons between prediction and observation for a wide variety of experiments, in the form of determinations of the running coupling. Implicit, of course, is that within any given type of experiment, the theory gives a successful account of the functional form of dependence on event topology, energy distribution, angular distribution, … . The Figure itself demonstrates the success of Eqns. 8,9, suitably generalized to include the effect of quark masses, as a description of Nature.
#### 2.2.3 Dimensional Transmutation and “Its From Bits”
Running of the coupling manifestly breaks the classical scale invariance $`_{\mathrm{scale}}^+`$. It is an almost inevitable result of quantum field theory. Indeed, quantum field theory predicts as its first consequence the existence of a space-filling medium of virtual particles, whose polarization makes the effective coupling strength depend on the distance at which it is measured, or in other words causes the coupling to run. (There are very special classes of supersymmetric theories in which different types of virtual particles give cancelling effects, and the coupling does not run.)
Because its coupling runs, in QCD the analogue of Pauli’s question:
> “Why is the value of the fine structure constant what it is?” –
receives a startling answer:
> “It’s anything you like – at some scale or other.”
We can simply declare it to be, say, $`\frac{1}{10}`$, thereby defining the correct distance at which to measure it – i.e., the distance where it is $`\frac{1}{10}`$! This is the phenomenon of dimensional transmutation. A dimensionless coupling constant has been transmuted into a length (or, equivalently, energy) scale.
In fact we can identify a scale explicitly, using Eqn. 8:
$$M=\underset{ϵ\mathrm{}}{lim}ϵe^{\frac{u}{\beta _0}}u^{\frac{\beta _1}{\beta _0^2}}$$
(12)
where
$$u\frac{1}{g(ϵ)^2}.$$
(13)
The form of Eqn. 8 insures that the limit exists and is finite.
Due to its formal scale invariance, QCD with massless quarks appears naively – that is, classically – to be a on-parameter family of theories, distinguished from one another by the value of the coupling $`g`$. And none of these theories, it appears, defines a scale of distance. Through the magic of dimensional transmutation, QCD turns out instead to be a family of perfectly identical clones, each of which does define a distance scale. Indeed, the clones differ only in the units they employ to measure distance. This difference in units enters into comparisons of purely QCD quantities to quantities outside of QCD, such as the ratio of the proton mass to the electron mass. But it does not affect dimensionless quantities within QCD itself, such as ratios of hadronic masses or of isomer splittings.
Using $`\mathrm{}`$ and $`c`$ as units, and no further inputs the truncated version of QCD including just the $`u`$ and $`d`$ quarks, with their masses set to zero – what I call “QCD Lite” – accounts pretty accurately for the low-lying spectrum of non-strange hadrons (demonstrably), nuclear physics (presumably), and much else besides. The only numerical inputs to QCD Lite are the number of colors – 3 (binary 11) – and the number of quarks –2 (binary 10). Thus QCD Lite provides a remarkable realization of Wheeler’s slogan “Getting Its From Bits”!
#### 2.2.4 Limitations
The most straightforward applications of asymptotic freedom are to processes involving large energy scales only. This constraint pretty much limits one to inclusive processes induced by electroweak currents. Any external hadron introduces a small energy scale, namely its mass. By using clever tricks to isolate calculable subprocesses, one can calculate much more, as conveyed in Figure 2. However these tricks will take you only so far, and there is no easy way to exploit the small effective coupling at short distances to address truly low-energy phenomena, or to calculate the spectrum. If you try to calculate such things directly, you will encounter infrared divergences – so at least the theory is kind enough to warn you off. Similarly, large temperature or large chemical potential introduce large “typical” energies, but it is not trivial (and usually not even true) that this fact in itself will allow access, via asymptotic freedom, to interesting spectral or transport properties.
### 2.3 Confinement
#### 2.3.1 Brute Facts and Crude Theory
An aura of mystery still seems to hover around the phenomenon of quark confinement. Historically, of course, it was a big surprise in world-modeling, and posed a major barrier both to the discovery and to the acceptance first of quarks, and then of modern QCD. And confinement is a genuinely profound and subtle dynamical phenomenon, as some of our later considerations will emphasize. Concerning the fact that QCD predicts confinement, however, there is no ambiguity. Our knowledge of the properties of the theory has moved far beyond abstract or hypothetical discussion of this point. Figure 3 exhibits the results of some direct calculations of the spectrum starting from the microscopic theory, with controlled errors, using the techniques of lattice gauge theory. There are neither massless flavor-singlet particles with long-range interactions, particles with quark or gluon quantum numbers, nor degenerate color multiplets. In fact the microscopic theory reproduces the observed spectrum extremely well, with no gratuitous additions.
Thus confinement is not a practical problem for modern QCD. Still, one would like to understand more precisely what it is, why it occurs and, particularly for these lectures, whether in can come undone in extreme conditions.
The simplest heuristic argument for confinement is due to Amati and Testa. It is based directly on Eqn. 1. If the coefficient $`\frac{1}{g^2}`$ of the gauge curvature term is taken to zero, then upon varying the action with respect to the gauge potential $`A_\mu `$ we find that the color current vanishes, including its zero component, the color charge density – which is to say, color is confined. We can combine this idea with the running of the coupling, to argue (still more heuristically) that low-frequency modes, associated with large effective couplings, are confined, while high-frequency modes can be dynamically active. This is broadly consistent with the observed behavior in Nature, that quarks and gluons become visible in hard processes, but are not accessible to soft probes.
Unfortunately it is a very singular operation to throw away the highest-derivative terms in any differential equation, because they will always dominate for sufficiently abrupt variation. In high Reynolds number hydrodynamics, one addresses this difficulty with boundary layer theory. In QCD, the only method we know to make the simple Amati-Testa argument the starting point for a systematic approximation is by employing an artful discretization, lattice gauge theory, as I’ll now discuss.
#### 2.3.2 Lattice Gauge Theory Basics
The great virtue of the lattice version of QCD are that it provides an ultraviolet cutoff and that it allows a convenient strong-coupling expansion, while preserving a very large local gauge symmetry. Its drawbacks are that it destroys translation and rotation symmetry, that it has an awkward weak-coupling expansion, and that it mutilates the ultraviolet behavior of the continuum theory. But let’s put these worries aside for the moment, and exploit the virtues. Also let’s consider the pure glue theory. The extension to quarks will appear in Lecture 2.
The fundamental operation in gauge theory is parallel transport. The basic objects of the theory are 3$`\times `$3 unitary matrices with unit determinant. They live on the oriented links of a cubic hyperlattice in four dimensions, and implement parallel transport. Thus the dynamical variable $`U_{r,\widehat{\mu }}`$ is a matrix associated with the link starting from lattice point $`r`$ and ending at $`r+\widehat{\mu }a`$, where $`a`$ is the lattice spacing. The matrix associated with the same link oriented in the opposite direction is the inverse, i.e. $`U_{r+\widehat{\mu }a,\mu }=U_{r,\widehat{\mu }}^1`$.
If there were an underlying continuum gauge potential $`A`$, the parallel transport would be
$$U_{r,\widehat{\mu }}\text{}=\text{}\overline{P}\mathrm{exp}i_r^{r+\widehat{\mu }a}A_\mu 𝑑x^\mu ,$$
(14)
where $`\overline{P}`$ denotes anti-path ordering. Indeed, this is the solution of the equation
$$_\mu U(_\mu +iA_\mu )U=0$$
(15)
having the indicated endpoints. In line with this underlying structure, which we would like to recover in an appropriate limit, local gauge transformations $`\mathrm{\Omega }(r)`$ should act on the lattice sites, and transform the $`U`$ matrices according to
$$\stackrel{~}{U}_{r,\mu }=\mathrm{\Omega }(r)U_{r,\mu }\mathrm{\Omega }(r+\widehat{\mu }a)^1.$$
(16)
The $`\mathrm{\Omega }(r)`$ are, of course, 3$`\times `$3 unitary matrices.
To form interaction terms invariant under Eqn. 16 one must take the trace of a product of $`U`$ matrices along a path of links forming a closed loop. The simplest possibility is simply the trace around a plaquette, e.g.
$$\mathrm{tr}\mathrm{𝑃𝑙}_{r,\widehat{x}\widehat{y}}\mathrm{tr}U_{r,\widehat{x}}U_{r+a\widehat{x},\widehat{y}}U_{r+a\widehat{x}+a\widehat{y},\widehat{x}}U_{r+a\widehat{y},\widehat{y}}.$$
(17)
Putting Eqn. 14 into Eqn. 17 and expanding for small $`a`$, we find the first non-trivial term
$$\mathrm{tr}\mathrm{𝑃𝑙}_{r,\widehat{x}\widehat{y}}\mathrm{tr}(1\frac{a^4}{2}G^{xy}G_{xy})+O(a^6).$$
(18)
The straightforward verification of Eqn. 18 is quite arduous, but one can restrict its form a priori by exploiting gauge invariance, and then evaluate on a simple configuration such as $`A_x=M(yr_y)`$. Note that a term $`a^2G^{xy}`$, which does appear in the plaquette product, vanishes when we take the trace.
Thus the simplest lattice gauge invariant action we can write, the Wilson action
$$S_W=\frac{1}{4g^2}\underset{\mathrm{plaquettes}}{}(3\mathrm{tr}\mathrm{𝑃𝑙}_{\mathrm{}})$$
(19)
reduces, formally, to the continuum action for very small lattice spacings. Hence we are invited to use it, and then attempt to justify the limit. (In the spirit of the Jesuit credo “It is more blessed to ask forgiveness than permission.”)
To evaluate a correlation function of operators $`𝒪_1𝒪_2\mathrm{}`$, then, we must evaluate the integral
$$\frac{𝒪_1𝒪_2\mathrm{}=𝑑\mathrm{exp}(S_W)𝒪_1𝒪_2\mathrm{}}{𝑑\mathrm{exp}(S_W)},$$
(20)
where
$$d=_{\mathrm{links}r,\widehat{\mu }}U_{r,\widehat{\mu }}^1dU_{r,\widehat{\mu }}$$
(21)
is the product of invariant integrals over the gauge group.
#### 2.3.3 Strong Coupling and Confinement
The lattice regularization permits one to formulate strong coupling perturbation theory in a simple, elegant way. When $`g`$ gets large, we can simply expand $`e^{S_W}`$ in a power series in $`\frac{1}{g^2}`$. The result is to “bring down” activated plaquettes, one for each inverse power of $`\frac{1}{g^2}`$. When we integrate over a link incident on an activated plaquette, we encounter an extra power of $`U`$ for that link.
To test for confinement, the traditional method is to study Wilson-Polyakov loops. These are simply the traces of products of $`U`$ matrices around loops, similar to the terms that appear in the action. But now we want to consider large loops, instead of very small ones.
The motivation for this method is as follows. Suppose we put a very heavy quark into the system. This will stay at a fixed point in space, so its world-line will be simply a straight line in the $`\widehat{\tau }`$ (Euclidian time) direction. The $`\mathrm{tr}jA`$ coupling of this color source will generate a product of $`U`$ matrices along the links of its world-line. So to measure the potential between a heavy quark and a heavy antiquark at distance $`R`$ we should measure the energy it takes to have a line like this and a similar line with $`U^1`$ matrices a distance $`R`$ away. If we allow this configuration to persist for a long Euclidian time $`T`$, the cost should go as $`e^{V(R)T}`$. Now to make the “measurement” clean we should imagine closing up the loop with short segments at the top and bottom. This corresponds to producing the quark-antiquark pair, letting them sit separated for a long time, and then annihilating them. With $`R<<T`$, by taking the negative of the log, we will extract the potential. In a formula
$$V(R)=\underset{T\mathrm{}}{lim}\frac{1}{T}\mathrm{ln}\frac{𝑑\mathrm{exp}(S_W)\mathrm{\Pi }}{𝑑\mathrm{exp}(S_W)}.$$
(22)
where $`\mathrm{\Pi }`$ is the trace of an ordered product of $`U`$ matrices along the perimeter of a long rectangle with sides of length $`R`$ and $`T`$, as shown in Figure 4a.
With this background in hand, it becomes very easy to understand how confinement arises in strong coupling. The integral of a single $`U`$ matrix over the group manifold, $`U^1𝑑UU`$, vanishes. (The change of variables $`UgU`$ leaves the measure invariant, so if $`U^1𝑑UU=k`$, then by changing variables we find $`gk=k`$ for any $`g`$, which of course means $`k=0`$.) So does $`U^1𝑑UU^1`$. Thus to find a non-zero contribution to Eqn. 22 in strong coupling we must at least pull down plaquettes to share sides with each of the links in the Wilson loop.
But now you see there’s a whole new set of interior links with single $`U`$ matrices and vanishing group integrals. Clearly, to get a non-zero contribution we must tile the whole area spanned by the Wilson loop, as in Figure 4b. This takes a number of plaquettes proportional to the area $`RT`$, at least. So the leading contribution to the Wilson loop in strong coupling goes as
$$V(R)\frac{1}{T}\mathrm{ln}(\frac{1}{g^2})^{RT}R.$$
(23)
The linear potential, of course, means that it is impossible to separate the color sources indefinitely, and so one has confinement.
#### 2.3.4 To the Continuum Limit
The strong coupling result forms the starting point for a convincing proof of confinement in (pure glue) QCD proper.
One first argues that the strong coupling perturbation theory has a finite radius of convergence. That can be done analytically. Then one investigates numerically whether there is a phase transition as a function of the coupling, as the coupling varies from strong to weak. It turns out there is a phase transition for $`U(1)`$, but not for $`SU(3)`$ (or $`SU(2)`$). When there is no phase transition, the theory remains in the same universality class, and its sharply defined qualitative properties cannot change.
Thus in the physically relevant $`SU(3)`$ case:
* Since the strong coupling perturbation expansion converges, the lowest non-trivial order governs the asymptotic behavior of the Wilson loop, and exhibits confinement.
* Since the strong coupling theory is in the same universality class as the weak coupling theory, the weak coupling theory also exhibits confinement.
* Since asymptotic freedom implies that the weak coupling lattice theory reproduces the continuum theory, the continuum theory exhibits confinement.
Now we see that is fortunate, and reassuring for this circle of ideas, that there is a phase transition for $`U(1)`$. Otherwise we’d have proved confinement in QED, which would be proving too much.
#### 2.3.5 Foundational Remarks
To round out this discussion I would like to emphasize the deep connections among renormalizability, asymptotic freedom, and lattice gauge theory. To construct a relativistic quantum theory, one typically introduces at intermediate stages a cutoff, which spoils the locality or relativistic invariance of the theory. Then one attempts to remove the cutoff, while adjusting the defining parameters, in order to achieve a finite, cutoff-independent limiting theory. Renormalizable theories are those for which this can be done, order by order in a perturbation expansion around free field theory. That formulation of the problem of constructing a quantum field theory, while convenient for mathematical analysis, obviously begs the question whether this perturbation theory converges. For interesting quantum field theories, it rarely does.
A more straightforward procedure, conceptually, is to regulate the theory as a whole by discretizing it. This involves approximating space-time by a lattice, and spoils the continuous space-time symmetries of the theory. Then one attempts to remove dependence on the discretization, by refining it, while if necessary adjusting the defining parameters, to achieve a finite limiting theory that does not depend on the discretization, and therefore has a chance to respect the space-time symmetries. The redefinition of parameters is necessary, because in refining the discretization one is introducing new degrees of freedom. The earlier, coarser theory results from integrating out these degrees of freedom. If it is to represent the same physics it must incorporate their effects, for example in vacuum polarization. Operationally, one can demand that some observable(s) measured at scales well beyond the lattice spacing stays fixed as the discretization is refined. This fixes the free coupling(s). The question is then whether, having fixed the available parameters, the calculated values of all observables have finite limits.
This is very hard to prove, in general. The only case in which it is straightforward arises when the effects of integrating out the additional short-wavelength modes, that are introduced with each refinement of the lattice, can be captured accurately by a re-definition of the coupling parameter(s) already present in the theory. That, in turn, will occur in a straightforward way only if these modes are weakly coupled. For then perturbation theory will show us how to take the limit for the renormalizable couplings, while assuring us that naive power counting can be applied to argue away all non-renormalizable ones. But of course the ultraviolet modes will be weakly coupled, if and only if the theory is asymptotically free.
Summarizing the argument, only those relativistic field theories which are asymptotically free can be argued in a straightforward way to exist. Furthermore, the only asymptotically free theories in four space-time dimensions involve nonabelian gauge symmetry, with highly restricted matter content. So the axioms of gauge symmetry and renormalizability which we invoked to define QCD are, in a certain sense, redundant. They are implicit in the mere existence of non-trivial interacting quantum field theories.
### 2.4 Chiral Symmetry Breaking
#### 2.4.1 Numerical and Laboratory Phenomena
The most direct evidence for chiral symmetry breaking in QCD comes form numerical simulation of the theory. One simply computes the expectation value
$$\overline{q_L}_iq_R^jv\delta _i^j0$$
(24)
in the ground state for the theory with massless quarks. This condensation, which breaks the chiral symmetry of the equations, is entirely analogous to the development of spontaneous magnetization in a ferromagnet. As in that case, for any finite sample (e.g., in any simulation) we must add an infinitesimal biasing field to stabilize a particular alignment.
In Eqn. 24 I’ve chosen to align in the flavor diagonal direction, but in the absence of a biasing field any chirally rotated configuration, with $`q_RUq_R`$, will have the same energy but $`\delta _j^iU_j^i`$ in the expectation value, for any $`UϵSU(f)`$. There are several technical issues in the simulations that arise and must be addressed, but the numerical evidence that chiral symmetry is spontaneously broken is unambiguous and overwhelming, at least for $`2f4`$. I’ll discuss this evidence in more detail in Lecture 2.
The historical path whereby spontaneous chiral symmetry was discovered as a property of Nature was of course quite different. Indeed, the discovery of chiral symmetry breaking in the strong interaction antedates by more than a decade the discovery of QCD as its microscopic theory.
The conceptual starting point for the historic development was the observation, coming into focus with the BCS theory of superconductivity, that if a symmetry is spontaneously broken there will be massless collective modes associated with this breakdown. (The “experimental” starting point was the Goldberger-Treiman relation; see below.) Quite generally, suppose the ground state of a physical system (e.g. a ferromagnet or the no-particle state of QCD at zero temperature) is characterized by the existence of a condensate
$$\eta |_b^a|\eta =\eta _b^a$$
(25)
that violates a continuous symmetry of the underlying equations (e.g. rotational or chiral symmetry, respectively). Let the symmetry $`g`$ of the underlying theory be implemented by the unitary operator $`U(g)`$, with $`U(g)^1MU(g)=\rho (g)M`$. Then if $`\rho (g)\eta \eta `$, which is the signature of symmetry breaking, the states
$$U(g)|\eta |\rho (g)\eta $$
(26)
will be physically distinct from the ground state, but energetically degenerate with it. By moving slowly within this manifold of states, as a function of space, we would expect to create states whose energy goes to zero as the wavelength of the variation goes to infinity. In a particle interpretation, the quanta of the field that creates such configurations will be massless. Furthermore, we have constructed a very specific realization of these quanta in terms of symmetry generators. This construction can be exploited to yield predictions for their properties. For example, if the broken symmetry is an internal symmetry, the quanta will be spin-0 particles.
Turning now specifically to the world of strong interactions, it is a striking fact that $`\pi `$ mesons are spin-0 particles which are much lighter than all other hadrons. This suggests the possibility that they are associated with the spontaneous breakdown of an approximate internal symmetry. Their pseudoscalar character, and the fact that they form an isotriplet, suggests the breaking pattern
$$SU(2)_L\times SU(2)_RSU(2)_{L+R}.$$
(27)
This very specific picture of pions as collective modes closely connected to broken chiral symmetry, besides explaining their quantum numbers and small mass, can be exploited to give many predictions about their low-energy behavior, as we shall discuss in Lecture 2. The phenomenological success of these predictions validates the hypothesis of spontaneously broken approximate chiral symmetry as a description of Nature.
Within QCD, this picture arises very naturally. If the $`u`$ and $`d`$ masses are small the basic equations of QCD will exhibit approximate chiral symmetry. And numerical work QCD spontaneously develops a symmetry-breaking condensate, as I mentioned. The general theoretical machinery for extracting predictions from spontaneous symmetry breaking remains valid and extremely valuable in modern QCD. Additionally, the specific form of intrinsic breaking in QCD, through small quark mass terms, has specific phenomenological consequences. I will spell out how all this works below in Lecture 2, when we discuss order parameters and effective Lagrangians. As we shall see, all these concepts take on additional twists, and become even more central, for QCD in extreme conditions.
#### 2.4.2 Ironic Aside
Ironically, the first generation of developments in high-energy physics to be inspired by modern superconducitivity theory were inspired by BCS pairing theory in the limit that the gauge coupling – that is, electromagnetism, and hence the phenomenon of superconductivity – is neglected. In that limit it is the global symmetry of electron number that is violated by the formation of a Cooper pair condensate, and there is a massless collective mode. Spontaneous breaking of a global symmetry turns out to be the appropriate, and fruitful, idea for chiral symmetry breaking in the strong interaction.
There was an interval of several years before the second generation of developments, when the gauge coupling was reinstated. Only then did the primary phenomenon of superconductivity itself – the Meissner effect – enter the picture. Rechristened in its new context as the Higgs mechanism, it of course became central to modern electroweak interaction theory.
#### 2.4.3 Pairing Heuristics
Just as for confinement, the fact of spontaneous chiral symmetry breaking in QCD is no longer negotiable. Still, just as for confinement, one would like to understand why and how it occurs, and whether there are circumstances in which it can come undone.
A heuristic model for chiral symmetry breaking was supplied by Nambu and Jona-Lasinio long before modern QCD. Amazingly, with some re-labeling of the players the concepts they introduced still apply. Indeed, as we shall see, at high density they come to look better than ever. We shall be discussing pairing theory in great detail in Lecture 4, so this is just a foretaste.
Suppose one has an attractive four-fermion interaction
$$_{\mathrm{int}.}=g(\overline{\psi }\psi )(\overline{\psi }\psi ).$$
(28)
Then one can imagine that it is energetically favorable to form a condensate
$$\overline{\psi }\psi =v0,$$
(29)
since this condensate generates negative interaction energy. Indeed, if the condensate is so large that we can ignore quantum fluctuations, we shall have the condensation energy density
$$\mathrm{\Delta }=\mathrm{\Delta }=g\overline{\psi }\psi ^2$$
(30)
To test this idea in the simplest crude way, write
$$(\overline{\psi }\psi )^2=(\overline{\psi }\psi v)^2+2v\overline{\psi }\psi v^2$$
(31)
and, in the interaction Lagrangian, discard the fluctuating first term (as is approximately valid at weak coupling). The other terms, when added to the standard kinetic energy term for massless fermions, generate the Lagrangian for free massive fermions. One can of course diagonalize this quadratic approximate Lagrangian and, by filling the negative energy sea, construct the appropriate zero fermion number density ground state (i.e., for a given value of the condensate). Then one can enforce consistency by calculating $`\overline{\psi }\psi `$ in this state, and demanding that it is equal to the originally assumed value $`v`$. This consistency equation is called the gap equation, in honor of its ancestor in BCS theory. If the gap equation has a non-trivial solution, one will have lowered the energy by forming a condensate.
In QCD, the one-gluon exchange interaction is quite attractive in the quark-antiquark color singlet channel. This is hardly surprising, since by forming a singlet one cancels the charge and eliminates field energy. To make a scalar condensate in this channel, one must pair left-handed antiquarks with right-handed quarks. So there are simple heuristic reasons to anticipate the possibility of spontaneous chiral symmetry breaking in QCD.
Whether spontaneous chiral symmetry breaking actually occurs, however, is a delicate question, because it involves a competition. The interaction energy one gains by pairing up occupied particle and antiparticle modes must compete against the kinetic energy lost in occupying them. The kinetic energy can become arbitrarily small, but only for a density of states that likewise vanishes – the tip of the Lorentz cone as shown in Figure 5. So whether the interaction energy ever wins out, or not, is a delicate dynamical issue. For example, there is some evidence that as the number of flavors $`f`$ grows the strength of chiral condensation (relative the the primary QCD scale) shrinks, and that for large enough $`f`$ (6? 7?) it’s gone. This comes about, presumably, because for larger $`f`$ the effective coupling does not run so quickly to large values at small energy.
It’s quite a different story at high density. In that case the density of states does not vanish, and spontaneous chiral symmetry breaking can appear at arbitrarily weak coupling, where we have excellent analytic control.
### 2.5 Chiral Anomalies and Instantons
#### 2.5.1 The Historic Case
The original discovery of the chiral anomaly involved what in hindsight is a fairly complicated example of the phenomenon. It came about not through abstract investigation of mathematical models, but in the process of analyzing a very specific physical process, the decay $`\pi ^0\gamma \gamma `$. This is obviously an electromagnetic decay, so to treat it we must consider QCD coupled to QED. The combined theory, though it violates isospin, still appears to be invariant under axial $`I_3`$ symmetry. (I will work, for simplicity, in the limit of vanishing $`u`$ and $`d`$ quark masses. This can be shown to be a good approximation for estimating the $`\pi \gamma \gamma `$ vertex, though of course the actual $`\pi ^0`$ mass must be inserted when we use this amplitude to calculate the decay rate.) If this were true, then the $`\pi ^0`$ would still be accurately a collective Nambu-Goldstone mode associated with the spontaneous breaking of axial $`I_3`$ symmetry. The coupling $`\pi ^0F^{\mu \nu }\stackrel{~}{F}_{\mu \nu }`$ would be forbidden, because at long wavelength such a Nambu-Goldstone mode is derivatively coupled to the corresponding current. Of course one can rewrite $`\pi ^0F^{\mu \nu }\stackrel{~}{F}_{\mu \nu }\frac{1}{2}^\mu \pi ^0A^\nu \stackrel{~}{F}_{\mu \nu }`$ inside the Lagrangian, after an integration by parts. However the electromagnetic term is not part of the axial $`I_3`$ symmetry current, classically.
The brilliant result of Adler, Bell and Jackiw is that when one investigates the situation more deeply, using the full resources of quantum field theory, such a term does in fact occur. (Again, the original analysis antedates QCD, and therefore was couched in rather different language. Its details are fascinating and of considerable historical interest, but I will not discuss them here.) Moreover its coefficient can, given an underlying theory of the strong interaction, be calculated precisely. When the coefficient is calculated in a free quark model – with colored, fractionally charged quarks – one finds agreement between the predicted rate for $`\pi ^0\gamma \gamma `$ and experiment. Remarkably, the free-quark result remains valid in QCD.
The mechanism whereby the extra term is generated is quite subtle. It is most readily seen in perturbation theory, though a non-perturbative derivation is possible.
To regulate the triangle graph and other loop graphs with circulating fermions it is convenient to follow the procedure of Pauli and Villars. In this procedure, one introduces into the theory fictitious boson fields $`\varphi `$ carrying all the same quantum numbers as the fermions (including spin $`\frac{1}{2}`$, and so flouting the spin-statistics theorem) but having a large mass $`M`$. Since the boson loops differ in sign from the fermion loops their contributions will cancel at very large virtual momentum, where the divergences arose. Thus one achieves finite integrals. Then one takes the limit $`M\mathrm{}`$, fixing a few low-energy amplitudes by setting them equal to their physical values. The basic result of renormalization theory is that in suitable (renormalizable) quantum field theories, having used a small number of low-energy amplitudes, to determine the couplings – which will be functions of the cutoff and a few physically determined parameters – all remaining physical amplitudes will approach a finite limit, order by order in perturbation theory. Of course the limiting amplitudes no longer contain contributions arising from the fictitious bosons as intermediate states, because those particles have been driven to infinite mass.
This is not the place to review the technicalities of renormalization theory. Fortunately, for present purposes we don’t need their details. Our focus is simply the triangle graph for the two-photon matrix element of the divergence of the axial current. This graph, by naive power counting, is linearly convergent in the ultraviolet by power counting. In more detail, there are three fermion propagators, and gauge invariance pulls out two powers of momentum to make $`F`$s from $`A`$s. However, one will only be able to use gauge invariance if one has employed a gauge invariant regulator, like Pauli-Villars.
By the same power counting, the leading $`M`$ dependence of the regulated integral, which arises from the Pauli-Villars boson loop, goes as $`\frac{1}{M}`$ for large $`M`$. This might seem to make it negligible. The large mass $`M`$, however, involves a large violation of chiral symmetry. Specifically, it implies a large coefficient $`M`$ in the divergence of the axial current:
$$_\mu (\overline{\varphi }\gamma ^\mu \gamma ^5\varphi )=2M\overline{\varphi }\gamma ^5\varphi .$$
(32)
(To have a regulated chiral symmetry that will behave sensibly as we remove the cutoff, the Pauli-Villars fields must transform in the same way as the quark fields they regulate.) This factor exactly cancels the convergence factor, leaving a finite residual contribution. The regulator, necessary to control the contribution of highly virtual quarks, spoils naive chiral symmetry. That is the essential mechanism of the chiral anomaly.
#### 2.5.2 An Easier Version
The basic mechanism leading to anomalies is nicely illustrated, in a context where it is relatively easy to understand, by the process $`hGG`$ of Higgs particle decay into two gluons. Since this process is of independent and timely interest, I hope you’ll forgive a brief diversion.
The coupling of the standard model Higgs boson to quarks is given as
$$_{\mathrm{int}}=2^{\frac{1}{4}}G_F^{\frac{1}{2}}hm_i\overline{q}_iq_i,$$
(33)
where $`G_F`$ is the Fermi coupling constant. The direct coupling to ordinary matter would seem to be extremely feeble, due to the very small masses of the $`u`$ and $`d`$ quarks. Ordinarily one would expect that the contribution from the heavy quarks would be suppressed, according to general “decoupling” theorems. Indeed, we’d be in pretty bad shape in physics if we always had to worry about big contributions to low-energy amplitudes from (potentially unknown) heavy particles. However there is an exception of sorts here, because the coupling grows with the mass. Thus the contribution to the dimension 5 induced interaction term
$$_{\mathrm{induced}}=\kappa h\mathrm{tr}G^{\mu \nu }G_{\mu \nu }$$
(34)
arising through a heavy quark loop of circulating heavy quarks in the triangle graph of Figure 6, which by power counting one might expect to be inversely proportional to the mass of the quark, is instead unsuppressed. (Of course, this time the external legs are a Higgs particle and two gluons.) This is very similar to what we saw for the triangle anomaly, but now with real heavy particles rather than virtual regulators. One finds
$$\kappa =2^{\frac{1}{4}}G_F^{\frac{1}{2}}\frac{g^2}{48\pi ^2}$$
(35)
per heavy quark, in the large mass limit ($`m_h<m_q`$). For values $`90\mathrm{Gev}<m_h<150\mathrm{Gev}`$ of the Higgs mass most interesting for ongoing searches this “anomalous” gluon coupling generates both the dominant mechanism for hadronic production of $`h`$ particles, and a significant decay mode for them.
#### 2.5.3 The Saga of Axial Baryon Number
For concreteness in this part I’ll mostly take $`f=2`$ and refer to the quarks as $`u`$ and $`d`$. This simplifies the notation, loses nothing essential, and is close to reality.
The spontaneous breakdown of approximate chiral $`SU(2)_L\times SU(2)_RSU(2)_{L+R}`$ is associated with an extensive, successful phenomenology. At the level of quarks in QCD, we understand it as the result of the development of a condensate $`\overline{u_L}u_R=\overline{d_L}d_R`$. This condensate also violates the axial baryon number symmetry under $`(u_L,d_L)e^{i\alpha }(u_L,d_L)`$, $`(u_R,d_R)e^{i\alpha }(u_R,d_R)`$, which is a symmetry of Eqn. 1. One might expect, then, a light Nambu-Goldstone boson with the associated properties – a light, flavor-singlet pseudoscalar with highly constrained couplings. Alas, there is no such particle. Thereby hangs a tale.
The first observation is that there is an “anomalous” contribution to the divergence of the axial baryon number current, arising from the triangle graph of Figure 6. There’s a new set of players – the axial baryon current instead of axial $`I_3`$, and gluons instead of photons – but they follow the same script. Thus we find
$$_\mu j_B^{\mu 5}_\mu (\overline{u}\gamma ^\mu \gamma ^5u+\overline{d}\gamma ^\mu \gamma ^5d)=\frac{g^2}{4\pi ^2}\mathrm{tr}\stackrel{~}{G}_{\mu \nu }G^{\mu \nu },$$
(36)
where $`\stackrel{~}{G}_{\mu \nu }\frac{1}{2}ϵ_{\mu \nu \alpha \beta }G^{\alpha \beta }`$ is the dual field strength. The current $`j_B^{\mu 5}`$ is not conserved. Naive axial baryon number symmetry, generated by the spatial integral of $`j_B^{05}`$, is spoiled by an anomaly.
However, the existence of this anomaly does not in itself remove the problematic mode. For one has
$$\mathrm{tr}G^{\mu \nu }\stackrel{~}{G}_{\mu \nu }=_\mu K^\mu $$
(37)
where
$$K_\mu =\frac{1}{2}ϵ_{\mu \alpha \beta \gamma }\mathrm{tr}(A^\alpha ^\beta A^\gamma +\frac{2}{3}A^\alpha A^\beta A^\gamma ).$$
(38)
Thus it would appear that a modified symmetry, generated by the charge associated with the modified conserved current $`j_\mu ^{A5}K_\mu `$, is spontaneously broken, leaving us not much better off than before, still with the mistaken prediction of an extra light flavor singlet pseudoscalar.
Fortunately, though, Eqn. 37 is itself problematic. The point is that $`K_\mu `$ is a gauge-dependent quantity, so that in principle it can be singular without implying the singularity of any physical observable.
We can be more specific about this in the context of a path integral treatment of the theory. In such a treatment we express quantum amplitudes as an integral of contributions from different classical field configurations. To have reasonable control of the functional integrals – to have a measure that is damped for large field strengths – we must consider the Euclidian form of the theory, rotating to imaginary values of the time. In this framework, consider the behavior of the gauge potentials $`A_\mu `$ at spatial infinity. For the integral of $`K_\mu `$ to acquire a non-vanishing surface term, and thereby violate the formal conservation law, requires that $`A_\mu (x)\frac{M_\mu (\widehat{x})}{|x|}`$. This is not allowed for generic forms of $`M_\mu (\widehat{x})`$, since a non-trivial field strength $`G\frac{1}{r^2}`$ would result, leading to a logarithmically divergent action. But for special values of the boundary conditions $`M_\mu (\widehat{x})`$ these can cancel, leaving behind a finite-action contribution to the functional integral that contributes to $`_\mu K^\mu `$.
In the weak-coupling limit, we look for the field configurations that contribute to the amplitude of interest that have the smallest possible action. For amplitudes that violate axial baryon number (non-vanishing $`G^{\mu \nu }\stackrel{~}{G}_{\mu \nu }`$) these configurations are instantons, much-studied classical solutions of the Euclidian Yang-Mills equations.
A very important consistency check is that the integral $`\mathrm{tr}G^{\mu \nu }\stackrel{~}{G}_{\mu \nu }`$, which according to the anomaly equation measure the violation of axial baryon number, must turn out to be quantized (that is, come in discrete units – it’s a c-number integral). Indeed, since axial baryon number is quantized, changes in axial baryon number had also better be quantized. The facts we have discussed, that this integral can be thrown onto the surface at infinity and that its finiteness requires special conspiracies, suggest a connection to topology and the possibility of its quantization. These features can indeed be demonstrated, but I won’t do that here.
In view of the easily proved inequality
$$\mathrm{tr}G^{\mu \nu }G_{\mu \nu }|\mathrm{tr}G^{\mu \nu }\stackrel{~}{G}_{\mu \nu }|$$
(39)
quantization of the right-hand side implies that the action of any configuration leading to axial baryon number violation is bounded below by a finite constant. Thus in the function integral such configurations are suppressed by a factor $`e^{\frac{1}{g^2}\mathrm{const}.}`$. In particular, they vanish to all orders in perturbation theory!
Instantons violate axial baryon number, but none of the other chiral flavor symmetries. This allows us to visualize their effect quite simply, at a heuristic level. The flavor structure must contain a product of determinant factors
$$_{\mathrm{inst}.}ϵ_{ij}q_L^iq_L^jϵ^{kl}\overline{q_R}_k\overline{q_R}_l.$$
(40)
For $`f`$ flavors, we’d have $`f`$ lefties in and $`f`$ righties out. The structure of the color and spin indices, and the form-factor accompanying this interaction, are more tricky to work out. To do it one must solve for the fermion zero-modes in the presence of an instanton. The resulting effective non-local “‘tHooft interaction” is represented pictorially in Figure 7: lots of fermions emerging from an extended gluon cloud.
Unfortunately, for reasons we have discussed before, the formal weak-coupling limit that leads us to focus on instantons is not under control for QCD in vacuum. So the arguments used in this section, while certainly suggestive, cannot be made quantitative in a convincing way. (Nevertheless, some reasonably successful phenomenology has been done by the saturation of the functional integral by superpositions of instantons and anti-instantons as a starting point.) Remarkably, under extreme conditions the theoretical situation for axial $`U_A(1)`$ breaking comes under much better control, with interesting consequences, as we’ll see.
## 3 Lecture 2: High Temperature QCD: Asymptotic Properties
### 3.1 Significance of High Temperature QCD
In this lecture I shall be discussing the behavior of QCD (and some closely related models) at high temperature and zero baryon number density. Since that is a mouthful I’ll just say high temperature.
The high temperature phase of QCD is of interest from many points of view. First of all, it is the answer to a fundamental question of obvious intrinsic interest: What happens to empty space, if you keep adding heat? Moreover, the high density phase of QCD was (almost certainly) the dominant form of matter during the earliest moments of the Big Bang. Moreover, it is a state of matter one can hope to approximate, and study systematically, in heavy ion collisions. Major efforts are being directed toward this goal and significant, encouraging results have already emerged. One can also simulate many aspects of the behavior with great flexibility and control, from first principles, using the techniques of lattice gauge theory. One can also make progress analytically. So there is a nice interplay among physical experiments, numerical experiments, and theory.
The fundamental theoretical result regarding the asymptotic high temperature phase is that it becomes quasi-free. That is, one can describe major features of this phase quantitatively by modeling it as a plasma of weakly interacting quarks and gluons. In this sense the fundamental degrees of freedom of the microscopic Lagrangian, ordinarily only indirectly and very fleetingly visible, become manifest (or at least, somewhat less fleetingly visible). Likewise the naive symmetry of the classical theory which, as we saw in Lecture 1, is vastly reduced in the familiar, low-temperature hadronic phase, gets restored asymptotically. In particular, chiral symmetry is restored, and confinement comes completely undone. Axial baryon number and scale symmetry, though never precisely restored, become increasingly accurate.
Since there are dramatic qualitative differences between the zero-temperature and the high-temperature phases, the question naturally arises whether there are sharp phase transitions separating them, and if so what is their nature. This turns out to be a rich and intricate story, whose answer depends in detail on the number of colors and light flavors. In the course of addressing it, we shall have to refine and modify common, rough intuitions about chiral symmetry and (especially) confinement. After an involved but I think interesting and coherent story, building up from the study of various idealizations, we shall find that there is, plausibly, a true phase transition in real QCD that we can converge upon from several directions – experimentally, numerically, and analytically.
### 3.2 Numerical Indications for Quasi-Free Behavior
For technical reasons it has been difficult until recently to simulate QCD including dynamical quarks with realistically small masses. That situation is changing, but it will be a few years before accurate quantitative results for thermodynamic quantities for QCD with light dynamical quarks become available.
Fortunately, we can already learn a lot from the existing simulations using pure glue or glue plus moderately massive quarks.
Representative results for the temperature dependence of the energy density and pressure in the two flavor theory are shown in Figure 8. Clearly, there is a rapid crossover in the behavior, with dramatic rises in the energy density and pressure (even when normalized to $`T^4`$) over a small range of temperatures around 150 Mev.
A notable feature of the numerical results is that while the energy density (divided by $`T^4`$) ascends rapidly to something close to its asymptotic value, the pressure appears much more sluggish. Thus the behavior of the plasma, even in regard to this basic bulk property, differs significantly from a free gas of massless particles. It is a worthy challenge to compute the corrections to free behavior analytically in weak coupling. This is not entirely straightforward, due to the absence of magnetic screening in perturbation theory (concerning which, more below). For recent progress see .
In reality the only hadrons light on the scale of the computed cross-over temperature are pions. Thus for temperatures significantly below this temperature (say $`T120\mathrm{Mev}`$) one has a rather dilute gas of pions, with 3 massive degrees of freedom for the three possible charge states of these spinless particles. Asymptotically, on the other hand, one has a gas with three different flavors of quarks, each of which comes with two spins, three colors, and antiquarks. Also there are eight gluons, each with two helicities. Thus the number of degrees of freedom is $`3\times 2\times 3\times 2+8\times 2=52`$, of which all but the strange quarks are essentially massless. Evidently, the difference is gigantic! Remarkably, the change from one regime to the other appears to occur largely within a narrow range of temperatures around 150 Mev, amazingly low if regarded from the hadron side.
### 3.3 Ideas About Quark-Gluon Plasma
The physics of quark-gluon plasma is already a big subject with a vast literature. It will bloom further as the RHIC and ALICE programs gather data. Let me briefly sketch a few of the characteristic phenomena that have been discussed.
* A fundamental foundational result is the observation that distributions of particle energies in the final state are well described by thermal distributions corresponding to a freeze-out temperature around 120 Mev. This observation makes it extremely plausible, as had already been anticipated from theoretical work, that approximate thermal equilibrium (at least, kinetic equilibrium) is established – at higher temperatures, of course – in the initial fireball. That’s very good news, both because it makes the theoretical analysis easier, and because it means that the collisions really are approximating the conditions which are of most fundamental interest.
* The most basic and profound prediction is what I have already mentioned, that one should have approximately the energy and pressure characteristic of the appropriate – large! – number of microscopic degrees of freedom. Qualitatively, this means among other things a steep rise in the specific heat, so that the rise of temperature with energy will slow markedly. Energy will go into particle production, not motion. In principle, the temperature is accessible either through measurement of the transverse momenta of hard leptons or photons emerging from the initial fireball. The entropy can be estimated from the final thermal particle distribution at freezeout, since the expansion and cooling should be roughly adiabatic until freezeout. Several more sophisticated flow diagnostics have been proposed to give handles on the full equation of state.
* Since strange quarks are expected to be much lighter than the lightest hadrons ($`K`$ mesons) in which they are found, one can anticipate a significant rise in the relative multiplicities of strange (and antistrange) particles, relative to normal hadronic collisions. There is already a striking phenomenon of this kind seen at SPS, with a dramatic rise (as much as a factor 15) in $`\mathrm{\Omega }`$ and $`\overline{\mathrm{\Omega }}`$ production.
* Perhaps the single most striking experimental result to emerge so far from the study of heavy ion collisions is the suppression of J/$`\psi `$, relative to Drell-Yan background, in lead-lead collisions at the highest energies. When the ratio is plotted as a function of atomic weight and energy, a clear break in its behavior, relative to lower energies or lower atomic numbers, leaps to the eye. No hadron-based model of the collision process anticipated this break, and none has been successful in reproducing it. On the other hand an effect of just this sort was anticipated, based on simple qualitative arguments, to mark the onset of quark-gluon plasma behavior.
The basic point is that free gluons are very effective in dissociating J/$`\psi `$ particles. In quark-gluon plasma there is an abundance of free gluons, while in the hadron phase there is a large mass gap for glue. Alternatively, we may say that in the plasma phase color screening prevents J/$`\psi `$ binding. Unfortunately, while some effect of this kind is very plausible, and evidently does occur, it seems difficult to refine the heuristic argument into a really precise calculation.
* The best hope for a rigorous characterization of quark-gluon plasma behavior is probably comparison of experimental measurements to calculated predictions for quantities that can be addressed using (sophisticated extensions of) perturbative QCD. Among the most promising candidates are hard probes, such as high transverse momentum jets, high-mass dileptons, and energetic photons. Assuming thermal equilibrium – or a definite model of quasi-equilibrium – one can formulate reasonably precise expectations for the rates and distributions of these phenomena, with many cross-checks. Their use is analogous to the use of radiative probes in traditional plasma diagnostics. Another characteristic signature of quark-gluon plasma is softening of quark jet distributions due to their passage through the medium.
A more conjectural possibility, which has received much attention under the name “disordered chiral condensate” or DCC, is that the return to equilibrium as the fireball cools is marked by collective relaxation. Then one might see gross deviations from equipartition in the collective modes.
Specifically, as we shall discuss at length below, one expects that the large spontaneous breaking of chiral symmetry which occurs in the ground state comes undone at high temperatures. The transition from chiral symmetry breaking to chiral symmetry restoration is described by equations very similar to the equations that describe the loss of magnetization when one heats a magnet past its Curie temperature.
To make the analogy accurate we must envision the demagnetization taking place in the presence of a tiny external field, representing the small intrinsic breaking of chiral symmetry due to non-zero $`u`$ and $`d`$ quark masses. Now if, after being disordered at high temperature, the magnet is cooled rapidly there are two extreme possibilities for how it might relax back down to the ground state. According to one extreme picture, each spin separately and independently settles down to align with the external field. According to the other extreme picture, the spins first align with each other in large clumps, usually in some wrong direction, before the clumps relax collectively, as units, toward the correct alignment. In the latter case, one will have significant correlation phenomena, and the possibility of coherent radiation. In QCD, this will take the form of “pion lasing” – an abnormally large number of pions occupying a small region of phase space, with a highly non-Gaussian distribution of charged to neutral multiplicity, will emerge.
Because it is an intrinsically non-equilibrium phenomenon, the likelihood of DCC formation is hard to assess theoretically. It has been observed in some idealized numerical simulations.
Evidently there is a plenitude of signature phenomena in quark-gluon plasma that can and presumably will be explored in heavy ion collisions. We can look forward to a many-faceted dialogue between theory and experiment in coming years.
For the remainder of this lecture and in the next one, however, I will focus on the specific, narrower theoretical question of equilibrium phase transitions. This emphasis brings the advantage that the conceptual issues become well-posed and precise, and support a rich theory; on the other hand application of the results derived to the complex realities of heavy ion collisions is not straightforward.
### 3.4 Screening Versus Confinement
One should not assume, that because the quark-gluon plasma at high temperatures is conveniently described using very different degrees of freedom from those we use to describe the hadronic gas at low temperatures, there must be a sharp phase transition separating them. Indeed, ordinary plasmas are very different from gases of atoms (so different, that at Princeton they are studied on different campuses), but it is well understood that no strict phase transition separates them. The fraction of ionized atoms rises smoothly, though rather abruptly, from nearly (but not quite) zero at low temperatures to nearly unity at high temperatures.
With this cautionary example in mind, let us revisit the question of confinement. Previously we discussed the pure glue theory, and were able to give a precise definition of confinement in terms of the asymptotic behavior of Wilson loops. We were even able to understand in a very simple way why confinement is not at all a bizarre or mysterious behavior, but quite a reasonable possibility for a strong-coupling (or asymptotically free) gauge theory. Actually, it was lack of confinement that required some explaining away–a failure of the strong-coupling expansion, or a phase transition.
Now let’s consider the theory with quarks.
The strong coupling expansion requires that we use the discretized lattice version of the theory. The basic idea of its extension to include quarks is quite simple, although there are great subtleties if one tries to do justice to chiral symmetries, and many algorithmic issues. These questions involve important, active areas of research. However they do not impact the basic issues of screening versus confinement, as discussed in this section.
To give the quarks dynamics, we need to supply a ‘hopping’ term. The sum over all links of
$$\mathrm{\Delta }_{\mathrm{hop}}=\psi (\widehat{r})U_{r,\mu }\overline{\psi }(\widehat{r}+\mu )$$
(41)
does the job, and reduces formally to the continuum action for $`a`$ small. It has an evident gauge invariance, generalizing Eqn. 16, whereby the $`\psi `$ variables, which live on vertices, are simply multiplied by the corresponding $`\mathrm{\Omega }`$s.
Revisiting the question of tiling the Wilson loop, we see that now it is possible to get a non-zero contribution by propagating a single quark line around the perimeter, as shown in Figure 4c. This is quite unlike the pure glue theory, where we were required to tile a whole area. The perimeter tiling corresponds to a potential which does not continue to grow at large distances, but rather saturates at a finite value. Physically, it corresponds to the production of a separated meson pair. The color sources, inserted by the two sides of the Wilson loop, can be saturated by a dynamical quark on one side, and a dynamical antiquark on the other. There is a finite energy to make the pair, but once it is made and combined with the sources into ‘mesons’, the mesons have only short-range residual interactions, and the total energy does not grow with the distance.
There is a simple heuristic way to understand the difference between the two cases. There is an additive quantum number modulo 3, triality, characterizing color charges. It is one for quarks, minus one for antiquarks, and zero for gluons. If we write $`SU(3)`$ indices on the fields, triality is simply the number of upper indices minus the number of lower ones. Because of the existence of the invariant epsilon symbol $`ϵ^{abc}`$ triality can jump in units of three by color invariant processes, but not in units of one or two. In the pure glue theory all the dynamical fields have zero triality, so a source of unit triality cannot be screened. Furthermore the presence, or not, of unit triality can be determined by measurements made at great distances. We saw this in the strong coupling expansion. A triality source generated a ‘live’ link that could be displaced by laying down plaquettes, but not cancelled. We have, therefore, a poor man’s version of Gauss’ law. If triality flux interferes with the correlations in the ground state, then as we separate source and antisource we will produce a finite change in vacuum energy per unit volume that extends over a growing volume, with confinement a conceivable outcome. By contrast, in the theory with dynamical quarks triality can be screened. In the absence of any strictly conserved quantity characterizing a source, it is difficult to imagine how its dynamical influence could extend to great distances. In fact, it would be hard to specify exactly what it is that is confined.
By the way, if the only dynamical quarks are extremely heavy ones then the area tiling can remain cheaper than the perimeter tiling up until very large values of the separation $`R`$. In this case one will have a linear interquark potential out to large $`R`$, supporting a spectrum of bound states up to an ionization threshold.
### 3.5 Models of Chiral Symmetry Breaking
To help ground our later discussions, I will now briefly discuss some basic elements of the phenomenology of chiral symmetry breaking in the observed strong interaction, and in QCD.
The circle of ideas around chiral symmetry breaking grew up around attempts to understand a remarkable formula discovered by Goldberger and Treiman. Their derivation of the formula made use of drastic and uncontrolled approximations, and is mainly of historical interest. The modern understanding starts from ideas introduced by Nambu and Gell Mann and Levy, and developed with great ingenuity by many physicists. Their hypotheses are fully justified within QCD. Indeed, nowadays it is appropriate to start from QCD, and to interpret the necessary hypotheses within the microscopic theory.
Interpreted within QCD, the hypothesis of chiral symmetry breaking has two parts:
i. The $`u`$ and $`d`$ quark masses are small, so that the corresponding fundamental interaction terms $`m_u\overline{u}u`$ and $`m_d\overline{d}d`$ in the Lagrangian may be treated as perturbations.
Thus we are invited to consider the properties of a zeroth-order theory with massless $`u`$ and $`d`$ quarks. In this limit, as we have discussed, there is an $`SU(2)_L\times SU(2)_R`$ chiral symmetry of the fundamental theory, rotating among the different helicities separately.
ii. In the absence of $`u`$ and $`d`$ quark masses, the $`SU(2)_L\times SU(2)_R`$ chiral symmetry is spontaneously broken, down to the diagonal vector subgroup $`SU(2)_{L+R}`$.
More precisely, the hypothesis is that a condensate
$$\overline{u}u=\overline{d}d=v0$$
(42)
develops.
One can also consider extending these hypotheses to the $`s`$ quark, but it is not entirely clear under what circumstances it is safe to treat $`m_s\overline{s}s`$ as a perturbation.
A consequence of these hypotheses is that one expects the existence of approximate Nambu-Goldstone bosons. If it were an exact symmetry that were spontaneously broken we would have exactly massless particles of this type; since there is some small intrinsic breaking, in addition to the larger spontaneous breaking, the corresponding Nambu-Goldstone particles acquire non-zero, but small, masses.
There are indeed particles within the observed hadron spectrum that are much lighter than any of their brethren, namely the $`\pi `$ mesons. Furthermore the quantum numbers of the $`\pi `$ mesons – $`J^{PC}=0^+`$, $`SU(2)_{L+R}`$ (isospin) triplet – are what one requires for Nambu-Goldstone bosons arising from $`SU(2)_L\times SU(2)_RSU(2)_{L+R}`$ breaking.
To see this, consider the physical origin of the Nambu-Goldstone bosons. They arise due to the possibility of obtaining low-energy field configurations by interpolating slowly, in space and time, among the energetically degenerate but inequivalent ground states one has due to spontaneous symmetry breaking. The inequivalent ground states are generated by three independent transformations of the type $`(q,q)`$ in the Lie algebra of $`SU(2)_L\times SU(2)_R`$, which manifestly form an isotriplet of odd parity. Furthermore there is no preferred space-time direction in the condensate, so the quanta are spin 0.
Although it is fundamentally a phenomenon of the strong interaction, much of the interest of chiral symmetry derives from its connection with the weak interaction. Specifically, the currents that generate the approximate chiral symmetry of the strong interaction also appear in the weak interaction. The prototype application, the Goldberger-Treiman relation, exploits this connection. The pion decay $`\pi ^+\mu ^+\nu `$ involves the hadronic matrix element of the axial vector current
$$0|A_\eta ^5|\pi ^+=F_\pi p_\eta $$
(43)
where $`p`$ is the momentum.
Thus $`F_\pi `$ is a directly measurable quantity. For the divergence of the axial current we find then
$$0|^\eta A_\eta ^5|\pi ^+=F_\pi m_\pi ^2.$$
(44)
We see here the connection between chiral symmetry and the mass of the pion: in the version of QCD with exact chiral symmetry the divergence would vanish, and so would the mass of the pion.
Now let us consider another matrix element that appears in describing another basic weak process, that is beta decay of the neutron. The nucleon matrix element of the axial current
$$N|A_\eta ^5|N=G_A\overline{u}\gamma _\eta \gamma _5u$$
(45)
at small momentum transfer for nucleons nearly at rest. $`G_A`$ is a quantity subject to strong-interaction corrections and it is therefore not the sort of thing we can normally expect, in the absence of special insight, to calculate easily. It is measured to be about 1.2. Taking again the divergence, we have on the right-hand side $`2MG_A\overline{u}\gamma _5u`$, not particularly small (beyond the kinematic suppression), whereas on the right-hand side we have the matrix element of a “small”, chiral-symmetry breaking operator. However there is a contribution to this matrix element arising from the nucleon coupling to a $`\pi `$ meson, which then communicates with the current divergence according to Eqn. 44. The Feynman graph for this is shown in Figure 9. The factors of $`m_\pi `$ cancel, and we find
$$g_YF_\pi =2m_NG_A,$$
(46)
which is the Goldberger-Treiman relation.
This logic of this derivation of the Goldberger-Treiman relation can be vastly generalized, to include matrix elements of low-momentum pions or currents between various states. It can be made systematic by using the technology of Ward identities. In this context one finds that in relating multi-current Green functions to multi-pion processes one must often evaluate current commutators. One obtains in this way a host of predictions for low-energy processes, which work remarkably well. (At high energies or momenta the saturation of axial currents with pions is no longer accurate.) The successful evaluation of such commutators, in agreement with experiment, using relations abstracted from free field theory (now justified in QCD), was a major step in the historical elucidation of the strong interaction, and in the revival of interest in quantum field theory in the late 1960s. When I reflect that this elaborate theoretical technology was developed before the microscopic theory, by working backward from nuggets of relevant data embedded within an overwhelming confusion of strong-interaction phenomena, I am lost in admiration.
Many of the results of Ward identity and current algebra gymnastics can be derived in an easier, more transparent way by writing down appropriate effective Lagrangians. If they embody the correct broken and unbroken symmetries, such Lagrangians will satisfy all the Ward identities that can be derived as consequences of these symmetries. Thus one can reproduce more transparently the valid results of the Green function analysis – and more. Generally the ‘more’ – relations derived from the effective Lagrangian away from the low-energy limit – will depend on non-generic features of the effective Lagrangian, and should be ignored.
### 3.6 More Refined Numerical Experiments
With this background, we are now in a better position to discuss two more refined diagnostics of finite temperature QCD.
The first is the so-called Polyakov loop. It is basically half a Wilson loop. Let me be more precise. A standard result in path integral theory states that one can set up the partition function at finite temperature by passing to imaginary values of the time variable and requiring periodicity of the fields (antiperiodicity for fermions) under $`\tau \tau +1/T`$. Now we can consider parallel transport around the circle of imaginary time:
$$L=\mathrm{Tr}\mathrm{exp}\{i_0^\beta A_\tau ^a\lambda ^a𝑑\tau \}.$$
(47)
By the same strong coupling argument as before, we anticipate that the expectation value of the Polyakov loop integral will vanish in a confining phase.
It is very pleasant that we can (following Polyakov) relate this anticipation to a symmetry principle. The action for the pure glue theory is left invariant if we multiply all the $`U`$ matrices connecting (say) $`\tau =0`$ to their temporal neighbor $`\tau =a`$ vertices by an element of the center of the group, since any plaquette product has either no such $`U`$ or one going up and one down. Thus for $`SU(2)`$ we can multiply by -1 (times the identity matrix), and we have the symmetry $`Z_2`$, whereas for $`SU(3)`$ we can multiply by $`\omega `$ or $`\omega ^2`$, where $`\omega =e^{2\pi i/3}`$ is the cube root of unity, and we have the group $`Z_3`$. On the other hand the Polyakov loop is not left invariant under these operations, but rather multiplied by the corresponding numerical factor. If the expectation value of the loop vanishes, as is characteristic of the confined phase, then the confinement symmetry is valid. However if the expectation value of the Polyakov loop does not vanish the confinement symmetry is spontaneously broken.
The discrete “confinement symmetry” of the pure glue theory is not valid for the action including quarks. Indeed, we can think of it as multiplying different triality sectors containing $`N`$ quarks, (modulo $`N`$ for $`SU(N)`$) by different phases. ‘Virtual’ quark world-lines winding around the imaginary time circle are not left invariant. Indeed, with the implementation above, terms in the action that hop quarks from $`\tau =0`$ to $`\tau =a`$ are not invariant. By redefining $`\psi (\tau =a)`$ by a phase you can move the changes in the action to the next time slice, … , but after winding around the circle you will just arrive back where you started.
This formal argument based on confinement symmetry agrees, of course, with our previous intuitive argument from consideration of conserved triality charge. In the pure glue theory there is a conserved flux that cannot be screened, an associated symmetry, and a strict criterion of confinement; in the theory with quarks all that structure is gone, and there is no strict definition of confinement to distinguish it from screening.
While there is no reason to expect the Polyakov loop strictly to vanish at the onset of deconfinement (since deconfinement is, as I’ve already belabored, an incoherent notion in this context), it remains a perfectly respectable observable. One might expect that if there is a crossover from a hadronic phase exhibiting pretty good confinement to a quark-gluon phase with very poor confinement the Polyakov loop should take a nose dive. This is indeed the behavior that shows up in the numerical simulation, as you see in Figure 10.
The second is the chiral condensate. While the notion of confinement gets fuzzy in the theory with quarks, the notion of chiral symmetry breaking is perfectly sharp (for massless quarks). And it exhibits very interesting dynamical behavior as a function of temperature. The simplest measure of chiral symmetry breaking is simply the expectation value $`\overline{\psi }\psi `$. This quantity is perfectly accessible to lattice gauge theory. In Figure 10, you see that this expectation value does indeed take a dive. In the simulation it does not reach zero, because the quarks are not truly massless, but the possibility of a smooth decrease to zero at a finite value of $`T`$ is certainly suggested. Studies with varying values of the quark masses, when extrapolated, further support this suggestion.
## 4 Lecture 3: High-Temperature QCD: Phase Transitions
### 4.1 Yoga of Phase Transitions and Order Parameters
Confinement, in the pure glue version of QCD (only), is a property we can associate with a definite symmetry, that is valid at low temperature but broken at high temperature. Chiral symmetry, in the versions of QCD with two or more massless quarks is, conversely, spontaneously broken at low temperatures (at least if the number of quarks is not too large) but restored at high temperatures. As emphasized by Landau, the presence or absence of a symmetry is a sharp, objective question, which in any given state of matter must have a yes or no answer. And if the answer is yes in one regime and no in another, passage from one regime to the other must be accompanied by a sharp phase transition. This situation is usually parameterized by some appropriate order parameter, that transforms non-trivially under the symmetry, and is zero on the unbroken side but non-zero on the broken side.
Phase transitions can occur without change of symmetry, or dynamical reasons – we shall see some simple examples below (where there is symmetry lurking just offstage). But by considering changes in symmetry, which must be associated with phase transitions, and behaviors of order parameters we will be able to say quite a lot, without doing any prohibitively difficult calculations.
#### 4.1.1 Second Order Transitions
There are two broad classes of phase transitions, which have quite different qualitative properties near the transition point. First order transition are characterized by a finite discontinuity in the generic thermodynamic parameter – i.e. basically in anything except the free energy, which of course must be equal for the two phases at the transition point. Second-order transitions, on the other hand, are characterized by continuous but nonanalytic behavior of thermodynamic quantities.
In Nature second order transitions are less common than first order transitions, but they are especially interesting. Near first order transitions the two phases are simply ‘different’, and are described by distinct expressions for the free energy (in terms of macroscopic variables). There is a wholesale reorganization of matter, even locally – there are jumps in intensive variables. Near second order transitions that is not the case. In a large but finite volume, a first order transition point will be marked by rare but sudden and drastic jumps from one phase to the other, going over in the infinite volume limit to hysteresis. In the same circumstances, a second order transition point will not exhibit any jumps, and the partition function will be a perfectly analytic function.
So how does the nonanalytic behavior arise? It can only arise from taking the infinite volume limit. This, in turn, implies that for a second order transition to occur there must be low-energy fluctuations of arbitrarily long wave length, since it is only such modes that can render the infinite volume limit subtle (otherwise the free energy must ultimately become simply additive in the volume, for large enough volumes). In terms of static quantities, there must be a diverging correlation length. In terms of particle physics, there must be massless particles. That is, if we quantized the modes under discussion, they would have massless quanta.
The hypothesis – or quasi-theorem, as motivated above – that nonanalytic behavior of thermodynamic quantities near a second order transition must arise from the dynamics of massless modes makes it possible, following Landau and Wilson, to make remarkably concrete and specific predictions about this behavior. The point is that it appears to be very difficult to construct consistent theories of massless particles, unless one considers small numbers of dimensions or large and exotic symmetry structures, that are inappropriate to the cases at hand. So if we specify the desired space dimension and symmetry we may find a unique theory of this kind, or none at all. Then the singular behavior of any possible second-order phase transitions with a given dimensionality and symmetry will be uniquely determined, independent of other details of the underlying microscopic theory. This is the hypothesis of universality.
Universality makes it possible to make rigorous predictions for the behavior of complicated physical systems – such as various versions of QCD – near second-order phase transitions by doing calculations in much simpler models.
Now let me give a few words of orientation about the formal aspects of such analyses, using the three dimensional Ising model as a prototype. We are interested in the singular behavior of thermodynamic functions near a possible second-order phase transition, where the magnetization decreases from a non-zero value to zero. The relevant low-energy, long-wavelength modes are gradual changes in the local average of the magnetization. Since we are working at long wavelengths it is appropriate to coarse grain, so the magnetization is described by a real three-dimensional scalar field $`\varphi (x)`$. We are interested in the singularity of the partition function induced by fluctuations of $`\varphi (x)`$. To find it, we need to construct the appropriate “universal” theory based on $`\varphi (x)`$.
We want this theory to be describing fluctuations that are small in magnitude and long in wavelength, so we should use a Lagrangian with the smallest possible powers of $`\varphi (x)`$ and of derivatives. Of course we need a quadratic term with two derivatives to get any non-trivial spatial behavior at all. To give this term its chance to shine, we will also need to put the mass equal to zero, since a mass term would always dominate the derivative term at long wavelength. Since we have $`\varphi \varphi `$ symmetry, the next possibility is a $`\varphi ^4`$ term. So our trial ‘Lagrangian’ (to be interpreted as $`H/T`$, the Hamiltonian divided by the temperature, in statistical mechanical language) is
$$d^3x=d^3x(\varphi )^2+\lambda \varphi ^4.$$
(48)
Now if we count dimensions we see that for the Lagrangian to be dimensionless $`\varphi `$ must have mass dimension $`1/2`$, and so $`\lambda `$ must have mass dimension 1. According to naive dimensional analysis, therefore, we are not getting a scale-invariant theory.
We know, however, that interacting field theories contain another source of non-trivial scaling behavior. Because there are an infinite number of degrees of freedom, we must regulate the theory, and define a renormalized coupling at some finite momentum scale, which we fix to a physical value independent of the cutoff. Then in favorable cases we will get cutoff-independent answers, in terms of the renormalized coupling, as we take the cutoff to infinity. We can get a scale invariant theory from this set-up if the bare coupling $`\lambda _b(\mathrm{\Lambda })`$ one needs to insure a fixed renormalized coupling scales as $`\lambda _b\mathrm{\Lambda }^{\frac{1}{2}}`$. The proportionality constant is then our dimensionless parameter.
That reasoning is quite abstract, so it is informative to consider the situation also from a radically different perspective, due to Wilson and Fisher. We consider, formally, extending the theory to $`4ϵ`$ dimensions. It is of course problematic to construct a field theory in non-integer dimensions, but we can perfectly well continue the perturbation theory integrals, which ought to be adequate if the relevant coupling turns out to be small. Now in $`4ϵ`$ dimensions the same counting as above shows that the mass dimension of $`\lambda `$ is $`ϵ`$. Thus as we rescale the momentum at which the coupling is defined there are two sources of variation. One is simple classical dimensional analysis, which tends to make the $`\varphi ^4`$ more relevant at small momenta; the other is the running due to fluctuations (loops), familiar in its quantum version from QED or QCD, where it is interpreted as vacuum polarization due to virtual particles. Of course here we are concerned with classical fluctuations, but the equations are just the same. Since $`\varphi ^4`$ in 4 dimensions is not asymptotically free, the two terms governing renormalization toward the infrared go as
$$\frac{d\stackrel{~}{\lambda }(p)}{dt}=\stackrel{~}{\lambda }ϵb\stackrel{~}{\lambda }^2,$$
(49)
here $`t=\mathrm{ln}(p_0/p)`$, with $`p_0`$ some reference momentum, and $`\stackrel{~}{\lambda }\lambda p^ϵ`$, and the equation is valid for $`p<<\mathrm{\Lambda }`$ the cutoff and small $`\lambda `$ and $`ϵ`$. $`b`$ is a calculable positive number. So as $`t\mathrm{}`$ $`\stackrel{~}{\lambda }ϵ/b`$, which is indeed small for small enough $`ϵ`$. We say there is a fixed-point coupling with this value. With $`\mathrm{\Lambda }`$ fixed, we approach a scale invariant theory for $`p<<\mathrm{\Lambda }`$. The funny dependence of the coupling $`\lambda `$ (not $`\stackrel{~}{\lambda }`$!) extrapolated to momenta approaching the cutoff – what we would call the bare coupling – on the cutoff $`\mathrm{\Lambda }`$ itself is just what we anticipated before, on abstract grounds.
This construction makes it plausible that there can be a scale-invariant theory, but also makes it clear that this theory will not be easy to find in three dimensions, where the fixed-point coupling cannot be small. One approach, which works remarkably well, is to calculate around $`4ϵ`$ dimensions and extrapolate to $`ϵ=1`$. Another is to work directly in three dimensions, calculate to high orders in perturbation theory, and join on to the form at high orders, which is known from sophisticated semiclassical techniques. Finally, one can simply simulate the theory directly numerically. Any of these techniques would be prohibitively difficult to use in high temperature QCD directly, but thanks to universality we can get rigorous quantitative answers (to carefully selected questions, of course!) using much simpler models.
#### 4.1.2 First Order Transitions
An important side-benefit of the analysis of how second-order transitions arise is that it alerts us to cases where this cannot occur. In the Ising model analysis, we found the scale invariant theory could arise when the mass parameter associated with the magnetization vanishes. Since we expect the effective mass parameter to be a function of temperature, $`m^2=m^2(T)`$, it is reasonable to expect that this can happen at one particular value of the temperature. So Eqn. 48 is a special case of the embedding set of Lagrangians
$$d^3x=d^3x(\varphi )^2+m^2(T)\varphi ^2+\lambda \varphi ^4$$
(50)
describing the dynamics of the fluctuating magnetization not only exactly at, but also near, the critical transition temperature. This is reasonable from another point of view as well: when $`m^2(T)<0`$ a non-zero expectation value for $`\varphi `$ will be preferred, whereas for $`m^2(T)>0`$ the expectation should vanish.
Now if we put the system in an external magnetic field, breaking the $`\varphi \varphi `$ symmetry, then $`\varphi `$ and $`\varphi ^3`$ terms are allowed. Let’s shift away the $`\varphi `$ term. There will still be a $`T`$-dependent $`m^2`$, and it can go through zero. But now that will generally not give rise to a second-order transition, because in the presence of a small $`\varphi ^3`$ term the expectation value will jump to a new minimum at a non-zero (positive) value of $`m^2`$. The special case where the cubic term vanishes simultaneously with $`m^2`$ can be accessed only if there is another control parameter available, in addition to the temperature. Then one has a so-called tricritical point. In the phase plane, the tricritical point appears as the terminus of a line along which there are weaker and weaker first-order transitions.
Even if the mean-field analysis allows a second-order transition, there will not be one if there is no suitable scale-invariant theory to represent the universality class. The mean field analysis ignored fluctuations, but as we have learned these are vital. In our discussion above, we saw that in order to construct the scale-invariant theory we needed to have a simple, finite limiting behavior of the effective coupling under renormalization group transformations toward the infrared. If there is no such limiting behavior, there cannot be a second-order transition. The physical interpretation of this outcome is simply that in such cases the fluctuations have grown out of control, resulting in a catastrophic rearrangement of the state – a first-order transition. Such an eventuality is, for obvious reasons, called a fluctuation driven first-order transition.
Since they are marked by finite discontinuities, first-order transitions are robust against small perturbations. Thus if we have a symmetry and an order parameter, whose change from a non-zero to a zero value forces the existence of a first-order transition according to either of the mechanisms I’ve just discussed, there will still be a first-order transition even if the symmetry is intrinsically slightly broken. There will be no strict order parameter, and thus we would not have been able to predict the necessity of a transition without referring to the nearby, unbroken variant of the theory.
Each and every one of the theoretical phenomena I have mentioned in these orienting sections plays a significant role in understanding the phase structure of QCD!
### 4.2 Application to Glue Theories
Let’s recall the basic facts. When there are no quarks at all, then there is the possibility of a true confinement-deconfinement transition. As we have discussed, such a transition is characterized by the Polyakov order parameter
$$L=\mathrm{Tr}\mathrm{exp}\{i_0^\beta A_\tau ^a\lambda ^a𝑑\tau \}.$$
(51)
Here the expectation value is taken over the thermal ensemble of field configurations periodic in imaginary time $`\tau `$ with period $`\beta =1/T`$, and $`\lambda `$ is the representation matrix for the fundamental representation. This loop inserts quark quantum numbers into the ensemble. In the pure glue theory, the operator inserts a flux that cannot be screened, and alters the state by an irreducible amount out to infinity. This costs a finite energy per unit volume, and therefore infinite energy altogether. The expectation value of the loop would therefore be expected to vanish in the confined phase, while it acquires a non-zero value in the unconfined phase. $`L`$ is multiplied by the appropriate complex root of unity when an element of the center of the gauge group is applied to the state. Thus symmetry under this discrete group (triality for color $`SU(3)`$, diality for $`SU(2)`$) is broken in the unconfined phase expected to exist at high temperature.
Since there is a simple order parameter with well-defined symmetry properties, one can entertain the possibility of a second order transition. Indeed there does seem to be a second order transition for $`SU(2)`$, in the universality class of the (inverted) 3d Ising model. ‘Inverted’ refers to the features that whereas in the Ising model the $`Z_2`$ symmetry is broken at low temperature, but restored at high temperature, the confinement $`Z_2`$ symmetry is valid at low temperature but broken at high temperature. This means the $`m^2(T)`$ goes through zero in the opposite direction, but of course one still has the same universal theory at the critical point and a simple correspondence away from it.
However for $`SU(3)`$ the appropriate model is different. It is something called the 3-state Potts model. In the field-theoretic version of that model we must use a complex scalar field $`\varphi `$ invariant under $`\varphi \omega \varphi `$, with $`\omega `$ the cube root of unity, to implement the symmetry. With such a field, cubic terms of the type
$$\mathrm{\Delta }=\kappa (\varphi ^3+\varphi ^3)$$
(52)
are allowed. The existence of a cubic invariant implies that the transition will be first order.
Both these predictions prove to be true in large-scale numerical simulations of pure glue QCD.
### 4.3 Application to Chiral Transitions
Again, let’s quickly recall the basics. With dynamical quarks there is no longer a confinement symmetry, but if we have $`f`$ flavors of massless quarks there is an additional symmetry under chiral transformations in the group $`SU(f)_L\times SU(f)_R\times U(1)_B`$ of independent special unitary rotations of the left- and right-handed fields, together with the overall vector baryon number symmetry. (The additional apparent axial baryon number symmetry, present at the classical level, is violated in quantum theory by the anomaly, as discussed earlier.) This chiral symmetry is believed on good grounds to break spontaneously down to vector $`SU(f)\times U(1)`$ at low temperatures; and to be restored at sufficiently high temperatures. Of course for $`f=1`$ the chiral symmetry is vacuous; but for $`f2`$ there is a phase transition associated with restoration of chiral symmetry. Since there is a simple order parameter for this phase transition – namely, for example, the expectation value of the quark bilinear
$$M_j^i=\overline{q}_{L}^{}{}_{}{}^{i}q_{R}^{}{}_{j}{}^{}$$
(53)
– one may again inquire concerning the possibility of a second order transition.
#### 4.3.1 Formulation of Models
In order to describe a possible second-order transition quantitatively, we must try to find a tractable model in the same universality class. For the chiral order parameter Eqn. 53 the relevant symmetries are independent unitary transformations of the left- and right-handed quark fields, under which
$$MU^{}MV.$$
(54)
These transformations generate an $`SU(f)_L\times SU(f)_R\times U(1)_V`$ symmetry, after the anomaly in the axial baryon number current is taken into account. At the phase transition, the true symmetry is broken to $`SU(f)_{L+R}\times U(1)_V`$.
To describe the critical behavior, it is sufficient to retain the degrees of freedom which develop long-range fluctuations at the critical point. It is natural to assume that these are associated with long-wavelength variations in the order parameter, whose magnitude is small and whose variations within the vacuum manifold therefore cost little energy near the transition. Thus the most plausible starting point for analyzing the critical behavior of a possible second-order phase transition in QCD is the Landau-Ginzburg free energy
$$=\mathrm{tr}_iM^{}_iM+\mu ^2\mathrm{tr}M^{}M+\lambda _1\mathrm{tr}(M^{}M)^2+\lambda _2(\mathrm{tr}M^{}M)^2.$$
(55)
Here $`\mu ^2`$ is the temperature-dependent renormalized (mass)<sup>2</sup>, which is negative below and positive above the critical point, while $`\lambda _1`$ and $`\lambda _2`$ parameterize the strength of the quartic couplings and are supposed to be smooth at the transition. The symmetry breaking pattern we want is $`M\mathrm{𝟏}`$ below the transition, which is what we shall find at the minimum of the potential, within a range of positive $`\lambda _1`$ and $`\lambda _2`$.
Actually Eqn. 55 is not quite what we want. It has a full $`U(f)\times U(f)`$ symmetry, which breaks down to $`U(f)`$. Thus it contains a massless Nambu-Goldstone boson for the axial baryon number symmetry, which is not present in the microscopic theory (QCD) we are trying to model.
For $`f=2`$ one can solve this problem very neatly by implementing the symmetry in a slightly different way. Instead of general complex matrices, which in this case form a reducible representation of the chiral symmetry, we can restrict ourselves to unitary matrices with positive real determinant. This restriction on $`M`$ is consistent with the transformation law Eqn. 54 as long as $`U`$ and $`V`$ have equal phases, but not if the phases are unequal. Thus the unwanted axial $`U(1)`$ symmetry is indeed removed. An important point is that for $`2\times 2`$ matrices the condition of being a multiple of a unitary matrix is a linear condition, so that we can enforce it while remaining within the domain of renormalizable field theories. (Nothing like this is true for larger matrices.) It is very convenient to parameterize the $`2\times 2`$ matrices in question in terms of four real parameters $`(\sigma ,\stackrel{}{\pi })`$ and the Pauli matrices as
$$M=\sigma +i\stackrel{}{\pi }\stackrel{}{\tau }.$$
(56)
In this way, we arrive back at the original model of Gell-Mann and Levy.
After this pruning, the order parameter variables that have been retained have the quantum numbers of the scalar isoscalar density $`\overline{q}^iq_i`$ and the pseudoscalar isovector densities $`\overline{q}^x\gamma _5\stackrel{}{\tau }q`$. Furthermore, the two a priori possible quartic couplings $`\lambda _1,\lambda _2`$ are not independent – a fact which proves to be very significant. In fact the model boils down to the theory of a four-component vector $`\varphi (\sigma ,\stackrel{}{\pi })`$ in internal space, that is to say the standard $`O(4)`$ invariant $`n=4`$ “Heisenberg magnet”. For smaller numbers of components, this sort of model is a much-studied model for the critical behavior of magnets, with the vector of course representing the magnetization (or staggered magnetization).
For larger values of $`f`$ the trick discussed above is no longer available. For $`f=3,4`$ one can break the unwanted $`U(1)`$ symmetry by adding an additional determinantal interaction $`\mathrm{det}M`$. Beyond $`f=4`$ that too is not entirely satisfactory, because it takes us outside the framework of renormalizable theories in four dimensions (which are suitable starting points for the construction of critical theories, using the $`ϵ`$ expansion). However, as will soon appear, in the present context it is almost certainly academic anyway.
#### 4.3.2 Fixed points
Now we may search for second-order transitions within each model, by the standard method of looking for infrared stable fixed points of the renormalization group. The vector model has been studied in great depth, for arbitrary $`n`$. The existence of a fixed point has been established by detailed analysis of perturbation theory – taken to high orders and supplemented with estimates of the asymptotic behavior directly in three dimensions – directly in three dimensions, and also via the $`ϵ`$ expansion. The calculated critical exponents have been successfully compared with appropriate experiments, for $`n3`$. Thus there can be no serious doubt that there is a model for a second-order QCD chiral phase transition for two massless quarks, consistent with the symmetry of that theory.
On the other hand for $`f3`$ the structure of the renormalization group for the appropriate model is more complicated – there are then two independent couplings, and so to speak more ways to go wrong. In the lowest order of perturbation theory in the $`ϵ`$ expansion, the calculations can be done quite simply. They indicate that there is a fixed point of the renormalization group, but that it is not infrared stable. In plain English, this means (if we can trust the $`ϵ`$ expansion!) that near the transition one would naively identify by taking $`\mu ^2`$ through zero in Eqn. 55 $`M`$ is actually subject to catastrophic fluctuations, that change the structure of the problem qualitatively. Thus the renormalization group fails to identify a consistent model for a second order transition, and indicates instead that the transition must be a first order transition. In important work Gausterer and Sanielevici have simulated the $`f=3`$ matrix models directly, both with and without the determinantal interaction, to search for second-order transitions. None was found; the transition is always first order. These authors also studied the $`f=2`$ model, and did find the expected second-order transition in that case. Thus the expectations drawn from simple $`ϵ`$ expansion analysis appear to be vindicated. Direct simulation of the $`n=4`$ magnet model, and comparison with $`f=2`$ QCD using the translation dictionary discussed below, is also an attractive possibility.
Existing numerical calculations for QCD, are consistent with the prediction that the proposed models for the universality classes of possible QCD chiral phase transitions are appropriate, and that there is a big difference between the nature of the chiral transition in QCD for $`f=2`$ and for larger $`f`$, with the former being second order and the latter first order.
### 4.4 Close Up on Two Flavors
If we accept that there is a second order transition for $`f=2`$, we can derive many precise, testable consequences.
As we have just discussed the most plausible starting point for analyzing the critical behavior of a second-order phase transition in QCD is the Landau-Ginzburg free energy
$$F=d^3x\left\{\frac{1}{2}^i\varphi ^\alpha _i\varphi _\alpha +\frac{\mu ^2}{2}\varphi ^\alpha \varphi _\alpha +\frac{\lambda }{4}(\varphi ^\alpha \varphi _\alpha )^2\right\},$$
(57)
of an $`n=4`$ component scalar field. Here $`\mu ^2`$ is the temperature-dependent renormalized (mass)<sup>2</sup>, which is negative below and positive above the critical point, while $`\lambda `$ is the strength of the quartic couplings and is supposed to be smooth at the transition. We neglect terms with higher powers of $`\varphi `$ since $`|\varphi |`$ is small near the transition. The symmetry breaking pattern we want is $`\mathrm{𝟏}`$ (equivalently, $`\sigma 0;\stackrel{}{\pi }=0`$) below the transition which is indeed what we find at the minimum of the potential for positive $`\lambda `$. This model has been studied in depth for arbitrary $`\eta `$ and spatial dimension $`d`$, and the existence of an infrared stable fixed point of the renormalization group for $`\eta =4`$, $`d=3`$ is firmly established. Hence, it is a model for a second order QCD chiral phase transition for two massless quarks.
When the free energy Eqn. 57 is written in terms of $`\sigma `$ and $`\stackrel{}{\pi }`$ it looks much like the original model of Gell-Mann and Levy. But there are two changes: there are no nucleon fields and only three (spatial) dimensions. These two changes reflect an important distinction. We are only proposing Eqn. 57 as appropriate near the second order phase transition point. This is because it is only there that we can appeal to universality – the long-wavelength behavior of the $`\sigma `$ and $`\stackrel{}{\pi }`$ fields is determined by the infrared fixed point of the renormalization group, and microscopic considerations are irrelevant to it. In Euclidian field theory at finite temperature, the integral over $`\omega `$ of zero temperature field theory is replaced by a sum over Matsubara frequencies $`\omega _n`$ given by $`2n\pi T`$ for bosons and $`(2n+1)\pi T`$ for fermions with $`n`$ an integer. Hence, one is left with a Euclidian theory in three spatial dimensions with massless fields from the $`n=0`$ terms in the boson sums and massive fields from the rest of the boson sums and the fermion sums. Hence, to discuss the massless modes of interest at the critical point, Eqn. 57 is sufficient. We do not need to introduce nucleon fields or constituent quark fields.
#### 4.4.1 Critical Exponents
The most important universal properties of the second order transition are the critical exponents, which we now define.
First, let us introduce the reduced temperature $`t=(TT_c)/T_c`$. The exponents $`\alpha `$, $`\beta `$, $`\gamma `$, $`\eta `$, and $`\nu `$ describe the singular behavior of the theory with strictly zero quark masses as $`t0`$. For the specific heat one finds
$$C(T)|t|^\alpha +\mathrm{less}\mathrm{singular}.$$
(58)
The behavior of the order parameter defines $`\beta `$.
$$|\varphi ||t|^\beta \mathrm{for}t<0.$$
(59)
$`\eta `$ and $`\nu `$ describe the behavior of the correlation length $`\xi `$ where
$$G_{\alpha \beta }(x)\varphi (x)_\alpha \varphi (0)_\beta \varphi _\alpha \varphi _\beta \delta _{\alpha \beta }\frac{A}{|x|}\mathrm{exp}(|x|/\xi )\mathrm{at}\mathrm{large}\mathrm{distances}.$$
(60)
$`A`$ is independent of $`|x|`$, but may depend on $`t`$. The correlation length exponent $`\nu `$ is defined by
$$\xi |t|^\nu .$$
(61)
Above $`T_c`$, where the correlation lengths are equal in the sigma and pion channels, the susceptibility exponent $`\gamma `$ is defined by
$$d^3xG_{\alpha \beta }(x)t^\gamma .$$
(62)
The exponent $`\eta `$ is defined through the behavior of the Fourier transform of the correlation function:
$$G_{\alpha \beta }(k0)k^{2+\eta }.$$
(63)
The last exponent, $`\delta `$ is related to the behavior of the system in a small magnetic field $`H`$ which explicitly breaks the $`O(4)`$ symmetry. Let us first show that in a QCD context, $`H`$ is proportional to a common quark mass $`m_u=m_dm_q`$. This common mass term may be represented by a $`2\times 2`$ matrix $`𝒟`$ given by $`m_q`$ times the identity matrix. We are now allowed to construct the free energy from invariants involving both $`𝒟`$ and $``$. The lowest dimension term linear in $`𝒟`$ is just $`\mathrm{tr}^{}𝒟=m_q\sigma `$, which in magnet language is simply the coupling of the magnetization to an external field $`Hm_q`$. In the presence of an external field, the order parameter is not zero at $`T_c`$. In fact,
$$|\varphi |(t=0,H0)H^{1/\delta }.$$
(64)
The six critical exponents defined above are related by four scaling relations. These are
$`\alpha `$ $`=2d\nu `$
$`\beta `$ $`={\displaystyle \frac{\nu }{2}}(d2+\eta )`$
$`\gamma `$ $`=(2\eta )\nu `$
$`\delta `$ $`={\displaystyle \frac{d+2\eta }{d2+\eta }}.`$ (65)
We therefore need values for $`\eta `$ and $`\nu `$ for the four component magnet in $`d=3`$. These were obtained in the remarkable work of Baker, Meiron and Nickel, who carried the perturbation theory to seven-loop order, and used information about the behavior of asymptotically large orders, and conformal mapping and Padé approximant techniques to obtain
$`\eta `$ $`=.03\pm .01`$
$`\nu `$ $`=.73\pm .02.`$ (66)
Using Eqn. 4.4.1, the remaining exponents are $`\alpha =0.19\pm .06`$, $`\beta =0.38\pm .01`$, $`\gamma =1.44\pm .04`$ and $`\delta =4.82\pm .05`$. Since $`\alpha `$ is negative there is a cusp in the specific heat at $`T_c`$, rather than a divergence.
A more powerful result which includes Eqn. 59 and Eqn. 64 as special cases is the critical equation of state:
$$H=M|M|^{\delta 1}\kappa _1g(\kappa _2t|M|^{\frac{1}{\beta }})$$
(67)
in which here $`Hm`$, $`M\overline{q}q=|\varphi |`$, $`t(TT_c)/T_c`$, $`g`$ is a universal function, and $`\kappa _1`$ and $`\kappa _2`$ are non-universal constants. This equation of state could, of course, be compared directly with sufficiently accurate numerical simulations.
#### 4.4.2 Tricritical Point
To this point I have been discussing a world with two massless quarks, and so we have implicitly been taking the strange quark mass to be infinite. On the other hand, I argued earlier that if the strange quark is massless, then the chiral phase transition is first order. Hence, as the strange quark mass is reduced from infinity to zero, at some point the phase transition must change from second order to first order. This point is called a tricritical point. There is numerical evidence that when the strange quark has its physical mass, and the two light quarks are taken strictly massless, the transition is second order, with the consequences discussed above. However this conclusion was controversial for many years, and even now may not be completely secure. In any case the physical strange quark mass is not drastically different from the critical value, so let’s indulge our idle curiosity a little further (this will pay off shortly).
In a lattice simulation, the strange quark mass can be tuned to just the right value to reach the tricritical point. Let’s discuss the critical exponents that would be observed in such a simulation.
Let us consider the effect of adding a massive but not infinitely massive strange quark to the two flavor theory. This will not introduce any new fields which become massless at $`T_c`$, and so the arguments leading to the free energy Eqn. 57 are still valid. The only effect of the strange quark, then, is to renormalize the couplings. Renormalizing $`\mu ^2`$ simply shifts $`T_c`$, as does renormalizing $`\lambda `$ unless $`\lambda `$ becomes negative. In that case, one can no longer truncate the Landau-Ginzburg free energy at fourth order. After adding a sixth order term, and keeping track of small light quark masses, the free energy becomes
$$F=d^3x\left\{\frac{1}{2}(\varphi )^2+\frac{\mu ^2}{2}\varphi ^2+\frac{\lambda }{4}(\varphi ^2)^2+\frac{\kappa }{6}(\varphi ^2)^3H\sigma \right\}.$$
(68)
While for positive $`\lambda `$, $`\varphi ^2`$ increases continuously from zero as $`\mu ^2`$ goes through zero, for negative $`\lambda `$, $`\varphi ^2`$ jumps discontinuously from zero to $`|\lambda |/(2\kappa )`$ when $`\mu ^2`$ goes through $`\lambda ^2/(4\kappa )`$. Hence, the phase transition has become first order. Thus at the value of $`m_s`$ where $`\lambda =0`$, the phase transition changes continuously from second order to first order.
The singularities of thermodynamic functions near tricritical points, like the singularities near ordinary critical points, are universal. Hence, it is natural to propose that QCD with two massless flavors of quarks and with $`T`$ near $`T_c`$ and $`m_s`$ near its tricritical value is in the universality class of the $`\varphi ^6`$ Landau-Ginzburg model Eqn. 68 . This model has been studied extensively. Because the $`\varphi ^6`$ interaction is strictly renormalizable in three dimensions, this model is much simpler to analyze than the $`\varphi ^4`$ model of the ordinary critical point. No $`ϵ`$ expansion is necessary, and the critical exponents all take their mean field values, up to calculable logarithmic corrections. Here I’ll be content to show you the mean field tricritical exponents.
In mean field theory, the correlation function in momentum space is simply $`G_{\alpha \beta }(k)=\delta _{\alpha \beta }(k^2+\mu ^2)^1`$. Since $`\mu ^2t`$, this gives the exponents $`\eta =0`$, $`\gamma =1`$ and $`\nu =1/2`$. To calculate $`\alpha `$ and $`\beta `$, we minimize $`F`$ for $`H=\lambda =\varphi =0`$, and find $`\alpha =1/2`$ and $`\beta =1/4`$. To calculate $`\delta `$, we minimize $`F`$ for $`t=\lambda =\varphi =0`$ and find $`\delta =5`$.
The result for the specific heat exponent $`\alpha `$ is particularly interesting, since it means that the specific heat diverges at the tricritical point, unlike at the ordinary critical point. This means that whereas for $`m_s`$ large enough that the transition is second order the specific heat $`C(T)`$ has a cusp but is finite at $`T=T_c`$, as $`m_s`$ is lowered to the tricritical value $`C(T_c)`$ should increase since at the tricritical point it diverges. This behavior should be seen in future lattice simulations.
Finally, at a tricritical point there is one more relevant operator than at a critical point, since two physical quantities ($`t`$ and $`m_s`$) must be tuned to reach a tricritical point. Hence, a new exponent $`\varphi _t`$, the crossover exponent, is required. For $`\lambda 0`$, tricritical behavior will be seen only for $`|t|>t^{}`$, while for $`|t|<t^{}`$, either ordinary critical behavior or first order behavior (depending on the sign of $`\lambda `$) results. $`t^{}`$ depends on $`\lambda `$ according to
$$t^{}\lambda ^{1/\varphi _t}$$
(69)
The mean field value of $`\varphi _t`$ is obtained by minimizing the free energy $`F`$ for $`H=\varphi =0`$, and is $`\varphi _t=1/2`$. These mean field tricritical exponents, $`\alpha =1/2`$, $`\beta =1/4`$, $`\gamma =1`$, $`\delta =5`$, $`\eta =0`$, $`\nu =1/2`$, and $`\varphi _t=1/2`$ would describe the real world if $`m_s`$ were smaller than it is, and will describe future lattice simulations with $`m_s`$ chosen appropriately.
#### 4.4.3 Robustness
If we now turn on small but non-zero $`u`$ and $`d`$ quark masses, the first-order line will persist, but the second-order transition will disappear, replaced by a crossover. And there will still be a true (tri)critical point, where the first-order line terminates! It will be in the universality class of a liquid-gas or Ising tricritical point, rather than the asymptotically free $`\varphi ^6`$-type theory.
### 4.5 A Genuine Critical Point! (?)
In reality, of course, we cannot vary the strange quark mass. If the strange quark mass is, as seems to be indicated, so large that for massless $`u`$ and $`d`$ quarks we would have a second-order transition, then after taking into account the non-zero mass of these quarks we are left with only a crossover. Moreover since the mass of the physical pions is not so different from the critical temperature, we cannot expect that correlation length ever gets very long. Thus we are left with a crossover, and perhaps not a very dramatic one. This is a definite prediction, though perhaps a disappointing one for experimentalists, of the form “nothing striking occurs” (in the way of sharp phase transitions). But fortunately, as emphasized in recent work of Stephanov, Rajagaopal, and Shuryak, that is not the end of the story.
Their basic insight is that although one cannot vary the strange quark mass experimentally, there is another control parameter that one can vary – the chemical potential. If the singularity in $`T`$ at slightly unphysical values of $`m_s`$ and zero $`\mu `$ continues into a singularity at the physical value of $`m_s`$ for slightly costive $`\mu `$, we will be back in business, with a true physical (tri)critical point, featuring diverging correlation lengths at a particular point in the $`\mu ,T`$ plane.
#### 4.5.1 Subnuclear Boiling
The same structure is also suggested by an entirely different line of thought. Consider now the behavior of two-flavor QCD as one varies the chemical potential $`\mu `$ at zero temperature. Models of the type we will be considering in the next two lectures suggest that in the case of two mass quarks there is a first-order transition from a state of broken chiral symmetry at $`\mu =0`$ to a state with chiral symmetry restored at large $`\mu `$. Since the transition is first order, it remains as a first-order transition in the presence of small symmetry-breaking perturbations (the quark masses), despite the absence of an order parameter. This idea is very much in the spirit of the phenomenologically successful MIT bag model, in which nucleons are pictured as droplets of chiral symmetry restored phase embedded in a broken symmetry vacuum. Indeed, if true, it comes very close to providing a microscopic justification of that model.
Physically, the suggested transition represents a sort of subnuclear boiling, to a phase in which quarks (or at least, as we shall see, a subset of them) are liberated to behave as nearly massless, weakly interacting particles.
If in fact there is such a transition, it becomes interesting to follow its fate as a function of temperature. At small temperatures, we shall still have a first-order transition at some value $`\mu _c(T)`$. We might expect that $`\mu _c(T)`$ decreases with $`T`$, since higher temperature should favor the higher-entropy ‘boiled’ state. This expectation can be convincingly justified on thermodynamic grounds. But we know that $`\mu _c(T)`$ cannot decrease to zero, since for $`\mu =0`$ we have only a crossover. Again, the simplest possibility to reconcile the behaviors along the axes is to suppose that the line of first-order transitions terminates at a definite $`(\mu _t,T_t)`$. The strength of the first-order transition weakens as a function of temperature, until at $`T_t`$ the strict distinction between hadron and quark phases has disappeared. This $`(\mu _t,T_t)`$ is the ripe fruit of our theoretical labors: the prediction of a true (tri)critical point in honest-to-god real world QCD, arrived at from qualitative, but deeply rooted, theoretical considerations!
#### 4.5.2 Physical Signatures
Stephanov, Rajagopal and Shuryak (SRS) have made quite specific and detailed proposals for experimental signatures of the phase transition. A proper discussion of this would take us far afield, but a few comments are in order.
In a heavy ion collision one produces a baryon number poor fireball in the central region, but toward the fragmentation regions the baryon number density increases. As the different parts of the fireball expand and cool, each part traces out a different history in the $`(\mu ,T)`$ plane. There will be critical fluctuations and long correlation lengths in a given region of phase space if – and only if – material in the region has passed close close to $`P=(\mu _t,T_t)`$ during its history. For that class of events and regions of rapidity space, one can expect enhanced multiplicities of, and correlations among, low-momentum pions. These enhancements will have the distinctive feature of being non-monotonic in the experimental control parameters, since one can miss the critical point on either of two sides.
The simplest observables to analyze are the event-by-event fluctuations of the mean transverse momentum of the charged particles in an event, $`p_T`$, and of the total charged multiplicity in an event, $`N`$. SRS calculate the magnitude of the effects of critical fluctuations on these and other observables, making predictions which, they hope, will allow experiments to find. As a necessary prelude, they analyze the contribution of noncritical thermodynamic fluctuations. They find that NA49 data (hep-ex/9904014) is consistent with the hypothesis that most of the event-by-event fluctuation observed in the data is thermodynamic in origin. This bodes well for the detectability of systematic changes in thermodynamic fluctuations near $`P`$.
As one example, consider the ratio of the width of the true event-by-event distributions of $`p_T`$ to the width of the distribution in a sample of mixed events. SRS call this ratio $`\sqrt{F}`$. NA49 has measured $`\sqrt{F}=1.002\pm 0.0002`$, which is consistent with expectations for noncritical thermodynamic fluctuations. A detailed calculation suggests that critical fluctuations can increase $`\sqrt{F}`$ by 10 - 20%, fifty times the statistical error in the present measurement. There are other observables which are even more sensitive to critical effects. For example, a $`\sqrt{F_{\mathrm{soft}}}`$ defined by using only the 10% softest pions in each event, might well be affected at the factor of two level.
NA49 data demonstrates very clearly that SPS collisions at $`\sqrt{s}`$ = 17 GeV do not freeze out near the critical point. $`P`$ has not yet been discovered. The nonmonotonic appearance and then disappearance (as $`\sqrt{s}`$ is varied) of any one of the signatures of the critical fluctuations mentioned above, or others, would be strong evidence for critical fluctuations. If nonmonotonic variation is seen in several of these observables, with the maxima in all signatures occurring at the same value of $`\sqrt{s}`$, it would turn strong evidence into an unambiguous discovery of the critical point. The quality of the present NA49 data, and the confidence with which we can use it to learn that collisions at $`\sqrt{s}`$ = 17 GeV 17 GeV do not freeze out near the critical point make it plausible that critical behavior, if present, could be discerned experimentally. If and when the critical point $`P`$ is discovered, it will appear prominently on the map of the phase diagram featured in any future textbook of QCD.
#### 4.5.3 Wanted: Numerical Data
The web of general arguments I have outlined above seems to me to make it quite plausible that there is a true critical point in QCD. We can invoke the canonical yoga of second-order phase transitions to make precise, quantitive predictions for the singular behavior of thermodynamics near this point. Arguments of this sort, however, cannot address the non-universal but vital question of precisely where $`(\mu _t,T_t)`$ is. It is a great challenge to locate this point theoretically, and of course such knowledge, even if approximate, would greatly simplify the experimentalist search.
I believe it ought to be possible to locate the tricritical point numerically. For while it is notoriously difficult to deal with large chemical potentials at small temperature numerically, there are good reasons to be optimistic about high temperatures and relatively small chemical potentials, which is our concern here. Some combination of extrapolating from data taken at imaginary values of the chemical potential, and/or using the real part of the fermion determinant as the measure, may be adequate to sense the singularity, especially if we dial the strange quark mass to bring it close to $`\mu =0`$ (and then extrapolate back to the physical strange quark mass).
If all these strands can be brought together, it will be a wonderful interweaving of theory, experiment, and numerics.
## 5 Lecture 4: High-Density QCD: Methods
The behavior of QCD at high density is intrinsically interesting, as the answer to the question: What happens to matter, if you keep squeezing it harder and harder? It is also directly relevant to the description of neutron star interiors, neutron star collisions, and events near the core of collapsing stars. Also, one might hope to obtain some insight into physics at “low” density – that is, ordinary nuclear density or just above – by approaching it from the high-density side.
### 5.1 Hopes, Doubts, and Fruition
Why might we anticipate QCD simplifies in the limit of high density? A crude answer is: “Asymptotic freedom meets the Fermi surface.” One might argue, formally, that the only external mass scale characterizing the problem is the large chemical potential $`\mu `$, so that if the effective coupling $`\alpha _s(\mu )`$ is small, as it will be for $`\mu \mathrm{\Lambda }_{QCD}`$, where $`\mathrm{\Lambda }_{QCD}200`$ Mev is the primary QCD scale, then we have a weak coupling problem. More physically, one might argue that at large $`\mu `$ the relevant, low-energy degrees of freedom involve modes near the Fermi surface, which have large energy and momentum. An interaction between particles in these modes will either barely deflect them, or will involve a large momentum transfer. In the first case we don’t care, while the second is governed by a small effective coupling.
These arguments are too quick, however. The formal argument is specious, if the perturbative expansion contains infrared divergences. And there are good reasons – two separate ones, in fact – to anticipate such divergences.
First, Fermi balls are generically unstable against the effect of attractive interactions, however weak, between pairs near the Fermi surface that carry equal and opposite momentum. This is the Cooper instability, which drives ordinary superconductivity in metals and the superfluidity of He3. It is possible to have an instability at arbitrarily weak coupling because occupied pair states can have very low energy, and they can all scatter into one another. Thus one is doing highly degenerate perturbation theory, and in such a situation even a very weak coupling can produce significant “nonperturbative” effects.
Second, nothing in our heuristic argument touches the gluons. To be sure the gluons will be subject to electric screening, but at zero frequency there is no magnetic screening, and infrared divergences do in fact arise, through exchange of soft magnetic gluons.
Fortunately, by persisting along this line of thought we find a path through the apparent difficulties. Several decades ago Bardeen, Cooper, and Schrieffer taught us, in the context of metallic superconductors, how the Cooper instability is resolved. We can easily adapt their methods to QCD. In electronic systems only rather subtle mechanisms can generate an attractive effective interaction near the Fermi surface, since the primary electron-electron interaction is Coulomb repulsion. In QCD, remarkably, it is much more straightforward. Even at the crudest level we find attraction. Indeed, two quarks, each a color triplet, can combine to form a single color antitriplet, thus reducing their total field energy.
The true ground state of the quarks is quite different from the naive Fermi balls. It is characterized by the formation of a coherent condensate, and the development of an energy gap. The condensation, which is energetically favorable, is inconsistent with a magnetic color field, and so such weak magnetic color fields are expelled. This is the color version of the famous Meissner effect in superconductivity, which is essentially identical to what is known as the Higgs phenomenon in particle physics. Magnetic screening of gluons, together with energy gaps for quark excitations, remove the potential sources of infrared divergences mentioned above. Thus we have good reasons to hope that a weak coupling – though, of course, nonperturbative – treatment of the high density state will be fully consistent and accurate.
The central result in recent developments is that this program can be carried to completion rigorously in QCD with sufficiently many (three or more) quark species. Thus the more refined, and fully adequate, answer to our earlier question is: “Asymptotic freedom meets the BCS groundstate.” Together, these concepts can render the behavior of QCD at asymptotically high density calculable.
### 5.2 Another Renormalization Group
To sculpt the problem, begin by assuming weak coupling, and focus on the quarks. Then the starting point is Fermi balls for all the quarks, and the low-energy excitations include states where some modes below the nominal Fermi surface are vacant and some modes above are occupied. The renormalization group, in a generalized sense, is a philosophy for dealing with problems involving nearly degenerate perturbation theory. In this approach, one attempts to map the original problem onto a problem with fewer degrees of freedom, by integrating out the effect of the higher-energy (or, in a relativistic theory, more virtual) modes. Then one finds a new formulation of the problem, in a smaller space, with new couplings. In favorable cases the reformulated problem is simpler than the original, and one can go ahead and solve it.
This account of the renormalization group might seem odd, at first sight, to high-energy physicists accustomed to using asymptotic freedom in QCD. That is because in traditional perturbative QCD one runs the procedure backward. When one integrates out highly virtual modes, one finds the theory becomes more strongly coupled. Simplicity arises when one asks questions that are somehow inclusive, so that to answer them one need not integrate out very much. It is then that the microscopic theory, which is ideally symmetric and constrained, applies directly. So one might say that the usual application of the renormalization group in QCD is fundamentally negative: it informs us how the fundamentally simple theory comes to look complicated at low energy, and helps us to identify situations where we can avoid the complexity.
Here, although we are still dealing with QCD, we are invoking quite a different renormalization group, one which conforms more closely to the Wilsonian paradigm. We consider the effect of integrating out modes whose energy is within the band $`(ϵ,\delta ϵ)`$ of the Fermi surface, on the modes of lower energy. This will renormalize the couplings of the remaining modes, due to graphs like those displayed in Figure 11. In addition the effect of higher-point interactions is suppressed, because the phase space for them shrinks, and it turns out that only four-fermion couplings survive unscathed (they are the marginal, as opposed to irrelevant, interactions). Indeed the most significant interactions are those involving particles or holes with equal and opposite three momenta, since they can scatter through many intermediate states. For couplings $`g_\eta `$ of this kind we find
$$\frac{dg_\eta }{d\mathrm{ln}\delta }=\kappa _\eta g_\eta ^2.$$
(70)
Here $`\eta `$ labels the color, flavor, angular momentum, … channel and in general we have a matrix equation – but let’s keep it simple, so $`\kappa _\eta `$ is a positive number. Then Eqn. 70 is quite simple to integrate, and we have
$$\frac{1}{g_\eta (1)}\frac{1}{g_\eta (\delta )}=\kappa _\eta \mathrm{ln}\delta .$$
(71)
Thus for $`g_\eta (1)`$ negative, corresponding to attraction, $`|g_\eta (\delta )|`$ will grow as $`\delta 0`$, and become singular when
$$\delta =e^{\frac{1}{\kappa _\eta g_\eta (1)}}.$$
(72)
Note that although the singularity occurs for arbitrarily weak attractive coupling, it is nonperturbative.
### 5.3 Pairing Theory
The renormalization group toward the Fermi surface helps us identify potential instabilities, but it does not indicate how they are resolved. The great achievement of BCS was to identify the form of the stable ground state the Cooper instability leads to. Their original calculation was variational, and that is still the most profound and informative approach, but simpler, operationally equivalent algorithms are now more commonly used. I will be very sketchy here, since this is textbook material.
The simplest and most beautiful results, luckily, occur in the version of QCD containing three quarks having equal masses. I say luckily, because this idealization applies to the real world, at densities so high that we can neglect the strange quark mass (yet not so high that we have to worry about charmed quarks). Until further notice, I’ll be focusing on this case.
Most calculations to date have been based on model interaction Hamiltonians, that are motivated, but not strictly derived, from microscopic QCD. They are chosen as a compromise between realism and tractability. For concreteness I shall here follow, and consider
$$\begin{array}{cc}H\hfill & =d^3x\overline{\psi }(x)(/\mu \gamma _0)\psi (x)+H_I,\hfill \\ H_I\hfill & =K\underset{\mu ,A}{}d^3x\overline{\psi }(x)\gamma _\mu T^A\psi (x)\overline{\psi }(x)\gamma ^\mu T^A\psi (x)\hfill \end{array}$$
(73)
Here the $`T^A`$ are the color $`SU(3)`$ generators, so the quantum numbers are those of one-gluon exchange. However instead of an honest gluon propagator we use an instantaneous contact interaction, modified by a form-factor $``$. $``$ is taken to be a product of several momentum dependent factors $`F(p)`$, one for each leg, and to die off at large momentum. One convenient possibility is $`F(p)=(\lambda ^2/(p^2+\lambda ^2))^\nu `$, where $`\lambda `$ and $`\nu `$ can be varied to study sensitivity to the location and shape of the cutoff. The qualitative effect of the form-factor is to damp the spurious ultraviolet singularities introduced by $`H_I`$; microscopic QCD, of course, does have good ultraviolet behavior. One will tend to trust conclusions that do not depend sensitively on $`\lambda `$ or $`\nu `$. In practice, one finds that the crucial results – the form and magnitude of gaps – are rather forgiving.
Given the Hamiltonian, we can study the possibilities for symmetry breaking condensations. The most favorable condensation possibility so far identified is of the form
$$\begin{array}{ccc}q_{La}^{i\alpha }(p)q_{Lb}^{j\beta }(p)=q_{Ra}^{i\alpha }(p)q_{Rb}^{j\beta }(p)=ϵ^{ij}(\kappa _1(p^2)\delta _a^\alpha \delta _b^\beta +\kappa _2(p^2)\delta _b^\alpha \delta _a^\beta ).\hfill & & \end{array}$$
(74)
Here we encounter the phenomenon of color-flavor locking. The ground state contains correlations whereby both color and flavor symmetry are spontaneously broken, but the diagonal subgroup, which applies both transformations simultaneously, remains valid. There are several good reasons to think that condensation of this form characterizes the true ground state, with lowest energy, at asymptotic densities. It corresponds to the most singular channel, in the renormalization group analysis discussed above. It produces a gap in all channels, and is perturbatively stable, so that it is certainly a convincing local minimum. It resembles the known order parameter for the B phase of superfluid He3. And it beats various more-or-less plausible competitors that have been investigated, by a wide margin.
Given the form of the condensate, one can fix the leading functional dependencies of $`\kappa _1(p^2,\mu )`$ and $`\kappa _2(p^2,\mu )`$ at weak coupling by a variational calculation. For present purposes, it is adequate to replace all possible contractions of the quark fields in Eqn. 73 having the quantum numbers of Eqn. 74 with their supposed expectation values, and diagonalize the quadratic part of the resulting Hamiltonian. The ground state is obtained, of course, by filling the lowest energy modes, up to the desired density. One then demands internal consistency, i.e. that the postulated expectation values are equal to the derived ones. Some tricky but basically straightforward algebra leads us to the result
$$\mathrm{\Delta }_{1,8}(p^2)=F(p)^2\mathrm{\Delta }_{1,8}$$
(75)
where $`\mathrm{\Delta }_1`$ and $`\mathrm{\Delta }_8`$ satisfy the coupled gap equations
$$\begin{array}{cc}\mathrm{\Delta }_8+\frac{1}{4}\mathrm{\Delta }_1\hfill & =\frac{16}{3}KG(\mathrm{\Delta }_1)\hfill \\ \multicolumn{2}{c}{}\\ \frac{1}{8}\mathrm{\Delta }_1\hfill & =\frac{16}{3}KG(\mathrm{\Delta }_8)\hfill \end{array}$$
(76)
where we have defined
$$G(\mathrm{\Delta })=\frac{1}{2}\underset{𝐤}{}\left\{\frac{F(k)^4\mathrm{\Delta }}{\sqrt{(k\mu )^2+F(k)^4\mathrm{\Delta }^2}}+\frac{F(k)^4\mathrm{\Delta }}{\sqrt{(k+\mu )^2+F(k)^4\mathrm{\Delta }^2}}\right\}$$
(77)
and
$$\begin{array}{cc}\kappa _1(p^2)\hfill & =\frac{1}{8K}(\mathrm{\Delta }_8(p^2)+\frac{1}{8}\mathrm{\Delta }_1(p^2))\hfill \\ \multicolumn{2}{c}{}\\ \kappa _2(p^2)\hfill & =\frac{3}{64K}\mathrm{\Delta }_1(p^2).\hfill \end{array}$$
(78)
The $`\mathrm{\Delta }`$ are defined so that $`F(p)^2\mathrm{\Delta }_{1,8}(p^2)`$ are the gaps for singlet or octet excitations at 3-momentum $`p`$. Eqn. 76 must be solved numerically.
Finally, to obtain quantitative estimates of the gaps, we must normalize the parameters of our model Hamiltonian. One can do this very crudely by using the model Hamiltonian in the manner originally pioneered by Nambu and Jona-Lasinio, that is as the basis for a variational calculation of chiral symmetry breaking at zero density. The magnitude of this chiral condensate can then be fixed to experimental or numerical results. In this application we have no firm connection between the model and microscopic QCD, because there is no large momentum scale (or weak coupling parameter) in sight. Nevertheless a very large literature following this approach encourages us to hope that its results are not wildly wrong, quantitatively. Upon adopting this normalization procedure, one finds that gaps of order several tens of Mev near the Fermi surface are possible at moderate densities.
While this model treatment captures major features of the physics of color-flavor locking, with a little more work it is possible to do a much more rigorous calculation, and in particular to normalize directly to the known running of the coupling at large momentum. I’ll now sketch .
### 5.4 Taming the Magnetic Singularity
A proper discussion of the fully microscopic calculation is necessarily quite technical, and would be out of place here, but the spirit of the thing – and one of the most striking results – can be conveyed simply.
When retardation or relativistic effects are important a Hamiltonian treatment is no longer appropriate. One must pass to Lagrangian and graphical methods. (Theoretical challenge: is it possible to systematize these in a variational approach?) The gap equation appears as a self-consistency equation for the assumed condensation, shown graphically in Figure 12.
With a contact interaction, and throwing away manifestly spurious ultraviolet divergences, we obtain a gap equation of the type
$$\mathrm{\Delta }g^2𝑑ϵ\frac{\mathrm{\Delta }}{\sqrt{ϵ^2+\mathrm{\Delta }^2}}.$$
(79)
The phase space transverse to the Fermi surface cancels against a propagator, leaving the integral over the longitudinal distance $`ϵ`$ to the Fermi surface. Note that the integral on the right diverges at small $`ϵ`$, so that as long as the proportionality constant is positive one will have non-trivial solutions for $`\mathrm{\Delta }`$, no matter how small is $`g`$. Indeed, one finds that for small $`g`$, $`\mathrm{\Delta }e^{\mathrm{const}/g^2}`$.
If we restore the gluon propagator, we will find a non-trivial angular integral, which diverges for forward scattering. That divergence will be killed, however, if the gluon acquires a mass $`g\mathrm{\Delta }`$ through the Meissner-Higgs mechanism. Thus we arrive at a gap equation of the type
$$\mathrm{\Delta }g^2𝑑ϵ\frac{\mathrm{\Delta }}{\sqrt{ϵ^2+\mathrm{\Delta }^2}}𝑑z\frac{\mu ^2}{\mu ^2+(g\mathrm{\Delta })^2}.$$
(80)
Now one finds $`\mathrm{\Delta }e^{\mathrm{const}/g}`$!
A proper discussion of the microscopic gap equation is considerably more involved than this, but the conclusion that the gap goes exponentially in the inverse coupling (rather than its square) at weak coupling still emerges. It has the amusing consequence, that at asymptotically high densities the gap becomes arbitrarily large! This is because asymptotic freedom insures that it is the microscopic coupling $`1/g(\mu )^2`$ which vanishes logarithmically, so that $`e^{\mathrm{const}/g(\mu )}`$ does not shrink as fast as $`1/\mu `$. Since the “dimensional analysis” scale of the gap is set by $`\mu `$, its linear growth wins out asymptotically.
## 6 Lecture 5: High-Density QCD: Color-Flavor Locking and Quark-Hadron Continuity
Color-flavor locking has many remarkable consequences. There is a gap for all colored excitations, including the gluons. This is, operationally, confinement. The photon picks up a gluonic component of just such a form as to ensure that all elementary excitations, including quarks, are integrally charged. Some of the gluons acquire non-zero, but integer-valued, electric charges. Baryon number is spontaneously broken, which renders the high-density material a superfluid.
If in addition the quarks are massless, then their chiral symmetry is spontaneously broken, by a new mechanism. The left-dynamics and the right-dynamics separately lock to color; but since color allows only vector transformations, left is thereby locked to right.
You may notice several points of resemblance between the low-energy properties calculated for the high-density color-flavor locked phase and the ones you might expect at low density, based on semi-phenomenological considerations such as the MIT bag model, or experimental results in real-world QCD. The quarks play the role of low-lying baryons, the gluons play the role of the low-lying vector mesons, and the Nambu-Goldstone bosons of broken chiral symmetry play the role of the pseudoscalar octet. All the quantum numbers match, and the spectrum has gaps – or not – in all the right places. In addition we have baryon number superfluidity, which extends the expected pairing phenomena in nuclei. Overall, there is an uncanny match between all the universal, and several of the non-universal, features of the calculable high-density and the expected low-density phase. This leads us to suspect that there is no phase transition between them!
### 6.1 Gauge Symmetry (non)-Breaking
An aspect of Eqn. 74 that might appear troubling at first sight, is its lack of gauge invariance. There are powerful general arguments that local gauge invariance cannot be broken. Indeed, local gauge invariance is really a tautology, stating the equality between redundant variables. Yet its ‘breaking’ is central to two of the most successful theories in physics, to wit BCS superconductivity theory and the standard model of electroweak interactions. In BCS theory we postulate a non-zero vacuum expectation value for the (electrically charged) Cooper pair field, and in the standard model we postulate a non-zero vacuum expectation value for the Higgs field, which violates both the weak isospin SU(2) and the weak hypercharge U(1).
In each case, we should interpret the condensate as follows. We are working in a gauge theory at weak coupling. It is then very convenient to fix a gauge, because after we have done so – but not before! – the gauge potentials will make only small fluctuations around zero, which we will be able to take into account perturbatively. Of course at the end of any calculation we must restore the gauge symmetry, by averaging over the gauge fixing parameters (gauge unfixing). Only gauge-invariant results will survive this averaging. In a fixed gauge, however, one might capture important correlations, that characterize the ground state, by specifying the existence of non-zero condensates relative to that gauge choice. These condensates need not, and generally will not, break any symmetries.
For example, in the standard electroweak model one employs a non-zero vacuum expectation value for a Higgs doublet field $`\varphi ^a=v\delta _1^a`$, which is not gauge invariant. One might be tempted to use the magnitude of its absolute square, which is gauge invariant, as an order parameter for the symmetry breaking, but $`\varphi ^{}\varphi `$ never vanishes, whether or not any symmetry is broken (and, of course, $`\varphi ^{}\varphi `$ breaks no symmetry). In fact there is no order parameter for the electroweak phase transition, and it has long been appreciated that one could, by allowing the $`SU(2)`$ gauge couplings to become large, go over into a ‘confined’ regime while encountering no sharp phase transition. The most important gauge-invariant consequences one ordinarily infers from the condensate, of course, are the non-vanishing W and Z boson masses. This absence of massless bosons and long-range forces is the essence of confinement, or of the Meissner-Higgs effect. Evidently, when used with care, the notion of spontaneous gauge symmetry breaking can be an extremely convenient fiction – so it proves for Eqn. 74.
### 6.2 Symmetry Accounting
The equations of our original model, QCD with three massless flavors, has the continuous symmetry group $`SU(3)^c\times SU(3)_L\times SU(3)_R\times U(1)_B`$. The Kronecker deltas that appear in the condensate Eqn. 74 are invariant under neither color nor left-handed flavor nor right-handed flavor rotations separately. Only a global, diagonal $`SU(3)`$ leaves the ground state invariant. Thus we have the symmetry breaking pattern
$$SU(3)^c\times SU(3)_L\times SU(3)_R\times U(1)_BSU(3)_{c+L+R}\times Z_2.$$
(81)
#### 6.2.1 Confinement
The breaking of local color symmetry implies that all the gluons acquire mass, according to the Meissner (or alternatively Higgs) effect. There are no long-range, $`1/r`$ interactions. There is no direct signature for the color degree of freedom – although, of course, in weak coupling one clearly perceives its avatars. It is veiled or, if you like, confined.
#### 6.2.2 Chiral Symmetry Breaking
If we make a left-handed chiral rotation then we must compensate it by a color rotation, in order to leave the left-handed condensate invariant. Color rotations being vectorial, we must then in addition make a right-handed chiral rotation, in order to leave the right-handed condensate invariant. Thus chiral symmetry is spontaneously broken, by a new mechanism: although the left- and right- condensates are quite separate (and, before we include instantons – see below – not even phase coherent), because both are locked to color they are thereby locked to one another.
The spontaneous breaking of global chiral $`SU(3)_L\times SU(3)_R`$ brings with it an octet of pseudoscalar Nambu-Goldstone bosons, collective modes interpolating, in space-time, among the condensates related by the lost symmetry. These massless modes, as is familiar, are derivatively coupled, and therefore they do not generate singular long-range interactions.
#### 6.2.3 Superfluidity and ‘Nuclear’ Pairing
Less familiar, and perhaps disconcerting at first sight, is the loss of baryon number symmetry. This does not, however, portend proton decay, any more than does the non-vanishing condensate of helium atoms in superfluid $`\mathrm{He4}`$. Given an isolated finite sample, the current divergence equation can be integrated over a surface surrounding the sample, and unambiguously indicates overall number conservation. To respect it, one should project onto states with a definite number of baryons, by integrating over states with different values of the condensate phase. This does not substantially alter the physics of the condensate, however, because the overlap between states of different phase is very small for a macroscopic sample. Roughly speaking, there is a finite mismatch per unit volume, so the overlap vanishes exponentially in the limit of infinite volume. The true meaning of the formal baryon number violation is that there are low-energy states with different distributions of baryon number, and easy transport among them. Indeed, the dynamics of the condensate is the dynamics of superfluidity: gradients in the Nambu-Goldstone mode are none other than the superfluid flow.
We know from experience that large nuclei exhibit strong even-odd effects, and an extensive phenomenology has been built up around the idea of pairing in nuclei. If electromagnetic Coulomb forces didn’t spoil the fun, we could confidently expect that extended nuclear matter would exhibit the classic signatures of superfluidity. In our 3-flavor version the Coulomb forces do not come powerfully into play, since the charges of the quarks average out to electric neutrality. Furthermore, the tendency to superfluidity exhibited by ordinary nuclear matter should be enhanced by the additional channels operating coherently. So one should expect strong superfluidity at ordinary nuclear density, and it becomes less surprising that we find it at asymptotically large density too.
#### 6.2.4 Global Order Parameters
I mentioned before that the Higgs mechanism as it operates in the electroweak sector of the standard model has no gauge-invariant signature. With color-flavor locking we’re in better shape, because global as well as gauge symmetries are broken. Thus there are sharp differences between the color-flavor locked phase and the free phase. There must be phase transitions – as a function, say, of temperature – separating them.
In fact, it is a simple matter to extract gauge invariant order parameters from our primary, gauge variant condensate at weak coupling. For instance, to form a gauge invariant order parameter capturing chiral symmetry breaking we may take the product of the left-handed version of Eqn. 74 with the right-handed version and saturate the color indices, to obtain
$$q_{La}^\alpha q_{Lb}^\beta \overline{q}_{R\alpha }^c\overline{q}_{R\beta }^dq_{La}^\alpha q_{Lb}^\beta \overline{q}_{R\alpha }^c\overline{q}_{R\beta }^d(\kappa _1^2+\kappa _2^2)\delta _a^c\delta _b^d+2\kappa _1\kappa _2\delta _a^d\delta _b^c$$
(82)
Likewise we can take a product of three copies of the condensate and saturate the color indices, to obtain a gauge invariant order parameter for superfluidity. These secondary order parameters will survive gauge unfixing unscathed. Unlike the primary condensate from which they were derived, they are not just convenient fictions, but measurable realities.
#### 6.2.5 A Subtlety: Axial Baryon Number
As it stands the chiral order parameter Eqn. 82 is not quite the usual one, but roughly speaking its square. It leaves invariant an additional $`Z_2`$, under which the left-handed quark fields change sign. Actually this $`Z_2`$ is not a legitimate symmetry of the full theory, but suffers from an anomaly.
Since we can be working at weak coupling, we can be more specific. Our model Hamiltonian Eqn. 73 was abstracted from one-gluon exchange, which is the main interaction among high-energy quarks in general, and so in particular for modes near our large Fermi surfaces. The instanton interaction is much less important, at least asymptotically, both because it is intrinsically smaller for energetic quarks, and because it involves six fermion fields, and hence (one can show) is irrelevant as one renormalizes toward the Fermi surface. However, it represents the leading contribution to axial baryon number violation. In particular, it is only $`U_A(1)`$ violating interactions that fix the relative phase of our left- and right- handed condensates. So a model Hamiltonian that neglects them will have an additional symmetry that is not present in the full theory. After spontaneous breaking, which does occur in the axial baryon number channel, there will be a Nambu-Goldstone boson in the model theory, that in the full theory acquires an anomalously (pun intended) small mass. Similarly, in the full theory there will be a non-zero tertiary chiral condensate of the usual kind, bilinear in quark fields, but it will be parametrically smaller than Eqn. 82.
### 6.3 Elementary Excitations
There are three sorts of elementary excitations. They are the modes produced directly by the fundamental quark and gluon fields, and the collective modes connected with spontaneous symmetry breaking.
The quark fields of course produce spin $`1/2`$ fermions. Some of these are true long-lived quasiparticles, since there is nothing for them to decay into. They form an octet and a singlet under the residual diagonal $`SU(3)`$. There is an energy gap for production of pairs above the ground state. Actually there are two gaps: a smaller one for the octet, and a larger one for the singlet.
The gluon fields produce an octet of spin $`1`$ bosons. As previously mentioned, they acquire a mass by the Meissner-Higgs phenomenon. We have already discussed the Nambu-Goldstone bosons, too.
### 6.4 A Modified Photon
The notion of ‘confinement’ I advertised earlier, phrased in terms of mass gaps and derivative interactions, might seem rather disembodied. So it is interesting to ask whether and how a more traditional and intuitive criterion of confinement – no fractionally charged excitations – is satisfied.
Before discussing electromagnetic charge we must identify the unbroken gauge symmetry, whose gauge boson defines the physical photon in our dense medium. The original electromagnetic gauge invariance is broken, but there is a combination of the original electromagnetic gauge symmetry and a color transformation which leaves the condensate invariant. Specifically, the original photon $`\gamma `$ couples according to the matrix
$$\left(\begin{array}{ccc}\frac{2}{3}& 0& 0\\ 0& \frac{1}{3}& 0\\ 0& 0& \frac{1}{3}\end{array}\right)$$
(83)
in flavor space, with strength $`e`$. There is a gluon $`G`$ which couples to the matrix
$$\left(\begin{array}{ccc}\frac{2}{3}& 0& 0\\ 0& \frac{1}{3}& 0\\ 0& 0& \frac{1}{3}\end{array}\right)$$
(84)
in color space, with strength $`g`$. Then the combination
$$\stackrel{~}{\gamma }=\frac{g\gamma +eG}{\sqrt{e^2+g^2}}$$
(85)
leaves the ‘locking’ Kronecker deltas in color-flavor space invariant. In our medium, it represents the physical photon. What happens here is similar to what occurs in the electroweak sector of the standard model, where both weak isospin and weak hypercharge are separately broken by the Higgs doublet, but a cunning combination remains unbroken, and defines electromagnetism.
#### 6.4.1 Integer Charges!
Now with respect to $`\stackrel{~}{\gamma }`$ the electron charge is
$$\frac{eg}{\sqrt{e^2+g^2}},$$
(86)
deriving of course solely from the $`\gamma `$ piece of Eqn. 85. The quarks have one flavor and one color index, so they pick up contributions from both pieces. In each sector we find the normalized charge unit $`\frac{eg}{\sqrt{e^2+g^2}}`$, and it is multiplied by some choice from among $`(2/3,1/3,1/3)`$ or $`(2/3,1/3,1/3)`$ respectively. The total, obviously, can be $`\pm 1`$ or $`0`$. Thus the excitations produced by the quark fields are integrally charged, in units of the electron charge. Similarly the gluons have an upper color and a lower anti-color index, so that one faces similar choices, and reaches a similar conclusion. In particular, some of the gluons have become electrically charged. The pseudoscalar Nambu-Goldstone modes have an upper flavor and a lower anti-flavor index, and yet again the same conclusions follow. The superfluid mode, of course, is electrically neutral.
#### 6.4.2 A Conceptual Jewel
It is fun to consider how a chunk of our color-flavor locked material would look. If the quarks were truly massless, then so would be Nambu-Goldstone bosons (at the level of pure QCD), and one might expect a rather unusual ‘bosonic metal’, in which low-energy electromagnetic response is dominated by these modes. Actually electromagnetic radiative corrections lift the mass of the charged Nambu-Goldstone bosons, creating a gap for the charged channel. The same effect would be achieved by turning on a common non-zero quark mass. Thus the color-flavor locked material forms a transparent insulator. Altogether it would resemble a diamond, reflecting portions of incident light waves, but allowing finite portions through and out again!
### 6.5 Quark-Hadron Continuity
The universal features of the color-flavor locked state: confinement, chiral symmetry breaking down to vector $`SU(3)`$, and superfluidity, are just what one would expect, based on standard phenomenological models and experience with real-world QCD at low density. Now we see that the low-lying spectrum likewise bears an uncanny resemblance to what one finds in the Particle Data Book (or rather what one would find, in a world of three degenerate quarks). It is hard to resist the inference that there is no phase transition separating them. Thus there need not be, and presumably is not, a sharp distinction between the low-density phase, where microscopic calculations are difficult but the convenient degrees of freedom are “obviously” hadrons, and the asymptotic high-density phase, where weak-coupling (but non-perturbative) calculations are possible, and the right degrees of freedom are elementary quarks and gluons plus collective modes associated with spontaneous symmetry breaking. We call this quark-hadron continuity. It might seem shocking that a quark can “be” a baryon, but remember that it is immersed in a sea of diquark condensate, wherein the distinction between one quark and three becomes negotiable.
### 6.6 Remembrance of Things Past
An entertaining aspect of the emergent structure is that two beautiful ideas from the pre-history of QCD, that were bypassed in its later development, have come very much back to center stage, now with microscopic validation. The quark-baryons of the color-flavor locked phase follow the charge assignments proposed by Han and Nambu. And the gluon-vector mesons derive from the Yang-Mills gauge principle – as originally proposed, for rho mesons!
### 6.7 More Quarks
For larger numbers of quarks, the story is qualitatively similar. Color symmetry is broken completely, and there is a gap in all quark channels, so the weak-coupling treatment is adequate. Color-flavor locking is so favorable that there seems to be a periodicity: if the number of quarks is a multiple of three, one finds condensation into 3$`\times `$3 blocks, while if it is 4+3k or 5+3k one finds k color-flavor locking blocks together with special patterns characteristic of 4 or 5 flavors.
There is an amusing point here. QCD with a very large number of massless quarks, say 16, has an infrared fixed point at very weak coupling. Thus it should be quasi-free at zero density, forming a nonabelian Coulomb phase, featuring conformal symmetry, no confinement, and no chiral symmetry breaking. To say the least, it does not much resemble real-world QCD. There are indications that this qualitative behavior may persist even for considerably fewer quarks (the critical number might be as small as 5 or 6). Nevertheless, at high density, we have discovered, these many-quark theories all support more-or-less normal-looking ‘nuclear matter’ – including confinement and chiral symmetry breaking!
### 6.8 Fewer Quarks, and Reality
One can perform a similar analysis for two quark flavors. A new feature is that the instanton interaction now involves four rather than six quark legs, so it remains relevant as one renormalizes toward the Fermi surface. Either the one-gluon exchange or the instanton interaction, treated in the spirit above, favors condensation of the form
$$\begin{array}{ccc}q_{La}^{i\alpha }(p)q_{Lb}^{j\beta }(p)=q_{Ra}^{i\alpha }(p)q_{Lb}^{j\beta }(p)=ϵ^{ij}\kappa (p^2)ϵ^{\alpha \beta 3}ϵ_{ab}.\hfill & & \end{array}$$
(87)
Formally, Eqn. 87 is quite closely related to Eqn. 82, since $`ϵ^{\alpha \beta I}ϵ_{abI}=2(\delta _a^\alpha \delta _b^\beta \delta _b^\alpha \delta _a^\beta )`$. Their physical implications, however, are quite different.
To begin with, Eqn. 87 does not lead to gaps in all quark channels. The quarks with color labels 1 and 2 acquire a gap, but quarks of the third color of quark are left untouched. Secondly, the color symmetry is not completely broken. A residual $`SU(2)`$, acting among the first two colors, remains valid. For these reasons, perturbation theory about the ground state defined by Eqn. 87 is not free of infrared divergences, and we do not have a fully reliable grip on the physics.
Nevertheless it is plausible that the qualitative features suggested by Eqn. 87 are not grossly misleading. The residual $`SU(2)`$ presumably produces confined glueballs of large mass, and assuming this occurs, the residual gapless quarks are weakly coupled.
Assuming for the moment that no further condensation occurs, for massless quarks we have the symmetry breaking pattern
$$SU(3)^c\times SU(2)_L\times SU(2)_R\times U(1)_BSU(2)^c\times SU(2)_L\times SU(2)_R\times \stackrel{~}{U}(1)_B$$
(88)
Here the modified baryon number acts only on the third color of quarks. It is a combination of the original baryon number and a color generator, that are separately broken but when applied together leave the condensate invariant. Comparing to the zero-density ground state, one sees that color symmetry is reduced, chiral symmetry is restored, and baryon number is modified. Only the restoration of chiral symmetry is associated with a legitimate order parameter, and only it requires a sharp phase transition.
In the real world, with the $`u`$ and $`d`$ quarks light but not strictly massless, there is no rigorous argument that a phase transition is necessary. It is (barely) conceivable that one might extend quark-hadron continuity to this case. Due to medium modifications of baryon number and electromagnetic charge the third-color $`u`$ and $`d`$ quarks have the quantum numbers of nucleons. The idea that chiral symmetry is effectively restored in nuclear matter, however, seems problematic quantitatively. More plausible, perhaps, is that there is a first-order transition between nuclear matter and quark matter. This is suggested by some model calculations, and is the basis for an attractive interpretation of the MIT bag model, according to which baryons are droplets wherein chiral symmetry is restored.
#### 6.8.1 Thresholds and Mismatches
In the real world there are two quarks, $`u`$ and $`d`$, whose mass is much less than $`\mathrm{\Lambda }_{QCD}`$, and one, $`s`$, whose mass is comparable to it. Two simple qualitative effects, that have major implications for the zero-temperature phase diagram, arise as consequences of this asymmetric spectrum. They are expected, whether one analyzes from the quark side or from the hadron side.
The first is that one can expect a threshold, in chemical potential (or pressure), for the appearance of any strangeness at all in the ground state. This will certainly hold true in the limit of large strange quark mass, and there is considerable evidence for it in the real world. This threshold is in addition to the threshold transitions at lower chemical potentials, from void to nuclear matter, and (presumably) from nuclear matter to two-flavor quark matter, as discussed above.
The second is that at equal chemical potential the Fermi surfaces of the different quarks will not match. This mismatch cuts off the Cooper instability in mixed channels. If the nominal gap is large compared to the mismatch, one can treat the mismatch as a perturbation. This will always be valid at asymptotically high densities, since the mismatch goes as $`m_s^2/\mu `$, whereas the gap eventually grows with $`\mu `$. If the nominal gap is small compared to the mismatch, condensation will not occur.
#### 6.8.2 Assembling the Pieces
With these complications in mind, we can identify three major phases in the plane of chemical potential and strange quark mass, that reflect the simple microscopic physics we have surveyed above. (There might of course be additional “minor” phases – notably including normal nuclear matter!) There is 2-flavor quark matter, with restoration of chiral symmetry, and zero strangeness. Then there is a 2+1-flavor phase, in which the strange and non-strange Fermi surfaces are badly mismatched, and one has independent dynamics for the corresponding low energy excitations. Here one expects strangeness to break spontaneously, by its own Fermi surface instability. Finally there is the color-flavor locked phase. A caricature version of the phase diagram in the $`(m_s,\mu )`$ plane, illustrating these features, is given in Figure 13.
#### 6.8.3 Reality
The progress reported here, while remarkable, mainly concerns the asymptotic behavior of high-density QCD. Its extrapolation to practical densities is at present semi-quantitative at best. To do real justice to the potential applications, we need to learn how to do more accurate analytical and numerical work at moderate densities.
As regards analytical work, we can take heart from some recent progress on the equation of state at high temperature. Here there are extensive, interesting numerical results, which indicate that the behavior is quasi-free, but that there are very significant quantitative corrections to free quark-gluon plasma results, especially for the pressure. Thus it is plausible a priori that some weak-coupling, but non-perturbative, approach will be workable, and this seems to be proving out. An encouraging feature here is that the analytical techniques used for high temperature appear to be capable of extension to finite density without great difficulty.
Numerical work at finite density, unfortunately, is plagued by poor convergence. This arises because the functional integral is not positive definite configuration by configuration, so that importance sampling fails, and one is left looking for a small residual from much larger canceling quantities.
There are cases in which this problem does not arise. It does not arise for two colors. Although low-density hadronic matter is quite different in a two-color world than a three-color world – the baryons are bosons, so one does not get anything like a shell structure for nuclei – I see no reason to expect that the asymptotic, high-density phases should be markedly different. It would be quite interesting to see Fermi-surface behavior arising for two colors at high density (especially, for the ground state pressure), and even more interesting to see the effect of diquark condensations.
Another possibility, that I have been discussing with David Kaplan, is to engineer lattice gauge theories whose low-energy excitations resemble those of finite density QCD near the Fermi surface, but which are embedded in a theory that is globally particle-hole symmetric, and so feature a positive-definite functional integral.
Aside from these tough quantitative issues, there are a number of directions in which the existing work should be expanded and generalized, that appear to be quite accessible. There is already a rich and important theory of the behavior of QCD at non-zero temperature and zero baryon number density. We should construct a unified picture of the phase structure as a function of both temperature and density; to make it fully illuminating, we should also allow at least the strange quark mass to vary. We should allow for the effect of electromagnetism (after all, this is largely what makes neutron stars what they are) and of rotation. We should consider other possibilities than a common chemical potential for all the quarks.
As physicists we should not, however, be satisfied with hoarding up formal, abstract knowledge. There are concrete experimental situations and astrophysical objects we must speak to. Hopefully, having mastered some of the basic vocabulary and grammar, we will soon be in a better position to participate in a two-way dialogue with Nature.
Background Material
* General background on quantum field theory and QCD:
1. T.-P. Cheng and L.-F. Li, Gauge Theory of Elementary Particles (Oxford University Press, London 1984),
2. M. Peskin and D. Schroder, Introduction to Quantum Field Theory (Addison-Wesley, Redwood City California, 1995).
* Lattice gauge theory:
1. M. Creutz, Quarks, Gluons, and Lattices (Cambridge University Press, Cambridge U.K., 1983).
* Chiral symmetry and current algebra:
1. S. Weinberg, Quantum Theory of Fields II chapter 19 (Cambridge University Press, Cambridge U.K., 1996).
* Renormalization group for critical phenomena:
1. D. Amit, Field Theory, the Renormalization Group, and Critical Phenomena (World Scientific, Singapore 1984).
* Renormalization group toward the Fermi surface:
1. J. Polchinski, hep-th/9210046,
2. R. Shankar, Rev. Mod. Phys. 66, 129 (1993).
* Superconductivity:
1. J. R. Schrieffer, Theory of Superconductivity (Benjamin-Cummins, Reading Mass., 1984).
* Tricritical points:
1. I. Lawrie and S. Sarbach, in Phase Transitions and Critical Phenomena 9 eds. C. Domb and J. Lebowitz (Academic, New York 1984).
* Quark-gluon plasma:
1. U. Heniz and M. Jacob, nucl-th/0002042.
Sources and Further Reading
* The perspective on QCD adopted in Lecture 1, as a series of interweaving stories of symmetry and dynamics, is taken from my monograph QCD in preparation for Princeton University Press.
* The material on high-temperature phase transitions in Lectures 2 and 3 is mostly taken from
1. F. Wilczek, Int. Jour. Mod. Phys. A7 3911, (1992).
2. K. Rajagopal and F. Wilczek, Nucl. Phys. B399) 395, (1993).
* The possibility of a true critical point was pointed out in M. Stephanov, K. Rajagopal, and E. Shuryak Phys. Rev. Lett. 81, 4816 (1998).
* Its possible experimental signatures are discussed in M. Stephanov, hep-ph/9906242.
* For recent lattice results see F. Karsch, hep-lat/9909006.
* Lectures 4 and 5 are based on F. Wilczek, hep-ph/9908480.
* For further developments see especially T. Schäfer, hep-ph/9909574.
These papers contain numerous additional references. Obviously, several topics discussed in the Lectures are under very active development, and you should consult the web for up-to-date results.
|
warning/0003/hep-th0003166.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
D-branes play important roles to describe solitonic modes in string theory. The physical observables of D-brane’s effective theories have dependences of moduli on compactified internal spaces or wrapped D-branes. In this paper, we focus on the central charge of the type II superstring compactified on Calabi-Yau manifold and study its properties from the point of view of topological sigma models (A- and B-models)-. The central charge is characterized by D-brane’s charges and periods of the B-model in the closed string channel. Together with the Kähler potential $`K`$, it allows us to construct a BPS mass formula of D-branes wrapped around cycles.
Recently there is a great advance to study properties of charges, boundary states, based on the Gepner model associated with the CY<sub>3</sub>. Also there appear many consistency checks about these charges, intersection forms of homology cycles by analyses both in the CFT and in the sigma models. They investigated three dimensional Calabi-Yau cases.
For these three fold cases, one can easily construct associated canonical bases of period integrals because there are prepotentials for the models. It makes analyses for three folds comparatively tractable. But for $`d`$-fold cases $`(d>3)`$, we do not know any existence of analogs of these convenient prepotentials and we cannot apply analogous recipes to $`d`$-fold cases $`(d>3)`$ directly. The aim of this paper is to develop a method to construct central charges $`Z`$ of Calabi-Yau $`d`$-folds and to investigate their properties in order to understand structures of the moduli spaces in the open string channel. We interpret constituent elements of the $`Z`$ as topological objects from the point of view of topological sigma models.
The paper is organized as follows. In section 2, we explain a mirror manifold paired with a Calabi-Yau $`d`$-fold embedded in $`CP^{d+1}`$. We also explain the results in , about a Kähler potential in order to fix notations. There we introduce a few sets of periods applicable either in the small or large complex structure regions. We construct a formula of the central charge $`Z`$ by using the periods. In section 3, we investigate the results by P. Candelas et al in our bases and develop a method to construct the $`Z`$. By generalizing a consideration in the quintic case, we apply the method to the $`5`$-fold case and construct the central charge $`Z`$ concretely in section 4. Also the monodromy matrices associated with singular points are investigated. In section 5, we review the Gepner model shortly and analyze D-brane’s charges both in large radius basis and in the Gepner basis with B-type boundary conditions. In section 6, cycles associated with the periods are constructed explicitly and intersection forms of the cycles are studied. In section 7, we propose formulae of the $`Z`$ applicable in the large volume region and make a consideration about an analogous structure to the Mukai vector in the $`Z`$. It is interpreted as a product of Chern characters of sheaves and a square root of the Todd class (or the A-roof genus) of the $`d`$-fold $`M`$. The set of sheaves is constructed as a dual basis of tautological line bundles of the ambient space $`CP^{N1}`$ by a restriction on $`M`$. Section 8 is devoted to conclusions and comments. In appendix A, we summarize several examples of the $`\sqrt{\widehat{A}}`$ in lower dimensional cases. In appendix B, we collect several data about monodromy properties of the quintic.
## 2 Periods and Kähler Potential
In our previous papers,, we determine the formula of the Kähler potential of the Calabi-Yau $`d`$-fold embedded in $`CP^{d+1}`$
$`M;p=X_1^N+X_2^N+\mathrm{}+X_N^NN\psi X_1X_2\mathrm{}X_N=0,`$
and that of the mirror partner with a moduli parameter $`\psi `$ of the complex structure
$`W;\widehat{\{p=0\}/𝐙_N^{(N1)}}.`$ (1)
The $`N`$ is related with the complex dimension $`d`$ of M, $`N=d+2`$. The $`G=𝐙_N^{(N1)}`$ is a maximally discrete group of the $`M`$. The formula is constructed by requiring consistency conditions with the results of the CFT at the Gepner point. In this paper, we investigate the central charge of the topological A-model associated with the $`M`$ (equivalently, the open string with B-type boundary conditions). The formula is important for BPS mass analyses of the D-branes.
First we review the results in ,. When one considers the Hodge structure of the $`G`$-invariant parts of the cohomology group H$`{}_{}{}^{d}(W)`$, the decomposition of the structure is controlled by one complex moduli parameter $`\psi `$. The Kähler potential $`K`$ in the B-model moduli of $`W`$ is constructed by combining a set of periods $`\stackrel{~}{\varpi }_k`$ quadratically
$`e^K={\displaystyle \underset{k=1}{\overset{N1}{}}}I_k\stackrel{~}{\varpi }_k^{}\stackrel{~}{\varpi }_k,`$
$`I_k={\displaystyle \frac{1}{\pi ^NN^{N+2}}}(1)^{k1}\left(\mathrm{sin}{\displaystyle \frac{\pi k}{N}}\right)^N,`$
$`\stackrel{~}{\varpi }_k(\psi )=\left[\mathrm{\Gamma }\left({\displaystyle \frac{k}{N}}\right)\right]^N{\displaystyle \frac{(N\psi )^k}{\mathrm{\Gamma }(k)}}`$
$`\times \left[{\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}\left[{\displaystyle \frac{\mathrm{\Gamma }\left(\frac{k}{N}+n\right)}{\mathrm{\Gamma }\left(\frac{k}{N}\right)}}\right]^N{\displaystyle \frac{\mathrm{\Gamma }(k)}{\mathrm{\Gamma }(Nn+k)}}(N\psi )^{Nn}\right].`$ (2)
The coefficients $`I_k`$ are determined in .
### 2.1 Periods
The formulae Eqs.(2) are valid in the small $`\psi `$ region because of the convergence of the series expansions. At a point $`\psi =0`$ in the B-model moduli space, there is a $`𝐙_N`$ symmetry which rotates the $`\psi \alpha \psi `$ ($`\alpha =e^{2\pi i/N}`$). A cyclic $`𝐙_N`$ monodromy transformation $`𝒜`$ is diagonalized on the set of this basis $`\stackrel{~}{\varpi }_k`$ ($`1kN1`$) with $`\alpha =e^{2\pi i/N}`$
$`𝒜\stackrel{~}{\varpi }_k(\psi )=\alpha ^k\stackrel{~}{\varpi }_k(\psi )(k=1,2,\mathrm{},N1).`$
For a later convenience, we also introduce a set of periods $`\varpi _j`$ ($`0jN1`$) as linear combinations of the $`\stackrel{~}{\varpi }_k`$s
$`\varpi _j={\displaystyle \frac{1}{N}}{\displaystyle \frac{1}{(2\pi i)^{N1}}}{\displaystyle \underset{k=1}{\overset{N1}{}}}\alpha ^{jk}(\alpha ^k1)^{N1}\stackrel{~}{\varpi }_k.`$
The $`𝐙_N`$ transformation acts on this basis cyclically
$`𝒜\varpi _j(\psi )=\varpi _{j+1}(\psi ),(j=0,1,2,\mathrm{},N1).`$
Here we identify the $`\varpi _N`$ with the $`\varpi _0`$. It is a redundant basis to represent the monodromy transformation because a linear relation is satisfied
$`{\displaystyle \underset{j=0}{\overset{N1}{}}}\varpi _j=0.`$
But the basis is useful because it is directly related to the Gepner basis of the CFT. These two sets of periods $`\stackrel{~}{\varpi }_k`$ and $`\varpi _j`$ are meaningful only in the small $`\psi `$ region because of the convergence of the series expansions. In order to describe the large complex structure region of $`W`$, we must introduce another set of periods $`\{\mathrm{\Omega }_m\}`$s ($`m=0,1,2,\mathrm{},N2`$). A generating function of the $`\mathrm{\Omega }_m`$ is defined by using a formal parameter $`\rho `$ with $`\rho ^{N1}=0`$
$`{\displaystyle \underset{m=0}{\overset{N2}{}}}\mathrm{\Omega }_m\rho ^m=\sqrt{\widehat{K}(\rho )}\varpi ({\displaystyle \frac{\rho }{2\pi i}};z),`$
$`\varpi (v)=z^v{\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{a(n+v)}{a(v)}}z^n,z=(N\psi )^N,`$ (3)
$`a(v)={\displaystyle \frac{\mathrm{\Gamma }(Nv+1)}{[\mathrm{\Gamma }(v+1)]^N}},`$
$`\widehat{K}(\rho ):=\mathrm{exp}\left[2{\displaystyle \underset{m=1}{}}{\displaystyle \frac{NN^{2m+1}}{2m+1}}\zeta (2m+1)\left({\displaystyle \frac{\rho }{2\pi i}}\right)^{2m+1}\right]`$
$`=1+2\zeta (3){\displaystyle \frac{c_3}{N}}\left({\displaystyle \frac{\rho }{2\pi i}}\right)^3+𝒪(\rho ^5).`$
The infinite series Eq.(3) converges around the large complex structure point $`z0`$ of W. We find that the two sets of the periods $`\stackrel{~}{\varpi }_k`$ and $`\mathrm{\Omega }_{\mathrm{}}`$ are related by a transformation matrix $`\stackrel{~}{M}`$ with components $`\stackrel{~}{M}_k\mathrm{}`$ through an analytic continuation into the large complex structure region
$`\stackrel{~}{\varpi }_k={\displaystyle \underset{\mathrm{}=0}{\overset{N2}{}}}\stackrel{~}{M}_k\mathrm{}\mathrm{\Omega }_{\mathrm{}}(k=1,2,\mathrm{},N1),`$
$`\stackrel{~}{M}_k\mathrm{}=(N)(2\pi i)^{N1}\times \left[\sqrt{\widehat{A}(\rho )}{\displaystyle \frac{\alpha ^k}{e^\rho \alpha ^k}}(\rho )^{\mathrm{}}\right]|_{\rho ^{N2}}`$
$`=(N)(2\pi i)^{N1}{\displaystyle \underset{m=0}{\overset{N2}{}}}G_{k,m}V_{m,\mathrm{}},`$
$`G_{k,m}={\displaystyle \frac{\alpha ^k}{(\alpha ^k1)^{m+1}}}(1kN1,\mathrm{\hspace{0.17em}\hspace{0.17em}0}mN2),`$
$`V_{m,\mathrm{}}=[\sqrt{\widehat{A}(\rho )}(e^\rho 1)^m(\rho )^{\mathrm{}}]|_{\rho ^{N2}}(0mN2,\mathrm{\hspace{0.17em}\hspace{0.17em}0}\mathrm{}N2).`$ (4)
The transformation matrix $`V`$ contains a square root of a topological invariant “A-roof genus” of the Calabi-Yau space
$`\widehat{A}(\rho )=\left({\displaystyle \frac{{\displaystyle \frac{\rho }{2}}}{\mathrm{sinh}{\displaystyle \frac{\rho }{2}}}}\right)^N\left({\displaystyle \frac{\mathrm{sinh}{\displaystyle \frac{N\rho }{2}}}{{\displaystyle \frac{N\rho }{2}}}}\right)`$
$`=\mathrm{exp}\left[+{\displaystyle \underset{m=1}{}}{\displaystyle \frac{(1)^mB_m}{(2m)!}}{\displaystyle \frac{NN^{2m}}{2m}}\rho ^{2m}\right]`$
$`=1+{\displaystyle \frac{1}{12}}{\displaystyle \frac{c_2}{N}}\rho ^2+𝒪(\rho ^4).`$
The $`B_m`$s are Bernoulli numbers and are defined in our convention as
$`{\displaystyle \frac{x}{e^x1}}=1{\displaystyle \frac{x}{2}}{\displaystyle \underset{n=1}{\overset{\mathrm{}}{}}}{\displaystyle \frac{(1)^nB_n}{(2n)!}}x^{2n},`$
$`B_1={\displaystyle \frac{1}{6}},B_2={\displaystyle \frac{1}{30}},B_3={\displaystyle \frac{1}{42}},B_4={\displaystyle \frac{1}{30}},\mathrm{}.`$
The expansion coefficients of the $`\widehat{A}(\rho )`$ in terms of the $`\rho `$ are represented as some combinations of Chern classes of the $`M`$. We summarize several concrete examples of the $`\sqrt{\widehat{A}(\rho )}`$ in the appendix A. Also we define auxiliary sets of periods $`\{\widehat{\mathrm{\Pi }}_n\}`$ and $`\{\widehat{\widehat{\mathrm{\Pi }}}_m\}`$ ($`n,m=0,1,2,\mathrm{},N2`$)
$`\widehat{\mathrm{\Pi }}_n={\displaystyle \underset{\mathrm{}=0}{\overset{N2}{}}}V_{n,\mathrm{}}\mathrm{\Omega }_{\mathrm{}},\widehat{\widehat{\mathrm{\Pi }}}_m={\displaystyle \underset{n=0}{\overset{N2}{}}}U_{m,n}\mathrm{\Pi }_n,`$
$`V_{n,\mathrm{}}=\left[\sqrt{\widehat{A}(\rho )}(e^\rho 1)^n(\rho )^{\mathrm{}}\right]_{\rho ^{N2}},U_{m,n}=\left(\begin{array}{c}m\\ n\end{array}\right)(1)^{mn}.`$
Here we introduced a symbol $`\left(\begin{array}{c}a\\ b\end{array}\right)`$ as a ratio of Euler’s gamma functions
$`\left(\begin{array}{c}a\\ b\end{array}\right):={\displaystyle \frac{\mathrm{\Gamma }(a+1)}{\mathrm{\Gamma }(b+1)\mathrm{\Gamma }(ab+1)}}.`$
The $`\widehat{\mathrm{\Pi }}`$ is related to the basis $`\varpi _j`$ through an analytic continuation
$`\varpi _j={\displaystyle \underset{n=0}{\overset{N2}{}}}P_{j,n}\widehat{\mathrm{\Pi }}_n,`$
$`P_{j,n}=\delta _{n,N2}N\left(\begin{array}{c}N2n\\ j1n\end{array}\right)\times (1)^{j1n},\left(\begin{array}{c}j=0,1,\mathrm{},N1\\ n=0,1,\mathrm{},N2\end{array}\right).`$ (5)
In discussing properties in the large $`\psi `$ region, we use the sets $`\widehat{\mathrm{\Pi }}_n`$ or $`\mathrm{\Omega }_m`$.
### 2.2 Kähler Potential
Under this preparation, we shall write down our results in the Kähler potential $`K`$ in these bases
$`e^K={\displaystyle \underset{\mathrm{}=1}{\overset{N1}{}}}\stackrel{~}{\varpi }_{\mathrm{}}^{}I_{\mathrm{}}\stackrel{~}{\varpi }_{\mathrm{}}`$
$`={\displaystyle \frac{(2\pi i)^{N2}}{N^{N+2}}}{\displaystyle \underset{ja}{}}{\displaystyle \underset{j^{}a}{}}\varpi _j^{}𝒦_{j,j^{}}\varpi _j^{}`$
$`={\displaystyle \frac{1}{(2\pi iN)^N}}{\displaystyle \frac{1}{N^2}}{\displaystyle \underset{m,n=0}{\overset{N2}{}}}\widehat{\mathrm{\Pi }}_m^{}_{m,n}\widehat{\mathrm{\Pi }}_n,`$
$`=(1)^N\left({\displaystyle \frac{2\pi i}{N}}\right)^{N2}{\displaystyle \frac{1}{N^2}}{\displaystyle \underset{\mathrm{},\mathrm{}^{}=0}{\overset{N2}{}}}\mathrm{\Omega }_{\mathrm{}}^{}\mathrm{\Sigma }_\mathrm{},\mathrm{}^{}\mathrm{\Omega }_{\mathrm{}^{}},`$
$`I_{\mathrm{}}={\displaystyle \frac{(1)^\mathrm{}1}{\pi ^NN^{N+2}}}\left(\mathrm{sin}{\displaystyle \frac{\pi \mathrm{}}{N}}\right)^N,(\mathrm{}=1,2,\mathrm{},N1),`$
$`𝒦_{j,j^{}}={\displaystyle \frac{i^N}{2^{N2}}}\times {\displaystyle \underset{\mathrm{}=1}{\overset{N1}{}}}(1)^{\mathrm{}}\left(\mathrm{sin}{\displaystyle \frac{\pi \mathrm{}}{N}}\right)^{(N2)}\times (\alpha ^j\mathrm{}\alpha ^a\mathrm{})(\alpha ^j^{}\mathrm{}\alpha ^a\mathrm{}),`$
$`(j,j^{}=0,1,2,\mathrm{},N1),`$
$`_{m,n}=2^Ni^N{\displaystyle \underset{k=1}{\overset{N1}{}}}{\displaystyle \frac{(1)^k\left(\mathrm{sin}{\displaystyle \frac{\pi k}{N}}\right)^N}{(\alpha ^k1)^{m+1}(\alpha ^k1)^{n+1}}},(m,n=0,1,2,\mathrm{},N2),`$
$`\mathrm{\Sigma }_\mathrm{},\mathrm{}^{}=(1)^{\mathrm{}}\delta _{\mathrm{}+\mathrm{}^{},N2},(\mathrm{},\mathrm{}^{}=0,1,2,\mathrm{},N2).`$
In the above formulae, the $`I_k`$, $`𝒦_{j,j^{}}`$, $`_{m,n}`$ and $`\mathrm{\Sigma }_\mathrm{},\mathrm{}^{}`$ are intersection matrices associated with homology cycles for the corresponding periods $`\stackrel{~}{\varpi }_k`$, $`\varpi _j`$, $`\widehat{\mathrm{\Pi }}_m`$ and $`\mathrm{\Omega }_{\mathrm{}}`$. The matrix $`I`$ has a diagonal form. In contrast, the $`𝒦_{j,j^{}}`$ and $`_{m,n}`$ are lower triangular matrices with non-vanishing components in the right lower entries with $`j+j^{}N2`$, $`m+nN2`$. Also the determinant of the $`𝒦`$ is not unit, but $`det𝒦=N^2`$. So the associated cyclic basis is not a canonical one. The set of the $`\mathrm{\Omega }_{\mathrm{}}`$s has an intersection matrix $`\mathrm{\Sigma }_\mathrm{},\mathrm{}^{}`$ with non-vanishing components at $`\mathrm{}+\mathrm{}^{}=N2`$. The form of the $`\mathrm{\Sigma }_\mathrm{},\mathrm{}^{}`$ means that the set of periods $`\mathrm{\Omega }_{\mathrm{}}`$s is a symplectic or an SO-invariant basis and associated homology cycles have appropriate intersection forms. But it is not an integral basis and the associated cycles belong to homology classes $`_{\mathrm{}=0}^d\text{H}_2\mathrm{}(M;𝐐)`$ or $`\text{H}_d(W;𝐐)`$ respectively in the A-, B-models. In order to obtain a set of canonical basis $`\{\mathrm{\Pi }_m\}`$ ($`m=0,1,2,\mathrm{},N2`$), we have to perform some linear transformation on the $`\mathrm{\Omega }_{\mathrm{}}`$
$`\mathrm{\Pi }_m={\displaystyle \underset{\mathrm{}=0}{\overset{N2}{}}}𝒩_{m,\mathrm{}}\mathrm{\Omega }_{\mathrm{}}(m=0,1,2,\mathrm{},N2).`$
The transformation matrix $`𝒩`$ generally has entries with fractional rational numbers. The basis $`\mathrm{\Pi }`$ is needed to discuss the D-brane charges $`Q_{\mathrm{}}`$s or a central charge $`Z`$ in the BPS mass formula.
### 2.3 Central Charge
In the B-model case in the open string channel<sup>1</sup><sup>1</sup>1So far we used the words “A-”, “B-”models for the closed string case. But the definition is exchanged when we consider properties in the open string channels. , there appear even dimensional D$`p`$-branes ($`p=0,2,4,\mathrm{},2d`$) which wrap around homology cycles $`\mathrm{\Sigma }_p`$ ($`p=0,2,4,\mathrm{},2d`$) of the Calabi-Yau $`d`$-fold $`M`$. The brane charge $`Q_{2d2\mathrm{}}`$ associated with the D-brane is defined by integrating a gauge field, more precisely a Mukai vector $`v()`$ associated with a bundle (sheaf) $``$ over the cycle $`\mathrm{\Sigma }_2\mathrm{}M`$ ($`\mathrm{}=0,1,2,\mathrm{},d`$)
$`Q_{2d2\mathrm{}}={\displaystyle _{\mathrm{\Sigma }_2\mathrm{}}}v()(\mathrm{}=0,1,2,\mathrm{},d).`$
By combining these charges $`Q_p`$ and the canonical basis $`\mathrm{\Pi }_{\mathrm{}}`$ with the B-type boundary conditions, we can construct the central charge $`Z`$
$`Z={\displaystyle \underset{\mathrm{}=0}{\overset{d}{}}}Q_2\mathrm{}\mathrm{\Pi }_{\mathrm{}}.`$
It is a constituent block of a BPS mass formula $`m_{BPS}e^{+K/2}|Z|`$ in the curved space. We can represent this canonical basis $`\mathrm{\Pi }_{\mathrm{}}`$ in the bases $`\mathrm{\Omega }_{\mathrm{}}`$, $`\widehat{\mathrm{\Pi }}_{\mathrm{}}`$ and $`\varpi _j`$
$`\mathrm{\Pi }_n={\displaystyle \underset{\mathrm{}=0}{\overset{N2}{}}}𝒩_{n,\mathrm{}}\mathrm{\Omega }_{\mathrm{}}={\displaystyle \underset{\mathrm{}=0}{\overset{N2}{}}}S_{n,\mathrm{}}\widehat{\mathrm{\Pi }}_{\mathrm{}}`$
$`={\displaystyle \underset{\stackrel{0jN1}{ja}}{}}m_{n,j}\varpi _j(0nN2).`$
There are two linear relations among the undetermined matrices $`𝒩`$, $`S`$ and $`m`$ with Eqs.(4),(5)
$`𝒩_n\mathrm{}={\displaystyle \underset{m=0}{\overset{N2}{}}}S_{nm}V_m\mathrm{},(0nN2;\mathrm{\hspace{0.17em}0}\mathrm{}N2),`$
$`𝒮_n\mathrm{}={\displaystyle \underset{\stackrel{0jN1}{ja}}{}}m_{nj}P_j\mathrm{},(0nN2;\mathrm{\hspace{0.17em}0}\mathrm{}N2).`$
If we can determine one of these matrices $`𝒩`$, $`S`$ and $`m`$, the canonical basis $`\mathrm{\Pi }_{\mathrm{}}`$ is fixed. Here the “canonical” condition means that the set $`\{\mathrm{\Pi }_{\mathrm{}}\}`$ is a symplectic (or an SO-invariant) and an integral basis of monodromy transformations. The symplectic (or SO-invariant) condition is reduced to that on the $`𝒩`$ with an appropriate integer $`\lambda `$ and the matrix $`\mathrm{\Sigma }`$
$`𝒩^t\mathrm{\Sigma }𝒩=\lambda \mathrm{\Sigma }.`$
But generally the matrix $`𝒩`$ may have fractional components.
## 3 Quintic
In order to exemplify our considerations, we shall study the result of the quintic. P. Candelas et al directly constructed the matrix $`m`$, which connects the two bases $`\mathrm{\Pi }_{\mathrm{}}`$ and $`\varpi _j`$
$`\mathrm{\Pi }:={}_{}{}^{t}\left(\begin{array}{cccc}\mathrm{\Pi }_0& \mathrm{\Pi }_1& \mathrm{\Pi }_2& \mathrm{\Pi }_3\end{array}\right),`$
$`\varpi :={}_{}{}^{t}\left(\begin{array}{cccc}\varpi _0& \varpi _1& \varpi _2& \varpi _4\end{array}\right),`$
$`m=\left(\begin{array}{cccc}1& 0& 0& 0\\ \frac{2}{5}& \frac{2}{5}& \frac{1}{5}& \frac{1}{5}\\ \frac{21}{5}& \frac{1}{5}& \frac{3}{5}& \frac{8}{5}\\ 1& 1& 0& 0\end{array}\right),m^1:=\left(\begin{array}{cccc}1& 0& 0& 0\\ 1& 0& 0& 1\\ 4& 8& 1& 3\\ 4& 3& 1& 1\end{array}\right).`$
These $`\mathrm{\Pi }_{\mathrm{}}`$s are canonical periods and are described by using a prepotential<sup>2</sup><sup>2</sup>2This “$`F`$” is not a gauge field “$`F`$” on the D-brane. I believe that there is no confusion in these notations. $`F`$
$`F={\displaystyle \frac{\kappa }{6}}t^3+{\displaystyle \frac{1}{2}}at^2+bt+{\displaystyle \frac{1}{2}}c+f,`$
$`\kappa =5,a={\displaystyle \frac{11}{2}},b={\displaystyle \frac{25}{12}},c={\displaystyle \frac{\chi \zeta (3)}{(2\pi i)^3}},\chi =200,`$
$`\mathrm{\Pi }=\left(\begin{array}{c}\mathrm{\Pi }_0\\ \mathrm{\Pi }_1\\ \mathrm{\Pi }_2\\ \mathrm{\Pi }_3\end{array}\right)=\left(\begin{array}{c}1\\ t\\ _tF\\ t_tF2F\end{array}\right)\times \mathrm{\Pi }_0,`$ (6)
$`2\pi it=\mathrm{log}z+{\displaystyle \frac{{\displaystyle \underset{n=1}{\overset{\mathrm{}}{}}}{\displaystyle \frac{(5n)!}{(n!)^5}}\left({\displaystyle \underset{m=n+1}{\overset{5n}{}}}{\displaystyle \frac{5}{m}}\right)z^n}{{\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{(5n)!}{(n!)^5}}z^n}},q=e^{2\pi it}.`$
The prepotential $`F`$ of $`M`$ is expressed as a sum of a polynomial part of $`t`$ and a non-perturbative part $`f`$. We choose a set of the $`\mathrm{\Omega }_{\mathrm{}}`$s in the $`N=5`$ case
$`\varpi \left({\displaystyle \frac{\rho }{2\pi i}}\right)\sqrt{\widehat{K}(\rho )}=:{\displaystyle \underset{\mathrm{}0}{}}\rho ^{\mathrm{}}\mathrm{\Omega }_{\mathrm{}},`$
$`\sqrt{\widehat{A}(\rho )}=1+{\displaystyle \frac{5}{12}}\rho ^2,\sqrt{\widehat{K}(\rho )}=1{\displaystyle \frac{40}{(2\pi i)^3}}\zeta (3)\rho ^3,`$
$`\widehat{c}={\displaystyle \frac{40}{(2\pi i)^3}}\zeta (3)={\displaystyle \frac{1}{5}}c,`$
$`\mathrm{\Omega }=\left(\begin{array}{c}\mathrm{\Omega }_0\\ \mathrm{\Omega }_1\\ \mathrm{\Omega }_2\\ \mathrm{\Omega }_3\end{array}\right)=\left(\begin{array}{c}1\\ t\\ \frac{1}{2}t^2+S_2(0,x_2)\\ \frac{1}{6}t^3\widehat{c}+tS_2(0,x_2)+S_3(0,x_2,x_3)\end{array}\right)\times \mathrm{\Omega }_0,`$ (7)
$`x_n={\displaystyle \frac{1}{(2\pi i)^n}}{\displaystyle \frac{1}{n!}}_\rho ^n\mathrm{log}\left[{\displaystyle \underset{m=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{\mathrm{\Gamma }(N(m+\rho )+1)}{\mathrm{\Gamma }(N\rho +1)}}\left({\displaystyle \frac{\mathrm{\Gamma }(\rho +1)}{\mathrm{\Gamma }(m+\rho +1)}}\right)^Nz^m\right]|_{\rho =0}.`$
By comparing Eq.(6) and Eq.(7), we can obtain a matrix $`𝒩`$
$`\mathrm{\Pi }=𝒩\mathrm{\Omega },𝒩=\left(\begin{array}{cccc}1& 0& 0& 0\\ 0& 1& 0& 0\\ \frac{25}{12}& \frac{11}{2}& 5& 0\\ 0& \frac{25}{12}& 0& 5\end{array}\right).`$ (8)
The $`𝒩`$ is a kind of a symplectic matrix and satisfies a relation
$`𝒩^t\mathrm{\Sigma }𝒩=(5)\mathrm{\Sigma }.`$
Also one can check this $`𝒩`$ and the $`m`$ satisfy an equation
$`𝒩=mPV,`$
by using definitions of the $`V`$ and $`P`$ in Eqs.(4),(5)
$`PV=\left(\begin{array}{cccc}1& 0& 0& 0\\ 1& \frac{25}{12}& 0& 5\\ \frac{23}{12}& \frac{15}{4}& 5& 15\\ \frac{23}{12}& \frac{55}{12}& 5& 5\end{array}\right).`$
In the $`\widehat{\mathrm{\Pi }}`$ basis, this result is also expressed as
$`\mathrm{\Pi }=S\widehat{\mathrm{\Pi }},S=\left(\begin{array}{cccc}0& 0& 0& 1\\ 0& 0& 1& 1\\ 0& 5& 8& 3\\ 5& 0& 0& 0\end{array}\right)=mP.`$
Then the intersection form in the $`\widehat{\mathrm{\Pi }}`$ basis is obtained as
$`S^t\mathrm{\Sigma }S=\left(\begin{array}{cccc}0& 0& 0& 5\\ 0& 0& 5& 5\\ 0& 5& 0& 5\\ 5& 5& 5& 0\end{array}\right),\mathrm{\Sigma }=\left(\begin{array}{cccc}0& 0& 0& 1\\ 0& 0& 1& 0\\ 0& 1& 0& 0\\ 1& 0& 0& 0\end{array}\right).`$
It coincides with the matrix $`_{m,n}`$ in our result for the $`N=5`$ case
$`S^t\mathrm{\Sigma }S=5,=\left(\begin{array}{cccc}0& 0& 0& 1\\ 0& 0& 1& 1\\ 0& 1& 0& 1\\ 1& 1& 1& 0\end{array}\right).`$
Conversely the result means that the matrix $``$ can be factorized into a form
$`={\displaystyle \frac{1}{5}}S^t\mathrm{\Sigma }S,`$
with a triangular matrix $`S`$ with non-vanishing components in the right lower entries. Generally we know the forms of the matrices $`V`$ and $`P`$ in Eqs.(4),(5), but the remaining ones $`m`$ and $`𝒩`$ are unknown. If we can determine the $`𝒩`$, then we obtain the $`m`$ from a relation
$`m=𝒩(PV)^1.`$ (9)
This $`𝒩`$ may contain invariants of K(M)-theory of the D-branes as its components. The $`𝒩`$ is a triangular matrix with non-vanishing components in the left lower blocks
$`𝒩=\left(\begin{array}{cccc}& 0& 0& 0\\ & & 0& 0\\ & & & 0\\ & & & \end{array}\right).`$
We multiply a charge vector $`Q`$ on the $`𝒩`$ from the left. The $`Q`$ encodes information of homology cycles of the Calabi-Yau around which D-branes wrap. For the quintic case, numbers in the last row of the $`𝒩`$ in Eq.(8) coincide with coefficients of a topological invariant $`(5)\sqrt{\widehat{A}(\rho )}`$
$`(5)\sqrt{\widehat{A}(\rho )}=5{\displaystyle \frac{25}{12}}\rho ^2.`$
In particular, the $`25/12`$ is related with a $`2`$nd Chern number $`c_2=50`$ of $`M`$
$`{\displaystyle \frac{1}{24}}c_2={\displaystyle \frac{25}{12}}.`$
It also appears in the $`1`$st column in the $`𝒩`$. The number $`5`$ is interpreted as a triple intersection number of a $`4`$-cycle in the $`M`$. The remaining entry $`11/2`$ is expected to be interpreted as some kind of invariant associated with a normal bundle of a world volume of a D-brane. But we do not precisely know the geometric characterization of this yet and cannot explain this number from the point of view of characteristic classes of some associated bundle.
We will return to the central charge $`Z`$. If we can find the matrix $`𝒩`$ and choose the basis $`\mathrm{\Pi }`$, the central charge is evaluated in the B-model in the open string channel
$`Z=Q\mathrm{\Pi }=(Q𝒩)(𝒩^1\mathrm{\Pi })=(Q𝒩)\mathrm{\Omega },`$
$`Q=\left(\begin{array}{ccccc}Q_{2d}& Q_{2d2}& \mathrm{}& Q_2& Q_0\end{array}\right).`$
When one uses a basis $`\mathrm{\Omega }_{\mathrm{}}`$ to construct the $`Z`$, the formula of a modified charge vector $`Q𝒩`$ is needed. But no systematic method to calculate the $`𝒩`$ is known yet. Also the basis $`\mathrm{\Pi }`$ has an ambiguity in the multiplication with a matrix $`L`$ from the left-side
$`\mathrm{\Pi }^{}=L\mathrm{\Pi }=L𝒩\mathrm{\Omega }.`$
When the $`L`$ is a symplectic (or an SO-invariant) matrix with integer components
$`L^t\mathrm{\Sigma }L=\mathrm{\Sigma },`$
the modified $`\mathrm{\Pi }^{}`$ gives us the same Kähler potential as that in the $`\mathrm{\Pi }`$ basis. It leads to the same results about properties of closed string moduli spaces. We do not know any principle to fix the ambiguity completely and do not touch on this here. In this paper, we choose a standard convention $`L=I`$ by choosing a pure D<sub>2d</sub> charge in the next section.
## 4 $`5`$-Fold
In the previous section, we analyzed the quintic. We started analyses by using the prepotential $`F`$ of the quintic and studied its monodromy properties. We cannot expect to use analogs of prepotentials for other $`d`$-fold cases $`(d>3)`$. But the essential part we learned in the previous section seems to be a factorizable property of the matrix $``$ by a (triangular) matrix $`S`$ and a matrix $`\mathrm{\Sigma }`$. Under this consideration, we will investigate the $`N=7`$ ($`d=5`$) case concretely. First the $``$ in the $`\widehat{\mathrm{\Pi }}`$ basis is defined as
$`=\left(\begin{array}{cccccc}0& 0& 0& 0& 0& 1\\ 0& 0& 0& 0& 1& 2\\ 0& 0& 0& 1& 1& 1\\ 0& 0& 1& 0& 2& 2\\ 0& 1& 1& 2& 0& 3\\ 1& 2& 1& 2& 3& 0\end{array}\right).`$
We find that the $``$ is factorized into a form with a matrix $`\mathrm{\Sigma }`$ and a triangular matrix $`S`$
$`7=S^t\mathrm{\Sigma }S,`$
$`\mathrm{\Sigma }=\left(\begin{array}{cccccc}0& 0& 0& 0& 0& 1\\ 0& 0& 0& 0& 1& 0\\ 0& 0& 0& 1& 0& 0\\ 0& 0& 1& 0& 0& 0\\ 0& 1& 0& 0& 0& 0\\ 1& 0& 0& 0& 0& 0\end{array}\right),S=\left(\begin{array}{cccccc}0& 0& 0& 0& 0& 1\\ 0& 0& 0& 0& 1& 2\\ 0& 0& 0& 1& 1& 1\\ 0& 0& 7& 7& 7& 7\\ 0& 7& 0& 14& 21& 7\\ 7& 0& 0& 0& 0& 0\end{array}\right).`$
This decomposition is not unique as we discussed at the end of the previous section. There are some possibilities for choosing the $`S`$. When we pick a cyclic basis $`\{\varpi _j\}`$ with an index “$`j`$” as
$`\varpi ={}_{}{}^{t}\left(\begin{array}{cccccc}\varpi _0& \varpi _1& \varpi _2& \varpi _3& \varpi _5& \varpi _6\end{array}\right),`$
a transformation matrix $`P`$ from the $`\widehat{\mathrm{\Pi }}`$ to the $`\varpi `$ $`(\widehat{\mathrm{\Pi }}=P\varpi )`$ is obtained
$`P=\left(\begin{array}{cccccc}0& 0& 0& 0& 0& 1\\ 7& 0& 0& 0& 0& 1\\ 35& 7& 0& 0& 0& 1\\ 70& 28& 7& 0& 0& 1\\ 35& 28& 21& 14& 7& 1\\ 7& 7& 7& 7& 7& 6\end{array}\right).`$
Now we know the matrix $`S`$, we can determine matrices $`m`$ and $`𝒩`$. By using the relation $`S=mP`$, we can obtain a transformation matrix $`m`$ from the $`\varpi `$ basis to the $`\mathrm{\Pi }`$ basis
$`\mathrm{\Pi }={}_{}{}^{t}\left(\begin{array}{cccccc}\mathrm{\Pi }_0& \mathrm{\Pi }_1& \mathrm{\Pi }_2& \mathrm{\Pi }_3& \mathrm{\Pi }_4& \mathrm{\Pi }_5\end{array}\right),`$
$`m=\left(\begin{array}{cccccc}1& 0& 0& 0& 0& 0\\ \frac{3}{7}& \frac{3}{7}& \frac{2}{7}& \frac{1}{7}& \frac{1}{7}& \frac{2}{7}\\ \frac{3}{7}& \frac{6}{7}& \frac{3}{7}& \frac{1}{7}& 0& \frac{1}{7}\\ 4& 4& 1& 0& 0& 1\\ 8& 4& 3& 1& 1& 4\\ 1& 1& 0& 0& 0& 0\end{array}\right).`$ (10)
Characteristic features of the $`m`$ appear in the $`1`$st and the last rows in $`m`$. The first row means that the $`\mathrm{\Pi }_0`$ is identified with the $`\varpi _0`$. The last row implies that the $`\mathrm{\Pi }_5`$ is represented as a linear combination of only $`\varpi _0`$ and $`\varpi _1`$ as
$`\mathrm{\Pi }_5=\varpi _0\varpi _1.`$
We believe that these structures are universal. Especially the $`\mathrm{\Pi }_d`$ might be represented as
$`\mathrm{\Pi }_d=\varpi _0\varpi _1.`$
It is related to the structure of a pure D<sub>2d</sub>-brane charge $`Q_{2d}`$.
Also we find the matrix $`𝒩`$ for the $`N=7`$ case by using an equation Eq.(9)
$`\mathrm{\Pi }=𝒩\mathrm{\Omega },`$
$`𝒩=\left(\begin{array}{cccccc}1& 0& 0& 0& 0& 0\\ 0& 1& 0& 0& 0& 0\\ \frac{7}{8}& \frac{1}{2}& 1& 0& 0& 0\\ 0& \frac{161}{24}& 0& 7& 0& 0\\ \frac{11711}{1920}& \frac{161}{48}& \frac{161}{24}& \frac{7}{2}& 7& 0\\ 0& \frac{147}{640}& 0& \frac{49}{8}& 0& 7\end{array}\right),`$
$`𝒩^t\mathrm{\Sigma }𝒩=(7)\mathrm{\Sigma }.`$
The entries in the last row of the $`𝒩`$ are topological numbers. They coincide with coefficients of the $`(7)\sqrt{\widehat{A}(\rho )}`$ for the $`N=7`$ case
$`(7)\sqrt{\widehat{A}(\rho )}=7{\displaystyle \frac{49}{8}}\rho ^2+{\displaystyle \frac{147}{640}}\rho ^4.`$
The entries $`\pm 7`$ is associated with an intersection number $`7`$ of five $`8`$-cycles in $`M`$. Possibly the other entries could be interpreted as invariants associated with D-branes from the point of view of K(M)-theory. We cannot precisely know any geometric characterization of numbers at the other entries yet. We do not touch on this topic here.
Next let us consider monodromy matrices associated with singular points in order to confirm our result for the $`N=7`$ case. The $`5`$-fold has singular points $`\psi =0,\mathrm{},e^{2\pi i\mathrm{}/7}`$ ($`\mathrm{}=0,1,2\mathrm{},6`$) in the B-model moduli space. At the point $`\psi =0`$, the monodromy transformation acts on the basis $`\varpi `$ cyclically and is realized as a matrix $`A_\varpi `$
$`𝒜\varpi =A_\varpi \varpi .`$
The action of $`𝒜`$ on other bases $`\widehat{\mathrm{\Pi }}`$, $`\mathrm{\Pi }`$ and $`\mathrm{\Omega }`$ is calculated as representation matrices $`A_{\widehat{\mathrm{\Pi }}}`$, $`A_\mathrm{\Pi }`$ and $`A_\mathrm{\Omega }`$ by using the $`𝒩`$ and $`m`$
$`A_\varpi =\left(\begin{array}{cccccc}0& 1& 0& 0& 0& 0\\ 0& 0& 1& 0& 0& 0\\ 0& 0& 0& 1& 0& 0\\ 1& 1& 1& 1& 1& 1\\ 0& 0& 0& 0& 0& 1\\ 1& 0& 0& 0& 0& 0\end{array}\right),A_{\widehat{\mathrm{\Pi }}}=\left(\begin{array}{cccccc}6& 1& 0& 0& 0& 0\\ 21& 1& 1& 0& 0& 0\\ 35& 0& 1& 1& 0& 0\\ 35& 0& 0& 1& 1& 0\\ 21& 0& 0& 0& 1& 1\\ 7& 0& 0& 0& 0& 1\end{array}\right),`$
$`A_\mathrm{\Pi }=\left(\begin{array}{cccccc}1& 0& 0& 0& 0& 1\\ 1& 1& 0& 0& 0& 1\\ 1& 1& 1& 0& 0& 1\\ 14& 0& 7& 1& 0& 14\\ 7& 14& 7& 1& 1& 7\\ 7& 7& 14& 0& 1& 6\end{array}\right),A_\mathrm{\Omega }=\left(\begin{array}{cccccc}1& \frac{147}{640}& 0& \frac{49}{8}& 0& 7\\ 1& \frac{787}{640}& 0& \frac{49}{8}& 0& 7\\ \frac{1}{2}& \frac{6737}{5120}& 1& \frac{539}{64}& 0& \frac{77}{8}\\ \frac{1}{6}& \frac{757}{1024}& 1& \frac{1033}{192}& 0& \frac{175}{24}\\ \frac{1}{24}& \frac{330779}{1228800}& \frac{1}{2}& \frac{26633}{15360}& 1& \frac{5999}{1920}\\ \frac{1}{120}& \frac{85451}{1228800}& \frac{1}{6}& \frac{3737}{15360}& 1& \frac{289}{1920}\end{array}\right).`$
Also the transformation around the $`\psi =\mathrm{}`$ point is evaluated on these bases as matrices $`T_\varpi `$, $`T_{\widehat{\mathrm{\Pi }}}`$, $`T_\mathrm{\Pi }`$ and $`T_\mathrm{\Omega }`$
$`T_\varpi =\left(\begin{array}{cccccc}1& 0& 0& 0& 0& 0\\ 2& 0& 0& 0& 0& 1\\ 6& 1& 0& 0& 0& 6\\ 15& 0& 1& 0& 0& 15\\ 14& 1& 1& 1& 1& 16\\ 6& 0& 0& 0& 1& 6\end{array}\right),T_{\widehat{\mathrm{\Pi }}}=\left(\begin{array}{cccccc}1& 1& 1& 1& 1& 1\\ 0& 1& 1& 1& 1& 1\\ 0& 0& 1& 1& 1& 1\\ 0& 0& 0& 1& 1& 1\\ 0& 0& 0& 0& 1& 1\\ 0& 0& 0& 0& 0& 1\end{array}\right),`$
$`T_\mathrm{\Pi }=\left(\begin{array}{cccccc}1& 0& 0& 0& 0& 0\\ 1& 1& 0& 0& 0& 0\\ 0& 1& 1& 0& 0& 0\\ 14& 7& 7& 1& 0& 0\\ 7& 14& 0& 1& 1& 0\\ 7& 7& 14& 1& 1& 1\end{array}\right),T_\mathrm{\Omega }=\left(\begin{array}{cccccc}1& 0& 0& 0& 0& 0\\ 1& 1& 0& 0& 0& 0\\ \frac{1}{2}& 1& 1& 0& 0& 0\\ \frac{1}{6}& \frac{1}{2}& 1& 1& 0& 0\\ \frac{1}{24}& \frac{1}{6}& \frac{1}{2}& 1& 1& 0\\ \frac{1}{120}& \frac{1}{24}& \frac{1}{6}& \frac{1}{2}& 1& 1\end{array}\right).`$
The forms of the matrices $`T_{\widehat{\mathrm{\Pi }}}`$ and $`T_\mathrm{\Omega }`$ are universal. We will explain these points in the section 7.
The $`\psi =1`$ is a conifold-like point and associated monodromy matrices are obtained for these bases as $`P_\varpi `$, $`P_\mathrm{\Pi }`$, $`P_{\widehat{\mathrm{\Pi }}}`$ and $`P_\mathrm{\Omega }`$
$`P_\varpi =\left(\begin{array}{cccccc}2& 1& 0& 0& 0& 0\\ 1& 0& 0& 0& 0& 0\\ 6& 6& 1& 0& 0& 0\\ 15& 15& 0& 1& 0& 0\\ 15& 15& 0& 0& 1& 0\\ 6& 6& 0& 0& 0& 1\end{array}\right),P_{\widehat{\mathrm{\Pi }}}=\left(\begin{array}{cccccc}1& 0& 0& 0& 0& 0\\ 7& 1& 0& 0& 0& 0\\ 14& 0& 1& 0& 0& 0\\ 21& 0& 0& 1& 0& 0\\ 14& 0& 0& 0& 1& 0\\ 7& 0& 0& 0& 0& 1\end{array}\right),`$
$`P_\mathrm{\Pi }=\left(\begin{array}{cccccc}1& 0& 0& 0& 0& 1\\ 0& 1& 0& 0& 0& 0\\ 0& 0& 1& 0& 0& 0\\ 0& 0& 0& 1& 0& 0\\ 0& 0& 0& 0& 1& 0\\ 0& 0& 0& 0& 0& 1\end{array}\right),P_\mathrm{\Omega }=\left(\begin{array}{cccccc}1& \frac{147}{640}& 0& \frac{49}{8}& 0& 7\\ 0& 1& 0& 0& 0& 0\\ 0& \frac{1029}{5120}& 1& \frac{343}{64}& 0& \frac{49}{8}\\ 0& 0& 0& 1& 0& 0\\ 0& \frac{3087}{409600}& 0& \frac{1029}{5120}& 1& \frac{147}{640}\\ 0& 0& 0& 0& 0& 1\end{array}\right).`$
Under these monodromy transformations, the set $`\mathrm{\Omega }_{\mathrm{}}`$ is not an integral basis because monodromy matrices have fractional rational numbers in their entries. In contrast, the sets $`\mathrm{\Pi }_{\mathrm{}}`$ and $`\widehat{\mathrm{\Pi }}_{\mathrm{}}`$ are integral bases under the monodromy transformations.
For the $`\widehat{\widehat{\mathrm{\Pi }}}`$ basis, the associated monodromy matrices $`A_{\widehat{\widehat{\mathrm{\Pi }}}}`$, $`P_{\widehat{\widehat{\mathrm{\Pi }}}}`$, $`T_{\widehat{\widehat{\mathrm{\Pi }}}}`$ are evaluated as
$`A_{\widehat{\widehat{\mathrm{\Pi }}}}=\left(\begin{array}{cccccc}7& 1& 0& 0& 0& 0\\ 28& 0& 1& 0& 0& 0\\ 84& 0& 0& 1& 0& 0\\ 210& 0& 0& 0& 1& 0\\ 462& 0& 0& 0& 0& 1\\ 925& 6& 15& 20& 15& 6\end{array}\right),P_{\widehat{\widehat{\mathrm{\Pi }}}}=\left(\begin{array}{cccccc}1& 0& 0& 0& 0& 0\\ 7& 1& 0& 0& 0& 0\\ 28& 0& 1& 0& 0& 0\\ 84& 0& 0& 1& 0& 0\\ 210& 0& 0& 0& 1& 0\\ 462& 0& 0& 0& 0& 1\end{array}\right),`$
$`T_{\widehat{\widehat{\mathrm{\Pi }}}}=\left(\begin{array}{cccccc}6& 15& 20& 15& 6& 1\\ 1& 0& 0& 0& 0& 0\\ 0& 1& 0& 0& 0& 0\\ 0& 0& 1& 0& 0& 0\\ 0& 0& 0& 1& 0& 0\\ 0& 0& 0& 0& 1& 0\end{array}\right).`$
We can study the other monodromy transformations around points $`\psi =e^{2\pi i\mathrm{}/7}`$ ($`\mathrm{}=1,2,\mathrm{},6`$) by combining the results for the $`\psi =1`$ and those for $`\psi =0`$. One can see that the monodromy matrices $`A_\mathrm{\Pi }`$, $`T_\mathrm{\Pi }`$ and $`P_\mathrm{\Pi }`$ in the canonical basis have integral entries. That is a necessary condition for the $`\mathrm{\Pi }`$ to be a canonical integral basis because the set of the associated D-brane (integral) charge vector $`Q`$ is transformed by these monodromy transformations into vectors $`QA_\mathrm{\Pi }`$, $`QT_\mathrm{\Pi }`$, and $`QP_\mathrm{\Pi }`$. In this $`5`$-fold case, we obtain the canonical basis $`\mathrm{\Pi }=𝒩\mathrm{\Omega }`$, but the choice of the matrix $`S`$ has ambiguities. They have their origins in the decomposition of the $``$ and the determination of the matrix $`S`$. But the data at the Gepner point allows us to calculate the exact formula of the central charge $`Z`$. We will explain this in sections 5 and 7.
## 5 Gepner Model
There are several analyses in 3-dimensional Calabi-Yau cases based on the Gepner model,. The boundary states, charges, and intersection forms are discussed. Shortly we review the Gepner model and expand the results, to the $`d`$-fold case.
A two dimensional $`N=2`$ minimal unitary model has a central charge
$`c={\displaystyle \frac{3k}{k+2}}.`$
Here the $`k`$ is a positive integer and is called the level. The primary fields of this model is labelled by a set of the conformal weight $`h`$ and the U$`(1)`$ charge $`q`$ as $`(h,q)`$
$`h={\displaystyle \frac{\mathrm{}(\mathrm{}+2)m^2}{4(k+2)}}+{\displaystyle \frac{s^2}{8}}(\text{mod.}\mathrm{\hspace{0.17em}1}),`$
$`q={\displaystyle \frac{m}{k+2}}{\displaystyle \frac{s}{2}}(\text{mod.}\mathrm{\hspace{0.17em}2}).`$
They are parametrized by a set of three integers $`(\mathrm{},m,s)`$. The standard range of the $`(\mathrm{},m,s)`$ is specified
$`0\mathrm{}k,|ms|\mathrm{},`$
$`s\{0,\pm 1\}\text{and}\mathrm{}+m+s2𝐙.`$
The NS sector is associated with the $`s=0`$ representation while the $`s=\pm 1`$ representations belong to the Ramond sector. The (anti-)chiral primary states are labelled by $`((\mathrm{},\mathrm{},0))`$ ($`\mathrm{},+\mathrm{},0`$) respectively in the NS sector. They are related to the Ramond ground states $`(\mathrm{},\pm \mathrm{},\pm 1)`$ by the spectral flow.
Let us consider a Gepner model realized by tensoring $`N`$ minimal models with the same level $`(N2)`$. The corresponding Landau-Ginzburg model is realized by a potential
$`X_1^N+X_2^N+\mathrm{}+X_N^N.`$
We restrict ourselves to (complex) odd dimensional Calabi-Yau cases with $`N=3,5,7`$ ($`d=1,3,5`$). Each minimal model has a $`𝐙_N\times 𝐙_2`$ symmetry whose generators $`(g,h)`$s act on the primary field $`\mathrm{\Phi }_{m,s}^{\mathrm{}}`$ as
$`g\mathrm{\Phi }_{m,s}^{\mathrm{}}=\alpha ^m\mathrm{\Phi }_{m,s}^{\mathrm{}}(\alpha =e^{2\pi i/N}),`$
$`h\mathrm{\Phi }_{m,s}^{\mathrm{}}=(1)^s\mathrm{\Phi }_{m,s}^{\mathrm{}}.`$
The $`𝐙_N`$-symmetry is correlated with the U(1) charge. The orbifold group of the Gepner model is generated by a diagonal $`𝐙_N`$ generator $`_{j=1}^Ng_j`$. Here the index “$`j`$” distinguishes the $`N`$ minimal models. The boundary states $`|\mathrm{\Xi }`$ which preserve $`N=2`$ worldsheet algebra are constructed in . According to the notations of Cardy, they are labelled by a set of integers
$`\mathrm{\Xi }=(L_j,M_j,S_j)(j=1,2,\mathrm{},N),`$
for each A-, B-type boundary condition. The action of the discrete symmetry $`𝐙_N\times 𝐙_2`$ is expressed on the state labelled by $`\mathrm{\Xi }`$ as
$`\{\begin{array}{cccc}M_j& & M_j+2& (𝐙_N\text{action})\\ S_j& & S_j+2& (𝐙_2\text{action})\end{array}.`$
Also it is known that a physically inequivalent choice for the $`S_j`$ together with the identification under the $`𝐙_2`$-action is given as
$`S={\displaystyle \underset{j=1}{\overset{N}{}}}S_j0\text{mod.}\mathrm{\hspace{0.17em}4}.`$
That is to say, it is enough to consider only boundary states with $`S=0`$. Also for B-type boundary states, it is known that the physically inequivalent choices of the $`M_j`$ ($`j=1,2,\mathrm{},N`$) can be described by a sum of the $`M_j`$s
$`M={\displaystyle \underset{j=1}{\overset{N}{}}}M_j.`$
It means that the B-type boundary states with fixed $`L=(L_1,L_2,\mathrm{},L_N)=:\{L_j\}`$ are described by the single integer $`M`$. We calculate an intersection matrix $`I_B`$ for an $`L=(0,0,\mathrm{},0)=:\{0\}`$ state in the B-type boundary condition with $`g^N=1`$
$`I_B=(1g^1)^N={\displaystyle \underset{\mathrm{}=0}{\overset{N}{}}}\left(\begin{array}{c}N\\ \mathrm{}\end{array}\right)(1)^{\mathrm{}}g^{\mathrm{}}.`$
Then an intersection form $`I_G`$ in the Gepner basis is related to the $`I_B`$ for general $`N=3,4,5,\mathrm{}`$
$`I_B=(1g)(1g^1)I_G,I_G=(g^1)(1g^1)^{N2}.`$
The $`I_G`$ coincides with an intersection matrix $`m^1\mathrm{\Sigma }(m^1)^t`$ in the sigma model when we choose the arrangement of the $`\varpi _j`$s appropriately
$`\varpi ={}_{}{}^{t}(\begin{array}{ccccccccc}\varpi _{a+1}& \varpi _{a+2}& \mathrm{}& \varpi _{N1}& \varpi _0& \varpi _1& \varpi _2& \mathrm{}& \varpi _{a1}\end{array}).`$
Let us recall that the basis $`\mathrm{\Pi }`$ in the large volume is related to the $`\varpi `$ through the $`m`$
$`\mathrm{\Pi }=m\varpi .`$
The charge vector $`Q_G`$ is related to the large volume charge vector $`Q_L`$ <sup>3</sup><sup>3</sup>3In this section and section 7, we put the super-, subscript “$`L`$” to the charge $`Q`$ in the large volume case to distinguish it from the $`Q_G`$ in the Gepner basis. as
$`Z=Q_L\mathrm{\Pi }=Q_G\varpi ,`$
$`Q_L=Q_Gm^1.`$
At a point $`\psi =1`$ in the moduli space of the $`W`$, a $`d`$-cycle in the mirror shrinks into a point. An associated cycle in the M is a pure $`2d`$-cycle around which a D$`2d`$-brane wraps. The associated charge is specified by a charge vector $`Q_L`$ in the large volume limit
$`Q_L=(\begin{array}{ccccc}Q_{2d}& Q_{2d2}& \mathrm{}& Q_2& Q_0\end{array})=(\begin{array}{cccccc}1& 0& 0& \mathrm{}& 0& 0\end{array})=:Q_L^{(0)}.`$
When we use a Gepner basis for $`\varpi `$ as
$`\varpi ={}_{}{}^{t}(\begin{array}{cccccccc}\varpi _0& \varpi _1& \varpi _2& \mathrm{}& \varpi _{a1}& \varpi _{a+1}& \mathrm{}& \varpi _{N1}\end{array}),`$
with $`a=2,3,4`$ for respectively $`N=3,5,7`$ cases, we can obtain a charge vector $`Q_G^{(0)}`$ in the Gepner basis corresponding to the $`Q_L^{(0)}`$
$`Q_G^{(0)}=\left(\begin{array}{ccccccc}1& 1& 0& 0& \mathrm{}& 0& 0\end{array}\right).`$
When we take a charge $`Q_G`$ of a B-type boundary state with $`|\{0\};0;0`$ to be the $`Q_G^{(0)}`$, this boundary state is identified with a pure D$`2d`$-brane with the charge $`Q_L^{(0)}`$. By acting a $`𝐙_N`$-monodromy matrix $`A^1`$, we can change the value $`M`$ into an $`M+2`$ and obtain a charge for the state $`|\{0\};M;0`$. Also the $`𝐙_2`$-action $`h`$ is implemented by reversing a sign of the charge $`QQ`$. The other charges $`Q_G`$ of states with $`L=(\begin{array}{cccc}L_1& L_2& \mathrm{}& L_N\end{array})`$ can be obtained by acting generators $`g_j`$s of the $`𝐙_N`$-symmetries on the $`Q_G^{(0)}`$
$`Q_G=Q_G^{(0)}{\displaystyle \underset{j=1}{\overset{N}{}}}\left({\displaystyle \underset{\mathrm{}_j=L_j/2}{\overset{L_j/2}{}}}g^\mathrm{}_j\right).`$ (11)
In our case, the $`N`$ is odd and there is a useful relation in calculating the $`Q_G`$ from the set of numbers $`\{L_j\}`$
$`g^{1/2}=g^{\frac{N1}{2}}.`$
Collecting all the relations, we calculate the charges for the $`5`$-fold case. We list the result for the $`N=7`$ ($`d=5`$) case in the table 1.
Here we use a basis $`\varpi `$
$`\varpi ={}_{}{}^{t}(\begin{array}{cccccc}\varpi _0& \varpi _1& \varpi _2& \varpi _3& \varpi _5& \varpi _6\end{array}).`$
for measuring the charges in the $`Q_G`$ and write down results for the $`L_j=0,1`$ ($`j=1,2,\mathrm{},7`$) cases for simplicity. We make a remark here: We used the $`m`$ in Eq.(10). But there is some arbitrariness in determining the $`m`$ and the vector $`Q_L`$ cannot be fixed uniquely. Probably the arbitrariness is reduced to some symplectic or SO transformation and they might possibly lead to equivalent physics. We do not know any clear explanations about this yet.
In contrast, the result here shows that we can calculate a charge vector $`Q_G`$ of the D-brane in the Gepner basis associated with a boundary state $`|\{L\};M;S`$
$`LQ_G,`$
when a set of numbers $`L=(\begin{array}{cccc}L_1& L_2& \mathrm{}& L_N\end{array})`$ is given as an input datum. So we can investigate relations between the $`L`$ and $`Q_G`$ more precisely. First we restrict ourselves to the states $`|\{L\};M=0;S=0`$ for simplicity. Then we obtain the charge vector $`Q_G`$ for $`N=3,4,5`$ ($`d=1,3,5`$) cases with respectively $`a=2,3,4`$
$`\varpi =(\begin{array}{cccccccc}\varpi _0& \varpi _1& \varpi _2& \mathrm{}& \varpi _{a1}& \varpi _{a+1}& \mathrm{}& \varpi _{N1}\end{array}),`$
$`Q_G:=(\begin{array}{cccccccc}Q_0^G& Q_1^G& Q_2^G& \mathrm{}& Q_{a1}^G& Q_{a+1}^G& \mathrm{}& Q_{N1}^G\end{array}),`$
$`Q_j^G={\displaystyle \frac{1}{N}}{\displaystyle \underset{k=1}{\overset{N1}{}}}\alpha ^{kj}(\alpha ^k1)^{N+1}\times \alpha ^{\frac{k}{2}_{j^{}=1}^NL_j^{}}\times {\displaystyle \underset{j^{\prime \prime }=1}{\overset{N}{}}}\left\{\alpha ^{k(L_{j^{\prime \prime }}+1)}1\right\},`$
$`(j=0,1,2,\mathrm{},a1,a+1,\mathrm{},N1).`$
It is this formula that connects the set $`\{L_j\}`$ and the charge vector $`Q_G`$. Let us recall that the D-brane charge vector $`Q_L`$ in the large radius volume with the B-type boundary condition is constructed from the $`Q_G`$ by the transformation matrix $`m`$ as $`Q_L=Q_Gm^1`$. The associated central charge (the BPS mass formula) $`Z`$ is expressed by the $`\mathrm{\Pi }_{\mathrm{}}`$ or $`\varpi `$ in the appropriate parameter regions of the moduli space
$`Z={\displaystyle \underset{\mathrm{}=0}{\overset{d}{}}}Q_2\mathrm{}^L\mathrm{\Pi }_{\mathrm{}}=Q_L\mathrm{\Pi }`$
$`=(Q_Lm)(m^1\mathrm{\Pi })`$
$`=Q_G\varpi ={\displaystyle \underset{\stackrel{0jN1}{ja}}{}}Q_j^G\varpi _j.`$
When the $`\psi `$ is not large (for $`|\psi |<1`$), the formula of the $`Z`$ is expressed in the $`\varpi _j`$ or $`\stackrel{~}{\varpi }_k`$ basis
$`Z={\displaystyle \frac{(1)^{d+1}}{N}}{\displaystyle \underset{j=0,1,\mathrm{},N1}{}}\varpi _j{\displaystyle \underset{k=1}{\overset{N1}{}}}\alpha ^{k(j+\frac{1}{2}_{j^{}=1}^NL_j^{})}\times (\alpha ^k1)^{N+1}\times {\displaystyle \underset{j^{\prime \prime }=1}{\overset{N}{}}}\left\{\alpha ^{k(L_{j^{\prime \prime }}+1)}1\right\}`$
$`={\displaystyle \frac{(1)^N(2i)^N}{(2\pi i)^{N1}}}\times {\displaystyle \underset{k=1}{\overset{N1}{}}}\stackrel{~}{\varpi }_k(1)^k\times {\displaystyle \underset{j=1}{\overset{N}{}}}\left\{\mathrm{sin}{\displaystyle \frac{\pi k(L_j+1)}{N}}\right\}.`$ (12)
Next the state $`|\{L\};M=2\mathrm{};0`$ $`(\mathrm{}0)`$ is obtained by acting the $`𝐙_N`$ monodromy matrix $`A^1`$ on the Z in the $`|\{L\};M=0;0`$ case
$`Z=Q(A_\mathrm{\Pi }^1)^{\mathrm{}}\mathrm{\Pi }.`$
The $`S0`$ case can be realized by acting the remaining $`𝐙_2`$ symmetry on the charge $`QQ`$.
## 6 Cycles
In this section, we study cycles associated with the periods $`\varpi _j`$ and their intersection forms. The cycles are susy $`d`$ cycles in the $`d`$ fold consistent with the A-type boundary condition.
### 6.1 Cycles
For the $`d`$-fold variety $`W`$
$`W;\widehat{\{p=0\}/𝐙_N^{(N1)}},`$
$`p=X_1^N+X_2^N+\mathrm{}+X_N^NN\psi X_1X_2\mathrm{}X_N=0,`$
we consider a set of $`d`$ chains $`V_j(\psi )`$ $`(j=0,1,2,\mathrm{},N1)`$
$`V_j(\psi )=\left\{\begin{array}{c}X_1,X_2,\mathrm{},X_{N1};\text{real positive}\hfill \\ X_N=1,\hfill \\ X_{N1}\text{s.t.}\text{arg}X_{N1}\pi {\displaystyle \frac{2j+1}{N}}\text{as}\psi 0\hfill \end{array}\right\}.`$ (16)
In this definition, the index “$`j`$” of the $`V_j`$ is defined modulo $`N`$, that is, $`V_{N+j}=V_j`$ for an arbitrary $`j𝐙`$. These are chains with the same boundary under an identification of the discrete symmetry $`𝐙_N^{(N1)}`$. But an arbitrary combination $`V_iV_j`$ ($`i,j=0,1,2,\mathrm{},N1`$) is a $`d`$-cycle. These cycles are related with the $`V_k`$s through an intersection matrix $`I_{j,k}`$
$`𝒞_j={\displaystyle \underset{k=0}{\overset{N1}{}}}I_{j,k}V_k,I_{j,k}={\displaystyle \frac{1}{N}}{\displaystyle \underset{\mathrm{}=0}{\overset{N1}{}}}\alpha ^{\mathrm{}(jk)}(1\alpha ^{\mathrm{}})^{N2}.`$ (17)
More explicitly, this expression is evaluated as a linear combination of $`d`$-cycles
$`𝒞_j={\displaystyle \underset{n=0}{\overset{N2}{}}}(1)^n\left(\begin{array}{c}N2\\ n\end{array}\right)V_{j+2}(j;\text{mod.}N)`$
$`={\displaystyle \underset{n=0}{\overset{N3}{}}}(1)^n\left(\begin{array}{c}N3\\ n\end{array}\right)(V_{j+n}V_{j+n+1})(j;\text{mod.}N).`$
Let us consider the $`N=`$odd cases for $`\psi =0`$. In the $`N=`$odd case, a phase of $`X_{N1}`$ in the $`V_j`$ tends to a definite value for $`\psi 0`$
$`\text{arg}X_{N1}\pi +{\displaystyle \frac{2\pi }{N}}\left(j{\displaystyle \frac{N1}{2}}\right)(N=\text{odd}).`$
The $`d`$ chain $`V_j`$ for $`N=`$odd is represented in this limit as
$`\text{Im}X_{\mathrm{}}=0(\mathrm{}=1,2,\mathrm{},N2,N),`$
$`\text{Im}\left(\alpha ^{\frac{N1}{2}j}X_{N1}\right)=0.`$
$`{\displaystyle \frac{N1}{2}}j𝐙,`$
At this special moduli point, there are several analyses about the cycles in . The associated susy $`d`$ chain for the orbifold point is obtained in as
$`\text{Im}\left(\alpha ^m_{\mathrm{}}X_{\mathrm{}}\right)=0,m_{\mathrm{}}𝐙(\mathrm{}=1,2,\mathrm{},N).`$ (18)
An arbitrary susy cycle is constructed as a difference of any two susy chains (18) at this moduli point. In our case, the $`V_j`$ evaluated at $`\psi =0`$ corresponds to a special kind of susy chains labelled by a set of numbers $`\{m_{\mathrm{}}\}`$ at the orbifold point
$`m_{\mathrm{}}=\left({\displaystyle \frac{N1}{2}}j\right)\delta _{\mathrm{},N1}.`$
But we obtain a formula of susy cycles in the $`d`$-fold $`W`$ at a specified moduli point $`\psi `$.
### 6.2 Periods
Next we introduce a set of periods $`q_j(\psi )`$ associated with the $`𝒞_j`$. These $`q_j(\psi )`$s are related with our periods $`\stackrel{~}{\varpi }_k(\psi )`$ as
$`q_j(\psi )={\displaystyle \frac{1}{2\pi iN^{N1}}}{\displaystyle \underset{k=1}{\overset{N1}{}}}\alpha ^{jk}(\alpha ^k1)^{N1}\stackrel{~}{\varpi }_k(\psi )(j=0,1,2,\mathrm{},N1).`$ (19)
In order to prove the Eq.(19), we introduce a set of one-cycles $`\gamma _j`$ as unions of half-lines
$`\gamma _j=\{\text{arg}X_j=0\}\{\text{arg}X_j={\displaystyle \frac{2\pi }{N}}\}(j=1,2,\mathrm{},N2).`$
From these $`d`$ one-cycles, we can construct $`d`$-cycles $`𝒞_j`$
$`𝒞_j=\{(X_1,X_2,\mathrm{},X_{N2},X_{N1},1);`$
$`(X_1,X_2,\mathrm{},X_{N2})\gamma _1\times \gamma _2\times \mathrm{}\times \gamma _{N2},`$
$`\text{branch of}X_{N1}\text{is specified by “}j\text{}\}(j=0,1,2,\mathrm{},N2).`$
We calculate a period $`q_0(\psi )`$ associated with the $`𝒞_0`$
$`q_0(\psi )={\displaystyle \underset{m=1}{\overset{\mathrm{}}{}}}u_m(N\psi )^m𝐈_m,`$
$`u_m={\displaystyle \frac{1}{N}}\alpha ^{m/2}{\displaystyle \frac{\mathrm{\Gamma }\left({\displaystyle \frac{N1}{N}}m\right)}{\mathrm{\Gamma }\left(1{\displaystyle \frac{m}{N}}\right)\mathrm{\Gamma }\left(m\right)}},`$
$`𝐈_m={\displaystyle _{\gamma _1\times \gamma _2\times \mathrm{}\times \gamma _{N2}}}𝑑X_1𝑑X_2\mathrm{}𝑑X_{N2}{\displaystyle \frac{(X_1X_2\mathrm{}X_{N2})^{m1}}{\mathrm{\Delta }^{\frac{N1}{N}m}}},`$
$`\mathrm{\Delta }:=1+X_1^N+X_2^N+\mathrm{}+X_{N2}^N.`$
This $`𝐈_m`$ is transformed into an integral associated with the $`V_0`$
$`𝐈_m=(1\alpha ^m)^{N2}{\displaystyle _0^{\mathrm{}}}𝑑X_1{\displaystyle _0^{\mathrm{}}}𝑑X_2\mathrm{}{\displaystyle _0^{\mathrm{}}}𝑑X_{N2}{\displaystyle \frac{(X_1X_2\mathrm{}X_{N2})^{m1}}{\mathrm{\Delta }^{\frac{N1}{N}m}}}.`$
When we recall the action of the $`𝐙_N`$ monodromy transformation $`𝒜`$ around the $`\psi =0`$
$`𝒜;(N\psi )^m\alpha ^m(N\psi )^m,`$
the $`q_0(\psi )`$ is expressed as
$`q_0(\psi )={\displaystyle \underset{m=1}{\overset{\mathrm{}}{}}}u_m(N\psi )^m{\displaystyle _{\gamma _1\times \gamma _2\times \mathrm{}\times \gamma _{N2}}}𝑑X_1𝑑X_2\mathrm{}𝑑X_{N2}{\displaystyle \frac{(X_1X_2\mathrm{}X_{N2})^{m1}}{\mathrm{\Delta }^{\frac{N1}{N}m}}}`$
$`=(1𝒜)^{N2}{\displaystyle \underset{m=1}{\overset{\mathrm{}}{}}}u_m(N\psi )^m{\displaystyle _0^{\mathrm{}}}𝑑X_1{\displaystyle _0^{\mathrm{}}}𝑑X_2\mathrm{}{\displaystyle _0^{\mathrm{}}}𝑑X_{N2}{\displaystyle \frac{(X_1X_2\mathrm{}X_{N2})^{m1}}{\mathrm{\Delta }^{\frac{N1}{N}m}}}.`$
From this formula, we obtain a relation between the $`𝒞_0`$ and the $`V_0`$
$`𝒞_0=(1𝒜)^{N2}V_0.`$
On the other hand, the $`q_0(\psi )`$ can be expressed in a series expansion explicitly
$`q_0(\psi )=\left({\displaystyle \frac{2\pi i}{N}}\right)^{N2}{\displaystyle \frac{1}{N}}(1)^{N1}`$
$`\times {\displaystyle \underset{m=1}{\overset{\mathrm{}}{}}}{\displaystyle \frac{\mathrm{\Gamma }\left({\displaystyle \frac{m}{N}}\right)}{\mathrm{\Gamma }(m)\left[\mathrm{\Gamma }\left(1{\displaystyle \frac{m}{N}}\right)\right]^{N1}}}\times (\alpha ^{\frac{N1}{2}}N\psi )^m.`$
It is related with a period $`\varpi _0(\psi )`$ as
$`q_0(\psi )=(1)^N\left({\displaystyle \frac{2\pi i}{N}}\right)^{N2}\varpi _0(\psi ),`$
$`\varpi _0(\psi )={\displaystyle \frac{1}{N}}{\displaystyle \underset{m=1}{\overset{\mathrm{}}{}}}{\displaystyle \frac{\mathrm{\Gamma }\left({\displaystyle \frac{m}{N}}\right)}{\mathrm{\Gamma }(m)\left[\mathrm{\Gamma }\left(1{\displaystyle \frac{m}{N}}\right)\right]^{N1}}}\times (\alpha ^{\frac{N1}{2}}N\psi )^m.`$
Also a relation $`q_j(\psi )=q_0(\alpha ^j\psi )`$ means that each cycle $`𝒞_j`$ is related with a period $`q_j`$. That completes the proof of the statement in Eq.(19).
We make a comment on the branch of the $`X_{N1}`$: First we introduce a variable $`v=N\psi \mathrm{\Delta }^{\frac{N1}{N}}X_1X_2\mathrm{}X_{N2}`$. In the small $`v`$ case ($`v=0`$), the $`X_{N1}`$ is evaluated as
$`X_{N1}=\mathrm{\Delta }^{1/N}\mathrm{exp}\left({\displaystyle \frac{\pi i}{N}}(2j^{}+1)\right)(j^{}=0,1,2,\mathrm{},N1).`$
The argument of $`X_{N1}`$ corresponds to the branch of the chain $`V_j^{}`$. We find an exact representation of the $`X_{N1}`$
$`\mathrm{\Delta }^{1/N}X_{N1}={\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{\alpha ^{(n+1)\left(j^{}+\frac{1}{2}\right)}}{N}}{\displaystyle \frac{\mathrm{\Gamma }\left({\displaystyle \frac{N1}{N}}(n+1)1\right)}{\mathrm{\Gamma }(n+1)\mathrm{\Gamma }\left(1{\displaystyle \frac{n+1}{N}}\right)}}v^n,`$
$`(j^{}=0,1,2,\mathrm{},N1).`$
On the other side, the $`X_{N1}`$ is evaluated in the large $`v`$ region
$`X_{N1}=\mathrm{\Delta }^{1/N}v^{\frac{1}{N1}}\mathrm{exp}\left({\displaystyle \frac{2\pi i}{N1}}j^{\prime \prime }\right)(j^{\prime \prime }=0,1,2,\mathrm{},N1).`$
When we move the parameter $`\psi `$ from $`0`$ to $`\mathrm{}`$ on the real line, two of the ($`N1`$) $`X_{N1}`$s with $`j=0`$ and $`N1`$ coincide at $`v=N(1N)^{\frac{1}{N}1}`$. Its value is evaluated as $`X_{N1}=\mathrm{\Delta }^{1/N}(N1)^{1/N}`$. The other $`X_{N1}`$s do not collide one another for the $`\psi 𝐑_+`$.
Next we will consider cycles associated with the $`\stackrel{~}{\varpi }_{\mathrm{}}`$. Because these $`q_j(\psi )`$ are related with our periods $`\stackrel{~}{\varpi }_k(\psi )`$ in Eq.(19), we can obtain cycles $`\stackrel{~}{C}_{\mathrm{}}`$ associated with the periods $`\stackrel{~}{\varpi }_{\mathrm{}}`$ as
$`\stackrel{~}{C}_{\mathrm{}}={\displaystyle \frac{1}{(1\alpha ^{\mathrm{}})^{N1}}}{\displaystyle \underset{j=0}{\overset{N1}{}}}\alpha ^j\mathrm{}𝒞_j`$
$`={\displaystyle \frac{1}{1\alpha ^{\mathrm{}}}}{\displaystyle \underset{k=0}{\overset{N1}{}}}\alpha ^\mathrm{}kV_k(1\mathrm{}N1).`$
These cycles $`\stackrel{~}{C}_{\mathrm{}}`$s diagonalize the action of the $`𝐙_N`$ monodromy $`\psi \alpha \psi `$
$`\stackrel{~}{C}_{\mathrm{}}\alpha ^{\mathrm{}}\stackrel{~}{C}_{\mathrm{}},`$
but the sets of $`𝒞_j`$ and $`V_k`$ change cyclically
$`Q_jQ_{j+1},V_kV_{k+1}.`$
The $`\stackrel{~}{C}_{\mathrm{}}`$s belong to H$`{}_{d}{}^{}(W;𝐂)`$. On the other hand, the $`𝒞_j`$s and $`V_k`$s are rational homology cycles in H$`{}_{d}{}^{}(W;𝐐)`$.
The results of the CFT at the Gepner point imply that an intersection matrix associated with the cycles $`𝒞_j`$ must be the $`I_B`$
$`𝒞_j𝒞_j^{}=(I_B)_{j,j^{}}={\displaystyle \frac{1}{N}}{\displaystyle \underset{r=0}{\overset{N1}{}}}\alpha ^{r(j+j^{}+1)}(1\alpha ^r)^{N2},(j,j^{}=0,1,2,\mathrm{},N1).`$
This fact allows us to calculate intersections of the $`V_k`$s formally by using the Eq.(17)
$`V_kV_k^{}={\displaystyle \frac{(1)^N}{N}}{\displaystyle \underset{r=1}{\overset{N1}{}}}{\displaystyle \frac{\alpha ^{r(k+k^{}1)}}{(1\alpha ^r)^{N2}}},(k,k^{}=0,1,2,\mathrm{},N1).`$
By using this relation, we can show that the $`d`$-cycles $`V_{k,\mathrm{}}:=V_kV_{\mathrm{}}`$s have intersection forms
$`V_k\mathrm{}V_k^{}\mathrm{}^{}={\displaystyle \frac{(1)^N}{N}}{\displaystyle \underset{r=1}{\overset{N1}{}}}{\displaystyle \frac{\alpha ^r}{(1\alpha ^r)^{N2}}}\times (\alpha ^{kr}\alpha ^\mathrm{}r)(\alpha ^{k^{}r}\alpha ^\mathrm{}^{}r),`$
$`(k,\mathrm{},k^{},\mathrm{}^{}=0,1,2,\mathrm{},N1).`$
## 7 Central Charge in Large Radius Region
In this section, we calculate the B-type central charge $`Z`$ for Calabi-Yau $`d`$-fold from the point of view of mirror symmetry. We analyze the structure of the $`Z`$ at a generic Kähler structure moduli point near the large radius region. Also we consider relations with this formula with the results in the Gepner model.
### 7.1 B-type Central Charge
In this subsection, we consider the B-type central charge $`Z`$ in the large radius region of the $`M`$. Let us recall that it is a product of a charge vector $`Q_L`$ and canonical basis $`\mathrm{\Pi }`$
$`Z=Q_L\mathrm{\Pi }=\stackrel{~}{Q}\mathrm{\Omega },`$
$`\stackrel{~}{Q}=Q_L𝒩.`$ (20)
The matrix $`𝒩`$ transforms the charges $`Q_L`$ into $`\stackrel{~}{Q}`$ by a fractional redefinition of the charge lattices. When we consider the $`d`$-fold $`M`$, the D<sub>2d-2p</sub> brane charge is given by integrating the $`F^p`$ (more precisely, $`2p`$ form parts of an associated Mukai vector $`v()`$ of a sheaf $``$) over the $`\mathrm{\Sigma }_{2p}`$ with the B-type boundary conditions ($`p=0,1,2,\mathrm{},d`$)
$`Q_{2d2p}={\displaystyle _{\mathrm{\Sigma }_{2p}}}e^F.`$
We prepare several notations and concrete formulae in order to write the central charge $`Z`$ explicitly. First the function $`\varpi \left(+\frac{\rho }{2\pi i}\right)`$ is defined by using a formal parameter $`\rho `$ with $`\rho ^{N1}=0`$
$`\varpi _0^1\times \varpi \left({\displaystyle \frac{\rho }{2\pi i}}\right):=e^{t\rho }\times \mathrm{exp}\left({\displaystyle \underset{n2}{}}\rho ^nx_n\right),`$
$`2\pi it=\mathrm{log}z+{\displaystyle \frac{{\displaystyle \underset{n=1}{\overset{\mathrm{}}{}}}{\displaystyle \frac{(Nn)!}{(n!)^N}}\left({\displaystyle \underset{m=n+1}{\overset{Nn}{}}}{\displaystyle \frac{N}{m}}\right)z^n}{{\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{(Nn)!}{(n!)^N}}z^n}},q=e^{2\pi it},`$
$`x_n={\displaystyle \frac{1}{(2\pi i)^n}}{\displaystyle \frac{1}{n!}}_\rho ^n\mathrm{log}\left[{\displaystyle \underset{m=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{\mathrm{\Gamma }(N(m+\rho )+1)}{\mathrm{\Gamma }(N\rho +1)}}\left({\displaystyle \frac{\mathrm{\Gamma }(\rho +1)}{\mathrm{\Gamma }(m+\rho +1)}}\right)^Nz^m\right]|_{\rho =0}.`$
Also the $`\widehat{K}(\rho )`$ is related to Riemann’s zeta functions
$`\widehat{K}(\rho )=\mathrm{exp}\left[2{\displaystyle \underset{m=1}{}}{\displaystyle \frac{NN^{2m+1}}{2m+1}}\zeta (2m+1)\left({\displaystyle \frac{\rho }{2\pi i}}\right)^{2m+1}\right].`$
Then the $`\widehat{\mathrm{\Pi }}_m`$ and $`\widehat{\widehat{\mathrm{\Pi }}}_n`$ is expressed as $`(0mN2;\mathrm{\hspace{0.17em}0}nN2)`$
$`\varpi _0^1\times \sqrt{\widehat{K}(\rho )}\varpi \left({\displaystyle \frac{\rho }{2\pi i}}\right)=e^{\rho t}\times \mathrm{exp}\left({\displaystyle \underset{n2}{}}(\rho )^ny_n\right),`$
$`\widehat{k}_{2m+1}={\displaystyle \frac{1}{(2\pi i)^{2m+1}}}{\displaystyle \frac{NN^{2m+1}}{2m+1}}\zeta (2m+1),`$
$`y_{2m}:=x_{2m},y_{2m+1}:=x_{2m+1}+\widehat{k}_{2m+1},(m=1,2,\mathrm{}),`$
$`\varpi _0^1\times \widehat{\mathrm{\Pi }}_m=\left[\sqrt{\widehat{A}(+\rho )}(e^\rho 1)^me^{\rho t}\times \mathrm{exp}\left({\displaystyle \underset{n2}{}}(\rho )^ny_n\right)\right]_{\rho ^{N2}},`$ (21)
$`\varpi _0^1\times \widehat{\widehat{\mathrm{\Pi }}}_n=\left[\sqrt{\widehat{A}(+\rho )}e^{n\rho }e^{\rho t}\times \mathrm{exp}\left({\displaystyle \underset{n2}{}}(\rho )^ny_n\right)\right]_{\rho ^{N2}}.`$ (22)
Now we will rewrite the formula Eq.(20) in the bases $`\widehat{\mathrm{\Pi }}`$ and $`\widehat{\widehat{\mathrm{\Pi }}}`$. The $`\widehat{\mathrm{\Pi }}`$ and $`\widehat{\widehat{\mathrm{\Pi }}}`$ are related to the $`\mathrm{\Pi }_n`$ and $`\mathrm{\Omega }_{\mathrm{}}`$ $`(0nN2;\mathrm{\hspace{0.17em}0}\mathrm{}N2)`$
$`\mathrm{\Pi }_n={\displaystyle \underset{m=0}{\overset{N2}{}}}S_{n,m}\widehat{\mathrm{\Pi }}_m,\widehat{\mathrm{\Pi }}_n={\displaystyle \underset{m=0}{\overset{N2}{}}}U_{n,m}\widehat{\widehat{\mathrm{\Pi }}}_m,(0nN2),`$
$`\widehat{\mathrm{\Pi }}_m={\displaystyle \underset{\mathrm{}=0}{\overset{N2}{}}}V_{m,\mathrm{}}\mathrm{\Omega }_{\mathrm{}}`$
$`U_{n,m}=\left(\begin{array}{c}n\\ m\end{array}\right)(1)^{nm},(0nN2;\mathrm{\hspace{0.17em}0}mN2),`$
$`V_{m,\mathrm{}}:=[\sqrt{\widehat{A}(+\rho )}(e^\rho 1)^m(\rho )^{\mathrm{}}]_{\rho ^{N2}},(0mN2;\mathrm{\hspace{0.17em}0}\mathrm{}N2).`$
We can obtain the central charge by using these equations Eqs.(21),(22)
$`Z={\displaystyle \underset{\mathrm{}=0}{\overset{N2}{}}}Q_2\mathrm{}^L\left[\sqrt{\widehat{A}(+\rho )}\left\{{\displaystyle \underset{m=0}{\overset{N2}{}}}S_{\mathrm{},m}(e^\rho 1)^m\right\}e^{\rho t}\times \mathrm{exp}\left({\displaystyle \underset{n2}{}}(\rho )^ny_n\right)\right]_{\rho ^{N2}}`$
$`={\displaystyle \underset{\mathrm{}=0}{\overset{N2}{}}}Q_2\mathrm{}^L\left[\sqrt{\widehat{A}(+\rho )}\left\{{\displaystyle \underset{m,n=0}{\overset{N2}{}}}S_{\mathrm{},m}U_{m,n}\right\}e^{n\rho }e^{\rho t}\times \mathrm{exp}\left({\displaystyle \underset{n2}{}}(\rho )^ny_n\right)\right]_{\rho ^{N2}}.`$ (23)
These formulae Eqs.(23) remind us the couplings of RR fields (potentials) $`C`$ to the gauge fields on the D-brane, that is to say, a Chern-Simons term in the B-model
$`{\displaystyle C}e^{}\sqrt{{\displaystyle \frac{\widehat{A}(M)}{\widehat{A}(N)}}}.`$
Here the $`N`$ is a normal bundle of the world volume. The $`C`$ couples with the Mukai vector
$`e^{}\sqrt{{\displaystyle \frac{\widehat{A}(M)}{\widehat{A}(N)}}}.`$
and the gauge field $`F`$ is combined with the Kalb-Ramond field $`B`$ into the $``$
$`=FB.`$
In the context of the topological sigma model, this $`B`$-field is combined into a complexified Kähler form
$`B+iJ=t[D]([D]\text{H}^2(M)).`$
$`B=\text{Re}(t)[D],J=\text{Im}(t)[D].`$
In our case, the “$`[D]`$” is a $`1`$st Chern class of a hyperplane bundle of $`CP^{N1}`$
$`[D]=c_1\left(𝒪_{CP^{N1}}(1)\right),`$ (24)
and then the $``$ is represented as
$`=FB=F\text{Re}(t)[D].`$
When one shifts the B-field, the real part of the $`t`$ changes, for an example, $`tt+1`$ and the $``$ changes. It implies a coupling
$`Ce^{Re(t)[D]}\sqrt{{\displaystyle \frac{\widehat{A}(M)}{\widehat{A}(N)}}}.`$
In our case, we see an analogous term in the $`Z`$ in Eqs.(23)
$`Z={\displaystyle \underset{\mathrm{}=0}{\overset{N2}{}}}Q_2\mathrm{}^L\left[\sqrt{\widehat{A}(\rho )}\mathrm{}e^{\rho t}\times \mathrm{}\right]_{\rho ^{N2}}.`$
Up to a multiplicative constant, this formal parameter $`\rho `$ could be identified with the divisor $`[D]`$ and the $`e^{t\rho }`$ is represented by using the $`[D]`$
$`e^{t\rho }=e^{t[D]}=e^{(iJ+B)}.`$
Then the $`Z`$ is expressed as integral formulae
$`Z={\displaystyle \underset{\mathrm{}=0}{\overset{N2}{}}}Q_2\mathrm{}^L{\displaystyle _M}[\sqrt{\widehat{A}([D])}\left\{{\displaystyle \underset{m=0}{\overset{N2}{}}}S_{\mathrm{},m}(e^{[D]}1)^m\right\}e^{t[D]}`$
$`\times \mathrm{exp}({\displaystyle \underset{n2}{}}([D])^ny_n)]\times ({\displaystyle _M}[D]^d)^1`$ (25)
$`={\displaystyle \underset{\mathrm{}=0}{\overset{N2}{}}}Q_2\mathrm{}^L{\displaystyle _M}[\sqrt{\widehat{A}([D])}\left\{{\displaystyle \underset{m,n=0}{\overset{N2}{}}}S_{\mathrm{},m}U_{m,n}\right\}e^{n[D]}e^{t[D]}`$
$`\times \mathrm{exp}({\displaystyle \underset{n2}{}}([D])^ny_n)]\times ({\displaystyle _M}[D]^d)^1.`$ (26)
Here the integral over the $`M`$ is defined with $`[H]:=N[D]`$
$`{\displaystyle _M}(\mathrm{})={\displaystyle _{CP^{N1}}}(\mathrm{})[H].`$
In the formula of the $`Z`$ in Eq.(26), there appears a term $`e^{n[D]}`$. When we identify the $`\rho `$ with the $`[D]`$, the $`n[D]`$ is identified with $`c_1\left(𝒪_{CP^{N1}}(n)^{}\right)`$. Then the term $`e^{n[D]}`$ in Eq.(26) is interpreted as a Chern character of the line bundle $`𝒪_{CP^{N1}}(n)^{}`$
$`e^{n[D]}=\text{ch}\left(𝒪_{CP^{N1}}(n)^{}\right).`$
We will explain this geometrical interpretation about $`Z`$ in the next subsection.
As a last topic in this subsection, we consider a monodromy transformation around large radius point $`\text{Im}t=\mathrm{}`$ (equivalently, $`\psi =\mathrm{}`$). When one performs a monodromy transformation around the $`\psi =\mathrm{}`$, the $`t`$ is shifted into $`t+1`$ and the $`B`$-field changes as $`BB+[D]`$. In the $`Z`$, the $`x_n`$s are invariant under the integral shift of the $`t`$, but the term $`e^{t[D]}`$ is transformed into $`e^{[D]}e^{t[D]}`$. This shift affects on the $`Z`$
$`Z={\displaystyle \underset{\mathrm{}=0}{\overset{N2}{}}}Q_2\mathrm{}^L\left[\mathrm{}{\displaystyle \underset{m=0}{\overset{N2}{}}}S_{\mathrm{},m}(e^{[D]}1)^me^{t[D]}\times \mathrm{}\right]`$
$`{\displaystyle \underset{\mathrm{}=0}{\overset{N2}{}}}Q_2\mathrm{}^L\left[\mathrm{}{\displaystyle \underset{m=0}{\overset{N2}{}}}S_{\mathrm{},m}{\displaystyle \underset{m^{}=0}{\overset{N2}{}}}(1)^m^{}(e^{[D]}1)^{m+m^{}}e^{t[D]}\times \mathrm{}\right],`$
$`Z={\displaystyle \underset{\mathrm{}=0}{\overset{N2}{}}}Q_2\mathrm{}^L\left[\mathrm{}\left\{{\displaystyle \underset{m,n=0}{\overset{N2}{}}}S_{\mathrm{},m}U_{m,n}\right\}e^{n[D]}e^{t[D]}\times \mathrm{}\right]`$
$`{\displaystyle \underset{\mathrm{}=0}{\overset{N2}{}}}Q_2\mathrm{}^L\left[\mathrm{}\left\{{\displaystyle \underset{m,n=0}{\overset{N2}{}}}S_{\mathrm{},m}U_{m,n}\right\}e^{(n1)[D]}e^{t[D]}\times \mathrm{}\right].`$
It induces rearrangements of the components of the matrices $`S`$ and $`U`$ and leads to a reshuffle of the charge vector. Also it gives us monodromy matrices on the basis $`\widehat{\mathrm{\Pi }}`$ and $`\widehat{\widehat{\mathrm{\Pi }}}`$. Especially monodromy matrices $`T_{\widehat{\mathrm{\Pi }}}`$ and $`T_\mathrm{\Omega }`$ around $`\psi =\mathrm{}`$ can be obtained in the $`\widehat{\mathrm{\Pi }}`$ and $`\mathrm{\Omega }`$ bases by this consideration
$`\widehat{\mathrm{\Pi }}_m=\varpi _0\times \left[\sqrt{\widehat{A}(\rho )}(e^\rho 1)^me^{\rho t}\times \mathrm{exp}\left({\displaystyle \underset{n2}{}}(\rho )^ny_n\right)\right]_{\rho ^{N2}}`$
$`\varpi _0\times \left[\sqrt{\widehat{A}(\rho )}{\displaystyle \underset{m^{}=0}{\overset{N2}{}}}(1)^m^{}(e^\rho 1)^{m+m^{}}e^{\rho t}\times \mathrm{exp}\left({\displaystyle \underset{n2}{}}(\rho )^ny_n\right)\right]_{\rho ^{N2}},`$
$`\mathrm{\Omega }_{\mathrm{}}=\varpi _0\times \left[e^{\rho t}\times \mathrm{exp}\left({\displaystyle \underset{n2}{}}\rho ^ny_n\right)\right]_\rho ^{\mathrm{}}`$
$`\varpi _0\times \left[e^\rho e^{\rho t}\times \mathrm{exp}\left({\displaystyle \underset{n2}{}}\rho ^ny_n\right)\right]_\rho ^{\mathrm{}},`$
$`T_{\widehat{\mathrm{\Pi }}}={\displaystyle \underset{\mathrm{}=0}{\overset{N2}{}}}(𝒯)^{\mathrm{}},T_\mathrm{\Omega }=e^{𝒯^t},`$
$`𝒯_\mathrm{},\mathrm{}^{}=\delta _{\mathrm{}+1,\mathrm{}^{}},(0\mathrm{}N2;\mathrm{\hspace{0.17em}0}\mathrm{}^{}N2).`$
The formulae of these $`T_{\widehat{\mathrm{\Pi }}}`$ and $`T_\mathrm{\Omega }`$ for $`N=7`$ case coincide with those in section 4.
### 7.2 Relation with Results in the Gepner Model
In this subsection, we investigate the $`Z`$ associated with the charge $`\{Q_j^G\}`$ at the Gepner point and study relations with the B-type boundary state $`|\{L\};M;S`$ in the Gepner model. The $`Z`$ is obtained in Eq.(12) by using the set $`\{Q_j^G\}`$ $`(j=0,1,2,\mathrm{},N1)`$. We can rewrite this formula as
$`Z=N^1{\displaystyle \underset{k=1}{\overset{N1}{}}}(\alpha ^k1)^{N1}\left({\displaystyle \underset{j=0}{\overset{N1}{}}}Q_j^G\alpha ^{kj}\right){\displaystyle _M}[{\displaystyle \frac{\alpha ^k}{e^{[D]}\alpha ^k}}\sqrt{\widehat{A}([D])}e^{t[D]}`$
$`\times \sqrt{\widehat{K}([D])}\times \mathrm{exp}({\displaystyle \underset{n2}{}}([D])^nx_n)]`$ (27)
$`=N^1{\displaystyle \underset{n=0}{\overset{N2}{}}}\{{\displaystyle \underset{\mathrm{}=n}{\overset{N2}{}}}{\displaystyle \underset{j=0}{\overset{N1}{}}}Q_j^GP_{j,\mathrm{}}\left(\begin{array}{c}\mathrm{}\\ n\end{array}\right)(1)^\mathrm{}n\}{\displaystyle _M}[e^{n[D]}\sqrt{\widehat{A}([D])}e^{t[D]}`$
$`\times \sqrt{\widehat{K}([D])}\times \mathrm{exp}({\displaystyle \underset{m2}{}}([D])^mx_m)],`$
$`P_{j,\mathrm{}}=\delta _{\mathrm{},N2}+N(1)^j\mathrm{}\left(\begin{array}{c}N2\mathrm{}\\ N1j\end{array}\right),\alpha =e^{2\pi i/N},`$
Here $`{\displaystyle \underset{j=0}{\overset{N1}{}}}Q_j^GP_{j,m}`$s are integers because the charges $`Q_j^G`$ and the $`P_{j,m}`$s are integers. Information about charges is encoded in a function $`R`$
$`R([D]):=N^1{\displaystyle \underset{k=1}{\overset{N1}{}}}(\alpha ^k1)^{N1}\left({\displaystyle \underset{j=0}{\overset{N1}{}}}Q_j^G\alpha ^{kj}\right){\displaystyle \frac{\alpha ^k}{e^{[D]}\alpha ^k}}.`$
Together with the $`\sqrt{\widehat{A}}`$, analogs of the Mukai vector appear in the $`Z`$
$`\sqrt{\widehat{A}([D])}R([D]),`$
$`\sqrt{\widehat{A}([D])}e^{n[D]}.`$
We investigate this structure in the geometrical point of view. The Calabi-Yau $`d`$-fold $`M`$ is embedded in the ambient projective space $`X=CP^{N1}`$. A line bundle $`𝒪_X(m)`$ of the $`X`$ is defined as an $`m`$ tensor product of a section of a hyperplane bundle. Here we introduce a set of line bundles $`R_a=𝒪_X(a[D])`$. An intersection paring $`I_{a,b}`$ between these bundles $`R_a`$ and $`R_b`$ is defined as a Euler characteristic $`\chi _X(R_a,R_b)`$
$`I_{a,b}=\chi _X(R_a,R_b)={\displaystyle _X}\text{ch}(R_a)^{}\text{ch}(R_b)\text{Td}(TX).`$
We can evaluate the $`I_{a,b}`$ and its inverse explicitly
$`I_{a,b}=\chi _X(R_a,R_b)=\left(\begin{array}{c}d+1+ab\\ ab\end{array}\right),`$
$`(I^1)_{a,b}=(1)^{ab}\left(\begin{array}{c}N\\ ab\end{array}\right),(N=d+2).`$
Next we define a set of dual basis $`\{S_a\}`$ for the bundles $`\{R_a\}`$
$`\text{ch}(S_a):={\displaystyle \underset{b}{}}(I^1)_{b,a}\text{ch}(R_b)={\displaystyle \underset{b}{}}(1)^{ba}\left(\begin{array}{c}N\\ ba\end{array}\right)\text{ch}(R_b),`$
$`\text{ch}(S_a^{}):={\displaystyle \underset{b}{}}(I^1)_{a,b}\text{ch}(R_b^{})={\displaystyle \underset{b}{}}(1)^{ab}\left(\begin{array}{c}N\\ ab\end{array}\right)\text{ch}(R_b^{}).`$
The two sets of bases are orthonormal with respect to the intersection paring
$`R_a,S_b=\chi _X(R_a,S_b)=\delta _{a,b},`$
$`S_a,R_b=\chi _X(S_a,R_b)=\delta _{a,b}.`$
The set of line bundles (sheaves) $`\{𝒪(a)\}`$ $`(a=0,1,2,\mathrm{},N1)`$ is an exceptional collection of the $`CP^{N1}`$ and also turns out to be a foundation of an associated helix of $`CP^{N1}`$. We can consider a left mutation $`𝐋`$ on the set $`\{𝒪(a)\}`$ because a condition $`\text{Ext}^0(𝒪(a1),𝒪(a))=H^0(𝒪(a1),𝒪(a))0`$ is satisfied. It leads to a relation of Chern characters of the bundles
$`\text{ch}(𝐋_{a1}𝒪(a))=\text{ch}(𝒪(a))\chi (𝒪(a1),𝒪(a))\text{ch}(𝒪(a1)),`$ (28)
where we introduce an abbreviated notation $`𝐋_{a1}𝒪(a):=𝐋_{𝒪(a1)}𝒪(a)`$. By using this formula Eq.(28) iteratively, we obtain a relation for the $`S_a^{}`$
$`\text{ch}(𝐋_0𝐋_1\mathrm{}𝐋_{a1}𝒪(a))={\displaystyle \underset{b=0}{\overset{a}{}}}(1)^{ab}\left(\begin{array}{c}N\\ ab\end{array}\right)\text{ch}(𝒪(b))=\text{ch}(S_a^{}).`$
Thus each element $`S_a^{}`$ of the dual basis can be constructed by acting iteratively left mutations on the $`R_a^{}=𝒪(a)`$. Because the Calabi-Yau $`M`$ is realized as a hypersurface in the $`CP^{N1}`$ , we shall restrict the sets of bundles $`\{S_a\}`$, $`\{R_a\}`$ on the $`M`$.
Now we introduce a function $`Z(S_m)`$ associated with each bundle $`S_m`$ on $`M`$ as
$`Z(S_m):=(1)^d{\displaystyle _M}\text{ch}(S_m^{})\sqrt{\widehat{A}(TM)}e^{t[D]}\times \sqrt{\widehat{K}([D])}\times \mathrm{exp}\left({\displaystyle \underset{m2}{}}([D])^mx_m\right)`$
$`(m=0,1,2,\mathrm{},N1).`$
It can be expressed as an integral formula on $`M`$
$`Z(S_m)=(1)^{d+1}N^1{\displaystyle \underset{k=1}{\overset{N1}{}}}\alpha ^{km}(\alpha ^k1)^N{\displaystyle _M}{\displaystyle \frac{\alpha ^k}{e^{[D]}\alpha ^k}}\sqrt{\widehat{A}(TM)}e^{t[D]}`$
$`\times \sqrt{\widehat{K}([D])}\times \mathrm{exp}({\displaystyle \underset{m2}{}}([D])^mx_m),(m=0,1,2,\mathrm{},N1).`$ (29)
We analyze a central charge of Calabi-Yau $`d`$-fold embedded in $`CP^{N1}`$ in the analysis based on the mirror symmetry. The central charge $`Z(\{Q^G\})`$ in Eq.(27) is labelled by a set of charges $`\{Q_j^G\}`$ at the Gepner point. They are transformed cyclically under the $`𝐙_N`$ action around $`\psi =0`$, $`Q_j^GQ_{j+1}^G`$ $`(Q_{j+N}^GQ_j^G)`$. They also satisfy a relation $`{\displaystyle \underset{j=0}{\overset{N1}{}}}Q_j^G=0`$ and we can decompose each $`Q_j^G`$ as $`Q_j^G=:q_j^Gq_{j1}^G`$. Then the central charge $`Z`$ is reexpressed as
$`Z=N^1(1){\displaystyle \underset{k=1}{\overset{N1}{}}}(\alpha ^k1)^N\left({\displaystyle \underset{j=0}{\overset{N1}{}}}q_j^G\alpha ^{jk}\right)`$
$`\times {\displaystyle _M}{\displaystyle \frac{\alpha ^k}{e^{[D]}\alpha ^k}}e^{t[D]}\sqrt{\widehat{A}(TM)}\times \sqrt{\widehat{K}([D])}\times \mathrm{exp}({\displaystyle \underset{m2}{}}([D])^mx_m).`$
By using the formula Eq.(29) for $`Z(S_m)`$, we obtain a formula of the $`Z`$ in terms of geometrical data
$`Z(\{q^G\})=(1)^N{\displaystyle \underset{m=0}{\overset{N1}{}}}q_m^GZ(S_m)`$
$`={\displaystyle \underset{m=0}{\overset{N1}{}}}q_m^G{\displaystyle _M}\text{ch}(S_m^{})e^{t[D]}\sqrt{\widehat{A}(TM)}`$
$`\times \sqrt{\widehat{K}([D])}\times \mathrm{exp}\left({\displaystyle \underset{m2}{}}([D])^mx_m\right).`$ (30)
It is a linear combination of the $`Z(S_m)`$’s. Each $`Z(S_m)`$ has information about a K-theory element of sheaves $`S_m`$ that is constructed through a restriction on the $`M`$. In the large radius region, fractional charges $`\{Q_n\}`$ in the fractional basis are defined as coefficients in an expansion of the above formula
$`Z(\{q^G\})={\displaystyle \underset{n=0}{\overset{d}{}}}{\displaystyle \frac{t^{dn}}{(dn)!}}(1)^nQ_{2(dn)},`$
$`Q_{2(dn)}=(1)^d{\displaystyle \underset{m=0}{\overset{N1}{}}}q_m^G{\displaystyle _M}\text{ch}(S_m^{})\sqrt{\widehat{A}(TM)}[D]^{dn}`$
$`=(1)^d{\displaystyle \underset{m=0}{\overset{N1}{}}}q_m^G{\displaystyle _{\gamma _{2n}}}\text{ch}(S_m^{})\sqrt{\widehat{A}(TM)},`$
where we introduce a cycle $`\gamma _{2n}`$ as a dual of the $`(2d2n)`$ form $`[D]^{dn}`$. We shall take an example with $`q_m^G=\delta _{m,\mathrm{}}`$, that is,
$`Z={\displaystyle _M}\text{ch}(S_{\mathrm{}}^{})\sqrt{\widehat{A}(TM)}e^{t[D]}\times \sqrt{\widehat{K}([D])}\times \mathrm{exp}\left({\displaystyle \underset{m2}{}}([D])^mx_m\right).`$
It is interesting that a pure D$`(2d)`$-brane is always labelled by a set of charges $`q_m^G=\delta _{m,0}`$, equivalently, $`Q_0^G=+1`$, $`Q_1^G=1`$, $`Q_j^G=0`$ $`(j0,1)`$.
Next we compare our result in the sigma model with boundary states associated with the Gepner model. The boundary states are labelled by a set of integers $`L_j`$ $`(j=1,2,\mathrm{},N)`$, $`M`$ and $`S`$ as $`|\{L\};M;S`$. The integer $`M`$ is transformed under the $`𝐙_N`$ monodromy transformation at the Gepner point as $`MM+2`$. On the other hand, the charge $`q_m^G`$ changes in such a transformation as $`q_m^Gq_{m+1}^G`$. So the integral number $`M`$ turns out to be related with the charge $`q_m^G`$. For a trivial state $`|\{L=0\};M=0;S=0`$, we put an ansatz that an associated bundle is trivial one $`𝒪`$. It leads to a condition $`q_m^G=\delta _{m,0}`$. Then by performing a $`𝐙_N`$ transformation, we can construct a central charge $`Z`$ for a boundary state $`|\{L=0\};M=2\mathrm{};S=0`$
$`Z(\{q_m^{G,0}\})={\displaystyle \underset{m=0}{\overset{N1}{}}}q_m^{G,0}{\displaystyle _M}\text{ch}(S_m^{})\sqrt{\widehat{A}(TM)}e^{t[D]}`$
$`\times \sqrt{\widehat{K}([D])}\times \mathrm{exp}\left({\displaystyle \underset{m2}{}}([D])^mx_m\right),`$
$`q_m^{G,0}=\delta _{m,\mathrm{}}.`$
For a boundary state $`|\{L\};M=2\mathrm{};S`$ $`(S=0,2)`$, we obtain an associated central charge in the sigma model
$`Z(\{q_m^G\})={\displaystyle \underset{m=0}{\overset{N1}{}}}q_m^G{\displaystyle _M}\text{ch}(S_m^{})\sqrt{\widehat{A}(TM)}e^{t[D]}`$
$`\times \sqrt{\widehat{K}([D])}\times \mathrm{exp}\left({\displaystyle \underset{m2}{}}([D])^mx_m\right),`$
$`q_m^G={\displaystyle \frac{1}{N}}(1)^{\frac{S}{2}}{\displaystyle \underset{n=0}{\overset{N1}{}}}q_n^{G,0}{\displaystyle \underset{k=1}{\overset{N1}{}}}\alpha ^{k(nm)}\times {\displaystyle \underset{j=1}{\overset{N}{}}}{\displaystyle \frac{\mathrm{sin}{\displaystyle \frac{\pi k(L_j+1)}{N}}}{\mathrm{sin}{\displaystyle \frac{\pi k}{N}}}},`$
$`q_n^{G,0}=\delta _{n,\mathrm{}}.`$
Here we used the formula Eq.(11) of the charges when we switched on the $`\{L_j\}`$. We can construct the central charge associated with the boundary state of the Gepner model.
Now let us return to the formula Eq.(27) and study quantum effects in the $`Z`$. They are contained (non-)perturbatively in the following terms
$`\sqrt{\widehat{A}([D])}R([D])\sqrt{\widehat{K}([D])}e^{t[D]}\times \mathrm{exp}\left({\displaystyle \underset{n2}{}}([D])^nx_n\right).`$ (31)
The term $`\sqrt{\widehat{A}}`$ describes topological features of the associated bundle over the curved space. On the other hand, the $`\widehat{K}`$ is expected to have its origin in the perturbative quantum corrections. The $`\widehat{A}`$ and the $`\widehat{K}`$ are combined into a function
$`\sqrt{\widehat{A}(+\rho )}\sqrt{\widehat{K}(\rho )}={\displaystyle \frac{\left[\mathrm{\Gamma }\left(1+{\displaystyle \frac{\rho }{2\pi i}}\right)\right]^N}{\mathrm{\Gamma }\left(1+{\displaystyle \frac{N\rho }{2\pi i}}\right)}}.`$
Also this relation can be generalized to other Calabi-Yau cases, for an example, a Calabi-Yau $`d`$-fold realized as complete intersections $`M`$ of $`\mathrm{}`$ hypersurfaces $`\{p_j=0\}`$ in products of $`k`$ projective spaces $``$
$`M:=\left(\begin{array}{c}𝐏^{n_1}(w_1^{(1)},\mathrm{},w_{n_1+1}^{(1)})\\ \mathrm{}\\ 𝐏^{n_k}(w_1^{(k)},\mathrm{},w_{n_k+1}^{(k)})\end{array}|\right|\begin{array}{c}d_1^{(1)}\mathrm{}d_{\mathrm{}}^{(1)}\\ \mathrm{}\\ d_1^{(k)}\mathrm{}d_{\mathrm{}}^{(k)}\end{array}).`$
The $`d_j^{(i)}`$ are degrees of the coordinates of $`𝐏^{n_i}(w_1^{(i)},\mathrm{},w_{n_i+1}^{(i)})`$ in the $`j`$-th polynomial $`p_j`$ ($`i=1,2,\mathrm{},k;j=1,2,\mathrm{},\mathrm{}`$). When we introduce a function $`a(v)`$ with variables $`v_i`$ $`(i=1,2,\mathrm{},k)`$
$`a(v):={\displaystyle \frac{{\displaystyle \underset{j=1}{\overset{\mathrm{}}{}}}\mathrm{\Gamma }\left(1+{\displaystyle \underset{i=1}{\overset{k}{}}}d_j^{(i)}v_i\right)}{{\displaystyle \underset{i=1}{\overset{k}{}}}{\displaystyle \underset{j^{}=1}{\overset{n_i+1}{}}}\mathrm{\Gamma }\left(1+w_j^{}^{(i)}v_i\right)}},`$
a generating function of an $`\widehat{A}`$ of the M is represented as
$`\widehat{A}(\lambda ):={\displaystyle \underset{i=1}{\overset{k}{}}}{\displaystyle \underset{j^{}=1}{\overset{n_i+1}{}}}\left({\displaystyle \frac{{\displaystyle \frac{\lambda w_j^{}^{(i)}\rho _i}{2}}}{\mathrm{sinh}{\displaystyle \frac{\lambda w_j^{}^{(i)}\rho _i}{2}}}}\right)\times {\displaystyle \underset{j=1}{\overset{\mathrm{}}{}}}\left({\displaystyle \frac{\mathrm{sinh}{\displaystyle \frac{\lambda (d_j\rho )}{2}}}{{\displaystyle \frac{\lambda (d_j\rho )}{2}}}}\right)`$
$`={\displaystyle \underset{m=1}{}}{\displaystyle \frac{(1)^mB_mX_{2m}}{(2m)!(2m)}}\lambda ^{2m},`$
$`d_j\rho :={\displaystyle \underset{i=1}{\overset{k}{}}}d_j^{(i)}\rho _i,`$
$`X_n:={\displaystyle \underset{i=1}{\overset{k}{}}}{\displaystyle \underset{j^{}=1}{\overset{n_i+1}{}}}(w_j^{}^{(i)}\rho _i)^n{\displaystyle \underset{j=1}{\overset{\mathrm{}}{}}}\left({\displaystyle \underset{i=1}{\overset{k}{}}}d_j^{(i)}\rho _i\right)^n,`$
$`d=\mathrm{}+{\displaystyle \underset{i=1}{\overset{k}{}}}n_i,(\text{dimension}).`$
On the other hand, an associated $`\widehat{K}`$ is defined by using the $`X_n`$s as
$`\widehat{K}(\lambda ):=\mathrm{exp}\left[+2{\displaystyle \underset{m=1}{}}{\displaystyle \frac{\zeta (2m+1)}{2m+1}}\left({\displaystyle \frac{\lambda }{2\pi i}}\right)^{2m+1}X_{2m+1}\right].`$
The $`\widehat{A}(\lambda )`$ and the $`\widehat{K}(\lambda )`$ satisfy a relation
$`\sqrt{\widehat{A}(+\lambda )}\sqrt{\widehat{K}(\lambda )}={\displaystyle \frac{{\displaystyle \underset{i=1}{\overset{k}{}}}{\displaystyle \underset{j^{}=1}{\overset{n_i+1}{}}}\mathrm{\Gamma }\left(1+{\displaystyle \frac{\lambda w_j^{}^{(i)}\rho _i}{2\pi i}}\right)}{{\displaystyle \underset{j=1}{\overset{\mathrm{}}{}}}\mathrm{\Gamma }\left(1+{\displaystyle \underset{i=1}{\overset{k}{}}}{\displaystyle \frac{\lambda d_j^{(i)}\rho _i}{2\pi i}}\right)}}.`$
In our previous paper, we propose a conjecture that the $`\sqrt{\widehat{K}}`$ could be interpreted as loop corrections of the sigma model. (For the quintic case, this statement is confirmed.) If it is true for generic cases, this term might contain effects of perturbative corrections. It seems interesting that these two terms are combined into a combination of Euler’s gamma functions.
The remaining very important term is the $`\mathrm{exp}\left({\displaystyle \underset{n2}{}}([D])^nx_n\right)`$ in Eq.(31). In order to study this function, we consider a set of functions $`\{\stackrel{ˇ}{\omega }_{\mathrm{}}\}`$ defined in an expansion
$`e^{tv}\mathrm{exp}\left({\displaystyle \underset{n2}{}}v^nx_n\right)={\displaystyle \underset{\mathrm{}=0}{}}v^{\mathrm{}}\stackrel{ˇ}{\omega }_{\mathrm{}}.`$
We collect these functions $`\stackrel{ˇ}{\omega }_{\mathrm{}}`$ $`(\mathrm{}=0,1,\mathrm{},N2)`$ into a vector $`\stackrel{~}{V}_0`$
$`\stackrel{~}{V}_0=\left(\begin{array}{cccc}\stackrel{ˇ}{\omega }_0& \stackrel{ˇ}{\omega }_1& \mathrm{}& \stackrel{ˇ}{\omega }_{N2}\end{array}\right).`$
Here the $`\stackrel{ˇ}{\omega }_{\mathrm{}}`$ can be interpreted as a paring of a B-cycle $`\stackrel{~}{\gamma }_{\mathrm{}}`$ and an A-model operator $`𝒪^{(0)}`$ associated with a $`0`$ form on the $`M`$
$`\stackrel{ˇ}{\omega }_{\mathrm{}}=\stackrel{~}{\gamma }_{\mathrm{}}|𝒪^{(0)}.`$
In our Calabi-Yau case, there are $`(d+1)`$ independent A-model operators $`𝒪^{(m)}`$ $`(m=0,1,\mathrm{},d)`$ and we can define an associated vector $`\stackrel{~}{V}_m`$ for each $`𝒪^{(m)}`$
$`\stackrel{~}{V}_m=\left(\begin{array}{cccc}\stackrel{~}{\gamma }_0|𝒪^{(m)}& \stackrel{~}{\gamma }_1|𝒪^{(m)}& \mathrm{}& \stackrel{~}{\gamma }_d|𝒪^{(m)}\end{array}\right),(m=0,1,\mathrm{},d).`$
Then we can introduce a matrix $`\stackrel{~}{\mathrm{\Pi }}`$ by collecting these $`(d+1)`$ vectors
$`\stackrel{~}{\mathrm{\Pi }}=\left(\begin{array}{c}\stackrel{~}{V}_0\\ \stackrel{~}{V}_1\\ \mathrm{}\\ \stackrel{~}{V}_d\end{array}\right),\stackrel{~}{\mathrm{\Pi }}_{m,\mathrm{}}=\stackrel{~}{\gamma }_{\mathrm{}}|𝒪^{(m)}.`$
This matrix $`\stackrel{~}{\mathrm{\Pi }}`$ satisfies a first order differential equation
$`_t\stackrel{~}{\mathrm{\Pi }}=K\stackrel{~}{\mathrm{\Pi }},`$ (32)
$`K:=\left(\begin{array}{ccccc}0& \kappa _0& & & \\ & 0& \kappa _1& & \\ & & \mathrm{}& \mathrm{}& \\ & & & 0& \kappa _{d1}\\ & & & & 0\end{array}\right).`$
The $`\kappa _{\mathrm{}}`$s are fusion couplings of A-model operators defined as
$`𝒪^{(1)}𝒪^{(\mathrm{}1)}=\kappa _{j1}𝒪^{(\mathrm{})}(1\mathrm{}d),`$
$`𝒪^{(1)}𝒪^{(d)}=0,`$ (33)
$`\kappa _m=_t{\displaystyle \frac{1}{\kappa _{m1}}}_t{\displaystyle \frac{1}{\kappa _{m2}}}_t\mathrm{}_t{\displaystyle \frac{1}{\kappa _1}}_t{\displaystyle \frac{1}{\kappa _0}}_t\stackrel{ˇ}{\omega }_{m+1}(1md1),`$
$`\kappa _0=1.`$ (34)
When we put an initial data at a specific moduli point $`t=t_i`$, the $`\stackrel{~}{\mathrm{\Pi }}`$ at a point $`t=t_f`$ is given by integrating the Eq.(32) formally
$`\stackrel{~}{\mathrm{\Pi }}(t_f)=\text{Pexp}\left({\displaystyle _{t_i}^{t_f}}𝑑sK(s)\right)\stackrel{~}{\mathrm{\Pi }}(t_i).`$
That is to say, the matrix $`K`$ plays a role of a connection on the moduli space. In other words, it induces a parallel transformation on the $`t`$-space. When we impose a boundary condition $`\stackrel{~}{\mathrm{\Pi }}(t)e^{t𝖱}`$ on the $`\stackrel{~}{\mathrm{\Pi }}`$ at $`t=t_{\mathrm{}}`$ $`(\text{Im}t_{\mathrm{}}=+\mathrm{})`$, we obtain the $`\stackrel{~}{\mathrm{\Pi }}`$ at a generic point $`t`$
$`\stackrel{~}{\mathrm{\Pi }}(t)=e^{t𝖱}\text{Pexp}\left({\displaystyle _t_{\mathrm{}}^t}𝑑se^{s𝖱}\stackrel{~}{K}(s)e^{+s𝖱}\right)I,`$
$`𝖱:=\left(\begin{array}{ccccc}0& 1& & & \\ & 0& 1& & \\ & & \mathrm{}& \mathrm{}& \\ & & & 0& 1\\ & & & & 0\end{array}\right),`$
$`\stackrel{~}{K}:=K𝖱.`$
Thus the $`\mathrm{exp}\left({\displaystyle \underset{n2}{}}([D])^nx_n\right)`$ in Eq.(31) is evaluated by substituting $`v=[D]`$ in the following formula
$`e^{tv}\mathrm{exp}\left({\displaystyle \underset{n2}{}}v^nx_n\right)={\displaystyle \underset{\mathrm{}=0}{}}v^{\mathrm{}}\stackrel{ˇ}{\omega }_{\mathrm{}}`$
$`=\left(\begin{array}{ccccc}1& t& {\displaystyle \frac{t^2}{2}}& \mathrm{}& {\displaystyle \frac{t^d}{d!}}\end{array}\right)\text{Pexp}\left({\displaystyle _t_{\mathrm{}}^t}𝑑se^{sR}\stackrel{~}{K}(s)e^{+sR}\right)\left(\begin{array}{c}1\\ v\\ v^2\\ \mathrm{}\\ v^d\end{array}\right).`$
As a conclusion, we can interpret the $`\mathrm{exp}\left({\displaystyle \underset{n2}{}}([D])^nx_n\right)`$ in Eq.(31) as an effect of a parallel transport through a path from the large radius point $`t_{\mathrm{}}`$ to a finite $`t`$ in the moduli space. Then the connection in the moduli space is the $`K`$ that are defined by fusion couplings of A-model operators.
## 8 Conclusions and Discussions
In this article, we develop a method to construct the central charge in the topological sigma model in the open string channel. First we analyzed quintic by using the associated prepotential $`F`$ and studied its monodromy properties. For this quintic case, one can easily construct associated canonical bases of period integrals because there is the prepotential for the model. But one cannot expect to find analogs of prepotentials for other $`d`$-fold cases $`(d>3)`$. But we find that the essential part we learned in the analysis seems to be a factorizable property of the matrix $``$ by a (triangular) matrix $`S`$ and a matrix $`\mathrm{\Sigma }`$. Under this consideration, we investigate the $`N=7`$ ($`d=5`$) case concretely in section 4.
The basis $`\mathrm{\Omega }`$ we pick here is a kind of symplectic one with an intersection matrix $`\mathrm{\Sigma }`$. In order to obtain a set of canonical basis $`\mathrm{\Pi }`$, we determine the matrix $`S`$ which connects the $`\mathrm{\Pi }`$ and $`\widehat{\mathrm{\Pi }}`$. It allows us to construct the matrix $`𝒩`$ that transforms the $`\mathrm{\Omega }`$ to the $`\mathrm{\Pi }`$. Some topological invariants appear in the entries of the $`𝒩`$. Some parts of them are characterized by an A-roof genus of the $`M`$. But we cannot give geometrical interpretations to other entries in $`𝒩`$ explicitly.
Together with the data of the Gepner model, we calculate the charge vectors of the D-branes and the central charge. When a set of numbers $`\{L_j\}`$ is specified, a boundary state is constructed in the Gepner model. By calculating an associated charge $`Q_G`$ in the Gepner basis, we construct a formula of the $`Z`$ labelled by the set $`\{L_j\}`$. It is related to the boundary state $`|\{L\};M;S`$.
In section 6, we investigate cycles associated with the sets of periods. At the orbifold point $`\psi =0`$, they coincide with the susy cycles analyzed by Becker et. al.. There we also analyze intersection forms among them. In section 7, we reexpress our result of the $`Z`$ applicable in the large volume region of the $`M`$. We find that the $`Z`$ contains terms analogous to the Mukai vectors. They are interpreted as a product of Chern characters of bundles (sheaves) $`S_m`$s and a square root of the Todd class of the $`M`$. The set of $`S_m`$s is constructed as a dual basis of tautological line bundles of the ambient space $`CP^{N1}`$ by a restriction on $`M`$. In addition, there appear terms that encode perturbative and non-perturbative quantum corrections. The non-perturbative part is induced by a parallel transport through a path from the large radius point $`t_{\mathrm{}}`$ to a finite $`t`$ in the moduli space. It turns out that the fusion coupling matrix $`K`$ of A-model operators plays a role of a connection on the moduli space.
## Acknowledgment
This work was supported by the Grant-in-Aid for Scientific Research from the Ministry of Education, Science and Culture 10740117.
## Appendix A Examples of $`\sqrt{\widehat{A}(\rho )}`$
We write down several examples of the $`\sqrt{\widehat{A}(\rho )}`$ for the $`d`$-fold $`M`$ concretely. A function $`\widehat{A}(\rho )`$ is defined as
$`\widehat{A}(\rho )=\left({\displaystyle \frac{{\displaystyle \frac{\rho }{2}}}{\mathrm{sinh}{\displaystyle \frac{\rho }{2}}}}\right)^N\left({\displaystyle \frac{\mathrm{sinh}{\displaystyle \frac{N\rho }{2}}}{{\displaystyle \frac{N\rho }{2}}}}\right)`$
$`=\mathrm{exp}\left({\displaystyle \underset{m=1}{}}(1)^m\rho ^{2m}{\displaystyle \frac{B_m}{(2m)!}}{\displaystyle \frac{X_{2m}}{2m}}\right).`$
The coefficients $`X_{\mathrm{}}=NN^{\mathrm{}}`$ are represented as some combinations of Chern classes $`c_{\mathrm{}}`$s of the $`M`$ with $`c_1=0`$
$`c(\rho )=1+{\displaystyle \underset{\mathrm{}1}{}}\rho ^{\mathrm{}}{\displaystyle \frac{c_{\mathrm{}}}{N}}=\mathrm{exp}\left({\displaystyle \underset{\mathrm{}1}{}}(1)^\mathrm{}1\rho ^{\mathrm{}}{\displaystyle \frac{X_{\mathrm{}}}{\mathrm{}}}\right),`$
$`X_1=0,X_2={\displaystyle \frac{2c_2}{N}},X_3={\displaystyle \frac{3c_3}{N}},`$
$`X_4={\displaystyle \frac{2c_{2}^{}{}_{}{}^{2}}{N^2}}{\displaystyle \frac{4c_4}{N}},X_5={\displaystyle \frac{5c_2c_3}{N^2}}+{\displaystyle \frac{5c_5}{N}},`$
$`X_6={\displaystyle \frac{2c_{2}^{}{}_{}{}^{3}}{N^3}}+{\displaystyle \frac{3c_{3}^{}{}_{}{}^{2}}{N^2}}+{\displaystyle \frac{6c_2c_4}{N^2}}{\displaystyle \frac{6c_6}{N}},`$
$`X_7={\displaystyle \frac{7c_{2}^{}{}_{}{}^{2}c_3}{N^3}}{\displaystyle \frac{7c_3c_4}{N^2}}{\displaystyle \frac{7c_2c_5}{N^2}}+{\displaystyle \frac{7c_7}{N}},`$
$`X_8={\displaystyle \frac{2c_{2}^{}{}_{}{}^{4}}{N^4}}{\displaystyle \frac{8c_2c_{3}^{}{}_{}{}^{2}}{N^3}}{\displaystyle \frac{8c_{2}^{}{}_{}{}^{2}c_4}{N^3}}+{\displaystyle \frac{4c_{4}^{}{}_{}{}^{2}}{N^2}}`$
$`+{\displaystyle \frac{8c_3c_5}{N^2}}+{\displaystyle \frac{8c_2c_6}{N^2}}{\displaystyle \frac{8c_8}{N}}.`$
In the central charge $`Z`$, there appears a function $`\sqrt{\widehat{A}(\rho )}`$. It is expanded in terms of the $`\rho `$ around $`\rho =0`$
$`\sqrt{\widehat{A}(\rho )}=1+{\displaystyle \underset{\mathrm{}=1}{\overset{d}{}}}\beta _{\mathrm{}}\rho ^{\mathrm{}},\beta _{2m+1}0,`$
$`\beta _2={\displaystyle \frac{c_2}{24N}},\beta _4={\displaystyle \frac{7c_{2}^{}{}_{}{}^{2}}{5760N^2}}{\displaystyle \frac{c_4}{1440N}},`$
$`\beta _6={\displaystyle \frac{31c_{2}^{}{}_{}{}^{3}}{967680N^3}}{\displaystyle \frac{c_{3}^{}{}_{}{}^{2}}{120960N^2}}{\displaystyle \frac{11c_2c_4}{241920N^2}}+{\displaystyle \frac{c_6}{60480N}},`$
$`\beta _8={\displaystyle \frac{127c_{2}^{}{}_{}{}^{4}}{154828800N^4}}{\displaystyle \frac{11c_2c_{3}^{}{}_{}{}^{2}}{14515200N^3}}{\displaystyle \frac{113c_{2}^{}{}_{}{}^{2}c_4}{58060800N^3}}`$
$`+{\displaystyle \frac{13c_{4}^{}{}_{}{}^{2}}{29030400N^2}}+{\displaystyle \frac{c_3c_5}{2419200N^2}}+{\displaystyle \frac{c_2c_6}{907200N^2}}{\displaystyle \frac{c_8}{2419200N}}.`$
We will summarize several examples of $`\sqrt{\widehat{A}(\rho )}`$ for lower dimensional cases
$`\widehat{A}^{1/2}(\rho ;N=3)=1,\widehat{A}^{1/2}(\rho ;N=4)=1+{\displaystyle \frac{\rho ^2}{4}},`$
$`\widehat{A}^{1/2}(\rho ;N=5)=1+{\displaystyle \frac{5\rho ^2}{12}},\widehat{A}^{1/2}(\rho ;N=6)=1+{\displaystyle \frac{5\rho ^2}{8}}{\displaystyle \frac{11\rho ^4}{384}},`$
$`\widehat{A}^{1/2}(\rho ;N=7)=1+{\displaystyle \frac{7\rho ^2}{8}}{\displaystyle \frac{21\rho ^4}{640}},`$
$`\widehat{A}^{1/2}(\rho ;N=8)=1+{\displaystyle \frac{7\rho ^2}{6}}{\displaystyle \frac{7\rho ^4}{240}}+{\displaystyle \frac{229\rho ^6}{1440}},`$
$`\widehat{A}^{1/2}(\rho ;N=9)=1+{\displaystyle \frac{3\rho ^2}{2}}{\displaystyle \frac{\rho ^4}{80}}+{\displaystyle \frac{3233\rho ^6}{10080}},`$
$`\widehat{A}^{1/2}(\rho ;N=10)=1+{\displaystyle \frac{15\rho ^2}{8}}+{\displaystyle \frac{3\rho ^4}{128}}+{\displaystyle \frac{38861\rho ^6}{64512}}{\displaystyle \frac{3542981\rho ^8}{3440640}}.`$
## Appendix B Quintic Case
We summarize monodromy matrices for the quintic case in the appendix for bases $`\varpi `$, $`\mathrm{\Pi }`$, $`\widehat{\mathrm{\Pi }}`$, $`\widehat{\widehat{\mathrm{\Pi }}}`$ and $`\mathrm{\Omega }`$
$`\mathrm{\Pi }=𝒩\mathrm{\Omega },\mathrm{\Pi }=m\varpi ,\varpi =P\mathrm{\Omega },`$
$`\mathrm{\Pi }=S\widehat{\mathrm{\Pi }},\widehat{\mathrm{\Pi }}=U\widehat{\widehat{\mathrm{\Pi }}},U_{m,n}=\left(\begin{array}{c}m\\ n\end{array}\right)(1)^{mn},`$
$`U=\left(\begin{array}{cccc}1& 0& 0& 0\\ 1& 1& 0& 0\\ 1& 2& 1& 0\\ 1& 3& 3& 1\end{array}\right),m=\left(\begin{array}{cccc}1& 0& 0& 0\\ \frac{2}{5}& \frac{2}{5}& \frac{1}{5}& \frac{1}{5}\\ \frac{21}{5}& \frac{1}{5}& \frac{3}{5}& \frac{8}{5}\\ 1& 1& 0& 0\end{array}\right),`$
$`PV=\left(\begin{array}{cccc}1& 0& 0& 0\\ 1& \frac{25}{12}& 0& 5\\ \frac{23}{12}& \frac{15}{4}& 5& 15\\ \frac{23}{12}& \frac{55}{12}& 5& 5\end{array}\right),P=\left(\begin{array}{cccc}0& 0& 0& 1\\ 5& 0& 0& 1\\ 15& 5& 0& 1\\ 5& 5& 5& 4\end{array}\right),`$
$`𝒩=mPV=\left(\begin{array}{cccc}1& 0& 0& 0\\ 0& 1& 0& 0\\ \frac{25}{12}& \frac{11}{2}& 5& 0\\ 0& \frac{25}{12}& 0& 5\end{array}\right),S=mP=\left(\begin{array}{cccc}0& 0& 0& 1\\ 0& 0& 1& 1\\ 0& 5& 8& 3\\ 5& 0& 0& 0\end{array}\right).`$
We write down monodromy matrices associated with singular points $`\psi =0,1,\mathrm{}`$ for various bases:
1. monodromy matrices associated with the $`\psi =0`$
$`A_\varpi =\left(\begin{array}{cccc}0& 1& 0& 0\\ 0& 0& 1& 0\\ 1& 1& 1& 1\\ 1& 0& 0& 0\end{array}\right),A_\mathrm{\Pi }=\left(\begin{array}{cccc}1& 0& 0& 1\\ 1& 1& 0& 1\\ 3& 5& 1& 3\\ 5& 8& 1& 4\end{array}\right),`$
$`A_{\widehat{\mathrm{\Pi }}}=\left(\begin{array}{cccc}4& 1& 0& 0\\ 10& 1& 1& 0\\ 10& 0& 1& 1\\ 5& 0& 0& 1\end{array}\right),A_\mathrm{\Omega }=\left(\begin{array}{cccc}1& \frac{25}{12}& 0& 5\\ 1& \frac{13}{12}& 0& 5\\ \frac{1}{2}& \frac{131}{144}& 1& \frac{55}{12}\\ \frac{1}{6}& \frac{103}{144}& 1& \frac{23}{12}\end{array}\right),`$
$`A_{\widehat{\widehat{\mathrm{\Pi }}}}=\left(\begin{array}{cccc}5& 1& 0& 0\\ 15& 0& 1& 0\\ 35& 0& 0& 1\\ 71& 4& 6& 4\end{array}\right).`$
2. monodromy matrices associated with the $`\psi =1`$
$`P_\varpi =\left(\begin{array}{cccc}2& 1& 0& 0\\ 1& 0& 0& 0\\ 4& 4& 1& 0\\ 4& 4& 0& 1\end{array}\right),P_\mathrm{\Pi }=\left(\begin{array}{cccc}1& 0& 0& 1\\ 0& 1& 0& 0\\ 0& 0& 1& 0\\ 0& 0& 0& 1\end{array}\right),`$
$`P_{\widehat{\mathrm{\Pi }}}=\left(\begin{array}{cccc}1& 0& 0& 0\\ 5& 1& 0& 0\\ 5& 0& 1& 0\\ 5& 0& 0& 1\end{array}\right),P_\mathrm{\Omega }=\left(\begin{array}{cccc}1& \frac{25}{12}& 0& 5\\ 0& 1& 0& 0\\ 0& \frac{125}{144}& 1& \frac{25}{12}\\ 0& 0& 0& 1\end{array}\right),`$
$`P_{\widehat{\widehat{\mathrm{\Pi }}}}=\left(\begin{array}{cccc}1& 0& 0& 0\\ 5& 1& 0& 0\\ 15& 0& 1& 0\\ 35& 0& 0& 1\end{array}\right).`$
3. monodromy matrices associated with the $`\psi =\mathrm{}`$ ($`tt+1`$)
$`T_\varpi =\left(\begin{array}{cccc}1& 0& 0& 0\\ 2& 0& 0& 1\\ 4& 1& 0& 4\\ 5& 1& 1& 3\end{array}\right),T_\mathrm{\Pi }=\left(\begin{array}{cccc}1& 0& 0& 0\\ 1& 1& 0& 0\\ 8& 5& 1& 0\\ 5& 3& 1& 1\end{array}\right),`$
$`T_{\widehat{\mathrm{\Pi }}}=\left(\begin{array}{cccc}1& 1& 1& 1\\ 0& 1& 1& 1\\ 0& 0& 1& 1\\ 0& 0& 0& 1\end{array}\right),T_\mathrm{\Omega }=\left(\begin{array}{cccc}1& 0& 0& 0\\ 1& 1& 0& 0\\ \frac{1}{2}& 1& 1& 0\\ \frac{1}{6}& \frac{1}{2}& 1& 1\end{array}\right),`$
$`T_{\widehat{\widehat{\mathrm{\Pi }}}}=\left(\begin{array}{cccc}4& 6& 4& 1\\ 1& 0& 0& 0\\ 0& 1& 0& 0\\ 0& 0& 1& 0\end{array}\right).`$
|
warning/0003/math0003064.html
|
ar5iv
|
text
|
# Abstract
## Abstract
In 1994, Sule presented the necessary and sufficient conditions of the feedback stabilizability of systems over unique factorization domains in terms of elementary factors and in terms of reduced minors. Recently, Mori and Abe have generalized his theory over commutative rings. They have introduced the notion of the generalized elementary factor, which is a generalization of the elementary factor, and have given the necessary and sufficient condition of the feedback stabilizability. In this paper, we present two generalization of the reduced minors. Using each of them, we state the necessary and sufficient condition of the feedback stabilizability over commutative rings. Further we present the relationship between the generalizations and the generalized elementary factors.
## Keywords
Linear systems, Feedback stabilization, Factorization approach, Systems over rings
## 1 Introduction
This paper is concerned with the coordinate-free approach to control systems. The coordinate-free approach is a factorization approach but does not require the coprime factorizations of plants.
The factorization approach was patterned after Desoer *et al.* and Vidyasagar *et al.*, which has the advantage that it embraces, within a single framework, numerous linear systems such as continuous-time as well as discrete-time systems, lumped as well as distributed systems, $`1`$-D as well as $`n`$-D (multidimensional) systems, etc.. In this approach, when problems such as feedback stabilization are studied, one can focus on the key aspects of the problem under study rather than be distracted by the special features of a particular class of linear systems. A transfer function of this approach is considered as the ratio of two stable causal transfer functions and the set of stable causal transfer functions forms a commutative ring. For a long time, the theory of the factorization approach had been founded on the coprime factorizability of transfer matrices, which is satisfied in the case where the set of stable causal transfer functions is such a commutative ring as a Euclidean domain, a principal ideal, or a Bézout domain.
However, Anantharam in showed that there exist models in which some stabilizable plants do not have right-/left-coprime factorizations. He considered the case where $`[\sqrt{5}\mathrm{i}]`$ ($`[x]/(x^2+5)`$) is the set of stable causal transfer functions, where $``$ is the ring of integers and $`\mathrm{i}`$ the imaginary unit. Using it, he showed that there exists a stabilizable plant which does not have right-/left-coprime factorizations. Further Mori in has recently considered the case where $`[z^2,z^3]`$ is the set of stable causal transfer functions, where $`z`$ denotes the unit delay operator and $``$ the real field. This set is corresponding to the discrete finite-time delay system which does not have the unit delay. He has presented that in the model, some stabilizable plants do not have right-/left-coprime factorizations. Both $`[\sqrt{5}\mathrm{i}]`$ and $`[z^2,z^3]`$ are not unique factorization domains.
Sule in has presented a theory of the feedback stabilization of multi-input multi-output strictly causal plants over commutative rings with some restrictions. This approach to the stabilization theory is called “coordinate-free approach” in the sense that the coprime factorizability of transfer matrices is not required.
In the case where the set of stable causal transfer functions is a unique factorization domain, Sule in introduced two notions, that is, elementary factors and reduced minors. Using each of them he gave the necessary and sufficient condition of the feedback stabilizability of the causal plants over commutative rings (Theorem 4 and Corollary 2 of ). Especially, using elementary factors, Sule presented a construction method of a stabilizing controller of a stabilizable plant. Recently, Mori and Abe in have generalized his theory over commutative rings. They have introduced the notion of the generalized elementary factor, which is a generalization of the elementary factor, and have given the necessary and sufficient condition of the feedback stabilizability. Further Lin in has presented the necessary and sufficient condition of the (structural) stabilizability of the multidimensional systems with the construction method of a stabilizing controller. In the case of the structural stability, it is known that the set of stable causal transfer functions is a unique factorization domain. Lin in introduced a notion “generating polynomial” about the plants and presented the necessary and sufficient condition of the stabilizability of the multidimensional systems with the construction method of a stabilizing controller. It is known that the notion of the generating polynomial is equivalent to the notion of the reduced minors.
In this paper we have two main objectives. The first one is to generalize the notion of the reduced minors and, using the generalizations, to state the necessary and sufficient condition of the feedback stabilizability over commutative rings since the original definition has been given on unique factorization domains. We will present two generalizations. The other is to present the relationship between the generalizations and the generalized elementary factors.
Historically the minors concerning the plants are much investigated (e.g. ). We will present that in the coordinate-free approach, the minors can play a role to state the feedback stabilizability, that is, the *projectivity* of the ideal generated by minors concerning the plant is a criterion of the feedback stability.
This paper is organized as follows. After this introduction, we begin on the preliminary in Section 2, in which we give mathematical preliminaries, set up the feedback stabilization problem and present the previous results. In Section 3, we present the previous results of the feedback stabilizability expressed with the elementary factors, its derivation, and the reduced minors. We present a generalization of the reduced minor in Section 4 and using it present the necessary and sufficient condition of the feedback stabilizability over commutative rings in Section 5. Then in Section 6 we present another generalization of the reduced minors and its relation to the generalized elementary factors.
## 2 Preliminaries
In the following we begin by introducing the notations of commutative rings, matrices, and modules used in this paper. Then we give the formulation of the feedback stabilization problem.
### 2.1 Notations
#### Commutative Rings
In this paper, we consider that any commutative ring has the identity $`1`$ different from zero. Let $``$ denote a (unspecified) commutative ring. The total ring of fractions of $``$ is denoted by $`()`$.
We will consider that *the set of stable causal transfer functions* is a commutative ring, which is denoted by $`𝒜`$ throughout this paper. Further, we will use the following rings of fractions.
1. The first one appears as the total ring of fractions of $`𝒜`$, which is denoted by $`(𝒜)`$ or simply by $``$; that is, $`=\{n/d|n,d𝒜,d\text{ is a nonzerodivisor}\}`$. This will be considered as *the set of all possible transfer functions*.
2. Let $`f`$ denote a nonzero (but possibly nonzerodivisor) element of $`𝒜`$. Given a set $`S_f=\{1,f,f^2,\mathrm{}\}`$, which is a multiplicative subset of $`𝒜`$, we denote by $`𝒜_f`$ the ring of fractions of $`𝒜`$ with respect to the multiplicative subset $`S_f`$; that is, $`𝒜_f=\{n/d|n𝒜,dS_f\}`$.
3. Let $`\mathrm{p}`$ denote a prime ideal of $`𝒜`$ and $`S`$ the complement of the prime ideal $`\mathrm{p}`$, that is, $`S=𝒜\backslash \mathrm{p}`$. Then $`S`$ is a multiplicative subset of $`𝒜`$. We denote by $`𝒜_\mathrm{p}`$ the ring of fractions of $`𝒜`$ with respect to the multiplicative subset $`S`$; that is, $`𝒜_\mathrm{p}=\{n/d|n𝒜,dS\}`$.
4. The last one is the total ring of fractions of $`𝒜_f`$ or $`𝒜_\mathrm{p}`$, which is denoted by $`(𝒜_f)`$ and $`(𝒜_\mathrm{p})`$; that is, $`(𝒜_f)=\{n/d|n,d𝒜_f,d\text{ is a nonzerodivisor of }𝒜_f\}`$ and $`(𝒜_\mathrm{p})=\{n/d|n,d𝒜_\mathrm{p},d\text{ is a nonzerodivisor of }𝒜_\mathrm{p}\}`$. If $`f`$ is a nonzerodivisor of $`𝒜`$, $`(𝒜_f)`$ coincides with the total ring of fractions of $`𝒜`$. Otherwise, they do not coincide.
In the case where $`𝒜`$ is a unique factorization domain, we call $`a`$ in $`𝒜`$ *the radical of* $`b`$ in $`𝒜`$ if $`a`$ has all nonunit factors of $`b`$ and is squarefree, that is, $`a`$ does not have duplicated nonunit factors. Note here that the radical defined here is unique up to any unit multiple.
For convenience, throughout the paper, if $`a𝒜`$ ($`a`$), then $`a`$ itself denotes $`a/1`$ in $`𝒜_f`$ and $`𝒜_\mathrm{p}`$ ($`a/1`$ in $`()`$). Moreover if $`a𝒜_f`$ or $`𝒜_\mathrm{p}`$ ($`a`$) and if there exists $`b𝒜`$ such that $`a=b/1`$ over $`𝒜_f`$ or $`𝒜_\mathrm{p}`$ (over $`()`$), then we regard $`a`$ as an element of $`𝒜`$ ($``$).
In the rest of the paper, we will use $``$ as an unspecified commutative ring and mainly suppose that $``$ denotes one of $`𝒜`$, $`𝒜_f`$, and $`𝒜_\mathrm{p}`$.
We will denote by $`Spec()`$ the set of all prime ideals of $``$ and by $`Max()`$ the set of all maximal ideals of $``$. Suppose that $`𝔞`$ is an ideal of $``$. Then we denote by $`𝔞_f`$ the ideal of fractions of $`𝔞`$ with respect to $`\{1,f,f^2,\mathrm{}\}`$ with $`f`$ (that is, $`𝔞_f=\{n/d|n𝔞,d\{1,f,f^2,\mathrm{}\}\}`$) and by $`𝔞_\mathrm{p}`$ the ideal of fractions of $`𝔞`$ with respect to $`\backslash \mathrm{p}`$ with $`\mathrm{p}Spec()`$ (that is, $`𝔞_\mathrm{p}=\{n/d|n𝔞,d\backslash \mathrm{p}\}`$). If $`𝔞`$ is an ideal of $``$ and if $`S`$ is a subset of $``$, then we denote by $`(𝔞:S)`$ the *quotient ideal* which is the set $`\{f|fS𝔞\}`$.
The reader is referred to Chapter 3 of for the ring of fractions.
#### Matrices
The set of matrices over $``$ of size $`x\times y`$ is denoted by $`^{x\times y}`$. Further, the set of square matrices over $``$ of size $`x`$ is denoted by $`()_x`$. The identity and the zero matrices are denoted by $`E_x`$ and $`O_{x\times y}`$, respectively, if the sizes are required, otherwise they are denoted by $`E`$ and $`O`$.
Matrix $`A`$ over $``$ is said to be *nonsingular* $`\mathbf{(}`$*singular*$`\mathbf{)}`$ *over $``$* if the determinant of the matrix $`A`$ is a nonzerodivisor $`\mathbf{(}`$a zerodivisor$`\mathbf{)}`$ of $``$. Matrices $`A`$ and $`B`$ over $``$ are *right-* *$`\mathbf{(}`$left-$`\mathbf{)}`$coprime over $``$* if there exist matrices $`X`$ and $`Y`$ over $``$ such that $`XA+YB=E`$ $`\mathbf{(}AX+BY=E\mathbf{)}`$ holds. Note that, in the sense of the above definition, two matrices which have no common right-$`\mathbf{(}`$left-$`\mathbf{)}`$factors except invertible matrices may not be right-$`\mathbf{(}`$left-$`\mathbf{)}`$coprime over $``$. Further, an ordered pair $`(N,D)`$ of matrices $`N`$ and $`D`$ is said to be a *right-coprime factorization over $``$* of $`P`$ if (i) $`D`$ is nonsingular over $``$, (ii) $`P=ND^1`$ over $`()`$, and (iii) $`N`$ and $`D`$ are right-coprime over $``$. As the parallel notion, the *left-coprime factorization over $``$* of $`P`$ is defined analogously. That is, an ordered pair $`(\stackrel{~}{D},\stackrel{~}{N})`$ of matrices $`\stackrel{~}{N}`$ and $`\stackrel{~}{D}`$ is said to be a *left-coprime factorization over $``$* of $`P`$ if (i) $`\stackrel{~}{D}`$ is nonsingular over $``$, (ii) $`P=\stackrel{~}{D}^1\stackrel{~}{N}`$ over $`()`$, and (iii) $`\stackrel{~}{N}`$ and $`\stackrel{~}{D}`$ are left-coprime over $``$. Note that the order of the “denominator” and “numerator” matrices is interchanged in the latter case. This is to reinforce the point that if $`(N,D)`$ is a right-coprime factorization over $``$ of $`P`$, then $`P=ND^1`$, whereas if $`(\stackrel{~}{D},\stackrel{~}{N})`$ is a left-coprime factorization over $``$ of $`P`$, then $`P=\stackrel{~}{D}^1\stackrel{~}{N}`$ according to . For short, we may omit “over $``$” when $`=𝒜`$, and “right” and “left” when the size of matrix is $`1\times 1`$. In the case where matrices are potentially used to express *left* fractional form and/or *left* coprimeness, we usually attach a tilde ‘$`\stackrel{~}{}`$’ to symbols; for example $`\stackrel{~}{N}`$, $`\stackrel{~}{D}`$ for $`P=\stackrel{~}{D}^1\stackrel{~}{N}`$ and $`\stackrel{~}{Y}`$, $`\stackrel{~}{X}`$ for $`\stackrel{~}{Y}N+\stackrel{~}{X}D=E`$.
#### Modules
Let $`M_r(X)`$$`\mathbf{(}M_c(X)\mathbf{)}`$ denote the $``$-module generated by rows $`\mathbf{(}`$columns$`\mathbf{)}`$ of a matrix $`X`$ over $``$. Let $`X=AB^1=\stackrel{~}{B}^1\stackrel{~}{A}`$ be a matrix over $`()`$, where $`A`$$`B`$$`\stackrel{~}{A}`$$`\stackrel{~}{B}`$ are matrices over $``$. It is known that $`M_r(\left[\begin{array}{cc}A^t& B^t\end{array}\right]^t)`$ $`\mathbf{(}M_c(\left[\begin{array}{cc}\stackrel{~}{A}& \stackrel{~}{B}\end{array}\right])\mathbf{)}`$ is unique up to an isomorphism with respect to any choice of fractions $`AB^1`$ of $`X`$$`\mathbf{(}\stackrel{~}{B}^1\stackrel{~}{A}`$ of $`X\mathbf{)}`$ (Lemma 2.1 of ). Therefore, for a matrix $`X`$ over $``$, we denote by $`𝒯_{X,}`$ and $`𝒲_{X,}`$ the modules $`M_r(\left[\begin{array}{cc}A^t& B^t\end{array}\right]^t)`$ and $`M_c(\left[\begin{array}{cc}\stackrel{~}{A}& \stackrel{~}{B}\end{array}\right])`$, respectively.
An $``$-module $`M`$ is called *free* if it has a basis, that is, a linearly independent system of generators. The *rank* of a free $``$-module $`M`$ is equal to the cardinality of a basis of $`M`$, which is independent of the basis chosen. An $``$-module $`M`$ is called *projective* if it is a direct summand of a free $``$-module, that is, there is a module $`N`$ such that $`MN`$ is free. The reader is referred to Chapter 2 of for the module theory.
We will consider occasionally ideals as modules in this paper. So, we will apply the words “projective,” “free,” and “isomorphic” to ideals. It is easy to check that an ideal which is free as a module is equivalent to a principal ideal whose generator is a nonzerodivisor.
### 2.2 Feedback Stabilization Problem
The stabilization problem considered in this paper follows that of Sule in , and Mori and Abe in , who consider the feedback system $`\mathrm{\Sigma }`$ \[20, Ch.5, Figure 5.1\] as in Figure 1.
For further details the reader is referred to . Throughout the paper, the plant we consider has $`m`$ inputs and $`n`$ outputs, and its transfer matrix, which is also called a *plant* itself simply, is denoted by $`P`$ and belongs to $`^{n\times m}`$. We can always represent $`P`$ in the form of a fraction $`P=ND^1`$ $`\mathbf{(}P=\stackrel{~}{D}^1\stackrel{~}{N}\mathbf{)}`$, where $`N𝒜^{n\times m}`$$`\mathbf{(}\stackrel{~}{N}𝒜^{n\times m}\mathbf{)}`$ and $`D(𝒜)_m`$$`\mathbf{(}\stackrel{~}{D}(𝒜)_n\mathbf{)}`$ with nonsingular $`D`$$`\mathbf{(}\stackrel{~}{D}\mathbf{)}`$.
###### Definition 1.
For $`P^{n\times m}`$ and $`C^{m\times n}`$, a matrix $`H(P,C)()_{m+n}`$ is defined as
$$H(P,C)=\left[\begin{array}{cc}(E_n+PC)^1& P(E_m+CP)^1\\ C(E_n+PC)^1& (E_m+CP)^1\end{array}\right]$$
(1)
provided that $`det(E_n+PC)`$ is a nonzerodivisor of $`𝒜`$. This $`H(P,C)`$ is the transfer matrix from $`\left[\begin{array}{cc}u_1^t& u_2^t\end{array}\right]^t`$ to $`\left[\begin{array}{cc}e_1^t& e_2^t\end{array}\right]^t`$ of the feedback system $`\mathrm{\Sigma }`$. If (i) $`det(E_n+PC)`$ is a nonzerodivisor of $`𝒜`$ and (ii) $`H(P,C)(𝒜)_{m+n}`$, then we say that the plant $`P`$ is *stabilizable*, $`P`$ is *stabilized* by $`C`$, and $`C`$ is a *stabilizing controller* of $`P`$.
Since the transfer matrix $`H(P,C)`$ of the stable causal feedback system has all entries in $`𝒜`$, we call the above notion *$`𝒜`$-stabilizability*. One can further introduce the notion of *$``$-stabilizability* with either $`=𝒜_f`$ or $`=𝒜_\mathrm{p}`$ as follows.
###### Definition 2.
Suppose that $``$ is either $`𝒜_f`$ with $`f𝒜\backslash \{0\}`$ or $`𝒜_\mathrm{p}`$ with $`\mathrm{p}Spec(𝒜)`$. If (i) $`det(E_n+PC)`$ is a nonzerodivisor of $``$ and (ii) $`H(P,C)()_{m+n}`$, then we say that the plant $`P`$ is *$``$-stabilizable*, $`P`$ is *$``$-stabilized* by $`C`$, and $`C`$ is an *$``$-stabilizing controller* of $`P`$.
The causality of transfer functions is an important physical constraint. We employ, in this paper, the definition of the causality from Vidyasagar *et al.*\[21, Definition 3.1\].
###### Definition 3.
Let $`𝒵`$ be a prime ideal of $`𝒜`$, with $`𝒵𝒜`$, including all zerodivisors. Define the subsets $`𝒫`$ and $`𝒫_\text{s}`$ of $``$ as follows:
$`𝒫`$ $`=`$ $`\{a/b|a𝒜,b𝒜\backslash 𝒵\},`$
$`𝒫_\text{s}`$ $`=`$ $`\{a/b|a𝒵,b𝒜\backslash 𝒵\}.`$
Then every transfer function in $`𝒫`$ $`\mathbf{(}𝒫_\text{s}\mathbf{)}`$ is called *causal* $`\mathbf{(}`$*strictly causal*$`\mathbf{)}`$. Analogously, if every entry of a transfer matrix $`F`$ is in $`𝒫`$ $`\mathbf{(}𝒫_\text{s}\mathbf{)}`$, the transfer matrix $`F`$ is called *causal* $`\mathbf{(}`$*strictly causal*$`\mathbf{)}`$. A matrix over $`𝒜`$ is said to be *$`𝒵`$-nonsingular* if the determinant is in $`𝒜\backslash 𝒵`$, and *$`𝒵`$-singular* otherwise.
Before proceeding the next section, we here introduce several symbols used throughout this paper. The symbol $``$ denotes the family of all sets of $`m`$ distinct integers between $`1`$ and $`m+n`$, and $`𝒥`$ the family of all sets of $`n`$ distinct integers between $`1`$ and $`m+n`$ (recall that $`m`$ and $`n`$ are the numbers of the inputs and the outputs, respectively). Normally, elements of $``$ ($`𝒥`$) will be denoted by $`I`$ ($`J`$) possibly with suffices. They will be used as suffices as well as sets. If $`I`$ is an element of $``$ and if $`i_1,\mathrm{},i_m`$ are elements of $`I`$ with ascending order, that is, $`i_a<i_b`$ if $`a<b`$, then the symbol $`\mathrm{\Delta }_I`$ denotes the $`m\times (m+n)`$ matrix whose $`(k,i_k)`$-entry is $`1`$ for $`i_kI`$ and zero otherwise. Analogously if $`J`$ is an element of $`𝒥`$ and if $`j_1,\mathrm{},j_n`$ are elements of $`J`$ with ascending order, then the symbol $`\mathrm{\Delta }_J`$ denotes the $`n\times (m+n)`$ matrix whose $`(k,j_k)`$-entry is $`1`$ for $`j_kJ`$ and zero otherwise.
## 3 Previous Results
In this section, we recall the previous results about the necessary and sufficient condition of the feedback stabilizability. First one is stated in terms of the elementary factors and the other in terms of the reduced minors.
### 3.1 Feedback Stabilizability in terms of Elementary Factors
To state the result, we first recall the notion of the elementary factors, which was defined under the assumption that $`𝒜`$ is a unique factorization domain.
###### Definition 4.
(Elementary Factors, \[18, p.1689\] Suppose that $`𝒜`$ is a unique factorization domain. Denote by $`T`$ and $`W`$ the matrices $`\left[\begin{array}{cc}N^t& dE_m\end{array}\right]^t`$ and $`\left[\begin{array}{cc}N& dE_n\end{array}\right]^t`$ over $`𝒜`$ with $`P=Nd^1`$. Further denote by $`^{}`$ $`\mathbf{(}𝒥^{}\mathbf{)}`$ the set of $`I`$’s in $``$ $`\mathbf{(}J`$’s in $`𝒥\mathbf{)}`$ such that $`\mathrm{\Delta }_IT`$ $`\mathbf{(}\mathrm{\Delta }_JW^t\mathbf{)}`$ is nonsingular. Then for each $`I^{}`$, let $`f_I`$ be the radical of the least common multiple of all the denominators of the matrix $`T(\mathrm{\Delta }_IT)^1`$ and for each $`J𝒥^{}`$, $`g_J`$ be the radical of the least common multiple of all the denominators of the matrix $`W^t(\mathrm{\Delta }_JW^t)^1`$. Then $`f_I`$ ($`g_J`$) is called *the elementary factor of the matrix $`T`$ $`\mathbf{(}W\mathbf{)}`$ with respect to $`I`$* $`\mathbf{(}J𝒥\mathbf{)}`$, $`F=\{f_I|I^{}\}`$ *the family of elementary factors of the matrix $`T`$*, $`G=\{g_J|J𝒥^{}\}`$ *the family of elementary factors of the matrix $`W`$*, and $`H=\{h_{IJ}:=f_Ig_J|I^{},J𝒥^{}\}`$ *the family of elementary factors of $`P`$*.
Then the necessary and sufficient condition of the feedback stabilizability is given as follows.
###### Theorem 5.
(Theorem 4 of Suppose that $`𝒜`$ is a unique factorization domain. Then the plant $`P`$ is stabilizable if and only if the elementary factors of $`P`$ are coprime, that is, $`_{I^{},J𝒥^{}}(h_{IJ})=𝒜`$.
In the proof of this theorem, Sule gave a method to construct a stabilizing controller of the plant.
The result above has been extended to include systems over commutative rings by Mori and Abe in as follows. They introduced the notion of the generalized elementary factors, which is a generalization of the elementary factors, and using it, stated the necessary and sufficient conditions of the feedback stabilizability over commutative rings.
###### Definition 6.
(Generalized Elementary Factors, Definition 3.1 of Denote by $`T`$ the matrix $`\left[\begin{array}{cc}N^t& D^t\end{array}\right]^t`$ over $`𝒜`$ with $`P=ND^1`$. For each $`I`$, an ideal $`\mathrm{\Lambda }_{PI}`$ over $`𝒜`$ is defined as
$$\mathrm{\Lambda }_{PI}=\{\lambda 𝒜|K𝒜^{(m+n)\times m}\lambda T=K\mathrm{\Delta }_IT\}.$$
We call the ideal $`\mathrm{\Lambda }_{PI}`$ the *generalized elementary factor* of the plant $`P`$ with respect to $`I`$. Further, the set of all $`\mathrm{\Lambda }_{PI}`$’s is denoted by $`_P`$, that is, $`_P=\{\mathrm{\Lambda }_{PI}|I\}`$.
In the case where $`𝒜`$ is a unique factorization domain, a generalized elementary factor with respect to $`I`$ is a principal ideal and the radical of its generator is an elementary factor of $`T`$ with respect to $`I`$ up to a unit multiple.
It is known that the generalized elementary factor of a plant $`P`$ is independent of the choice of fractions $`ND^1=P`$ (Lemma 3.3 of ).
The following is the necessary and sufficient conditions of the feedback stabilizability.
###### Theorem 7.
(Theorem 3.2 of Consider a causal plant $`P`$. Then the following statements are equivalent:
1. The plant $`P`$ is stabilizable.
2. $`𝒜`$-modules $`𝒯_{P,𝒜}`$ and $`𝒲_{P,𝒜}`$ are projective.
3. The set of all generalized elementary factors of $`P`$ generates $`𝒜`$; that is, $`_P`$ satisfies:
$$\underset{\mathrm{\Lambda }_{PI}_P}{}\mathrm{\Lambda }_{PI}=𝒜.$$
(2)
Provided that we can check (2) and that we can construct the right-coprime factorizations over $`𝒜_{\lambda _I}`$ of the given causal plant, where $`\lambda _I`$ is a nonzero element of $`𝒜`$, Mori and Abe have given a method to construct a causal stabilizing controller of a causal stabilizable plant, which has been given in the proof of “(iii)$``$(i)” of Theorem 3.2 of .
### 3.2 Feedback Stabilizability in terms of Reduced Minors
We first recall the definition of the reduced minors and then state the necessary and sufficient conditions of the feedback stabilizability in terms of the reduced minors. We suppose in this subsection that $`𝒜`$ is a unique factorization domain.
###### Definition 8.
(Reduced Minors, \[18, p.1690\] Let $`P`$ be a plant of $`^{n\times m}`$, $`N`$ a matrix of $`𝒜^{n\times m}`$, and $`d`$ an element of $`𝒜`$ such that $`P=Nd^1`$. Denote by $`T`$ and $`W`$ the matrices $`\left[\begin{array}{cc}N^t& dE_m\end{array}\right]^t`$ and $`\left[\begin{array}{cc}N& dE_n\end{array}\right]`$. Let $`t_I=det(\mathrm{\Delta }_IT)`$ $`\mathbf{(}w_J=det(\mathrm{\Delta }_JW^t)\mathbf{)}`$, which is a full-size minor of the matrix $`T`$ $`\mathbf{(}W\mathbf{)}`$, for $`I`$ $`\mathbf{(}J𝒥\mathbf{)}`$. Let $`d_t`$ $`\mathbf{(}d_w\mathbf{)}`$ be the greatest common factor of $`t_I`$’s $`\mathbf{(}w_J`$’s$`\mathbf{)}`$ and $`a_I=t_I/d_t`$ for $`I`$ $`\mathbf{(}b_J=w_J/d_w`$ for $`J𝒥\mathbf{)}`$. Then $`a_I`$ $`\mathbf{(}b_J\mathbf{)}`$ is called the *reduced minor of the matrix $`T`$ ($`W`$) with respect to $`I`$ $`\mathbf{(}J𝒥\mathbf{)}`$*, the set $`\{a_I|I\}`$ $`\mathbf{(}\{b_J|J𝒥\}\mathbf{)}`$ *the family of reduced minors of $`T`$ $`\mathbf{(}W\mathbf{)}`$*.
It is known that the families of reduced minors of $`T`$ and of $`W`$ are identical modulo units (Lemma 5 of ).
Now, Corollary 2 of including its comments can be stated as follows:
###### Theorem 9.
(cf. Corollary 2 of Suppose that $`𝒜`$ is a unique factorization domain. A plant $`P^{m\times n}`$ is stabilizable if and only if the family of the reduced minors of $`T`$ (and also of $`W`$) generates $`𝒜`$.
The theorem above can be rewritten directly as follows.
###### Corollary 10.
Let $`t_I`$ and $`w_J`$ be as in Definition8. Then the following are equivalent:
1. A plant $`P^{m\times n}`$ is stabilizable.
2. The ideal $`_I(t_I)`$ is principal, or equivalently free as an $`𝒜`$-module.
3. The ideal $`_{J𝒥}(w_J)`$ is principal, or equivalently free as an $`𝒜`$-module.
## 4 Full-Size Minor Ideal
On the statements concerning the elementary factors and the reduced minors in Subsections 3.1 and 3.2, we have considered that the denominator matrices of the plant is expressed as $`dE_m`$ or $`dE_n`$ rather than general nonsingular matrices. This may be considered as a restriction on the expression of the plant. Thus we rather consider that $`P`$ is expressed as either $`P=ND^1`$ with $`N𝒜^{n\times m}`$ and $`D(𝒜)_m`$ or $`P=\stackrel{~}{D}^1\stackrel{~}{N}`$ with $`\stackrel{~}{N}𝒜^{n\times m}`$ and $`\stackrel{~}{D}(𝒜)_n`$. Now we redefine the matrices $`T`$, $`W`$ as $`T=\left[\begin{array}{cc}N^t& D^t\end{array}\right]^t`$ and $`W=\left[\begin{array}{cc}\stackrel{~}{N}& \stackrel{~}{D}\end{array}\right]`$. Further we consider that $`t_I`$’s and $`w_J`$’s are defined with the matrices $`T`$ and $`W`$ here. In the rest of this paper, we will use these notations unless otherwise stated.
We now introduce a notion to state the feedback stabilizability over commutative rings.
###### Definition 11.
(Full-Size Minor Ideals) The ideal generated by $`t_I`$’s for $`I`$ is called the *full-size minor ideal* of the plant $`P`$. We denote it by $`_I(t_I)`$ or simply $`𝔱`$.
We can also consider the ideal generated by $`w_J`$’s for $`J𝒥`$, denoted by $`_{J𝒥}(w_J)`$ or simply $`𝔴`$. The ideals $`𝔱`$ and $`𝔴`$ depend on the fractional representation of the plant $`P=ND^1=\stackrel{~}{D}^1\stackrel{~}{N}`$. However, this is *not* a problem from the following reason. To state the feedback stabilizability in terms of the full-size minor ideals, we will regard them as modules. Further, when these ideals are considered as modules, both the ideals $`𝔱`$ and $`𝔴`$ are uniquely determined as modules up to isomorphism with respect to any choice of fractions $`ND^1`$ and $`\stackrel{~}{N}^1\stackrel{~}{D}`$ of $`P`$ as shown below.
###### Lemma 12.
Let $`P`$ be in $`()^{n\times m}`$, where $``$ is one of $`𝒜`$, $`𝒜_f`$ with a nonzero $`f𝒜`$, and $`𝒜_\mathrm{p}`$ with a prime ideal $`\mathrm{p}`$ in $`Spec(𝒜)`$. For $`x=1,2`$ let $`N_x`$, $`D_x`$, $`\stackrel{~}{N}_x`$, $`\stackrel{~}{D}_x`$ be matrices over $``$ with $`P=N_xD_x^1=\stackrel{~}{D}_x^1\stackrel{~}{N}_x`$ over $`()`$, $`T_x=\left[\begin{array}{cc}N_x^t& D_x^t\end{array}\right]^t`$ and $`W_x=\left[\begin{array}{cc}\stackrel{~}{N}_x& \stackrel{~}{D}_x\end{array}\right]`$. Further for $`x=1,2`$ and for $`I`$, $`J𝒥`$, let $`t_{xI}=det(\mathrm{\Delta }_IT_x)`$, and $`w_{xJ}=det(\mathrm{\Delta }_JW_x^t)`$. Then the ideals $`_I(t_{1I})`$, $`_I(t_{2I})`$, $`_{J𝒥}(w_{1J})`$, and $`_{J𝒥}(w_{2J})`$ are isomorphic to one another as $``$-modules.
###### Proof.
We show first (i) $`_I(t_{1I})_I(t_{2I})`$ and then (ii) $`_I(t_{1I})_{J𝒥}(w_{1J})`$. The isomorphism $`_{J𝒥}(w_{1J})_{J𝒥}(w_{2J})`$ can be proved analogously to (i) and so is omitted.
(i). Observe that in the case where $`\left[\begin{array}{cc}N_2^t& D_2^t\end{array}\right]^t=\left[\begin{array}{cc}N_1^t& D_1^t\end{array}\right]^tX`$ holds with some nonsingular matrix $`X`$ over $``$, the statement of the lemma obviously holds. Hence by considering $`\left[\begin{array}{cc}N_2^t& D_2^t\end{array}\right]^tadj(D_2)`$ as $`\left[\begin{array}{cc}N_2^t& D_2^t\end{array}\right]^t`$, we can assume without loss of generality that $`D_2`$ is expressed as $`d_2E_m`$ with nonzero $`d_2`$. Observe now that $`\left[\begin{array}{cc}N_1^t& D_1^t\end{array}\right]^td_2=\left[\begin{array}{cc}N_2^t& d_2E_m\end{array}\right]D_1`$ holds. From this relation and the first observation, we now have (i).
(ii). It is sufficient to consider the case $`P=Nd^1`$ with $`N^{n\times m}`$ and $`d`$ as in (i). In the case $`P=ND^1`$, one can consider $`P=(Nadj(D))det(D)^1`$.
First we define a bijective mapping $`\tau `$ from $``$ to $`𝒥`$. For convenience we decompose $`I`$ into $`I_N`$ and $`I_d`$ as follows
$$I_N=\{i|in,iI\},I_d=\{i|i>n,iI\}.$$
Corresponding to $`I_N`$ and $`I_d`$, we define $`J_N`$ and $`J_d`$ as
$$J_N=[1,m]\backslash \{in|iI_d\},J_d=\{i+m|i[1,n]\backslash I_N\}.$$
We now define the mapping $`\tau :𝒥`$ as
$$\tau :I_NI_dJ_NJ_d.$$
Since $`I_N`$ and $`I_d`$ can be expressed by $`J_N`$ and $`J_d`$ as $`I_N=[1,n]\backslash \{jm|jJ_d\}`$, $`I_d=\{j+n|j[1,m]\backslash J_N\}`$, the inverse mapping $`\tau ^1:𝒥`$ can be defined naturally. Hence, the map $`\tau `$ is bijective.
Now let $`T=\left[\begin{array}{cc}N^t& dE_m\end{array}\right]^t`$ and $`W=\left[\begin{array}{cc}N& dE_n\end{array}\right]`$. By the straightforward calculation with noting that $`dE_m`$ and $`dE_n`$ are diagonal, we obtain the following relations:
$$det(\mathrm{\Delta }_IT)=\pm det(\mathrm{\Delta }_{\tau (I)}W^t)d^{mn}.$$
Thus $`t_{1I}=\pm w_{1\tau (I)}d^{mn}`$ for all $`I`$. It follows that the ideals $`_I(t_{1I})`$ and $`_{J𝒥}(w_{1J})`$ are isomorphic to each other. ∎
###### Note 13.
The reduced minors are derived from $`t_I`$’s and $`w_J`$’s in Definition8. Thus $`t_I`$’s and $`w_J`$’s can be considered more primitive than the reduced minors. Nevertheless since we will present in Theorem 15 that $`t_I`$’s and $`w_J`$’s (or the ideals $`𝔱`$ and $`𝔴`$ generated by them) have the capability to state feedback stabilizability over commutative rings, we here consider that the full-size minor ideal $`𝔱`$ (or the ideal $`𝔴`$) is a generalization of the reduced minors.
## 5 Feedback Stabilizability in terms of Full-Size Minor Ideal
In this section, we present the necessary and sufficient condition of the feedback stabilizability over commutative rings in terms of the full-size minor ideal.
Let us consider the case where the set $`𝒜`$ of the stable causal transfer functions is not a unique factorization domain. Then it is not sufficient to use the family of reduced minors in order to state the feedback stabilizability. To see this, let us consider the result given by Anantharam in <sup>1</sup><sup>1</sup>1The author wishes to thank to Dr. A. Quadrat (Centre d’Enseignement et de Recherche en Mathématiques, Informatique et Calcul Scientifique, ENPC, France) who introduced him to the paper of Anantharam..
###### Example 14.
In , Anantharam considered the case where $`[\sqrt{5}\mathrm{i}]`$ ($`[x]/(x^2+5)`$) is the set of stable causal transfer functions, where $``$ is the ring of integers and $`\mathrm{i}`$ the imaginary unit; that is, $`𝒜=[\sqrt{5}\mathrm{i}]`$. The set of all possible transfer functions is given as the field of fractions of $`𝒜`$; that is, $`=(\sqrt{5}\mathrm{i})`$. In this case we have multiple factorizations $`23=(1+\sqrt{5}\mathrm{i})(1\sqrt{5}\mathrm{i})`$ over $`𝒜`$, so that $`𝒜`$ is not a unique factorization domain. Anantharam in considered the single-input single-output case and showed that the plant $`p=(1+\sqrt{5}\mathrm{i})/2`$ does not have its coprime factorization over $`𝒜`$ but is stabilizable.
Now let $`T=\left[\begin{array}{cc}1+\sqrt{5}\mathrm{i}& 2\end{array}\right]^t`$. Since the plant $`p`$ is of the single-input single-output ($`m=n=1`$), we have $`=\{\{1\},\{2\}\}`$. Thus let $`I_1=\{1\}`$ and $`I_2=\{2\}`$ so that $`=\{I_1,I_2\}`$. The full-size minors of the matrix $`T`$ are $`t_{I_1}=det(\mathrm{\Delta }_{I_1}T)=1+\sqrt{5}\mathrm{i}`$ and $`t_{I_2}=det(\mathrm{\Delta }_{I_2}T)=2`$. If Theorem 9 (or equivalently Corollary 10) could be applied even over a general commutative ring, the ideal $`(t_{I_1},t_{I_2})`$ should be principal. However, the ideal $`(t_{I_1},t_{I_2})`$ is not principal since $`p`$ does not have its coprime factorization.
In order to involve even such an example as a system over commutative ring, we extend Theorem 9. Since we cannot use the reduced minors to state the feedback stabilizability in general, we alternatively employ the full-size minor ideal $`𝔱`$ rather than the reduced minors. The extension is the first main result of this paper and stated as follows.
###### Theorem 15.
Let $`P`$ be a causal plant of $`𝒫^{n\times m}`$. Then the plant $`P`$ is stabilizable if and only if the full-size minor ideal $`𝔱`$ of the plant $`P`$ is projective. Further when $`𝔱`$ is projective, it is of rank $`1`$.
By virtue of Lemma 12, the above theorem can be also stated with the ideal $`𝔴`$ instead of the full-size minor ideal $`𝔱`$.
In the case where $`𝒜`$ is a unique factorization domain, as in Theorem 9, the condition of feedback stabilizability is that the full-size minor ideal is free. On the other hand, in Theorem 15, the condition is that the ideal is projective. They are equivalent to each other in the case where $`𝒜`$ is a unique factorization domain as follows.
###### Proposition 16.
Let $``$ be a unique factorization domain. Then the ideal generated by finite elements of $``$ is projective if and only if it is free.
This proof will be given after finishing the proof of Theorem 15.
Now that we have presented the statement of Theorem 15, the main objective of the remainder of this section is to carry out the proof of Theorem 15. To do so, we prepare two main intermediate results. The first one is about the existence of right-/left-coprime factorizations of stabilizable plants over local rings, which will be presented in Subsection 5.1. The other is about the local-global principle of the feedback stabilizability, which will be presented in Subsection 5.2. Then we will prove Theorem 15. After the proof of Theorem 15 we will prove Proposition 16. Before finishing this section, we will present the relationship among the full-size minor ideals of $`P`$, $`C`$, and $`H(P,C)`$.
### 5.1 Right-/Left-Coprime Factorizations over Local Rings
The following is the first intermediate result of Theorem 15 about the existence of right-/left-coprime factorizations of stabilizable plants over local rings.
###### Proposition 17.
Let $`P`$ be a plant in $`^{n\times m}`$. Suppose that $``$ is $`𝒜_\mathrm{p}`$ with a prime ideal $`\mathrm{p}`$ in $`Spec(𝒜)`$. Then the following statements are equivalent:
* The plant $`P`$ is $``$-stabilizable.
* There exists a right-coprime factorization over $``$ of $`P`$.
* There exists a left-coprime factorization over $``$ of $`P`$.
The proof of this proposition will be presented after giving several its intermediate results.
We here recall the notion of Hermite used in <sup>2</sup><sup>2</sup>2It should be noted that this definition of “Hermite” is different from . , which can characterize the existence of both right-/left-coprime factorizations of transfer matrices.
###### Definition 18.
(\[20, p.345\] Let $``$ be a commutative ring and $`A`$ a matrix over $``$ of size $`x\times y`$ with $`x<y`$. Then we say that the matrix $`A`$ can be *complemented* if there exists a unimodular matrix in $`()_y`$ containing the matrix $`A`$ as a submatrix. A row $`\left[\begin{array}{ccc}a_1& \mathrm{}& a_y\end{array}\right]^{1\times y}`$ is said to be a *unimodular row* if $`a_1,\mathrm{},a_y`$ together generate $``$. A commutative ring $``$ is said to be *Hermite* if every unimodular row can be complemented.
The following result was given in provided that $``$ is an integral domain.
###### Theorem 19.
(cf. Theorem 8.1.66 of Let $``$ be a commutative ring. The following three statements are equivalent:
* The commutative ring $``$ is Hermite.
* If a matrix over $`()`$ has a right-coprime factorization over $``$, it has also a left-coprime factorization over $``$.
* If a matrix over $`()`$ has a left-coprime factorization over $``$, it has also a right-coprime factorization over $``$.
The “integral domain” version of this theorem was given as Theorem 8.1.66 of . Even in the case of commutative rings, the proof is similar with that of Theorem 8.1.66 of and so is omitted.
The following result is the intermediate result of Proposition 17, which makes the result above applicable to the proof of the proposition.
###### Lemma 20.
Any local ring is Hermite.
###### Proof.
Suppose that $``$ is a local ring and $`\left[\begin{array}{c}a_1,\mathrm{},a_y\end{array}\right]^{1\times y}`$ is a unimodular row. Thus there exist $`b_1,\mathrm{},b_y`$ such that
$$a_1b_1+\mathrm{}+a_yb_y=1.$$
(3)
Since $``$ is local, the set of all nonunits is an ideal. From (3), there exists an $`i`$ with $`1iy`$ such that $`a_i`$ is a unit. We assume without loss of generality that $`a_1`$ is a unit. If $`y=1`$, then $`a_1`$ is a unit, which can be considered as a unimodular matrix of $`()_1`$. In the following we consider the case $`y>1`$. Then we can construct a unimodular matrix $`U=(u_{ij})()_y`$:
uij={
ajif =i1,a1-1if i=j=2,1if i=j>2,0otherwise. u_{ij}=\left\{\mbox{\begin{tabular}[]{ll}$a_{j}$&if $i=1$,\\
$a_{1}^{-1}$&if $i=j=2$,\\
$1$&if $i=j>2$,\\
$0$&otherwise.\end{tabular}
}\right.
This $`U`$ contains the row $`\left[\begin{array}{c}a_1,\mathrm{},a_y\end{array}\right]`$ as a submatrix and hence every unimodular row can be complemented. Therefore $``$ is Hermite. ∎
We prepare one more result which will help us present a nonsingular denominator matrix of a stabilizing controller
###### Lemma 21.
Let $``$ be a commutative ring and $`\mathrm{p}`$ a prime ideal of $``$. Suppose that there exist matrices $`A`$, $`B`$, $`C_1`$, $`C_2`$ over $``$ such that the determinant of the following square matrix is in $`\backslash \mathrm{p}`$:
$$\left[\begin{array}{cc}A& C_1\\ B& C_2\end{array}\right],$$
(4)
where the matrix $`A`$ is square and the matrices $`A`$ and $`B`$ have same number of columns. Then there exists a matrix $`R`$ over $``$ such that the determinant of the matrix $`A+RB`$ is in $`\backslash \mathrm{p}`$.
Before starting the proof, it is worth reviewing some easy facts about a prime ideal.
###### Remark 5.22.
Suppose that $`\mathrm{p}`$ is a prime ideal of $``$. (i) If $`a`$ is in $`\backslash \mathrm{p}`$ and expressed as $`a=b+c`$ with $`b,c`$, then at least one of $`b`$ and $`c`$ is in $`\backslash \mathrm{p}`$. (ii) If $`a`$ is in $`\backslash \mathrm{p}`$ and $`b`$ in $`\mathrm{p}`$, then the sum $`a+b`$ is in $`\backslash \mathrm{p}`$. (iii) Every factor in $``$ of an element of $`\backslash \mathrm{p}`$ belongs to $`\backslash \mathrm{p}`$ (that is, if $`a,b`$ and $`ab\backslash \mathrm{p}`$, then $`a,b\backslash \mathrm{p}`$).
Proof of Lemma 21. This proof mainly follows that of Lemma 4.4.21 of .
If $`det(A)`$ is in $`\backslash \mathrm{p}`$, then we can select the zero matrix as $`R`$. Thus we assume in the following that $`det(A)`$ is in $`\mathrm{p}`$.
Since the determinant of (4) is in $`\backslash \mathrm{p}`$, there exists a full-size minor of $`\left[\begin{array}{cc}A^t& B^t\end{array}\right]^t`$ in $`\backslash \mathrm{p}`$ by Laplace’s expansion of (4) and by Remark 5.22(i,iii). Let $`a`$ be such a full-size minor of $`\left[\begin{array}{cc}A^t& B^t\end{array}\right]^t`$ having as few rows from $`B`$ as possible.
We here construct a matrix $`R`$ such that $`det(A+RB)=\pm a+z`$ with a $`z\mathrm{p}`$. Since $`det(A)\mathrm{p}`$, the full-size minor $`a`$ must contain at least one row of $`B`$ from the matrix $`\left[\begin{array}{cc}A^t& B^t\end{array}\right]^t`$. Suppose that $`a`$ is obtained by excluding the rows $`i_1,\mathrm{},i_k`$ of $`A`$ and including the rows $`j_1,\mathrm{},j_k`$ of $`B`$. Now define $`R=(r_{ij})`$ by $`r_{i_1j_1}=\mathrm{}=r_{i_kj_k}=1`$ and $`r_{ij}=0`$ for all other $`i`$$`j`$. Observe that $`det(A+RB)`$ is expanded in terms of full-size minors of the matrices $`\left[\begin{array}{cc}E& R\end{array}\right]`$ and $`\left[\begin{array}{cc}A^t& B^t\end{array}\right]^t`$ from the factorization $`A+RB=\left[\begin{array}{cc}E& R\end{array}\right]\left[\begin{array}{cc}A^t& B^t\end{array}\right]^t`$ by the Binet-Cauchy formula. Every minor of $`\left[\begin{array}{cc}E& R\end{array}\right]`$ containing more than $`k`$ columns of $`R`$ is zero. By the method of choosing the rows from $`\left[\begin{array}{cc}A^t& B^t\end{array}\right]^t`$ for the full-size minor $`a`$, every full-size minor of $`\left[\begin{array}{cc}A^t& B^t\end{array}\right]^t`$ having less than $`k`$ rows of $`B`$ is in $`\mathrm{p}`$. There is only one nonzero minor of $`\left[\begin{array}{cc}E& R\end{array}\right]`$ containing exactly $`k`$ columns of $`R`$, which is obtained by excluding the columns $`i_1,\mathrm{},i_k`$ of the identity matrix $`E`$ and including the columns $`j_1,\mathrm{},j_k`$ of $`R`$; it is equal to $`\pm 1`$. From the Binet-Cauchy formula the corresponding minor of $`\left[\begin{array}{cc}A^t& B^t\end{array}\right]^t`$ is $`a`$. As a result, $`det(A+RB)`$ is given as a sum of $`\pm a`$ and elements in $`\mathrm{p}`$. By Remark 5.22(ii), the sum is in $`\backslash \mathrm{p}`$ and so is $`det(A+RB)`$.
Now that we have the result above, we can prove Proposition 17.
Proof of Proposition 17. Since $``$ is local, (ii) and (iii) are equivalent by Theorem 19 and Lemma 20. Thus we only prove (i)$``$(ii) and *vice versa*.
(i)$``$(ii). Suppose that $`P`$ is $``$-stabilizable. Then the $``$-module $`𝒯_{P,}`$ is projective by Proposition 2.1 of . Further it is free by Corollary 3.5 of \[7, Ch.IV\]. Let $`N`$ and $`D`$ be matrices over $``$ with $`P=ND^1()`$. Then the $``$-module $`M_r(\left[\begin{array}{cc}N^t& D^t\end{array}\right]^t)`$ is free of rank $`m`$ since $`D`$ is nonsingular over $``$. Let $`v_1,v_2,\mathrm{},v_m^m`$ be a basis of the module $`M_r(\left[\begin{array}{cc}N^t& D^t\end{array}\right]^t)`$ and $`V`$ the matrix of $`()_m`$ whose rows are $`v_1,v_2,\mathrm{},v_m`$. Then, the matrix $`\left[\begin{array}{cc}N^t& D^t\end{array}\right]^t`$ can be written in the form $`\left[\begin{array}{cc}N^t& D^t\end{array}\right]^t=\left[\begin{array}{cc}N_0^t& D_0^t\end{array}\right]^tV`$ by uniquely choosing the matrices $`N_0`$ in $`^{n\times m}`$ and $`D_0`$ in $`()_m`$. Because of $`det(D)=det(D_0V)`$, $`det(D_0)`$ is a nonzerodivisor. It follows that $`P=N_0D_0^1`$ over $`()`$. In the following we show that the matrices $`N_0`$ and $`D_0`$ are right-coprime over $``$. Since $`v_1,\mathrm{},v_m`$ belong to $`M_r(\left[\begin{array}{cc}N^t& D^t\end{array}\right]^t)`$, there exist matrices $`\stackrel{~}{Y}`$ in $`^{m\times n}`$ and $`\stackrel{~}{X}`$ in $`()_m`$ such that $`V=\left[\begin{array}{cc}\stackrel{~}{Y}& \stackrel{~}{X}\end{array}\right]\left[\begin{array}{cc}N^t& D^t\end{array}\right]^t`$. So we have $`V=(\stackrel{~}{Y}N_0+\stackrel{~}{X}D_0)V`$. Since $`V`$ is nonsingular, we obtain $`\stackrel{~}{Y}N_0+\stackrel{~}{X}D_0=E_m`$ over $``$. Thus $`(N_0,D_0)`$ is a right-coprime factorization over $``$ of $`P`$.
(ii)$``$(i). Suppose that there exists a right-coprime factorization over $``$ of the plant $`P`$; that is, there exist the matrices $`N`$, $`D`$, $`\stackrel{~}{Y}`$, $`\stackrel{~}{X}`$ over $``$ with $`\stackrel{~}{Y}N+\stackrel{~}{X}D=E_m`$ and $`P=ND^1`$. If $`det(\stackrel{~}{X})`$ is a nonzerodivisor of $``$, it is obvious that $`\stackrel{~}{X}^1\stackrel{~}{Y}`$ is an $``$-stabilizing controller. Thus in the following we suppose that $`det(\stackrel{~}{X})`$ is a zerodivisor of $``$.
By the equivalence between (ii) and (iii), there also exists a left-coprime factorization over $``$ of $`P`$; that is, there exist the matrices $`\stackrel{~}{N}`$, $`\stackrel{~}{D}`$, $`Y`$, $`X`$ over $``$ with $`\stackrel{~}{N}Y+\stackrel{~}{D}X=E_n`$ and $`P=\stackrel{~}{D}^1\stackrel{~}{N}`$. Thus we have the following matrix equation:
$$\left[\begin{array}{cc}\stackrel{~}{X}& \stackrel{~}{Y}\\ \stackrel{~}{N}& \stackrel{~}{D}\end{array}\right]\left[\begin{array}{cc}D& Y\\ N& X\end{array}\right]=\left[\begin{array}{cc}E_m& \stackrel{~}{X}Y\stackrel{~}{Y}X\\ O& E_n\end{array}\right].$$
(5)
Observe that the determinant of the right-hand side of the matrix equation above is in $`\backslash 𝒵_\mathrm{p}`$, where $`𝒵_\mathrm{p}`$ denotes the localization of the prime ideal $`𝒵`$ at $`\mathrm{p}`$ (Note that $`𝒵_\mathrm{p}`$ is also a prime ideal of $``$). Hence the determinant of the first matrix in (5) is in $`\backslash 𝒵_\mathrm{p}`$ again. Applying Lemma 21 to the first matrix, we have a matrix $`R`$ over $``$ such that the determinant of the matrix $`\stackrel{~}{X}+R\stackrel{~}{N}`$ is in $`\backslash 𝒵_\mathrm{p}`$. Now $`(\stackrel{~}{X}+R\stackrel{~}{N})^1(\stackrel{~}{Y}R\stackrel{~}{D})`$ is an $``$-stabilizing controller.
### 5.2 Local-Global Principle in Stabilizability
Next we present the local-global principle below about the feedback stabilizability as the second intermediate result of this section.
###### Proposition 5.23.
Suppose that the plant $`P`$ is causal. Then the following statements are equivalent:
1. $`P`$ is stabilizable.
2. $`P`$ is $`𝒜_\mathrm{p}`$-stabilizable for each prime ideal $`\mathrm{p}`$ in $`Spec(𝒜)`$.
3. $`P`$ is $`𝒜_\mathrm{m}`$-stabilizable for each maximal ideal $`\mathrm{m}`$ in $`Max(𝒜)`$.
4. For every prime ideal $`\mathrm{p}`$ in $`Spec(𝒜)`$, $`P`$ has either its right- or left-coprime factorization over $`𝒜_\mathrm{p}`$.
5. For every maximal ideal $`\mathrm{m}`$ in $`Max(𝒜)`$, $`P`$ has either its right- or left-coprime factorization over $`𝒜_\mathrm{m}`$.
Further, if $`P`$ is stabilizable, then there exists a causal stabilizing controller of $`P`$.
Note here that by virtue of Proposition 17, if (iv) holds $`\mathbf{(}`$if (v) holds$`\mathbf{)}`$, then the plant $`P`$ has both right-/left-coprime factorizations over $`𝒜_\mathrm{p}`$ $`\mathbf{(}`$over $`𝒜_\mathrm{m}\mathbf{)}`$.
We consider that this is a generalization of Proposition 2 of in which the strict causality of the plant is assumed (see for the definition of the strict causality). On the other hand, we assume only that the plant is causal.
Now we begin to prove Proposition 5.23.
Proof of Proposition 5.23. Since the following implications are obvious:
by virtue of Proposition 17, we only show that (v) implies (i).
Suppose that (v) holds. Let $`N`$, $`D`$, $`\stackrel{~}{N}`$, and $`\stackrel{~}{D}`$ be matrices over $`𝒜`$ with $`P=ND^1=\stackrel{~}{D}^1\stackrel{~}{N}`$ such that $`D`$ and $`\stackrel{~}{D}`$ are $`𝒵`$-nonsingular (recall that $`P`$ is causal). By Proposition 17, $`P`$ has both right-/left-coprime factorizations over $`𝒜_\mathrm{m}`$ with $`\mathrm{m}Max(𝒜)`$. As in the proof of Proposition 17, for each $`\mathrm{m}`$ in $`Max(𝒜)`$, there exist matrices $`Y_\mathrm{m}`$, $`X_\mathrm{m}`$, $`\stackrel{~}{Y}_\mathrm{m}`$, $`\stackrel{~}{X}_\mathrm{m}`$, $`N_\mathrm{m}`$, $`D_\mathrm{m}`$, $`\stackrel{~}{N}_\mathrm{m}`$, $`\stackrel{~}{D}_\mathrm{m}`$, $`V_\mathrm{m}`$, and $`W_\mathrm{m}`$ over $`𝒜_\mathrm{m}`$ such that
$`\left[\begin{array}{c}N\\ D\end{array}\right]=\left[\begin{array}{c}N_\mathrm{m}\\ D_\mathrm{m}\end{array}\right]V_\mathrm{m},\left[\begin{array}{cc}\stackrel{~}{N}& \stackrel{~}{D}\end{array}\right]=W_\mathrm{m}\left[\begin{array}{cc}\stackrel{~}{N}_\mathrm{m}& \stackrel{~}{D}_\mathrm{m}\end{array}\right],`$ (6)
$`\stackrel{~}{Y}_\mathrm{m}N_\mathrm{m}+\stackrel{~}{X}_\mathrm{m}D_\mathrm{m}=E_m,\stackrel{~}{N}_\mathrm{m}Y_\mathrm{m}+\stackrel{~}{D}_\mathrm{m}X_\mathrm{m}=E_n`$ (7)
hold over $`𝒜_\mathrm{m}`$. For each $`\mathrm{m}Max(𝒜)`$ let $`q_\mathrm{m}`$ be an arbitrary but fixed element of $`𝒜\backslash \mathrm{m}`$ such that the six matrices $`q_\mathrm{m}N_\mathrm{m}\stackrel{~}{Y}_\mathrm{m}`$, $`q_\mathrm{m}N_\mathrm{m}\stackrel{~}{X}_\mathrm{m}`$, $`q_\mathrm{m}D_\mathrm{m}\stackrel{~}{Y}_\mathrm{m}`$, $`q_\mathrm{m}D_\mathrm{m}\stackrel{~}{X}_\mathrm{m}`$, $`q_\mathrm{m}\stackrel{~}{D}_\mathrm{m}`$, and $`q_\mathrm{m}\stackrel{~}{N}_\mathrm{m}`$ are over $`𝒜`$.
For a subset $``$ of $`𝒜`$, denote by $`\mathrm{\Gamma }()`$ the set of all maximal ideals $`\mathrm{m}`$ of $`𝒜`$ with $`\mathrm{m}`$, that is, $`\mathrm{\Gamma }()=\{\mathrm{m}Max(𝒜)|\mathrm{m}\}`$. Since $`q_\mathrm{m}𝒜\backslash \mathrm{m}`$, we have $`\mathrm{m}\mathrm{\Gamma }(𝒜q_\mathrm{m})`$. Thus $`Max(𝒜)=_{\mathrm{m}Max(𝒜)}\mathrm{\Gamma }(𝒜q_\mathrm{m})`$. Recall that $`Max(𝒜)`$ is compact (see Theorem IV.1 of ). Hence there are a finite number of $`\mathrm{m}_1,\mathrm{},\mathrm{m}_t`$ of maximal ideals such that $`Max(𝒜)=_{i=1}^t\mathrm{\Gamma }(𝒜q_{\mathrm{m}_i})`$. It follows that $`Max(𝒜)=\mathrm{\Gamma }(_{i=1}^t𝒜q_{\mathrm{m}_i})`$ and, consequently, $`𝒜=_{i=1}^t𝒜q_{\mathrm{m}_i}`$. Therefore there exist $`r_1,\mathrm{},r_t`$ in $`𝒜`$ with $`1=r_1q_{\mathrm{m}_1}+\mathrm{}+r_tq_{\mathrm{m}_t}`$.
Next we want to consider that at least one of $`\mathrm{m}_1,\mathrm{},\mathrm{m}_t`$ contains $`𝒵`$. In the case where every $`\mathrm{m}_i`$ in $`\mathrm{m}_1,\mathrm{},\mathrm{m}_t`$ does not contain $`𝒵`$, we reconstruct $`t`$, $`r_i`$’s, and $`q_{\mathrm{m}_i}`$’s as follows. We first pick an $`\mathrm{m}_{t+1}Max(𝒜)`$ with $`\mathrm{m}_{t+1}𝒵`$. Then we let $`r_i`$ be $`(1q_{\mathrm{m}_{t+1}})r_i`$ for $`1it`$ and $`r_{t+1}=1`$. We now let $`t:=t+1`$. Then we have again $`1=r_1q_{\mathrm{m}_1}+\mathrm{}+r_tq_{\mathrm{m}_t}`$ and, in this case, $`\mathrm{m}_t𝒵`$. Hence we can assume without loss of generality that at least one of $`\mathrm{m}_1,\mathrm{},\mathrm{m}_t`$, say $`\mathrm{m}_1`$, contains $`𝒵`$.
Observe then that the following equality holds:
$`\begin{array}{c}\hfill 1=(r_1q_{\mathrm{m}_1}+r_11)q_{\mathrm{m}_1}+(r_2q_{\mathrm{m}_1}+r_2)q_{\mathrm{m}_2}\\ \hfill +\mathrm{}+(r_tq_{\mathrm{m}_1}+r_t)q_{\mathrm{m}_t}.\end{array}`$ (8)
At least one of $`r_1q_{\mathrm{m}_1}+r_11`$ and $`r_1`$ must be in $`𝒜\backslash 𝒵`$. Thus in the case $`r_1𝒵`$, we can reassign $`r_i`$’s as in (8), so that $`r_1`$ is in $`𝒜\backslash 𝒵`$. Therefore we can assume without loss of generality that $`r_1q_{\mathrm{m}_1}𝒜\backslash 𝒵`$.
Consider here the following matrix
$`\left[\begin{array}{cc}E_n_{i=1}^tr_iq_{\mathrm{m}_i}N_{\mathrm{m}_i}\stackrel{~}{Y}_{\mathrm{m}_i}& _{i=1}^tr_iq_{\mathrm{m}_i}N_{\mathrm{m}_i}\stackrel{~}{X}_{\mathrm{m}_i}\\ _{i=1}^tr_iq_{\mathrm{m}_i}D_{\mathrm{m}_i}\stackrel{~}{Y}_{\mathrm{m}_i}& _{i=1}^tr_iq_{\mathrm{m}_i}D_{\mathrm{m}_i}\stackrel{~}{X}_{\mathrm{m}_i}\end{array}\right],`$ (9)
which is over $`𝒜`$. For short we partition (9) as
$$\left[\begin{array}{cc}H_{11}& H_{12}\\ H_{21}& H_{22}\end{array}\right].$$
In the case where $`H_{22}`$ is $`𝒵`$-nonsingular, letting $`C=H_{22}^1H_{21}𝒫^{m\times n}`$ we can check that $`H(P,C)`$ is equal to (9), which implies that $`P`$ is stabilized by $`C`$. Hence in the rest of this proof we show that if $`H_{22}`$ is $`𝒵`$-singular, then $`H_{22}`$ can be made $`𝒵`$-nonsingular by reassigning $`\stackrel{~}{X}_{\mathrm{m}_i}`$ and $`\stackrel{~}{Y}_{\mathrm{m}_i}`$ for an $`i`$.
First we show the $`𝒵`$-nonsingularity of the matrices $`r_1q_{\mathrm{m}_1}D_{\mathrm{m}_1}`$ and $`r_1q_{\mathrm{m}_1}\stackrel{~}{D}_{\mathrm{m}_1}`$. Since $`r_1q_{\mathrm{m}_1}𝒜\backslash 𝒵`$, we have $`det(r_1q_{\mathrm{m}_1}D)𝒜\backslash 𝒵`$. From the first matrix equation of (6), we have $`det(r_1q_{\mathrm{m}_1}D)=det(r_1q_{\mathrm{m}_1}D_{\mathrm{m}_1})det(V_{\mathrm{m}_1})`$. Hence $`r_1q_{\mathrm{m}_1}D_{\mathrm{m}_1}`$ is $`𝒵`$-nonsingular. Analogously, from the second matrix equation of (6), $`r_1q_{\mathrm{m}_1}\stackrel{~}{D}_{\mathrm{m}_1}`$ is $`𝒵`$-nonsingular.
Next consider the following matrix equation over $`𝒜`$:
$`\left[\begin{array}{cc}_{i=1}^tr_iq_{\mathrm{m}_i}D_{\mathrm{m}_i}\stackrel{~}{X}_{\mathrm{m}_i}& _{i=1}^tr_iq_{\mathrm{m}_i}D_{\mathrm{m}_i}\stackrel{~}{Y}_{\mathrm{m}_i}\\ r_1q_{\mathrm{m}_1}det(r_1q_{\mathrm{m}_1}D_{\mathrm{m}_1})\stackrel{~}{N}_{\mathrm{m}_1}& r_1q_{\mathrm{m}_1}det(r_1q_{\mathrm{m}_1}D_{\mathrm{m}_1})\stackrel{~}{D}_{\mathrm{m}_1}\end{array}\right]\times `$
$`\left[\begin{array}{cc}D& O\\ N& E_n\end{array}\right]=\left[\begin{array}{cc}D& _{i=1}^tr_iq_{\mathrm{m}_i}D_{\mathrm{m}_i}\stackrel{~}{Y}_{\mathrm{m}_i}\\ O& r_1q_{\mathrm{m}_1}det(r_1q_{\mathrm{m}_1}D_{\mathrm{m}_1})\stackrel{~}{D}_{\mathrm{m}_1}\end{array}\right].`$ (10)
Since the matrices $`D`$, $`r_1q_{\mathrm{m}_1}D_{\mathrm{m}_1}`$, and $`r_1q_{\mathrm{m}_1}\stackrel{~}{D}_{\mathrm{m}_1}`$ are $`𝒵`$-nonsingular, so is the right-hand side of (10). Thus the first matrix of (10) is also $`𝒵`$-nonsingular. By Lemma 21 and the first matrix of (10), there exists a matrix $`R_{\mathrm{m}_1}^{}`$ of $`𝒜^{m\times n}`$ such that the following matrix is $`𝒵`$-nonsingular:
$$\underset{i=1}{\overset{t}{}}r_iq_{\mathrm{m}_i}D_{\mathrm{m}_i}\stackrel{~}{X}_{\mathrm{m}_i}r_1q_{\mathrm{m}_1}det(r_1q_{\mathrm{m}_1}D_{\mathrm{m}_1})R_{\mathrm{m}_1}^{}\stackrel{~}{N}_{\mathrm{m}_1}.$$
Now let $`R_{\mathrm{m}_1}`$ be $`r_1q_{\mathrm{m}_1}adj(r_1q_{\mathrm{m}_1}D_{\mathrm{m}_1})R_{\mathrm{m}_1}^{}`$. Further we let $`\stackrel{~}{X}_{\mathrm{m}_1}`$ be the matrix $`\stackrel{~}{X}_{\mathrm{m}_1}R_{\mathrm{m}_1}\stackrel{~}{N}_{\mathrm{m}_1}`$ and $`\stackrel{~}{Y}_{\mathrm{m}_1}`$ the matrix $`\stackrel{~}{Y}_{\mathrm{m}_1}+R_{\mathrm{m}_1}\stackrel{~}{D}_{\mathrm{m}_1}`$, which are consistent with (7). Thus we can now consider without loss of generality that the matrix $`_{i=1}^tr_iq_{\mathrm{m}_i}D_{\mathrm{m}_i}\stackrel{~}{X}_{\mathrm{m}_i}`$ is $`𝒵`$-nonsingular and so is $`H_{22}`$.
### 5.3 Proof of Theorem 15
Before proving Theorem 15, we should prepare a small result.
###### Lemma 5.24.
Let $`a𝒜`$ and $`\mathrm{p}Spec(𝒜)`$. Then $`(a)_\mathrm{p}`$ and $`(a/1)`$ are isomorphic to each other as $`𝒜_\mathrm{p}`$-modules, where $`(a)_\mathrm{p}`$ denotes the localization, at $`\mathrm{p}`$, of the principal ideal generated by $`a𝒜`$ and $`(a/1)`$ the principal ideal generated by $`a/1𝒜_\mathrm{p}`$.
The proof of the lemma is elementary and is omitted.
Now we start to prove the first result of this paper.
Proof of Theorem 15. We show first the “Only If” part and then the “If” part.
(Only If). Suppose that $`P`$ is stabilizable. Then by Proposition 5.23, for every prime ideal $`\mathrm{p}`$ in $`Spec(𝒜)`$, $`P`$ is $`𝒜_\mathrm{p}`$-stabilizable. By Proposition 17, $`P`$ has both its right-/left-coprime factorizations over $`𝒜_\mathrm{p}`$. Suppose that $`\stackrel{~}{Y}_\mathrm{p}N_\mathrm{p}+\stackrel{~}{X}_\mathrm{p}D_\mathrm{p}=E_m`$ holds over $`𝒜_\mathrm{p}`$ with $`P=N_\mathrm{p}D_\mathrm{p}^1`$, where the matrices $`N_\mathrm{p}`$, $`D_\mathrm{p}`$, $`\stackrel{~}{Y}_\mathrm{p}`$, and $`\stackrel{~}{X}_\mathrm{p}`$ are over $`𝒜_\mathrm{p}`$. Then let $`T_\mathrm{p}=\left[\begin{array}{cc}N_\mathrm{p}^t& D_\mathrm{p}^t\end{array}\right]^t`$. By Binet-Cauchy formula we have $`_I(det(\mathrm{\Delta }_IT_\mathrm{p}))=𝒜_\mathrm{p}`$. Thus by virtue of Lemmas 12 and 5.24, the ideal $`𝔱_\mathrm{p}`$ is free (recall that $`𝔱_\mathrm{p}`$ denotes the localization of the full-size minor ideal $`𝔱`$ at $`\mathrm{p}`$), which is also finitely generated. This holds for every prime ideal $`\mathrm{p}`$. From Theorem IV.32 of , the full-size minor ideal $`𝔱`$ is projective.
(If). Suppose that the full-size minor ideal $`𝔱`$ is projective. Let $`\mathrm{p}`$ be a prime ideal in $`Spec(𝒜)`$. Then $`𝔱_\mathrm{p}`$ is free by Theorem IV.32 of again. Thus there exist $`g`$, $`a_I`$, and $`r_I`$ in $`𝒜_\mathrm{p}`$ with $`g=_Ir_It_I`$ and $`t_I=a_Ig`$ for every $`I`$. Since $`g=_Ir_Ia_Ig`$ and $`g`$ is a nonzerodivisor, we have $`_Ir_Ia_I=1`$. Recall here that $`𝒜_\mathrm{p}`$ is local. Hence the set of all nonunits in $`𝒜_\mathrm{p}`$ is an ideal. Thus there exists $`I_0`$ such that $`r_{I_0}a_{I_0}`$ is a unit of $`𝒜_\mathrm{p}`$. This implies that $`a_{I_0}`$ is a unit of $`𝒜_\mathrm{p}`$ and further that every $`t_I`$ has a factor $`t_{I_0}`$ over $`𝒜_\mathrm{p}`$ (that is, $`t_{I_0}`$ and $`g`$ are associate). Now let $`T^{}=Tadj(\mathrm{\Delta }_{I_0}T)`$ and $`t_I^{}=det(\mathrm{\Delta }_IT^{})`$ for every $`I`$. Then $`t_I^{}=t_Idet(adj(\mathrm{\Delta }_{I_0}T))`$ and $`\mathrm{\Delta }_{I_0}T^{}=t_{I_0}E_m`$ hold. Since $`det(adj(\mathrm{\Delta }_{I_0}T))=t_{I_0}^{m1}`$, every $`t_I^{}`$ has a common factor $`t_{I_0}^m`$.
Suppose that $`i`$ is an integer with $`iI_0`$ and $`1im+n`$. Suppose further that $`i_{01},\mathrm{},i_{0m}`$ are elements in $`I_0`$ with ascending order. Now let $`I=\{i,i_{01},i_{02},\mathrm{},i_{0k1},`$ $`i_{0k+1},\mathrm{},i_{0m}\}`$. Then $`t_I`$ is expressed as $`\pm t_{ik}t_{I_0}^{m1}`$ where $`t_{ik}`$ is the $`(i,k)`$-entry of the matrix $`T^{}`$. Since $`t_I^{}`$ has a factor $`t_{I_0}^m`$, $`t_{ik}`$ has a factor $`t_{I_0}`$. This fact holds for all $`i`$ between $`1im+n`$ but $`iI_0`$. As a result, $`t_{I_0}`$ is a common factor of all entries of $`T^{}`$.
Let $`T^{\prime \prime }=T^{}/t_{I_0}`$ over $`𝒜_\mathrm{p}`$. Since $`\mathrm{\Delta }_{I_0}T^{\prime \prime }`$ is the identity matrix, the matrix $`\mathrm{\Delta }_{I_0}`$ itself is a left inverse of $`T^{\prime \prime }`$. Let $`\stackrel{~}{Y}_{I_0}`$ and $`\stackrel{~}{X}_{I_0}`$ be matrices with $`\left[\begin{array}{cc}\stackrel{~}{Y}_{I_0}& \stackrel{~}{X}_{I_0}\end{array}\right]=\mathrm{\Delta }_{I_0}`$. Further we let $`N_{I_0}`$ and $`D_{I_0}`$ be matrices over $`𝒜_\mathrm{p}`$ with $`T^{\prime \prime }=\left[\begin{array}{cc}N_{I_0}^t& D_{I_0}^t\end{array}\right]^t`$. Then we obtain $`\stackrel{~}{Y}_{I_0}N_{I_0}+\stackrel{~}{X}_{I_0}D_{I_0}=E_m`$ over $`𝒜_\mathrm{p}`$, which is a right-coprime factorization over $`𝒜_\mathrm{p}`$ of the plant $`P`$. Therefore by Proposition 5.23, $`P`$ is stabilizable.
### 5.4 Proof of Proposition 16
Now we prove Proposition 16. We first prepare the following local-global principle on ideals.
###### Lemma 5.25.
Let $``$ be a commutative ring. Let $`𝔞_1,\mathrm{},𝔞_k`$ be ideals of $``$. Then the following statements are equivalent:
* $`𝔞_1+\mathrm{}+𝔞_k=`$.
* $`𝔞_{1\mathrm{p}}+\mathrm{}+𝔞_{k\mathrm{p}}=_\mathrm{p}`$ for all prime ideal $`\mathrm{p}Spec(𝒜)`$.
* $`𝔞_{1\mathrm{m}}+\mathrm{}+𝔞_{k\mathrm{m}}=_\mathrm{m}`$ for all maximal ideal $`\mathrm{m}Max(𝒜)`$.
###### Proof 5.26.
It is obvious that (i) implies (ii) and (ii) implies (iii). Hence we only show that (iii) implies (i).
(iii)$``$(i). Suppose that (iii) holds. Let $`\mathrm{m}`$ be a maximal ideal of $`𝒜`$. Since $`_\mathrm{m}`$ is local, the set of all nonunits in $`_\mathrm{m}`$ is an ideal. Hence there exists an $`i_\mathrm{m}`$ with $`1i_\mathrm{m}k`$ such that $`𝔞_{i_\mathrm{m}\mathrm{m}}=_\mathrm{m}`$. Thus there exists $`s_\mathrm{m}`$ in $`\backslash \mathrm{m}`$ such that $`s_\mathrm{m}𝔞_{i_\mathrm{m}}`$.
Recalling the proof of Proposition 5.23, we have a finite number of $`\mathrm{m}_1,\mathrm{},\mathrm{m}_t`$ in $`Max()`$ and $`r_1,\mathrm{},r_t`$ such that $`1=r_1s_{\mathrm{m}_1}+\mathrm{}+r_ts_{\mathrm{m}_t}`$ over $``$. For every $`l=1`$ to $`t`$, $`r_ls_{\mathrm{m}_l}`$ is an element of $`𝔞_i`$ with $`i=i_{\mathrm{m}_l}`$. Therefore we have (i).
Proof of Proposition 16. Suppose that $``$ is a unique factorization domain. Since the “If” part is obvious, we prove only the “Only If” part.
(Only If). Let $`a_1,\mathrm{},a_k`$ be in $``$. Suppose that $`(a_1,\mathrm{},a_k)`$ is projective. If all $`a_1,\mathrm{},a_k`$ are zero, the proof is obvious. Thus in the following we suppose that at least one of $`a_1,\mathrm{},a_k`$ is nonzero. Since $``$ is a unique factorization domain, there exists a nonzero greatest common factor of $`a_i`$’s, denoted by $`g`$. Thus there exist $`b_i`$’s in $`𝒜`$ with $`b_ig=a_i`$. Then $`(b_1,\mathrm{},b_k)`$ is projective again. For any prime ideal $`\mathrm{p}`$ in $`Spec()`$, $`(b_1,\mathrm{},b_k)_\mathrm{p}`$ is free of rank $`1`$. Since there is no nonunit common factor among $`b_i`$’s over $``$, $`(b_1,\mathrm{},b_k)_\mathrm{p}=_\mathrm{p}`$. By Lemma 5.25, $`(b_1,\mathrm{},b_k)=`$. Hence $`(a_1,\mathrm{},a_k)=(g)`$, which is free.
### 5.5 Full-Size Minor Ideals of $`P`$, $`C`$, and $`H(P,C)`$
Now that we have obtained Theorem 15, we know that the projectivity of the full-size minor ideal of the plant connects with the feedback stabilizability of the plant. Since $`P`$, $`C`$, and $`H(P,C)`$ are transfer matrices over $``$, we can define the full-size minor ideals of $`C`$ and $`H(P,C)`$ analogously to that of $`P`$.
We present here the relationship among the full-size minor ideals of $`P`$, $`C`$, and $`H(P,C)`$.
###### Proposition 5.27.
Let $`𝔱_P`$, $`𝔱_C`$, $`𝔱_{H(P,C)}`$ be the full-size minor ideals of $`P`$, $`C`$, and $`H(P,C)`$, respectively. Then $`𝔱_{H(P,C)}`$ is isomorphic (as an $`𝒜`$-module) to the ideal generated by $`t_1t_2`$’s for all $`t_1𝔱_P`$ and all $`t_2𝔱_C`$.
This proposition holds even if $`C`$ is not a stabilizing controller of $`P`$. Before proving this proposition, we present a preliminary lemma.
###### Lemma 5.28.
Let $`A`$ and $`B`$ are matrices over $``$ such that $`B=UA`$, where $`U`$ is a unimodular matrix over $``$. Then the ideal generated by the full-size minors of $`A`$ is equal to that of $`B`$.
The proof of this lemma is straightforward and omitted.
Proof of Proposition 5.27. By virtue of Lemma 12, we suppose without loss of generality that $`N`$ and $`N_c`$ are matrices over $`𝒜`$ and $`d`$ and $`d_c`$ in $`𝒜`$ with $`P=Nd^1`$ and $`C=N_cd_c^1`$. Let $`A`$ and $`B`$ be the following matrices:
$`A=\left[\begin{array}{cc}N_c& O\\ d_cE_n& O\\ O& N\\ O& dE_m\end{array}\right],B=\left[\begin{array}{c}Q\\ S\end{array}\right],\text{ where}`$
$`Q=\left[\begin{array}{cc}d_cE_n& N\\ N_c& dE_m\end{array}\right],S=\left[\begin{array}{cc}d_cE_n& O\\ O& dE_m\end{array}\right].`$
Then we can see that there exists a unimodular matrix $`U`$ with $`B=UA`$ and that $`H(P,C)=SQ^1`$. Let $`𝔞`$ be the ideal generated by the full-size minors of $`A`$ and $`𝔱_{P,C}`$ be the ideal generated by $`t_1t_2`$’s for all $`t_1𝔱_P`$ and all $`t_2𝔱_C`$. Then by Lemma 5.28, $`𝔱_{H(P,C)}`$ is isomorphic to $`𝔞`$ as $`𝒜`$-modules. Also by Binet-Cauchy formula, $`𝔞𝔱_{P,C}`$. Hence we obtain $`𝔱_{H(P,C)}𝔱_{P,C}`$.
## 6 Stabilizability in terms of Coprimeness of Quotient Ideals
In this section, we present one more necessary and sufficient condition of the feedback stabilizability which is given in terms of quotient ideals.
###### Theorem 6.29.
Let $`P`$ be a causal plant of $`𝒫^{n\times m}`$. Then the plant $`P`$ is stabilizable if and only if the ideal
$$\underset{I}{}((t_I):𝔱)$$
(11)
is equal to $`𝒜`$.
The ideal of (11) will be considered as another generalization of the reduced minors. This will be presented later as Proposition 6.34.
We note that the result above can be considered as a generalization of Theorem 2.1.1 in given by Shankar and Sule as well as a generalization of Theorem 9. They considered the single-input single-output case. In Theorem 2.1.1 of , they stated the feedback stabilizability of the given plant in terms of the coprimeness of the ideal quotients as (11). As a result, Theorem 6.29 can be considered as a multi-input multi-output version of Theorem 2.1.1 of .
In order to prove Theorem 6.29, we prepare a relationship between projective modules and quotient ideals as follows.
###### Theorem 6.30.
Let $``$ be a commutative ring and $`a_1,\mathrm{},a_k`$. Then $`(a_1,\mathrm{},a_k)`$, that is, the ideal generated by $`a_1,\mathrm{},a_k`$ is projective if and only if the following equation holds:
$$\underset{i=1}{\overset{k}{}}((a_i):(a_1,\mathrm{},a_k))=.$$
(12)
Once we obtain Theorem 6.30, the proof of Theorem 6.29 is directly obtained from Theorems 15 and 6.30. Thus we will present only the proof of Theorem 6.30, which will be given after showing intermediate results (Lemmas 6.31 and 6.33).
###### Lemma 6.31.
Let $``$ be a commutative ring and $`a_1,\mathrm{},a_k`$. If $`(a_1,\mathrm{},a_k)`$ is free, then (12) holds.
###### Proof 6.32.
As in the proof of Proposition 16, if all $`a_1,\mathrm{},a_k`$ are zero, the proof is obvious. Thus in the following we assume that at least one of $`a_1,\mathrm{},a_k`$ is nonzero. Then there exist a nonzero $`g`$ in $``$ and $`b_i`$ in $``$ for $`i=1`$ to $`k`$ such that $`(g)=(a_1,\mathrm{},a_k)`$ and $`a_i=b_ig`$. Thus there exist $`r_i`$ for $`i=1`$ to $`k`$ with $`g=r_1a_1+\mathrm{}+r_ka_k`$. If $`g`$ was a zerodivisor, the principal ideal $`(g)`$ could not be free. Hence $`g`$ is a nonzerodivisor. Now we have
$$r_1b_1+\mathrm{}+r_kb_k=1.$$
(13)
Since $`b_i(a_1,\mathrm{},a_k)(a_i)`$ for all $`i`$, we have $`b_i((a_i):(a_1,\mathrm{},a_k))`$. It follows from (13) that we now have (12).
###### Lemma 6.33.
Let $``$ be a commutative ring, $`𝔞,𝔟`$ ideals of $``$, and $`\mathrm{p}`$ a prime ideal of $``$. Denote by $`(𝔞:𝔟)_\mathrm{p}`$ the localization of the quotient ideal $`(𝔞:𝔟)`$ at $`\mathrm{p}`$. Further let $`(𝔞_\mathrm{p}:𝔟_\mathrm{p})`$ be the quotient ideal of $`_\mathrm{p}`$, where $`𝔞_\mathrm{p}`$ and $`𝔟_\mathrm{p}`$ are localizations of ideals $`𝔞`$ and $`𝔟`$ at $`\mathrm{p}`$, respectively. Then $`(𝔞:𝔟)_\mathrm{p}=(𝔞_\mathrm{p}:𝔟_\mathrm{p})`$ holds.
Now we are in a position to prove Theorem 6.30.
Proof of Theorem 6.30. By the same reason as in the proofs of Proposition 16 and Lemma 6.31, we assume that at least one of $`a_1,\mathrm{},a_k`$ is nonzero.
(If). Suppose that (12) holds. Then there exist $`x_i((a_i):(a_1,\mathrm{},a_k))`$ for $`i=1`$ to $`k`$ such that $`1=_{i=1}^kx_i`$. By appropriate changes of $`a_1,\mathrm{},a_k`$, we assume without loss of generality that all $`x_1,\mathrm{},x_k^{}`$ are nonzero with $`1k^{}k`$ and all $`x_{k^{}+1},\mathrm{},x_k`$ are zero subject to $`k^{}<k`$. Observe that for each $`i`$ between $`1`$ and $`k^{}`$, $`(a_1,\mathrm{},a_k)_{x_i}=(a_i)_{x_i}`$ over $`𝒜_{x_i}`$, where $`(a_1,\mathrm{},a_k)_{x_i}`$ and $`(a_i)_{x_i}`$ denote the localizations of $`(a_1,\mathrm{},a_k)`$ and $`(a_i)`$ at $`x_i`$, respectively. Hence for each $`i`$ between $`1`$ and $`k^{}`$, $`(a_1,\mathrm{},a_k)_{x_i}`$ is free over $`𝒜_{x_i}`$. Therefore by Theorem IV.32 of , $`(a_1,\mathrm{},a_k)`$ is projective as $``$-module.
(Only If). Suppose that $`(a_1,\mathrm{},a_k)`$ is projective. Then again by Theorem IV.32 of , for each $`\mathrm{p}`$ in $`Spec()`$, $`(a_1,\mathrm{},a_k)_\mathrm{p}`$ is free over $`_\mathrm{p}`$. By Lemma 6.31, we have
$$\underset{i=1}{\overset{k}{}}((a_i)_\mathrm{p}:(a_1,\mathrm{},a_k)_\mathrm{p})=_\mathrm{p}$$
(14)
for each $`\mathrm{p}`$ in $`Spec()`$. Then (14) can be rewritten as follows by Lemma 6.33:
$$\underset{i=1}{\overset{k}{}}((a_i):(a_1,\mathrm{},a_k))_\mathrm{p}=_\mathrm{p}.$$
(15)
Since this holds for every $`\mathrm{p}`$ in $`Spec(𝒜)`$, applying Lemma 5.25 to (15) we obtain (12).
We now connect the reduced minors with the quotient ideal of (11) provided that $`𝒜`$ is a unique factorization domain.
###### Proposition 6.34.
Suppose that $`𝒜`$ is a unique factorization domain. Let $`a_I`$ denote the reduced minor of the matrix $`T`$ with respect to $`I`$. Then $`(a_I)=((t_I):𝔱)`$ holds for every $`I`$.
###### Proof 6.35.
We first show (i) $`(a_I)((t_I):𝔱)`$ and then (ii) the opposite inclusion.
(i). For every $`I^{}`$, $`a_It_I^{}=a_I^{}t_I`$ holds, which implies that $`a_I((t_I):(t_I^{}))`$. Hence $`a_I((t_I):𝔱)`$.
(ii). Suppose that $`\lambda _I`$ is an element of the quotient ideal $`((t_I):𝔱)`$. Then for every $`I^{}`$, there exists $`\nu _I^{}𝒜`$ such that $`\lambda _It_I^{}=\nu _I^{}t_I`$ holds and so $`\lambda _Ia_I^{}=\nu _I^{}a_I`$. Since this equality holds for every $`I^{}`$, $`\lambda _I`$ has a factor $`a_I`$. Hence $`\lambda _I(a_I)`$.
From the result above, the reduced minor of the matrix $`T`$ with respect to $`I`$ is equal to the quotient ideal $`((t_I):𝔱)`$ up to a unit multiple of $`𝒜`$ provided that $`𝒜`$ is a unique factorization domain.
Now that we have shown a new criterion (11) of the feedback stabilizability, in the following we present the relationship between generalized elementary factors and (11) by using radicals of ideals.
###### Theorem 6.36.
Let $`\mathrm{\Lambda }_{PI}`$ denote the generalized elementary factor of the plant $`P`$ with respect to $`I`$ in $``$. Then the radical of $`\mathrm{\Lambda }_{PI}`$ is equal to the radical of $`((t_I):𝔱)`$.
Before proving this result, we present an analogous result of Lemma 6.33.
###### Lemma 6.37.
Let $``$ be a commutative ring, $`𝔞,𝔟`$ ideals of $``$, and $`f`$. Denote by $`(𝔞:𝔟)_f`$ the localization of the quotient ideal $`(𝔞:𝔟)`$ at $`f`$. Further let $`(𝔞_f:𝔟_f)`$ be the quotient ideal of $`_f`$, where $`𝔞_f`$ and $`𝔟_f`$ are localizations of principal ideals $`𝔞`$ and $`𝔟`$ at $`f`$, respectively. Then $`(𝔞:𝔟)_f=(𝔞_f:𝔟_f)`$ holds.
Analogously to Lemma 6.33, the proof of this lemma is omitted.
Proof of Theorem 6.36. Let $`I`$ be fixed. We first show (i) $`\mathrm{\Lambda }_{PI}\sqrt{((t_I):𝔱)}`$ and then (ii) $`\sqrt{\mathrm{\Lambda }_{PI}}((t_I):𝔱)`$. They are sufficient to prove this theorem.
(i). Let $`\lambda `$ be an arbitrary but fixed element of $`\mathrm{\Lambda }_{PI}`$. Then there exists a matrix $`K`$ over $`𝒜`$ with $`\lambda T=K\mathrm{\Delta }_IT`$. Then for every $`I^{}`$, we have $`\lambda \mathrm{\Delta }_I^{}T=\mathrm{\Delta }_I^{}K\mathrm{\Delta }_IT`$, so that $`\lambda ^mt_I^{}=det(\mathrm{\Delta }_I^{}K)t_I`$. This implies $`\lambda ^m((t_I):(t_I^{}))`$. Hence we have $`\lambda ^m_I^{}((t_I):(t_I^{}))=((t_I):_I^{}(t_I^{}))`$.
(ii). Let $`\lambda `$ be an arbitrary but fixed element of $`((t_I):𝔱)`$. Then $`((t_I):𝔱)_\lambda =𝒜_\lambda `$ and hence $`((t_I)_\lambda :𝔱_\lambda )=𝒜_\lambda `$ by Lemma 6.37. This implies that $`(t_I)_\lambda =𝔱_\lambda `$ and further that every full-size minor of $`T`$ has a factor $`t_I`$ over $`𝒜_\lambda `$. Since $`t_I`$ is a factor of $`det(D)`$, it is a nonzerodivisor of $`𝒜_\lambda `$. Now let $`T^{}=T(adj(\mathrm{\Delta }_IT))`$ and $`t_I^{}^{}=det(\mathrm{\Delta }_I^{}T^{})`$ for every $`I^{}`$. Then $`t_I^{}^{}=t_I^{}det(adj(\mathrm{\Delta }_IT))`$ and $`\mathrm{\Delta }_IT^{}=t_IE_m`$ hold. Since $`det(adj(\mathrm{\Delta }_IT))=t_I^{m1}`$, every $`t_I^{}^{}`$ has a common factor $`t_I^m`$.
Analogously to the proof of Theorem 15, we can show that every entry of $`T^{}`$ has a factor $`t_I`$. Let $`T^{\prime \prime }=T^{}/t_I`$ over $`𝒜_\lambda `$. Then $`T=T^{\prime \prime }\mathrm{\Delta }_IT`$ holds over $`𝒜_\lambda `$. Hence there exists an integer $`\omega `$ such that $`\lambda _I^\omega T^{\prime \prime }`$ can be considered over $`𝒜`$ and further $`\lambda _I^\omega T=\lambda _I^\omega T^{\prime \prime }\mathrm{\Delta }_IT`$ holds over $`𝒜`$. Now letting $`K=\lambda _I^\omega T^{\prime \prime }\mathrm{\Delta }_I`$, we have that $`\lambda _I^\omega `$ is an element of $`\mathrm{\Lambda }_{PI}`$ and hence $`\lambda _I\sqrt{\mathrm{\Lambda }_{PI}}`$.
In the case where $`𝒜`$ is a unique factorization domain, we obtain the following result which connects Theorems 7 and 9.
###### Theorem 6.38.
Suppose that $`𝒜`$ is a unique factorization domain. Let $`P`$ be a causal plant and $`I`$ in $``$. Then the radical of the elementary factor of the matrix $`T`$ with respect to $`I`$ is equal to the radical of the reduced minor of $`T`$ with respect to $`I`$ up to a unit multiple.
###### Proof 6.39.
Let $`f_I`$ denote the elementary factor of the matrix $`T`$ with respect to $`I`$. Also let $`a_I`$ denote the reduced minor of $`T`$ with respect to $`I`$.
In the case where $`𝒜`$ is a unique factorization domain, the generalized elementary factor of the plant $`P`$ with respect to $`I`$ is equal to the principal ideal $`(f_I)`$. Thus, by Theorem 6.36, $`\sqrt{(f_I)}=\sqrt{((t_I):𝔱)}`$. By virtue of Proposition 6.34, we have $`\sqrt{(f_I)}=\sqrt{(a_I)}`$.
## 7 Concluding Remarks
We have presented two generalization of the reduced minors. One is the full-size minor ideal. Its projectivity is a criterion of the feedback stabilizability(Theorem 15). The other is quotient ideals in (12). Their coprimeness is a criterion of the feedback stabilizability(Theorem 6.29).
|
warning/0003/math0003200.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
Denote by $``$ the ring of integers. A unimodular (linear) lattice is a finite-rank free abelian group with a symmetric integer-valued $``$-bilinear form whose associated symmetric matrix is of determinant $`\pm 1`$. Unimodular lattices are important objects in geometry of numbers. In sphere packings, many packings provided by unimodular lattices are of higher covering density (cf. \[CS3\]). The second cohomology group of a 4-manifold over $``$ modulo the torsion subgroup forms a unimodular lattice with respect to its intersection matrix (cf. \[F\]). The vertex operator superalgebra associated with a positive definite unimodular lattice enjoys the property of having a unique locally-finite irreducible module, that is, itself. The well-known example of Leech lattice is directly related to important simple finite groups, the Conway groups (cf. \[C\]). The moonshine representation of the Monster group was constructed by orbifold construction through the vertex operator algebra associated with the Leech lattice and some of their twisted modules (cf. \[B2\], \[FLM\]). Classification of positive definite unimodular lattices has been done up to rank twenty-six (cf. \[B1\], \[CS2\]). In \[X1\] (also cf. \[X2\], \[X3\]), the author gave some explicit constructions of infinite families of positive definite unimodular lattices.
The arithmatic content of a positive definite unimodular lattice is given by its theta series, the generating function of the numbers of lattice points on spheres of integral square radius. A classical Hecke’s theorem says that the theta series of a positive definite even unimodular lattice is a polynomial of the well-known Essenstein series $`E_4(z)`$ and Ramanujan series $`\mathrm{\Delta }_{24}(z)`$ (cf. \[H\]). However, there is still a little known on how to write the theta series of a positive definite even unimodular lattice as the polynomials of $`E_4(z)`$ and $`\mathrm{\Delta }_{24}(z)`$. In particular, there are no results on the theta series of infinite families of positive definite unimodular lattices. A natural question is what kind polynomials of $`E_4(z)`$ and $`\mathrm{\Delta }_{24}(z)`$ could be the theta series of positive definite even unimodular lattices. In this paper, we find two combinatorial identities on the theta series of the root lattices of the finite-dimensional simple Lie algebras of type $`D_{4n}`$ and the cosets in their integral duals, in terms of the Essenstein series $`E_4(z)`$ and Ramanujan series $`\mathrm{\Delta }_{24}(z)`$. Using these two identities, we determine the theta series of certain infinite families of postive definite even unimodular lattices with a sublattice of the same rank that is isomorphic to the direct sum of finite copies of the root lattices of the finite-dimensional simple Lie algebras of type $`D_{2n}`$. It turns out that these theta series are weighted symmetric polynomials in $`\mathrm{\Delta }_{24}(z)`$ and two fixed families of polynomials of $`E_4(z)`$ and $`\mathrm{\Delta }_{24}(z)`$. Below, we shall give a more detailed technical introduction.
The fundamental arithematic functions used in counting lattice points are the following theta functions:
$$\vartheta _2(z)=\underset{m}{}q^{(m+1/2)^2},\vartheta _3(z)=\underset{m}{}q^{m^2},\vartheta _4(z)=\underset{m}{}(q)^{m^2},$$
$`(1.1)`$
where
$$q=e^{\pi \sqrt{1}z}\text{with}\text{Im}(z)>0.$$
$`(1.2)`$
An important identity among them is
$$\vartheta _2^4(z)+\vartheta _4^4(z)=\vartheta _3^4(z).$$
$`(1.3)`$
The Essenstein $`E_4(z)`$ series is defined by
$$E_4(z)=\frac{1}{2}(\vartheta _2^8(z)+\vartheta _3^8(z)+\vartheta _4^8(z))$$
$`(1.4)`$
and the Ramanujan series is defined by
$$\mathrm{\Delta }_{24}(z)=\left(\frac{\vartheta _2(z)\vartheta _3(z)\vartheta _4(z)}{2}\right)^8.$$
$`(1.5)`$
As $`q`$-powers,
$$E_4(z)=1+240\underset{m=1}{\overset{\mathrm{}}{}}(\underset{d|m}{}d^3)q^{2m}$$
$`(1.6)`$
and
$$\mathrm{\Delta }_{24}(z)=\underset{m=1}{\overset{\mathrm{}}{}}\tau (m)q^{2m}$$
$`(1.7)`$
whose coefficients $`\tau (m)`$ are called Ramanujan numbers. We refer to Table 4.9 in \[CS3\] for the first one hundred of coefficients of the $`q`$-powers in $`E_4(z)`$ and the reference \[L\] for first three hundred of the Ramanujan numbers.
Denote by $``$ the field of real numbers. For a positive integer $`n`$, we denote by $`^n`$ the $`n`$-dimensional Eucleadean space with the inner product
$$\stackrel{}{\alpha },\stackrel{}{\beta }=\underset{i=1}{\overset{n}{}}\alpha _i\beta _i\text{for}\stackrel{}{\alpha }=(\alpha _1,\mathrm{},\alpha _n),\stackrel{}{\beta }=(\beta _1,\mathrm{},\beta _n)^n.$$
$`(1.8)`$
A positive definite integral lattice $`L`$ of rank $`n`$ is an additive subgroup $`L`$ of $`^n`$ generated by a basis of $`^n`$ such that
$$\stackrel{}{\alpha },\stackrel{}{\beta }\text{for}\stackrel{}{\alpha },\stackrel{}{\beta }L.$$
$`(1.9)`$
The lattice $`L`$ is called even if
$$\stackrel{}{\alpha },\stackrel{}{\alpha }2\text{for}\stackrel{}{\alpha }L,$$
$`(1.10)`$
and is unimodular if and only if
$$L=\{\stackrel{}{\alpha }^n\stackrel{}{\alpha },\stackrel{}{\beta }\text{for}\stackrel{}{\beta }L\}.$$
$`(1.11)`$
For any coset $`L^{}`$ of $`L`$ in $`^n`$, we define the theta series of $`L^{}`$ by
$$\mathrm{\Theta }_L^{}(z)=\underset{\stackrel{}{\alpha }L^{}}{}q^{\stackrel{}{\alpha },\stackrel{}{\alpha }}.$$
$`(1.12)`$
Hecke \[H\] proved that $`\mathrm{\Theta }_L(z)`$ is a polynomial of $`E_4(z)`$ and $`\mathrm{\Delta }_{24}(z)`$ for any positive definite even unimodular lattice $`L`$. Thus a fundamental question in positive definite even unimodular lattices is to determine the coefficients of the theta series of the lattices as polynomials of $`E_4(z)`$ and $`\mathrm{\Delta }_{24}(z)`$. One may ask the question in another way that what kind polynomials of $`E_4(z)`$ and $`\mathrm{\Delta }_{24}(z)`$ could be the theta series of positive definite even unimodular lattices. To this author’s best knowledge, only a few of theta series of positive definite even unimodular lattices are known.
For a nonnegative integer $`n`$, we let
$$h_n(z)=\vartheta _2^{8n}(z)+\vartheta _3^{8n}(z)+\vartheta _4^{8n}(z),$$
$`(1.13)`$
$$\rho _n(z)=\frac{\vartheta _3^{8(n+1)+4}(z)\vartheta _2^{8(n+1)+4}(z)\vartheta _4^{8(n+1)+4}(z)}{(\vartheta _2(z)\vartheta _3(z)\vartheta _4(z))^4}.$$
$`(1.14)`$
By (1.3), the coefficients of the $`q`$-powers in $`\rho _n(z)`$ are positive integers. Denote by $`_+`$ the set of positive integers and by $``$ the set of nonnegative integers. In this paper, we shall prove that for any $`n_+`$, the following combinatorial identities hold:
$$h_n(z)=2E_4^n(z)+\underset{i=1}{\overset{[|n/3|]}{}}\frac{n}{i}(\begin{array}{c}ni1\\ 2i1\end{array})2^{8i}\mathrm{\Delta }_{24}^i(z)E_4^{n3i}(z),$$
$`(1.15)`$
$$\rho _n(z)=\underset{i=0}{\overset{[|n/3|]}{}}\frac{2n+3}{2i+1}(\begin{array}{c}ni\\ 2i\end{array})2^{8i}\mathrm{\Delta }_{24}^i(z)E_4^{n3i}(z).$$
$`(1.16)`$
The above two identities can be viewed as higher-order analogues of the identity (1.3).
For a postive intger $`n`$, the type-D root lattice $`R_{D_n}`$ is defined by
$$R_{D_n}=\{\alpha =(\alpha _1,\alpha _2,\mathrm{},\alpha _n)^n\underset{i=1}{\overset{n}{}}\alpha _i2\}^n$$
$`(1.17)`$
with the symmetric form (1.8). We define weights by
$$\text{wt}\mathrm{\Delta }_{24}(z)=3,\text{wt}h_n(z)=\text{wt}\rho _n(z)=n\text{for}n_+.$$
$`(1.18)`$
In this paper, we shall determine the theta series of certain infinite families of even unimodular lattices that have a sublattice of the same rank and isomorphic to the direct sum of finite copies of the lattices $`R_{D_n}`$ with various $`n`$. They are the following functions
$`2^{2\mathrm{}1}[{\displaystyle \underset{1j_1j_2;j_1+j_2\mathrm{}}{}}\text{sym}\{h_{m_1+\mathrm{}+m_{2j_1}}(z)h_{m_{2j_1+1}+\mathrm{}+m_{2(j_1+j_2)}}(z)h_{m_{2(j_1+j_2)+1}+\mathrm{}+m_{2\mathrm{}+1}}(z)\}`$
$`{\displaystyle \underset{0j_1j_2\mathrm{}j_1j_21}{}}\text{sym}\{h_{m_1+\mathrm{}+m_{2j_1+1}}(z)h_{m_{2j_1+2}+\mathrm{}+m_{2(j_1+j_2+1)}}(z)`$
$`\times h_{m_{2(j_1+j_2)+3}+\mathrm{}+m_{2\mathrm{}+1}}(z)\}+{\displaystyle }_{j=1}^{\mathrm{}}(34^\mathrm{}j)\text{sym}\{h_{m_1+\mathrm{}+m_{2j}}(z)h_{m_{2j+1}+\mathrm{}+m_{2\mathrm{}+1}}(z)\}`$
$`+(434^{\mathrm{}}+{\displaystyle \frac{2}{3}}{\displaystyle \underset{j_1+j_2\mathrm{}1}{}}(\begin{array}{c}2\mathrm{}+1\\ 2j_1+1,2j_2+1\end{array}))h_{m_1+\mathrm{}+m_{2\mathrm{}+1}}(z)](1.19)`$
for $`\mathrm{}`$ and $`m_1,\mathrm{},m_{2\mathrm{}+1}_+`$;
$`2^2\mathrm{}[{\displaystyle \underset{1j_1j_2\mathrm{}j_1j_2}{}}\text{sym}\{h_{m_1+\mathrm{}+m_{2j_1}+j_1}(z)h_{m_{2j_1+1}+\mathrm{}+m_{2(j_1+j_2)}+j_2}(z)`$
$`\times h_{m_{2(j_1+j_2)+1}+\mathrm{}+m_2\mathrm{}+\mathrm{}j_1j_2}(z)\}2^8\mathrm{\Delta }_{24}(z){\displaystyle }_{1j_1,j_2;j_1+j_2\mathrm{}2}\text{sym}\{\rho _{m_1+\mathrm{}+m_{2j_1+1}+j_11}(z)`$
$`\times \rho _{m_{2j_1+2}+\mathrm{}+m_{2(j_1+j_2+1)}+j_21}(z)h_{m_{2(j_1+j_2)+3}+\mathrm{}+m_2\mathrm{}+\mathrm{}j_1j_21}(z)\}`$
$`+3{\displaystyle \underset{j=1}{\overset{[|\mathrm{}/2|]}{}}}\text{sym}\{h_{m_1+\mathrm{}+m_{2j}+j}(z)h_{m_{2j+1}+\mathrm{}+m_2\mathrm{}+\mathrm{}j}(z)\}+2^8\mathrm{\Delta }_{24}(z)`$
$`\times {\displaystyle \underset{j=0}{\overset{[|(\mathrm{}1)/2|]}{}}}(2^{2j}+2^{2(\mathrm{}j1)}3)\text{sym}\{\rho _{m_1+\mathrm{}+m_{2j+1}+j1}(z)\rho _{m_{2j+2}+\mathrm{}+m_2\mathrm{}+\mathrm{}j2}(z)\}`$
$`+(4{\displaystyle \frac{2}{3}}{\displaystyle \underset{j_1+j_2\mathrm{}}{}}(\begin{array}{c}2\mathrm{}\\ 2j_1,2j_2\end{array}))h_{m_1+\mathrm{}+m_2\mathrm{}+\mathrm{}}(z)](1.20)`$
for $`\mathrm{}_+`$ and $`m_1,\mathrm{},m_2\mathrm{}`$;
$`{\displaystyle \frac{1}{2}}h_{m_1+m_2+m_3+m_4+2ϵ+1}(z)32\mathrm{\Delta }_{24}(z)[\rho _{m_1+m_2+ϵ1}(z)\rho _{m_3+m_4+ϵ1}(z)`$
$`+\rho _{m_1+m_3+ϵ1}(z)\rho _{m_2+m_4+ϵ1}(z)+\rho _{m_1+m_4+ϵ1}(z)\rho _{m_2+m_3+ϵ1}(z)](1.21)`$
for $`m_1,m_2,m_3,m_4`$ and $`ϵ=0,1`$. Here we have treated $`\rho _1(z)=0`$ in (1.20), and (1.21), and “$`\text{sym}\{\}`$” means the symmetric polynomial with respect to the integrable variables $`m_i`$ and the expression in the braces is a representative term. For instance,
$`\text{sym}\{h_{m_1+m_2+1}(z)h_{m_3+m_4+1}(z)\}`$
$`=`$ $`h_{m_1+m_2+1}(z)h_{m_3+m_4+1}(z)+h_{m_1+m_3+1}(z)h_{m_2+m_4+1}(z)`$
$`+h_{m_1+m_4+1}(z)h_{m_2+m_3+1}(z),(1.22)`$
$`\text{sym}\{h_{m_1+m_2+1}(z)\rho _{m_31}(z)\rho _{m_41}(z)\}`$
$`=`$ $`h_{m_1+m_2+1}(z)\rho _{m_31}(z)\rho _{m_41}(z)+h_{m_1+m_3+1}(z)\rho _{m_21}(z)\rho _{m_41}(z)`$
$`+h_{m_1+m_4+1}(z)\rho _{m_21}(z)\rho _{m_31}(z)+h_{m_3+m_4+1}(z)\rho _{m_11}(z)\rho _{m_21}(z)`$
$`+h_{m_2+m_4+1}(z)\rho _{m_11}(z)\rho _{m_31}(z)+h_{m_2+m_3+1}(z)\rho _{m_11}(z)\rho _{m_41}(z).(1.23)`$
These theta series are weighted symmetric polynomials of the series $`\{\mathrm{\Delta }_{24}(z),h_n(z),\rho _n(z)n_+\}`$. Thus we essentially determine the theta series of these lattices as polynomials of the well-known Essenstein series $`E_4(z)`$ and Ramanujan series $`\mathrm{\Delta }_{24}(z)`$. We speculate that the theta series of the other infinite families of positive definite even unimodular lattices may relate to the invariants of the other finite groups.
The results in this paper could be useful in study of modular forms and partition functions of the conformal field theories related to positive definite even unimodular lattices. The covering densities of the sphere packings of these lattices can be calculated by our formulae.
In Section 2, we shall prove (1.15) and (1.16). In Section 3, the theta series of our concerned lattice will be determined.
## 2 Proofs of the Combinatorial Identities
In this section, we shall present the proofs of the combinatorial identities (1.15) and (1.16).
Set
$$a=\vartheta _2^4(z),b=\vartheta _4^4(z).$$
$`(2.1)`$
By (1.3), we have
$$\vartheta _3^4(z)=a+b.$$
$`(2.2)`$
Moreover, (1.13) and (1.14) become
$$h_n=a^{2n}+b^{2n}+(a+b)^{2n},\rho _n=\frac{(a+b)^{2n+3}a^{2n+3}b^{2n+3}}{ab(a+b)}.$$
$`(2.3)`$
We allow $`n=0`$ in the above equations.
For convenience, we let
$$E=E_4(z)=\frac{1}{2}(a^2+b^2+(a+b)^2)=a^2+b^2+ab,$$
$`(2.4)`$
$$\mathrm{\Delta }=2^8\mathrm{\Delta }_{24}(z)=(\vartheta _2(z)\vartheta _3(z)\vartheta _4(z))^8=a^2b^2(a+b)^2.$$
$`(2.5)`$
Note that
$$h_0=3,h_1=2E.$$
$`(2.6)`$
Moreover,
$`E^2`$ $`=`$ $`(a^2+b^2+ab)^2`$
$`=`$ $`a^4+b^4+(ab)^2+2a^2b^2+2a^2ab+2abb^2`$
$`=`$ $`a^4+b^4+3a^2b^2+2ab(a^2+b^2).(2.7)`$
Hence,
$`h_2`$ $`=`$ $`a^4+b^4+(a+b)^4`$
$`=`$ $`a^4+b^4+a^4+b^4+4ab(a^2+b^2)+6a^2b^2`$
$`=`$ $`2[a^4+b^4+3a^2b^2+2ab(a^2+b^2)]`$
$`=`$ $`2E^2.(2.8)`$
Lemma 2.1. For $`3n`$, we have:
$$h_n=2Eh_{n1}E^2h_{n2}+\mathrm{\Delta }h_{n3}.$$
$`(2.9)`$
Proof. Note that
$`2Eh_{n1}`$ $`=`$ $`(a^2+b^2+(a+b)^2)(a^{2(n1)}+b^{2(n1)}+(a+b)^{2(n1)})`$
$`=`$ $`a^{2n}+b^{2n}+a^2b^2(a^{2(n2)}+b^{2(n2)})+(a^2+b^2)(a+b)^{2(n1)}`$
$`+(a+b)^2(a^{2(n1)}+b^{2(n1)})+(a+b)^{2n}`$
$`=`$ $`h_n+a^2b^2(a^{2(n2)}+b^{2(n2)})+(a^2+b^2)(a+b)^{2(n1)}`$
$`+(a+b)^2(a^{2(n1)}+b^{2(n1)}).(2.10)`$
Moreover,
$`(a+b)^2(a^2+b^2)+a^2b^2`$
$`=`$ $`(a^2+b^2+2ab)(a^2+b^2)+a^2b^2`$
$`=`$ $`(a^2+b^2)^2+2ab(a^2+b^2)+a^2b^2`$
$`=`$ $`a^4+b^4+2a^2b^2+2ab(a^2+b^2)+a^2b^2`$
$`=`$ $`E^2.(2.11)`$
Hence, we obtain
$`E^2h_{n2}`$
$`=`$ $`((a+b)^2(a^2+b^2)+a^2b^2)(a^{2(n2)}+b^{2(n2)}+(a+b)^{2(n2)})`$
$`=`$ $`(a+b)^2[a^{2(n1)}+b^{2(n1)}+a^2b^2(a^{2(n3)}+b^{2(n3)})]+a^2b^2(a^{2(n2)}+b^{2(n2)})`$
$`+(a^2+b^2)(a+b)^{2(n1)}+a^2b^2(a+b)^{2(n2)}`$
$`=`$ $`(a+b)^2(a^{2(n1)}+b^{2(n1)})+\mathrm{\Delta }(a^{2(n3)}+b^{2(n3)})+a^2b^2(a^{2(n2)}+b^{2(n2)})`$
$`+(a^2+b^2)(a+b)^{2(n1)}+\mathrm{\Delta }(a+b)^{2(n3)}`$
$`=`$ $`(a+b)^2(a^{2(n1)}+b^{2(n1)})+a^2b^2(a^{2(n2)}+b^{2(n2)})`$
$`+(a^2+b^2)(a+b)^{2(n1)}+\mathrm{\Delta }h_{n3}.(2.12)`$
Thus we have
$$2Eh_{n1}E^2h_{n2}=h_n\mathrm{\Delta }h_{n3},$$
$`(2.13)`$
which is equivalent to (2.9). $`\mathrm{}`$
By Lemma 2.1, we calculate
$$h_3=2E^3+3\mathrm{\Delta },h_4=2E^4+8\mathrm{\Delta }E,h_5=2E^5+15\mathrm{\Delta }E^2,$$
$`(2.14)`$
$$h_6=2E^6+24\mathrm{\Delta }E^3+3\mathrm{\Delta }^2,h_7=2E^7+35\mathrm{\Delta }E^4+14\mathrm{\Delta }^2E,$$
$`(2.15)`$
$$h_8=2E^8+48\mathrm{\Delta }E^5+40\mathrm{\Delta }^2E^2,h_9=2E^9+63\mathrm{\Delta }E^6+90\mathrm{\Delta }^2E^3+3\mathrm{\Delta }^3,$$
$`(2.16)`$
$$h_{10}=2E^{10}+80\mathrm{\Delta }E^7+175\mathrm{\Delta }^2E^4+20\mathrm{\Delta }^3E.$$
$`(2.17)`$
In fact, we had directly calculated (2.14)-(2.17), and then observed (2.9) from them. Analyzing the coefficients in (2.14)-(2.17), we speculated the following theorem.
Theorem 2.2. For $`n_+`$, we have
$$h_n=2E^n+\underset{i=1}{\overset{[|n/3|]}{}}\frac{n}{i}(\begin{array}{c}ni1\\ 2i1\end{array})\mathrm{\Delta }^iE^{n3i},$$
$`(2.18)`$
which is equivalent to (1.15) by (2.4) and (2.5).
Proof. We shall prove (2.18) by (2.9) and induction on $`n`$. First for $`m,k_+`$ such that $`km1`$, we have
$`{\displaystyle \frac{2m}{k}}(\begin{array}{c}mk1\\ 2k1\end{array}){\displaystyle \frac{m1}{k}}(\begin{array}{c}mk2\\ 2k1\end{array})+{\displaystyle \frac{m2}{k1}}(\begin{array}{c}mk2\\ 2k3\end{array})`$
$`=`$ $`{\displaystyle \frac{1}{2k(k1)(2k1)}}(\begin{array}{c}mk2\\ 2k3\end{array})[2m(mk1)(m+13k)`$
$`(m1)(m+13k)(m3k)+2k(m2)(2k1)]`$
$`=`$ $`{\displaystyle \frac{1}{2k(k1)(2k1)}}(\begin{array}{c}mk2\\ 2k3\end{array})[2m(m+1)(mk1)6mk(mk1)`$
$`(m1)(m+1)(m3k)+3(m1)k(m3k)+2k(m2)(2k1)]`$
$`=`$ $`{\displaystyle \frac{1}{2k(k1)(2k1)}}(\begin{array}{c}mk2\\ 2k3\end{array})[(m+1)(2m(mk1)(m1)(m3k))`$
$`+k(6m(mk1)+3(m1)(m3k)+2(2k1)(m2)]`$
$`=`$ $`{\displaystyle \frac{1}{2k(k1)(2k1)}}(\begin{array}{c}mk2\\ 2k3\end{array})[(m+1)(m^2+kmm3k)`$
$`+k(3m^2+km+m+k+4)]`$
$`=`$ $`{\displaystyle \frac{1}{2k(k1)(2k1)}}(\begin{array}{c}mk2\\ 2k3\end{array})[(m+1)(m^2+kmm3k)`$
$`+k(m+1)(k+43m)]`$
$`=`$ $`{\displaystyle \frac{1}{2k(k1)(2k1)}}(\begin{array}{c}mk2\\ 2k3\end{array})(m+1)(m^2+kmm3k+k^2+4k3km)`$
$`=`$ $`{\displaystyle \frac{1}{2k(k1)(2k1)}}(\begin{array}{c}mk2\\ 2k3\end{array})(m+1)(m^2+k^22kmm+k)`$
$`=`$ $`{\displaystyle \frac{(m+1)(mk)(mk1)}{2k(k1)(2k1)}}(\begin{array}{c}mk2\\ 2k3\end{array})`$
$`=`$ $`{\displaystyle \frac{m+1}{k}}(\begin{array}{c}mk\\ 2k1\end{array}).(2.19)`$
When $`n=1`$, (2.18) holds by (2.6). Let $`m_+`$. Suppose that (2.18) holds for $`nm`$. By (2.9),
$`h_{m+1}`$ $`=`$ $`2Eh_mE^2h_{m1}+\mathrm{\Delta }h_{m2}`$
$`=`$ $`2E^{m+1}+[2m(m2)(m1)(m3)+2]\mathrm{\Delta }E^{m2}+2{\displaystyle \underset{i=2}{\overset{[|m/3|]}{}}}{\displaystyle \frac{m}{i}}(\begin{array}{c}mi1\\ 2i1\end{array})`$
$`\times \mathrm{\Delta }^iE^{m+13i}{\displaystyle \underset{i=2}{\overset{[|(m1)/3|]}{}}}{\displaystyle \frac{m1}{i}}(\begin{array}{c}mi2\\ 2i1\end{array})\mathrm{\Delta }^iE^{m+13i}`$
$`+\mathrm{\Delta }{\displaystyle \underset{i=1}{\overset{[|(m2)/3|]}{}}}{\displaystyle \frac{m2}{i}}(\begin{array}{c}mi3\\ 2i1\end{array})\mathrm{\Delta }^iE^{m23i}`$
$`=`$ $`2E^{m+1}+(m+1)(m1)\mathrm{\Delta }E^{m2}+2{\displaystyle \underset{i=2}{\overset{[|m/3|]}{}}}{\displaystyle \frac{m}{i}}(\begin{array}{c}mi1\\ 2i1\end{array})\mathrm{\Delta }^iE^{m+13i}`$
$`{\displaystyle \underset{i=2}{\overset{[|(m1)/3|]}{}}}{\displaystyle \frac{m1}{i}}(\begin{array}{c}mi2\\ 2i1\end{array})\mathrm{\Delta }^iE^{m+13i}`$
$`+\mathrm{\Delta }{\displaystyle \underset{i=2}{\overset{[|(m2)/3|]+1}{}}}{\displaystyle \frac{m2}{i1}}(\begin{array}{c}mi2\\ 2i3\end{array})\mathrm{\Delta }^iE^{m+13i}.(2.20)`$
Case 1. $`m=3l`$ with $`l_+`$.
$`h_{m+1}`$ $`=`$ $`2E^{m+1}+(m+1)(m1)\mathrm{\Delta }E^{m2}+2{\displaystyle \underset{i=2}{\overset{l}{}}}{\displaystyle \frac{m}{i}}(\begin{array}{c}mi1\\ 2i1\end{array})\mathrm{\Delta }^iE^{m+13i}`$
$`{\displaystyle \underset{i=2}{\overset{l1}{}}}{\displaystyle \frac{m1}{i}}(\begin{array}{c}mi2\\ 2i1\end{array})\mathrm{\Delta }^iE^{m+13i}+\mathrm{\Delta }{\displaystyle \underset{i=2}{\overset{l}{}}}{\displaystyle \frac{m2}{i1}}(\begin{array}{c}mi2\\ 2i3\end{array})\mathrm{\Delta }^iE^{m+13i}`$
$`=`$ $`2E^{m+1}+(m+1)(m1)\mathrm{\Delta }E^{m2}+{\displaystyle \underset{i=2}{\overset{l1}{}}}[{\displaystyle \frac{2m}{i}}(\begin{array}{c}mi1\\ 2i1\end{array}){\displaystyle \frac{m1}{i}}(\begin{array}{c}mi2\\ 2i3\end{array})`$
$`+{\displaystyle \frac{m2}{i1}}(\begin{array}{c}mi2\\ 2i3\end{array})]\mathrm{\Delta }^iE^{m+13i}+[{\displaystyle \frac{2m}{l}}(\begin{array}{c}2l1\\ 2l1\end{array})+{\displaystyle \frac{m2}{l2}}(\begin{array}{c}2l2\\ 2l3\end{array})]\mathrm{\Delta }^lE`$
$`=`$ $`2E^{m+1}+(m+1)(m1)\mathrm{\Delta }E^{m2}+{\displaystyle \underset{i=2}{\overset{l1}{}}}{\displaystyle \frac{m+1}{i}}(\begin{array}{c}mi\\ 2i1\end{array})\mathrm{\Delta }^iE^{m+13i}`$
$`+2(m+1)\mathrm{\Delta }^lE`$
$`=`$ $`2E^{m+1}+{\displaystyle \underset{i=1}{\overset{l1}{}}}{\displaystyle \frac{m+1}{i}}(\begin{array}{c}mi\\ 2i1\end{array})\mathrm{\Delta }^iE^{m+13i}+{\displaystyle \frac{m+1}{l}}(\begin{array}{c}ml\\ 2l1\end{array})\mathrm{\Delta }^lE`$
$`=`$ $`2E^{m+1}+{\displaystyle \underset{i=1}{\overset{l}{}}}{\displaystyle \frac{m+1}{i}}(\begin{array}{c}mi\\ 2i1\end{array})\mathrm{\Delta }^iE^{m+13i}`$
$`=`$ $`2E^{m+1}+{\displaystyle \underset{i=1}{\overset{[|(m+1)/3|]}{}}}(\begin{array}{c}mi\\ 2i1\end{array})\mathrm{\Delta }^iE^{m+13i}.(2.21)`$
Case 2. $`m=3l+1`$ with $`l`$.
$`h_{m+1}`$ $`=`$ $`2E^{m+1}+(m+1)(m1)\mathrm{\Delta }E^{m2}+{\displaystyle \underset{i=2}{\overset{l}{}}}[{\displaystyle \frac{2m}{i}}(\begin{array}{c}mi1\\ 2i1\end{array})`$
$`{\displaystyle \frac{m1}{i}}(\begin{array}{c}mi2\\ 2i1\end{array})+{\displaystyle \frac{m2}{i1}}(\begin{array}{c}mi2\\ 2i3\end{array})]\mathrm{\Delta }^iE^{m+13i}`$
$`=`$ $`2E^{m+1}+{\displaystyle \underset{i=1}{\overset{[|(m+1)/3|]}{}}}{\displaystyle \frac{m+1}{i}}(\begin{array}{c}mi\\ 2i1\end{array})\mathrm{\Delta }^iE^{m+13i}.(2.22)`$
Case 3. $`m=3l+2`$ with $`l`$.
$`h_{m+1}`$ $`=`$ $`2E^{m+1}+(m+1)(m1)\mathrm{\Delta }E^{m2}+{\displaystyle \underset{i=2}{\overset{l}{}}}[{\displaystyle \frac{2m}{i}}(\begin{array}{c}mi1\\ 2i1\end{array}){\displaystyle \frac{m1}{i}}(\begin{array}{c}mi2\\ 2i1\end{array})`$
$`+{\displaystyle \frac{m2}{(i1)}}(\begin{array}{c}mi2\\ 2i3\end{array})]\mathrm{\Delta }^iE^{m+13i}+{\displaystyle \frac{m2}{l}}(\begin{array}{c}2l1\\ 2l1\end{array})\mathrm{\Delta }^{l+1}`$
$`=`$ $`2E^{m+1}+{\displaystyle \underset{i=1}{\overset{l}{}}}{\displaystyle \frac{(m+1)}{i}}(\begin{array}{c}mi\\ 2i1\end{array})\mathrm{\Delta }^iE^{m+13i}+3\mathrm{\Delta }^{l+1}`$
$`=`$ $`2E^{m+1}+{\displaystyle \underset{i=1}{\overset{l}{}}}{\displaystyle \frac{(m+1)}{i}}(\begin{array}{c}mi\\ 2i1\end{array})\mathrm{\Delta }^iE^{m+13i}+{\displaystyle \frac{m+1}{l+1}}(\begin{array}{c}m(l+1)\\ 2(l+1)1\end{array})\mathrm{\Delta }^{l+1}`$
$`=`$ $`2E^{m+1}+{\displaystyle \underset{i=1}{\overset{l+1}{}}}{\displaystyle \frac{(m+1)}{i}}(\begin{array}{c}mi\\ 2i1\end{array})\mathrm{\Delta }^iE^{m+13i}`$
$`=`$ $`2E^{m+1}+{\displaystyle \underset{i=1}{\overset{[|(m+1)/3|]}{}}}{\displaystyle \frac{(m+1)}{i}}(\begin{array}{c}mi\\ 2i1\end{array})\mathrm{\Delta }^iE^{m+13i}.(2.23)`$
Thus (2.18) holds for $`n=m+1`$. Therefore, (2.18) holds for any $`n_+`$ by induction on $`n.\mathrm{}`$
Next we consider $`\rho _n`$ (cf. (2.3)). Note that
$$\rho _0=\frac{(a+b)^3a^3b^3}{ab(a+b)}=\frac{3a^2b+3ab^2}{ab(a+b)}=3,$$
$`(2.24)`$
$`\rho _1`$ $`=`$ $`{\displaystyle \frac{(a+b)^5a^5b^5}{ab(a+b)}}`$
$`=`$ $`(ab)^1[(a+b)^4a^4b^4+ab(a^2+b^2)a^2b^2]`$
$`=`$ $`(ab)^1[4ab(a^2+b^2)+6a^2b^2+ab(a^2+b^2)a^2b^2]`$
$`=`$ $`5(a^2+b^2+ab)`$
$`=`$ $`5E,(2.25)`$
$`\rho _2`$ $`=`$ $`{\displaystyle \frac{(a+b)^7a^7b^7}{ab(a+b)}}`$
$`=`$ $`(ab)^1[(a+b)^6a^6b^6+ab(a^4+b^4)a^2b^2(a+b)+a^3b^3]`$
$`=`$ $`(ab)^1[6ab(a^4+b^4)+15a^2b^2(a+b)+20a^3b^3+ab(a^4+b^4)a^2b^2(a+b)+a^3b^3]`$
$`=`$ $`7(a^4+b^4+2ab(a+b)+3a^2b^2)`$
$`=`$ $`7E^2(2.26)`$
by (2.4) and (2.7).
Lemma 2.3. For $`3n`$, we have:
$$\rho _n=2E\rho _{n1}E^2\rho _{n2}+\mathrm{\Delta }\rho _{n3}.$$
$`(2.27)`$
Proof. Note that
$`2E\rho _{n1}`$ $`=`$ $`(ab(a+b))^1(a^2+b^2+(a+b)^2)[(a+b)^{2n+1}a^{2n+1}b^{2n+1}]`$
$`=`$ $`(ab(a+b))^1[(a^2+b^2)(a+b)^{2n+1}a^{2n+3}b^{2n+3}`$
$`a^2b^2(a^{2n1}+b^{2n1})+(a+b)^{2n+3}(a+b)^2(a^{2n+1}+b^{2n+1})]`$
$`=`$ $`\rho _n+(ab(a+b))^1[(a^2+b^2)(a+b)^{2n+1}a^2b^2(a^{2n1}+b^{2n1})`$
$`(a+b)^2(a^{2n+1}+b^{2n+1})],(2.28)`$
$`E^2\rho _{n2}`$
$`=`$ $`(ab(a+b))^1((a+b)^2(a^2+b^2)+a^2b^2)((a+b)^{2n1}a^{2n1}b^{2n1})`$
$`=`$ $`(ab(a+b))^1[(a+b)^{2n+1}(a^2+b^2)(a+b)^2(a^{2n+1}+b^{2n+1}+a^2b^2(a^{2n3}+b^{2n3}))`$
$`+a^2b^2(a+b)^{2n1}a^2b^2(a^{2n1}+b^{2n1})]`$
$`=`$ $`(ab(a+b))^1[(a+b)^{2n+1}(a^2+b^2)(a+b)^2(a^{2n+1}+b^{2n+1})\mathrm{\Delta }(a^{2n3}+b^{2n3})`$
$`+\mathrm{\Delta }(a+b)^{2n3}a^2b^2(a^{2n1}+b^{2n1})]`$
$`=`$ $`(ab(a+b))^1[(a+b)^{2n+1}(a^2+b^2)(a+b)^2(a^{2n+1}+b^{2n+1})`$
$`a^2b^2(a^{2n1}+b^{2n1})]+\mathrm{\Delta }\rho _{n3}(2.29)`$
by (2.11). Thus we have
$$2E\rho _{n1}E^2\rho _{n2}=\rho _n\mathrm{\Delta }\rho _{n3},$$
$`(2.30)`$
which is equivalent to (2.27). $`\mathrm{}`$
By Lemma 2.3, we calculate
$$\rho _3=9E^3+3\mathrm{\Delta },\rho _4=11E^4+11\mathrm{\Delta }E,\rho _5=13E^5+26\mathrm{\Delta }E^2,$$
$`(2.31)`$
$$\rho _6=15E^6+50\mathrm{\Delta }E^3+3\mathrm{\Delta }^2,\rho _7=17E^7+85\mathrm{\Delta }E^4+17\mathrm{\Delta }^2E,$$
$`(2.32)`$
$$\rho _8=19E^8+133\mathrm{\Delta }E^5+57\mathrm{\Delta }^2E^2,\rho _9=21E^9+196\mathrm{\Delta }E^6+147\mathrm{\Delta }^2E^3+3\mathrm{\Delta }^3,$$
$`(2.33)`$
$$\rho _{10}=23E^{10}+276\mathrm{\Delta }E^7+322\mathrm{\Delta }^2E^4+23\mathrm{\Delta }^3E.$$
$`(2.34)`$
Analyzing the coefficients in (2.31)-(2.34), we speculated the following theorem.
Theorem 2.4. For $`n_+`$, we have
$$\rho _n=\underset{i=0}{\overset{[|n/3|]}{}}\frac{2n+3}{2i+1}(\begin{array}{c}ni\\ 2i\end{array})\mathrm{\Delta }^iE^{n3i},$$
$`(2.35)`$
which is equivalent to (1.16) by (2.4) and (2.5).
Proof. We shall prove (2.35) by (2.27) and induction on $`n`$. First for $`m,k_+`$ such that $`km1`$, we have
$`{\displaystyle \frac{2(2m+3)}{2k+1}}(\begin{array}{c}mk\\ 2k\end{array}){\displaystyle \frac{2m+1}{2k+1}}(\begin{array}{c}mk1\\ 2k\end{array})+{\displaystyle \frac{2m1}{2k1}}(\begin{array}{c}mk1\\ 2k2\end{array})`$
$`=`$ $`{\displaystyle \frac{1}{2k(2k+1)(2k1)}}(\begin{array}{c}mk1\\ 2k2\end{array})[2(2m+3)(mk)(m3k+1)`$
$`(2m+1)(m3k+1)(m3k)+2k(2k+1)(2m1)]`$
$`=`$ $`{\displaystyle \frac{1}{2k(2k+1)(2k1)}}(\begin{array}{c}mk1\\ 2k2\end{array})[2(2m+3)(mk)(m3k+1)`$
$`(2m+1)(m3k+1)(mk)+2k(2m+1)(m3k+1)+2k(2k+1)(2m1)]`$
$`=`$ $`{\displaystyle \frac{1}{2k(2k+1)(2k1)}}(\begin{array}{c}mk1\\ 2k2\end{array})[(2(2m+3)(2m+1))(m3k+1)(mk)`$
$`+2k((2m+1)(m3k+1)+(2k+1)(2m1))]`$
$`=`$ $`{\displaystyle \frac{1}{2k(2k+1)(2k1)}}(\begin{array}{c}mk1\\ 2k2\end{array})[(2m+5)(m3k+1)(mk)`$
$`+2k(2m^26km+2m+m3k+1+4km2k+2m1)]`$
$`=`$ $`{\displaystyle \frac{1}{2k(2k+1)(2k1)}}(\begin{array}{c}mk1\\ 2k2\end{array})[(2m+5)(m3k+1)(mk)`$
$`+2k(2m^22km+5m5k)]`$
$`=`$ $`{\displaystyle \frac{1}{2k(2k+1)(2k1)}}(\begin{array}{c}mk1\\ 2k2\end{array})[(2m+5)(m3k+1)(mk)`$
$`+2k(2m+5)(mk)]`$
$`=`$ $`{\displaystyle \frac{(2m+5)(mk+1)(mk)}{2k(2k+1)(2k1)}}(\begin{array}{c}mk1\\ 2k2\end{array})`$
$`=`$ $`{\displaystyle \frac{2m+5}{2k+1}}(\begin{array}{c}m+1k\\ 2k\end{array}).(2.36)`$
Note that when $`k=0`$, $`(_{\mathrm{\hspace{0.33em}\hspace{0.33em}2}k2}^{mk1})=0`$ and
$$\frac{2(2m+3)}{2k+1}(\begin{array}{c}mk\\ 2k\end{array})\frac{2m+1}{2k+1}(\begin{array}{c}mk1\\ 2k\end{array})=2(2m+3)(2m+1)=2m+5.$$
$`(2.37)`$
When $`n=1`$, (2.35) holds by (2.25). Let $`m_+`$. Suppose that (2.35) holds for $`nm`$. By (2.35), we have the following cases.
Case 1. $`m=3l`$ with $`l_+`$.
$`\rho _{m+1}`$ $`=`$ $`2{\displaystyle \underset{i=0}{\overset{l}{}}}{\displaystyle \frac{2m+3}{2i+1}}(\begin{array}{c}mi\\ 2i\end{array})\mathrm{\Delta }^iE^{m+13i}{\displaystyle \underset{i=0}{\overset{l1}{}}}{\displaystyle \frac{2m+1}{2i+1}}(\begin{array}{c}mi1\\ 2i\end{array})\mathrm{\Delta }^iE^{m+13i}`$
$`+{\displaystyle \underset{i=0}{\overset{l1}{}}}{\displaystyle \frac{2m1}{2i+1}}(\begin{array}{c}mi2\\ 2i\end{array})\mathrm{\Delta }^{i+1}E^{m3i2}`$
$`=`$ $`2{\displaystyle \underset{i=0}{\overset{l}{}}}{\displaystyle \frac{2m+3}{2i+1}}(\begin{array}{c}mi\\ 2i\end{array})\mathrm{\Delta }^iE^{m+13i}{\displaystyle \underset{i=0}{\overset{l1}{}}}{\displaystyle \frac{2m+1}{2i+1}}(\begin{array}{c}mi1\\ 2i\end{array})\mathrm{\Delta }^iE^{m+13i}`$
$`+{\displaystyle \underset{i=0}{\overset{l}{}}}{\displaystyle \frac{2m1}{2i1}}(\begin{array}{c}mi1\\ 2i2\end{array})\mathrm{\Delta }^iE^{m+13i}`$
$`=`$ $`{\displaystyle \underset{i=0}{\overset{l1}{}}}\left[{\displaystyle \frac{2(2m+3)}{2i+1}}(\begin{array}{c}mi\\ 2i\end{array}){\displaystyle \frac{2m+1}{2i+1}}(\begin{array}{c}mi1\\ 2i\end{array})+{\displaystyle \frac{2m1}{2i1}}(\begin{array}{c}mi2\\ 2i2\end{array})\right]`$
$`\times \mathrm{\Delta }^iE^{m+13i}+\left[{\displaystyle \frac{2(6l+3)}{2l+1}}(\begin{array}{c}2l\\ 2l\end{array})+{\displaystyle \frac{6l1}{2l1}}(\begin{array}{c}3ll1\\ 2l2\end{array})\right]\mathrm{\Delta }^lE^{m+13l}`$
$`=`$ $`{\displaystyle \underset{i=0}{\overset{l1}{}}}{\displaystyle \frac{2m+5}{2i+1}}(\begin{array}{c}m+1i\\ 2i\end{array})\mathrm{\Delta }^iE^{m+13i}+(6l+5)\mathrm{\Delta }^lE^{m+13l}`$
$`=`$ $`{\displaystyle \underset{i=0}{\overset{l1}{}}}{\displaystyle \frac{2m+5}{2i+1}}(\begin{array}{c}m+1i\\ 2i\end{array})\mathrm{\Delta }^iE^{m+13i}+{\displaystyle \frac{2m+5}{2l+1}}(\begin{array}{c}m+1l\\ 2l\end{array})\mathrm{\Delta }^lE^{m+13l}`$
$`=`$ $`{\displaystyle \underset{i=0}{\overset{l}{}}}{\displaystyle \frac{2m+5}{2i+1}}(\begin{array}{c}m+1i\\ 2i\end{array})\mathrm{\Delta }^iE^{m+13i}`$
$`=`$ $`{\displaystyle \underset{i=0}{\overset{[|(m+1)/3|]}{}}}{\displaystyle \frac{2m+5}{2i+1}}(\begin{array}{c}m+1i\\ 2i\end{array})\mathrm{\Delta }^iE^{m+13i}(2.38)`$
Case 2. $`m=3l+1`$ with $`l`$.
$`\rho _{m+1}`$ $`=`$ $`2{\displaystyle \underset{i=0}{\overset{l}{}}}{\displaystyle \frac{2m+3}{2i+1}}(\begin{array}{c}mi\\ 2i\end{array})\mathrm{\Delta }^iE^{m+13i}{\displaystyle \underset{i=0}{\overset{l}{}}}{\displaystyle \frac{2m+1}{2i+1}}(\begin{array}{c}mi1\\ 2i\end{array})\mathrm{\Delta }^iE^{m+13i}`$
$`+{\displaystyle \underset{i=0}{\overset{l1}{}}}{\displaystyle \frac{2m1}{2i+1}}(\begin{array}{c}mi2\\ 2i\end{array})\mathrm{\Delta }^{i+1}E^{m3i2}`$
$`=`$ $`{\displaystyle \underset{i=0}{\overset{l}{}}}{\displaystyle \frac{2m+5}{2i+1}}(\begin{array}{c}m+1i\\ 2i\end{array})\mathrm{\Delta }^iE^{m+13i}`$
$`=`$ $`{\displaystyle \underset{i=0}{\overset{[|(m+1)/3|]}{}}}{\displaystyle \frac{2m+5}{2i+1}}(\begin{array}{c}m+1i\\ 2i\end{array})\mathrm{\Delta }^iE^{m+13i}(2.39)`$
Case 3. $`m=3l+2`$ with $`l`$.
$`\rho _{m+1}`$ $`=`$ $`2{\displaystyle \underset{i=0}{\overset{l}{}}}{\displaystyle \frac{2m+3}{2i+1}}(\begin{array}{c}mi\\ 2i\end{array})\mathrm{\Delta }^iE^{m+13i}{\displaystyle \underset{i=0}{\overset{l}{}}}{\displaystyle \frac{2m+1}{2i+1}}(\begin{array}{c}mi1\\ 2i\end{array})\mathrm{\Delta }^iE^{m+13i}`$
$`+{\displaystyle \underset{i=0}{\overset{l}{}}}{\displaystyle \frac{2m1}{2i+1}}(\begin{array}{c}mi2\\ 2i\end{array})\mathrm{\Delta }^{i+1}E^{m3i2}`$
$`=`$ $`2{\displaystyle \underset{i=0}{\overset{l}{}}}{\displaystyle \frac{2m+3}{2i+1}}(\begin{array}{c}mi\\ 2i\end{array})\mathrm{\Delta }^iE^{m+13i}{\displaystyle \underset{i=0}{\overset{l}{}}}{\displaystyle \frac{2m+1}{2i+1}}(\begin{array}{c}mi1\\ 2i\end{array})\mathrm{\Delta }^iE^{m+13i}`$
$`+{\displaystyle \underset{i=0}{\overset{l+1}{}}}{\displaystyle \frac{2m1}{2i1}}(\begin{array}{c}mi1\\ 2i2\end{array})\mathrm{\Delta }^iE^{m+13i}`$
$`=`$ $`{\displaystyle \underset{i=0}{\overset{l}{}}}{\displaystyle \frac{2m+5}{2i+1}}(\begin{array}{c}m+1i\\ 2i\end{array})\mathrm{\Delta }^iE^{m+13i}+{\displaystyle \frac{6l+3}{2l+1}}(\begin{array}{c}2l\\ 2l\end{array})\mathrm{\Delta }^{l+1}E^{m3l2}`$
$`=`$ $`{\displaystyle \underset{i=0}{\overset{l}{}}}{\displaystyle \frac{2m+5}{2i+1}}(\begin{array}{c}m+1i\\ 2i\end{array})\mathrm{\Delta }^iE^{m+13i}+3\mathrm{\Delta }^{l+1}E^{m3l2}`$
$`=`$ $`{\displaystyle \underset{i=0}{\overset{l}{}}}{\displaystyle \frac{2m+5}{2i+1}}(\begin{array}{c}m+1i\\ 2i\end{array})\mathrm{\Delta }^iE^{m+13i}`$
$`+{\displaystyle \frac{2m+5}{2(l+1)+1}}(\begin{array}{c}m+1(l+1)\\ 2(l+1)\end{array})\mathrm{\Delta }^{l+1}E^{m+13(l+1)}`$
$`=`$ $`{\displaystyle \underset{i=0}{\overset{l+1}{}}}{\displaystyle \frac{2m+5}{2i+1}}(\begin{array}{c}m+1i\\ 2i\end{array})\mathrm{\Delta }^iE^{m+13i}`$
$`=`$ $`{\displaystyle \underset{i=0}{\overset{[|(m+1)/3|]}{}}}{\displaystyle \frac{2m+5}{2i+1}}(\begin{array}{c}m+1i\\ 2i\end{array})\mathrm{\Delta }^iE^{m+13i}.(2.40)`$
Thus (2.35) holds for $`n=m+1`$. Therefore, (2.34) holds for any $`n_+`$ by induction on $`n.\mathrm{}`$
## 3 Theta Series and Weighted Symmetric Polynomials
In this section, we shall determine the theta series of certain infinite families of positive definite even unimodular lattices containing a sublattice of the same rank and isomorphic to the direct sum of finite copies of the lattices $`R_{D_{2n}}`$ with various $`n`$. The theta series of these lattices are weighted symmetric polynomials of the functions of $`\{\mathrm{\Delta }_{24}(z),h_n(z),\rho _n(z)n_+\}`$. By (1.15) and (1.16), we essentially determine the theta series of these lattices as polynomials of the well-known Essenstein series $`E_4(z)`$ and Ramanujan series $`\mathrm{\Delta }_{24}(z)`$.
Let $`k`$ be a positive integer. Suppose that $`\{n_1,\mathrm{},n_k\}_+`$ is a subset. Set
$$\overline{n}_0=0,\overline{n}_i=\underset{j=1}{\overset{i}{}}n_i,n=\overline{n}_k.$$
$`(3.1)`$
Recall the Euclidean space $`^n`$ with inner product (1.8). All the lattices in this section have the symmetric bilinear form inherited from this inner product. Set
$$Q=\{\stackrel{}{\alpha }=(\alpha _1,\mathrm{},\alpha _n)^n\underset{j=1}{\overset{n_i}{}}\alpha _{\overline{n}_{i1}+j}2\text{for}i\overline{1,k}\}^n.$$
$`(3.2)`$
Then $`Q`$ is a lattice that isomorphic to $`R_{D_{n_1}}\mathrm{}R_{D_{n_k}}`$. For $`j`$, we denote
$$\mathrm{𝟎}_j=(0,\mathrm{},0),\mathbf{\hspace{0.33em}\hspace{0.33em}1}_j=(1,\mathrm{},1)^j.$$
$`(3.3)`$
Set
$$x_1^i=\frac{1}{2}(\mathrm{𝟎}_{\overline{n}_{i1}},\mathrm{𝟏}_{n_i},\mathrm{𝟎}_{n\overline{n}_i}),x_2^i=(\mathrm{𝟎}_{\overline{n}_{i1}},1,\mathrm{𝟎}_{n\overline{n}_{i1}1}),x_3^i=x_1^i+x_2^i^n$$
$`(3.4)`$
for $`i=1,\mathrm{},k`$. When $`k=1`$, we have
$$\mathrm{\Theta }_Q(z)=\mathrm{\Theta }_{R_{D_{n_1}}}=\frac{1}{2}(\vartheta _3(z)^{n_1}+\vartheta _4(z)^{n_1}),\mathrm{\Theta }_{x_1^1+Q}(z)=\mathrm{\Theta }_{x_3^1+Q}(z)=\frac{1}{2}\vartheta _2(z)^{n_1},$$
$`(3.5)`$
$$\mathrm{\Theta }_{x_2^1+Q}(z)=\frac{1}{2}(\vartheta _3(z)^{n_1}\vartheta _4(z)^{n_1})$$
$`(3.6)`$
(cf. (1.12) and Chapter 4 of \[CS3\]). Moreover,
$$2x_1^1,2x_2^1Q$$
$`(3.7)`$
when $`n_1`$ is even. In fact, we have
$$\{\stackrel{}{\alpha }^n\stackrel{}{\alpha },\stackrel{}{\beta }\text{for}\stackrel{}{\beta }Q\}=\underset{i=1}{\overset{k}{}}(x_1^i+x_2^i)+Q$$
$`(3.8)`$
(eg., cf. \[CS3\] or \[X3\]).
Set
$$L=\underset{i=1}{\overset{k}{}}(x_1^i+x_2^1+\mathrm{}+x_2^{i1}+x_2^{i+1}+\mathrm{}+x_2^k)+Q.$$
$`(3.9)`$
When $`k=1`$,
$$L=x_1^1+Q.$$
$`(3.10)`$
Theorem 3.1. If $`k=2\mathrm{}+1`$ is an odd positive integer and $`n_i=8m_i`$ with $`\mathrm{}`$ and $`m_i_+`$, then $`L`$ is a positive definite even unimodular lattice with the theta series
$`2^{2\mathrm{}+1}\mathrm{\Theta }_L(z)`$
$`=`$ $`{\displaystyle \underset{1j_1j_2;j_1+j_2\mathrm{}}{}}\text{sym}\{h_{m_1+\mathrm{}+m_{2j_1}}(z)h_{m_{2j_1+1}+\mathrm{}+m_{2(j_1+j_2)}}(z)h_{m_{2(j_1+j_2)+1}+\mathrm{}+m_{2\mathrm{}+1}}(z)\}`$
$`{\displaystyle \underset{0j_1j_2\mathrm{}j_1j_21}{}}\text{sym}\{h_{m_1+\mathrm{}+m_{2j_1+1}}(z)h_{m_{2j_1+2}+\mathrm{}+m_{2(j_1+j_2+1)}}(z)`$
$`\times h_{m_{2(j_1+j_2)+3}+\mathrm{}+m_{2\mathrm{}+1}}(z)\}+{\displaystyle }_{j=1}^{\mathrm{}}(34^\mathrm{}j)\text{sym}\{h_{m_1+\mathrm{}+m_{2j}}(z)h_{m_{2j+1}+\mathrm{}+m_{2\mathrm{}+1}}(z)\}`$
$`+\left(434^{\mathrm{}}+{\displaystyle \frac{2}{3}}{\displaystyle \underset{j_1+j_2\mathrm{}1}{}}(\begin{array}{c}2\mathrm{}+1\\ 2j_1+1,2j_2+1\end{array})\right)h_{m_1+\mathrm{}+m_{2\mathrm{}+1}}(z)(3.11)`$
If $`k=2\mathrm{}`$ is an even positive integer and $`n_i=8m_i+4`$ with $`\mathrm{}_+`$ and $`m_i`$, then $`L`$ is a positive definite even unimodular lattice with the theta series
$`2^2\mathrm{}\mathrm{\Theta }_L(z)`$
$`=`$ $`{\displaystyle \underset{1j_1j_2\mathrm{}j_1j_2}{}}\text{sym}\{h_{m_1+\mathrm{}+m_{2j_1}+j_1}(z)h_{m_{2j_1+1}+\mathrm{}+m_{2(j_1+j_2)}+j_2}(z)`$
$`\times h_{m_{2(j_1+j_2)+1}+\mathrm{}+m_2\mathrm{}+\mathrm{}j_1j_2}(z)\}2^8\mathrm{\Delta }_{24}(z){\displaystyle }_{1j_1,j_2;j_1+j_2\mathrm{}2}\text{sym}\{\rho _{m_1+\mathrm{}+m_{2j_1+1}+j_11}(z)`$
$`\times \rho _{m_{2j_1+2}+\mathrm{}+m_{2(j_1+j_2+1)}+j_21}(z)h_{m_{2(j_1+j_2)+3}+\mathrm{}+m_2\mathrm{}+\mathrm{}j_1j_21}(z)\}`$
$`+3{\displaystyle \underset{j=1}{\overset{[|\mathrm{}/2|]}{}}}\text{sym}\{h_{m_1+\mathrm{}+m_{2j}+j}(z)h_{m_{2j+1}+\mathrm{}+m_2\mathrm{}+\mathrm{}j}(z)\}+2^8\mathrm{\Delta }_{24}(z)`$
$`\times {\displaystyle \underset{j=0}{\overset{[|(\mathrm{}1)/2|]}{}}}(2^{2j}+2^{2(\mathrm{}j1)}3)\text{sym}\{\rho _{m_1+\mathrm{}+m_{2j+1}+j1}(z)\rho _{m_{2j+2}+\mathrm{}+m_2\mathrm{}+\mathrm{}j2}(z)\}`$
$`+\left(4{\displaystyle \frac{2}{3}}{\displaystyle \underset{j_1+j_2\mathrm{}}{}}(\begin{array}{c}2\mathrm{}\\ 2j_1,2j_2\end{array})\right)h_{m_1+\mathrm{}+m_2\mathrm{}+\mathrm{}}(z)(3.12)`$
Proof. First we consider the case that $`k=2\mathrm{}+1`$ is an odd positive number and $`n_i=8m_i`$ with $`\mathrm{}`$ and $`m_i_+`$. The lattice $`L`$ is unimodular by (1.11) and (3.7)-(3.9). Before we prove (3.11), we need some combinatorial facts as follows. For $`j_+`$, we have
$$\underset{i=0}{\overset{[|j/2|]}{}}(\begin{array}{c}j\\ 2i\end{array})\pm \underset{i=0}{\overset{[|(j1)/2|]}{}}(\begin{array}{c}j\\ 2i+1\end{array})=(1\pm 1)^j,$$
$`(3.13)`$
which implies
$$\underset{i=0}{\overset{[|j/2|]}{}}(\begin{array}{c}j\\ 2i\end{array})=\underset{i=0}{\overset{[|(j1)/2|]}{}}(\begin{array}{c}j\\ 2i+1\end{array})=2^{j1}.$$
$`(3.14)`$
Thus
$$\underset{i=1}{\overset{[|j/2|]}{}}(\begin{array}{c}j\\ 2i\end{array})=2^{j1}1.$$
$`(3.15)`$
Moreover,
$$\underset{i=0}{\overset{j}{}}(\begin{array}{c}j\\ 2i\end{array})2^{j2i}=\frac{(1+2)^j+(1+2)^j}{2}=\frac{3^j+1}{2}.$$
$`(3.16)`$
In particular,
$$\underset{i=1}{\overset{\mathrm{}}{}}(\begin{array}{c}2\mathrm{}+1\\ 2i\end{array})4^\mathrm{}i=\frac{3^{2\mathrm{}+1}+1}{4}4^{\mathrm{}}.$$
$`(3.17)`$
Furthermore,
$`{\displaystyle \underset{j_1+j_2\mathrm{}}{}}(\begin{array}{c}2\mathrm{}+1\\ 2j_1,2j_2\end{array})+{\displaystyle \underset{j_1+j_2\mathrm{}1}{}}(\begin{array}{c}2\mathrm{}+1\\ 2j_1+1,2j_2+1\end{array})`$ $`=`$ $`{\displaystyle \frac{(1+1+1)^{2\mathrm{}+1}+(11+1)^{2\mathrm{}+1}}{2}}`$
$`=`$ $`{\displaystyle \frac{3^{2\mathrm{}+1}1}{2}}.(3.18)`$
The following countings are also needed. For $`\mathrm{}>j_+`$, we have
$`\text{the number of minomials in the expression}{\displaystyle \underset{i=0}{\overset{[|(j1)/2|]}{}}}\text{sym}\{h_{m_1+\mathrm{}+m_{2i+1}}h_{m_{2i+2}++m_{2j}}\}`$
$`=`$ $`{\displaystyle \frac{1}{2}}{\displaystyle \underset{i=0}{\overset{j1}{}}}(\begin{array}{c}2j\\ 2i+1\end{array}),(3.19)`$
$`\text{the number of minomials in the expression}{\displaystyle \underset{i=1}{\overset{[|j/2|]}{}}}\text{sym}\{h_{m_1+\mathrm{}+m_{2i}}h_{m_{2i+1}++m_{2j}}\}`$
$`=`$ $`{\displaystyle \frac{1}{2}}{\displaystyle \underset{i=1}{\overset{j1}{}}}(\begin{array}{c}2j\\ 2i\end{array}).(3.20)`$
Moreover,
the number of minomials in the expression
$`{\displaystyle \underset{1j_1j_2;j_1+j_2\mathrm{}}{}}\text{sym}\{h_{m_1+\mathrm{}+m_{2j_1}}(z)h_{m_{2j_1+1}+\mathrm{}+m_{2(j_1+j_2)}}(z)h_{m_{2(j_1+j_2)+1}+\mathrm{}+m_{2\mathrm{}+1}}(z)\}`$
$`=`$ $`{\displaystyle \frac{1}{2}}{\displaystyle \underset{1j_1,j_2;j_1+j_2\mathrm{}}{}}(\begin{array}{c}2\mathrm{}+1\\ 2j_1,2j_2\end{array}),(3.21)`$
$`\text{the number of minomials in the expression}{\displaystyle \underset{0j_1j_2\mathrm{}j_1j_21}{}}`$
$`\text{sym}\{h_{m_1+\mathrm{}+m_{2j_1+1}}(z)h_{m_{2j_1+2}+\mathrm{}+m_{2(j_1+j_2+1)}}(z)h_{m_{2(j_1+j_2)+3}+\mathrm{}+m_{2\mathrm{}+1}}(z)\}`$
$`=`$ $`{\displaystyle \frac{1}{6}}{\displaystyle \underset{j_1+j_2\mathrm{}1}{}}(\begin{array}{c}2\mathrm{}+1\\ 2j_1+1,2j_2+1\end{array}).(3.22)`$
Furthermore,
$`{\displaystyle \underset{j_1+j_2\mathrm{}}{}}(\begin{array}{c}2\mathrm{}+1\\ 2j_1,2j_2\end{array})`$
$`=`$ $`2{\displaystyle \underset{i=0}{\overset{\mathrm{}}{}}}(\begin{array}{c}2\mathrm{}+1\\ 2i\end{array})1+{\displaystyle \underset{1j_1j_2;j_1+j_2\mathrm{}}{}}(\begin{array}{c}2\mathrm{}+1\\ 2j_1,2j_2\end{array})`$
$`=`$ $`2^{2\mathrm{}+1}1+{\displaystyle \underset{1j_1j_2;j_1+j_2\mathrm{}}{}}(\begin{array}{c}2\mathrm{}+1\\ 2j_1,2j_2\end{array}).(3.23)`$
By (1.12), (3.5)-(3.7), (3.13)-(3.23) and calculating the theta series of the following cosets of $`Q`$:
$$Q,\underset{i=1}{\overset{2j}{}}(x_1^i+x_2^1+\mathrm{}+x_2^{i1}+x_2^{i+1}+\mathrm{}+x_2^k)+Q,$$
$`(3.24)`$
and
$$\underset{i=1}{\overset{2j^{}+1}{}}(x^i+x_2^1+\mathrm{}+x_2^{i1}+x_2^{i+1}+\mathrm{}+x_2^k)+Q$$
$`(3.25)`$
for $`1j\mathrm{}`$ and $`0j^{}\mathrm{}`$, we obtain
$`2^{2\mathrm{}+1}\mathrm{\Theta }_L(z)`$
$`=`$ $`{\displaystyle \underset{j=0}{\overset{\mathrm{}}{}}}\text{sym}\{\vartheta _2(z)^{n_1+\mathrm{}+n_{2j}}{\displaystyle \underset{p=2j+1}{\overset{2\mathrm{}+1}{}}}(\vartheta _3(z)^{n_p}+\vartheta _4(z)^{n_p})\}`$
$`+{\displaystyle \underset{j=0}{\overset{\mathrm{}}{}}}\text{sym}\{\vartheta _2(z)^{n_1+\mathrm{}+n_{2j+1}}{\displaystyle \underset{p=2j+2}{\overset{2\mathrm{}+1}{}}}(\vartheta _3(z)^{n_p}\vartheta _4(z)^{n_p})\}`$
$`=`$ $`{\displaystyle \underset{1j_1j_2;j_1+j_2\mathrm{}}{}}[\text{sym}\{\vartheta _2(z)^{n_1+\mathrm{}+n_{2j_1}}\vartheta _3(z)^{n_{2j_1+1}+\mathrm{}+n_{2(j_1+j_2)}}\vartheta _4(z)^{n_{2(j_1+j_2)+1}+\mathrm{}+n_{2\mathrm{}+1}}\}`$
$`+\text{sym}\{\vartheta _3(z)^{n_1+\mathrm{}+n_{2j_1}}\vartheta _4(z)^{n_{2j_1+1}+\mathrm{}+n_{2(j_1+j_2)}}\vartheta _2(z)^{n_{2(j_1+j_2)+1}+\mathrm{}+n_{2\mathrm{}+1}}\}`$
$`+\text{sym}\{\vartheta _4(z)^{n_1+\mathrm{}+n_{2j_1}}\vartheta _2(z)^{n_{2j_1+1}+\mathrm{}+n_{2(j_1+j_2)}}\vartheta _3(z)^{n_{2(j_1+j_2)+1}+\mathrm{}+n_{2\mathrm{}+1}}\}]`$
$`{\displaystyle \underset{0j_1j_2\mathrm{}j_1j_21}{}}\text{sym}\{\vartheta _2(z)^{n_1+\mathrm{}+n_{2j_1+1}}\vartheta _3(z)^{n_{2j_1+2}+\mathrm{}+n_{2(j_1+j_2+1)}}`$
$`\times \vartheta _4(z)^{n_{2(j_1+j_2)+3}+\mathrm{}+n_{2\mathrm{}+1}}\}+{\displaystyle }_{j=1}^{\mathrm{}}[\text{sym}\{\vartheta _2(z)^{n_1+\mathrm{}+n_{2j}}(\vartheta _3(z)^{n_{2j+1}+\mathrm{}+n_{2\mathrm{}+1}}`$
$`+\vartheta _4(z)^{n_{2j+1}+\mathrm{}+n_{2\mathrm{}+1}})\}+\text{sym}\{\vartheta _3(z)^{n_1+\mathrm{}+n_{2j}}(\vartheta _2(z)^{n_{2j+1}+\mathrm{}+n_{2\mathrm{}+1}}`$
$`+\vartheta _4(z)^{n_{2j+1}+\mathrm{}+n_{2\mathrm{}+1}})\}+\text{sym}\{\vartheta _4(z)^{n_1+\mathrm{}+n_{2j}}(\vartheta _3(z)^{n_{2j+1}+\mathrm{}+n_{2\mathrm{}+1}}`$
$`+\vartheta _2(z)^{n_{2j+1}+\mathrm{}+n_{2\mathrm{}+1}})\}]+\vartheta _2(z)^{n_1+\mathrm{}+n_{2\mathrm{}+1}}+\vartheta _3(z)^{n_1+\mathrm{}+n_{2\mathrm{}+1}}+\vartheta _4(z)^{n_1+\mathrm{}+n_{2\mathrm{}+1}}`$
$`=`$ $`{\displaystyle \underset{1j_1j_2;j_1+j_2\mathrm{}}{}}\text{sym}\{(\vartheta _2(z)^{n_1+\mathrm{}+n_{2j_1}}+\vartheta _3(z)^{n_1+\mathrm{}+n_{2j_1}}+\vartheta _4(z)^{n_1+\mathrm{}+n_{2j_1}})`$
$`\times (\vartheta _2(z)^{n_{2j_1+1}+\mathrm{}+n_{2(j_1+j_2)}}+\vartheta _3(z)^{n_{2j_1+1}+\mathrm{}+n_{2(j_1+j_2)}}+\vartheta _4(z)^{n_{2j_1+1}+\mathrm{}+n_{2(j_1+j_2)}})`$
$`\times (\vartheta _2(z)^{n_{2(j_1+j_2)+1}+\mathrm{}+n_{2\mathrm{}+1}}+\vartheta _3(z)^{n_{2(j_1+j_2)+1}+\mathrm{}+n_{2\mathrm{}+1}}+\vartheta _4(z)^{n_{2(j_1+j_2)+1}+\mathrm{}+n_{2\mathrm{}+1}})\}`$
$`{\displaystyle \underset{0j_1j_2\mathrm{}j_1j_21}{}}\text{sym}\{(\vartheta _2(z)^{n_1+\mathrm{}+n_{2j_1+1}}+\vartheta _3(z)^{n_1+\mathrm{}+n_{2j_1+1}}+\vartheta _4(z)^{n_1+\mathrm{}+n_{2j_1+1}})`$
$`\times (\vartheta _2(z)^{n_{2j_1+2}+\mathrm{}+n_{2(j_1+j_2+1)}}+\vartheta _3(z)^{n_{2j_1+2}+\mathrm{}+n_{2(j_1+j_2+1)}}+\vartheta _4(z)^{n_{2j_1+2}+\mathrm{}+n_{2(j_1+j_2+1)}})`$
$`\times (\vartheta _2(z)^{n_{2(j_1+j_2)+3}+\mathrm{}+n_{2\mathrm{}+1}}+\vartheta _3(z)^{n_{2(j_1+j_2)+3}+\mathrm{}+n_{2\mathrm{}+1}}+\vartheta _4(z)^{n_{2(j_1+j_2)+3}+\mathrm{}+n_{2\mathrm{}+1}})\}`$
$`+{\displaystyle \underset{j=1}{\overset{\mathrm{}}{}}}\left(1+{\displaystyle \frac{1}{2}}{\displaystyle \underset{i=0}{\overset{j1}{}}}(\begin{array}{c}2j\\ 2i+1\end{array}){\displaystyle \frac{1}{2}}{\displaystyle \underset{i=1}{\overset{j1}{}}}(\begin{array}{c}2j\\ 2i\end{array}){\displaystyle \underset{i=1}{\overset{\mathrm{}j}{}}}(\begin{array}{c}2(\mathrm{}j)+1\\ 2i\end{array})\right)`$
$`\times [\text{sym}\{\vartheta _2(z)^{n_1+\mathrm{}+n_{2j}}(\vartheta _3(z)^{n_{2j+1}+\mathrm{}+n_{2\mathrm{}+1}}+\vartheta _4(z)^{n_{2j+1}+\mathrm{}+n_{2\mathrm{}+1}})\}`$
$`+\text{sym}\{\vartheta _3(z)^{n_1+\mathrm{}+n_{2j}}(\vartheta _2(z)^{n_{2j+1}+\mathrm{}+n_{2\mathrm{}+1}}+\vartheta _4(z)^{n_{2j+1}+\mathrm{}+n_{2\mathrm{}+1}})\}`$
$`+\text{sym}\{\vartheta _4(z)^{n_1+\mathrm{}+n_{2j}}(\vartheta _3(z)^{n_{2j+1}+\mathrm{}+n_{2\mathrm{}+1}}+\vartheta _4(z)^{n_{2j+1}+\mathrm{}+n_{2\mathrm{}+1}})]`$
$`+\left(1{\displaystyle \frac{1}{2}}{\displaystyle \underset{1j_1,j_2;j_1+j_2\mathrm{}}{}}(\begin{array}{c}2\mathrm{}+1\\ 2j_1,2j_2\end{array})+{\displaystyle \frac{1}{6}}{\displaystyle \underset{j_1+j_2\mathrm{}1}{}}(\begin{array}{c}2\mathrm{}+1\\ 2j_1+1,2j_2+1\end{array})\right)`$
$`\times (\vartheta _2(z)^{n_1+\mathrm{}+n_{2\mathrm{}+1}}+\vartheta _3(z)^{n_1+\mathrm{}+n_{2\mathrm{}+1}}+\vartheta _4(z)^{n_1+\mathrm{}+n_{2\mathrm{}+1}})`$
$`=`$ $`{\displaystyle \underset{1j_1j_2;j_1+j_2\mathrm{}}{}}\text{sym}\{h_{m_1+\mathrm{}+m_{2j_1}}(z)h_{m_{2j_1+1}+\mathrm{}+m_{2(j_1+j_2)}}(z)h_{m_{2(j_1+j_2)+1}+\mathrm{}+m_{2\mathrm{}+1}}(z)\}`$
$`{\displaystyle \underset{0j_1j_2\mathrm{}j_1j_21}{}}\text{sym}\{h_{m_1+\mathrm{}+m_{2j_1+1}}(z)h_{m_{2j_1+2}+\mathrm{}+m_{2(j_1+j_2+1)}}(z)`$
$`\times h_{m_{2(j_1+j_2)+3}+\mathrm{}+m_{2\mathrm{}+1}}(z)\}+{\displaystyle }_{j=1}^{\mathrm{}}(34^\mathrm{}j)\text{sym}\{(\vartheta _2(z)^{n_1+\mathrm{}+n_{2j}}+\vartheta _3(z)^{n_1+\mathrm{}+n_{2j}}`$
$`+\vartheta _4(z)^{n_1+\mathrm{}+n_{2j}})(\vartheta _2(z)^{n_{2j+1}+\mathrm{}+n_{2\mathrm{}+1}}+\vartheta _3(z)^{n_{2j+1}+\mathrm{}+n_{2\mathrm{}+1}}+\vartheta _4(z)^{n_{2j+1}+\mathrm{}+n_{2\mathrm{}+1}})\}`$
$`+[2^2\mathrm{}+{\displaystyle \frac{1}{2}}{\displaystyle \frac{1}{2}}{\displaystyle \underset{j_1+j_2\mathrm{}}{}}(\begin{array}{c}2\mathrm{}+1\\ 2j_1,2j_2\end{array})+{\displaystyle \frac{1}{6}}{\displaystyle \underset{j_1+j_2\mathrm{}1}{}}(\begin{array}{c}2\mathrm{}+1\\ 2j_1+1,2j_2+1\end{array}){\displaystyle \underset{j=1}{\overset{\mathrm{}}{}}}(\begin{array}{c}2\mathrm{}+1\\ 2j\end{array})`$
$`\times (34^\mathrm{}j)](\vartheta _2(z)^{n_1+\mathrm{}+n_{2\mathrm{}+1}}+\vartheta _3(z)^{n_1+\mathrm{}+n_{2\mathrm{}+1}}+\vartheta _4(z)^{n_1+\mathrm{}+n_{2\mathrm{}+1}})`$
$`=`$ $`{\displaystyle \underset{1j_1j_2;j_1+j_2\mathrm{}}{}}\text{sym}\{h_{m_1+\mathrm{}+m_{2j_1}}(z)h_{m_{2j_1+1}+\mathrm{}+m_{2(j_1+j_2)}}(z)h_{m_{2(j_1+j_2)+1}+\mathrm{}+m_{2\mathrm{}+1}}(z)\}`$
$`{\displaystyle \underset{0j_1j_2\mathrm{}j_1j_21}{}}\text{sym}\{h_{m_1+\mathrm{}+m_{2j_1+1}}(z)h_{m_{2j_1+2}+\mathrm{}+m_{2(j_1+j_2+1)}}(z)`$
$`\times h_{m_{2(j_1+j_2)+3}+\mathrm{}+m_{2\mathrm{}+1}}(z)\}+{\displaystyle }_{j=1}^{\mathrm{}}(34^\mathrm{}j)\text{sym}\{h_{m_1+\mathrm{}+m_{2j}}(z)h_{m_{2j+1}+\mathrm{}+m_{2\mathrm{}+1}}(z)\}`$
$`+\left(434^{\mathrm{}}+{\displaystyle \frac{2}{3}}{\displaystyle \underset{j_1+j_2\mathrm{}1}{}}(\begin{array}{c}2\mathrm{}+1\\ 2j_1+1,2j_2+1\end{array})\right)h_{m_1+\mathrm{}+m_{2\mathrm{}+1}}(z)(3.26)`$
Next we consider the case that $`k=2\mathrm{}`$ is an even positive number and $`n_i=8m_i+4`$. $`L`$ is unimodular again by (1.11) and (3.7)-(3.9). We need the combinatorial facts as follows. First,
$`{\displaystyle \underset{j=0}{\overset{\mathrm{}1}{}}}(\begin{array}{c}2\mathrm{}\\ 2j+1\end{array})(2^{2j}+2^{2(\mathrm{}j1)})={\displaystyle \underset{j=0}{\overset{\mathrm{}1}{}}}(\begin{array}{c}2\mathrm{}\\ 2j+1\end{array})2^{2j+1}`$
$`=`$ $`{\displaystyle \frac{(1+2)^2\mathrm{}(1+2)^2\mathrm{}}{2}}={\displaystyle \frac{3^2\mathrm{}1}{2}}.(3.27)`$
Moreover,
$`{\displaystyle \underset{j_1+j_2\mathrm{}}{}}(\begin{array}{c}2\mathrm{}\\ 2j_1,2j_2\end{array})`$ $`=`$ $`3{\displaystyle \underset{j=0}{\overset{\mathrm{}}{}}}(\begin{array}{c}2\mathrm{}\\ 2j\end{array})3+{\displaystyle \underset{1j_1,j_2;j_1+j_2\mathrm{}1}{}}(\begin{array}{c}2\mathrm{}\\ 2j_1,2j_2\end{array})`$
$`=`$ $`3(2^{2\mathrm{}1}1)+{\displaystyle \underset{1j_1,j_2;j_1+j_2\mathrm{}1}{}}(\begin{array}{c}2\mathrm{}\\ 2j_1,2j_2\end{array}),(3.28)`$
$`{\displaystyle \underset{j_1+j_2\mathrm{}1}{}}(\begin{array}{c}2\mathrm{}\\ 2j_1+1,2j_2+1\end{array})`$ $`=`$ $`{\displaystyle \underset{j=0}{\overset{\mathrm{}1}{}}}(\begin{array}{c}2\mathrm{}\\ 2j+1\end{array})+{\displaystyle \underset{j_1+j_2\mathrm{}2}{}}(\begin{array}{c}2\mathrm{}\\ 2j_1+1,2j_2+1\end{array})`$
$`=`$ $`2^{2\mathrm{}1}+{\displaystyle \underset{j_1+j_2\mathrm{}2}{}}(\begin{array}{c}2\mathrm{}\\ 2j_1+1,2j_2+1\end{array})(3.29)`$
by (3.14). Furthermore,
$`{\displaystyle \underset{j_1+j_2\mathrm{}}{}}(\begin{array}{c}2\mathrm{}\\ 2j_1,2j_2\end{array})+{\displaystyle \underset{j_1+j_2\mathrm{}1}{}}(\begin{array}{c}2\mathrm{}\\ 2j_1+1,2j_2+1\end{array})`$ $`=`$ $`{\displaystyle \frac{(1+1+1)^2\mathrm{}+(11+1)^2\mathrm{}}{2}}`$
$`=`$ $`{\displaystyle \frac{3^2\mathrm{}1}{2}}.(3.30)`$
The following countings are also needed. Note
$`\text{the number of minomials in the expression}{\displaystyle \underset{1j_1j_2;j_1+j_2\mathrm{}1}{}}`$
$`\text{sym}\{h_{m_1+\mathrm{}+m_{2j_1}}(z)h_{m_{2j_1+1}+\mathrm{}+m_{2(j_1+j_2)}}(z)h_{m_{2(j_1+j_2)+1}+\mathrm{}+m_2\mathrm{}}(z)\}`$
$`=`$ $`{\displaystyle \frac{1}{6}}{\displaystyle \underset{1j_1,j_2;j_1+j_2\mathrm{}1}{}}(\begin{array}{c}2\mathrm{}\\ 2j_1,2j_2\end{array}),(3.31)`$
$`\text{the number of minomials in the expression}{\displaystyle \underset{0j_1j_2;j_1+j_2\mathrm{}2}{}}`$
$`\text{sym}\{h_{m_1+\mathrm{}+m_{2j_1+1}}(z)h_{m_{2j_1+2}+\mathrm{}+m_{2(j_1+j_2+1)}}(z)h_{m_{2(j_1+j_2)+3}+\mathrm{}+m_2\mathrm{}}(z)\}`$
$`=`$ $`{\displaystyle \frac{1}{2}}{\displaystyle \underset{j_1+j_2\mathrm{}2}{}}(\begin{array}{c}2\mathrm{}\\ 2j_1+1,2j_2+1\end{array}).(3.32)`$
By (3.5), (3.6), (3.13)-(3.15), (3.19), (3.20), (3.27)-(3.32) and calculating the theta series of the cosets in $`L/Q`$ by (3.5)-(3.7), we have
$`2^{2\mathrm{}+1}\mathrm{\Theta }_L(z)`$
$`=`$ $`{\displaystyle \underset{j=0}{\overset{\mathrm{}}{}}}\text{sym}\{\vartheta _2(z)^{n_1+\mathrm{}+n_{2j}}{\displaystyle \underset{p=2j+1}{\overset{2\mathrm{}}{}}}(\vartheta _3(z)^{n_p}+\vartheta _4(z)^{n_p})\}`$
$`+{\displaystyle \underset{j=0}{\overset{\mathrm{}1}{}}}\text{sym}\{\vartheta _2(z)^{n_1+\mathrm{}+n_{2j+1}}{\displaystyle \underset{p=2j+2}{\overset{2\mathrm{}}{}}}(\vartheta _3(z)^{n_p}\vartheta _4(z)^{n_p})\}`$
$`=`$ $`{\displaystyle \underset{0j_1j_2\mathrm{}j_1j_2}{}}\text{sym}\{\vartheta _2(z)^{n_1+\mathrm{}+n_{2j_1}}\vartheta _3(z)^{n_{2j_1+1}+\mathrm{}+n_{2(j_1+j_2)}}\vartheta _4(z)^{n_{2(j_1+j_2)+1}+\mathrm{}+n_2\mathrm{}}\}`$
$`+{\displaystyle \underset{0j_1j_2;j_1+j_2\mathrm{}1}{}}\text{sym}\{\vartheta _2(z)^{n_1+\mathrm{}+n_{2j_1+1}}\vartheta _3(z)^{n_{2j_1+2}+\mathrm{}+n_{2(j_1+j_2+1)}}\vartheta _4(z)^{n_{2(j_1+j_2)+3}+\mathrm{}+n_2\mathrm{}}\}`$
$`{\displaystyle \underset{0j_1j_2;j_1+j_2\mathrm{}1}{}}\text{sym}\{\vartheta _2(z)^{n_1+\mathrm{}+n_{2j_1+1}}\vartheta _4(z)^{n_{2j_1+2}+\mathrm{}+n_{2(j_1+j_2+1)}}\vartheta _3(z)^{n_{2(j_1+j_2)+3}+\mathrm{}+n_2\mathrm{}}\}`$
$`+{\displaystyle \underset{0j_1j_2;j_1+j_2\mathrm{}1}{}}\text{sym}\{\vartheta _3(z)^{n_1+\mathrm{}+n_{2j_1+1}}\vartheta _4(z)^{n_{2j_1+2}+\mathrm{}+n_{2(j_1+j_2+1)}}\vartheta _2(z)^{n_{2(j_1+j_2)+3}+\mathrm{}+n_2\mathrm{}}\}`$
$`=`$ $`{\displaystyle \underset{1j_1j_2\mathrm{}j_1j_2}{}}\text{sym}\{(\vartheta _2(z)^{n_1+\mathrm{}+n_{2j_1}}+\vartheta _3(z)^{n_1+\mathrm{}+n_{2j_1}}+\vartheta _4(z)^{n_1+\mathrm{}+n_{2j_1}})`$
$`\times (\vartheta _2(z)^{n_{2j_1+1}+\mathrm{}+n_{2(j_1+j_2)}}+\vartheta _3(z)^{n_{2j_1+1}+\mathrm{}+n_{2(j_1+j_2)}}+\vartheta _4(z)^{n_{2j_1+1}+\mathrm{}+n_{2(j_1+j_2)}})`$
$`\times (\vartheta _2(z)^{n_{2(j_1+j_2)+1}+\mathrm{}+n_2\mathrm{}}+\vartheta _3(z)^{n_{2(j_1+j_2)+1}+\mathrm{}+n_2\mathrm{}}+\vartheta _4(z)^{n_{2(j_1+j_2)+1}+\mathrm{}+n_2\mathrm{}})\}`$
$`{\displaystyle \underset{0j_1j_2;j_1+j_2\mathrm{}2}{}}\text{sym}\{(\vartheta _3(z)^{n_1+\mathrm{}+n_{2j_1+1}}\vartheta _2(z)^{n_1+\mathrm{}+n_{2j_1+1}}\vartheta _4(z)^{n_1+\mathrm{}+n_{2j_1+1}})`$
$`\times (\vartheta _3(z)^{n_{2j_1+2}+\mathrm{}+n_{2(j_1+j_2+1)}}\vartheta _2(z)^{n_{2j_1+2}+\mathrm{}+n_{2(j_1+j_2+1)}}\vartheta _4(z)^{n_{2j_1+2}+\mathrm{}+n_{2(j_1+j_2+1)}})`$
$`\times (\vartheta _2(z)^{n_{2(j_1+j_2)+3}+\mathrm{}+n_2\mathrm{}}+\vartheta _3(z)^{n_{2(j_1+j_2)+3}+\mathrm{}+n_2\mathrm{}}+\vartheta _4(z)^{n_{2(j_1+j_2)+3}+\mathrm{}+n_2\mathrm{}})\}`$
$`+3{\displaystyle \underset{j=1}{\overset{[|\mathrm{}/2|]}{}}}\text{sym}\{(\vartheta _2(z)^{n_1+\mathrm{}+n_{2j}}+\vartheta _3(z)^{n_1+\mathrm{}+n_{2j}}+\vartheta _4(z)^{n_1+\mathrm{}+n_{2j}})(\vartheta _2(z)^{n_{2j+1}+\mathrm{}+n_2\mathrm{}}`$
$`+\vartheta _3(z)^{n_{2j+1}+\mathrm{}+n_2\mathrm{}}+\vartheta _4(z)^{n_{2j+1}+\mathrm{}+n_2\mathrm{}})\}+{\displaystyle }_{j=0}^{[|(\mathrm{}1)/2|]}(2^{2j}+2^{2(\mathrm{}j1)}3)`$
$`\times \text{sym}\{(\vartheta _3(z)^{n_1+\mathrm{}+n_{2j+1}}\vartheta _2(z)^{n_1+\mathrm{}+n_{2j+1}}\vartheta _3(z)^{n_1+\mathrm{}+n_{2j+1}})(\vartheta _3(z)^{n_{2j+2}+\mathrm{}+n_2\mathrm{}}`$
$`\vartheta _2(z)^{n_{2j+2}+\mathrm{}+n_2\mathrm{}}\vartheta _4(z)^{n_{2j+2}+\mathrm{}+n_2\mathrm{}})\}+[1{\displaystyle \frac{1}{6}}{\displaystyle \underset{1j_1,j_2;j_1+j_2\mathrm{}1}{}}(\begin{array}{c}2\mathrm{}\\ 2j_1,2j_2\end{array})`$
$`+{\displaystyle \frac{1}{2}}{\displaystyle \underset{0j_1,j_2;j_1+j_2\mathrm{}2}{}}(\begin{array}{c}2\mathrm{}\\ 2j_1+1,2j_2+1\end{array}){\displaystyle \frac{3}{2}}{\displaystyle \underset{j=1}{\overset{\mathrm{}1}{}}}(\begin{array}{c}2\mathrm{}\\ 2j\end{array}){\displaystyle \frac{1}{2}}{\displaystyle \underset{i=0}{\overset{\mathrm{}1}{}}}(\begin{array}{c}2\mathrm{}\\ 2j+1\end{array})(2^{2j}`$
$`+2^{2(\mathrm{}j1)}3)](\vartheta _2(z)^{n_1+\mathrm{}+n_2\mathrm{}}+\vartheta _3(z)^{n_1+\mathrm{}+n_2\mathrm{}}+\vartheta _4(z)^{n_1+\mathrm{}+n_2\mathrm{}})`$
$`=`$ $`{\displaystyle \underset{1j_1j_2\mathrm{}j_1j_2}{}}\text{sym}\{h_{m_1+\mathrm{}+m_{2j_1}+j_1}(z)h_{m_{2j_1+1}+\mathrm{}+m_{2(j_1+j_2)}+j_2}(z)`$
$`\times h_{m_{2(j_1+j_2)+1}+\mathrm{}+m_2\mathrm{}+\mathrm{}j_1j_2}(z)\}2^8\mathrm{\Delta }_{24}(z){\displaystyle }_{1j_1,j_2;j_1+j_2\mathrm{}2}\text{sym}\{\rho _{m_1+\mathrm{}+m_{2j_1+1}+j_11}(z)`$
$`\times \rho _{m_{2j_1+2}+\mathrm{}+m_{2(j_1+j_2+1)}+j_21}(z)h_{m_{2(j_1+j_2)+3}+\mathrm{}+m_2\mathrm{}+\mathrm{}j_1j_21}(z)\}`$
$`+3{\displaystyle \underset{j=1}{\overset{[|\mathrm{}/2|]}{}}}\text{sym}\{h_{m_1+\mathrm{}+m_{2j}+j}(z)h_{m_{2j+1}+\mathrm{}+m_2\mathrm{}+\mathrm{}j}(z)\}+2^8\mathrm{\Delta }_{24}(z)`$
$`\times {\displaystyle \underset{j=0}{\overset{[|(\mathrm{}1)/2|]}{}}}(2^{2j}+2^{2(\mathrm{}j1)}3)\text{sym}\{\rho _{m_1+\mathrm{}+m_{2j+1}+j1}(z)\rho _{m_{2j+2}+\mathrm{}+m_2\mathrm{}+\mathrm{}j2}(z)\}`$
$`+(4{\displaystyle \frac{2}{3}}{\displaystyle \underset{j_1+j_2\mathrm{}}{}}(\begin{array}{c}2\mathrm{}\\ 2j_1,2j_2\end{array}))h_{m_1+\mathrm{}+m_2\mathrm{}+\mathrm{}}(z).\mathrm{}(3.33)`$
According to the above theorem, we have the following theta series of the lattice (3.9) with the assumption on $`n_i`$ as in the theorem:
$$\mathrm{\Theta }_L(z)=\frac{1}{2}h_{m_1}(z)$$
$`(3.34)`$
when $`k=1`$;
$$\mathrm{\Theta }_L(z)=\frac{1}{2}h_{m_1+m_2+1}(z)64\mathrm{\Delta }_{24}(z)\rho _{m_11}(z)\rho _{m_21}(z)$$
$`(3.35)`$
when $`k=2`$;
$`\mathrm{\Theta }_L(z)`$ $`=`$ $`{\displaystyle \frac{1}{2}}h_{m_1+m_2+m_3}(z)+{\displaystyle \frac{1}{4}}(h_{m_1}(z)h_{m_2+m_3}(z)+h_{m_2}(z)h_{m_1+m_3}(z)`$
$`+h_{m_3}(z)h_{m_1+m_2}(z)){\displaystyle \frac{1}{8}}h_{m_1}(z)h_{m_2}(z)h_{m_3}(z)(3.36)`$
when $`k=3`$;
$`\mathrm{\Theta }_L(z)`$
$`=`$ $`16\mathrm{\Delta }_{24}(z)[2\text{sym}\{\rho _{m_11}(z)\rho _{m_2+m_3+m_4}(z)\}\text{sym}\{h_{m_1+m_2+1}(z)\rho _{m_31}(z)\rho _{m_41}(z)\}]`$
$`+{\displaystyle \frac{1}{16}}[3\text{sym}\{h_{m_1+m_2+1}(z)h_{m_3+m_4+1}(z)\}10h_{m_1+m_2+m_3+m_4+2}(z)](3.37)`$
when $`k=4`$;
$`\mathrm{\Theta }_L(z)`$ $`=`$ $`{\displaystyle \frac{1}{32}}[\text{sym}\{h_{m_1}(z)h_{m_2+m_3}(z)h_{m_4+m_5}(z)\}\text{sym}\{h_{m_1}(z)h_{m_2}(z)h_{m_3+m_4+m_5}(z)\}`$
$`2\text{sym}\{h_{m_1}(z)h_{m_2+m_3+m_4+m_5}(z)\}\text{sym}\{h_{m_1+m_2}(z)h_{m_3+m_4+m_5}(z)\}`$
$`4h_{m_1+m_2+m_4+m_5}(z)](3.38)`$
when $`k=5`$;
$`\mathrm{\Theta }_L(z)`$
$`=`$ $`{\displaystyle \frac{1}{64}}[118h_{m_1+m_2+m_3+m_4+m_5+m_6+3}(z)+3\text{sym}\{h_{m_1+m_2+1}(z)h_{m_3+m_4+m_5+m_6+2}(z)\}`$
$`+\text{sym}\{h_{m_1+m_2+1}(z)h_{m_3+m_4+1}(z)h_{m_5+m_6+1}(z)\}]4\mathrm{\Delta }_{24}(z)[\text{sym}\{h_{m_1+m_2+1}(z)`$
$`\times \rho _{m_31}(z)\rho _{m_4+m_5+m_6}(z)\}+\text{sym}\{h_{m_1+m_2+m_3+m_4+2}(z)\rho _{m_51}(z)\rho _{m_61}(z)\}`$
$`14\text{sym}\{\rho _{m_11}(z)\rho _{m_2+m_3+m_4+m_5+m_6+1}(z)\}`$
$`5\text{sym}\{\rho _{m_1+m_2+m_3}(z)\rho _{m_4+m_5+m_6}(z)\}](3.39)`$
when $`k=6`$. In particular, we have
$$\text{the theta series of Niemeier lattice of type}D_{24}=E_4^3(z)+384\mathrm{\Delta }_{24}(z)$$
$`(3.40)`$
by (3.34);
$$\text{the theta series of Niemeier lattice of type}D_{12}^2=E_4^3(z)192\mathrm{\Delta }_{24}(z)$$
$`(3.41)`$
by (3.35);
$$\text{the theta series of Niemeier lattice of type}D_8^3=E_4^3(z)384\mathrm{\Delta }_{24}(z)$$
$`(3.42)`$
by (3.36). We refer to \[N\] for the Niemeier lattices.
Remark 3.2. In the case $`k=2\mathrm{}`$ and $`n_1=n_2=\mathrm{}=n_2\mathrm{}=4`$, the lattice $`L`$ in Theorem 3.1 contains a sublattice isomorphic to $`R_{D_8\mathrm{}}`$, and $`\mathrm{\Theta }_L(z)=h_{\mathrm{}}(z)/2`$. The theta series $`\mathrm{\Theta }_L(z)`$ in (3.12) is different from that in (3.11) if not all $`m_i`$ are zero.
We go back to the settings in (3.1)-(3.8). Assume that $`k=4`$ and $`n_i=8m_i+4ϵ+2`$ with $`m_i`$ and $`ϵ=0,1`$. Set
$$L=(x_2^1+x_3^3+x_3^4)+(x_2^2+x_3^4+x_3^1)+(x_2^3+x_3^1+x_3^2)+(x_2^4+x_3^2+x_3^3)+Q$$
$`(3.43)`$
(cf. (3.4)).
Theorem 3.3. The lattice $`L`$ in (3.43) is a positive definite even unimodular lattice with the theta series
$`\mathrm{\Theta }_L(z)`$ $`=`$ $`{\displaystyle \frac{1}{2}}h_{m_1+m_2+m_3+m_4+2ϵ+1}(z)32\mathrm{\Delta }_{24}(z)(\rho _{m_1+m_2+ϵ1}(z)\rho _{m_3+m_4+ϵ1}(z)`$
$`+\rho _{m_1+m_3+ϵ1}(z)\rho _{m_2+m_4+ϵ1}(z)+\rho _{m_1+m_4+ϵ1}(z)\rho _{m_2+m_3+ϵ1}(z)).(3.44)`$
Proof. $`L`$ is unimodular by (1.11) and (3.7)-(3.9). By (3.5)-(3.7) and calculating the theta series of the cosets in $`L/Q`$, we obtain
$`2^4\mathrm{\Theta }_L(z)`$
$`=`$ $`\vartheta _2(z)^{n_3+n_4}(\vartheta _3(z)^{n_1}\vartheta _4(z)^{n_1})(\vartheta _3(z)^{n_2}+\vartheta _4(z)^{n_2})+\vartheta _2(z)^{n_1+n_4}(\vartheta _3(z)^{n_2}\vartheta _4(z)^{n_2})`$
$`\times (\vartheta _3(z)^{n_3}+\vartheta _4(z)^{n_3})+\vartheta _2(z)^{n_1+n_2}(\vartheta _3(z)^{n_3}\vartheta _4(z)^{n_3})(\vartheta _3(z)^{n_4}+\vartheta _4(z)^{n_4})`$
$`+\vartheta _2(z)^{n_2+n_3}(\vartheta _3(z)^{n_4}\vartheta _4(z)^{n_4})(\vartheta _3(z)^{n_1}+\vartheta _4(z)^{n_1})+\vartheta _2(z)^{n_1+n_3}(\vartheta _3(z)^{n_2}\vartheta _4(z)^{n_2})`$
$`\times (\vartheta _3(z)^{n_4}+\vartheta _4(z)^{n_4})+\vartheta _2(z)^{n_1+n_2+n_3+n_4}+\vartheta _2(z)^{n_2+n_4}(\vartheta _3(z)^{n_1}\vartheta _4(z)^{n_1})`$
$`\times (\vartheta _3(z)^{n_3}+\vartheta _4(z)^{n_3})+\vartheta _2(z)^{n_2+n_4}(\vartheta _3(z)^{n_3}\vartheta _4(z)^{n_3})(\vartheta _3(z)^{n_1}+\vartheta _4(z)^{n_1})`$
$`+\vartheta _2(z)^{n_1+n_2+n_3+n_4}+\vartheta _2(z)^{n_1+n_3}(\vartheta _3(z)^{n_4}\vartheta _4(z)^{n_4})(\vartheta _3(z)^{n_2}+\vartheta _4(z)^{n_2})+\vartheta _2(z)^{n_2+n_3}`$
$`\times (\vartheta _3(z)^{n_1}\vartheta _4(z)^{n_1})(\vartheta _3(z)^{n_4}+\vartheta _4(z)^{n_4})+\vartheta _2(z)^{n_1+n_2}(\vartheta _3(z)^{n_4}\vartheta _4(z)^{n_4})`$
$`\times (\vartheta _3(z)^{n_3}+\vartheta _4(z)^{n_3})+\vartheta _2(z)^{n_1+n_4}(\vartheta _3(z)^{n_3}\vartheta _4(z)^{n_3})(\vartheta _3(z)^{n_2}+\vartheta _4(z)^{n_2})`$
$`+\vartheta _2(z)^{n_3+n_4}(\vartheta _3(z)^{n_2}\vartheta _4(z)^{n_2})(\vartheta _3(z)^{n_1}+\vartheta _4(z)^{n_1})+(\vartheta _3(z)^{n_1}\vartheta _4(z)^{n_1})(\vartheta _3(z)^{n_2}`$
$`\vartheta _4(z)^{n_2})(\vartheta _3(z)^{n_3}\vartheta _4(z)^{n_3})(\vartheta _4(z)^{n_1}\vartheta _4(z)^{n_4})+(\vartheta _3(z)^{n_1}+\vartheta _4(z)^{n_1})`$
$`\times (\vartheta _3(z)^{n_2}+\vartheta _4(z)^{n_2})(\vartheta _3(z)^{n_3}+\vartheta _4(z)^{n_3})(\vartheta _4(z)^{n_1}+\vartheta _4(z)^{n_4})`$
$`=`$ $`\vartheta _2(z)^{n_3+n_4}[(\vartheta _3(z)^{n_1}\vartheta _4(z)^{n_1})(\vartheta _3(z)^{n_2}+\vartheta _4(z)^{n_2})+(\vartheta _3(z)^{n_2}\vartheta _4(z)^{n_2})`$
$`\times (\vartheta _3(z)^{n_1}+\vartheta _4(z)^{n_1})]+\vartheta _2(z)^{n_1+n_4}[(\vartheta _3(z)^{n_2}\vartheta _4(z)^{n_2})(\vartheta _3(z)^{n_2}+\vartheta _4(z)^{n_3})`$
$`+(\vartheta _3(z)^{n_3}\vartheta _4(z)^{n_3})(\vartheta _3(z)^{n_2}+\vartheta _4(z)^{n_2})]+\vartheta _2(z)^{n_1+n_2}[(\vartheta _3(z)^{n_3}\vartheta _4(z)^{n_3})`$
$`\times (\vartheta _3(z)^{n_4}+\vartheta _4(z)^{n_4})+(\vartheta _3(z)^{n_4}\vartheta _4(z)^{n_4})(\vartheta _3(z)^{n_3}+\vartheta _4(z)^{n_3})]`$
$`+\vartheta _2(z)^{n_2+n_3}[(\vartheta _3(z)^{n_1}\vartheta _4(z)^{n_1})(\vartheta _3(z)^{n_4}+\vartheta _4(z)^{n_4})+(\vartheta _3(z)^{n_4}\vartheta _4(z)^{n_4})`$
$`\times (\vartheta _3(z)^{n_1}+\vartheta _4(z)^{n_1})]+\vartheta _2(z)^{n_1+n_3}[(\vartheta _3(z)^{n_2}\vartheta _4(z)^{n_2})(\vartheta _3(z)^{n_4}+\vartheta _4(z)^{n_4})`$
$`+(\vartheta _3(z)^{n_4}\vartheta _4(z)^{n_4})(\vartheta _3(z)^{n_2}+\vartheta _4(z)^{n_2})]+\vartheta _2(z)^{n_2+n_4}[(\vartheta _3(z)^{n_1}\vartheta _4(z)^{n_1})`$
$`\times (\vartheta _3(z)^{n_3}+\vartheta _4(z)^{n_3})+(\vartheta _3(z)^{n_3}\vartheta _4(z)^{n_3})(\vartheta _3(z)^{n_1}+\vartheta _4(z)^{n_1})]`$
$`+2\vartheta _2(z)^{n_1+n_2+n_3+n_4}+(\vartheta _3(z)^{n_1}\vartheta _4(z)^{n_1})(\vartheta _3(z)^{n_2}\vartheta _4(z)^{n_2})(\vartheta _3(z)^{n_3}\vartheta _4(z)^{n_3})`$
$`\times (\vartheta _4(z)^{n_1}\vartheta _4(z)^{n_4})+(\vartheta _3(z)^{n_1}+\vartheta _4(z)^{n_1})(\vartheta _3(z)^{n_2}+\vartheta _4(z)^{n_2})`$
$`\times (\vartheta _3(z)^{n_3}+\vartheta _4(z)^{n_3})(\vartheta _4(z)^{n_1}+\vartheta _4(z)^{n_4})`$
$`=`$ $`2[(\vartheta _3(z)^{n_1+n_2}\vartheta _2(z)^{n_1+n_2}\vartheta _4(z)^{n_1+n_2})(\vartheta _3(z)^{n_3+n_4}\vartheta _2(z)^{n_3+n_4}\vartheta _4(z)^{n_3+n_4})`$
$`(\vartheta _3(z)^{n_1+n_3}\vartheta _2(z)^{n_1+n_3}\vartheta _4(z)^{n_1+n_3})(\vartheta _3(z)^{n_2+n_4}\vartheta _2(z)^{n_2+n_4}\vartheta _4(z)^{n_2+n_4})`$
$`(\vartheta _3(z)^{n_1+n_4}\vartheta _2(z)^{n_1+n_4}\vartheta _4(z)^{n_1+n_4})(\vartheta _3(z)^{n_2+n_3}\vartheta _2(z)^{n_2+n_3}\vartheta _4(z)^{n_2+n_3})`$
$`+4(\vartheta _2(z)^{n_1+n_2+n_3+n_4}+\vartheta _3(z)^{n_1+n_2+n_3+n_4}+\vartheta _4(z)^{n_1+n_2+n_3+n_4})]`$
$`=`$ $`2[4h_{m_1+m_2+m_3+m_4+2ϵ+1}2^8\mathrm{\Delta }_{24}(z)\text{sym}\{\rho _{m_1+m_2+ϵ1}\rho _{m_3+m_4+ϵ1}\}],(3.45)`$
which implies (3.44).$`\mathrm{}`$
By (3.44) with $`m_1=m_2=m_3=m_4=0`$ and $`ϵ=1`$, we have
$$\text{the theta series of Niemeier lattice of type}D_6^4=E_4(z)^3480\mathrm{\Delta }_{24}(z)$$
$`(3.46)`$
(cf. \[N\]).
Remark 3.4. (a) The theta series in (3.44) is different from those in (3.11) and (3.12) if $`(m_1,m_2,m_3,m_4,ϵ)(0,0,0,0,0)`$.
(b) The above families of lattices are the infinite families of positive definite even unimodular lattices whose theta series are weighted symmetric polynomials of the functions of $`\{\mathrm{\Delta }_{24}(z),h_n(z),\rho _n(z)n_+\}`$ that we can find so far. We speculate that the theta series of the other infinite families of positive definite even-unimodular lattices may be related to the invariants of the other finite groups.
References
R. E. Borcherds, The Leech lattice and other lattices, Ph.D. Dissertation, Univ. of Cambridge, 1983.
—, Vertex algebras, Kac-Moody algebras, and the Monster, Proc. Natl. Acad. Sci. USA 83 (1986), 3068-3071.
—, Three lectures on exceptional groups, in: Finite simple groups, ed. by G. Higman and M. B. Powell, Chap.7, Academic Pless, London/New York, 215-247, 1971.
J. H. Conway and N. J. A. Sloane, Twenty-three constructions of the Leech lattice, Proc. Roy. Soc. London A 381 (1982), 275-283.
—, On the enumeration of lattices of determinant one, J. Number Theory 15 (1982), 83-93.
—, Sphere Packings, Lattices and Groups, Springer-Verlag, 1988.
E. Hecke, Analytische arithmetik der positiven quadratischen formen, Kgl. Danske Vid. Selskab. Mat.-fys. Medd. 13 No. 12 (1940).
M. H. Freedman, The topolgy of 4-dimensional manifolds, J. Diff. Geom. 17 (1982), 357-453.
I. B. Frenkel, J. Lepowsky and A. Meurman, Vertex Operator Algebras and the Monster, Pure and Applied Mathematics 134, Academic Press Inc., Boston, 1988.
D. H. Lehmer, Ramanujan’s function $`\tau (n)`$, Duke Math. J. 10 (1943), 483-492.
H. V. Niemeier, Definite quadratische formen der dimension 24 und diskriminante 1, J. Number Theory 5 (1973), 142-178.
N. J. A. Sloane, codes over $`GF(4)`$ and complex lattices, J. Algebra 52 (1978), 168-181.
—, Self-dual codes and lattices, Proc. Symposia in Pure Math. 34 (1979), 273-307.
X. Xu, Untwisted and twisted gluing techniquaes for constructing self-dual lattices, Ph.D. Dissertation, Rutgers University, 1992.
—, Self-dual lattices of type A, Acta Math. 175 (1995), 123-150.
—, Introduction to Vertex Operator Superalgebras and Their Modules, Kluwer Academic Publishers, Dordrecht/Boston/London, 1998.
|
warning/0003/astro-ph0003459.html
|
ar5iv
|
text
|
# 1 HIGH ENERGY COSMIC NEUTRINO SCIENCE
## 1 HIGH ENERGY COSMIC NEUTRINO SCIENCE
### 1.1 Neutrinos from Interactions of Ultrahigh Energy Cosmic Rays with the 3 K Cosmic Background Radiation
Measurements from the COBE (Cosmic Background Explorer) convincingly proved that the universe is filled with radiation having the character of a near perfect 3 K black body, which is the cooled down remnant of the hot big bang. Extremely high energy protons (above $`10^{20}`$ eV) will collide with photons of this radiation, producing pions. Ultrahigh energy neutrinos (greater than $`10^{19}`$ eV) are the main result of the decay of these pions. Using present measurements of the flux of such high energy protons, it can be shown that measurable numbers of high energy neutrinos can be detected using imaging optics aboard satellites looking down at the luminous tracks produced in the atmosphere by showers of charged particles produced when these neutrinos hit the nuclei of atoms in the atmosphere. Such satellite arrays have been proposed with the names OWL (Orbiting Wide-angle Light Collectors) and Airwatch.
### 1.2 Fossils of Grand Unified Theories and Inflation from the Early Universe
The modern scenario for the early history of the big bang, takes account of the work of particle theorists to unify the forces of nature in the framework of Grand Unified Theories (GUTs). This concept extends the very successful work of Nobel Laureates Glashow, Weinberg, and Salam in unifying the electromagnetic and weak nuclear forces of nature. In GUTs, these forces become unified with the strong nuclear force at very high energies of $`10^{25}`$ eV which occurred only $`10^{35}`$ seconds after the big bang. The fossil remnants of this unification are predicted to be very heavy “topological defects” in the vacuum of space. These are localized regions where extremely high densities of mass-energy are trapped. Such defects go by designations such as cosmic strings, monopoles, walls, necklaces (string bounded by monopoles), and textures, depending on their geometrical and topological properties. Inside a topological defect, the vestiges of the early universe may be preserved to the present day. Topological defects are expected to produce very heavy particles that decay to produce ultrahigh energy neutrinos. One can learn about the high energy physics unachievable to collider experiments by studying the neutrino signal from the topological defects.The annihilation and decay of these structures and particles is predicted to produce large numbers of neutrinos with energies approaching the energy of grand unification. It has been suggested that this process may have produced the highest energy cosmic rays yet observed. The discovery of such large flux of neutrinos with energies near the GUT energy scale would be prima facie evidence for grand unification. The GUT energy scale is over 12 orders of magnitude higher than the energies currently available at terrestrial accelerators. It is difficult to imagine that terrestrial accelerators which ever reach this energy scale.
Relicts of an early inflationary phase in the history of the universe can also lead to the production of ultrahigh energy neutrinos. The homogeneity and flatness of the present universe may imply that a period of very rapid expansion, called inflation, took place shortly after the big bang. During inflation, the universe is cold but, when inflation is over, coherent oscillations of the inflation field reheat it to a high temperature. While the inflation is oscillating, a non-thermal production of very heavy particles may take place. These particles may survive to the present as a part of dark matter. Their decays can give origin to the highest energy cosmic rays, either by emission of hadrons and photons, or through production of ultrahigh energy neutrinos. Observation of such ultrahigh energy neutrinos may teach us about the dark matter of the universe as well as its inflationary history.
### 1.3 Neutrinos from Active Galactic Nuclei
Quasars and other active galactic nuclei (AGN) are most powerful continuous emitters of energy in the known universe. These remarkable objects are fueled by the gravitational energy released by matter falling into a supermassive black hole at the center of the quasar core. The infalling matter accumulates in an accretion disk which heats up to temperatures high enough to emit large amounts of UV and soft X-radiation. The mechanism responsible for the efficient conversion of gravitational energy to observed luminous energy in not yet completely understood. If this conversion occurs partly through the acceleration of particles to relativistic energies, perhaps by the shock formed at the inner edge of the accretion disk, then the interactions of the resulting high energy cosmic rays with the intense photon fields produced by the disk at the quasar cores can lead to the copious production of mesons. The subsequent decay of these mesons will then produce large fluxes of high energy neutrinos. Since the gamma-rays and high energy cosmic rays deep in the intense radiation field of the AGN core will lose their energy rapidly and not leave the source region, these AGN core sources will only be observable as high energy neutrino sources.
Radio loud quasars contain jets of plasma streaming out from the vicinity of the black hole, in many cases with relativistic velocities approaching the speed of light. In a subcategory of quasars, known as blazars, these jets are pointed almost directly at us with their observed radiation, from radio to gamma-ray wavelengths, beamed toward us. It has been found that most of these blazars actually emit the bulk of their energy in the high energy gamma-ray range. If, as has been suggested, the gamma-radiation from these objects is the result of interactions of relativistic nuclei, then high energy neutrinos will be produced with energy fluxes comparable to the gamma-ray fluxes from these objects. On the other hand, if the blazar gamma-radiation is produced by purely electromagnetic processes involving only high energy electrons, then no neutrino flux will result.
### 1.4 Neutrinos from Gamma-Ray Bursts
Gamma-ray Bursts (GRBs) are nature’s most energetic transient phenomenon. In a very short time of $``$ 0.1 to 100 seconds, these bursts can release an energy in gamma-rays alone of the order of $`10^{52}`$ erg. They are detected at a rate of about a thousand per year by present instruments. It has been proposed that particles can be accelerated in these bursts to energies in excess of $`10^{20}`$ eV, either by shocks or perhaps by photonically driven plasma waves.
It is now known that most bursts are at cosmological distances corresponding to moderate redshifts ($`z1`$). If cosmic-rays are accelerated in them to ultrahigh energies, interactions with gamma-rays in the sources leading to the production of pions has been suggested as a mechanism for producing very high energy neutrinos as well. These neutrinos would also arrive at the Earth in a burst coincident with the gamma-rays. This is particularly significant since the ultrahigh energy cosmic rays from moderate redshifts are attenuated by interactions with the 3 K microwave radiation from the big-bang and are not expected to reach the Earth themselves in significant numbers. This attenuation of cosmic rays of ultrahigh energy is known as the Greisen-Zatsepin-Kuzmin (GZK) cutoff.
### 1.5 Neutrino Oscillations and Neutrino Observatons
Recent observations of the disappearance of atmospheric muon-neutrinos relative to electron neutrinos by the Kamiokande group, and also the zenith angle distribution of this effect, may be interpreted as evidence of the oscillation of this weakly interacting neutrino state (“flavor”) into another neutrino flavor, either tau neutrinos or sterile neutrinos. A corollary of such a conclusion is that at least one neutrino state has a finite mass. This has very important consequences for our basic theoretical understanding of the nature of neutrinos and may be evidence for the grand unification of electromagnetic, weak and strong interactions.
If muon neutrinos oscillate into tau neutrinos with the parameterization implied by the Super-Kamiokande measurements, then the fluxes of these two neutrino flavors observed from astrophysical sources should be equal. This is because cosmic neutrinos arrive from such large distances that many oscillations are expected to occur during their journey, equalizing the fluxes in both flavor states.
On the other hand, if these oscillations do not occur, the fluxes of tau neutrinos from such sources should be much less than those of muon neutrinos. This is because muon neutrinos are produced abundantly in the decay of pions which are easily produced in cosmic sources, whereas tau neutrinos are not.
Thus, by looking for upward moving showers from tau neutrinos, which can propagate thorugh the Earth through regeneration at energies above $`10^{14}`$ eV, one can test for the existence of neutrino oscillations. For large mixing angles between neutrino states, the detection of $`10^{14}`$ eV tau-neutrino induced upward-moving atmospheric showers from the direction of a cosmic source such as an active galaxy or gamma-ray burst at a distance of 1 Gpc would occur for a difference of the squares of the mass states as small as $`10^{17}`$ eV<sup>2</sup>, providing an extremely sensitive test for oscillations between neutrino states with extremely small mass differences.
Another important signature of ultrahigh energy tau neutrinos is the “double bang” which they would produce. The first shower is produced by the original interaction which creates a tau particle and a hadronic shower. This is followed by the decay of the tau which produces the second shower bang. The two bamgs are separated by a distance of $``$ 91.4 $`\mu `$m times the Lotentz factor of the tau.
### 1.6 Z-bursts from Neutrinos and Ultrahigh Energy Cosmic Rays
#### 1.6.1 Ultrahigh Energy Cosmic Rays
The quest for higher and higher energy cosmic rays goes forward undeterred by the expectation that protons above $`10^{20}`$ eV from sources farther away than $`100`$ Mpc should strongly depleted by interactions with the 3 K photons which make up the cosmic background radiation from the big-bang. Some cosmic rays at these energies have already been detected with ground-based observatories such as AGASA (Akeno Giant Air Shower Array) and Fly’s Eye. Satellite observatories such as OWL/Airwatch would increase the number to thousands and may allow the detection of cosmic rays above $`10^{21}`$ eV, if such cosmic rays exist. The detection of thousands of events would enable OWL/Airwatch to obtain an energy spectrum covering an interval where different acceleration and topological defect scenarios make different predictions of the energy spectrum.
OWL/Airwatch also has the unmatched capacity to map the arrival directions of cosmic rays over the entire sky and thus to reveal the locations of strong nearby sources and large-scale anisotropies, this owing to the magnetic stiffness of charged particles of such high energy. Thus, OWL/Airwatch can investigate energy spectra of any detected sources and also time correlations with high energy neutrinos and gamma-rays.
#### 1.6.2 Z-burst Neutrinos
An exciting possibility is that the highest energy cosmic-ray events observed above the GZK cutoff energy provides indirect detection of the relic neutrinos predicted in standard big bang cosmology. The observed thermal 3 K cosmic microwave background (CMB) permeates the universe as a relic of the big-bang is accompanied by a 2 K cosmic neutrino background of the same thermal big-bang origin. It has been proposed that high energy neutrinos interacting within the GZK attenuation distance with the copious 2 K blackbody neutrinos and annihilating at the Z-boson resonance energy can produce the observed “trans-GZK” air-shower events. This is a possible solution to the problem of how can cosmic rays of “trans GZK” energies (above the GZK cutoff), which are observed to interact in our atmosphere and produce giant air showers, get here from the extragalactic sources which may be many absorption lengths away, (i.e., hundreds of megaparsecs). This is because high energy neutrinos, which may originate at such large distances throughout the Universe, can reach the Earth or nearby parts of the Universe and then annihilate to produce ultrahigh energy cosmic rays. The neutrinos themselves interact too weakly with nuclei in our atmosphere to produce the observed high altitude air showers, but they can annihilate with CNB neutrinos through the Z-boson resonance to produce “local” photons and protons, with a probability of order $`10^{2\pm 1}`$. The resulting Z-boson then decays to produce a shower of hadrons and leptons, a “Z-burst”. The products of the Z decay, as measured at the CERN and SLAC colliders, include on average 20 photons and 2 nucleons. These photons and protons are the candidate particles for initiating the observed trans-GZK air-showers. Because the annihilation process is resonant, the event energy is unique. It is $`E_{Zburst}=4\times 10^{21}[m_\nu ]^1`$ eV in terms of the neutrino mass (in units of eV). Each nucleon and photon in the burst carries on average an energy which is a few percent of the Z-burst energy.
The Z-burst hypothesis based on the assumption that there exists a significant flux of neutrinos at $`E10^{22}`$ eV, perhaps from topological defects. Some predictive consequences of this hypothesis are (a) that the direction of the air showers should be close to the directions of their cosmological sources, (b) that there may be multiple events coming from the directions of the strongest sources, (c) that there exists a relationship between the maximum shower energy attainable and the terrestrially-measured neutrino mass, and (d) that there may be an observable large-scale anisotropy caused by a clustering of 2 K neutrinos within the cosmic-ray attenuation distance.
## 2 NEUTRINO FLUX PREDICTIONS
Figure 1 illustrates the high energy neutrino flux predictions from various astrophysical sources as a function of neutrino energy. Note that curves show the differential neutrino flux multiplied by $`E_\nu ^2`$ which is equivalent to an energy flux. In the energy range of $`10^{14}to10^{17}`$ eV, the AGN neutrino flux is predicted to dominate over other sources. However, neutrinos from individual gamma-ray bursts may be observable via their directionality and short, intense time characteristics. The time-averaged background flux from all bursts is shown in the figure. In the energy range $`E10^{18}`$ eV, neutrinos are produced from photomeson interactions of ultrahigh enrgy cosmic rays with the 3 K background photons. The highest energy neutrinos ($`E10^{20}`$ eV) are presumed to arise from more the speculative physics of topological defects and Z-bursts.
The proposed high energy neutrinos sources also have different signatures in terms of other observables which include coincidences with other observations (GRB’s), anisotropy (Z-burst’s), and specific relations to the number of hadronic or photonic air showers also induced by the phenomena (topological defectss, Z-burst’s, and 3 K photomeson neutrinos). The distiguishing characteristics of these neutrino sources are summarized in Table 1.
## 3 EXPERIMENTAL SIGNATURES AND RATES
### 3.1 Air Fluorescence Events
A space-based experiment viewing $`10^6`$ km<sup>2</sup> of the surface of the Earth also monitors $`10^{13}`$ metric tons of atmosphere. This large target mass opens the possibility of observing the interactions of ultra-high energy (UHE) neutrinos in this atmospheric volume. The distribution of atmospheric depth of neutrino interactions is approximately uniform due to the extremely long interaction path of neutrinos in the atmosphere. This offers a unique signature of neutrino-induced airshowers as a significant portion of the neutrino interactions will be deep in the atmosphere, i.e. near horizontal, and well separated from airshowers induced by hadrons and photons. A space-based experiment observing the fluorescence signal of airshowers will have a segmented detector plane (pixels) in order to measure the spatial development of the airshowers. Additionally, the signals will be recorded with a very fine time resolution in order to measure the temporal development of the showers. This experimental configuration yields a multiple pixel signature for near horizontal airshowers which translates into straight-forward detection with sufficient angular resolution to guarantee separation of neutrino induced events from the hadronic (or electromagnetic) airshower background.
At ultrahigh energies, the cross sections for neutrino and antineutrino interactions with quarks become equivalent (Ghandhi, et al. 1998, Phys Rev D, 58, 093009). The kinematics of UHE neutrino interactions ($`E_\nu >10^9`$ GeV) leads to the condition that average energy of the resulting lepton will be approximately 80% of the incident neutrino energy. The remaining 20% will materialize in the form of a hadronic cascade from the neutrino interaction point. Charged current neutrino interactions will, on average, yield an UHE charged lepton and a hadronic airshower. At these energies, electrons will generate electronic airshowers while muons and taus will make airshowers with reduced particle densities and, thus, fluorescence signals. The taus offer an additional, “double-bang” signature because of their lifetime and decay modes. At $`10^{10}`$ GeV, $`\gamma c\tau =500`$ km for a tau after which the particle will decay inducing a second airshower separated from the first, hadronic airshower at the neutrino interaction point.
As a first step in quantifing the rates of neutrino airshower observation, we will focus on the charged current electron neutrino interaction. These deposit 100% of the incident neutrino energy into an airshower and will yield the highest air fluorescence signal for a given neutrino energy of the possible flavor channels. Preliminary Monte Carlo simulation of an OWL/AirWatch instrument have indicated that charged current electron neutrino interactions can be identified with a neutrino aperture of 20 km<sup>2</sup>-ster at a threshold energy of $`3\times 10^{10}`$ GeV and this aperture grows with the $`E_\nu ^{0.363}`$ assumed increase in neutrino cross section. Event rates can be obtained by convolving this neutrino aperture with neutrino flux predictions and integrating. Note that the neutrino interaction cross section is included in the definition of neutrino aperture. Assuming a 10% duty cycle of the experiment, Table 2 lists the electron neutrino event rates from several possible UHE neutrino sources: neutrinos from the interaction of UHE protons with the microwave background (p$`\gamma _{2.7K}`$, Stecker et al. 1991, Phys. Rev. Letters 66, 2697) and the more speculative sources of topological defects (Sigl et al. 1999, Phys. Rev. D 59, 043504) and the interaction of UHE massive neutrinos with the 2 K relic neutrino background (Z<sub>Burst</sub>, Yoshida et al. 1998, Phys. Rev. Letters 81, 5505).
### 3.2 Upward Cherenkov Events
The ensemble of charged particles in an airshower will produce a large photon signal from Cherenkov radiation which is strongly peaked in the forward direction and which is much stronger than the signal due to air fluorescence at a given energy. This translates into a much reduced energy threshold for observing airshowers via Cherenkov radiation. As this signal is highly directional, an orbitting instrument will only observe those events where the airshower is moving towards the experiment with the instrument located in the field of the narrow, Cherenkov cone.
Virtually all particles, including neutrinos with $`E40`$ TeV, are attenuated by the Earth. However, tau neutrinos will regenerate themselves, albeit at a lower energy, due to the fact that both charged and neutral current interactions will have a tau neutrino in the eventual, final state (see section 1.5). Thus, the use of the Earth as a tau neutrino filter leads to the possibility of a cosmological, long-baseline tau neutrino appearance experiment. Moreover, an experiment which monitors $`10^6`$ km<sup>2</sup> of the Earth’s surface samples an incredible target mass of the Earth’s crust.
At high energies ($`E>10^6`$ GeV), neutrinos and antineutrinos interact with approximately equal cross sections. Furthermore, the average energy in the lepton resulting from a neutrino interaction is greater than 70% of the incident neutrino energy. Tau leptons produced in charged current, neutrino interactions will have a flight path of length $`\gamma c\tau (50\mathrm{m}`$ at $`10^6`$ GeV), after which they will decay producing high energy cascades for most of the branching ratio. For interactions in the Earth’s crust, those events which occur at a depth less than $`\gamma c\tau `$ will have a tau coming out of the Earth and generating an airshower. For a target area of 10<sup>6</sup> km<sup>2</sup>, this yields a target mass of $`10^8\times (E_\nu /\mathrm{GeV}`$) metric tons, e.g. $`10^{14}`$ metric tons at an energy of $`10^6`$ GeV.
Preliminary investigation of the response of an OWL/AirWatch instrument has indicated that the experiment would have a threshold energy $`10^5`$ GeV to upward, Cherenkov airshowers. Assuming that the Super-Kamiokande atmospheric neutrino results are due to $`\nu _\mu \nu _\tau `$ oscillations, the predicted AGN muon neutrino flux (Stecker & Salamon 1996, Space Sci. Rev. 75, 341) indicates that OWL/AirWatch could observe several hundred tau events per year. Thus OWL/AirWatch would measure the flux of AGN neutrinos and observe their oscillations.
## 4 Conclusion
An OWL/Airwatch array of satellite-based optics optimized to monitor a significant volume of Earth’s atmosphere for the purpose of studying the characteristics of giant atmospheric air showers induced by ultrahigh energy neutrinos (as well as ultrahigh energy cosmic rays) can open a new window of astronomy and physics which will explore the ultrahigh energy neutrino universe.
## 5 Acknowledgments
This white paper is the result of a workshop on Observing Ultrahigh Energy Neutrinos with OWL/Airwatch which was held at UCLA on November 1-3, 1999. Contributers to this paper were (in alphabetical order) D.B. Cline, G. Gelmini, J. Krizmanic, A. Kusenko, J. Linsley, F.W. Stecker, Y. Takahashi and T. Weiler.
|
warning/0003/hep-ph0003200.html
|
ar5iv
|
text
|
# Scalar Mesons and Glueballs in a Chiral 𝑈(3)×𝑈(3) Quark Model with ’t Hooft Interaction.
## 1. Introduction
The self-interaction of gluons, a peculiarity of QCD, gave an idea that gluons can form bound states that can propagate particlelike in the space. Unfortunately, because of theoretical problems, there is still no exact answer on do these states really exist or not. However, from recent lattice simulations one can conclude that it is most probably that glueballs are the reality of our world. Having assumed that glueballs exist, one can try to construct a model to describe their interaction with other mesons, their properties, such as, e.g., the mass and the decay width, and to identify them with observed resonances.
An exact microscopic description of bound gluon states cannot be done systematically in the framework of QCD. In this situation, QCD-motivated phenomenological models are the tool which can help to deal with glueballs as well as with quarkonia which form the most of observed meson states. However, using these models to describe glueballs, we encounter many difficulties concerning, e.g., the ambiguity of the ways of including glueballs into models and identification of experimentally observed meson states. This justifies the variety of points of view on this problem.
First of all, we do not know the exact mass of a glueball. From the quenched QCD lattice simulations, Weingarten (see e.g. ) concluded that the lightest scalar glueball is expected around 1.7 GeV. Amsler considered the state $`f_0(1500)`$ as a candidate for the scalar glueball. QCD sum rules and K-matrix method showed that both $`f_0(1500)`$ and $`f_0(1710)`$ are mixed states with large admixture of the glueball component.
All the bound isoscalar $`q\overline{q}`$ states are subject to mixing with glueballs, and their spectrum has many interpretations made by different authors. For instance, Palano suggested a scenario, in which the states $`a_0(980)`$, $`K_0^{}(1430)`$, $`f_0(980)`$, and $`f_0(1400)`$ form a nonet. The state $`f_0(1500)`$ is considered as the scalar glueball. Törnqvist et al. looked upon the states $`f_0(980)`$ and $`f_0(1370)`$ as manifestations of the ground and excited $`s\overline{s}`$ states, and the state $`f_0(4001200)`$ as the ground $`u\overline{u}`$ state. Eef van Beveren et al. considered the states $`f_0(4001200)`$ and $`f_0(1370)`$ as ground $`u\overline{u}`$ states, and the states $`f_0(980)`$ and $`f_0(1500)`$ as ground $`s\overline{s}`$ states. Two states for each $`q\overline{q}`$ system occur due to pole doubling, which takes place for scalar mesons in their model. Shakin et al. obtained from a nonlocal confinement model that the $`f_0(980)`$ resonance is the ground $`u\overline{u}`$ state and $`f_0(1370)`$ is the ground $`s\overline{s}`$ state. The state $`f_0(1500)`$ is considered as a radial excitation of $`f_0(980)`$. They believe the mass of scalar glueball to be 1770 MeV.
In our recent papers , following the methods given in Refs. , we showed that all experimentally observed scalar meson states with masses in the interval from 0.4 GeV till 1.7 GeV can be interpreted as members of two scalar meson nonets — the ground meson nonet and its first radial excitation. We considered all scalar mesons as $`q\overline{q}`$ states and took into account the singlet-octet mixing caused by the ’t Hooft interaction. However, we did not describe the scalar state $`f_0(1500)`$. We supposed that this state contained a significant component of the scalar glueball (see ).
In this work, we introduce the glueball field into our effective Lagrangian as a dilaton and describe it and its mixing with the scalar isoscalar $`q\overline{q}`$ states. We consider only ground $`q\overline{q}`$ states. As a result, we describe three scalar mesons $`f_0(4001200)`$, $`f_0(980)`$, and $`f_0(1500)`$ (or $`f_0(1710)`$). Let us note that the model we present here is just a tentative one. In this model, we probe one of the possible ways of including the scalar glueball into a model that would describe the glueball state and quarkonia simultaneously.
In various works (see e.g. ), the authors introduced the glueball into meson Lagrangians using the principle of scale invariance and the dilaton model. For example, $`SU(2)\times SU(2)`$ models were considered and the $`U(3)\times U(3)`$ model, however, without the ’t Hooft interaction was investigated in . Here we follow these works and introduce the scalar glueball state into the Lagrangian constructed in .
In this paper, we introduce the dilaton field into the phenomenological meson Lagrangian obtained from the effective quark interaction of the NJL type after bosonization. We take the meson Lagrangian in a chirally symmetric form (before the spontaneous breaking of chiral symmetry), where only quadratic meson terms are of scale-noninvariant form, therefore, we introduce the dilaton field only into these (massive) terms. To this Lagrangian, we add the potential describing the dilaton self-interaction in the form given in Ref. and dilaton kinetic term.
After introducing the glueball into our effective Lagrangian we describe the mixing of three scalar mesons: two scalar isoscalar $`(q\overline{q})`$ states and the glueball. As a result, we obtain the mass spectrum of these scalar meson states and also describe their main strong decays.
We emphasize once more that we do not claim this model to give a quantitative description of experimental data. We do not take into account the states $`f_0(1370)`$ and one of the states $`f_0(1500)`$, $`f_0(1710)`$ which could not be neglected if the real meson spectrum were our goal. We intend to obtain rather qualitative estimates that will allow us to choose a more reasonable way of including the scalar glueball field into the effective meson Lagrangian.
In Section 2, we discuss an effective meson Lagrangian with the ’t Hooft interaction. In Section 3, we introduce a dilaton field into the Lagrangian. In Section 4, we derive gap equations for the masses of $`u(d)`$ and $`s`$ quarks. We also deduce a set of equations allowing us to bind two of the three new parameters connected with the dilaton field. In Section 5, we obtain mass formulae for scalar mesons and the dilaton and describe the singlet-octet mixing among quarkonia and the mixing of the dilaton with two isosinglet scalar mesons. Here we fix the new model parameters and give their numerical estimates as well as the estimates for the masses of scalar mesons and the glueball. In Section 6, we calculate the widths of main strong decays of the scalar mesons and the glueball. In the Conclusion, we discuss obtained results.
## 2. Chiral effective Lagrangian with ’t Hooft interaction
A $`U(3)\times U(3)`$ chiral Lagrangian with the ’t Hooft interaction was investigated in paper . It consists of three terms as shown in formula (1). The first term represents the free quark Lagrangian, the second is composed of four-quark vertices as in the NJL model, and the last one describes the six-quark ’t Hooft interaction that is necessary to solve the $`U_A(1)`$ problem.
$`L`$ $`=`$ $`\overline{q}(i\widehat{}m^0)q+{\displaystyle \frac{G}{2}}{\displaystyle \underset{a=0}{\overset{8}{}}}[(\overline{q}\lambda _aq)^2+(\overline{q}i\gamma _5\lambda _aq)^2]`$ (1)
$`K\left\{det[\overline{q}(1+\gamma _5)q]+det[\overline{q}(1\gamma _5)q]\right\}.`$
Here $`\lambda _a(a=1,\mathrm{},8)`$ are the Gell-Mann matrices and $`\lambda _0=\sqrt{2/3}`$1, with 1 being the unit matrix; $`m^0`$ is a current quark mass matrix with diagonal elements $`m_u^0`$, $`m_d^0`$, $`m_s^0`$ $`(m_u^0m_d^0)`$.
The standard treatment for local quark models consists in replacing the four-quark vertices with bosonic fields (bosonization). The final effective bosonic Lagrangian appears as a result of the calculation of the quark determinant. First of all, using the method described in , we go from Lagrangian (1) to a Lagrangian which contains only four-quark vertices<sup>2</sup><sup>2</sup>2 In addition to the one-loop corrections of constants of four-quark vertices, we allow for two-loop contributions from the ’t Hooft interaction that modify the current quark masses. Thus, we avoid the problem of double counting of the ’t Hoot contribution in gap equations which was encountered by the author in .
$`L=\overline{q}(i\widehat{}\overline{m}^0)q+{\displaystyle \frac{1}{2}}{\displaystyle \underset{a,b=1}{\overset{9}{}}}[G_{ab}^{()}(\overline{q}\tau _aq)(\overline{q}\tau _bq)+G_{ab}^{(+)}(\overline{q}i\gamma _5\tau _aq)(\overline{q}i\gamma _5\tau _bq)],`$ (2)
where
$`\tau _a=\lambda _a(a=1,\mathrm{},7),\tau _8=(\sqrt{2}\lambda _0+\lambda _8)/\sqrt{3},`$
$`\tau _9=(\lambda _0+\sqrt{2}\lambda _8)/\sqrt{3},`$
$`G_{11}^{(\pm )}=G_{22}^{(\pm )}=G_{33}^{(\pm )}=G\pm 4Km_sI_1^\mathrm{\Lambda }(m_s),`$
$`G_{44}^{(\pm )}=G_{55}^{(\pm )}=G_{66}^{(\pm )}=G_{77}^{(\pm )}=G\pm 4Km_uI_1^\mathrm{\Lambda }(m_u),`$
$`G_{88}^{(\pm )}=G4Km_sI_1^\mathrm{\Lambda }(m_s),G_{99}^{(\pm )}=G,`$
$`G_{89}^{(\pm )}=G_{98}^{(\pm )}=\pm 4\sqrt{2}Km_uI_1^\mathrm{\Lambda }(m_u)`$
$`G_{ab}=0(ab;a,b=1,\mathrm{},7),`$ (3)
$`\overline{m}_u^0`$ $`=`$ $`m_u^032Km_um_sI_1^\mathrm{\Lambda }(m_u)I_1^\mathrm{\Lambda }(m_s),`$ (4)
$`\overline{m}_s^0`$ $`=`$ $`m_s^032Km_u^2I_1^\mathrm{\Lambda }(m_u)^2.`$ (5)
Here $`m_u`$ and $`m_s`$ are constituent quark masses and the integrals
$`I_n^\mathrm{\Lambda }(m_a)={\displaystyle \frac{N_c}{(2\pi )^4}}{\displaystyle d_e^4k\frac{\theta (\mathrm{\Lambda }^2k^2)}{(k^2+m_a^2)^n}}`$ (6)
are calculated (for simplicity) in the Euclidean metric and regularized by a simple $`O(4)`$-symmetric ultra-violet cutoff $`\mathrm{\Lambda }`$. The mass $`m_a`$ can be $`m_u`$ or $`m_s`$.
Now it is necessary to bosonize Lagrangian (2). After bosonization and renormalization of the meson fields, we obtain
$`L(\sigma ,\varphi )={\displaystyle \underset{a,b=1}{\overset{9}{}}}{\displaystyle \frac{g_ag_b}{2}}\left(\sigma _a(G^{()})_{ab}^1\sigma _b+Z\varphi _a(G^{(+)})_{ab}^1\varphi _b\right)`$
$`i\mathrm{Tr}\mathrm{ln}\left\{1+{\displaystyle \frac{1}{i\widehat{}m}}{\displaystyle \underset{a=1}{\overset{9}{}}}g_a\tau _a(\sigma _a+i\gamma _5\sqrt{Z}\varphi _a)\right\},`$ (7)
where $`(G^{(\pm )})^1`$ is the inverse of $`G^{(\pm )}`$, and $`\varphi _a`$ and $`\sigma _a`$ are the pseudoscalar and scalar fields, respectively.
$`g_1^2=g_2^2=g_3^2=g_8^2=g_u^2=[4I_2^\mathrm{\Lambda }(m_u)]^1,`$
$`g_4^2=g_5^2=g_6^2=g_7^2=[4I_2^\mathrm{\Lambda }(m_u,m_s)]^1,`$
$`g_9^2=g_s^2=[4I_2^\mathrm{\Lambda }(m_s)]^1,`$
$`I_2^\mathrm{\Lambda }(m_u,m_s)={\displaystyle \frac{N_c}{(2\pi )^4}}{\displaystyle d_e^4k\frac{\theta (\mathrm{\Lambda }^2k^2)}{(k^2+m_u^2)(k^2+m_s^2)}}=`$
$`={\displaystyle \frac{3}{(4\pi )^2(m_s^2m_u^2)}}\left[m_s^2\mathrm{ln}\left({\displaystyle \frac{\mathrm{\Lambda }^2}{m_s^2}}+1\right)m_u^2\mathrm{ln}\left({\displaystyle \frac{\mathrm{\Lambda }^2}{m_u^2}}+1\right)\right],`$ (8)
$`Z=\left(1{\displaystyle \frac{6m_u}{M_{A_1}^2}}\right)^11.44,`$ (9)
where $`M_{A_1}`$ is the mass of the axial-vector meson.
In the one-loop approximation the meson Lagrangian looks as follows<sup>3</sup><sup>3</sup>3 The expression under Tr in (10) is written formally for the reason of being concise and is not ready for practical use. One will find which $`g_a`$ should be taken, considering the corresponding quark-loop diagrams.
$`L(\sigma ,\varphi )={\displaystyle \frac{1}{2}}{\displaystyle \underset{a=1}{\overset{9}{}}}[(_\mu \sigma _a)^2+(_\mu \varphi _a)^2]+`$
$`+\left[(m_um_u^0)(G^{()})_{88}^1{\displaystyle \frac{m_sm_s^0}{\sqrt{2}}}(G^{()})_{98}^18m_uI_1^\mathrm{\Lambda }(m_u)\right]g_8\sigma _8`$
$`+\left[(m_u\overline{m}_u^0)(G^{()})_{89}^1{\displaystyle \frac{m_s\overline{m}_s^0}{\sqrt{2}}}(G^{()})_{99}^1+8{\displaystyle \frac{m_s}{\sqrt{2}}}I_1^\mathrm{\Lambda }(m_s)\right]g_9\sigma _9`$
$`{\displaystyle \frac{1}{2}}{\displaystyle \underset{a,b=1}{\overset{9}{}}}g_ag_b\left[\sigma _a(G^{()})_{ab}^1\sigma _b+Z\varphi _a(G^{(+)})_{ab}^1\varphi _b\right]+`$
$`{\displaystyle \frac{1}{4}}\mathrm{Tr}\{g^2[8I_1^\mathrm{\Lambda }(m)(\sigma ^2+Z\varphi ^2)`$
$`(\sigma ^2{\displaystyle \frac{\{M,\sigma \}_+}{g}}+Z\varphi ^2)^2+[(\sigma {\displaystyle \frac{M}{g}}),\varphi ]_{}^2]\},`$ (10)
$`C_a=0(a=1,\mathrm{},7),C_8=1,C_9={\displaystyle \frac{1}{\sqrt{2}}}.`$ (11)
It is useful to notice that Lagrangian (10) can be given a chirally symmetric form. For this purpose, we introduce fields $`\sigma ^{}=\sigma mg^1`$ that have nonzero expectation values, $`<\sigma ^{}>_00`$ . Here
$$g=g_8\tau _8\frac{g_9\tau _9}{\sqrt{2}};m=m_u\tau _8+\frac{m_s\tau _9}{2}.$$
(12)
This is the form of the effective Lagrangian in the case of unbroken chiral symmetry:
$`L(\sigma ^{},\varphi )={\displaystyle \frac{1}{2}}{\displaystyle \underset{a=1}{\overset{9}{}}}[(_\mu \sigma _{}^{}{}_{a}{}^{})^2+(_\mu \varphi _a)^2]`$
$`{\displaystyle \frac{1}{2}}{\displaystyle \underset{a,b=1}{\overset{9}{}}}\left[\left(g_a\sigma _a^{}+\mu _a\right)(G^{()})_{ab}^1\left(g_b\sigma _b^{}+\mu _b\right)+Zg_ag_b\varphi _a(G^{(+)})_{ab}^1\varphi _b\right]+`$
$`+{\displaystyle \frac{1}{4}}\mathrm{Tr}\left\{8g^2(I_1^\mathrm{\Lambda }(m)+m^2I_2^\mathrm{\Lambda }(m))(\sigma ^2+Z\varphi ^2)g^2[(\sigma ^2+Z\varphi ^2)^2Z[\sigma ^{},\varphi ]_{}^2]\right\},`$ (13)
$`\mu _a=0(a=1,\mathrm{},7),\mu _8=\overline{m}_u^0,\mu _9={\displaystyle \frac{\overline{m}_s^0}{\sqrt{2}}},`$ (14)
where $`\sigma ^{}=_{a=1}^9\sigma _a\tau ^a`$ and $`\varphi =_{a=1}^9\varphi _a\tau ^a`$. From (13), one can see that, in the one-loop approximation, our Lagrangian does not violate chiral symmetry if $`m^0=0`$, and the ’t Hooft interaction is switched off. Let us require the chiral and scale invariance both to be properties of the effective Lagrangian. We remind that the QCD Lagrangian satisfies this requirement in the chiral limit ($`m^0=0`$). The scale invariance of the effective meson Lagrangian is restored by means of a dilaton field introduced into the Lagrangian so that it provides a proper dimension for each Lagrangian term. The dilaton field is introduced only into the mass terms. We also introduce the dilaton potential intended to reproduce QCD scale anomaly .
## 3. Nambu–Jona-Lasinio model with dilaton
Following earlier works devoted to dilatons, we introduce a color singlet dilaton field $`\chi `$ that experiences a potential
$`V(\chi )=B\left({\displaystyle \frac{\chi }{\chi _0}}\right)^4[\mathrm{ln}\left({\displaystyle \frac{\chi }{\chi _0}}\right)^41].`$ (15)
It has a minimum at $`\chi =\chi _0`$, and the parameter $`B`$ represents the vacuum energy, when there are no quarks. The curvature of the potential at its minimum determines the bare glueball mass
$`m_g={\displaystyle \frac{4\sqrt{B}}{\chi _0}}.`$ (16)
To introduce the dilaton field into the effective meson Lagrangian (13), we use the following principle. Insofar as the QCD Lagrangian, with the current quark masses equal to zero, is chiral and scale invariant, we suppose that our effective meson Lagrangian, motivated by QCD, has also to be chiral and scale invariant in the case when the current quark masses and the ’t Hooft interaction are equal to zero. Then, instead of Lagrangian (13), we obtain
$`L(\sigma ^{},\varphi ,\chi )={\displaystyle \frac{1}{2}}{\displaystyle \underset{a=1}{\overset{9}{}}}[(_\mu \sigma _{}^{}{}_{a}{}^{})^2+(_\mu \varphi _a)^2]`$
$`\{{\displaystyle \frac{1}{2}}{\displaystyle \underset{a,b=1}{\overset{9}{}}}[(g_a\sigma _a^{}+\mu _a)(G^{()})_{ab}^1(g_b\sigma _b^{}+\mu _b)+Zg_ag_b\varphi _a(G^{(+)})_{ab}^1\varphi _b]`$
$`{\displaystyle \frac{1}{4}}\mathrm{Tr}\left\{8g^2(I_1^\mathrm{\Lambda }(m)+m^2I_2^\mathrm{\Lambda }(m))(\sigma ^2+Z\varphi ^2)\right\}\left\}\right({\displaystyle \frac{\chi }{\chi _c}})^2`$
$`{\displaystyle \frac{1}{4}}\mathrm{Tr}\left\{g^2[(\sigma ^2+Z\varphi ^2)^2Z[\sigma ^{},\varphi ]_{}^2]\right\}+{\displaystyle \frac{1}{2}}(_\mu \chi )^2V(\chi ),`$ (17)
where $`\chi _c`$ is the vacuum expectation value of dilaton fields $`\chi =\overline{\chi }+\chi _c`$, $`<\chi >_0=\chi _c`$ and $`<\overline{\chi }>_0=0`$.<sup>4</sup><sup>4</sup>4 Since the mass of current quark explicitly breaks the scale invariance of the model, there is no need to make these terms scale invariant, using the dilaton fields.
By rewriting this Lagrangian in terms of quantum fields $`\sigma `$ and $`\overline{\chi }`$ with vacuum expectations equal zero, we finally obtain
$`L(\sigma ,\varphi ,\overline{\chi })=L(\sigma ,\varphi )+\mathrm{\Delta }L(\sigma ,\varphi ,\overline{\chi }),`$ (18)
where $`L(\sigma ,\varphi )`$ equals Lagrangian (10), and $`\mathrm{\Delta }L(\sigma ,\varphi ,\overline{\chi })`$ has the form
$`\mathrm{\Delta }L(\sigma ,\varphi ,\overline{\chi })={\displaystyle \frac{1}{2}}(_\mu \overline{\chi })^2V^{}(\overline{\chi }+\chi _c)+{\displaystyle \frac{\overline{\chi }}{\chi _c}}\left(2+{\displaystyle \frac{\overline{\chi }}{\chi _c}}\right)`$
$`\times \{8g_8m_u^3I_2^\mathrm{\Lambda }(m_u)\sigma _8+8g_9{\displaystyle \frac{m_s^3I_2^\mathrm{\Lambda }(m_s)}{\sqrt{2}}}\sigma _9`$
$`{\displaystyle \frac{1}{2}}{\displaystyle \underset{a,b=1}{\overset{9}{}}}g_ag_b\left[\sigma _a(G^{()})_{ab}^1\sigma _b+Z\varphi _a(G^{(+)})_{ab}^1\varphi _b\right]`$
$`+4{\displaystyle \underset{a=1}{\overset{9}{}}}g_a^2𝒥_a(\sigma _a^2+Z\varphi _a^2)\},`$ (19)
where
$`𝒥_a=I_1^\mathrm{\Lambda }(m_u)+m_u^2I_2^\mathrm{\Lambda }(m_u),(a=1,2,3,8),`$
$`𝒥_a={\displaystyle \frac{1}{2}}\left(I_1^\mathrm{\Lambda }(m_u)+I_1^\mathrm{\Lambda }(m_s)+{\displaystyle \frac{(m_u+m_s)^2}{4}}I_2^\mathrm{\Lambda }(m_u,m_s)\right),(a=4,5,6,7),`$
$`𝒥_9=I_1^\mathrm{\Lambda }(m_s)+m_s^2I_2^\mathrm{\Lambda }(m_s),`$ (20)
and we used gap equations (25) in the terms linear over $`\sigma _a`$. The potential $`V^{}(\chi )=V(\chi )`$ \+ $`(\chi /\chi _c)^2A`$, where
$`A`$ $`=`$ $`{\displaystyle \frac{1}{2}}{\displaystyle \underset{a,b=1}{\overset{8}{}}}\sigma _a(G^{()})_{ab}^1\sigma _b4𝒥_8m_u^22𝒥_9m_s^2.`$ (21)
From this Lagrangian, one can obtain the system of equations determining constituent quark masses (gap equations) and dilaton potential parameters $`\chi _0`$, $`\chi _c`$ and $`B`$.
## 4. Equations for the quark masses (gap equations) and dilaton potential parameters $`\chi _0`$, $`\chi _c`$ and $`B`$
The conditions for linear terms being absent in our Lagrangian
$`{\displaystyle \frac{\delta L}{\delta \sigma _8}}|_{\varphi ,\sigma ,\overline{\chi }=0}=0,{\displaystyle \frac{\delta L}{\delta \sigma _9}}|_{\varphi ,\sigma ,\overline{\chi }=0}=0,{\displaystyle \frac{\delta L}{\delta \chi }}|_{\varphi ,\sigma ,\overline{\chi }=0}=0`$ (22)
lead us to the following equations
$`(m_u\overline{m}_u^0)(G^{()})_{88}^1{\displaystyle \frac{m_s\overline{m}_s^0}{\sqrt{2}}}(G^{()})_{89}^18m_uI_1^\mathrm{\Lambda }(m_u)`$ $`=`$ $`0,`$ (23)
$`(m_s\overline{m}_s^0)(G^{()})_{99}^1\sqrt{2}(m_u\overline{m}_u^0)(G^{()})_{98}^18m_sI_1^\mathrm{\Lambda }(m_s)`$ $`=`$ $`0,`$ (24)
$`4B\left({\displaystyle \frac{\chi _c}{\chi _0}}\right)^3{\displaystyle \frac{1}{\chi _0}}\mathrm{ln}\left({\displaystyle \frac{\chi _c}{\chi _0}}\right)^4{\displaystyle \frac{2A}{\chi _c}}`$ $`=`$ $`0.`$ (25)
An additional equation follows from the relation between the divergence of dilaton current $`S_\mu `$ and the gluon condensate
$`_\mu S^\mu _0`$ $`=`$ $`\left(\chi {\displaystyle \frac{V(\sigma ^{},\chi )}{\chi }}+{\displaystyle \underset{a=8}{\overset{9}{}}}\sigma _a^{}{\displaystyle \frac{V(\sigma ^{},\chi )}{\sigma _a^{}}}4V(\sigma ^{},\chi )\right)|_{\begin{array}{c}\chi =\chi _c\hfill \\ \sigma _a^{}=\stackrel{~}{\mu }_a/g_a\hfill \end{array}}`$ (28)
$`=`$ $`𝒞_g2m_u^0\overline{u}u_0m_s^0\overline{s}s_0,`$ (29)
$`𝒞_g`$ $`=`$ $`\left({\displaystyle \frac{11}{24}}N_c{\displaystyle \frac{1}{12}}N_f\right){\displaystyle \frac{\alpha }{\pi }}G_{\mu \nu }^2_0,`$ (30)
where $`V(\sigma ^{},\chi )`$ is the potential corresponding to Lagrangian (17) at $`\varphi =0`$, and
$`𝒞_g2m_u^0\overline{u}u_0m_s^0\overline{s}s_0`$ $`=`$ $`4B\left({\displaystyle \frac{\chi _c}{\chi _0}}\right)^4+{\displaystyle \underset{a,b=8}{\overset{9}{}}}(\stackrel{~}{\mu }_a\mu _a)\left(G^{()}\right)_{ab}^1\mu _b`$ (31)
where
$$\stackrel{~}{\mu }_8=m_u,\stackrel{~}{\mu }_9=\frac{m_s}{\sqrt{2}}.$$
(32)
The terms proportional to current quark masses on the right hand side of (31) cancel the quark condansate contribution on the left hand side, therefore we have
$`𝒞_g`$ $`=`$ $`4B\left({\displaystyle \frac{\chi _c}{\chi _0}}\right)^4.`$ (33)
Using (4) and (5), one can rewrite the gap equations (23) and (24) in a well-known form
$`m_u^0`$ $`=`$ $`m_u8Gm_uI_1^\mathrm{\Lambda }(m_u)32Km_um_sI_1^\mathrm{\Lambda }(m_u)I_1^\mathrm{\Lambda }(m_s),`$ (34)
$`m_s^0`$ $`=`$ $`m_s8Gm_sI_1^\mathrm{\Lambda }(m_s)32K(m_uI_1^\mathrm{\Lambda }(m_u))^2.`$ (35)
To define all three parameters of the dilaton potential $`(\chi _0,\chi _c,B)`$, we have to use, in addition to equations (23)–(30), the equation for the bare glueball mass to be be given in the next Section.
## 5. Mass formulae for scalar isoscalar mesons and glueball
The free part of the Lagrangian (18) has the form
$`L^{(2)}(\sigma ,\varphi ,\overline{\chi })`$ $`=`$ $`{\displaystyle \frac{1}{2}}g_8^2[(G^{()})_{88}^18I_1^\mathrm{\Lambda }(m_u)+4m_u^2]\sigma _8^2`$ (36)
$``$ $`{\displaystyle \frac{1}{2}}g_9^2[(G^{()})_{99}^18I_1^\mathrm{\Lambda }(m_s)+4m_s^2]\sigma _9^2`$
$``$ $`g_8g_9(G^{()})_{89}^1\sigma _8\sigma _98(𝒞_gA)\left({\displaystyle \frac{\overline{\chi }}{\chi _c}}\right)^2`$
$``$ $`{\displaystyle \frac{16}{\chi _c}}\left[g_8m_u^3I_2^\mathrm{\Lambda }(m_u)\sigma _8g_9{\displaystyle \frac{m_s^3I_2^\mathrm{\Lambda }(m_s)}{\sqrt{2}}}\sigma _9\right]\overline{\chi }.`$
The dilaton and its interaction with quarkonia does not change the model parameters $`m_u`$, $`m_s`$, $`\mathrm{\Lambda }`$, $`G`$, and $`K`$ fixed in our earlier paper
$`m_u=280\text{MeV},m_s=420\text{MeV},\mathrm{\Lambda }=1.25\text{ GeV},`$
$`G=4.38\text{GeV}^2,K=11.2\text{GeV}^5.`$ (37)
After the dilaton field is introduced into our model, there appear three new parameters: $`\chi `$, $`\chi _c`$, and $`B`$. To determine these parameters, we use two equations (25) and (30) and the bare glueball mass
$$m_g^2=4\frac{(𝒞_gA)}{\chi _c^2}.$$
(38)
We adjust it so that, in the output, the mass of the heaviest meson would be 1500 MeV or 1710 MeV and thereby fix $`\chi _c`$. For the glueball condensate, we use the value $`(390MeV)^4`$ . The result of our fit is presented in Table 1 where we show the spectrum of three physical scalar isoscalar states $`\sigma _I`$, $`\sigma _{II}`$ and $`\sigma _{III}`$.
The parameters $`\chi _0`$ and $`B`$ are fixed by the gluon condensate and constituent quark masses
$$\chi _0=\chi _c\mathrm{exp}\left(\frac{A}{2𝒞_g}\right),$$
(39)
$$B=\frac{𝒞_g}{4}\mathrm{exp}\left(\frac{2A}{𝒞_g}\right).$$
(40)
It is worth noting that the mixing of the glueball with quarkonia shifts the quarkonia mass spectrum. They become lighter, whereas the glueball becomes heavier. This is good for our model because, in previous paper , the mass of $`\sigma _{II}`$ associated with $`f_0(980)`$ was exaggerated.
## 6. Decay widths
Once all parameters are fixed, we estimate decay widths for the major strong decay modes of the scalar mesons: $`\sigma _a\pi \pi `$ and $`\sigma _aKK`$. We neglect decays into $`\eta \eta `$ and $`\eta \eta ^{}`$ as they are small. The results are displayed below. The state $`\sigma _I`$ that we identify with $`f_0(4001200)`$ decays mostly into a pair of pions, and this process determines the width of $`\sigma _I`$:
$$\mathrm{\Gamma }_{\sigma _I\pi \pi }600\text{ MeV}.$$
(41)
The decay of the state $`\sigma _{II}`$ that we identify with $`f_0(980)`$ into pions is noticeably enhanced by the glueball component because of mixing with the $`s\overline{s}`$ quarkonium. We obtain
$$\mathrm{\Gamma }_{\sigma _{II}\pi \pi }=140\text{ MeV}$$
(42)
if $`\sigma _{III}f_0(1500)`$ and
$$\mathrm{\Gamma }_{\sigma _{II}\pi \pi }=120\text{ MeV}$$
(43)
if $`\sigma _{III}f_0(1710)`$. From experiment, we know that its decay width lies within the interval from 40 MeV to 100 MeV.
The decay width of $`\sigma _{III}`$ is slightly different for both cases. In case $`\sigma _{III}`$ is $`f_0(1500)`$, we have
$$\mathrm{\Gamma }_{\sigma _{III}\pi \pi }=96\text{MeV},\mathrm{\Gamma }_{\sigma _{III}KK}=176\text{MeV},$$
(44)
and in the other case ($`\sigma _{III}f_0(1710)`$)
$$\mathrm{\Gamma }_{\sigma _{III}\pi \pi }=120\text{MeV},\mathrm{\Gamma }_{\sigma _{III}KK}=160\text{MeV}.$$
(45)
As one can see, we obtained reasonable values for the states which are mostly quarkonia and overestimated decay width for the state $`\sigma _{III}`$. This can be explained by that the mixing of $`s\overline{s}`$ quarkonium with the glueball in this type of model is too large (it is proportional to the cubed mass of strange quark, see (36)). As a result, the decay $`\sigma _{III}KK`$ is large. This mixing becomes a bit less, when we fit the paramers for a higher glueball mass.
Let us note also that we do not include the decay into $`4\pi `$. This process is not dominant in our model (contrary to Ref. ). Our estimates are based on the assumption that the process $`\sigma _{III}4\pi `$ occurs through two intermediate $`\sigma `$ resonances $`\sigma _{III}\sigma \sigma 4\pi `$. We found that $`\mathrm{\Gamma }_{\sigma _{III}4\pi }`$ does not exceed 20 MeV. Therefore, in both the cases, the total width of $`\sigma _{III}`$ is approximately $`300`$ MeV.
## 7. Conclusion
In this work, we investigated a possible way of including the glueball into an effective chirally symmetric meson Lagrangian. This Lagrangian was studied in Ref. where the masses and strong decay widths of the ground scalar meson states were estimated in the NJL model with the ’t Hooft interaction taken into account. Now, following Ref. , we considered the interaction of the glueball with the ground scalar isoscalar $`q\overline{q}`$ states $`f_0(4001200)`$ and $`f_0(980)`$. The mixing of the glueball with radially excited states $`f_0(1370)`$ and $`f_0(1710)`$ (if $`\sigma _{III}f_0(1500)`$) was not taken into account. However, their mixing is very important and will be considered in subsequent works. Therefore, the results obtained here are tentative and are not claimed for a quantitative explanation of experimental data on scalar resonances.
As it was mentioned in the Introduction, nowadays there are many papers devoted to the description of the scalar glueball in the framework of an effective meson Lagrangian . The way, we introduce the scalar glueball, is closer to that used in Ref. , but our work differs in two points. First, we take into account the ’t Hooft interaction leading to the singlet-octet mixing of the scalar isoscalar quarkonia. The glueball is also involved in this mixing and changes it. Next, we introduce the glueball as a dilaton field into the Lagrangian written in a chirally symmetric form corresponding to the phase with chiral symmetry not broken spontaneously. Thereby, the Lagrangian is given a highly symmetric form that keeps both the chiral symmetry and scale invariance. In this form the dilaton fields are introduced only into the mass terms of meson Lagrangian. The rest of Lagrangian terms are scale invariant except the term with the current quark mass which explicitly breaks both the chiral symmetry and scale invariance.
As a result, we obtain reasonable estimates for the masses of the scalar mesons $`f_0(4001200)`$, $`f_0(980)`$<sup>5</sup><sup>5</sup>5In Ref. , the ground $`q\overline{q}`$ scalar states were presented by $`f_0(980)`$ and $`f_0(1300)`$ as $`u\overline{u}`$ and $`s\overline{s}`$ quarkonia mixed with the glueball., and the glueball $`f_0(1500)`$ (or $`f_0(1710)`$) and also for their strong decay widths. But we have to point out that the width of the $`f_0(1500)`$ (or $`f_0(1710)`$) resonance is, possibly, overestimated ($`300`$ MeV), however, it can change after the states $`f_0(1370)`$ and $`f_0(1710)`$ (or $`f_0(1500)`$) are included into the whole picture.
The results that we obtained in this work, as one can see, are not enough to answer the question: which of the states $`f_0(1500)`$ and $`f_0(1710)`$ is the scalar glueball? We hope to make closer towards the solution of this problem in our further works, where alternative ways of including the glueball into an effective meson Lagrangian will be investigated, radially excited meson states will be considered as well as the ground states, and the mixing of five scalar isoscalar states $`f_0(4001200)`$, $`f_0(980)`$, $`f_0(1370)`$, $`f_0(1500)`$, and $`f_0(1710)`$ will be taken into account.
## Acknowledgement
We are grateful to Prof. S.B. Gerasimov and Dr. A.Dorokhov for useful discussions. The work is supported by RFBR Grant 00-02-17190, Heisenberg-Landau program 2000 and by the Slovak Grant Agency for Science, Grant 2/7157/20.
|
warning/0003/astro-ph0003193.html
|
ar5iv
|
text
|
# Small Deviations from Gaussianity and The Galaxy Cluster Abundance Evolution
## 1 Introduction
Generally, the problem of structure formation is associated to the gravitational growth of small density fluctuations generated by physical processes in the very early universe. Also, these fluctuations are supposed to build a Gaussian random field (GRF), where the Fourier components $`\delta _k`$ have independent, random and uniformly distributed phases. Such a condition means that phases are non-correlated in space and assures the statistical properties of the GRF are completely specified by the two-point correlation function or, equivalently, by the power spectrum $`P(k)=|\delta _k|^2`$, which contains information on the density fluctuation amplitude of each scale $`k`$. This makes the choice of a GRF the simplest initial condition for structure formation studies from the mathematical point of view. At the same time, the GRF simplicity is vindicated by a great number of inflationary models that predict a nearly scale-invariant spectrum of Gaussian density perturbations from quantum-mechanical fluctuations in the field that drives inflation (Guth & Pi 1982). Likewise, the central limit theorem guarantees a GRF if a wide range of random physical processes acts on the distribution of matter in the early universe.
However, a number of mechanisms can generate non-Gaussian density fluctuations. For instance, they arise in some inflation models with multiple scalar fields (e.g. Salopek, Bond & Bardeen 1989); or after phase transitions when different types of topological defects can be formed (Kibble 1976); still, by any discrete, random distributed seed masses like primordial black holes and soliton objects (Sherrer & Bertschinger 1991); as well as in astrophysical processes during the non-linear regime where early generations of massive stars produce shocks which sweep material on to giant blast waves triggering formation of large-scale structure (Ostriker & Cowie 1981). Thus, in order to better understand the process of structure formation, it is necessary to investigate the possibility of the non-Gaussian statistics contribution to the density fluctuation field.
Due to the difficulty to work with generic statistical models, the usual approach is to examine specific classes of non-Gaussian distributions. Examples of these efforts are the studies carried out by Weinberg & Cole (1992) that studied non-Gaussian initial conditions generated by a range of specific local transformations of an underlying Gaussian field; Moscardini et al. (1991) investigated whether non-Gaussian initial conditions can help to reconcile the CDM models with observations; and Kayama, Soda & Taruya (1999), who used data on the abundance of clusters at three different redshifts to establish constraints on structure formation models based on chi-squared non-Gaussian fluctuations generated during inflation.
In this work, we propose a new approach to this problem, exploring the hypothesis that initial conditions for structure formation do not build a single GRF, but a combination of different fields, produced by different physical mechanisms, whose resultant effect presents an arbitrarily small departure from the strict Gaussianity. The paper is organized as follows: in Section 2, we introduce the statistical analysis of finite mixture distributions and present a two-component mixture model; in Sections 3, we apply the model to the cluster abundance evolution; in Section 4 we summarize and discuss our results.
## 2 Mixture Distributions Models: The Positive Skewness Case
Suppose the density fluctuations field, given by the density contrast $`\delta =(\rho (r)\overline{\rho })/\overline{\rho }`$, is a random variable which takes values in a sample space $`\mathrm{}`$, and that its distribution can be represented by a probability density function of the form
$$p(\delta )=\alpha _1f_1(\delta )+\mathrm{}+\alpha _kf_k(\delta )(\delta \mathrm{})$$
$`(2.1)`$
where
$$\alpha _j>0,j=1,\mathrm{},k;\alpha _1+\mathrm{}+\alpha _k=1$$
and
$$f_j(\delta )0,_{\mathrm{}}f_j(\delta )𝑑\delta =1,j=1,\mathrm{},k.$$
When this happens, we say that $`\delta `$ has a finite mixture distribution defined by (2.1), where the components of the mixture are $`f_1(\delta ),\mathrm{},f_k(\delta )`$ and the mixing weights are $`\alpha _1,\mathrm{},\alpha _k`$ (e.g. Titterrington, Smith & Makov 1985). Note that we are not using here the central limit theorem. Mathematically, this will be valid only when $`k\mathrm{}`$ and the weights have similar values, so that one process has no more importance than the others. We are not making these hypotheses here and, consequently, the summation of processes will not necessarily converge to a Gaussian.
Statistical evidence for a small level of non-Gaussianity in the anisotropy of the cosmic background radiation temperature has been found in the COBE 4 year maps (e.g. Ferreira, Magueijo & Gósrki 1998; Pando, Valls-Gabaud & Fang 1998; Magueijo 1999). Non-Gaussian statistics is also expected in the framework of biased models of galaxy formation (Bardeen et al. 1986). In this case, analytical arguments show that non-Gaussian behaviour corresponds to a threshold effect superimposed on the Gaussian background (Politzer & Wise 1984; Jensen & Szalay 1986). In the same way, hybrid models show that it is possible for structure to be seeded by a weighted combination of adiabatic perturbations produced during inflation and active isocurvature pertubations produced by topological defects generated at the end of the inflationary epoch (e.g. Battye & Weller 1998). Thus, a very compelling way to simplify our model is to apply (2.1) to the combination of only two fields: a GRF plus a second field, where the latter will represent a small departure from the strict Gaussianity. This can be posed as
$$p(\delta )=\alpha f_1(\delta )+(1\alpha )f_2(\delta )$$
$`(2.2)`$
The first field will be always the Gaussian component and a possible effect of the second component is to modify the GRF to have positive and/or negative tails. The parameter $`\alpha `$ in (2.2) allow us to modulate the relative importance between the two components. It represents an arbitrarily small departure from the strict Gaussianity and can be due to some primordial mechanism acting on the energy distribution. Such a two-component random field can be generated by taking $`\delta _k^2=P(k)\nu ^2`$, where $`\nu `$ is a random number with distribution given by (2.2). Then we have
$$\delta ^2(r)=\frac{V}{(2\pi )^3}_kP(k)\left[_\nu [\alpha f_1(\nu )+(1\alpha )f_2(\nu )]\nu ^2𝑑\nu \right]d^3k$$
$`(2.3)`$
where $`V`$ is the volume of an arbitrarily large region of the universe. The quantity in the brackets will be defined as the mixture term
$$T^{mix}_\nu [\alpha f_1(\nu )+(1\alpha )f_2(\nu )]\nu ^2𝑑\nu $$
$`(2.4)`$
so that $`P(k)^{mix}P(k)T^{mix}`$, for the case where $`\alpha `$ is not scale-dependent. In the same way, the $`rms`$ mass overdensity within a certain scale $`R`$ will be $`\sigma ^2(R)^{mix}\sigma ^2(R)T^{mix}`$.
As an ilustration, in this work we explore the case of a positive skewness model, where the second field adds to the Gaussian component a positive tail representing a number of rare peaks in the density fluctuation field. A simple way to obtain this effect is to take the well known lognormal distribution as the second component. Besides its mathematical simplicity, this distribution seems to play an important role over the non-linear regime for a wide range of physical scales (e.g. Coles & Jones 1991, Plionis & Vardarini 1995, Bi & Davidsen 1997). Accordingly, our mixture becomes
$$f_1(\nu )=\frac{1}{\sqrt{2\pi }}e^{\frac{1}{2}\nu ^2}\mathrm{and}f_2(\nu )=\frac{1}{\nu \sqrt{2\pi }}e^{\frac{1}{2}(ln\nu )^2}$$
$`(2.5)`$
(for the case of mean zero). Introducing (2.5) in (2.4), we find
$$T^{mix}=_\nu \left[\frac{\alpha }{\sqrt{2\pi }}e^{\frac{1}{2}\nu ^2}+\frac{(1\alpha )}{\nu \sqrt{2\pi }}e^{\frac{1}{2}(ln\nu )^2}\right]\nu ^2𝑑\nu $$
$`(2.6)`$
Resolving this integral we have
$$T^{mix}=\left[\alpha +\frac{e^2}{2}(1\alpha )\right]$$
$`(2.7)`$
Hence, if $`\alpha 1`$, then $`P(k)^{mix}P(k)`$ and $`\sigma ^2(R)^{mix}\sigma ^2(R)`$, which means that a sufficiently small contribution of the second field leaves the amplitude and shape of the power spectrum and the mass fluctuation practically unchanged.
## 3 Cluster Abundance Evolution
The correct framework to describe the evolution of non-linear objects in the context of this model requires a generalization of the Press & Schechter formalism (Press & Schechter 1974) in order to take into account the second field. Assuming that only regions with $`\nu >\nu _c`$ will form gravitationally bound objects with mass larger than $`M`$ by the time $`t`$, the fraction of these objects can be calculated through
$$F(M)=_{\nu _c}^{\mathrm{}}p(\nu )𝑑\nu ,$$
$`(3.1)`$
where $`\nu =\delta /\sigma _R`$. This quantity is transformed into the comoving number density of objects with mass between $`M`$ and $`M+dM`$ by taking $`F/M`$ and dividing it by $`(M/\rho _b)`$. Thus,
$$n(M)dM=2\left(\frac{\rho _b}{M}\right)\frac{}{M}\left[_{\nu _c}^{\mathrm{}}p(\nu )𝑑\nu \right]dM$$
$`(3.2)`$
where $`\rho _b`$ is the background density and the number 2 comes from the correction factor $`[_0^{\mathrm{}}p(\nu )𝑑\nu ]^1=2`$, which takes into account all the mass of the universe. If $`p(\nu )`$ is given by (2.2), then (3.2) can be written as
$$n(M)dM=2\left(\frac{\rho _b}{M}\right)\frac{}{M}\left[_{\nu _c}^{\mathrm{}}[\alpha f_1(\nu )+(1\alpha )f_2(\nu )]𝑑\nu \right]dM$$
$`(3.3)`$
Now, introducing (2.5) in (3.3) we have
$$n(M)dM=\sqrt{\frac{2}{\pi }}\left(\frac{\rho _b}{M}\right)\left[\alpha \left(\frac{\nu _c}{M}\right)e^{\frac{\nu _c^2}{2}}+(1\alpha )\left(\frac{\mathrm{ln}\nu _c}{M}\right)e^{\frac{(\mathrm{ln}\nu _c)^2}{2}}\right]dM$$
$`(3.4)`$
Following Sasaki (1994), we rewrite (3.4) to give the density of objects with mass in the range $`dM`$ about $`M`$ which virialize at the redshift $`z`$ and survive until the present epoch without merging with other systems. It becomes
$$n(M,z)=F(\mathrm{\Omega })\left(\frac{M}{M_{}(z)}\right)^{\frac{(n+3)}{3}}\sqrt{\frac{2}{\pi }}\left(\frac{\rho _b}{M^2}\right)\frac{(n+3)}{6}\left[\alpha A(M,z)+(1\alpha )B(M,z)\right]$$
$`(3.5)`$
where
$$F(\mathrm{\Omega })=\frac{5}{2}\mathrm{\Omega }\left[\frac{(1+\frac{3}{2}\mathrm{\Omega })}{(1+\frac{3}{2}\mathrm{\Omega }+\frac{5}{2}\mathrm{\Omega }z)^2}\right],A(M,z)=\left(\frac{M}{M_{}(z)}\right)^{\frac{(n+3)}{6}}exp\left[\frac{1}{2}\left(\frac{M}{M_{}(z)}\right)^{\frac{(n+3)}{3}}\right],$$
$$B(M,z)=exp[\frac{1}{2}\mathrm{ln}\left(\frac{M}{M_{}(z)}\right)^{\frac{(n+3)}{6}}]^2\mathrm{and}M_{}(z)=M_{}(1+z)^{6/(n+3)}$$
Eq.(3.5) allows us to compare the cluster abundance evolution with observational data. Clusters, as the most massive collapsed structures, correspond to rare peaks in the primordial density field and so their abundance is sensitive to the occurence of non-Gaussianity in the density fluctuation distribution. Also, cluster evolution provides a constraint on the amplitude of the mass fluctuation at 8 $`h^1`$ Mpc scale, $`\sigma _8`$, and on the cosmological density parameter, $`\mathrm{\Omega }_m`$, through the relation $`\sigma _8\mathrm{\Omega }_m^{0.5}0.5`$ (e.g. Henry & Arnaud 1991; Pen 1998). In a recent work, Bahcall (1999) shows that several independent methods based on clusters data indicate a low mass densitiy in the universe, $`\mathrm{\Omega }_m0.2`$ and, in consequence, $`\sigma _81.2`$, breaking the degeneracy between these parameters.
Here, we compare the behaviour of the cluster abundance evolution given by (3.5) with data compiled by Bahcall & Fan (1998). As an example, we plot in Figure 1a some fits to the observational data for two different values of $`\mathrm{\Omega }_m`$ (0.2 and 1.0). Note that our model is very sensitive to the parameter $`\alpha `$. Even for $`(1\alpha )10^310^4`$ (i.e., almost Gaussian initial conditions), the curves diverge significantly from the strict Gaussian cases. This means that even very small deviations from Gaussianity may introduce a significant change in the cluster abundance. Actually, the presence of the second field tends to slow down the cluster abundance evolution at high redshifts. In the case of $`\mathrm{\Omega }_m=1.0`$ this effect is dramatic for $`z>0.3`$, while in the case of $`\mathrm{\Omega }_m=0.2`$ the difference is less pronounced and it is clearer for $`z0.6`$. Indeed, by plotting the 68% confidence limits around the curve $`\mathrm{\Omega }_m=0.2`$, we see that Gaussian and non-Gaussian models are not clearly distinguishable for $`z1`$ (see Figure 1b). This is associated to the small number of observational points and, possibly, to the simplicity of our model. However, even considering these caveats, our results seem to indicate that small deviations from the strict Gaussianity may play an important role in the cluster abundance evolution.
## 4 Summary and Discussion
We presented the first results of a study concerning small deviations from Gaussianity in the primordial density field. Using very simple arguments, we developed a model based on the combination of two random fields in order to take into account the non-Gaussianity effects. This model is physically motivated in the context of hybrid models, as well as in the framework of biased scenarios for structure formation. The weighted combined field involves a parameter $`\alpha `$ which modulates the relative importance of its components. For $`\alpha 1`$, we preserve the amplitude and shape of $`P(k)`$ and $`\sigma (R)`$ almost the same as in the Gaussian case. At the same time, our results suggest that even very small values of $`(1\alpha )`$ can introduce a significant change in the cluster abundance evolution. This effect seems to be stronger in high density universes (at $`z1`$) than in low density universes where the effect probably turns more important at higher redshifts.
The model has some drawbacks. Firstly, it depends on the choice and amplitude of the second component of the combined field. Our choice of the lognormal function had a mathematical criterion of simplicity. A detailed investigation of the use of different distribution functions as the second component will be the subject of future works. However, the reasonable agreement between the model and the data gives some support to our arbitrary choice. Other possible limitation of this work comes from the use of the analytical approximation to the density of non-linear objects following Sasaki (1994). A more accurate description of the cluster abundance evolution requires the utilization of numerical methods. But Blain & Longair (1993), also working in the Press & Schecter framework, found results numerically similar to Sasaki’s, so it seems that using this analytical approximation does not introduce any systematical error. Finally, we should keep in mind that our results are preliminary and both theoretical and observational efforts are necessary in order to confirm or disproof the hypothesis that the primordial density field can be described as a slightly non-Gaussian distribution.
We thank the anonymous referee for useful suggestions. A.L.B. Ribeiro and P.S. Letelier thank the support of FAPESP. P.S. Letelier also thanks the support of CNPq.
Fig. 1a - Galaxy cluster abundance in the two-component model for $`\mathrm{\Omega }_m=0.2`$ with $`\alpha =0.9990`$ (solid) and $`\alpha =1`$ (dashed) and for $`\mathrm{\Omega }_m=1.0`$ with $`\alpha =0.9999`$ (solid) and $`\alpha =1`$ (dashed). The curves, normalized at $`z=0`$, correspond to $`n=1.0`$, $`M_{}=10^{14}h^1M_{\mathrm{}}`$ and $`M>8\times 10^{14}h^1M_{\mathrm{}}`$. The observational points were taken from Bahcall & Fan (1998).
Fig. 1b - Galaxy cluster abundance in the two-component model for $`\mathrm{\Omega }_m=0.2`$ with $`\alpha =0.9990`$ (solid) and $`\alpha =1`$ (dashed). The dotted lines are the 68% confidence limits around the non-Gaussian fit. The observational points were taken from Bahcall & Fan (1998).
|
warning/0003/cond-mat0003114.html
|
ar5iv
|
text
|
# Quenched Spin Tunneling and Diabolical Points in Magnetic Molecules: I. Symmetric Configurations
## I Introduction
In several previous papers , we have discussed the tunneling of a spin described by the model Hamiltonian
$$=k_2J_z^2+(k_1k_2)J_x^2g\mu _B𝐉𝐇,$$
(1)
where $`𝐉`$ is dimensionless spin operator with components $`J_x`$, $`J_y`$, and $`J_z`$, $`𝐇`$ is an external magnetic field, and $`k_1>k_2>0`$. Since most of the earlier work has been for the special case where $`𝐇\widehat{𝐱}`$, let us first limit ourselves to that. Viewed as a function of a classical angular momentum vector $`𝐉`$ of fixed length $`J`$, this Hamiltonian has two degenerate minima for $`H_x<H_c=2k_1J/g\mu _B`$. On general grounds we expect quantum mechanical tunneling to lift the degeneracy, and split the energies. The surprise is that the ground state tunnel splitting, $`\mathrm{\Delta }_0`$, is an oscillatory function of $`H_x`$, vanishing exactly whenever
$$\frac{H_x}{H_c}=\frac{\sqrt{1\lambda }}{J}\left[Jn\frac{1}{2}\right],$$
(2)
where $`\lambda =k_2/k_1`$, and $`n=0,1,\mathrm{},2J1`$.
The quenching of $`\mathrm{\Delta }_0`$ was first studied purely as a theoretical curiosity, and explained in terms of instantons (a). Since then, this effect has been observed in the magnetic molecule \[(tacn)<sub>6</sub>Fe<sub>8</sub>O<sub>2</sub>(OH)<sub>12</sub>\]<sup>8+</sup> (or just Fe<sub>8</sub> for short), which is approximately described by the Hamiltonian (1), with $`J=10`$, and $`k_10.33`$ K, and $`k_20.22`$ K . Motivated by a desire to use only elementary methods of analysis, the problem was restudied (d) using the discrete phase integral (DPI) method (also known as the discrete Wentzel-Kramers-Brillouin method). An approximate version of this method was developed and applied to Fe<sub>8</sub> independently by Villain and Fort . The results of these later studies confirm Eq. (2) for the ground pair quenching points, and also find additional quenching points, as we discuss next.
To help grasp the full richness of the spectrum of the Hamiltonian (1), we show in Fig. 1 the results of numerical calculation of the energies as a function of $`H_x`$, for $`J=3`$, for three different values of $`H_z`$. In all three cases, $`H_y=0`$. In part (a), $`H_z=0`$, and we have the symmetric situation mentioned above. Note that (i) the lowest two energy level curves cross six times (including negative values of $`H_x`$), and (ii) the crossing points are perfectly periodically spaced, in complete accord with Eq. (2). Wernsdorfer and Sessoli have shown the existence of analogous crossings for Fe<sub>8</sub>. To quantitatively account for the observed period, one must include higher order anisotropies in the Hamiltonian. This does not change the basic physics. In addition, Fig. 1(a) also shows a number of crossings of higher energy levels, of which the analog in Fe<sub>8</sub> has not been seen yet.
In Fig. 1(b), $`H_z`$ has a specific non-zero value. The problem is no longer symmetric, and one of the classical minima is lower than the other. Correspondingly, we see that the lowest quantum mechanical state is always non degenerate. However, ignoring tunneling for the moment, the first excited state in the deeper well can have an energy equal to that of the lowest state in the shallower well if $`H_z`$ is correctly chosen. And indeed, we see from the figure that the second and third energy levels do cross a number of times. These crossings, when $`𝐇`$ has an easy component, were not anticipated in Ref. (a), and were discovered by Wernsdorfer and Sessoli experimentally. As seen in the experiments, the crossings in Fig. 1(b) are shifted by half a period from those in Fig. 1(a). Note that as in part (a), Fig. 1(b) displays crossings between yet higher energy levels (the fourth and fifth, e.g.), which have also not been seen experimentally yet.
This pattern continues as $`H_z`$ is increased still further \[Fig. 1(c)\]. Now the lowest two levels in the deeper well are nondegenerate, and the lowest crossings are between levels 3 and 4. Comapred to Fig. 1(b), these crossings are shifted by yet another half-period, just as seen experimentally. Again, there are crossings between higher pairs of levels, and again only those between levels 3 and 4 have been seen in Fe<sub>8</sub>.
It is clearly interesting to understand the structure in the energy spectrum analytically, and numerical diagonalization alone cannot provide this. When $`J`$ is of order 10, as it is for Fe<sub>8</sub>, a semiclassical analysis is natural, and it is profitable to think of the energy differences amongst low lying levels in terms of tunneling. Such an analysis was done in Ref. (a–d). In this paper, we shall elaborate on our earlier DPI analysis (d), and provide several results that are more generally applicable to Hamiltonians other than Eq. (1). We shall limit ourselves, however, to problems which are analogous to symmetric double-well potentials in the continuum WKB case. In the context of Eq. (1), this means that $`𝐇\widehat{𝐱}`$. The cases where $`H_y`$ or $`H_z`$ are also nonzero correspond to an asymmetric potential, and will be considered in a second paper.
Before describing the results of our analysis, however, let us digress to make two points. The first is the issue of degeneracy and its connection with symmetry in light of the von Neumann-Wigner theorem. When $`𝐇\widehat{𝐱}`$, or $`𝐇\widehat{𝐳}`$, $``$ is invariant under a $`180^{}`$ rotation about $`\widehat{𝐱}`$ or $`\widehat{𝐳}`$, so energies of levels that are odd and even under this operation can intersect. The quenchings for $`𝐇\widehat{𝐱}`$ \[Fig. 1(a)\] can be understood as instances of this phenomenon (b). When $`𝐇`$ has both $`\widehat{𝐱}`$ and $`\widehat{𝐳}`$ components, however, $``$ has no symmetry, and the level crossings \[Fig. 1(b), (c)\] are nontrivial instances of conical intersections or diabolical points . Viewed in the larger $`H_x`$$`H_z`$ plane, or the full three-dimensional space of magnetic fields $`𝐇`$, however, all points of degeneracy are diabolical.
The second point is that there are several other special features in the spectrum, which are evident from numerical analysis for several different $`J`$, and can also be seen in Fig. 1. First, the successive half-period shifts in the crossing fields as we go from (a) to (b) to (c) in Fig. 1 mean that the diabolical points form part of a centered rectangular lattice in the $`H_x`$$`H_z`$ plane. The length of the rectangular unit cell along $`H_x`$ can be read off Eq. (2), while that along $`H_z`$ is given by
$$\mathrm{\Delta }H_z=\frac{\lambda ^{1/2}}{J}H_c,$$
(3)
where $`\lambda =k_2/k_1`$. Second, at a diabolical point, we often find simultaneous degeneracy of more than one pair of levels to very high accuracy if not exactly. All these facts are captured by the DPI analysis. In fact, in the case $`𝐇\widehat{𝐱}`$, all the available evidence to date — exact diagonalization for small $`J`$, perturbation theory in $`\lambda k_2/k_1`$, numerics — indicates that the simultaneity of the degeneracy of many pairs of levels, as well as the values of the degeneracy fields, are exactly given by the leading semiclassical analysis, i.e., Eq. (2) (e). These facts point to the existence of a higher dynamical symmetry, but that is not yet established. Further, when higher order anisotropy terms are included in the Hamiltonian to obtain quantitative agreement with experimentally observed period , the numerical evidence indicates that although the simultaneous degeneracy of several pairs of levels and the perfect lattice of diabolical points are no longer exact properties, they continue to hold to rather good approximation .
The plan of the paper is as follows. In Sec. II we outline the DPI approach. The basic idea is that in the $`J_z`$ basis, Schrödinger’s equation for Eq. (1) has the form of a recursion relation or difference equation, as opposed to a differential equation for a massive particle in a one dimensional potential $`V(x)`$. This difference equation can be solved in close analogy with the continuum WKB approximation. We will see that compared to previous DPI studies new types of turning points arise in the study of Eq. (1), because the recursion relation has five terms as opposed to three in the earlier studies. These turning points have no continuum analogue. Our present discussion will rely on physical arguments and correspondence with the continuum case. A more formal discussion is given in Ref. .
In Sec. III we develop an analogue of Herring’s formula for problems leading to five term recursion relations. In the continuum case, for a symmetric double well potential \[$`V(x)=V(x)`$\], this formula expresses the splitting for the $`n`$th pair of levels in terms of the $`x=0`$ values of the wavefunction $`\psi _n(x)`$ and its derivative $`\psi _n^{}(x)`$ for the $`n`$th state localized in one of the wells. In Sec. IV we will use the DPI method to find the analogous discrete wavefunction near the center of the potential, and use our Herring formula to obtain a completely general formula \[See Eq. (67)\] that applies to any eigenvalue problem in the form of a recursion relation. This latter formula is written in terms of an action integral that runs between turning points, in close analogy with the continuum case. This formula is inconvenient for practical applications, however, and so in subsection E of Sec. IV, we will transform it into another result \[Eq. (68)\] that only requires the evaluation of a small number of much simpler integrals. The second formula is also completely general, and has the advantage of making the $`J\mathrm{}`$ asymptotic structure of the splittings transparent. In Sec. V, we will apply this lattermost formula to the Fe<sub>8</sub> problem, and obtain the splitting for all pairs of levels. We will discuss the quenching points and several other aspects of our results, including comparison with numerics, work by other authors , and the features that appear to be exact.
## II Summary of the DPI Method
The starting point is to write Schrödinger’s equation in the $`J_z`$ basis. Suppose $`|\psi `$ is an eigenstate of $``$ with energy $`E`$. Then with $`J_z|m=m|m`$, $`m|\psi =C_m`$, $`m||m=w_m`$, and $`m||m^{}=t_{m,m^{}}`$ ($`mm^{}`$), we have
$$\stackrel{}{}_{n=m2}^{m+2}t_{m,n}C_n+w_mC_m=EC_m,$$
(4)
where the prime on the sum indicates that the term $`n=m`$ is to be omitted. The diagonal terms ($`w_m`$) arise from the $`J_z^2`$ part of $``$, the $`t_{m,m\pm 1}`$ terms from the $`J_xH_x`$ part, and the $`t_{m,m\pm 2}`$ terms from the $`J_x^2`$ part.
We can think of Eq. (4) as a tight binding model for an electron in a one-dimensional lattice with sites labelled by $`m`$, and slowly varying on-site energies ($`w_m`$), nearest-neighbor ($`t_{m,m\pm 1}`$), and next-nearest-neighbor ($`t_{m,m\pm 2}`$) hopping terms. Since we can think of dynamics in this model in terms of wavepackets, it is clear that there is a generalization of the usual continuum quasiclassical or phase integral method to the lattice case. This is the DPI method.
Previous work with the DPI method has been limited to the case where the recursion relation has only three terms. New features arise when five or more terms are considered. In particular, we encounter nonclassical turning points, i.e., turning points at $`m`$ values other than those at the limits of the classically allowed motion. It is these turning points that give rise to oscillatory tunnel splittings, so that this effect is absent in systems described by three-term recursion relations.
The general formalism of this method and the extension to five terms is discussed at length elsewhere , so here we will only give a brief summary. The fundamental requirement for a quasiclassical approach to work is that $`w_m`$ and $`t_{m,m\pm \alpha }`$ ($`\alpha =1,2`$) vary slowly enough with $`m`$ that we can find smooth continuum approximants $`w(m)`$ and $`t_\alpha (m)`$, such that whenever $`m`$ is an eigenvalue of $`J_z`$, we have
$`w(m)`$ $`=`$ $`w_m,`$ (5)
$`t_\alpha (m)`$ $`=`$ $`(t_{m,m+\alpha }+t_{m,m\alpha })/2,\alpha =1,2.`$ (6)
We further demand that
$$\frac{dw}{dm}=O\left(\frac{w(m)}{J}\right),\frac{dt_\alpha }{dm}=O\left(\frac{t_\alpha (m)}{J}\right),$$
(7)
with $`m/J`$ being treated as quantity of order 1. We will see that for Eq. (1), these conditions will hold in the semiclassical limit $`J1`$.
Given these conditions, the basic approximation, which readers will recognize from the continuum case, is to write the wavefunction as a linear combination of the quasiclassical forms
$$C_m\frac{1}{\sqrt{v(m)}}\mathrm{exp}\left(i^mq(m^{})𝑑m^{}\right),$$
(8)
where $`q(m)`$ and $`v(m)`$ obey the equations
$`E`$ $`=`$ $`w(m)+2t_1(m)\mathrm{cos}q+2t_2(m)\mathrm{cos}(2q)_{\mathrm{sc}}(q,m),`$ (9)
$`v(m)`$ $`=`$ $`_{\mathrm{sc}}/q=2\mathrm{sin}q(m)\left(t_1(m)+4t_2(m)\mathrm{cos}q(m)\right).`$ (10)
Equations (9) and (10) are the lattice analogs of the eikonal and transport equations. Equation (8) represents the first two terms in an expansion of $`\mathrm{log}C_m`$ in powers of $`1/J`$.
As in the continuum case, the approximate DPI wavefunction is invalid at turning points. These points arise whenever the velocity $`v(m)`$ vanishes for given energy $`E`$, for then the approximation (8) diverges. We see from Eq. (10) that $`v(m)`$ can vanish either because $`\mathrm{sin}q=0`$, i.e., $`q=0`$ or $`q=\pi `$, or because $`q=q_{}\mathrm{cos}^1(t_1/4t_2)`$. Substituting these values of $`q`$ in the eikonal equation, we see that a turning point is obtained whenever
$$E=U_0(m),U_\pi (m),\mathrm{or}U_{}(m),$$
(11)
where
$`U_0(m)`$ $`=`$ $`_{\mathrm{sc}}(0,m)=w(m)+2t_1(m)+2t_2(m),`$ (12)
$`U_\pi (m)`$ $`=`$ $`_{\mathrm{sc}}(\pi ,m)=w(m)2t_1(m)+2t_2(m),`$ (13)
$`U_{}(m)`$ $`=`$ $`_{\mathrm{sc}}(q_{},m)=w(m)2t_2(m){\displaystyle \frac{t_1^2(m)}{4t_2(m)}}.`$ (14)
Note that at a turning point, both $`m`$ and $`q`$ are determined. If we denote the values of these quantities generically by $`m_c`$ and $`q_c`$, $`m_c`$ may be regarded as being fixed by Eq. (11), and $`q_c`$ by the corresponding condition $`q_c=0`$, $`q_c=\pi `$, or $`q_c=q_{}(m_c)`$.
To understand the nature of these turning points, let us assume that $`t_1<0`$, and $`t_2>0`$. \[This is the case for the Hamiltonian (1). We can always arrange for $`t_1`$ to be negative by means of the gauge transformation $`C_m(1)^mC_m`$. Thus there is only one other case to be considered, namely, $`t_1<0`$, $`t_2<0`$. This is discussed in Ref. .\] It then follows that $`U_\pi >U_0`$, and that
$$U_0(m)U_{}(m)=\frac{1}{4t_2(m)}\left(t_1(m)+4t_2(m)\right)^20.$$
(15)
Secondly, let us think of $`(q,m)`$ for fixed $`m`$ as an energy band curve. Then $`U_\pi `$ is always the upper band edge, while the lower band edge is either $`U_0`$ or $`U_{}`$ according as whether $`t_1/4t_2`$ is greater than or lesser than 1. To deal with this possibility, it pays to introduce a dual labelling scheme for all three curves $`U_0`$, $`U_\pi `$, and $`U_{}`$. We write $`U_\pi (m)U_+(m)`$, and
$`U_0(m)=U_i(m),U_{}(m)=U_{}(m),\mathrm{if}q_{}(0,\pi ),`$ (16)
$`U_0(m)=U_{}(m),U_{}(m)=U_f(m),\mathrm{if}q_{}(0,\pi ).`$ (17)
The subscripts $`+`$ and $``$ denote upper and lower band edges, while the subscripts i and f denote internal and forbidden respectively, since in the first case above, $`U_0`$ lies inside the energy band, while in the second case, $`U_{}`$ lies outside. As examples of these curves for a symmetric recursion relation, we show those for Fe<sub>8</sub> in Fig. 2. A magnified view of the lower left hand portion of this diagram is given in Fig. 3.
Turning points where $`E=U_+`$, or $`E=U_{}`$ when $`U_{}=U_0`$, are analogous to those encountered in the continuum quasiclassical method, since the energy lies at a limit of the classically allowed range for the value of $`m`$ in question. Points where $`E=U_{}`$ when $`U_{}=U_{}`$ are physically analogous, but mathematically different since the value of $`q_c`$ is neither $`0`$ nor $`\pi `$. Points where $`E=U_i`$ (see the energy $`E_1`$ in Fig. 2, e.g.) are novel in that the energy is inside the classically allowed range for $`m_c`$, but the mathematical form of the connection formulas is identical to the case $`E=U_{}=U_0`$ since $`q_c=0`$. Most interesting are the turning points with $`E=U_f`$ (the point $`m=m_1`$ in Fig. 3, for instance), since now the energy is outside the allowed range for $`m=m_c`$, and the value of $`q_c`$ is therefore necessarily complex. These points lie “under the barrier” and turn out to be the ones of importance for understanding oscillatory tunnel splittings.
The above discussion shows that the curves $`U_0`$, $`U_\pi `$, and $`U_{}`$ collectively play the same role as the potential energy in the continuum quasiclassical method. We refer to them as critical curves. We have already noted that $`U_\pi >U_0U_{}`$. Let us suppose that the case of equality in Eq. (15) occurs at $`m=m^{}`$. Clearly $`t_1(m^{})/4t_2(m^{})=1`$, which is precisely the condition found above for the lower band edge to change from $`q=0`$ to $`q=q_{}`$. Secondly, expanding $`t_1`$ and $`t_2`$ about $`m`$, we see that $`U_0`$ and $`U_{}`$ have a common tangent when they meet.
## III Herring’s formula for five-term recursion relations
The problem of computing tunnel splittings in a symmetric double well potential in the continuum case is greatly simplified by use of Herring’s formula . An entirely analogous formula can be derived in the discrete case (c) following the simplified treatment of Landau and Lifshitz .
We have already noted the importance of the critical curves. For low lying energy levels, in particular, the curve $`U_{}`$ is very much like the potential energy in the continuum case, and it is clear that we will have an entire series of approximate energy eigenstates with wavefunctions localized in any one of the two wells, in the vicinity of $`\pm m_0`$, the minima of $`U_{}(m)`$. (See Fig. 2.) Let $`C_m`$ be the $`n`$th such wavefunction localized in the right hand well, normalized to unit total probability, and let it satisfy Schrödinger’s equation (4) with an energy $`E_0`$ for all values of $`m`$ well to the right of the left well, including in particular the region around $`m=0`$. More precisely, we take $`C_m`$ to decay away from the right well in both directions. Such a function could be obtained, e.g., as the energy eigenfunction of a modified problem in which the on-site energy is increased by a large positive amount for all $`m<m_a`$, where $`m_0m_a0`$, it being understood that $`m_a`$ is far away from all turning points for the energy concerned. However, this problem need not be solved explicitly, as the exact behavior of $`C_m`$ near $`m=m_0`$ is never needed, and therefore need not be examined too closely.
Given such a function, Herring shows that the true symmetric and antisymmetric eigenfunctions, $`s_j`$ and $`a_j`$, with energies $`E_1`$ and $`E_2`$ respectively, are given very accurately by
$$\begin{array}{ccc}\hfill a_m& =& \frac{1}{\sqrt{2}}(C_mC_m),\hfill \\ \multicolumn{3}{c}{}\\ \hfill s_m& =& \frac{1}{\sqrt{2}}(C_m+C_m).\hfill \end{array}$$
(18)
The product $`C_mC_m`$ is exponentially small everywhere, these functions are normalized to unit total probaility to exponentially high accuracy.
The Schrödinger equations obeyed by $`C_m`$ and $`a_m`$ are
$`(w_mE_0)C_m+{\displaystyle \stackrel{}{}_{n=m2}^{m+2}}t_{m,n}C_n`$ $`=`$ $`0,`$ (19)
$`(w_mE_1)a_m+{\displaystyle \stackrel{}{}_{n=m2}^{m+2}}t_{m,n}a_n`$ $`=`$ $`0.`$ (20)
Let us now define $`m_r`$ to be $`1`$ if $`J`$ is integral, and $`1/2`$ when $`J`$ is half-integral. Multiplying Eq. (19) by $`a_m`$, Eq. (20) by $`C_m`$, and summing over $`m`$ from $`m_r`$ to $`J`$, we get
$$(E_1E_0)\underset{m=m_r}{}C_ma_m+\mathrm{\Sigma }_1\mathrm{\Sigma }_2=0,$$
(21)
where
$`\mathrm{\Sigma }_1`$ $`=`$ $`{\displaystyle \underset{m=m_r}{\overset{J}{}}}{\displaystyle \stackrel{}{}_{n=m2}^{m+2}}a_mt_{m,n}C_n,`$ (22)
$`\mathrm{\Sigma }_2`$ $`=`$ $`{\displaystyle \underset{m=m_r}{\overset{J}{}}}{\displaystyle \stackrel{}{}_{n=m2}^{m+2}}C_mt_{m,n}a_n.`$ (23)
To simplify Eq. (21), we first note that by Eq. (18)
$$\underset{m=m_r}{}C_ma_m\frac{1}{\sqrt{2}}\underset{m=m_r}{}C_m^2\frac{1}{\sqrt{2}},$$
(24)
since the product $`C_mC_m`$ is everywhere exponentially small, and since $`C_m^2`$ is concentrated almost completely in the right well. Secondly, most of the terms in the sums $`\mathrm{\Sigma }_1`$ and $`\mathrm{\Sigma }_2`$ can be seen to be identical by shifting the summation indices in various terms suitably, and making use of the symmetry $`t_{m,n}=t_{n,m}`$. For example, the difference between the terms in $`\mathrm{\Sigma }_1`$ with $`n=m+2`$, and those in $`\mathrm{\Sigma }_2`$ with $`n=m2`$ equals
$`{\displaystyle \underset{m=m_r}{\overset{J}{}}}\left(a_mt_{m,m+2}C_{m+2}C_mt_{m,m2}a_{m2}\right)`$ $`=`$ $`{\displaystyle \underset{m=m_r}{\overset{J}{}}}a_mt_{m,m+2}C_{m+2}{\displaystyle \underset{m=m_r2}{\overset{J2}{}}}C_{m+2}t_{m+2,m}a_m`$ (25)
$`=`$ $`a_{m_r1}t_{m_r1,m_r+1}C_{m_r+1}a_{m_r2}t_{m_r2,m_r}C_{m_r},`$ (26)
where we have made use of the obvious facts that $`t_{J,J+2}`$ and $`t_{J1,J+1}`$ are identically zero. The differences between the other terms in $`\mathrm{\Sigma }_1`$ and $`\mathrm{\Sigma }_2`$ can be similarly evaluated, and reduce to a small number of terms involving the product of an $`a`$ with a $`C`$, which can then be written entirely in terms of $`C`$’s using Eq. (18). Finally, we can see that $`E_1E_0`$ = $`E_0E_2`$ = $`\pm \mathrm{\Delta }/2`$, and the net result is that upto an irrelevant over all sign,
$$\mathrm{\Delta }=\{\begin{array}{cc}2\left[t_{0,1}C_0(C_1C_1)+t_{0,2}C_0(C_2C_2)+t_{1,1}(C_1^2C_1^2)\right],\hfill & \\ integer\text{J},\hfill & \\ 2t_{\frac{1}{2},\frac{1}{2}}\left(C_{\frac{1}{2}}^2C_{\frac{1}{2}}^2\right)+4t_{\frac{3}{2},\frac{1}{2}}\left(C_{\frac{1}{2}}C_{\frac{3}{2}}C_{\frac{1}{2}}C_{\frac{3}{2}}\right),\hfill & \text{half-integer }J\text{.}\hfill \end{array}$$
(27)
Herring gives a more careful justification of his formula by employing the Temple-Kato error bound on energy eigenvalues . His argument can be adapted word for word to the present problem, and shows that the error in the splitting as calculated via Eq. (27) is exponentially smaller than the splitting itself, by a factor such as $`e^{cJ}`$ where $`c>0`$. As $`J\mathrm{}`$, therefore, Eq. (27) is asymptotically correct.
We remind readers that Eq. (27) is not limited to the ground state splitting.
## IV General Formula for Tunnel Splitting in terms of action integrals
To apply Herring’s formula (27) to the Hamiltonian (1), we will use the DPI approximation for the wavefunction. Actually, we will take $`C_m`$ in Eq. (27) to be localized in the left well. This can only change the answer by a sign, which is not of interest to us anyway.
### A DPI form near potential well minimum
Let us first take up the problem of finding the DPI approximation to $`C_m`$ in somewhat general terms. Step 1 is to find $`C_m`$ in the classically allowed region near $`m_0`$, the minimum of $`U_{}(m)`$. (See Fig. 3.) We assume, as will be seen to be true for Eq. (1), that in this region $`U_{}=U_0`$. For energies close to $`U_{}(m_0)`$, and $`m`$ close to $`m_0`$, the eikonal equation can only be satisfied if $`q`$ is close to zero. We can therefore expand $`_{\mathrm{sc}}`$ in powers of $`m+m_0`$ and $`q`$:
$$_{\mathrm{sc}}(q,m)U_{}(m_0)+\frac{1}{2M}q^2+\frac{1}{2}M\omega _0^2(m+m_0)^2+\mathrm{}$$
(28)
where
$`M`$ $`=`$ $`\left[2t_1(m_0)+8t_2(m_0)\right]^1>0,`$ (29)
$`\omega _0^2`$ $`=`$ $`2(t_1+4t_2){\displaystyle \frac{^2U_{}}{m^2}}|_{m=m_0}.`$ (30)
Note that by virtue of Eq. (7), and its natural extension to second derivatives, $`\omega _0`$ is of order $`1/J`$ relative to $`t_1`$ and $`t_2`$.
The allowed eigenvalues and eigenfunctions can now be written down very simply by noting that the eikonal equation is also the Hamilton-Jacobi equation with $`q=\mathrm{\Phi }/m`$, $`\mathrm{\Phi }`$ being the action. Thus the problem is identical to that of a harmonic oscillator. (Alternatively, we could arrive at the same result by approximating the original recurrence relation by a differential equation in the vicinity of $`m_0`$.) For the $`n`$th state, therefore,
$$E_0=U_{}(m_0)+\left(n+\frac{1}{2}\right)\omega _0,$$
(31)
and
$$C_m=\left(2^{2n}(n!)^2\pi \xi ^2\right)^{1/4}e^{x^2/2\xi ^2}H_n(x/\xi ),$$
(32)
where $`x=m+m_0`$, $`H_n`$ is the $`n`$th Hermite polynomial, and $`\xi =(M\omega _0)^{1/2}`$. The wavefunction is already normalized, and the additional tails from the forbidden region only modify the normalization by an exponentially small amount.
It is apparent that the expansion (28) is invalid unless the point $`m_0`$ is sufficiently far from the edge $`m=J`$. Since the width of the wavefunction (32) is $`\sqrt{n}\xi `$, a necessary condition for the validity of our procedure is
$$Jm_0\sqrt{n}\xi .$$
(33)
If this condition does not hold, then the recursion relation must be solved near the edge by a different method, which is tantamount to using the Holstein-Primakoff or Bogoliubov transformation. An example of the latter approach is given in Sec. IV of Ref. .
From the viewpoint of the DPI method, we have two turning points very close to $`m_0`$, one to the left, and one to the right, since the condition $`E=U_{}(m)`$ is then satisfied. The one to the left has been discussed above. Let us now consider the one to the right, and denote it by $`m_t`$. We have
$$m_t+m_0=\left[\frac{2n+1}{M\omega _0}\right]^{1/2}(nJ)^{1/2}.$$
(34)
The neglected terms in Eq. (28), on the other hand, are of relative orders $`q^4`$, $`(m+m_0)^3/J^3`$, and $`(m+m_0)q^2/J`$, and thus smaller than $`n\omega _0`$ for $`x(nJ^2)^{1/3}`$. Thus, provided $`nJ`$, the solution (32) holds well past $`m_t`$, and can be matched directly onto the DPI solution under the barrier, without any need of connection formulas at $`m=m_t`$ . This argument is given at greater length in Sec. V of Ref. .
### B DPI form in ordinary forbidden region
Step 2 is to consider the DPI solution for $`m>m_t`$. Since we want this solution to decay as $`m`$ increases, we take it as
$$C_m=\frac{B}{\sqrt{|v(m)|}}\mathrm{exp}\left(_{m_t}^m\kappa (m^{})𝑑m^{}\right),$$
(35)
where $`\kappa (m)=\mathrm{Im}q(m)>0`$. This solution must be matched on to (32) to determine $`B`$. We can continue to use the harmonic oscillator approximation (28) to $`_{\mathrm{sc}}`$ for this purpose, and a simple calculation , which may in fact be traced back to Furry , leads to
$`B`$ $`=`$ $`\left({\displaystyle \frac{\omega _0g_n}{2\pi }}\right)^{1/2};`$ (36)
$`g_n`$ $`=`$ $`{\displaystyle \frac{\sqrt{2\pi }}{n!}}\left(n+\frac{1}{2}\right)^{n+\frac{1}{2}}e^{(n+\frac{1}{2})}.`$ (37)
The quantity $`g_n`$ is defined so that $`g_n1`$ as $`n\mathrm{}`$; $`g_0=(\pi /e)^{1/2}1.075`$, $`g_11.028`$, $`g_21.017`$, $`\mathrm{}`$.
### C DPI form in central region
Step 3 is to find the wavefunction in the central region near $`m=0`$. This is already done if there are no turning points between $`m_t`$ and $`m=0`$. For the Hamiltonian (1), it turns out that we encounter another turning point where $`E=U_f(m)`$ (the only possibility) at an intermediate point $`m=m_1`$ (see Fig. 3). The solutions for $`m<m_1`$ and $`m>m_1`$ must therefore be related by a connection formula. To understand this turning point, we note that the eikonal equation (9) may be solved for $`\mathrm{cos}q`$ as
$$\mathrm{cos}q(m)=\frac{t_1(m)\pm [t_1^2(m)4t_2(m)f(m)]^{1/2}}{4t_2(m)},$$
(38)
where $`f(m)=w(m)2t_2(m)E`$. Since $`\mathrm{cos}q=t_1/4t_2`$ at $`m=m_1`$, the discriminant in Eq. (38) must vanish, and we conclude that as we cross $`m_1`$, $`\mathrm{cos}q`$ changes from real to complex, and $`q`$ changes from imaginary to complex. \[Incidentally, it may be verified that the condition for vanishing discriminant, i.e.,
$$t_1^2(m)=4t_2(m)\left(w(m)2t_2(m)E\right),$$
(39)
is identical to $`E=U_{}(m)`$.\] Since the recursion relation is real, the solution $`C_m`$ must also be real for all $`m`$. A single DPI solution for $`m>m_1`$ cannot meet this demand, and so we must take a linear combination of two DPI forms with complex conjugate $`q`$’s. If we write these as
$$q_{1,2}(m)=i\kappa (m)\pm \chi (m),$$
(40)
with $`\kappa `$ and $`\chi `$ both real, then we must still have $`\kappa >0`$ in order that $`C_m`$ continue decaying, and we may also take $`\chi >0`$. Let us further write the solution (35) for $`m<m_1`$ as
$`C_m`$ $`=`$ $`{\displaystyle \frac{A}{2\sqrt{|v(m)|}}}\mathrm{exp}\left({\displaystyle _{m_1}^m}\kappa (m^{})𝑑m^{}\right),`$ (41)
$`A`$ $`=`$ $`2B\mathrm{exp}\left({\displaystyle _{m_t}^{m_1}}\kappa (m^{})𝑑m^{}\right).`$ (42)
Then, as shown in Ref. , the DPI solution for $`m>m_1`$ is given by
$$C_m=\mathrm{Re}\frac{A}{\sqrt{s_1(m)}}\mathrm{exp}\left(i_{m_1}^mq_1(m^{})𝑑m^{}\right),$$
(43)
where $`s_1(m)=iv\left(q_1(m)\right)`$ .
### D Herring’s formula with DPI approximation
The solution (43) is ripe for substitution into the Herring formula (27). To do this, we first note that
$`\mathrm{cosh}\kappa \mathrm{cos}\chi `$ $`=`$ $`t_1/4t_2,`$ (44)
$`\mathrm{sinh}\kappa \mathrm{sin}\chi `$ $`=`$ $`\left(4t_2ft_1^2\right)^{1/2}/4t_2.`$ (45)
It then follows that
$$s_1=8t_2(m)\mathrm{sinh}\kappa (m)\mathrm{sin}\chi (m)\mathrm{sin}q_1(m).$$
(46)
We now substitute Eqs. (43)–(46) into Herring’s formula, Eq. (27). In doing this, we may neglect the variation of quantities $`t_\alpha (m)`$, $`q(m)`$, and $`v(m)`$ among the sites near the center of the lattice, since the number of sites involved is of order $`1`$, and so the variation leads to higher order corrections in powers of $`1/J`$. To save writing, we denote quantities evaluated at $`m=0`$ by a bar: $`q_1(0)\overline{q}_1`$, $`\kappa (0)\overline{\kappa }`$, etc. We thus get
$$C_m=\mathrm{Re}A_2\frac{e^{i(\mathrm{\Omega }+m\overline{q}_1)}}{\sqrt{\mathrm{sin}\overline{q}_1}},$$
(47)
where,
$`\mathrm{\Omega }`$ $`=`$ $`{\displaystyle _{m_1}^0}q_1(m^{})𝑑m^{},`$ (48)
$`A_2`$ $`=`$ $`(8\overline{t}_2\mathrm{sinh}\overline{\kappa }\mathrm{sin}\overline{\chi })^{1/2}A.`$ (49)
The cases of integer and half-integer $`J`$ are best tackled separately. Doing the former first, we have
$`C_1C_1`$ $`=`$ $`iA_2[e^{i\mathrm{\Omega }}\sqrt{\mathrm{sin}\overline{q}_1}\mathrm{c}.\mathrm{c}.],`$ (50)
$`C_1+C_1`$ $`=`$ $`A_2[e^{i\mathrm{\Omega }}\sqrt{{\displaystyle \frac{\mathrm{cos}^2\overline{q}_1}{\mathrm{sin}\overline{q}_1}}}+\mathrm{c}.\mathrm{c}.],`$ (51)
$`C_0(C_1C_1)`$ $`=`$ $`i{\displaystyle \frac{A_2^2}{2}}[(e^{2i\mathrm{\Omega }}e^{2\mathrm{I}\mathrm{m}\mathrm{\Omega }}\sqrt{{\displaystyle \frac{\mathrm{sin}\overline{q}_1^{}}{\mathrm{sin}\overline{q}_1}}})\mathrm{c}.\mathrm{c}.],`$ (52)
$`C_2C_2`$ $`=`$ $`2iA_2[e^{i\mathrm{\Omega }}\mathrm{cos}\overline{q}_1\sqrt{\mathrm{sin}\overline{q}_1}\mathrm{c}.\mathrm{c}.],`$ (53)
$`C_0(C_2C_2)`$ $`=`$ $`iA_2^2[(e^{2i\mathrm{\Omega }}\mathrm{cos}\overline{q}_1e^{2\mathrm{I}\mathrm{m}\mathrm{\Omega }}\mathrm{cos}\overline{q}_1^{}\sqrt{{\displaystyle \frac{\mathrm{sin}\overline{q}_1^{}}{\mathrm{sin}\overline{q}_1}}})\mathrm{c}.\mathrm{c}.],`$ (54)
$`C_1^2C_1^2`$ $`=`$ $`iA_2^2[\mathrm{cos}\overline{q}_1(e^{2i\mathrm{\Omega }}\mathrm{cos}\overline{q}_1e^{2\mathrm{I}\mathrm{m}\mathrm{\Omega }}\sqrt{{\displaystyle \frac{\mathrm{sin}\overline{q}_1^{}}{\mathrm{sin}\overline{q}_1}}})\mathrm{c}.\mathrm{c}.].`$ (55)
Substituting these and the formula $`\overline{t}_1=4\overline{t}_2\mathrm{cosh}\overline{\kappa }\mathrm{cos}\overline{\chi }`$ into Eq. (27), we get
$`\mathrm{\Delta }`$ $`=`$ $`8A_2^2\overline{t}_2\mathrm{Im}\left(e^{2i\mathrm{\Omega }}\mathrm{\Theta }e^{2\mathrm{I}\mathrm{m}\mathrm{\Omega }}\sqrt{{\displaystyle \frac{\mathrm{sin}\overline{q}_1^{}}{\mathrm{sin}\overline{q}_1}}}\mathrm{Re}\mathrm{\Theta }\right);`$ (56)
$`\mathrm{\Theta }`$ $`=`$ $`\mathrm{cos}\overline{q}_1\mathrm{cosh}\overline{\kappa }\mathrm{cos}\overline{\chi }.`$ (57)
But, it follows from Eq. (40) that
$$\mathrm{cos}\overline{q}_1=\mathrm{cosh}\overline{\kappa }\mathrm{cos}\overline{\chi }i\mathrm{sinh}\overline{\kappa }\mathrm{sin}\overline{\chi },$$
(58)
so the second term in Eq. (56) vanishes altogether, and
$`\mathrm{\Delta }`$ $`=`$ $`4A_2^2\overline{t}_2\mathrm{sinh}\overline{\kappa }\mathrm{sin}\overline{\chi }(e^{2i\mathrm{\Omega }}+e^{2i\mathrm{\Omega }^{}})`$ (59)
$`=`$ $`\frac{1}{2}A^2(e^{2i\mathrm{\Omega }}+e^{2i\mathrm{\Omega }^{}}),`$ (60)
where we have used Eq. (49) in the last step.
For half-integer $`J`$, we get
$`C_{\pm 1/2}^2`$ $`=`$ $`{\displaystyle \frac{A_2^2}{4}}[({\displaystyle \frac{e^{2i\mathrm{\Omega }}}{\mathrm{sin}\overline{q}_1}}e^{\pm i\overline{q}_1}+{\displaystyle \frac{e^{2\mathrm{I}\mathrm{m}\mathrm{\Omega }}}{|\mathrm{sin}\overline{q}_1|}})+\mathrm{c}.\mathrm{c}.],`$ (61)
$`C_{\pm 1/2}C_{\pm 3/2}`$ $`=`$ $`{\displaystyle \frac{A_2^2}{4}}[({\displaystyle \frac{e^{2i\mathrm{\Omega }}}{\mathrm{sin}\overline{q}_1}}e^{\pm 2i\overline{q}_1}+{\displaystyle \frac{e^{2\mathrm{I}\mathrm{m}\mathrm{\Omega }}}{|\mathrm{sin}\overline{q}_1|}}e^{i\overline{q}_1})+\mathrm{c}.\mathrm{c}.].`$ (62)
Thus,
$`C_{1/2}^2C_{1/2}^2`$ $`=`$ $`{\displaystyle \frac{i}{2}}A_2^2(e^{2i\mathrm{\Omega }}\mathrm{c}.\mathrm{c}.),`$ (63)
$`C_{1/2}C_{3/2}C_{1/2}C_{3/2}`$ $`=`$ $`iA_2^2(\mathrm{cos}\overline{q}_1e^{2i\mathrm{\Omega }}\mathrm{c}.\mathrm{c}.),`$ (64)
and
$`\mathrm{\Delta }`$ $`=`$ $`iA_2^2[(\overline{t}_1+4\overline{t}_2)e^{2i\mathrm{\Omega }}\mathrm{c}.\mathrm{c}.]`$ (65)
$`=`$ $`4A_2^2\overline{t}_2\mathrm{sinh}\overline{\kappa }\mathrm{sin}\overline{\chi }(e^{2i\mathrm{\Omega }}+e^{2i\mathrm{\Omega }^{}}),`$ (66)
which leads, once again, to Eq. (60).
Collecting Eqs. (36), (42), (48) and (60), and making use of the symmetry of the problem, we may write the tunnel splitting for both integer and half-integer $`J`$ as
$$\mathrm{\Delta }=\frac{\omega _0g_n}{2\pi }[\mathrm{exp}\left(i_{m_t}^{m_t}q(m^{})dm^{}\right)+\mathrm{c}.\mathrm{c}.].$$
(67)
Here $`q(m^{})`$ is chosen to have a positive real part $`\chi `$ in the first term. We note once again that this result applies to higher pairs of excited states, and not just the ground pair. The essential dependence on $`n`$, the excitation number, enters through the $`n`$ dependence of $`m_t`$, the turning point.
The similarity of Eq. (67) to the final result in Ref. is striking , and one can ask whether one should not have anticipated it right away. For the ground state pair, the instanton approach (a) makes it very easy to understand the presence of two complex conjugate tunneling actions, and the fact that they should be superposed, but does not give the prefactor. The action integrals in the instanton approach, however, run not from turning point to turning point but from one minimum of the energy to the other. Further, properly justifiying the prefactor using instantons has proven very difficult . Purely as a recipe for calculations, however, a hybrid approach, in which one adds the tunneling actions from all equivalent instantons, and uses the DPI approach to determine the form of the prefactor, would appear to be valid for all problems. Thus, we strongly suspect that Eq. (67) is correct even when the recursion relation has seven or more terms.
### E Extraction of singular parts of action integrals
While the formula (67) is very general, it has the disadvantage that the action integral runs between turning points. The integrand is therefore close to a singularity, and for low lying states, this gives rise to terms in the action that depend on $`\mathrm{ln}J`$. Hence the formula does not reveal the asymptotic behavior as a function of $`J`$ in a transparent way.
In this subsection, we will show that we can write the splitting very simply in a way that does not suffer from the above drawback. The final result is
$$\mathrm{\Delta }_n=\frac{1}{n!}\sqrt{\frac{8}{\pi }}\omega _0F^{n+\frac{1}{2}}e^{\mathrm{\Gamma }_0}\mathrm{cos}\mathrm{\Lambda }_n,$$
(68)
where,
$`\mathrm{\Gamma }_0`$ $`=`$ $`2{\displaystyle _{m_0}^0}\kappa _0(m)𝑑m,`$ (69)
$`\mathrm{\Lambda }_n`$ $`=`$ $`2{\displaystyle _{m_1}^0}\left(\chi _0+(n+\frac{1}{2})\omega _0\chi _0^{}\right)𝑑m,`$ (70)
$`F`$ $`=`$ $`2M\omega _0(mm_1)^2\mathrm{exp}\left(2\left(Q_1+\omega _0{\displaystyle _{m_1}^0}\kappa _0^{}𝑑m\right)\right),`$ (71)
$`Q_1`$ $`=`$ $`{\displaystyle _{m_0}^{m_1}}\left({\displaystyle \frac{\omega _0B_0^{}}{\sqrt{B_0^21}}}+{\displaystyle \frac{1}{m+m_0}}\right)𝑑m.`$ (72)
In Eqs. (6972), the irregular turning points $`\pm m_1`$ may be evaluated by setting $`E=U_{}(\pm m_0)`$, and it should be recalled that $`\pm m_0`$ are the minima of $`U_{}(m)`$. Further,
$`\kappa _0`$ $`=`$ $`\kappa (m,ϵ=0);\kappa _0^{}=.{\displaystyle \frac{\kappa (m,ϵ)}{ϵ}}|_{ϵ=0},`$ (73)
$`\chi _0`$ $`=`$ $`\chi (m,ϵ=0);\chi _0^{}=.{\displaystyle \frac{\chi (m,ϵ)}{ϵ}}|_{ϵ=0},`$ (74)
$`B_0`$ $`=`$ $`\mathrm{cos}q(m,ϵ=0);B_0^{}=.{\displaystyle \frac{\mathrm{cos}q(m,ϵ)}{ϵ}}|_{ϵ=0},`$ (75)
with
$$ϵEU_{}(m_0).$$
(76)
The problem of finding the low level splittings is thus reduced to the evaluation of a handful of integrals. The proliferation of notation masks the actual simplicty of these formulas.
To derive these results, we follow a procedure similar to that used for the continuum case in Ref. . We begin by defining
$$\mathrm{\Phi }(ϵ)=i_{m_t(ϵ)}^0q(m,ϵ)𝑑m,$$
(77)
where the energy dependence is made explicit. The splitting for the $`n`$th pair of states is then given by
$$\mathrm{\Delta }_n=\frac{\omega _0g_n}{2\pi }(e^{2\mathrm{\Phi }(ϵ_n)}+\mathrm{c}.\mathrm{c}.),$$
(78)
with $`ϵ_n=(n+\frac{1}{2})\omega _0`$. Writing $`x=m+m_0`$ as in Eq. (32), the integrand in $`\mathrm{\Phi }`$ behaves as $`(x^2x_t^2)^{1/2}`$ near the lower limit, with $`x_t=m_t+m_0ϵ^{1/2}`$. Thus there is a singular part in $`\mathrm{\Phi }`$ of the form $`ϵ\mathrm{ln}ϵ`$, which it is our goal to extract. To this end, we differentiate Eq. (77) to get
$$\mathrm{\Phi }^{}(ϵ)=\frac{d\mathrm{\Phi }}{dϵ}=i_{m_t(ϵ)}^0\frac{q}{ϵ}𝑑m.$$
(79)
Note that the term arising from differentiating the lower limit vanishes, nor is there any explicit contribution from the singular behavior $`q(m+m_c)^{1/2}`$ for $`m`$ near $`m_c`$.
Next, let us divide $`\mathrm{\Phi }^{}(ϵ)`$ into two integrals, $`\mathrm{\Phi }_1^{}`$, in which the limits of integration are $`m_t`$ and $`m_1`$, and $`\mathrm{\Phi }_2^{}`$, which runs from $`m_1`$ to $`0`$. Defining
$$B_ϵ(m)=\mathrm{cos}\left(q(m,ϵ)\right),$$
(80)
we have
$$\mathrm{\Phi }_1^{}(ϵ)=_{m_t(ϵ)}^0\frac{B_ϵ^{}}{\sqrt{B_ϵ^2(m)1}}𝑑m,$$
(81)
where $`B_ϵ^{}=B_ϵ/ϵ`$. It follows from Eq. (28) that near $`m=m_t`$,
$$B_ϵ1+\left(\frac{1}{2}M\omega ^2x^2ϵ\right)M+\mathrm{},$$
(82)
so the integrand in Eq. (81) behaves as $`1/\omega _0(x^2x_t^2)^{1/2}`$. If we add and subtract the integral of this expression, we obtain
$$\mathrm{\Phi }_1^{}(ϵ)=\frac{1}{\omega _0}_{x_t}^{x_1}\frac{dx}{\sqrt{x^2x_t^2}}+_{x_t}^{x_1}\left[\frac{B_ϵ^{}}{\sqrt{B_ϵ^2(m)1}}+\frac{1}{\omega _0\sqrt{x^2x_t^2}}\right]𝑑x,$$
(83)
where $`x_1=m_0m_1`$. The first integral can be evaluated exactly. In the second integral we can put $`ϵ=0`$ both in the limits and in the integrand, since we are not interested in terms of $`O(ϵ)`$. Ignoring terms of this order throughout, and making use of the relation
$$x_t^2=2ϵ/M\omega _0^2,$$
(84)
we obtain
$$\mathrm{\Phi }_1^{}(ϵ)=\frac{1}{2\omega _0}\left[\mathrm{ln}\frac{ϵ}{2M\omega _0(m_0m_1)^2}+2Q_1\right],$$
(85)
where $`Q_1`$ is given by Eq. (72). Also, we can evaluate $`m_1`$ at $`ϵ=0`$.
The remaining contribution to $`\mathrm{\Phi }^{}(ϵ)`$, $`\mathrm{\Phi }_2^{}(ϵ)`$, can be evaluated simply by putting $`ϵ=0`$, since the neglected part is $`O(ϵ)`$. Recalling the definitions (73) and (74), we have
$$\mathrm{\Phi }_2^{}(ϵ)_{m_1}^0(\kappa _0^{}i\chi _0^{})𝑑m.$$
(86)
We now integrate the expression for $`\mathrm{\Phi }^{}(ϵ)`$ and obtain $`\mathrm{\Phi }`$. It is useful to separate the real and imaginary parts of the answer at this stage. For the real part, we get
$$\mathrm{\Gamma }=2\mathrm{Re}\mathrm{\Phi }=\mathrm{\Gamma }_0+\frac{ϵ}{\omega _0}\left[2Q_11+\mathrm{ln}\frac{ϵ}{2M\omega _0(m_0m_1)^2}+2\omega _0_{m_1}^0\kappa _0^{}𝑑m\right],$$
(87)
with $`\mathrm{\Gamma }_0`$ given by Eq. (69), while for the imaginary part, $`\mathrm{\Lambda }_n2\mathrm{Im}\mathrm{\Phi }`$, we get Eq. (70).
Substituting Eqs. (87)–(70), and the definition (37) of $`g_n`$ in the formula (78) for $`\mathrm{\Delta }_n`$, and recalling that $`ϵ_n=(n+\frac{1}{2})\omega _0`$, we finally obtain the answer quoted at the start, Eq. (68).
## V Application to the Fe<sub>8</sub> problem
We now apply our general formulas to the specific problem of Fe<sub>8</sub>, as described by the Hamiltonian (1). The various matrix elements of this Hamiltonian are given by
$`w_m`$ $`=`$ $`{\displaystyle \frac{1}{2}}(k_1+k_2)[J(J+1)m^2],`$ (88)
$`t_{m,m+1}`$ $`=`$ $`{\displaystyle \frac{1}{2}}g\mu _BH_x[J(J+1)m(m+1)]^{1/2},`$ (89)
$`t_{m,m+2}`$ $`=`$ $`{\displaystyle \frac{1}{4}}(k_1k_2)\left[[J(J+1)m(m+1)][J(J+1)(m+1)(m+2)]\right]^{1/2}.`$ (90)
We must now replace these by continuous functions $`w(m)`$, $`t_1(m)`$, and $`t_2(m)`$. Since our formalism requires knowing the first two terms in the action in an expansion in powers of $`1/J`$, it follows that we need only determine the functions $`w(m)`$ etc. to the same order. Furthermore, this determination need not be made in the form of a power series, and any functional representation that gives the first two terms correctly will be adequate. The most convenient way to do this is to replace the combination $`J(J+1)`$ in the above expressions by $`\overline{J}^2`$, where
$$\overline{J}=J+\frac{1}{2}.$$
(91)
The evaluation of the integrals (79)–(72) is then lengthy, but straightforward. We will present and discuss the final results first, and give the details of the analysis later.
### A Tunnel splittings for Fe<sub>8</sub>
The final result for the splitting of the $`n`$th pair of levels is
$$\mathrm{\Delta }_n=\frac{1}{n!}\sqrt{\frac{8}{\pi }}\omega _0F^{n+\frac{1}{2}}e^{\mathrm{\Gamma }_0}\mathrm{cos}\mathrm{\Lambda }_n,$$
(92)
where,
$`\omega _0`$ $`=`$ $`2J[k_1k_2(1h_{x0}^2)]^{1/2},`$ (93)
$`F`$ $`=`$ $`8J{\displaystyle \frac{\lambda ^{1/2}(1h_x^2)^{3/2}}{1\lambda h_x^2}},`$ (94)
$`\mathrm{\Gamma }_0`$ $`=`$ $`\overline{J}\left[\mathrm{ln}\left({\displaystyle \frac{\sqrt{1h_x^2}+\sqrt{\lambda }}{\sqrt{1h_x^2}\sqrt{\lambda }}}\right){\displaystyle \frac{h_x}{\sqrt{1\lambda }}}\mathrm{ln}\left({\displaystyle \frac{\sqrt{(1h_x^2)(1\lambda )}+h_x\sqrt{\lambda }}{\sqrt{(1h_x^2)(1\lambda )}h_x\sqrt{\lambda }}}\right)\right],`$ (95)
$`\mathrm{\Lambda }_n`$ $`=`$ $`\mathrm{max}\{0,\pi J\left(1{\displaystyle \frac{H_x}{\sqrt{1\lambda }H_c}}\right)n\pi \}.`$ (96)
In Eqs. (9396), $`\lambda =k_2/k_1`$, and
$$h_x=\frac{JH_x}{\overline{J}H_c},h_{x0}=\frac{H_x}{H_c}.$$
(97)
Recall that $`H_c=2k_1J/g\mu _B`$.
Let us now turn to the discussion of these results. The first point concerns the fields where the $`n`$th tunnel splitting vanishes. Taking account of the fact that $`\mathrm{\Lambda }_n`$ is necessarily positive as indicated by Eq. (96), we see that this happens whenever (d,e)
$$\frac{H_x}{H_c}=\frac{\sqrt{1\lambda }}{J}\left[J\mathrm{}\frac{1}{2}\right],$$
(98)
with $`\mathrm{}=n`$, $`n+1`$, $`\mathrm{}`$, $`2Jn1`$, yielding $`2(Jn)`$ quenching points in all for $`\mathrm{\Delta }_n`$. When $`n=0`$, these are the results quoted in Sec. I.
In Fig. 4 we compare Eq. (92) with the numerically evaluated splittings for the first three pairs of levels. Within our numerical precision, we always find the zeros of $`\mathrm{\Delta }_n`$ to agree with Eq. (98). Note, however, that for other values of $`H_c`$, the discrepancy between the numerics and Eq. (92) is well outside our numerical error, so that Eq. (92) is not exact, even though as an asymtotic estimate of the splitting it is rather good. This means that the leading semiclassical approximation does not give the eigenvalues themselves exactly, and only the quenching points appear to be so reproduced. The second point to note is that for $`n=1`$ (the pair of first excited states in each well), the highest field quenching point is lost, for $`n=2`$, the highest two points are lost, and so on, exactly as indicated by Eq. (98).
Next, let us compare our answers with previous work. Let us consider the Gamow factor $`\mathrm{\Gamma }_0`$ first. Except for the replacement of $`J`$ by $`\overline{J}`$ and $`h_{x0}`$ by $`h_x`$, this is precisely the action in Eq. (3.10) of Ref. (c). This agreement is unsurprising, because if we write $`\mathrm{\Delta }_n`$ in the form of a prefactor $`c_1`$ times a Gamow factor $`\mathrm{exp}(Jc_2)`$ where $`c_2=O(1)`$, then the $`J\overline{J}`$, and $`h_{x0}h_x`$ corrections in Eq. (95) represent terms that should be included in the prefactor $`c_1`$, which we did not seek to find in Ref. (c). The detailed form of the prefactor is perhaps more interesting. Up to multiplicative terms of order $`J^0`$, our answer for $`\mathrm{\Delta }_n`$ agrees precisely with that in Ref. . We do not understand, however, how these papers have succeeded in sidestepping the difficulties in the path integral treatment that were noted by Enz and Schilling , and by Belinicher, Providencia, and da Providencia . In Ref. , for instance, the problem is treated by writing the spin coherent state expectation value of the Hamiltonian (1) in spherical polar coordinates, and integrating out $`\mathrm{cos}\theta `$ (the $`J_z`$ projection) exactly, and then addressing the resulting effective Lagrangian for the $`\varphi `$ coordinate exactly as for a massive particle in one dimension. In performing the integration over $`\theta `$, however, it is not clear to us why $`S^2`$ is replaced by $`S(S+1)`$ in the scalar potential $`V(\varphi )`$ \[see Eq. (12) there\], but not in the vector potential $`\mathrm{\Theta }(\varphi )`$.
A related point, which is relatively minor, but has scope for creating confusion, is that if the Gamow factor is written as $`\mathrm{exp}(Jc_2)`$ with $`c_2=O(1)`$, then it is safest to write the $`J`$ dependence of the prefactor as $`\omega _0J^{n+1/2}`$, since $`\omega _0`$ depends on parameters such as $`k_1`$ and $`k_2`$, whose scaling with $`J`$ is a matter of choice, at least as far as model Hamiltonians are concerned.
One further check is obtained by considering the limiting case $`H_x=0`$, answers for which are known. \[See, e.g., Eq. (16) of , Eq. (48) of , or .\] Transcribing Eqs. (4.30) and (4.31) from in terms of the present parameters, we get
$`\mathrm{\Delta }_n`$ $`=`$ $`{\displaystyle \frac{1}{n!}}F_0^n\mathrm{\Delta }_0;`$ (99)
$`F_0`$ $`=`$ $`8J{\displaystyle \frac{\sqrt{\lambda }}{1\lambda }},`$ (100)
$`\mathrm{\Delta }_0`$ $`=`$ $`8\omega _{00}\left({\displaystyle \frac{J}{\pi }}\right)^{1/2}{\displaystyle \frac{\lambda ^{1/4}}{1+\sqrt{\lambda }}}\left({\displaystyle \frac{1\sqrt{\lambda }}{1+\sqrt{\lambda }}}\right)^J,`$ (101)
with $`\omega _{00}=2J(k_1k_2)^{1/2}`$. It follows from Eqs. (9396) that as $`H_x0`$, $`\omega _0\omega _{00}`$, $`FF_0`$ \[see Eq. (94)\], $`\mathrm{cos}\mathrm{\Lambda }_n\pm 1`$, and
$$\mathrm{\Gamma }_0\left(J+\frac{1}{2}\right)\mathrm{ln}\frac{1+\sqrt{\lambda }}{1\sqrt{\lambda }}.$$
(102)
It is then easy to see that our present answers for $`\mathrm{\Delta }_n`$ go over precisely into Eqs. (99)–(101).
### B Evaluation of Action Integrals
To carry out the evaluation of Eqs. (79)–(72), it is convenient to measure energies (including $`\omega _0`$) in units of $`k_1\overline{J}^2`$, and introduce the scaled variable $`\mu =m/\overline{J}`$. In terms of these variables, we have
$`w(m)`$ $`=`$ $`(1+\lambda )(1\mu ^2)/2,`$ (103)
$`t_1(m)`$ $`=`$ $`h_x(1\mu ^2)^{1/2},`$ (104)
$`t_2(m)`$ $`=`$ $`(1\lambda )(1\mu ^2)/4.`$ (105)
The turning points $`\mu _0=m_0/\overline{J}`$, and $`\mu _1=m_1/\overline{J}`$ (for $`ϵ=0`$) are given by
$`\mu _0`$ $`=`$ $`(1h_x^2)^{1/2},`$ (106)
$`\mu _1`$ $`=`$ $`[(1\lambda h_x^2)/(1\lambda )]^{1/2}.`$ (107)
It is most convenient to express everything in terms of $`\mu _0`$ and $`\mu _1`$, so we give inverse formulas as well:
$`h_x=(1\mu _0^2)^{1/2},`$ (108)
$`\lambda =(\mu _0^2\mu _1^2)/(1\mu _1^2).`$ (109)
The mass and the small oscillation frequency are given by
$`M={\displaystyle \frac{1}{2\lambda h_x^2}}={\displaystyle \frac{1}{2}}{\displaystyle \frac{1\mu _1^2}{(1\mu _0^2)(\mu _0^2\mu _1^2)}},`$ (110)
$`\omega _0=2[\lambda (1h_x^2)]^{1/2}/\overline{J}={\displaystyle \frac{2\mu _0}{\overline{J}}}\left({\displaystyle \frac{\mu _0^2\mu _1^2}{1\mu _1^2}}\right)^{1/2}.`$ (111)
To evaluate the integrals, we need expressions for $`\kappa _0`$, $`\chi _0`$, $`\kappa _0^{}`$, etc. in the ranges $`\mu _1<\mu <\mu _0`$, and $`0<\mu <\mu _1`$. The requisite calculations are straightforward so we give the main results only. First, in the range $`\mu _1<\mu <\mu _0`$, we get
$`B_0`$ $`=`$ $`\mathrm{cosh}\kappa _0={\displaystyle \frac{1\mu _1^2[(\mu _0^2\mu _1^2)(\mu ^2\mu _1^2)]^{1/2}}{[(1\mu _0^2)(1\mu ^2)]^{1/2}}},`$ (112)
$`\sqrt{B_0^21}`$ $`=`$ $`\mathrm{sinh}\kappa _0={\displaystyle \frac{\left(\sqrt{\mu _0^2\mu _1^2}\sqrt{\mu ^2\mu _1^2}\right)\sqrt{1\mu _1^2}}{[(1\mu _0^2)(1\mu ^2)]^{1/2}}},`$ (113)
$`B_0^{}`$ $`=`$ $`{\displaystyle \frac{1}{2}}{\displaystyle \frac{1\mu _1^2}{[(1\mu _0^2)(1\mu ^2)(\mu _0^2\mu _1^2)(\mu ^2\mu _1^2)]^{1/2}}},`$ (114)
$`{\displaystyle \frac{\omega _0B_0^{}}{\sqrt{B_0^21}}}`$ $`=`$ $`{\displaystyle \frac{\mu _0}{\overline{J}}}{\displaystyle \frac{1}{\sqrt{\mu ^2\mu _1^2}\left(\sqrt{\mu _0^2\mu _1^2}\sqrt{\mu ^2\mu _1^2}\right)}}.`$ (115)
Next, in the range, $`0<\mu <\mu _1`$, we first put $`ϵ=0`$ in Eqs. (44) and (45), and solve, to obtain
$`\mathrm{cosh}\kappa _0=[(1\mu _1^2)/(1\mu _0^2)]^{1/2},`$ (116)
$`\mathrm{cos}\chi _0=[(1\mu _1^2)/(1\mu ^2)]^{1/2}.`$ (117)
$`\mathrm{sin}\chi _0=[(\mu _1^2\mu ^2)/(1\mu ^2)]^{1/2}`$ (118)
We then differentiate Eqs.(44) and (45), and set $`ϵ=0`$ to obtain the equations
$`\mathrm{sinh}\kappa _0\mathrm{cos}\chi _0\kappa _0^{}\mathrm{cosh}\kappa _0\mathrm{sin}\chi _0\chi _0^{}`$ $`=`$ $`0,`$ (119)
$`\mathrm{cosh}\kappa _0\mathrm{sin}\chi _0\kappa _0^{}+\mathrm{sinh}\kappa _0\mathrm{cos}\chi _0\chi _0^{}`$ $`=`$ $`1/8t_2\mathrm{sinh}\kappa _0\mathrm{sin}\chi _0.`$ (120)
Solving these, we obtain
$$\left(\begin{array}{c}\kappa _0^{}\\ \chi _0^{}\end{array}\right)=\frac{(1\mu _1^2)^{1/2}}{2(\mu _0^2\mu ^2)}\left(\begin{array}{c}(\mu _0^2\mu _1^2)^{1/2}\\ (\mu _1^2\mu ^2)^{1/2}\end{array}\right).$$
(121)
The first integral that we wish to evaluate is $`\mathrm{\Gamma }_0`$, which will give us the dominant WKB or Gamow factor in the tunnel splitting. We divide the integral into two parts by breaking the integration range at $`m_1`$. From the right-hand part, an integration by parts gives
$`\mathrm{\Gamma }_{01}`$ $`=`$ $`2\overline{J}{\displaystyle _{\mu _1}^{\mu _0}}\kappa _0𝑑\mu `$ (122)
$`=`$ $`2\overline{J}[\kappa _0(\mu )\mu |_{\mu _1}^{\mu _0}{\displaystyle _{\mu _1}^{\mu _0}}{\displaystyle \frac{\mu }{\mathrm{sinh}\kappa _0}}{\displaystyle \frac{dB_0(\mu )}{d\mu }}d\mu ],`$ (123)
while from the left-hand part we get
$$\mathrm{\Gamma }_{02}=2\overline{J}_0^{\mu _1}\kappa _0𝑑\mu =2\overline{J}\kappa _0(\mu _1)\mu _1,$$
(124)
as $`\kappa _0`$ is a constant in this range. Since $`\kappa _0(\mu _0)=0`$, $`\mathrm{\Gamma }_{02}`$ cancels the first term in Eq. (123), leaving us only with the second for $`\mathrm{\Gamma }_0`$. Using Eqs. (112) and (113), we find
$$\mathrm{\Gamma }_0=2\overline{J}(1\mu _1^2)^{1/2}_{\mu _1}^{\mu _0}\frac{d\mu }{(1\mu ^2)(\mu ^2\mu _1^2)^{1/2}}.$$
(125)
The integration is now elementary, and the result, expressed back in terms of $`\lambda `$ and $`h_x`$ is Eq. (95).
The second integral to be evaluated is $`\mathrm{\Lambda }_n`$. For the first term in Eq. (70), we integrate by parts, and use Eqs. (117) and (118):
$`2\overline{J}{\displaystyle _0^{\mu _1}}\chi _0(\mu )𝑑\mu `$ $`=`$ $`2\overline{J}{\displaystyle _0^{\mu _1}}{\displaystyle \frac{\mu }{\mathrm{sin}\chi _0}}{\displaystyle \frac{d}{d\mu }}\mathrm{cos}\chi _0d\mu `$ (126)
$`=`$ $`2\overline{J}{\displaystyle _0^{\mu _1}}{\displaystyle \frac{\mu ^2}{(1\mu ^2)(\mu _1^2\mu ^2)^{1/2}}}𝑑\mu `$ (127)
$`=`$ $`\pi \overline{J}[1(1\mu _1^2)^{1/2}].`$ (128)
For the second term in Eq. (70), we have, with $`ϵ=(n+\frac{1}{2})\omega _0`$, and Eqs. (111) and (121),
$`2ϵ\overline{J}{\displaystyle _0^{\mu _1}}\kappa _0^{}𝑑\mu `$ $`=`$ $`(2n+1)\mu _0(\mu _0^2\mu _1^2)^{1/2}{\displaystyle _0^{\mu _1}}{\displaystyle \frac{d\mu }{(\mu _0^2\mu ^2)(\mu _1^2\mu ^2)^{1/2}}}`$ (129)
$`=`$ $`\left(n+\frac{1}{2}\right)\pi .`$ (130)
Adding together the parts, and rewriting the result in terms of $`H_x`$ and $`\lambda `$, we get Eq. (96). The restriction that $`\mathrm{\Lambda }_n`$ be positive follows from the fact that we chose $`q(m)`$ to have a positive real part in Eq. (67). Thus $`\mathrm{\Lambda }`$ is necessarily positive as defined in Eq. (70). If $`H_x`$ is so large as to yield a negative value for the function of $`H_x`$ that results after doing the integral, that means that in fact there are no irregular turning points in the problem. Both terms in Eq. (67) are then equal, and the formula reduces to the expected one when there are only regular turning points.
Note that unlike Eq. (95), what appears in Eq. (96) is the ratio $`H_x/H_c`$, i.e., $`h_{x0}`$, not $`h_x`$. This fact is important for the location of the diabolical points.
The third integral we need is that of $`\kappa _0^{}`$ from $`0`$ to $`\mu _1`$. Using Eqs. (111) and (121), we get
$$2\omega _0\overline{J}_0^{\mu _1}\kappa _0^{}𝑑\mu =2\mu _0_0^{\mu _1}\frac{d\mu }{\mu _0^2\mu ^2}=\mathrm{ln}\frac{\mu _0\mu _1}{\mu _0+\mu _1}.$$
(131)
The fourth and last integral needed is $`Q_1`$. Substituting Eq. (115) in Eq. (72), we obtain
$$Q_1=_{\mu _0}^{\mu _1}\left[\frac{\mu _0}{\sqrt{\mu ^2\mu _1^2}\left(\sqrt{\mu _0^2\mu _1^2}\sqrt{\mu ^2\mu _1^2}\right)}\frac{1}{\mu _0\mu }\right]𝑑\mu .$$
(132)
The integrand is now nonsingular at $`\mu =\mu _0`$. We can make this manifest by rationalizing the difference of square roots in the first term. Some simple algebra yields
$$Q_1=_{\mu _0}^{\mu _1}\frac{1}{\sqrt{\mu ^2\mu _1^2}}\frac{\mu _0^2+\mu ^2\mu _1^2}{\mu _0\sqrt{\mu _0^2\mu _1^2}+\mu \sqrt{\mu ^2\mu _1^2}}𝑑\mu .$$
(133)
We now make the substitution $`\mu =\mu _1\mathrm{cosh}z`$, and define
$$\mathrm{cosh}z_0=\mu _0/\mu _1.$$
(134)
This yields
$`Q_1`$ $`=`$ $`{\displaystyle _0^{z_0}}{\displaystyle \frac{\mathrm{cosh}2z_0+\mathrm{cosh}2z}{\mathrm{sinh}2z_0+\mathrm{sinh}2z}}𝑑z`$ (135)
$`=`$ $`{\displaystyle _0^{z_0}}{\displaystyle \frac{\mathrm{cosh}(z+z_0)}{\mathrm{sinh}(z+z_0)}}𝑑z=\mathrm{ln}(2\mathrm{cosh}z_0)`$ (136)
$`=`$ $`\mathrm{ln}{\displaystyle \frac{2\mu _0}{\mu _1}}.`$ (137)
We now have all the ingredients needed to calculate the quantity $`F`$. Substituting Eqs. (110), (111), (131), and (137) in Eq. (71), and writing the result in terms of $`\lambda `$ and $`h_x`$, we obtain Eq. (94). Note that in writing down the final answer, we have replaced $`\overline{J}`$ by $`J`$ and $`h_x`$ by $`h_{x0}`$ in this formula. This is because $`F`$ is part of the pre-exponential factor in $`\mathrm{\Delta }_n`$, which is determined only to leading order in $`1/J`$. Keeping higher order corrections by distinguishing between $`\overline{J}`$ and $`J`$ or $`h_x`$ and $`h_{x0}`$ is not justified.
The final answer (92) for $`\mathrm{\Delta }_n`$ is obtained by substituting Eqs. (95), (96), and (94) in Eq. (68).
###### Acknowledgements.
This work is supported by the NSF via grant number DMR-9616749. I am indebted to Wolfgang Wernsdorfer and Jacques Villain for useful discussions and correspondence about Fe<sub>8</sub>.
|
warning/0003/math0003089.html
|
ar5iv
|
text
|
# The intersection of spheres in a sphere and a new geometric meaning of the Arf invariant
## 1 Introduction and Main results
Let $`S_i^3`$ be a 3-sphere embedded in the 5-sphere $`S^5`$ ($`i=1,2`$). Let $`S_1^3`$ and $`S_2^3`$ intersect transversely. Then the intersection $`C=S_1^3S_2^3`$ is a disjoint collection of circles. Then $`C`$ in $`S_i^3`$ is a 1-link ($`i=1,2`$). Note that the orientation of $`C`$ is induced by that of $`S_1^3`$ , that of $`S_2^3`$ and that of $`S^5`$. Thus we obtain a pair of 1-links, $`C`$ in $`S_i^3`$ ($`i=1,2`$), and a pair of 3-knots, $`S_i^3`$ in $`S^5`$ ($`i=1,2`$).
Conversely let $`(L_1,L_2)`$ be a pair of 1-links and $`(X_1,X_2)`$ be a pair of 3-knots. It is natural to ask whether the pair of 1-links $`(L_1,L_2)`$ is obtained as the intersection of the 3-knots $`X_1`$ and $`X_2`$ as above. We give a complete answer to this question. (Theorem 1.1.)
To state our results we need some definitions.
An (oriented) (ordered) $`m`$-component n-(dimensional) link is a smooth, oriented submanifold $`L=\{K_1,\mathrm{},K_m\}`$ of $`S^{n+2}`$, which is the ordered disjoint union of $`m`$ manifolds, each PL homeomorphic to the $`n`$-sphere. If $`m=1`$, then $`L`$ is called a knot. (See , , , . )
We say that n-links $`L_1`$ and $`L_2`$ are equivalent if there exists an orientation preserving diffeomorphism $`f:`$ $`S^{n+2}`$ $``$ $`S^{n+2}`$ such that $`f(L_1)`$=$`L_2`$ and $`f|_{L_1}:`$ $`L_1`$ $``$ $`L_2`$ is an orientation preserving diffeomorphism.
Definition $`(L_1,L_2,X_1,X_2)`$ is called a 4-tuple of links if the following conditions (1), (2) and (3) hold.
(1) $`L_i=(K_{i1},\mathrm{},K_{im_i})`$ is an oriented ordered $`m_i`$-component 1-dimensional link $`(i=1,2).`$
(2) $`m_1=m_2.`$
(3) $`X_i`$ is a 3-knot.
Definition A 4-tuple of links $`(L_1,L_2,X_1,X_2)`$ is said to be realizable if there exists a smooth transverse immersion $`f:S_1^3S_2^3S^5`$ with the following properties. We assume that the orientations of $`S_1^3,S_2^3`$ and $`S^5`$ are given.
(1) $`f|S_i^3`$ is a smooth embedding. $`f(S_i^3)`$ in $`S^5`$ is equivalent to the 3-knot $`X_i(i=1,2).`$
(2) For $`C=f(S_1^3)f(S_2^3)`$, the inverse image $`f^1(C)`$ in $`S_i^3`$ is equivalent to the 1-link $`L_i(i=1,2).`$ Here, the orientation of $`C`$ is induced naturally from the preferred orientations of $`S_1^3,S_2^3,`$ and $`S^5`$, and an arbitrary order is given to the components of $`C`$.
The following theorem characterizes the realizable 4-tuples of links.
Theorem 1.1 A 4-tuple of links $`(L_1,L_2,X_1,X_2)`$ is realizable if and only if $`(L_1,L_2,X_1,X_2)`$ satisfies one of the following conditions (1) and (2).
(1) Both $`L_1`$ and $`L_2`$ are proper links, and
$`\mathrm{Arf}(L_1)=\mathrm{Arf}(L_2).`$
(2) Neither $`L_1`$ nor $`L_2`$ is a proper link, and
$`\mathrm{lk}(K_{1j},L_1K_{1j})\mathrm{lk}(K_{2j},L_2K_{2j})\mathrm{mod2}`$ for all $`j`$.
In the case where $`L_i`$ is a 1-component link, that is, $`L_i`$ is a knot $`K_i`$, we have the following corollary.
Corollary 1.2 For 1-knots $`K_1`$ and $`K_2`$, a 4-tuple of links $`(K_1,K_2,X_1,X_2)`$ is realizable if and only if
$`\mathrm{Arf}(K_1)=\mathrm{Arf}(K_2).`$
Note. Theorem 1.1 and Corollary 1.2 give a new geometric meaning of the Arf invariant.
Note 1.2.1. The problem (26) in says: Investigate ordinary sense slice 1-links, where ordinary sense slice 1-links are 1-links which are obtained as follows: Let $`S^2`$ be in $`R^4=R^3\times R.`$ Then $`S^2[R^3\times \{0\}]`$ in $`R^3\times \{0\}`$ is a 1-link. By using Theorem 1.1, the author gives an answer to this problem. The answer is: for every ordinary sense slice 1-link we can define the Arf invariant and it is zero (see ).
Let $`f:S^3S^5`$ be a smooth transverse immersion. Then the self-intersection $`C`$ consists of double points. Suppose that $`C`$ is a single circle in $`S^5`$. Then the $`f^1(C)`$ in $`S^3`$ is a 1-knot or a 2-component 1-link. There is a similar realization problem. We consider which 1-knots (resp. 2-component 1-links ) we obtain as above. We give complete answers.
Theorem 1.3 Let $`f:S^3S^5`$ be a smooth transverse immersion. Then the self-intersection $`C`$ consists of double points. Suppose that $`C`$ is a single circle in $`S^5`$.
(1) Any 2-component 1-link is realizable as $`f^1(C)`$ in $`S^3`$ for an immersion $`f`$.
(2) Any 1-knot is realizable as $`f^1(C)`$ in $`S^3`$ for an immersion $`f`$.
Remark. Suppose that $`K_1`$ is the trivial 1-knot, $`K_2`$ is the trefoil 1-knot, and $`X_1`$ and $`X_2`$ are 3-knots. Suppose that $`L=(L_1,L_2)`$ is a split 1-link such that $`L_1`$ is the trivial 1-knot and $`L_2`$ is the trefoil knot. Then, by Corollary 1.2, a 4-tuple of links $`(K_1,K_2,X_1,X_2)`$ is not realizable. But, by Theorem 1.3 (1), the two component split link $`L`$ is realizable as in Theorem 1.3 (1).
In ,,, the author discussed some topics which are related to this paper. In he discussed the intersection of three 4-spheres in a 6-sphere. In he discussed the intersection of two $`(n+2)`$-spheres in an $`(n+4)`$-sphere. In he discussed the following: Let $`L=(K_1,K_2)`$ be a 2-link in $`S^4=B^5`$. Take a slice disc $`D_i^3`$ in $`B^5`$ for each component $`K_i`$. He discussed the intersection of two slice discs $`D_1^3`$ and $`D_2^3`$ in the 5-ball $`B^5`$. In he applied Theorem 1.1 to Fox’s problem as we state in Note 1.2.1.
Problem 1.4. Suppose each of $`S_1^3,S_2^3`$, and $`S_3^3`$ is a 3-sphere embedded in $`S^5`$. Suppose $`S_i^3`$ and $`S_j^3`$ intersect transversely. Suppose each of $`S_1^3S_2^3`$, $`S_2^3S_3^3`$, and $`S_3^3S_1^3`$ is a single circle. Then we have a triple of 1-links, $`L_i`$=($`S_i^3S_j^3`$, $`S_i^3S_k^3`$) in $`S_i^3`$, where $`(i,j,k)=(1,2,3),(2,3,1),(3,1,2)`$.
Which triple of 1-links do we obtain like this?
Do we characterize such triple by the Arf invariants, the linking numbers, and the Saito-Sato-Levine invariants? (See for the definition of the Saito-Sato-Levine invariant.)
In the author discussed a higher dimensional version of Problem 1.4.
This paper is organized as follows. In §2 we review spin cobordism and the Arf invariant. In §3 we discuss a necessary condition for the realization of 4-tuple of links. We find the obstruction for the realization in the spin cobordism group $`\mathrm{\Omega }_{}^{\mathrm{spin}}.`$ In §4 we discuss a sufficient condition for the realization of 4-tuple of links. We carry out surgeries of submanifolds to carry out an (un)knotting operation. Theorem 1.1 is deduced from §3 and §4. In §5 we prove Theorem 1.3. In §6 we give a problem.
## 2 Spin cobordism and the Arf invariant
In this section we review some results on the Arf invariant and spin cobordism. See and for the Arf invariant. See for spin structures and spin cobordism.
We suppose that, when we say $`M`$ is a spin manifold, $`M`$ is oriented.
Recall that a proper link is an $`m`$-component 1-link $`L=\{K_1,\mathrm{},K_m\}`$ such that $`\mathrm{lk}(K_j,LK_j)`$ $`=_{1im,ij}`$ $`\mathrm{lk}(K_j,K_i)`$ is an even number for each $`K_j`$.
Let $`L=(K_1,\mathrm{},K_m)`$ be an $`m`$-component 1-link. Let $`F`$ be a Seifert surface for $`L`$. We induce a spin structure $`\sigma `$ on $`F`$ from the unique one on $`S^3`$. We induce a spin structure $`\sigma _i`$ on $`K_i`$ from $`\sigma `$ on $`F`$. Then we have:
Proposition 2.1 Under the above condition, for each $`i`$,
mod 2 $`\mathrm{lk}`$ $`(K_i,LK_i)=[(K_i,\sigma _i)]\mathrm{\Omega }_1^{\mathrm{spin}}`$.
In particular, $`L`$ is a proper link if and only if each $`[(K_i,\sigma _i)]=0`$.
Suppose that $`L`$ is a proper link. Take $`(F,\sigma )`$ as above. Let $`\widehat{F}`$ be the closed surface obtained from $`F`$ by attaching disks to the boundaries. Let $`\widehat{\sigma }`$ be the unique extension of $`\sigma `$ over $`\widehat{F}`$. Then we have:
Proposition 2.2 Under the above condition, $`\mathrm{Arf}`$ $`(L)`$= $`[(\widehat{F},\widehat{\sigma })]`$ $`\mathrm{\Omega }_2^{\mathrm{spin}}`$.
Although they may be folklore, the author gives a proof of Proposition 2.1 and that of Proposition 2.2 in the appendix.
## 3 A necessary condition for the realization of 4-tuple of links
In this section we discuss a necessary condition for the realization of a 4-tuple of links. That is, we prove the following two propositions.
Proposition 3.1 If $`(L_1,L_2,X_1,X_2)`$ is realizable then
$`\mathrm{lk}(K_{1j},L_1K_{1j})\mathrm{lk}(K_{2j},L_2K_{2j})\mathrm{mod2}`$ for all $`j`$.
In particular, $`L_1`$ is a proper link if and only if $`L_2`$ is a proper link.
Proposition 3.2 Let $`L_1`$ and $`L_2`$ be proper links. If $`(L_1,L_2,X_1,X_2)`$ is realizable then
$$\mathrm{Arf}(L_1)=\mathrm{Arf}(L_2).$$
In order to prove them, we prepare a lemma.
Let $`M_1`$ and $`M_2`$ be codimension one submanifolds of an $`n`$-dimensional compact spin manifold $`N`$. Suppose that $`M_1`$ and $`M_2`$ are compact oriented manifolds. Suppose that $`M_1`$, $`M_2`$, and $`N`$ may have the boundary and the corner. Let $`M_1`$ and $`M_2`$ intersect transversely. Suppose $`M_i`$ may be embedded in the boundary (resp. the corner ) of $`N`$.
We induce a spin structure $`\sigma _i`$ on $`M_i`$ from $`N`$. We induce a spin structure $`\xi _i`$ on $`M_1M_2`$ from $`\sigma _i`$ on $`M_i`$. ($`i=1,2`$). Then it is easy to prove:
Lemma $`\xi _1`$ and $`\xi _2`$ are same.
The spin structure $`\xi _1`$=$`\xi _2`$ on $`M_1M_2`$ is called the unique spin structure induced by $`M_1`$, $`M_2`$ and $`N`$.
Proof of Proposition 3.1. Let $`f:S_1^3S_2^3S^5`$ be an immersion to realize the 4-tuple of links $`(L_1,L_2,X_1,X_2)`$. Let $`C=C_i`$ denote $`f(S_1^3)f(S_2^3)`$.
We abbreviate $`f(S_i^3)`$ to $`S_i^3.`$ Let $`V_i`$ be a Seifert hypersurface for $`X_i`$, i.e.,
$$S_i^3=V_iV_iS^5.$$
We make $`V_1`$ and $`V_2`$ intersect transversely. We induce a spin structure $`v_i`$ on $`V_i`$ from the unique one on $`S^5`$.
Put $`W=V_1V_2`$. We have:
We give $`W`$ the unique spin structure $`w`$ induced by $`V_1`$, $`V_2`$ and $`W`$.
Here, we see that $`W=(V_1V_2)(V_1V_2).`$ Put $`F_1`$=$`V_1V_2`$. Put $`F_2`$=$`V_1V_2`$. Then $`F_1`$ ( resp. $`F_2`$ ) is a Seifert surface for $`L_1`$ ( resp. $`L_2`$ ).
We induce a spin structure on $`V_i=S_i^3`$ from $`v_i`$ on $`V_i`$. Note that it is the unique one on $`S_i^3`$.
We have:
We give $`F_1`$ the unique spin structure $`\sigma _1`$ induced by $`S_1^3`$, $`W`$ and $`V_1`$.
We have:
We give $`F_2`$ the unique spin structure $`\sigma _2`$ induced by $`S_2^3`$, $`W`$ and $`V_2`$.
We have:
We give $`C_j`$ the unique spin structure $`\tau _j`$ induced by $`F_1`$, $`F_2`$ and $`W`$.
Then mod2 lk$`(K_{1j},L_1K_{1j})`$=mod2 lk$`(K_{2j},L_2K_{2j})`$ =$`[(C_j,\tau _j)]`$ $`\mathrm{\Omega }_1^{\mathrm{spin}}`$ for all $`j`$. This completes the proof of Proposition 3.1.
We confirm that we have:
Proof of Proposition 3.2. We can make the corner of $`W`$ smooth. Note $`(W,w)`$ = $`(F_1,\sigma _1)(F_2,\sigma _2`$. Take $`\widehat{F_i}`$ and $`\widehat{\sigma _i}`$ as above. Then we have:
$`\mathrm{Arf}(L_1)+\mathrm{Arf}(L_2)=\mathrm{Arf}(L_1)+\mathrm{Arf}(L_2)=[(\widehat{F_1},\widehat{\sigma _1})]+[(\widehat{F_2},\widehat{\sigma _2})]=[(W,w)]=0\mathrm{\Omega }_2^{\mathrm{spin}}Z_2.`$
Therefore $`\mathrm{Arf}(L_1)=\mathrm{Arf}(L_2).`$
## 4 A sufficient condition for the realization of 4-tuple of links
In this section we discuss a sufficient condition for the realization of a 4-tuple of links. That is, we prove the following proposition.
Proposition 4.1 A 4-tuple of links $`(L_1,L_2,X_1,X_2)`$ is realizable if $`(L_1,L_2,X_1,X_2)`$ satisfies one of the conditions (1) and (2) of Theorem 1.1.
It is easy to prove that Proposition 4.1 is equivalent to the following Proposition 4.2.
Proposition 4.2 Let $`X_1`$ and $`X_2`$ be the trivial 3-knots. Let $`L_1`$ and $`L_2`$ be 1-links. Suppose that $`L_1`$ and $`L_2`$ satisfies one of the conditions (1) and (2) of Theorem 1.1. Then the 4-tuple of links $`(L_1,L_2,X_1,X_2)`$ is realizable.
We prove
Lemma 4.3 Let $`X_1`$ and $`X_2`$ be the trivial 3-knots. Let $`L`$ be a 1-link. Then the 4-tuple of links $`(L,L,X_1,X_2)`$ is realizable.
Proof of Lemma 4.3. Let $`f:S_1^3S_2^3S^5`$ be an embedding such that $`f(S_1^3S_2^3)`$ in $`S^5`$ is equivalent to the trivial 3-link. We take a chart $`(U,\varphi )`$ of $`S^5`$ with the following properties (1) and (2).
(1) $`\varphi :UR^5=\{(x,y,z,u,v)|x,y,z,u,vR\}`$ =$`R^3\times R_u\times R_v`$.
(2) $`Uf(S_1^3)=\{(x,y,z,u,v)|u=0,v=0\}=R_1^3`$
$`Uf(S_2^3)=\{(x,y,z,u,v)|u=1,v=0\}=R_2^3`$
$$\mathrm{Figure1}$$
Obviously Lemma 4.3 follows from Lemma 4.4.
Lemma 4.4 There exists an immersion $`g:S_1^3S_2^3S^5`$ with the following conditions.
(1) $`g|_{S_2^3}=f|_{S_2^3}`$.
(2) $`g|_{S_1^3}`$ is isotopic to $`f|_{S_1^3}`$.
(3) $`g|_{S_1^3g^1(U)}`$ = $`f|_{S_1^3g^1(U)}`$. Hence $`g(S_1^3)g(S_2^3)`$ $`U`$.
(4) $`g(S_1^3)g(S_2^3)`$ in $`g(S_1^3)`$ and that in $`g(S_2^3)`$ are both equivalent to the 1-link $`L`$.
(5) $`g(S_1^3)\{(x,y,z,u,v)|u=0,vR\}`$= $`g(S_1^3)\{(x,y,z,u,v)|u=0,v=0\}`$.
Proof of Lemma 4.4. We modify the embedding $`f`$ to construct an immersion $`g`$.
In $`R_1^3`$ we take the 1-link $`L`$ and a Seifert surface $`F`$ for $`L`$. Let $`N(F)`$
=$`F\times \{t|1t1\}`$ be a tubular neighborhood of $`F`$ in $`R_1^3`$.
We define a subset $`E`$ of $`N(F)\times R_u\times R_v`$ $`=\{(p,t,u,v)|pF,1t1,uR,vR\}`$ so that
$`E=\{(p,t,u,v)|pF,`$ $`0u\frac{\pi }{2},t=k\mathrm{cos}u,v=k\mathrm{sin}u,`$ $`1k1\}`$.
Put $`P`$=$`\overline{E(Ef(S_1^3))}`$. Put $`Q`$=$`\overline{f(S_1^3)(Ef(S_1^3))}`$. Note that $`P`$=$`Q`$=$`N(F)`$. Put $`\mathrm{\Sigma }=`$$`PQ`$. Then, by the construction, $`\mathrm{\Sigma }`$ is a 3-sphere embedded in $`S^5`$ and is the trivial 3-knot.
Note. In $`U`$ the following hold.
(1) $`g(S_1^3)\{(x,y,z,u,v)|u=0,vR\}`$ $`=g(S_1^3)\{(x,y,z,u,v)|u=0,v=0\}`$ $`=\overline{g(S_1^3)N(F)}`$
(2) $`g(S_1^3)\{(x,y,z,u,v)|0<u\frac{\pi }{2},vR\}`$ is Int $`N(F)`$.
(3) Let $`0<u^{}<\frac{\pi }{2}`$.
$`g(S_1^3)\{(x,y,z,u,v)|u=u^{},v>0\}`$ is diffeomorphic to Int $`F`$.
(4) Let $`0<u^{}<\frac{\pi }{2}`$.
$`g(S_1^3)\{(x,y,z,u,v)|u=u^{},v<0\}`$ is diffeomorphic to Int $`F`$.
(5) Let $`0<u^{}<\frac{\pi }{2}`$.
$`g(S_1^3)\{(x,y,z,u,v)|u=u^{},v=0\}`$ is diffeomorphic to $`L`$.
(6) Let $`0<u^{}<\frac{\pi }{2}`$.
$`g(S_1^3)\{(x,y,z,u,v)|u=u^{},vR\}`$ is diffeomorphic to $`(N(F))`$.
Let $`F_i`$ be diffeomorphic to $`F`$ ($`i=1,2`$). Recall $`F`$ is a compact oriented surface with boundary. We identify $`F_1`$ with $`F_2`$ to obtain $`F_0=F_1F_2`$. Note $`F_0`$ is diffeomorphic to $`(N(F))`$.
(7) $`g(S_1^3)\{(x,y,z,u,v)|u=\frac{\pi }{2},vR\}`$ is diffeomorphic to $`N(F)`$.
(8) $`g(S_1^3)\{(x,y,z,u,v)|u=\frac{\pi }{2},v=0\}`$ is diffeomorphic to $`F`$.
(9) $`g(S_1^3)\{(x,y,z,u,v)|0<u\frac{\pi }{2},v=0\}`$ is diffeomorphic to Int $`F`$.
In Figure 2, 3, 4, $`\mathrm{\Sigma }U`$ and $`g(S_2^3)U`$ are drawn. There, we replace $`R^3\times R_u\times R_v`$ with $`R^2\times R_u\times R_v`$.
$$\mathrm{Figure2}$$
$$\mathrm{Figure3}$$
$$\mathrm{Figure4}$$
By the construction, $`\mathrm{\Sigma }`$ $`g(S_2^3)`$ in $`\mathrm{\Sigma }`$ and that in $`g(S_2^3)`$ are both equivalent to $`L`$. Define $`g|_{S_1^3}`$ so that $`g(S_1^3)`$=$`\mathrm{\Sigma }`$.
This completes the proof of Lemma 4.4 and therefore Lemma 4.3.
Note. Lemma 4.3 gives an alternative proof of the results of and Lemma 1 of .
In order to prove Proposition 4.2, we review the pass-moves. See and for detail.
Definition (See .) Two 1-links are pass-move equivalent if one is obtained from the other by a sequence of pass-moves. See Figure 5 for an illustrations of the pass-move.
$$\mathrm{Figure5}$$
In $`B_a^3`$ there are four arcs. Each of four arcs may belong to different components of the 1-link. We do not assume the four arcs belong to one component of the 1-link.
The following propositions are essentially proved in . A proof is written in the appendix of .
Proposition 4.5 (See .) Let $`L_1`$ and $`L_2`$ be 1-links. Then $`L_1`$ and $`L_2`$ are pass-move equivalent if and only if $`L_1`$ and $`L_2`$ satisfy one of the conditions (1) and (2) of Theorem 1.1.
Proposition 4.6 (See .) Let $`L_1`$ and $`L_2`$ be 1-links. Suppose that $`L_1`$ is pass-move equivalent to $`L_2`$. Let $`F`$ be an oriented Seifert surface for $`L_2`$ such that the genus of $`F`$ is not less than the genus of $`L_1`$. Then there exists a disjoint union of 3-balls $`B^3`$ such that $`B^3F`$ is as in Figure 6 and the pass-moves in all $`B^3`$ change $`L_2`$ to $`L_1`$
$$\mathrm{Figure6}$$
Let $`L_1`$ and $`L_2`$ be 1-links. Suppose $`L_1`$ and $`L_2`$ satisfy one of the conditions (1) and (2) of Theorem 1.1. Then by Proposition 4.5 $`L_1`$ is obtained from $`L_2`$ by a sequence of pass-moves. We choose a Seifert surface $`F`$ for $`L_2`$ and the disjoint union of 3-balls $`B^3`$ in $`S^3`$ as in Proposition 4.6.
Take a 1-link ($`Y_1`$, $`Y_2`$) in each $`B^3`$ as in Figure 7. By considering the Kirby moves of framed links (in ), it is easy to prove:
$$\mathrm{Figure7}$$
Lemma 4.7 We carry out 0-framed surgeries along all $`Y_1`$ and all $`Y_2`$. After these surgeries, we have the following. (1) $`S^3`$ becomes a 3-sphere again. (2) each $`B^3`$ becomes a 3-ball again, (3) $`L_2`$ in the old sphere $`S^3`$ changes to $`L_1`$ in the new 3-sphere.
We go back to the proof of Proposition 4.2.
As in Lemma 4.4, take an immersion $`g:S_1^3S_2^3S^5`$ to realize $`(L_2,L_2,X_1,X_2)`$, where $`X_i`$ is the trivial 3-knot.
Obviously Proposition 4.2 follows from Lemma 4.8.
Lemma 4.8 There exists an immersion $`h:S_1^3S_2^3S^5`$ with the following conditions.
(1) $`h|_{S_2^3}=g|_{S_2^3}`$.
(2) $`h(S_1^3)`$ in $`S^5`$ is equivalent to the trivial 3-knot.
(3) $`h|_{S_1^3g^1(U)}`$ = $`g|_{S_1^3g^1(U)}`$. Hence $`h(S_1^3)h(S_2^3)`$ $`U`$.
(4) $`h(S_1^3)h(S_2^3)`$ in $`h(S_1^3)`$ is equivalent to $`L_1`$. $`h(S_1^3)h(S_2^3)`$ in $`h(S_2^3)`$ is equivalent to $`L_2`$.
(5) $`g(S_1^3)\{u0,vR\}`$=$`h(S_1^3)\{u0,vR\}`$
Proof of Lemma 4.8. We modify the immersion $`g`$ to define an immersion $`h`$ as follows.
As in the proof of Lemma 4.4, we take $`L_2`$ and a Seifert surface $`F`$ for $`L_2`$ in $`g(S_1^3)`$. Suppose $`L_2R_1^3`$ and $`FU`$. Suppose the genus of $`F`$ satisfies the condition in Proposition 4.6. Take 3-balls $`B^3`$ in $`g(S_1^3)`$ as in Proposition 4.6. Take ($`Y_1`$, $`Y_2`$) in $`B^3`$ as in Lemma 4.7. See Int $`N(F)`$ in $`g(S_1^3)`$. Int $`N(F)`$ is not in $`R_1^3`$ and is in $`\{u>0\}`$. But by Proposition 4.6, we can suppose $`Y_1`$ and $`Y_2`$ are in $`R_1^3`$.
Take an embedded 2-disc $`h_1^{}_{}{}^{}2`$ in $`\{(x,y,z,u,v)|u=0,v0\}`$ so that $`h_1^{}_{}{}^{}2`$ meets $`\{(x,y,z,u,v)|u=0,v=0\}`$ at $`Y_1`$ transversely. Suppose that $`h_1^{}_{}{}^{}2`$ is embedded trivially.
Take an embedded 2-disc $`h_2^{}_{}{}^{}2`$ in $`\{(x,y,z,u,v)|u=0,v0\}`$ so that $`h_2^{}_{}{}^{}2`$ meets $`\{(x,y,z,u,v)|u=0,v=0\}`$ at $`Y_2`$ transversely. Suppose that $`h_2^{}_{}{}^{}2`$ is embedded trivially.
See Figure 8.
$$\mathrm{Figure8}$$
Let $`h_1^2`$ be a tubular neighborhood of $`h_1^{}_{}{}^{}2`$ in $`\{(x,y,z,u,v)|u=0,v0\}`$.
Let $`h_2^2`$ be a tubular neighborhood of $`h_2^{}_{}{}^{}2`$ in $`\{(x,y,z,u,v)|u=0,v0\}`$.
Then $`h_1^2`$ and $`h_2^2`$ are diffeomorphic to the 4-ball. Furthermore we can regard $`h_i^2`$ as a 4-dimensional 2-handle attached to $`g(S_1^3)`$ along $`Y_i`$ with 0-framing.
Put $`R=g(S_1^3)`$ $`\{g(S_1^3)h_1^2\}`$ $`\{g(S_1^3)h_2^2\}`$. Put $`S_1=`$ $`h_1^2`$ $`\{g(S_1^3)h_1^2\}`$. Put $`S_2=`$ $`h_2^2`$ $`\{g(S_1^3)h_2^2\}`$. Here, we consider all $`h_i^2`$. Put $`\mathrm{\Lambda }`$=$`RS_1S_2`$.
Then we can regard $`\mathrm{\Lambda }`$ as the result of 0-framed surgeries on $`g(S_1^3)`$ along all $`Y_1`$ and $`Y_2`$. Then the pass-moves are carried out in all $`B^3`$.
Therefore $`\mathrm{\Lambda }`$ is an embedded 3-sphere in $`S^5`$. Furthermore $`\mathrm{\Lambda }g(S_2^3)`$ in $`\mathrm{\Lambda }`$ is equal to the 1-link $`L_1`$ and $`\mathrm{\Lambda }g(S_2^3)`$ in $`g(S_2^3)`$ is equal to the 1-link $`L_2`$.
Put $`h|_{S_1^3}`$ so that $`h(S_1^3)`$=$`\mathrm{\Lambda }`$. By the above construction, $`h`$ satisfies the conditions (1)(3)(4)(5) in Lemma 4.8.
We prove:
Lemma 4.9 $`h|(S_1^3)`$ is equal to the trivial 3-knot.
Proof. By the construction, $`h(S_1^3)`$ bounds a 4-manifold represented by the disjoint union of some copies of the framed link in Figure 9.
$$\mathrm{Figure9}$$
See for dot-circles. Hence this 4-manifold is diffeomorphic to the 4-ball. Therefore $`h(S_1^3)`$ bounds a 4-ball. This completes the proof of Lemma 4.9 and therefore Lemma 4.8. This completes the proof of Proposition 4.2 and therefore Proposition 4.1.
## 5 The proof of Theorem 1.3
Lemma 5.1.1 Let $`L`$ be the trivial 2-component 1-link. There exists a self-transverse immersion $`f:S^3B^5`$ with the following properties.
(1) The singular point set (in $`B^5`$) is a single circle $`C`$.
(2) $`f^1(C)`$ in $`S^3`$ is equivalent to $`L`$.
(3) $`f(\overline{S^3N(L)})`$ $`B^5`$ and $`f(\mathrm{Int}N(L))`$ $``$ $`\mathrm{Int}B^5`$, where $`N(L)`$ is a tubular neighborhood of $`L`$ in $`S^3`$.
Lemma 5.1.2 Let $`L`$ be the Hopf link. There exists a self-transverse immersion $`f:S^3B^5`$ satisfying with the properties (1), (2) and (3) in Lemma 5.1.1.
Lemma 5.1.3 Let $`L`$ be the trivial 1-knot. There exists a self-transverse immersion $`f:S^3B^5`$ satisfying with the properties (1), (2) and (3) in Lemma 5.1.1.
Proof of Lemma 5.1.1 Take a chart of $`(U,\varphi )`$ of $`B^5`$ such that $`\varphi (U)`$= $`\{(x,y,z,v,t)|`$ $`x,y,z,vR,t0\}.`$
Put $`F`$= $`\{(x,y,z,v,t)|`$ $`x,yR,z0,v=0,t0\}.`$
Put $`A`$= $`\{(x,y,z,v,t)|`$ $`x,yR,z=0,v=0,t0\}.`$
We can regard $`U`$ as the result of rotating $`F`$ around the axis $`A`$.
Take an immersed 2-disc $`D`$ in $`F`$ as in Figure 10, 11.
$$\mathrm{Figure10}$$
$$\mathrm{Figure11}$$
Note. In $`F`$ the following hold.
(1) $`\{(x,y,z,v,t)|`$ $`x,yR,z0,v=0,t=0\}D`$ is diffeomorphic to $`\overline{D(two2discs)}`$.
(2) $`\{(x,y,z,v,t)|`$ $`x,yR,z0,v=0,t<0\}D`$ is a union of the interior of two 2-discs, where the intersection of the two 2-discs is one point. The point is $`p=(1,0,0,0,1)`$.
(3) Let $`1<t^{}<0`$.
$`\{(x,y,z,v,t)|`$ $`x,yR,z0,v=0,t=t^{}\}D`$
is the Hopf link in
$`\{(x,y,z,v,t)|`$ $`x,yR,z0,v=0,t=t^{}\}`$.
(4) $`\{(x,y,z,v,t)|`$ $`x,yR,z0,v=0,t=1\}D`$ is a union of two circles, where the intersection of the two circles is one point. The point is $`p=(1,0,0,0,1)`$.
(5) Let $`2<t^{}<1`$.
$`\{(x,y,z,v,t)|`$ $`x,yR,z0,v=0,t=t^{}\}D`$
is the trivial link in
$`\{(x,y,z,v,t)|`$ $`x,yR,z0,v=0,t=t^{}\}`$.
(6) $`\{(x,y,z,v,t)|`$ $`x,yR,z0,v=0,t=2\}D`$
is a disjoint union of two 2-discs.
As we rotate $`F`$ as above, we rotate $`D`$ as well. We obtain an immersed 3-sphere $`X`$.
Take $`f`$ so that $`f(S^3)`$=$`X`$.
This completes the proof of Lemma 5.1.1.
Proof of Lemma 5.1.2 Take $`B^5,F,A,`$ and $`D`$ as above.
Put $`G`$= $`\{(x,y,z,v,t)|`$ $`x=0,y=0,z0,v=0,t0\}.`$
In Figure 10, 11 we suppose: If, in $`F`$, we rotate $`D`$ around $`G`$ by any angle, then $`DA`$ does not change.
As we rotate $`F`$ as above, we rotate $`D`$ around $`A`$ so that in $`F`$ we rotate $`D`$ around $`G`$ one time. We obtain an immersed 3-sphere $`X`$.
Take $`f`$ so that $`f(S^3)`$=$`X`$.
This completes the proof of Lemma 5.1.2.
Proof of Lemma 5.1.3 Take $`B^5,F,A,D,`$ and $`G`$ as above.
In Figure 10, 11 we suppose: If we rotate $`D`$ around $`G`$ half time, then the resultant $`D`$ coincides with the original $`D`$.
As we rotate $`F`$ as above, we rotate $`D`$ around $`A`$ so that in $`F`$ we rotate $`D`$ around $`G`$ half time. We obtain an immersed 3-sphere $`X`$.
Take $`f`$ so that $`f(S^3)`$=$`X`$.
This completes the proof of the proof of Lemma 5.1.3.
Lemma 5.2.1 Let $`L`$ be a 2-component 1-link whose linking number is even. There exists a self-transverse immersion $`g:S^3\mathrm{}^lS^2\times B^3`$, for a non-negative integer $`l`$, with the following properties.
(1) The singular point set (in $`\mathrm{}^lS^2\times B^3`$) is a single circle $`C`$.
(2) $`g^1(C)`$ in $`S^3`$ is equivalent to $`L`$.
(3) $`g(\overline{S^3N(L)})`$ $`(\mathrm{}^lS^2\times B^3)=`$ $`\mathrm{}^lS^2\times S^2`$ and $`g(\mathrm{Int}N(L))`$ $``$ $`\mathrm{Int}(\mathrm{}^lS^2\times B^3)`$, where $`N(L)`$ is a tubular neighborhood of $`L`$ in $`S^3`$.
Here, we suppose that $`\mathrm{}^0S^2\times B^3`$ means $`B^5`$.
Lemma 5.2.2 Let $`L`$ be a 2-component 1-link whose linking number is odd. There exists a self-transverse immersion $`g:S^3\mathrm{}^lS^2\times B^3`$, for a non-negative integer $`l`$, satisfying with the properties (1), (2) and (3) in Lemma 5.2.1.
Lemma 5.2.3 Let $`L`$ be a 1-knot. There exists a self-transverse immersion
$`g:S^3\mathrm{}^lS^2\times B^3`$, for a non-negative integer $`l`$, satisfying with the properties (1), (2) and (3) in Lemma 5.2.1.
We prove:
Claim 5.2.4 Lemma 5.2.1, 5.2.2, and 5.2.3 imply Theorem 1.3.
There is an embedding $`h:\mathrm{}^lS^2\times B^3S^5`$. Then $`hg`$ is an immersion in Theorem 1.3. This completes the proof.
In order to prove Lemma 5.2.1, 5.2.2, and 5.2.3, we review the $`\mathrm{}`$-moves. See and for detail.
Definition () Two 1-links are $`\mathrm{}`$-move equivalent if one is obtained from the other by a sequence of $`\mathrm{}`$-moves. See Figure 12 for an illustrations of the $`\mathrm{}`$-move. The $`\mathrm{}`$-move is different from the pass-move by the orientation.
$$\mathrm{Figure12}$$
In each 3-ball in Figure 12 there are four arcs. Each of four arcs may belong to different components of the 1-link. We do not assume the four arcs belong to one component of the 1-link.
The following proposition 5.3 is proved in the appendix of .
Proposition 5.3 ()(1) Let $`L`$ be a 2-component link. Then $`L`$ is $`\mathrm{}`$-move equivalent to the trivial link if and only if the linking number is even.
(2) Let $`L`$ be a 2-component link. Then $`L`$ is $`\mathrm{}`$-move equivalent to the Hopf link if and only if the linking number is odd.
(3) Any 1-knot is $`\mathrm{}`$-move equivalent to the trivial knot.
We have:
Proposition 5.4 ()Let $`L_1`$ and $`L_2`$ be 1-links. Suppose that $`L_1`$ is $`\mathrm{}`$-move equivalent to $`L_2`$. Take $`L_1`$ in $`S^3`$. Then there exists a disjoint union of 3-balls $`B^3`$ in $`S^3`$ such that $`L_1B^3`$ is as in Figure 12 and that the $`\mathrm{}`$-moves in all $`B^3`$ change $`L_1`$ to $`L_2`$.
Let $`L_1`$ and $`L_2`$ be 1-links. Suppose $`L_1`$ is $`\mathrm{}`$-move equivalent to $`L_2`$. Take $`L_1`$ in $`S_1^3`$. Take a disjoint union of 3-balls $`B^3`$ in $`S^3`$ as in Proposition 5.3. Take a 1-link ($`Y_1`$, $`Y_2`$) in each $`B^3`$ as in Figure 13. Suppose that $`Y_i`$ bounds a 2-disc $`B_i^2`$ as in Figure 13.
$$\mathrm{Figure13}$$
By considering the Kirby moves of framed links (in ), it is easy to prove:
Lemma 5.5 We carry out 0-framed surgeries on all $`Y_1`$ and all $`Y_2`$. After these surgeries, we have the following. (1) each $`B^3`$ becomes a 3-ball again, (2) $`S^3`$ becomes a 3-sphere again. (3) $`L_1`$ in the old sphere $`S^3`$ changes to $`L_2`$ in the new 3-sphere.
We go back to the proof of Lemma 5.2.1, 5.2.2, 5.2.3.
Proof of Lemma 5.2.1 ( resp. 5.2.2, 5.2.3 )
By Proposition 5.3, a sequence of $`\mathrm{}`$-moves changes $`L`$ into the trivial 2-component 1-link ( resp. the Hopf link, the trivial 1-knot ).
As in Lemma 5.1.1 ( resp. 5.1.2, 5.1.3 ), take $`f:S^3B^5`$. See the 1-link $`f^1(C)`$ in $`S^3`$. It is equivalent to the 1-link $`L`$. As in Proposition 5.4, take 3-balls $`B^3`$. As in Lemma 5.5, take $`Y_1`$ and $`Y_2`$ in $`B^3`$. Suppose that $`f(Y_1)`$ and $`f(Y_2)`$ are in $`B^5`$.
Attach 4-dimensional 2-handles $`h^2`$ to $`f(S^3N(L))`$ in $`B^5`$ along $`f(Y_i)`$ with 0-framing. Here, 0-framing means the following. When attaching the 4-dimensional 2-handles $`h^2`$ to $`f(S^3N(L))`$, we can attach 4-dimensinal 2-handles to $`S^3`$ naturally. These attaching maps are 0-framing.
When attaching the 4-dimensional 2-handles $`h^2`$ to $`f(S^3N(L))`$, we can attach 5-dimensional 2-handles $`h^2\times [1,1]`$ to $`B^5`$ along $`f(Y_i)`$ naturally. Of course the attached parts are in $`B^5`$. We obtained a 5-manifold. Call it $`M`$.
Put $`P=`$ $`f(S^3)(f(S^3)h^2)`$. Put $`Q=`$ $`h^2(f(S^3)h^2)`$. Put $`\mathrm{\Sigma }=PQ`$. Then $`\mathrm{\Sigma }`$ is an immersed 3-sphere in $`M`$.
We prove Lemma 5.6. Before the proof of Lemma 5.6, we prove Lemma 5.7.
Lemma 5.6 Let $`M`$ be as above. $`M`$=$`\mathrm{}^lS^2\times B^3`$, for a non-negative integer $`l`$.
Lemma 5.7 Lemma 5.6 implies Lemma 5.2.1, 5.2.2 and 5.2.3.
Proof of Lemma 5.7. Take a self-transverse immersion $`g:S^3M=\mathrm{}^lS^2\times B^3`$ so that $`g(S^3)=\mathrm{\Sigma }`$.
Proof of Lemma 5.6. Recall: $`J=`$ (one 5-dimensional 0-handle)$``$(one 5-dimensional 2-handle) is $`S^2\times B^3`$ or $`S^2\stackrel{~}{\times }B^3`$. $`J`$ is $`S^2\times B^3`$ if and only if the attaching map of the 2-handle is spin preserving diffeomorphism map.
It suffices to prove that the attaching maps of the 5-dimensional 2-handles are spin-preserving diffeomorphism maps.
We give a spin structure on $`f(S^3N(L))`$ from the unique one on $`B^5`$.
We give a spin structure on $`f(B_i^2)B^5`$ from the spin structure on $`f(S^3N(L))`$.
We give a spin structure $`\xi `$ on $`f(Y_i)`$ from the spin structure on $`f(B_i^2)B^5`$.
Put $`S_a^1S_b^1`$= $`f(B_i^2)N(L)`$. We give a spin structure $`\alpha `$ (resp. $`\beta `$) on $`S_a^1`$ (resp. $`S_b^1`$ ) from the spin structure on $`f(B_i^2)B^5`$. By the construction, $`\alpha `$ and $`\beta `$ are the $`S_{Lie}^1`$ spin structure.
Since $`(f(Y_i),\xi )`$ is spin cobordant to
$`(S_a^1,`$ the $`S_{Lie}^1`$ spin structure $`)`$ $``$ $`(S_b^1,`$ the $`S_{Lie}^1`$ spin structure $`)`$,
$`\xi `$ is the $`S_{bd}^1`$ spin structure.
See for the $`S_{bd}^1`$ spin structure and the $`S_{Lie}^1`$ spin structure.
This completes the proof of Lemma 5.6.
This completes the proof of Lemma 5.2.1 ( resp. 5.2.2, 5.2.3 ). By Claim 5.2.4, Theorem 1.3 holds.
Appendix. The proof of Proposition 2.1 and 2.2
Firstly we prove Proposition 2.1.
Let $`L=(K_1,\mathrm{},K_m)`$ be a 1-link. Let $`F`$ be a Seifert surface. Let $`K_i\times [0,1](F)`$ be a collar neighborhood of $`K_i`$ in $`F`$. Let $`K_i\times \{0\}`$=$`K_i`$. Then lk$`(K_i,LK_i)`$=lk$`(K_i\times \{0\},K_i\times \{1\})`$.
See for spin structures.
Let $`\epsilon ^2`$ be the trivial bundle over $`S^1`$. Let $`(e_p^1,e_p^2)`$ (for each $`pS^1`$) denote the trivialization. Let $`e_p^0`$ (for each $`pS^1`$) denote the trivialization on $`TS^1`$. Spin structures on $`S^1`$ are defined by spin structures on $`TS^1\epsilon ^2`$. defines that the spin structure defined by $`(e_p^0,e_p^1,e_p^2)`$ is the $`S_{Lie}^1`$ spin structure. The other is the $`S_{bd}^1`$ spin structure.
Let $`(M,\sigma )`$ and $`(N,\tau )`$ be spin manifolds. Let $`f:MN`$ be an orientation preserving diffeomorphism. $`f`$ is called a spin preserving diffeomorphism if $`dfid:TM\epsilon ^pTN\epsilon ^p`$ carries $`\sigma `$ to $`\tau `$, $`id`$ carries the trivialization to the trivialization.
We give $`F`$ a spin structure $`\alpha `$ induced from the unique spin structure $`\beta `$ on $`S^3`$. We give $`K_i`$ a spin structure $`\gamma `$ induced from $`\alpha `$. Let $`K_i\times D^2`$ be the tubular neighborhood of $`K_i`$ in $`S^3`$. Regard $`K_i\times D^2`$ as the product $`D^2`$ bundle over $`K_i`$. We give $`K_i\times D^2`$ a trivialization such that $`e_p^1`$ is in $`K_i\times [0,1]`$ and that $`e_p^2`$ is perpendicular to $`K_i\times [0,1]`$ for each $`pK_i`$. Then we can regard $`\alpha `$ as a spin structure on the trivialized bundle $`TK_i(K_i\times [0,1])`$.
Let $`G`$ be any Seifert surface of $`K_i`$. Let $`g`$ be a spin structure on $`G`$ induced from the unique one $`S^3`$. Let $`\mu `$ be a spin structure on $`K_i`$ induced from $`(G,g)`$. Since $`(G,g)=(K_i,\mu )`$, $`[(K_i,\mu )]=0`$. We give $`K_i\times D^2`$ a trivialization such that $`g_p^1`$ is in $`G`$ and $`g_p^2`$ is perpendicular to $`G`$. Then the spin structure on $`K_i`$ induced by the framing $`(e_p^0,g_p^1,g_p^2)`$ is $`\mu `$.
The homotopy class of the framing $`(e_p^0,e_p^1,e_p^2)`$ coincides with the homotopy class of the framing $`(e_p^0,g_p^1,g_p^2)`$ if and only if lk$`(K_i\times \{0\},K_i\times \{1\})`$ is even.
Therefore Proposition 2.1 holds.
Next we prove Proposition 2.2.
Under the above condition, suppose $`L`$ is a proper 1-link.
Let $`\widehat{F}`$ be the closed oriented surface $`F(_{i=1}^mD^2)`$. Since $`\gamma `$ is the $`S_{Lie}^1`$ spin structure, $`\alpha `$ extends to $`\widehat{F}`$, say $`\widehat{\alpha }`$. Let $`a_1,\mathrm{},a_g,b_1,\mathrm{},b_g`$ be circles in $`F`$ representing symplectic basis of $`H_1(\widehat{F};Z)`$. Let $`a_i\times [1,1]`$, $`b_i\times [1,1]`$ be the tubular neighborhood in $`F`$. Put mod 2 lk ($`a_i\times \{1\},a_i\times \{1\}`$)=$`x_i`$ and mod 2 lk ($`b_i\times \{1\},b_i\times \{1\}`$)=$`y_i`$. Recall Arf$`L=`$$`\mathrm{\Sigma }x_iy_i`$.
We give a spin structure $`\sigma _i`$, (resp. $`\tau _i`$) on $`a_i`$ (resp. $`b_i`$). Then \[($`a_i,\sigma _i`$)\]$`\mathrm{\Omega }_1^{spin}=`$ $`x_i`$ and \[($`b_i,\tau _i`$)\]$`\mathrm{\Omega }_1^{spin}=`$ $`y_i`$. By P.36 of , \[($`\widehat{F},\widehat{\alpha }`$)\]$`\mathrm{\Omega }_2^{spin}`$ is $`\mathrm{\Sigma }x_iy_i`$ .
Hence Proposition 2.2 holds.
The shaded part is $`FB^3`$.
Figure 10
Figure 11
|
warning/0003/quant-ph0003012.html
|
ar5iv
|
text
|
# New experimental test of Bell inequalities by the use of a non-maximally entangled photon state
Abstract
We report on the realisation of a new test of Bell inequalities using the superposition of type I parametric down conversion produced in two different non-linear crystals pumped by the same laser, but with different polarisations.
We discuss the advantages and the possible developments of this configuration.
PACS: 03.65.Bz
Keywords: Bell inequalities, non-locality, hidden variable theories
Many experiments have already been devoted to a test of Bell inequalities , leading to a substantial agreement with quantum mechanics and disfavouring realistic local hidden variable theories. However, due to the low total detection efficiency (the so-called ”detection loophole”) no experiment has yet been able to exclude definitively realistic local hidden variable theories, for it is necessary a further additional hypothesis , stating that the observed sample of particles pairs is a faithful subsample of the whole. This problem is known as detection or efficiency loophole. Considering the extreme relevance of a conclusive elimination of local hidden variable theories, the research for new experimental configurations able to overcome the detection loophole is of a great interest.
A very important theoretical step in this direction has been achieved recognising that for non maximally entangled pairs a total efficiency larger than 0.67 (in the limit of no background) is required to obtain an efficiency-loophole free experiment, whilst for maximally entangled pairs this limit rises to 0.81. However, it must be noticed that for non-maximally entangled states the largest discrepancy between quantum mechanics and local hidden variable theories is reduced: thus a compromise between a lower total efficiency and a still sufficiently large value of this difference will be necessary in a realisation of an experiment addressed to overcome the detection loophole.
On the experimental side, in recent years the largest improvements have been obtained by using parametric down conversion (PDC) processes.
This technique has been employed, since its discovery, for generations of ”entangled” photon pairs, i.e. pairs of photons described by a common wave function which by no means can be factored up into the product of two distinct wave functions pertaining to separated photons. It is due essentially to L. Mandel and collaborators the idea of using such a state of the electromagnetic field to perform tests of quantum mechanics, more specifically to test Bell’s inequalities.
The generation of entangled states by parametric down conversion is alternative to other techniques, such as the radiative decay of atomic excited states, as it was in the celebrated experiment of A. Aspect et al. , and overcomes some former limitations in the direction of propagation of the conjugated photons. In fact the poor angular correlation of atomic cascade photons is at the origin of a small total efficiency of this set up, leading to a selection of a small subsample of the produced photons, which makes impossible to eliminate the detection loophole in this case. On the other hand, a very good angular correlation (better than 1 mrad) of the two photons of the pair is obtained in the PDC process, giving the possibility, at least in principle, to overcome the previous problem.
The entanglement on phase and momentum, which is directly produced in Type I parametric down conversion can be used for a test of Bell inequalities using two spatially separated interferometers , as realised by . The use of beam splitters, however, strongly reduces the total quantum efficiency.
In alternative, one can generate a polarisation entangled state . It appears, however, that the creation of couples of photons entangled from the point of view of polarisation, which is by far the most diffuse case due to the easy experimental implementation, still suffers severe limitations, as it was pointed out recently in the literature. The essence of the problem is that in generating this state, half of the initial photon flux is lost (in most of the used configurations), and one is, of necessity, led to assume that the photon’s population actually involved in the experiment is a faithful sample of the original one, without eliminating the efficiency loophole.
A scheme which allows no postselection of the photons has been realised recently, using Type II PDC, where a polarisation entangled state is directly generated. This scheme has effectively permitted, at the price of delicate compensations for having identical arrival time of the ordinary and extraordinary photon, a much higher total efficiency than the previous ones. It is, however, still far from the value $`0.81`$ for the total efficiency. Also, some recent experiments studying equalities among correlations functions rather than Bell inequalities are far by solving these problems . A large interest remains therefore for new experiments increasing total quantum efficiency in order to reduce and finally overcome the efficiency loophole.
Our experiment follows and develops an idea by Hardy and contemplates the creation of a polarisation (non maximally-) entangled states of the form
$$|\psi =\frac{|H|H+f|V|V}{\sqrt{(1+|f|^2)}}$$
(1)
(where $`H`$ and $`V`$ indicate horizontal and vertical polarisations respectively) via the superposition of the spontaneous fluorescence emitted by two non-linear crystals driven by the same pumping laser. The crystals are put in cascade along the propagation direction of the pumping laser and the superposition is obtained by using an appropriate optics.
We describe more in details the experimental set-up, with reference to the sketch of figure 1 below. The two crystals of LiIO3 (10x10x10 mm) are 250 mm apart, a distance smaller than the coherence length of the pumping laser. This guarantees indistinguishibility in the creation of a couple of photons in the first or in the second crystal. A couple of planoconvex lenses of 120 mm focal length centred in between focalises the spontaneous emission from the first crystal into the second one maintaining exactly, in principle, the angular spread. A hole of 4 mm diameter is drilled into the centre of the lenses in order to allow transmission of the pump radiation without absorption and, even more important, without adding stray-light, because of fluorescence and diffusion of the UV radiation. This configuration, which performs as a so-called ”optical condenser”, was chosen among others as a compromise between minimisation of aberrations (mainly spherical and chromatic) and losses due to the number of optical components. The pumping beam at the exit of the first crystal is displaced from its input direction by birefringence: the small quartz plate (5 x5 x5 mm) in front of the first lens of the condensers compensates this displacement, so that the input conditions are prepared to be the same for the two crystals, apart from alignment errors. Finally, a half-wavelength plate immediately after the condenser rotates the polarisation of the Argon beam and excites in the second crystal a spontaneous emission cross-polarised with respect to the first one. The dimensions and positions of both plates are carefully chosen in order not to intersect the spontaneous emissions at 633 and 789 nm. With a phase matching angle of $`51^o`$ they are emitted at $`3.5^o`$ and $`4^o`$ respectively.
The precise value of $`f`$, which determines how far from a maximally entangled state ($`f=1`$) the produced state is, can be easily tuned by modifying the pump intensity between the two crystals. This is a fundamental property, which could permit to select the most appropriate state for the practical realisation of a detection loophole free experiment.
The alignment (which is of fundamental importance for having a high visibility and in principle constitutes the main problem of such a configuration) has been profitably improved using an optical amplifier scheme, where a solid state laser is injected into the crystals together with the pumping laser, an argon laser at 351 nm wavelength (see fig.1). Such a technique had already proved its validity and it is of great help in the recognition of the directions of propagation of correlated wavelengths photon and applied in our laboratory for metrological studies.
The proposed scheme is well suited for leading to a further step toward a conclusive experimental test of non-locality in quantum mechanics. The analysis of the experiments realised up to now shows in fact that visibility of the wanted effect (essentially visibility of interference fringes) and overall quantum efficiency of detection are the main parameters in such experiments. One first advantage of the proposed configuration with respect to most of the previous experimental set-ups is that all the entangled pairs are selected (and not only $`50\%`$ as with beams splitters).
For the moment, the results which we are going to discuss are still far from a definitive solution of the detection loophole; nevertheless, being the first test of Bell inequalities using a non-maximally entangled state, they represents an important step in this direction. Furthermore, this configuration permits to use any pair of correlated frequencies and not only the degenerate ones. We have thus realised this test using for a first time two different wave lengths (at $`633`$ and $`789`$ nm).
A somehow similar set-up has been realised recently in ref. . The main difference between the two experiments is that in the two crystals are very thin and in contact with orthogonal optical axes: this permits a ”partial” superposition of the two emissions with opposite polarisation. This overlapping is mainly due to the finite dimension of the pump laser beam, which reflects into a finite width of each wavelength emission. We think that the arrangement, described in the present paper is better, in that by fine tuning the crystals’ and optics’ positions a perfect superposition can be obtained for at least a couple of wavelength in PDC emission. Moreover, the parametric amplifier trick allows to approach this ideal situation very closely. Furthermore, in the experiment of ref. the value of $`f`$ is in principle tunable by rotating the polarisation of the pump laser, however this reduce the power of the pump producing PDC already in the first crystal, while in our case the whole pump power can always be used in the first crystal, tuning the PDC produced in the second one.
As a first check of our apparatus, we have measured the interference fringes, varying the setting of one of the polarisers, leaving the other fixed. We have found a high visibility, $`V=0.973\pm 0.038`$.
Our results are summarised by the value obtained for the Clauser-Horne sum,
$$CH=N(\theta _1,\theta _2)N(\theta _1,\theta _2^{})+N(\theta _1^{},\theta _2)+N(\theta _1^{},\theta _2^{})N(\theta _1^{},\mathrm{})N(\mathrm{},\theta _2)$$
(2)
which is strictly negative for local realistic theory. In (2), $`N(\theta _1,\theta _2)`$ is the number of coincidences between channels 1 and 2 when the two polarisers are rotated to an angle $`\theta _1`$ and $`\theta _2`$ respectively ($`\mathrm{}`$ denotes the absence of selection of polarisation for that channel)
For quantum mechanics $`CH`$ can be larger than zero, e.g. for a maximally entangled state the largest value is obtained for $`\theta _1=67^o.5`$ , $`\theta _2=45^o`$, $`\theta _1^{}=22^o.5`$ , $`\theta _2^{}=0^o`$ and correspond to a ratio
$$R=[N(\theta _1,\theta _2)N(\theta _1,\theta _2^{})+N(\theta _1^{},\theta _2)+N(\theta _1^{},\theta _2^{})]/[N(\theta _1^{},\mathrm{})+N(\mathrm{},\theta _2)]$$
(3)
equal to 1.207.
For non-maximally entangled states the angles for which CH is maximal are somehow different and the largest value smaller. The angles corresponding to the maximum can be evaluated maximising Eq. 2 with
$`\begin{array}{c}N[\theta _1,\theta _2]=[ϵ_1^{||}ϵ_2^{||}(Sin[\theta _1]^2Sin[\theta _2]^2)+\hfill \\ ϵ_1^{}ϵ_2^{}(Cos[\theta _1]^2Cos[\theta _2]^2)\hfill \\ (ϵ_1^{}ϵ_2^{||}Sin[\theta _1]^2Cos[\theta _2]^2+ϵ_1^{||}ϵ_2^{}Cos[\theta _1]^2Sin[\theta _2]^2)\hfill \\ +|f|^2(ϵ_1^{}ϵ_2^{}(Sin[\theta _1]^2Sin[\theta _2]^2)+ϵ_1^{||}ϵ_2^{||}(Cos[\theta _1]^2Cos[\theta _2]^2)+\hfill \\ (ϵ_1^{||}ϵ_2^{}Sin[\theta _1]^2Cos[\theta _2]^2+\hfill \\ ϵ_1^{}ϵ_2^{||}Cos[\theta _1]^2Sin[\theta _2]^2)\hfill \\ +(f+f^{})((ϵ_1^{||}ϵ_2^{||}+ϵ_1^{}ϵ_2^{}ϵ_1^{||}ϵ_2^{}ϵ_1^{}ϵ_2^{||})(Sin[\theta _1]Sin[\theta _2]Cos[\theta _1]Cos[\theta _2])]/(1+|f|^2)\hfill \end{array}.`$ (11)
where (for the case of non-ideal polariser) $`ϵ_i^{||}`$ and $`ϵ_i^{}`$ correspond to the transmission when the polariser (on the branch $`i`$) axis is aligned or normal the polarisation axis respectively.
In our case we have a state with $`f0.4`$ which corresponds, for $`\theta _1=72^o.24`$ , $`\theta _2=45^o`$, $`\theta _1^{}=17^o.76`$ and $`\theta _2^{}=0^o`$, to $`R=1.16`$.
Our experimental result is $`CH=512\pm 135`$ coincidences per second, which is almost four standard deviations different from zero and compatible with the theoretical value predicted by quantum mechanics. In terms of the ratio (3), our result is $`1.082\pm 0.031`$.
It must be noticed that results which are by far of many more standard deviations above zero have been obtained. Nevertheless, to our knowledge, this is the first measurement of the Clauser-Horne sum (or other Bell inequalities) using a non-maximally entangled state and thus represents and interesting result as a first step in the direction of eliminating the detection loophole. This is also the first measurement of a polarisation two photon entanglemend state, where the two photons have different wavelengths. Further developments in this sense are the purpose of this collaboration.
Acknowledgements We would like to acknowledge support of Italian Space Agency under contract LONO 500172.
Figure caption
Sketch of the source of polarisation entangled photons. CR1 and CR2 are two $`LiIO_3`$ crystals cut at the phase-matching angle of $`51^o`$. L1 and L2 are two identical piano-convex lenses with a hole of 4 mm in the centre. P is a 5 x 5 x 5 mm quartz plate for birefringence compensation and $`\mathrm{\Lambda }/2`$ is a first order half wave-length plate at 351 nm. U.V. identifies the pumping radiation at 351 nm. The infrared beam (I.R.) at 789 nm is generated by a diode laser and is used for system alignment only. The parametric amplifier scheme, described in the text, is shown as well. The dashed line identifies the idler radiation at 633 nm. A second half-wave plate on the I.R. beam (not shown in the figure) allows amplified idler emission from the second crystals too. The figure is not in scale.
|
warning/0003/hep-ph0003158.html
|
ar5iv
|
text
|
# Positive strangeness contribution to the nucleon magnetic moment in a relativistic chiral potential model
## Abstract
The strangeness contribution to the nucleon magnetic moment is calculated at the one-loop level in a relativistic SU(3) chiral potential model and is found to be positive, that is, with an opposite sign to the nucleon strangeness polarization. It is the “Z” diagram that violates the usual relation between spin and magnetic moment. The positive value is due to the contribution from the intermediate excited quark states, while the intermediate ground state gives a negative contribution. Our numerical results agree quite well with the new measurement of the SAMPLE Collaboration.
PACS numbers: 12.39.Ki, 12.39.Fe, 12.39.Pn, 13.88.+e
The new determination of the strangeness magnetic form factor of the nucleon at $`Q^2=0.1`$GeV<sup>2</sup> by the SAMPLE collaboration confirmed its previous result and indicates a significantly positive value:
$$G_M^s(0.1\mathrm{GeV}^2)=+0.61\pm 0.17\pm 0.21\pm 0.19\mu _N.$$
(1)
This result implies new challenges to our understanding of the nucleon structure since most theoretical calculations typically generate negative values for $`\mu _sG_M^s(Q^2=0)`$ (see and references therein). A positive value for $`\mu _s`$ is also intuitively difficult to understand with respect to the usual magnetic moment-spin ($`\mu `$-$`s`$) relation, since the strange quark polarization of the nucleon is confirmed to be negative both by experiments and by lattice QCD calculations . (Note that the negative charge of the strange quark has been extracted in the definition of $`G_M^s`$.)
In a recent paper , by using a relativistic chiral potential model, we have successfully reproduced the experimental result of the strange quark polarization ($`\mathrm{\Delta }s`$) of the nucleon. In this paper we report on the corresponding result for the nucleon strangeness magnetic moment, especially in comparison to the nucleon strangeness polarization.
The nucleon magnetic moment $`\mu _N`$ is defined by its interaction with a static, external magnetic field $`\stackrel{}{B}`$:
$$N|\underset{q}{}iQ_qd^3x\overline{\psi }_q\gamma ^\mu \psi _qA_\mu |N\stackrel{}{\mu }_N\stackrel{}{B},$$
(2)
It can be shown that $`\mu _N`$ is related to the electromagnetic form factor by $`\mu _N=_qQ_qG_M^q(0)`$ , where $`G_M^q(k^2)F_1^q(k^2)+F_2^q(k^2)`$ and $`F_1`$ and $`F_2`$ are defined through
$$N|\overline{\psi }_q\gamma ^\mu \psi _q|N=\overline{u}_N\left(F_1^q\gamma ^\mu +\frac{i}{2M_N}F_2^q\sigma ^{\mu \nu }k_\nu \right)u_N.$$
(3)
The contribution of the quark flavor $`q`$ ($`q=u,d,s`$) to the nucleon magnetic moment is usually defined as $`\mu _qG_M^q(0)`$. Equivalently, $`\mu _q`$ can be evaluated as the expectation value of the magnetic moment operator:
$$\mu _q=N|d^3x\psi _q^{}(\stackrel{}{x}\times \stackrel{}{\alpha })_3\psi _q|N,$$
(4)
which follows directly from Eq. (2) . Eq. (4) is especially suitable for model calculations. In the following we will perform a perturbative calculation of $`\mu _s`$ in a chiral potential model.
Our starting point is the chiral Lagrangian
$``$ $`=`$ $`\overline{\psi }[i/S(r)\gamma ^0V(r)]\psi `$ (7)
$`{\displaystyle \frac{1}{2F_\pi }}\overline{\psi }[S(r)(\sigma +i\gamma ^5\lambda ^i\varphi _i)+(\sigma +i\gamma ^5\lambda ^i\varphi _i)S(r)]\psi +`$
$`{\displaystyle \frac{1}{2}}(_\mu \sigma )^2+{\displaystyle \frac{1}{2}}(_\mu \varphi _i)^2{\displaystyle \frac{1}{2}}m_\sigma ^2\sigma ^2{\displaystyle \frac{1}{2}}m_{\varphi _i}^2\varphi _i^2.`$
The model Lagrangian is derived from the $`\sigma `$ model in which meson fields are introduced to restore chiral symmetry . The flavor and color indices for the quark field $`\psi `$ are suppressed; the scalar term $`S(r)=cr+m`$ represents the the linear scalar confinement potential $`cr`$ and the quark mass matrix $`m`$; $`V(r)=\alpha /r`$ is the Coulomb type vector potential and $`F_\pi `$=93MeV is the pion decay constant. $`\sigma `$ and $`\varphi _i`$ ($`i`$ runs from $`1`$ to $`8`$) are the scalar and pseudoscalar meson fields, respectively and $`\lambda _i`$ are the Gell-Mann matrices. The quark-meson interaction term of Eq.(7) is symmetrized since the mass matrix $`m`$ does not commute with all $`\lambda _i`$ for different quark masses.
At zeroth order the nucleon is described by the usual SU(6) three-quark ground state of the Hamiltonian
$$H_q=d^3x\psi ^{}[\stackrel{}{\alpha }\frac{1}{i}\stackrel{}{}+\beta S(r)+V(r)]\psi .$$
(8)
The lowest order contribution to $`\mu _s`$ arises from the one-loop diagrams of Fig. 1, which we now evaluate.
The meson propagator, given by the Lagrangian of Eq. (7), is the free propagator. Since the non-perturbative confinement is included in $`H_q`$ the quark propagator has to be obtained numerically, and in practise we have to work with time-ordered perturbation theory. We write the solution of $`H_q`$ as
$$\psi (x)=\underset{\alpha }{}u_\alpha (x)a_\alpha +\underset{\beta }{}v_\beta (x)b_\beta ^{},$$
(9)
where $`u_\alpha (x)=e^{iE_\alpha t}u_\alpha (\stackrel{}{x})\tau _\alpha `$, $`v_\beta (x)=e^{iE_\beta t}v_\beta (\stackrel{}{x})\tau _\beta `$; $`\tau `$ is the flavor wavefunction and the spatial wavefunction is:
$$u_\alpha (\stackrel{}{x})=\left(\begin{array}{c}g_{njl}(r)\\ i\stackrel{}{\sigma }\widehat{\stackrel{}{r}}f_{njl}(r)\end{array}\right)Y_{jl}^m(\widehat{\stackrel{}{r}}),$$
(10)
where $`g`$ and $`f`$ are real functions, $`n`$ is the radial quantum number, and $`Y_{jl}^m(\widehat{\stackrel{}{r}})`$ are the vector spherical harmonics. For computational convenience, we will take the same form for $`v_\beta (\stackrel{}{x})`$.
In correspondence to Eq. (9), the quark propagator is
$`D(x_1,x_2)`$ $``$ $`0|T\{\psi (x_1),\overline{\psi }(x_2)\}|0`$ (11)
$`=`$ $`\theta (t_1t_2){\displaystyle \underset{\alpha }{}}u_\alpha (x_1)\overline{u}_\alpha (x_2)`$ (13)
$`\theta (t_2t_1){\displaystyle \underset{\beta }{}}v_\beta (x_1)\overline{v}_\beta (x_2).`$
Applying the propagators to Fig. 1, we get the contribution for a single quark line (with the initial and final states denoted as $`u_i`$ and $`u_f`$, respectively):
$`\mu _s`$ $`=`$ $`{\displaystyle \frac{1}{F_\pi ^2}}{\displaystyle }d^4x_1d^4x_2\overline{u}_f(x_2)S(r_2)\gamma ^5\lambda ^i\times `$ (19)
$`[\theta (t_2t)\theta (tt_1){\displaystyle \underset{\alpha \alpha ^{}}{}}u_\alpha (x_2)\mathrm{\Gamma }_{\alpha \alpha ^{}}\overline{u}_\alpha ^{}(x_1)+`$
$`\theta (t_1t)\theta (tt_2){\displaystyle \underset{\beta \beta ^{}}{}}v_\beta (x_2)\mathrm{\Gamma }_{\beta \beta ^{}}\overline{v}_\beta ^{}(x_1)`$
$`\theta (t_2t)\theta (t_1t){\displaystyle \underset{\alpha \beta ^{}}{}}u_\alpha (x_2)\mathrm{\Gamma }_{\alpha \beta ^{}}\overline{v}_\beta ^{}(x_1)`$
$`\theta (tt_2)\theta (tt_1){\displaystyle \underset{\beta \alpha ^{}}{}}v_\beta (x_2)\mathrm{\Gamma }_{\beta \alpha ^{}}\overline{u}_\alpha ^{}(x_1)]\times `$
$`S(r_1)\gamma ^5\lambda ^iu_i(x_1){\displaystyle \frac{i}{(2\pi )^4}}{\displaystyle d^4k\frac{\delta _{ij}e^{ik(x_1x_2)}}{k^2m_{\varphi _i}^2+iϵ}},`$
where $`\mathrm{\Gamma }_{\alpha \alpha ^{}}d^3xu_\alpha ^{}(\stackrel{}{x}\times \stackrel{}{\alpha })_3u_\alpha ^{}`$, and similarly for $`\mathrm{\Gamma }_{\beta \beta ^{}}`$ etc. The four time-ordered terms in Eq.(19) correspond to the time-ordered diagrams of Fig. 2.
We omit here the details for the evaluation of $`\mu _s`$ of Eq. (19). The integrals of Eq. (19) can be reduced analytically to radial integrations at the vertex points ($`r_1`$ and $`r_2`$) and of the loop momentum $`|\stackrel{}{k}|`$. The the remaining integrations are carried out numerically. To obtain $`\mu _s`$ for the whole nucleon, one still has to multiply a spin-isospin factor which can be straightforwardly calculated to be 2.
When calculating $`\mu _s`$ we allow strong variations of the model parameters entering in the Lagrangian fo Eq. (7). Since $`F_\pi =93`$ MeV and $`m_K=495`$ MeV are fixed by experiment, our model contains four free parameters: the two quark masses $`m_{u,d}`$, $`m_s`$ and the two strength constants of the scalar and vector potential denoted by $`c`$ and $`\alpha `$. The parameter $`\alpha `$ is fixed by the long-wavelength, transverse fluctuations of the QCD based static-source flux-tube picture . It was obtained to be $`0.26`$ in and $`0.30`$ in , while a much larger value of about $`0.52`$ was used by the Cornell group . Recent lattice calculation got a value around $`0.32`$ in the quenched approximation, and suggested that relaxing the quenched approximation may lead to $`\alpha 0.40`$. Effective quark masses and confinement strength are rather uncertain quantities. We therefore choose in our calculation five different sets of parameters (see Table I), including both current and constituent quark masses.
Fig. 3 gives the numerical results of $`\mu _s`$ in units of $`\mu _N`$. The intermediate quark states are consistently summed up to a given energy. In the last column of Table I we list the contribution of the intermediate ground state. As evident from Fig. 3, for all choices of parameter sets, $`\mu _s`$ turns out to be positive, as long as enough excited quark states are taken into account. However, as already indicated in Table I the intermediate quark ground state always give a negative contribution. This explains why in many calculations at the baryon level, where the quarks are restricted to the ground state and the intermediate baryon is truncated to be the ground state octet or decuplet baryons, a negative value of $`\mu _s`$ is usually obtained.
We see in Fig. 3 that the summation over the quark intermediate states is divergent. This is because in the chiral Lagrangian of Eq. (7), the electromagnetic current of the strange quark is not separately conserved. To obtain a meaningful finite result, we must renormalize the composite, non-conserved magnetic moment operator of the strange quark. Analogous to the lattice renormalization, we cut the quark intermediate states at a certain energy, which should roughly correspond to the inverse of the lattice spacing ($`a^11.7`$GeV) in the lattice QCD calculation of nucleon properties . The cutoff point is indicated in Fig. 3.
Fig. 3 shows that a larger quark mass or a stronger confinement (which is effectively a static mass) always reduce the magnetic moment, as expected. However, the variations of the vector potential do not affect $`\mu _s`$ too much.
To analyze how the positive value for $`\mu _s`$ arises, in Fig. 5 we give the separate contributions to $`\mu _s`$ from the time-ordered diagrams of Fig. 2 for the second set of parameters. Correspondingly in Fig. 6 we indicate the separate contributions of the time-ordered diagrams to strange quark polarization $`\mathrm{\Delta }s`$ of the nucleon .
The results of Figs. 5 and 6 indicate why $`\mu _s`$ and $`\mathrm{\Delta }s`$ have opposite sign: the intermediate quark states give a contribution of the same sign (both negative) to $`\mu _s`$ and $`\mathrm{\Delta }s`$, which is as expected; the antiquark states contribute a positive amount to $`\mathrm{\Delta }s`$, but a negative amount to $`\mu _s`$. This is also reasonable since the antiquark has an opposite charge to the quark. However, one would not expect the usual relation between spin and magnetic moment for the contributions from the “Z” diagrams in which a quark-antiquark pair is created or annihilated; an evident obstacle is that we do not know what sign of the charge we should attribute to these diagrams. Figs. 5 and 6 show that the “Z” diagrams give a negative contribution to the polarization while they generate a positive contribution to magnetic moment. They are also the dominating contributions (note that there are two “Z” diagrams); so eventually we get for the whole nucleon a negative strangeness polarization but a positive strangeness magnetic moment. This agrees well with the experimental results.
In summary, we found by a standard perturbative calculation that the SU(3) chiral potential model predicts a positive nucleon strangeness magnetic moment for a wide range of model parameters. Here the contributions from the intermediate excited states and the “Z” diagrams are most important. If one restricts the intermediate state to the quark ground state an opposite result is obtained. Further investigation of the time-ordered diagrams reveals that the positive-energy and negative-energy states (Figs. 2A and 2B) contribute to polarization and magnetic moment in the usual way that respects the relation between spin and magnetic moment; however this is not true for the “Z” diagrams, whose contribution is found to be dominant and therefore determines the overall sign of $`\mu _s`$ and $`\mathrm{\Delta }s`$ for the nucleon. Since in our calculations both for $`\mu _s`$ and $`\mathrm{\Delta }s`$ we have assumed only one Lagrangian, the success of our calculations can be regarded as a strong support of our approach and the chiral Lagrangian.
This work is supported by the CNSF (19675018), CSED, CSSTC, the DFG (FA67/25-1), and the DAAD.
|
warning/0003/hep-lat0003011.html
|
ar5iv
|
text
|
# S and P-wave heavy-light mesons in lattice NRQCD
## I INTRODUCTION
Calculations in QCD can be performed numerically on a discrete space-time lattice, provided the lattice spacing is small enough to accommodate all of the revelant physical distance scales. In the presence of a heavy quark the lattice spacing must be small relative to the inverse quark mass, resulting in large computational requirements, unless an appropriate “effective theory” is used. In particular, for the case of a hadron containing exactly one heavy quark, the dynamics can be expanded in powers of the inverse heavy quark mass using the well-established technique of heavy quark effective theory or nonrelativistic QCD (NRQCD). With the heavy quark expansion, the lattice spacing can be much coarser and the computational requirements are correspondingly smaller.
In the present work, two issues will be addressed within quenched lattice NRQCD. Firstly, is the charm quark sufficiently heavy to permit the use of lattice NRQCD for charmed meson spectroscopy? Secondly, what are the mass splittings (magnitude and sign) between P-wave mesons containing a single heavy quark?
The first issue is clearly of interest due to the computational efficiency of lattice NRQCD. The heavy quark expansion is known to work well for bottom mesons, but previous research has demonstrated that $`O(1/M^2)`$ and $`O(1/M^3)`$ contributions can be comparable in magnitude for the S-wave spin splitting of charmed mesons. The present work provides an extension of this investigation by considering new simulations that incorporate a number of changes in method. For example, Ref. used the fourth root of an elementary plaquette to define the tadpole improvement factor, $`U_0`$, whereas the present work uses the mean link in Landau gauge. Since the Landau definition is advantageous in other contexts, including the velocity expansion for the charmonium spectrum of lattice NRQCD, it might be expected to improve the convergence of the heavy quark expansion for the charmed heavy-light spectrum as well. Also, the light quark was described by the Sheikholeslami-Wohlert action in Ref. , but the present work uses a D234 action which has smaller lattice spacing errors classically. The simulations in the present work differ from those of Ref. in various other ways as well, including Dirichlet versus periodic boundary conditions for light quark propagation, differing discretizations for heavy quark propagation, and the introduction of an anisotropic lattice with a smaller temporal lattice spacing than spatial spacing. Despite these modifications, the present simulations produce a conclusion similar to that of Ref. : the $`O(1/M^2)`$ and $`O(1/M^3)`$ contributions are comparable in magnitude for the S-wave charmed spin splitting; in the present work, each is roughly 20% of the $`O(1/M)`$ result for $`D^{}D`$, though the $`O(1/M^3)`$ contribution is closer to 10% for $`D_s^{}D_s`$.
The second issue under discussion relates to the spectrum of P-wave mesons containing a single heavy quark. The relative orderings of the P-wave bottom mesons have only recently come under direct experimental scrutiny, and the complete picture is not yet clear. Meanwhile theoretical predictions differ from one another even at a qualitative level. The traditional expectation of a hydrogen-like spectrum that arises from a number of model calculations has been questioned long ago by Schnitzer and very recently by Isgur and by Ebert, Galkin and Faustov.
The calculation is difficult within lattice QCD because the P-wave splittings are not large in comparison to the typical scale of nonperturbative QCD, and because of operator mixing for the pair of P-wave states having $`J=1`$. Some previous attempts have been made. Unfortunately, the uncertainties are often substantial, and results are not always as consistent with one another as might have been hoped. The present work represents a further comment on this situation. In particular, the $`D_{s2}^{}D_{s0}^{}`$ mass splitting is found to be positive as in the traditional hydrogen-like ordering, and to be substantially less than 100 MeV, while the $`D_2^{}D_0^{}`$, $`B_{s2}^{}B_{s0}^{}`$ and $`B_2^{}B_0^{}`$ splittings are even smaller. These splittings are consistent with a number of model calculations, but are somewhat smaller than the lattice NRQCD calculation of Ref. .
## II ACTION
The lattice action has three terms: gauge action, light quark action and heavy quark action. The entire action is classically and tadpole-improved with the tadpole factors, $`u_s`$ and $`u_t`$, defined as the mean links in Landau gauge in a spatial and temporal direction, respectively.
The gauge field action is
$`S_G(U)`$ $`=`$ $`{\displaystyle \frac{5\beta }{3}}[{\displaystyle \frac{1}{u_s^4\xi }}{\displaystyle \underset{\mathrm{ps}}{}}(1{\displaystyle \frac{1}{3}}\mathrm{ReTr}U_{\mathrm{ps}}){\displaystyle \frac{1}{20u_s^6\xi }}{\displaystyle \underset{\mathrm{rs}}{}}(1{\displaystyle \frac{1}{3}}\mathrm{ReTr}U_{\mathrm{rs}})`$ (3)
$`+{\displaystyle \frac{\xi }{u_s^2u_t^2}}{\displaystyle \underset{\mathrm{pt}}{}}\left(1{\displaystyle \frac{1}{3}}\mathrm{ReTr}U_{\mathrm{pt}}\right){\displaystyle \frac{\xi }{20u_s^4u_t^2}}{\displaystyle \underset{\mathrm{rst}}{}}\left(1{\displaystyle \frac{1}{3}}\mathrm{ReTr}U_{\mathrm{rst}}\right)`$
$`{\displaystyle \frac{\xi }{20u_s^2u_t^4}}{\displaystyle \underset{\mathrm{rts}}{}}(1{\displaystyle \frac{1}{3}}\mathrm{ReTr}U_{\mathrm{rts}})],`$
where $`\xi a_s/a_t`$. The subscripts “ps” and “rs” denote spatial plaquettes and spatial planar 1$`\times `$2 rectangles respectively. Plaquettes in the temporal-spatial planes are denoted by “pt”, while rectangles with the long side in a spatial(temporal) direction are labeled by “rst”(“rts”). The leading classical errors of this action are quartic in lattice spacing.
For light quarks, a D234 action is used with parameters set to their tadpole-improved classical values. Its leading classical errors are cubic in lattice spacing.
$`S_F(\overline{q},q;U)`$ $`=`$ $`{\displaystyle \frac{4\kappa }{3}}{\displaystyle \underset{x,i}{}}\left[{\displaystyle \frac{1}{u_s\xi ^2}}D_{1i}(x){\displaystyle \frac{1}{8u_s^2\xi ^2}}D_{2i}(x)\right]+{\displaystyle \frac{4\kappa }{3}}{\displaystyle \underset{x}{}}\left[{\displaystyle \frac{1}{u_t}}D_{1t}(x){\displaystyle \frac{1}{8u_t^2}}D_{2t}(x)\right]`$ (6)
$`+{\displaystyle \frac{2\kappa }{3u_s^4\xi ^2}}{\displaystyle \underset{x,i<j}{}}\overline{\psi }(x)\sigma _{ij}F_{ij}(x)\psi (x)+{\displaystyle \frac{2\kappa }{3u_s^2u_t^2\xi }}{\displaystyle \underset{x,i}{}}\overline{\psi }(x)\sigma _{0i}F_{0i}(x)\psi (x)`$
$`{\displaystyle \underset{x}{}}\overline{\psi }(x)\psi (x),`$
where
$`D_{1i}(x)`$ $`=`$ $`\overline{\psi }(x)(1\xi \gamma _i)U_i(x)\psi (x+\widehat{i})+\overline{\psi }(x+\widehat{i})(1+\xi \gamma _i)U_i^{}(x)\psi (x),`$ (7)
$`D_{1t}(x)`$ $`=`$ $`\overline{\psi }(x)(1\gamma _4)U_4(x)\psi (x+\widehat{t})+\overline{\psi }(x+\widehat{t})(1+\gamma _4)U_4^{}(x)\psi (x),`$ (8)
$`D_{2i}(x)`$ $`=`$ $`\overline{\psi }(x)(1\xi \gamma _i)U_i(x)U_i(x+\widehat{i})\psi (x+2\widehat{i})`$ (9)
$`+`$ $`\overline{\psi }(x+2\widehat{i})(1+\xi \gamma _i)U_i^{}(x+\widehat{i})U_i^{}(x)\psi (x),`$ (10)
$`D_{2t}(x)`$ $`=`$ $`\overline{\psi }(x)(1\gamma _4)U_4(x)U_4(x+\widehat{t})\psi (x+2\widehat{t})`$ (11)
$`+`$ $`\overline{\psi }(x+2\widehat{t})(1+\gamma _4)U_4^{}(x+\widehat{t})U_4^{}(x)\psi (x),`$ (12)
$`gF_{\mu \nu }(x)`$ $`=`$ $`{\displaystyle \frac{1}{2i}}\left(\mathrm{\Omega }_{\mu \nu }(x)\mathrm{\Omega }_{\mu \nu }^{}(x)\right){\displaystyle \frac{1}{3}}\mathrm{Im}\left(\mathrm{Tr}\mathrm{\Omega }_{\mu \nu }(x)\right),`$ (13)
$`\mathrm{\Omega }_{\mu \nu }`$ $`=`$ $`{\displaystyle \frac{1}{4}}[U_\mu (x)U_\nu (x+\widehat{\mu })U_\mu ^{}(x+\widehat{\nu })U_\nu ^{}(x)`$ (17)
$`+U_\nu (x)U_\mu ^{}(x\widehat{\mu }+\widehat{\nu })U_\nu ^{}(x\widehat{\mu })U_\mu (x\widehat{\mu })`$
$`+U_\mu ^{}(x\widehat{\mu })U_\nu ^{}(x\widehat{\mu }\widehat{\nu })U_\mu (x\widehat{\mu }\widehat{\nu })U_\nu (x\widehat{\nu })`$
$`+U_\nu ^{}(x\widehat{\nu })U_\mu (x\widehat{\nu })U_\nu (x+\widehat{\mu }\widehat{\nu })U_\mu ^{}(x)].`$
The heavy quark action is NRQCD, which is discretized to give the following Green’s function propagation:
$`G_{\tau +1}`$ $`=`$ $`\left(1{\displaystyle \frac{a_tH_B}{2}}\right)\left(1{\displaystyle \frac{a_tH_A}{2n}}\right)^n{\displaystyle \frac{U_4^{}}{u_t}}\left(1{\displaystyle \frac{a_tH_A}{2n}}\right)^n\left(1{\displaystyle \frac{a_tH_B}{2}}\right)G_\tau ,`$ (18)
with $`n=5`$ chosen for this work. Separation of the Hamiltonian into two terms, $`H=H_A+H_B`$, is important for ensuring stability of the discretization. For example, recall the discussion in Ref. of a large nonzero vacuum expectation value for the term containing $`c_{10}`$ in the Hamiltonian (see Eq. (25)). This issue will be discussed further in Sec. IV.
The following Hamiltonian, written in terms of the bare heavy quark mass $`M`$, is complete to $`O(1/M^3)`$ in the classical continuum limit:
$`H`$ $`=`$ $`H_0+\delta H,`$ (19)
$`H_0`$ $`=`$ $`{\displaystyle \frac{\mathrm{\Delta }^{(2)}}{2M}},`$ (20)
$`\delta H`$ $`=`$ $`\delta H^{(1)}+\delta H^{(2)}+\delta H^{(3)}+O(1/M^4)`$ (21)
$`\delta H^{(1)}`$ $`=`$ $`{\displaystyle \frac{c_4}{u_s^4}}{\displaystyle \frac{g}{2M}}𝝈\stackrel{~}{𝐁}+c_5{\displaystyle \frac{a_s^2\mathrm{\Delta }^{(4)}}{24M}},`$ (22)
$`\delta H^{(2)}`$ $`=`$ $`{\displaystyle \frac{c_2}{u_s^2u_t^2}}{\displaystyle \frac{ig}{8M^2}}(\stackrel{~}{𝚫}\stackrel{~}{𝐄}\stackrel{~}{𝐄}\stackrel{~}{𝚫}){\displaystyle \frac{c_3}{u_s^2u_t^2}}{\displaystyle \frac{g}{8M^2}}𝝈(\stackrel{~}{𝚫}\times \stackrel{~}{𝐄}\stackrel{~}{𝐄}\times \stackrel{~}{𝚫})c_6{\displaystyle \frac{a_s(\mathrm{\Delta }^{(2)})^2}{16n\xi M^2}},`$ (23)
$`\delta H^{(3)}`$ $`=`$ $`c_1{\displaystyle \frac{(\mathrm{\Delta }^{(2)})^2}{8M^3}}{\displaystyle \frac{c_7}{u_s^4}}{\displaystyle \frac{g}{8M^3}}\{\stackrel{~}{\mathrm{\Delta }}^{(2)},𝝈\stackrel{~}{𝐁}\}{\displaystyle \frac{c_9ig^2}{8M^3}}𝝈\left({\displaystyle \frac{\stackrel{~}{𝐄}\times \stackrel{~}{𝐄}}{u_s^4u_t^4}}+{\displaystyle \frac{\stackrel{~}{𝐁}\times \stackrel{~}{𝐁}}{u_s^8}}\right)`$ (25)
$`{\displaystyle \frac{c_{10}g^2}{8M^3}}\left({\displaystyle \frac{\stackrel{~}{𝐄}^2}{u_s^4u_t^4}}+{\displaystyle \frac{\stackrel{~}{𝐁}^2}{u_s^8}}\right)c_{11}{\displaystyle \frac{a_s^2(\mathrm{\Delta }^{(2)})^3}{192n^2\xi ^2M^3}}.`$
The coefficients of the Hamiltonian are chosen so the dimensionless parameters, $`c_i`$, are unity at the classical level. As will be discussed below, computations have been performed with the $`c_i`$ set to unity or zero in various combinations, including separate computations at $`O(1/M)`$, $`O(1/M^2)`$ and $`O(1/M^3)`$ to allow discussions of convergence for the $`1/M`$ expansion. Throughout this work, $`H_0`$ is always placed in $`H_A`$ of Eq. (18) and all of the remaining terms except the $`c_{10}`$ term are only placed in $`H_B`$. The difference between having the $`c_{10}`$ term in $`H_A`$ or $`H_B`$ will be discussed explicitly, since it has the nonzero vacuum expectation value.
A tilde on any quantity indicates that the leading discretization errors have been removed. In particular,
$`\stackrel{~}{E}_i`$ $`=`$ $`\stackrel{~}{F}_{4i},`$ (26)
$`\stackrel{~}{B}_i`$ $`=`$ $`{\displaystyle \frac{1}{2}}ϵ_{ijk}\stackrel{~}{F}_{jk},`$ (27)
where
$`\stackrel{~}{F}_{\mu \nu }(x)`$ $`=`$ $`{\displaystyle \frac{5}{6}}F_{\mu \nu }(x){\displaystyle \frac{1}{6u_\mu ^2}}U_\mu (x)F_{\mu \nu }(x+\widehat{\mu })U_\mu ^{}(x){\displaystyle \frac{1}{6u_\mu ^2}}U_\mu ^{}(x\widehat{\mu })F_{\mu \nu }(x\widehat{\mu })U_\mu (x\widehat{\mu })`$ (29)
$`(\mu \nu ).`$
The various spatial lattice derivatives are defined as follows:
$`a_s\mathrm{\Delta }_iG(x)`$ $`=`$ $`{\displaystyle \frac{1}{2u_s}}[U_i(x)G(x+\widehat{ı})U_i^{}(x\widehat{ı})G(x\widehat{ı})],`$ (30)
$`a_s\mathrm{\Delta }_i^{(+)}G(x)`$ $`=`$ $`{\displaystyle \frac{U_i(x)}{u_s}}G(x+\widehat{ı})G(x),`$ (31)
$`a_s\mathrm{\Delta }_i^{()}G(x)`$ $`=`$ $`G(x){\displaystyle \frac{U_i^{}(x\widehat{ı})}{u_s}}G(x\widehat{ı}),`$ (32)
$`a_s^2\mathrm{\Delta }_i^{(2)}G(x)`$ $`=`$ $`{\displaystyle \frac{U_i(x)}{u_s}}G(x+\widehat{ı})2G(x)+{\displaystyle \frac{U_i^{}(x\widehat{ı})}{u_s}}G(x\widehat{ı}),`$ (33)
$`\stackrel{~}{\mathrm{\Delta }}_i`$ $`=`$ $`\mathrm{\Delta }_i{\displaystyle \frac{a_s^2}{6}}\mathrm{\Delta }_i^{(+)}\mathrm{\Delta }_i\mathrm{\Delta }_i^{()},`$ (34)
$`\mathrm{\Delta }^{(2)}`$ $`=`$ $`{\displaystyle \underset{i}{}}\mathrm{\Delta }_i^{(2)},`$ (35)
$`\stackrel{~}{\mathrm{\Delta }}^{(2)}`$ $`=`$ $`\mathrm{\Delta }^{(2)}{\displaystyle \frac{a_s^2}{12}}\mathrm{\Delta }^{(4)},`$ (36)
$`\mathrm{\Delta }^{(4)}`$ $`=`$ $`{\displaystyle \underset{i}{}}\left(\mathrm{\Delta }_i^{(2)}\right)^2.`$ (37)
This NRQCD action has quadratic classical lattice spacing errors.
## III METHOD
The data presented here come from 2000 gauge field configurations on 10$`{}_{}{}^{3}\times `$30 lattices at $`\beta =2.1`$ with a bare aspect ratio of $`\xi a_s/a_t=2`$. Two light quark masses are used, corresponding to $`\kappa =0.23`$ and 0.24. Fixed time boundaries are used for the light quark propagators so they fit naturally into a heavy-light meson, since the NRQCD heavy quark propagator is also not periodic in the temporal direction.
A calculation of the string tension from these gauge field configurations provides a determination of the renormalized anisotropy:
$$\xi a_s/a_t=1.96(2).$$
(38)
Using light quarks only, the lightest pseudoscalar and vector meson masses are easily obtained from local creation operators. By linear interpolation and extrapolation in $`1/\kappa `$, the critical ($`\kappa _c`$) and strange ($`\kappa _s`$) hopping parameters and the temporal lattice spacing are found to be
$`\kappa _c`$ $`=`$ $`0.243025(41),`$ (39)
$`\kappa _s`$ $`=`$ $`\{\begin{array}{cc}\text{0.2344(11)}\hfill & \text{from }m_\varphi \text{,}\hfill \\ \text{0.2356(3)}\hfill & \text{from }m_K\text{,}\hfill \end{array}`$ (42)
$`a_t`$ $`=`$ $`0.1075(23)\mathrm{fm}\mathrm{from}m_\rho .`$ (43)
For the hopping parameters used in explicit computations, $`\kappa =0.23`$ and 0.24, the ratio of pseudoscalar to vector meson masses is “$`m_\pi /m_\rho `$” = 0.815(3) and 0.517(8) respectively. No exceptional configurations were encountered at these $`\kappa `$ values. One might expect a systematic uncertainty on $`a_t`$ to account for deviations from the linear relationship between $`a_tm_\rho `$ and $`1/\kappa `$. The uncertainty is presumably a few percent, but cannot be estimated using only the two $`\kappa `$ values studied here.
A heavy-light meson is created by the following operator,
$$\underset{\stackrel{}{x}}{}Q^{}(\stackrel{}{x})\mathrm{\Omega }(\stackrel{}{x})\mathrm{\Gamma }(\stackrel{}{x})q(\stackrel{}{x}),$$
(44)
where $`\mathrm{\Omega }(\stackrel{}{x})`$ is given in Table I and the smearing operator is
$$\mathrm{\Gamma }(\stackrel{}{x})=[1+c_s\mathrm{\Delta }^{(2)}(\stackrel{}{x})]^{n_s}.$$
(45)
All plots shown here use $`(c_s,n_s)=(0.15,10)`$ at the source and a local sink. The source is fixed at timestep 4, which is a distance $`3a_t`$ from the lattice boundary.
Because NRQCD is an expansion in the inverse bare heavy quark mass, all meson mass differences can be obtained from correlation functions at $`𝐩=\mathrm{𝟎}`$, but the absolute meson mass itself cannot be obtained directly. One way to determine the mass is to compute the change in energy when a meson is boosted,
$$E_𝐩E_0=\frac{𝐩^\mathrm{𝟐}}{2M_{\mathrm{kin}}}.$$
(46)
This defines the kinetic mass, $`M_{\mathrm{kin}}`$, which is interpreted as the meson’s physical mass. For the present work, $`E_𝐩`$ is computed only for the $`{}_{}{}^{1}S_{0}^{}`$ state, with $`𝐩=(0,0,2\pi /L_s)`$ where $`L_s=N_sa_s`$ is the spatial extent of the lattice in physical units. Solving for the kinetic mass gives
$$M_{\mathrm{kin}}=\frac{2\pi ^2}{N_s^2\xi ^2a_t[a_t(E_𝐩E_0)]}.$$
(47)
Some justification for Eq. (46) comes from consideration of the next correction term, giving
$$E_𝐩E_0=\frac{𝐩^\mathrm{𝟐}}{2M_{\mathrm{kin}}}\frac{𝐩^\mathrm{𝟒}}{8M_{\mathrm{kin}}^3}.$$
(48)
We have verified that the extra term shifts the kinetic mass by an amount which is smaller than the uncertainties for every value of $`M_{\mathrm{kin}}`$ reported in this work.
In the case of S-wave mesons, a plateau containing ample timesteps is clearly evident in all effective mass plots; some examples are shown in Fig. 1. A more detailed discussion of the P-wave plateaus is deferred to Sec. V. In this paper, the plateau region for each mass is defined by the maximum value of
$$Q\frac{\mathrm{\Gamma }(N/21,\chi ^2/2)}{\mathrm{\Gamma }(N/21,0)}$$
(49)
where
$$\mathrm{\Gamma }(a,x)=_x^{\mathrm{}}dtt^{a1}\mathrm{exp}(t),$$
(50)
and $`N`$ is the number of timesteps in the proposed plateau region. $`\chi ^2`$ is obtained from a single exponential fit to each meson correlation function. A correlated fit is done with the covariance matrix inverted by singular value decomposition. Statistical uncertainties are obtained from the analysis of 5000 bootstrap ensembles. All plateaus are ended at timestep 22(20) for simulations with $`\kappa =0.23(0.24)`$, except for an infinitely-heavy quark due to excessive noise at these larger times. The examples in Table II demonstrate the quality of the fits.
## IV S-WAVE SPECTRUM
Calculations were performed for $`a_sM`$ = 1.2, 1.5, 5, 6 and $`\mathrm{}`$, where $`M`$ represents the bare heavy quark mass in the NRQCD action. Figure 2 shows the simulation energy of the ground state as a function of $`a_sM`$ for $`\kappa =0.24`$. The huge $`O(1/M^3)`$ effect at smaller $`M`$ values is due to the vacuum expectation value of the $`c_{10}`$ term. This large correction to the unphysical simulation energy does not discredit the convergence of NRQCD, but special care must be taken to ensure that the large vacuum value is incorporated into the heavy quark propagation appropriately. In particular, previous work has shown the linear approximation to be insufficient for the $`c_{10}`$ term, which contains the vacuum expectation value, in computations of the S-wave spin splitting in the charm region. Fig. 2 explicitly shows the error introduced by placing the $`c_{10}`$ term in $`H_B`$ rather than $`H_A`$ in Eq. (18).
It will also be noted from Fig. 2 that the simulation energy is negative for $`a_sM=1.2`$ at third order in NRQCD. Of course the absolute energy scale is unphysical in NRQCD due to omission of the large heavy quark mass term from the leading order action, and physical quantities (i.e. mass differences) are independent of the absolute energy scale. If the simulation energy were large and negative, it might signal a poor $`1/M`$ expansion and/or a problem for the discretization of heavy quark propagation, but a small negative result presents no problem.
Table III shows a calculation of the kinetic mass, and indicates that the bottom quark requires $`a_sM5.5`$. The charm quark seems to want $`1.2<a_sM<1.5`$, although the data suggest that the $`1/M`$ expansion might not be converging in this region. It is interesting to notice that the vacuum expectation value does not affect the calculation of this observable significantly.
A stronger statement about convergence comes from the splitting between the spin-singlet and spin-triplet S-wave states, since the uncertainties are smaller. Table IV suggests a nice convergence in the bottom region, but no guarantee of convergence for charm. It will be noted that an incorrect treatment of the vacuum expectation value (i.e. putting the $`c_{10}`$ term into $`H_B`$) can actually lead to a small $`O(1/M^3)`$ contribution, but this is incidental.
Tables III and IV correctly accommodate the vacuum expectation value of the $`c_{10}`$ term by placing it in $`H_A`$. In Ref. , this method was found to give the same numerical results, within statistical uncertainties, for the S-wave kinetic mass and spin splitting as was obtained by computing the vacuum expectation value directly and subtracting it from the Hamiltonian. For the present calculation, a similar check was performed: the vacuum expectation value was computed from 400 of the gauge field configurations, and the S-wave kinetic mass and spin splitting were computed from 200 configurations using the Hamiltonian with the vacuum value explicitly removed. As expected, the results agree within statistics with Tables III and IV when the $`c_{10}`$ term is in $`H_A`$.
Interpolating these data so that $`M_{\mathrm{kin}}`$ is the physical mass of a bottom meson, one finds a spin splitting which is only about 55% of the experimental value. This is typical of quenched lattice calculations (see for eg. Refs. ). Even an unquenched NRQCD calculation did not reproduce the experimental $`B^{}B`$ splitting so perhaps the tadpole-improved classical values, which were used for the coefficients $`c_i`$ in the NRQCD Hamiltonian, account for the residual discrepancy.
The data reported in Ref. display a $`1/M`$ expansion for charmed mesons in lattice NRQCD in which $`O(1/M^3)`$ terms were as significant as $`O(1/M^2)`$ terms, so convergence of the expansion could not be assured. In that work, it was hoped that a more convergent expansion might be obtained via changes in the lattice method. In particular, replacement of the average plaquette tadpole factor by the mean link in Landau gauge was suggested to hold some promise. The present work has made this modification plus a number of others including: a more aggressively-improved light quark action, asymmetric lattices with temporal spacing reduced by a factor of two, Dirichlet temporal boundaries for light quarks rather than periodic ones, smeared meson sources rather than local sources, and a symmetric dependence on $`H_B`$ in Eq. (18). Despite these changes, convergence of the $`1/M`$ expansion remains uncomfirmed for S-wave charmed mesons. It is possible that a study of terms beyond $`O(1/M^3)`$ would reveal that the series really is well-behaved, but this remains unexplored.
## V P-WAVE SPECTRUM
Each of the four P-wave operators from Table I leads to a visually-identifiable plateau; an example is shown in Fig. 3. The method of maximum $`Q`$, discussed in Sec. III, can be used to define precise plateau boundaries and the resultant $`{}_{}{}^{3}P_{0}^{}{}_{}{}^{1}S_{0}^{}`$ splitting is shown in Table V. In contrast to the S-wave splitting discussed in the previous section, the $`1/M`$ corrections to the $`{}_{}{}^{3}P_{0}^{}{}_{}{}^{1}S_{0}^{}`$ splitting are not large relative to the statistical uncertainties, even in the charm region. The splittings given in Table V are consistent with the available experimental data, as will be discussed below.
Tables VI and VII contain lattice results for splittings which involve the other P-wave mesons. In the charm region, the $`\kappa =0.23`$ data produce a $`{}_{}{}^{3}P_{2}^{}`$ meson which is heavier than the $`{}_{}{}^{3}P_{0}^{}`$ meson, but at $`\kappa =0.24`$ the $`{}_{}{}^{3}P_{2}^{}{}_{}{}^{3}P_{0}^{}`$ splitting is consistent with zero. The magnitudes of the splittings decrease in the bottom region, as expected.
A comparison of Tables V, VI and VII plus the effective mass plots in Fig. 4 provide some indication of the systematic uncertainty which arises from the choice of plateau region. In particular, it will be noted that the effective mass plots are monotonically decreasing near the source, so if a plateau region is chosen too near the source, it will produce a mass splitting which is too large. Thus one arrives at an upper bound for the $`{}_{}{}^{3}P_{2}^{}{}_{}{}^{3}P_{0}^{}`$ splitting, as presented in Ref. .
The correlation functions constructed using $`{}_{}{}^{1}P_{1}^{}`$ and $`{}_{}{}^{3}P_{1}^{}`$ operators contain some combination of the physical $`J=1`$ mesons. Both calculations should lead to the same (lighter) physical mass at large Euclidean times if both operators have a substantial overlap with the less massive $`J=1`$ state. In principle, the masses of the physical states can be obtained using the $`{}_{}{}^{1}P_{1}^{}`$/$`{}_{}{}^{3}P_{1}^{}`$ operator basis by also calculating the mixing matrix elements, but the effect was too small to be observed in these data. In practice, for Euclidean times which can be used in our lattice simulation, no significant energy difference is observed between the $`{}_{}{}^{1}P_{1}^{}`$ and $`{}_{}{}^{3}P_{1}^{}`$ channels.
Predictions for the physical mesons are displayed alongside experimental data in Table VIII. The $`D_sD`$ and $`B_sB`$ mass differences depend only mildly on $`O(1/M^2)`$ and $`O(1/M^3)`$ terms, and they are close to experiment. In contrast, the S-wave spin splittings are significantly smaller than experiment (as is typical of quenched lattice QCD). Furthermore, the $`O(1/M^2)`$ and $`O(1/M^3)`$ corrections to $`D^{}D`$ are each about 20% of the leading order result. The results in Table VIII use the value of $`\kappa _s`$ determined from $`m_K`$ in Eq. (42). Use of the other determination in Eq. (42) shifts the central values of the $`D_s^{}D_s`$ and $`B_s^{}B_s`$ mass splittings by 1 MeV or less, and shifts the splittings among P-waves ($`D_s^{}`$ and $`B_s^{}`$) by 7 MeV or less.
A detailed comparison of P-wave results would be somewhat premature, since the experimental data are rather incomplete and often rely on theoretical models for input, while the lattice calculation is quenched and lacks a firm connection between the physical mesons and the $`J=1`$ operators. Nevertheless, Table VIII shows a general consistency between the experimental and computed P-wave masses.
It is instructive to compare lattice P-wave masses to the predictions of models, such as quark models, a Bethe-Salpeter study, a chromodynamic potential model, a bag model and a Blankenbecler-Sugar approach. Many of these present results for $`J=1`$ directly in the $`{}_{}{}^{1}P_{1}^{}/{}_{}{}^{3}P_{1}^{}`$ basis which simplifies the comparison to lattice data, and the quark models may in general be more closely related to the quenched approximation than to experiment.
The models of Refs. predict the traditional hydrogen-like ordering of P-waves, where the $`{}_{}{}^{3}P_{0}^{}`$ is the lightest meson, the $`{}_{}{}^{3}P_{2}^{}`$ is the heaviest, and the $`P_1`$ states lie in between. The authors of Ref. find, from lightest to heaviest, $`{}_{}{}^{3}P_{0}^{}`$, $`P_1(3/2)`$, $`{}_{}{}^{3}P_{2}^{}`$, $`P_1(1/2)`$, where the arguments represent the angular momentum of the light degrees of freedom. Ref. predicts a dramatic inversion where the $`{}_{}{}^{3}P_{0}^{}`$ is heavier than the $`{}_{}{}^{3}P_{2}^{}`$ by 100(150) MeV for $`D^{}`$($`B^{}`$) mesons, but the $`P_1`$ states are the absolute lightest and heaviest. Ref. claims very small splittings (tens of MeV) where the $`P(1/2)`$ and $`P(3/2)`$ doublets overlap to produce different orderings for the $`D`$, $`D_s`$ and $`B`$ systems (the $`B_s`$ is ordered like the $`D`$).
Fig. 5 shows the $`D_2^{}D_0^{}`$ and $`B_2^{}B_0^{}`$ splittings from lattice QCD and from the models just mentioned. Despite the range of model predictions, it is evident that some general consistency exists between our results and a number of the models. Notice in particular that our results are numerically distinct from the large inversion of Ref. .
Fig. 6 compares the $`{}_{}{}^{3}P_{0}^{}{}_{}{}^{1}S_{0}^{}`$ splittings as obtained from lattice QCD and the models.
Of special importance is the comparison with Ref. , where the spectrum was also computed from quenched lattice NRQCD. The discrepancy between the two lattice calculations is exemplified by Fig. 5. While there are many differences in method between the two computations, it is difficult to identify a compelling reason for the disagreement. Fig. 4 indicates that our data must satisfy $`M(B_2^{})M(B_0^{})<100`$ MeV for any chosen plateau region, and this is not consistent with Ref. . It is hoped that further lattice efforts will improve this situation. Unfortunately, a recent lattice NRQCD study of the heavy-light meson spectrum has statistical uncertainties which are too large to resolve the discrepancy between our results and those of Ref. .
## VI CONCLUSIONS
The masses of S and P-wave heavy-light mesons have been calculated in the quenched approximation, using lattice NRQCD for the heavy quark and a highly-improved action for the light degrees of freedom. Calculations at first, second and third order in the heavy quark mass expansion were used as a test of convergence, and it was concluded that the $`O(1/M^2)`$ and $`O(1/M^3)`$ contributions to the $`D^{}D`$ splitting are both near 20%. The $`D_s^{}D_s`$ splitting receives a 20% correction at $`O(1/M^2)`$ and a 10% correction at $`O(1/M^3)`$. These results might represent a convergent $`1/M`$ expansion if terms beyond $`O(1/M^3)`$ are decreasing appropriately, but convergence cannot be ascertained from the present study. This same conclusion was reached in Ref. by a computational method which differed from the present one in some details; most notably, Ref. used the average plaquette tadpole definition whereas the present work uses the mean link in Landau gauge.
No convergence problem is found for P-wave charmed masses, and the P-wave spectrum for both charmed and bottom mesons is predicted. The $`{}_{}{}^{3}P_{2}^{}`$ is heavier than the $`{}_{}{}^{3}P_{0}^{}`$ and $`P_1`$ states in the $`D_s^{}`$ system, with $`M(D_{s2}^{})M(D_{s0}^{})=55\pm 17`$ MeV. For the $`D^{}`$, $`B^{}`$ and $`B_s^{}`$ systems the $`P_2P_0`$ splittings are smaller and consistent with zero. These conclusions agree with a number of model calculations and are compatible with the available experimental data.
###### Acknowledgements.
The authors thank Howard Trottier for a critical reading of the manuscript, and R.L. thanks Niranjan Venugopal for useful discussions. This work was supported in part by the Natural Sciences and Engineering Research Council of Canada.
|
warning/0003/hep-ph0003064.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
The production of $`10^710^8`$ top quark pairs per year at the CERN Large Hadron Collider (LHC) will allow to probe the top couplings to both known and new particles involved in possible top decay channels different from the main $`tbW`$. Thanks to the large top mass, there are several decays that can be considered, even involving the presence of on-shell heavy vector bosons or heavy new particles in the final states. On a purely statistical basis, one should be able to detect a particular decay channel whenever its branching ratio (B) is larger than about $`10^610^7`$. In practice, background problems and systematics will lower this potential by a few orders of magnitude, the precise reduction being dependent of course from the particular signature considered. We will see, that the final detection threshold for each channel will not allow the study of many possible final states predicted in the standard model (SM), unless new stronger couplings come into play.
After discussing in section 2 the main top quark rare decay channels predicted by the SM, we review in section 3 the expectations of a large class of possible extensions of the SM, such as the inclusion of a 4<sup>th</sup> fermion family, the two-Higgs-doublet models, and the minimal supersymmetric standard model (MSSM). Finally, results of the analysis of the LHC potential in the top rare decays field are reported in section 4.
## 2 Standard Model Decays
In this section, we give an overview of the decay channels of the top quark in the framework of the SM. In the SM the decay $`tbW`$ is by far the dominant one. The top total width is then given by:
$$\mathrm{\Gamma }_T\mathrm{\Gamma }(tbW)0.17\mathrm{GeV}|V_{tb}|^2\frac{m_t^3}{M_W^3}1.55\mathrm{GeV}.$$
The rates for other decay channels are predicted to be smaller by a few orders of magnitude in the SM. The second most-likely decays are the Cabibbo-Kobajashi-Maskawa (CKM) non-diagonal decays $`tsW`$and $`tdW`$. Assuming $`|V_{ts}|0.04`$ and $`|V_{td}|0.01`$, respectively , one gets;
$$B(tsW)1.6\times 10^3\mathrm{and}B(tdW)1\times 10^4,$$
(1)
in the SM with three families. From now on, $`B(t\mathrm{})`$ for a generic decay channel, stands for the quantity:
$$B(t\mathrm{})=\frac{\mathrm{\Gamma }(t\mathrm{})}{\mathrm{\Gamma }(tbW)}.$$
(2)
The two-body tree-level decay channels are the only ones that LHC could be able to detect in the framework of the SM. The next less-rare processes indeed turn out to have rates not larger than $`10^6`$.
At tree level, the decay $`tbWZ`$(fig. 1) has some peculiar features, since the process occurs near the kinematical threshold ($`m_tM_W+M_Z+m_b`$). This fact makes crucial the $`W`$ and $`Z`$ finite-width effects in the theoretical prediction of the corresponding width. In (see also ), an estimate of $`B(tbWZ)`$ was given on the basis of the definition:
$$\mathrm{\Gamma }(tbWZ)\frac{\mathrm{\Gamma }(tb\mu \nu _\mu ee)}{B(W\mu \nu _\mu )B(Zee)}$$
(3)
involving experimentally well-observable decays. This definition involves 14 diagrams contributing to the $`b\mu \nu _\mu ee`$ final state in the SM. The estimate for the corresponding branching ratio is:
$$B_{cut}(tbWZ)6\times 10^7,$$
(4)
for $`m_t=175`$ GeV, assuming a minimum cut of $`0.8M_Z`$ on the $`ee`$-pair invariant mass. This cut tries to cope with the contribution of background graphs where the $`ee`$ pair rises not from a $`Z`$ boson but from a photon. Recently, in , it has been argued that $`B(tbWZ)`$ should be defined by including in the definition in eq. (3) only the three (gauge-invariant) diagrams that correspond to the process where the final state $`b\mu \nu _\mu ee`$ is truly mediated by a $`W`$ and a $`Z`$. Then one obtains, for $`m_t=175`$ GeV:
$$B(tbWZ)=B_{res}(tbWZ)=2\times 10^6.$$
(5)
Such an increase in $`B(tbWZ)`$ is partly due to the negative interference effects with the graphs where a photon replaces the $`Z`$ boson (which are present in the previous estimate of the quantity), and partly to the absence of any kinematical cut (which for the resonant graphs is not needed).
One could also find less ambiguous definitions for $`B(tbWZ)`$ than the one involved in eq. (3). For instance, the new definition:
$$\stackrel{~}{\mathrm{\Gamma }}(tbWZ)\frac{\mathrm{\Gamma }(tb\mu \nu _\mu \nu _e\nu _e)}{B(W\mu \nu _\mu )B(Z\nu _e\nu _e)}$$
(6)
would not require any kinematical cut, would involve a negligible background, and would give for $`B(tbWZ)`$ the same result as in eq.(5). Of course, the signature $`b\mu \nu _\mu \nu _e\nu _e`$ would not be practical from an experimental point of view, also because of the tiny rates involved in this decay channel.
Then, the authors in agree that an experimental measurement of the decay rate of $`tbWZ`$should go through the definition in eq. (3), where the final state $`b\mu \nu _\mu ee`$ includes all the possible backgrounds. They confirm the result in eq. (4) as an experimental effective quantity.
If the Higgs boson is light enough, one could also have the decay $`tbWH`$(fig. 1), although the present limits on $`m_H`$ strongly suppress its rate. For $`m_H>\mathrm{\hspace{0.33em}100}`$ GeV, one gets (see also ):
$$B(tbWH)<\mathrm{\hspace{0.33em}7}\times 10^8.$$
(7)
Finally, the decay $`tcWW`$is very much suppressed by a GIM factor $`\frac{m_b^2}{M_W^2}`$ in the amplitude. One than gets :
$$B(tcWW)10^{13}.$$
(8)
One can also consider the radiative three-body decays $`tbWg`$and $`tbW\gamma `$. These channels suffer from infrared divergences, and the evaluation of their rate requires a full detector simulation, including for instance the effects of the detector resolution and the jet isolation algorithm. In an idealized situation where the rate is computed in the $`t`$ rest frame with a minimum cut of 10 GeV on the gluon or photon energies, one finds (see also ):
$`B(tbWg)`$ $``$ $`0.3,`$ (9)
$`B(tbW\gamma )`$ $``$ $`3.5\times 10^3.`$ (10)
The FCNC decays $`tcg`$, $`tc\gamma `$and $`tcZ`$occur at one loop, and are also GIM suppressed by a factor $`\frac{m_b^2}{M_W^2}`$ in the amplitude. Hence, the corresponding rates are very small :
$`B(tcg)`$ $``$ $`5\times 10^{11}`$ (11)
$`B(tc\gamma )`$ $``$ $`5\times 10^{13}`$ (12)
$`B(tcZ)`$ $``$ $`1.3\times 10^{13}`$ (13)
For a light Higgs boson, one can consider also the FCNC decay $`tcH`$(fig. 2). A previous evaluation of its rates has now been corrected. For $`m_H100(160)`$ GeV, one gets then :
$$B(tcH)0.9\times 10^{13}(4\times 10^{15}).$$
(14)
We conclude this section, by presenting in table 1 a summary of the expected decay rates for the main top decay channels in the SM.
## 3 Beyond the Standard Model (BSM) decays
The fact that a measurement of the top width is not available and the branching ratio $`B(tbW)`$ is a model dependent quantity makes the present experimental constraints on the top BSM decays quite weak. Hence, the large value of $`m_t`$ allows to consider the possibility of $`t`$ decays into new massive states with branching fraction of order $`B(tbW)`$. Apart from the production of new final states with large branching fractions, we will see that new physics could also give rise to a considerable increase in the rates of many decay channels that in the SM framework are under the threshold of observability.
### 3.1 $`4^{th}`$ fermion family
Extending the SM with a $`4^{th}`$ fermion family can alter considerably a few $`t`$ decay channels. First of all, when adding a $`4^{th}`$ family to the CKM matrix the present constraints on the $`|V_{tq}|`$ elements are considerably relaxed. In particular, $`|V_{ts}|`$ and $`|V_{td}|`$ can grow up to about 0.5 and 0.1, respectively . Correspondingly, assuming $`|V_{tb}|1`$ for the sake of normalization, one can have up to :
$$B_4(tsW)0.25\mathrm{and}B_4(tdW)0.01,$$
(15)
to be confronted with the SM expectations in eq. (1).
The presence of a $`4^{th}`$ fermion family could also show up in the $`t`$ direct decay into a heavy $`b^{}`$ quark with a relatively small mass ($`m_b^{}100`$ GeV) . This decay would contribute to the $`tcWW`$channel. The corresponding rate would be:
$`\mathrm{B}(tW^+b^{}(W^{}c))10^3(10^7)\mathrm{at}m_b^{}=100(300)\mathrm{GeV},`$ (16)
to be confronted with the SM prediction in eq. (8).
### 3.2 Two Higgs Doublet models (2HDM’s)
The possibility that the electroweak symmetry breaking involves more than one Higgs doublet is well motivated theoretically. In particular, three classes of two Higgs doublet models have been examined in connection with rare top decays, called model I, II and III. The first two are characterized by an ad hoc discrete symmetry which forbid tree-level FCNC , that are strongly constrained in the lightest quarks sector. In model I and model II, the up-type quarks and down-type quarks couple to the same scalar doublet and to two different doublets, respectively (the Higgs sector of the MSSM is an example of model II). In model III , the above discrete symmetry is dropped and tree-level FCNC are allowed. In particular, a tree-level coupling $`tcH`$ is predicted with a coupling constant $`<\frac{\sqrt{m_tm_c}}{v}`$ (where $`v`$ is the Higgs vacuum expectation value).
Since enlarging the Higgs sector automatically implies the presence of charged Higgs bosons in the spectrum, one major prediction of these new frameworks is the decay $`tbH^+`$, with possibly competitive rates with $`B(tbW)`$ for $`m_{H^+}<\mathrm{\hspace{0.33em}170}`$ GeV. In the MSSM, one has $`B(tbH^+)1`$, both at small and large values of $`\mathrm{tan}\beta `$. If $`m_H<m_tm_b`$, one expects $`H^+\tau ^+\nu `$ (favored for large $`\mathrm{tan}\beta `$) and/or $`H^+c\overline{s}`$ (favored for small $`\mathrm{tan}\beta `$) to be the dominant decays. Hence, for $`\mathrm{tan}\beta >1`$ and $`H^+\tau ^+\nu `$ dominant, one can look for the channel $`tbH^+`$ by studying a possible excess in the $`\tau `$ lepton signature from the $`t`$ pair production . On the other hand, if $`\mathrm{tan}\beta <2`$ and $`m_H>130`$ GeV, the large mass (or coupling) of the $`t`$-quark causes $`\mathrm{B}(H^+t^{}\overline{b}W^+b\overline{b})`$ to exceed $`\mathrm{B}(H^+c\overline{s})`$ (see for details). As a consequence, new interesting signatures at LHC like leptons plus multi-jet channels with four $`b`$-tags, coming from the gluon-gluon fusion process $`ggt\overline{b}H^{}`$, followed by the $`H^{}\overline{t}b`$ decay, have been studied . These processes could provide a viable signature over a limited but interesting range of the parameter space.
One should recall however that both $`B(tbH^+)`$ and $`B(H^+W^+b\overline{b})`$ are very sensitive to higher-order corrections, which are highly model dependent .
In model III, the tree-level FCNC decay $`tch`$ can occur with branching ratios up to 10<sup>-2</sup> . In , the rate for the channel $`tchcWW(cZZ,c\gamma \gamma )`$ has been studied. Accordingly, $`B(tcWW)`$ can be enhanced by several orders of magnitude with respect to its SM value. In particular, for an on-shell decay with $`2M_W<m_h<m_t`$, one can have up to $`B(tcWW)10^4`$ from this source. The same process was considered in a wider range of models, where the decay $`tcWW`$can occur not only through a scalar exchange but also through a fermion or vector exchange . In this framework, the fermion exchange too could lead to detectable rates for $`tcWW`$, as in eq. (16).
In 2HDM’s, the prediction for the FCNC decays $`tcg`$, $`tc\gamma `$and $`tcZ`$can also be altered. While in models I and II the corresponding branching fractions can not anyway approach the detectability threshold , in model III values up to $`B(tcg)10^5`$, $`B(tc\gamma )10^7`$ and $`B(tcZ)10^6`$ are predicted .
By further extending the 2HDM’s sector and including Higgs triplets, one can give rise to a vertex $`HWZ`$ at tree level in a consistent way . Accordingly, the $`tbWZ`$decay can be mediated by a charged Higgs (coupled with $`m_t`$) that can enhance the corresponding branching fraction up to $`B(tbWZ)10^2`$. Large enhancements can also be expected in similar models for the channels $`tsWZ`$and $`tdWZ`$.
### 3.3 Minimal Supersymmetric Standard Model (MSSM)
Supersymmetry could affect the $`t`$ decays in different ways \[here, we assume the MSSM framework , with (or without, when specified) $`R`$ parity conservation\].
First of all, two-body decays into squarks and gauginos, such as $`t\stackrel{~}{t}_1\stackrel{~}{g}`$, $`t\stackrel{~}{b}_1\stackrel{~}{\chi }_1^+`$, $`t\stackrel{~}{t}_1\stackrel{~}{\chi }_1^0`$, could have branching ratios of order $`B(tbW)`$, if allowed by the phase space (see, i.e., for references). QCD corrections to the channel $`t\stackrel{~}{t}_1\stackrel{~}{g}`$ have been computed in and found to increase its width up to values even larger than $`\mathrm{\Gamma }(tbW)`$. Three-body $`t`$ decays in supersymmetric particles were surveyed in .
The presence of light top and bottom squarks, charginos and neutralinos in the MSSM spectrum could also give rise to a $`CP`$ asymmetry of the order $`10^3`$ in the partial widths for the decays $`tbW^+`$ and $`\overline{t}\overline{b}W^{}`$ .
Explicit $`R`$-parity violating interactions could provide new flavor-changing $`t`$ decays, both at tree-level (as in the channels $`t\stackrel{~}{\tau }b`$ and $`t\tau b\stackrel{~}{\chi }_1^0`$ ) and at one loop (as in $`tc\stackrel{~}{\nu }`$ ), with observable rates. For instance, $`B(tc\stackrel{~}{\nu })10^410^3`$ in particularly favorable cases.
Another sector where supersymmetric particles could produce crucial changes concerns the one-loop FCNC decays $`tcg`$, $`tc\gamma `$, $`tcZ`$and $`tcH`$, which in the SM are unobservally small. In the MSSM with universal soft breaking the situation is not much affected, while, by relaxing the universality with a large flavor mixing between the 2<sup>nd</sup> and 3<sup>rd</sup> family only, one can reach values such as :
$`B_{MSSM}(tcg)`$ $``$ $`10^6`$ (17)
$`B_{MSSM}(tc\gamma )`$ $``$ $`10^8`$ (18)
$`B_{MSSM}(tcZ)`$ $``$ $`10^8,`$ (19)
which anyhow are still not observable. The introduction of $`B`$-violating couplings in broken $`R`$-parity models could on the other hand give large enhancements , and make some of these channels observable. The corresponding upper limits on branching ratios get then:
$`B_{R/}(tcg)`$ $``$ $`10^3`$ (20)
$`B_{R/}(tc\gamma )`$ $``$ $`10^5`$ (21)
$`B_{R/}(tcZ)`$ $``$ $`10^4.`$ (22)
A particularly promising channel is the FCNC decay $`tch`$ in the framework of MSSM, where $`h=h^0,H^0,A^0`$ is any of the supersymmetric neutral Higgs bosons . By including the leading MSSM contributions to these decays (including gluino-mediated FCNC couplings), one could approach the detectability threshold, especially in the case of the light CP-even Higgs boson, for which one can get up to:
$$B_{MSSM}(tch^0)10^4.$$
(23)
## 4 LHC potential for top rare decays
An extensive analysis of the LHC potential for detecting top rare decays has been performed by both the LHC experiments ATLAS and CMS . In the framework of the SM, the top rare decays (that is any channel different from $`tqW`$) turn out to be definitely below the threshold for an experimental analysis at LHC. On the other hand, LHC experiments will be able to probe quite a few predictions of possible extensions of the SM.
An extended Higgs sector will be looked for through the tree-level decay $`tbH^+`$. ATLAS estimates its sensitivity to this channel in the MSSM, through an excess in the tau lepton signal, to be around B$`(tH^+b)`$ = 3% (that is almost 4 times better than what expected from Run 2 at the Tevatron). This would allow to probe all values of $`m_{H^\pm }`$ below $`m_t20`$ GeV over most of the $`\mathrm{tan}\beta `$ range. For low $`\mathrm{tan}\beta `$, the complementary decay mode $`H^\pm cs`$ has been considered. In the mass range 110 $`<H^\pm <`$ 130 GeV, the $`H^\pm `$ peak can be reconstructed and separated from the dominant $`Wjj`$ background.
For CMS, using the tau excess signature, the expected 5$`\sigma `$ discovery range for 10 fb<sup>-1</sup> in the MSSM ($`m_A,\mathrm{tan}\beta `$) parameter space is $`m_A<`$ 110 GeV, for all $`\mathrm{tan}\beta `$ values, and somewhat extended ($`m_A<`$ 140), for $`\mathrm{tan}\beta <`$ 2.
Other interesting signatures like $`H^\pm hW^{}`$, $`H^\pm AW^{}`$ and $`H^\pm bt^{}bbW`$ are very promising in particular parameter ranges, but have not yet been thoroughly investigated.
ATLAS has studied its sensitivity to the radiative decay $`tWbZ`$. This has been found to be at most of the order 10<sup>-4</sup>, that is insufficient for the study of a SM signal ($`10^6`$), but possibly useful for exploring the predictions of some extended Higgs-sector model, for which B($`tWqZ)<\mathrm{\hspace{0.33em}10}^2`$. On the other hand, the radiative Higgs decay $`tWbH`$ seems out of the reach of LHC in any realistic model.
The LHC reach for the FCNC decays $`tqZ`$, $`tq\gamma `$ and $`tqg`$ has also been thoroughly investigated. Apart from the $`tqg`$, which is completely overwhelmed by the hadronic background, both ATLAS and CMS have a sensitivity of about $`2\times 10^4`$ to the $`tqZ`$ channel, while the CMS reach for the $`tq\gamma `$ channel is about $`3.4\times 10^5`$, that is slightly better than the ATLAS sensitivity ($`1.0\times 10^4`$), assuming an integrated luminosity of 100 fb<sup>-1</sup>. These thresholds are too high to test the predictions of the models reviewed here. On the other hand, they could be largely sufficient to detect some manifestation of possible FCNC anomalous couplings in the top sector .
|
warning/0003/physics0003071.html
|
ar5iv
|
text
|
# 0.0.1 References
## 0.0.1 References
> Aksnes G., Asaad A.N. Influence of the water structure on chemical reactions in water. A study of proton-catalyzed acetal hydrolysis. Acta Chem. Scand. $`1989,43,726734`$.
>
> Aksnes G., Libnau O. Temperature dependence of esther hydrolysis in water. Acta Chem.Scand. $`1991,45,463467`$.
>
> Aliotta F., Fontanella M.E., Magazu S. Sound propagation in thyxotropic strucures. Phys. Chem. Liq. 1990,
>
> Benassi P., D’Arrigo G., Nardone M. Brilouin light scattering in low temperature water-ethanol solutions. J.Chem.Phys. $`1988,89,44694477`$.
>
> Bertolini D., Cassetari M., Grigolini P., Salvetti G. and Tani A. The mesoscopic systems of water and other complex systems. J.Mol.Liquids, $`1989,\mathrm{\hspace{0.33em}41},\mathrm{\hspace{0.17em}251}`$.
>
> Bertolini D., Cassetari M., Salvetti G., Tombari E., Veronesi S., Squadrito G. Il nuovo Cim. $`1992,\mathrm{\hspace{0.17em}14}D,\mathrm{\hspace{0.33em}199}`$.
>
> Cantor C.R., Schimmel P.R. Biophysical Chemistry. W.H.Freemen and Company, San Francisco, 1980.
>
> Clegg J. S. On the physical properties and potential roles of intracellular water. Proc.NATO Adv.Res.Work Shop. 1985.
>
> Clegg J.S. and Drost-Hansen W. On the biochemistry and cell physiology of water. In: Hochachka and Mommsen (eds.). Biochemistry and molecular biology of fishes. Elsevier Science Publ. vol.1, Ch.1, pp.1-23, 1991.
>
> D’Aprano A., Donato Ines., Liveri V.T. Molecular association of n- alcohols in nonpolar solvents. Excess volumes and viscosities of n- pentanol+n-octane mixtures at 0, 5, 25, 35 and $`45^0`$C. J.Solut.Chem. $`1990a,\mathrm{𝟏𝟗},711720`$.
>
> D’.Aprano A., Donato I., Liveri V.T. Molecular interactions in 1- pentanol + 2-butanol mixtures: static dielectric constant, viscosity and refractive index investigations at 5, 25, and $`45^0`$C. J.Solut.Chem. $`1990b,18,785793`$.
>
> D’Aprano A. and Donato I. Dielectric polarization and Polarizability of 1-pentanol + n-octane mixtures from static dielectric constant and refractive index data at 0, 25 and $`45^0`$. J.Solut.Chem. $`1990c,\mathrm{\hspace{0.17em}19},\mathrm{\hspace{0.17em}883}892`$.
>
> D’Arrigo G., Paparelli A. Sound propagation in water-ethanol mixtures at low temperatures. I.Ultrasonic velocity. J.Chem.Phys. $`1988a,\mathrm{\hspace{0.17em}88}`$, No.$`1,405415`$.
>
> D’Arrigo G., Paparelli A. Sound propagation in water-ethanol mixtures at low temperatures. II.Dynamical properties. J.Chem.Phys. $`1988b,\mathrm{\hspace{0.17em}88}`$, No.$`12,76877697`$.
>
> D’Arrigo G., Paparelli A. Anomalous ultrasonic absorption in alkoxyethanls aqeous solutions near their critical and melting points. J.Chem.Phys. 1989, 91, No.$`4,\mathrm{\hspace{0.17em}2587}2593`$.
>
> D’Arrigo G., Texiera J. Small-angle neutron scattering study of $`D_2O`$-alcohol solutions. J.Chem.Faraday Trans. $`1990,\mathrm{\hspace{0.17em}86},\mathrm{\hspace{0.17em}1503}1509`$.
>
> Del Giudice E., Dogulia S., Milani M. and Vitello G. A quantum field theoretical approach to the collective behaviour of biological systems. Nuclear Physics $`1985,`$ $`B251[FS13],375400`$.
>
> Del Guidice E. Doglia S., Milani M. Spontaneous symmetry breaking and electromagnetic interactions in biological systems. Physica Scripta. $`1988,38,505507`$.
>
> Drost-Hansen W. In: Colloid and Interface Science. Ed. Kerker M. Academic Press, New York, 1976, p.267.
>
> Drost-Hansen W., Singleton J. Lin. Our aqueous heritage: evidence for vicinal water in cells. In: Fundamentals of Medical Cell Biology, v.3A, Chemisrty of the living cell, JAI Press Inc.,1992, p.157-180.
>
> Etzler F.M., Conners J.J. Structural transitions in vicinal water: pore size and temperature dependence of the heat capacity of water in small pores. Langmuir 1991, 7, 2293-2297.
>
> Etzler F.M., White P.J. The heat capacity of water in silica pores. J. Colloid and Interface Sci., 1987, 120, 94-99.
>
> Farsaci F., Fontanella M.E., Salvato G., Wanderlingh F. and Giordano R., Wanderlingh U. Dynamical behaviour of structured macromolecular solutions. Phys.Chem. Liq. $`1989,\mathrm{\hspace{0.17em}20},\mathrm{\hspace{0.17em}205}220`$.
>
> Fontaine A. et al., Phys Rev. Lett. $`1978,\mathbf{\hspace{0.17em}41},504`$.
>
> Fröhlich H. Phys.Lett. 51 (1975) 21.
>
> Fröhlich H. Proc. Nat. Acad. Sci. USA $`1975,\mathrm{\hspace{0.17em}72},\mathrm{\hspace{0.17em}4211}`$.
>
> Giordano R., Fontana M.P., Wanderlingh F. J.Chem.Phys. $`\mathrm{\hspace{0.17em}1981}a,\mathrm{\hspace{0.17em}74},\mathrm{\hspace{0.17em}2011}`$.
>
> Giordano R. et al. Phys.Rev. $`\mathrm{\hspace{0.17em}1983}b,A28,\mathrm{\hspace{0.17em}3581}`$.
>
> Giordano R., Salvato G., Wanderlingh F., Wanderlingh U. Quasielastic and inelastic neutron scattering in macromolecular solutions. Phys.Rev.A. $`\mathrm{\hspace{0.17em}\hspace{0.17em}1990},41,689696`$.
>
> Glansdorf P., Prigogine I. Thermodynamic theory of structure, stability and fluctuations. Wiley and Sons, N.Y., 1971.
>
> Gordeyev G.P., Khaidarov T. In: Water in biological systems and their components. Leningrad University Press, Leningrad, 1983, p.3 (in Russian).
>
> Haken H. Information and selforganization. Springer, Berlin, 1988.
>
> Haken H. Synergetics, computers and cognition. Springer, Berlin, 1990.
>
> Ise N. and Okubo T. Accounts of Chem. Res. $`1980,13,303`$.
>
> Ise N., Matsuoka H., Ito K., Yoshida H. Inhomogenity of solute distribution in ionic systems. Faraday Discuss. Chem. Soc. $`1990,\mathbf{\hspace{0.17em}90},\mathrm{\hspace{0.17em}153}162`$.
>
> Ito K., Yoshida H., Ise N. Void Structure in colloid Dispersion. Science, $`\mathrm{\hspace{0.17em}1994},\mathrm{\hspace{0.17em}263},\mathrm{\hspace{0.17em}66}`$.
>
> Käiväräinen A.I. Solvent-dependent flexibility of proteins and principles of their function. D.Reidel Publ.Co., Dordrecht, Boston, Lancaster, 1985, pp.290.
>
> Käiväräinen A.I. The noncontact interaction between macromolecules revealed by modified spin-label method. Biofizika (USSR$`)\mathrm{\hspace{0.33em}1987},\mathrm{\hspace{0.17em}32},\mathrm{\hspace{0.17em}536}`$.
>
> Käiväräinen A.I. Thermodynamic analysis of the system: water-ions-macromolecules. Biofizika (USSR$`),\mathrm{\hspace{0.17em}1988},\mathrm{\hspace{0.17em}33},\mathrm{\hspace{0.17em}549}`$.
>
> Käiväräinen A.I. Theory of condensed state as a hierarchical system of quasiparticles formed by phonons and three-dimensional de Broglie waves of molecules. Application of theory to thermodynamics of water and ice. J.Mol.Liq. $`1989a,\mathrm{\hspace{0.17em}41},\mathrm{\hspace{0.17em}53}60`$.
>
> Käiväräinen A.I. Mesoscopic theory of matter and its interaction with light. Principles of selforganization in ice, water and biosystems. University of Turku, Finland 1992, pp.275.
>
> Käiväräinen A., Fradkova L., Korpela T. Separate contributions of large- and small-scale dynamics to the heat capacity of proteins. A new viscosity approach. Acta Chem.Scand. $`1993,47,456460`$.
>
> Lagrage P., Fontaine A., Raoux D., Sadoc A., Miglardo P. J.Chem. Phys. $`\mathrm{\hspace{0.17em}1980},\mathbf{\hspace{0.17em}72},3061`$.
>
> Lumry R. and Gregory R.B. Free-energy managment in protein reactions: concepts, complications and compensations. In book: The fluctuating enzyme. A Wiley-Interscience publication. 1986, p. 341- 368.
>
> Nemethy G., Scheraga H.A. J.Chem.Phys. $`1962,\mathbf{\hspace{0.33em}36},\mathrm{\hspace{0.17em}3382}`$.
>
> Peschel G. Adlfinger K.H. J. Coll. Interface Sci., 1970, v.34 (4), p.505.
>
> Sadoc A., Fountaine A., Lagarde D., Raoux D. J.Am.Chem.Soc. $`1981,`$ $`\mathrm{𝟏𝟎𝟑},6287`$.
>
> Tereshkevitch M.O., Kuprin F.V., Kuratova T.S., Ivishina G.A. J. Phys. Chem. (USSR$`)\mathrm{\hspace{0.33em}1974},\mathrm{\hspace{0.17em}48},\mathrm{\hspace{0.17em}2498}`$.
>
> Watterson J. Solvent cluster size and colligative properties. Phys. Chem.Liq. $`1987,16,317320`$.
>
> Watterson J. The role of water in cell architecture. Mol.Cell.Biochem. $`1988,\mathrm{𝟕𝟗},101105`$.
>
> Yoshida H., Ito K., and Ise N. Colloidal crystal growth. J. Chem. Soc. Faraday Trans., $`1991,87(3),371378`$.
|
warning/0003/hep-th0003252.html
|
ar5iv
|
text
|
# Q-ball Dynamics
## 1 Introduction
One of the most fascinating areas of inter-disciplinary research at the interface between mathematics and physics is the study of solitons. This word has as many definitions as there are people who study them, but in general terms they are stable, localized energy distributions. From a purely mathematical perspective, solitons are described as extended solutions to a set of hyperbolic or parabolic partial differential equations, which can travel without dissipation at a uniform velocity and maintain, at least asymptotically, their shape during collisions; often these properties of solitons are attributable to the existence of an infinite number of conserved quantities, connected with the notions integrability, and radiation-free soliton collisions can be constructed.
In the context of particle physics, which is our main interest here, the usage of the word soliton is less rigorous and any kind of localized energy distribution falls under this broad umbrella. Only in very specialised circumstances will soliton collisions not generate radiation and solutions which radiate substantially during, for example, a highly relativistic 2-soliton collision are included. Of course very little exact analytic progress is possible in these more general settings since radiative processes are notoriously difficult to model analytically, thus numerical simulations are necessary to probe more complicated situations. The usual development of understanding in this subject follows an intricate interplay between the two, with analytic work putting on solid ground more qualitative observations from simulations. The domain of validity of analytic approximations can then be checked and extended by further simulations. Examples of the classes of solitons which have been investigated in this way are vortices , monopoles and skyrmions , whose existence and stability is essentially due to conserved topological currents and charges, along with energy bounds related to the charge and stable scaling laws. In these examples the topological features constrain the amount of radiation produced in low energy collisions and allows approximations to be applied which ignore the radiative effects.
The subject of this paper is a particular class of solitons known as Q-balls . These are different in many ways to the topological solitons mentioned above. Firstly, they are time-dependent with a rotating internal phase. Secondly, the conserved charge associated with their stability, Noether charge (Q), is not topological and therefore their stability is also a dynamical issue. As we will see these two features lead to a much more complicated variety of interaction properties than seen in the study of topological solitons. The main difference is that the charge is quantized in topological models, it usually being scaled to be an exact integer, whereas we shall see that the charge Q can take any value (in a specified range) allowing for the possibility of charge transfer between solitons and/or fission during the interaction process.
Although the concepts associated with Q-balls are extremely general and they are likely to exist in a wide variety of physical contexts (for example, see ref. ), the main motivation for the current study is the recent realization that they are a generic consequence of the Minimal Supersymmetric Standard Model (MSSM) due to the existence of D-flat directions in the effective potential created by tri-linear couplings. In this context the conserved charge is that associated with the U(1) symmetries of Baryon and Lepton number conservation and the relevant U(1) fields correspond to either squark or slepton particles. Therefore, the Q-balls can be thought of as condensates of either a large number of squarks or sleptons. It has been suggested that such condensates can be involved in baryogenesis via the Affleck-Dine mechanism after an epoch of inflation in the early universe. If this is the case then there are two interesting possibilities. If the Q-balls can avoid evaporation into lighter, stable particles such as protons, then it might be possible for them to be important cosmologically as cold dark matter . Whereas if they are unstable, they would decay in a non-trivial way into baryons and could lead to observable isocurvature baryon fluctuations .
Underpinning these interesting suggestions are assumptions as to how Q-balls actually interact and it is our intention here to make an exhaustive study of this issue. Our approach will be to identify numerically the important dynamical processes that can occur in general situations of two interacting Q-balls, which we will then explain qualitatively, leaving a more detailed analytic exposition of the dynamics and the cosmological implications to subsequent papers. In the next section we will discuss the basic properties of static U(1) Q-balls. Then we will present a detailed and extensive study of Q-balls on the line. As we will see, many of the properties of interest such as fission and charge transfer are observed in one-dimension and given the simplicity of simulations, it seems sensible to make the most exhaustive study there. In the subsequent sections on planar Q-balls and fully three-dimensional Q-balls we will show to what extent the one-dimensional simulations can be used to understand the dynamics in higher dimensions and what effects are clearly of higher dimensional origin. In a penultimate section we will discuss the interactions of Q-balls with anti-Q-balls which have an equal and opposite rotation, before making a concluding summary in the final section.
We should note that there is a disparate literature on Q-balls in which some (but by no means all) of the processes we will discuss have already been noted, but not necessarily completely understood. In particular we should mention recent work which presented results for Q-balls in one and two dimensions. There it was suggested that the right-angle scattering of solitons seen in two-dimensional topological soliton models is also prevalent in these non-topological models. At the relevant points we will point out in what ways we disagree with their explanation of this phenomena, and demonstrate that it is by no means general.
## 2 U(1) Q-balls
Given our motivation for studying Q-balls it seems sensible to work with the U(1) Goldstone model, although Q-balls can exist in a variety of field theoretic models. To be precise, the model we consider is that of a single complex scalar field $`\varphi `$ in $`D=1,2,3`$ spatial dimensions with a potential $`U(|\varphi |).`$ Explicitly, the Lagrangian is
$$=\frac{1}{2}_\mu \varphi ^\mu \overline{\varphi }U(|\varphi |),$$
(2.1)
with the key feature being the fact that the potential is only a function of $`|\varphi |`$. The model has a global $`U(1)`$ symmetry and the associated conserved Noether current $`J_\mu `$ is given by
$$J_\mu =\frac{1}{2i}\left(\overline{\varphi }_\mu \varphi \varphi _\mu \overline{\varphi }\right),$$
(2.2)
whose covariant conservation $`^\mu J_\mu =0`$ leads to the existence of the conserved Noether charge Q, given by
$$Q=\frac{1}{2i}\left(\overline{\varphi }_t\varphi \varphi _t\overline{\varphi }\right)d^Dx=\text{Im}(\overline{\varphi }_t\varphi )d^Dx.$$
(2.3)
A stationary Q-ball solution has the form
$$\varphi =e^{i\omega t}f(r),$$
(2.4)
where $`f(r)`$ is a real radial profile function which satisfies the ordinary differential equation
$$\frac{d^2f}{dr^2}=\frac{(1D)}{r}\frac{df}{dr}\omega ^2f+U^{}(f),$$
(2.5)
with the boundary conditions that $`f(\mathrm{})=0`$ and $`\frac{df}{dr}(0)=0.`$
This equation can either be interpreted as describing the motion of a point particle moving in a potential with friction , or in terms of Euclidean bounce solutions ; in each case the effective potential being $`U_{\mathrm{eff}}(f)=\omega ^2f^2/2U(f)`$. This leads to constraints on the potential $`U(f)`$ and the frequency $`\omega `$ in order for a Q-ball solution to exist. Firstly, the effective mass of $`f`$ must be negative. If we consider a potential $`U(f)`$ which is non-negative and satisfies $`U(0)=U^{}(0)=0`$, $`U^{\prime \prime }(0)=\omega _+^2>0`$, then one can deduce that $`\omega <\omega _+`$. Furthermore, the minimum of $`U(f)/f^2`$ must be attained at some positive value of $`f`$, say $`0<f_0<\mathrm{},`$ and existence of the solution requires that $`\omega >\omega _{}`$ where
$$\omega _{}^2=2U(f_0)/f_0^2.$$
(2.6)
Hence, Q-ball solutions exist for all $`\omega `$ in the range $`\omega _{}<|\omega |<\omega _+`$. Note (i) that solutions exist for positive and negative values of $`\omega `$, the negative ones being termed anti-Q-balls, (ii) it is often interesting to think of the Q-balls as being akin to charged bubbles; their profiles being very similar.
The classical stability of the solutions is a more sensitive issue. For sufficiently large $`Q`$ these solutions are guaranteed to be stable, as can be seen using the ‘thin wall limit’ , where the profile function can be approximated by a smoothed-out step function. For small Q it is necessary to perform a full stability analysis using the second variation of the action. In general, the results depend on the details of the potential, but it can be shown that arbitrarily small Q-balls are stable for certain potentials . For a rigorous approach to the classical stability of Q-balls see ref. and references therein. From the quantum mechanical point of view, the solutions are always stable for large enough $`Q`$ since the energy per unit charge approaches $`\omega _{}`$, which is always less than that for the $`\varphi `$ particle itself, which is $`\omega _+`$.
In choosing a simple potential which admits Q-balls there are three natural classes which have been considered, although there are obviously many other possibilities,
$`I`$ $`:U(f)=\alpha _1f^2+\alpha _2f^4+\alpha _3f^6,`$ (2.7)
$`II`$ $`:U(f)=\beta _1f^2+\beta _2f^3+\beta _3f^4,`$ (2.8)
$`III`$ $`:U(f)=\gamma _1f^2(1\gamma _2\mathrm{log}(\gamma _3f^2))+\gamma _4f^{2p}.`$ (2.9)
In each case it is possible to remove two of the parameters by rescaling the units of energy and time. Note therefore that the potentials of type I and II have one free parameter, while potentials of type III have two free parameters for a fixed value of $`p`$.
The type I potential is the simplest allowed potential which is a polynomial in $`f^2`$, while type II is the simplest which is polynomial in $`f`$. Finally, those of type III mimic the D-flat direction in the MSSM. Here $`p6`$ is some integer that ensures the growth of the potential for large $`f`$, but does not destroy the flatness property for intermediate values of $`f.`$ We should note that none of these types of potential are the kind which might be associated with a renormalizable quantum field theory, but are typical of effective theories incorporating radiative or finite temperature corrections to a bare potential.
In this paper we will mainly be concerned with the type I potential, although we have also studied the type II case. We should note that although the existence and stability properties of Q-balls with these potentials are somewhat different, the qualitative features of the dynamics appears to be almost independent of the potential. The reason for this is that the main interaction processes that we will describe, charge transfer and fission, are due mainly to the time-dependent nature of the solution, rather than the precise profile function.
Explicitly, we shall choose our potential so that
$$U(f)=f^2(1+(1f^2)^2),$$
(2.10)
and therefore in terms of the earlier notation we have that $`w_+=2`$ and $`w_{}=\sqrt{2}`$, so that stable Q-balls exist for $`\sqrt{2}<\omega <2`$. To illustrate the important features of Q-ball solutions we shall focus on the case of one dimension where the profile function equation (2.5) can be solved exactly to give
$$f_\omega (r)=\sqrt{\frac{4\omega ^2}{2+\sqrt{2\omega ^24}\mathrm{cosh}(2r\sqrt{4\omega ^2})}}.$$
(2.11)
The associated energy $`E_\omega `$ and charge $`Q_\omega `$ can then be computed to be
$$Q_\omega =\sqrt{2}\omega \mathrm{tanh}^1\left(\frac{2\sqrt{2\omega ^24}}{\sqrt{2}\sqrt{4\omega ^2}}\right),E_\omega =\frac{\sqrt{4\omega ^2}}{2}+\frac{(\omega ^2+2)}{2\omega }Q_\omega .$$
(2.12)
In figure 1 we plot the charge $`Q_\omega `$ and energy $`E_\omega `$ for $`\omega `$ in the allowed range $`\sqrt{2}<\omega <2.`$ and we see that both $`Q_\omega `$ and $`E_\omega `$ are monotonically decreasing functions of $`\omega .`$ From (2.11) we can deduce that
$$f_\omega (0)=\sqrt{(4\omega ^2)/(2+\sqrt{2\omega ^24)}},$$
(2.13)
and therefore $`f_\omega (0)`$ increases with the charge $`Q_\omega `$, since it is a decreasing function of $`\omega .`$ In figure 2 we display the energy per unit charge $`E_\omega /Q_\omega `$ as a function of the charge $`Q_\omega `$, from which it can be seen that $`E_\omega /Q_\omega `$ is a decreasing function of the charge. Recall that the asymptotic limit as $`Q_\omega \mathrm{}`$ is $`E_\omega /Q_\omega =\omega _{}=\sqrt{2}.`$ Thus, these Q-balls are stable against decay into a number of smaller Q-balls preserving the total charge.
Most of the general properties of Q-balls in any dimension and for differing choices of the potential are captured by this one-dimensional example, where explicit formulae are available. However, there are some slight differences, for example if $`D=3`$ then there is a lower bound on the charge of a Q-ball, whereas in the $`D=1`$ case considered above Q-balls can have an arbitrarily small charge. This is not a generic feature of every potential and, for example, it has been shown that for $`D=3`$ arbitrarily small Q-balls can be found with a potential of type II using a ‘thick wall limit’ in ref. . These kind of details are easily determined by solving the profile function equation (2.5) numerically and a complete treatment of these issues can be found in ref. . But, as we have already noted, we don’t believe that they are important for the dynamical processes which we focus on in the subsequent sections.
## 3 Q-ball dynamics on the line
In this section we shall investigate the dynamics of Q-balls in one-dimension; even for $`D=1`$ we shall see that multi-Q-ball dynamics is a complicated issue, there being a rich variety of phenomena associated with the non-quantization of the charge and the time-dependent phase.
To investigate the dynamics of Q-balls we numerically solve the field equations which follow from the Lagrangian (2.1) with the potential (2.10), namely
$$\ddot{\varphi }^2\varphi +2\varphi (24|\varphi |^2+3|\varphi |^4)=0,$$
(3.1)
which is valid for any value of D. The numerical methods we use are simple finite difference schemes involving either second or fourth order accurate spatial derivatives and a second order leapfrog algorithm for the time evolution with 1000 points, the spatial step size $`\mathrm{\Delta }x=0.1`$ and the time step size $`\mathrm{\Delta }t=0.02`$. We apply absorbing boundary conditions, which allows radiation to leave the grid and therefore simulates an infinite domain (see refs. for details on how to apply these boundary conditions).
As initial conditions to describe two well-separated Q-balls we use the ansatz
$$\varphi =e^{i\omega _1t+i\alpha }f_{\omega _1}(|x+a|)+e^{i\omega _2t}f_{\omega _2}(|xa|),$$
(3.2)
in one dimension, which can be trivially generalized to higher dimensions. This ansatz describes a Q-ball with frequency $`\omega _1`$ at the position $`x=a`$ and a second Q-ball with frequency $`\omega _2`$ at the position $`x=a.`$ The $`U(1)`$ symmetry of the theory means that for a single Q-ball the phase of $`\varphi `$ can be set to zero at $`t=0`$ without loss of generality. However, for multi-Q-ball configurations only the initial overall phase can be removed and so for a 2-Q-ball configuration the relative phase, $`\alpha `$, remains as an important parameter.
The total charge of this configuration is
$$Q=Q_{\omega _1}+Q_{\omega _2}+(\omega _1+\omega _2)\mathrm{cos}\alpha _{\mathrm{}}^{\mathrm{}}f_{\omega _1}(|x+a|)f_{\omega _2}(|xa|)𝑑x.$$
(3.3)
The final term in the above expression is exponentially small in the separation parameter $`a`$, since the profile functions have an exponential fall-off. However, we see that the relative phase $`\alpha `$ does affect the value of the total charge. Thus, strictly speaking, it is not valid to substitute this ansatz into the energy functional to determine how the energy depends on the relative phase, since this would involve comparing configurations with differing values of $`Q.`$ The same remark also applies to any attempt to determine how the potential energy depends upon the relative separation $`a,`$ which would help to determine the nature of the interaction force between Q-balls. This issue and its resolution using the methods of ref. will be discussed in ref. .
To begin with we consider configurations in which the two Q-balls have the same charge, that is $`\omega _1=\omega _2=\omega .`$ In figure 3, and in all subsequent figures illustrating the dynamics of Q-balls, we plot the charge density (the integrand in equation (2.3)) for the initial conditions and at later times<sup>1</sup><sup>1</sup>1We could have made similar plots of the energy density which are not equivalent. We believe the charge density gives a better representation of the dynamics.. The parameter values used for this simulation are $`\omega =1.5`$ and $`a=4`$, with the Q-balls initially in phase so that $`\alpha =0.`$ The two Q-balls slowly attract and coalesce to form one larger Q-ball which has a charge which is slightly less than the sum of the charges of the two original Q-balls; the charge deficit being carried away by the fission of two additional Q-balls which, in figure 3, can just be seen moving away from the almost stationary large Q-ball at the origin. The attraction of the two Q-balls is simple to explain; it being a consequence of the ratio $`E/Q`$ decreasing as $`Q`$ increases. However, the process of fission is less intuitive and is a novel concept to those who might have studied the dynamics of topological solitons in an attractive potential. In the topological case the charge is an integer and so the fission of higher charge solitons can only be achieved when the solitons are moving sufficiently fast for the kinetic energy to overcome the attraction and release a soliton. But here the charge of an isolated Q-ball can have any value, arbitrarily close to zero, and so one might imagine that the energy barrier to fission at low interaction speeds is small, particularly when the charge is high.
The fission of Q-balls is a process which can occur when a Q-ball suffers a large distortion, for example, during a collision. This can be demonstrated by taking a single Q-ball and squashing it by applying the scale transformation
$$x\lambda x,\varphi \sqrt{\lambda }\varphi ,$$
(3.4)
where $`\lambda >1`$ is a scale factor. The scaling of the $`\varphi `$ field is required in order that the charge of the Q-ball is not changed by the scaling. For small enough values of $`\lambda `$ the Q-ball oscillates, but then eventually returns to the original un-squashed Q-ball corresponding to $`\lambda =1`$, which is to be expected since these Q-balls are stable. However, for a sufficiently large distortion the Q-ball will break up into smaller Q-balls. This is illustrated in figure 4, where we take a Q-ball with charge $`Q=8.4`$ and perturb it with a scale $`\lambda =1.6.`$
To consider the efficiency of the fission process as a function of the charge, we define the quantity
$$\mathrm{\Delta }(Q)=(2E(Q/2)E(Q))/E(Q),$$
(3.5)
where $`E(Q)`$ denotes the energy of a Q-ball with charge $`Q.`$ $`\mathrm{\Delta }(Q)`$ is the fractional increase in energy required to allow a charge $`Q`$ Q-ball to fission into two charge $`Q/2`$ Q-balls. This is a monotonically decreasing function of $`Q`$ with $`\mathrm{\Delta }(\mathrm{})=0`$, with the limit $`Q\mathrm{}`$ being a Bogomolny-like limit in which the energy is proportional to the charge. Thus we expect that the fission of Q-balls is more easily stimulated when the charge is large. To verify this we apply the same distortion factor $`\lambda =1.6`$ as displayed in figure 4, but this time we take a smaller Q-ball with charge $`Q=5.6`$, and the results are displayed in figure 5. The Q-ball performs a breather-like motion in which two structures initially begin to form but then recombine. This motion persists for many cycles with a slowly decreasing amplitude until it eventually settles down to a configuration which is very close to the original Q-ball without a distortion (the configuration at $`t=3000`$ is almost identical to the stationary $`Q=5.6`$ Q-ball).
The above expectations of the fission of Q-balls are confirmed by performing simulations, as in figure 3, with two stationary Q-balls and varying the charge (increasing or decreasing the value of $`\omega `$). The charge of the additional Q-balls produced decreases as the charge of the initial Q-balls is reduced and for small enough Q-balls no fission takes place; rather the sole Q-ball formed oscillates for some time, with a decreasing amplitude.
If the two Q-balls are initially Lorentz boosted toward one another, each with a velocity $`v`$ say, then if $`v`$ is large enough the two Q-balls can be made to pass through each other. In figure 6 we display the charge density for a simulation with $`\omega _1=\omega _2=1.5`$, $`a=10`$ and $`v=0.3.`$
In this case the two Q-balls pass through each other, although they do lose some charge via radiation during the interaction process and their velocities are reduced. For a slightly lower value of the velocity the two Q-balls pass through each other, but do not escape to infinite separation. Rather they subsequently recombine, forming a stationary Q-ball at the origin and producing two additional Q-balls in the same manner as described above. In figure 7 we plot the positions of the two main Q-balls (determined as the location of the maximum of the charge density) as a function of time for the velocity $`v=0.28.`$
To study how the relative phase affects the interaction we consider the same initial configuration used to produce figure 3 except that we set $`\alpha =\pi `$, so that the two Q-balls are exactly out of phase. In this case the resulting evolution is very different and the two Q-balls simply drift apart with no change in their shape or charge, even though the crude arguments based on $`E/Q`$ suggest that they should attract. The effects of changing the overall phase are similar in many ways to the overall isospin rotations possible in 2-skyrmion interactions. There it is possible to make the skyrmions attract or repel by an internal SU(2) rotation about the line joining the two soliton centres. Another way of understanding the relative phase is as a current between two charged bubbles, if the Q-balls are thought of as bubbles. The results of this repulsive interaction channel are displayed in figure 8.
As we have seen, for $`\alpha =0`$ the Q-balls attract and for $`\alpha =\pi `$ they repel. However, for intermediate values of $`\alpha `$ the dynamics is much more complicated. We display the evolution for $`\alpha =\pi /9`$ in figure 9, illustrating a novel process which we shall call charge transfer. The Q-balls initially move very slightly towards each other, but they eventually repel and separate off to infinity. The most interesting aspect is that the charge of the first Q-ball has clearly decreased and that of the second Q-ball has increased, despite the fact that the Q-balls remain two distinct objects with only a very small overlap throughout the time evolution. As they separate the smaller Q-ball moves at a greater speed than the larger one.
In figure 10 we plot the total charge $`Q`$ in the right half of the line, $`x>0,`$ as a function of time for the simulation displayed in figure 9, where $`\alpha =\pi /9`$, and also for the cases $`\alpha =\pi /4`$ and $`\alpha =\pi /2.`$ We see that as the relative phase is decreased the rate at which the charge transfer initially takes place is reduced but the total charge transfered is increased. The in-phase limit $`\alpha =0,`$ where the two Q-balls form a single larger Q-ball, is a smooth limit if we interpret it as a total charge transfer. In the out-of-phase limit $`\alpha =\pi `$, as we have already seen, there is no charge transfer. If $`\alpha <0`$ then the same amount of charge is transferred as in the case of a phase $`\alpha `$, but this time it is the Q-ball in the left half of the interval which increases in charge.
It may seem surprising that there is charge transfer, but this result can be understood, at least qualitatively, by considering a simplified mechanical analogue of the field dynamics associated with two well-separated Q-balls. Consider two equal charge Q-balls, fixed at the positions $`x=\pm a`$, then the ansatz (3.2) may be written in the form
$$\varphi =e^{i\theta _1}f(|x+a|)+e^{i\theta _2}f(|xa|),$$
(3.6)
where $`\theta _1\theta _1(t),\theta _2\theta _2(t)`$ are the time-dependent phases of the two Q-balls and $`f`$ is a profile function. To leading order in the separation $`a`$, corresponding to the limit of large separation, the contribution to the Lagrangian is given by
$$=\frac{1}{2}M(\dot{\theta }_1^2+\dot{\theta }_2^2)ϵ^2\mathrm{cos}(\theta _1\theta _2)4M,$$
(3.7)
where $`M=_{\mathrm{}}^{\mathrm{}}f(|x|)^2𝑑x`$ is treated as a constant moment of inertia and $`ϵ^2=4_{\mathrm{}}^{\mathrm{}}f(|x+a|)f(|xa|)𝑑x`$ is a small interaction coefficient. To derive this Lagrangian we have made the assumption that the profile function is time independent, which obviously has a very limited range of validity as we shall discuss further below; it is nonetheless instructive. The equations of motion which follow from (3.7) are
$$\ddot{\theta }_1+\ddot{\theta }_2=0,\ddot{\theta }_1\ddot{\theta }_2=\frac{2ϵ^2}{M}\mathrm{sin}(\theta _1\theta _2).$$
(3.8)
The first of these equations simply represents the fact that the sum of the rotation frequencies $`\dot{\theta }_1+\dot{\theta }_2`$ is conserved, and the second equation determines the dynamics of the relative phase. There are symmetric solutions, $`\theta _1=\theta _2`$, and $`\theta _1=\theta _2+\pi `$, where the phase difference remains constant, corresponding to the two Q-balls being exactly in-phase or exactly out-of-phase for all time, but for general values of the initial relative phase, $`\alpha =\theta _1(0)\theta _2(0)`$, there will be a non-trivial time dependence. For all $`\alpha (0,\pi )`$ there will be an initial positive acceleration in the relative phase, so that $`\dot{\theta }_1>\dot{\theta }_2`$ for small $`t>0.`$ Thus the first Q-ball will have a higher frequency than the second Q-ball and, since we know that the charge of a Q-ball decreases with increasing frequency, then this corresponds to the charge of the first Q-ball decreasing and the charge of the second Q-ball increasing. This simple analysis also predicts that the initial rate of charge transfer will be greatest for a relative phase $`\alpha =\pi /2`$ and will decrease as $`\alpha `$ decreases. This agrees with the observation we made earlier by an examination of the plots in figure 10 for small times.
However, what we are clearly not able to study with our simple restricted mechanical model is the whole charge transfer process for later times. One reason for this is that we assumed that the profile function $`f`$ was fixed when of course we know that it is highly dependent on the rotation frequency (see equation (2.11)). In particular this dependence constrains the rotation frequencies to satisfy $`\omega _{}<\dot{\theta }_1,\dot{\theta }_2<\omega _+`$ and as either of these limits are approached our simple model breaks down. One might be tempted to improve our simple model to deal with this issue by including the known frequency dependence of the profile function, but it is not obvious how to do this since the profile function depends upon the frequency $`\dot{\theta }`$ so using such an ansatz in the Lagrangian would lead to a Lagrangian for a mechanical system with second order derivatives $`\ddot{\theta }`$ and hence a fourth order equation of motion. Furthermore, we have assumed that the positions of the Q-balls are fixed when in fact the results of the full simulations show that they eventually drift apart. This effect will also serve to cut-off the relative phase dynamics since it will correspond to reducing the $`ϵ^2`$ coefficient in our simple mechanical model.
In summary, we have shown that a simple mechanical model is useful in understanding the qualitative features of the charge transfer process, but a more sophisticated analysis is required to explain the quantitative behaviour found. The analysis of relative phase dynamics in mechanical systems, such as discrete breathers, has been studied in some detail and the phase space trajectories are well understood . These methods can be extended to study the more complicated relative phase dynamics, and hence charge transfer, of Q-balls .
So far we have only considered initial conditions in which the two Q-balls have the same charge. For two Q-balls which have different charges the initial relative phase does not have the same importance as for Q-balls of the same charge, due to the fact that the Q-balls have different frequencies and so the initial relative phase will not be preserved, even with no interaction. This can easily been seen using the simple mechanical system above.
In figure 11 we plot the charge density for the initial conditions $`\omega _1=1.8,\omega _2=1.5,a=3,\alpha =0.`$ It can be seen that the two Q-balls repel and there is virtually no charge transfer since the solitons never get close enough for the charge transfer process to become important. Similarly, if a non-zero relative phase is introduced then virtually no charge transfer takes place, although the rate of separation does vary very slightly. However, if the Q-balls are Lorentz boosted towards each other with a sufficiently large velocity $`v`$ so that they collide and pass through each other, it is possible to induce charge transfer as illustrated in figure 12 for the parameters $`\omega _1=1.8,\omega _2=1.5,a=6,\alpha =0,v=0.2.`$ The amount of charge transferred depends on the value of the relative phase as the Q-balls collide, as can be verified by changing the initial phase. It can be checked that this is equivalent to varying the initial separation, since the time to collision is then altered and hence the relative phase is different by an amount equal to the change in collision time multiplied by the frequency difference. Just using the simple mechanical analogy, one might have naively expected that an initial difference in the rotation speeds would be on a similar footing to an initial phase difference, but this is clearly not the case. It is evident that there is a non-trivial interaction between the relative dynamics of the Q-balls and the charge transfer process.
## 4 Planar Q-balls
The main features of one-dimensional Q-balls which we have described in the previous section, such as charge transfer and the dependence of the interaction force on the relative phase, carry through to the two-dimensional case. We demonstrate this by again performing numerical simulations of the field equations via an equivalent finite difference scheme to the one dimensional case. We find that a grid containing $`200^2`$ points with $`\mathrm{\Delta }x=0.2`$ and $`\mathrm{\Delta }t=0.05`$ gives an accurate representation of the dynamics in this case. In contrast to the one-dimensional case an exact solution is not known for the profile function in two-dimensions, but it is a simple matter to numerically obtain the profile function using a standard shooting method.
One might assume that head-on collisions of Q-balls with a small charge (for example, $`\omega =1.6`$) are equivalent to those in one-dimension with attraction, repulsion and charge transfer taking place as before and indeed this is the case as we will discuss later. The phase-dependent force is, however, a more general concept which we will illustrate in fully two dimensional interactions with a non-zero impact parameter. In particular, we consider the collision of two equal charge Q-balls ($`\omega _1=\omega _2=1.6`$) with a non-zero impact parameter, where the initial positions of the Q-balls are $`(x_1,x_2)=\pm (6,3)`$, each is boosted along the $`x_1`$-axis with a velocity $`v=0.05`$, and the relative phase $`\alpha `$ is set to zero. In figure 13 we plot the charge density for this interaction at $`t=0,104,112,3200.`$
As in the one-dimensional case (where collisions are head-on) the two Q-balls attract and form a single larger Q-ball, although the large Q-ball has some angular momentum due to the fact that the collision was not head-on. Figure 14 displays the results of the same simulation except that the two Q-balls are exactly out-of-phase, that is $`\alpha =\pi ,`$ where the Q-balls clearly repel. Finally, in figure 15 we display the simulation with a relative phase $`\alpha =\pi /4`$, where there is an initial attraction, followed charge transfer and finished off by a repulsion which forces the two Q-balls apart. We conclude, therefore, that while the head-on collisions of small Q-balls in two-dimensions can be thought of as being effectively one-dimensional, the same dynamical processes are active in the case of a non-zero impact parameter.
Next, we turn our attention to head-on collisions of Q-balls with higher charge, where — based on the intuition of the one-dimensional interactions — one would expect things to be slightly different. We take two Q-balls with $`\omega _1=\omega _2=1.5`$ at positions $`(x_1,x_2)=\pm (10,0)`$ with each Lorentz boosted towards the other with a velocity $`v=0.4.`$ Figure 16 displays the charge density at $`t=0,20,24,28,32,36,40,44,52`$ for the in-phase case, $`\alpha =0.`$ As can be seen from the figure, there is a very complicated interaction process involving the charge being strongly deformed and the emission of some radiation. Eventually, two Q-balls emerge from the interaction region at right angles to the initial direction of approach. Naively one may think that this is a simple $`90^{}`$ scattering process as seen in a number of topological soliton models and suggested for Q-balls in ref. <sup>2</sup><sup>2</sup>2We should note that the work presented in ref. uses a potential of type II, not type I as used in our work, but that the qualitative nature of this process is independent of the choice of potential.. However, the scattering of Q-balls is a complicated dynamical issue rather than being topological, and the underlying mechanisms are very different. In particular, there is no associated geometry of a moduli space which forces the Q-balls to scatter at right angles. Rather, during collisions the Q-matter becomes highly deformed with huge charge densities and it is this deformation, and its associated pressures, that lies at the heart of the interaction process and the fission of Q-balls in the plane perpendicular to the incident direction. As we demonstrated in the one-dimensional case, a sufficient distortion of a Q-ball will induce fission, and it this same process which is responsible for this more complicated phenomena in two-dimensions.
This point can be illustrated immediately by considering the same scattering process, but this time we set the Q-balls to be exactly out-of-phase, that is $`\alpha =\pi `$ with the results displayed in figure 17. Although the initial conditions look identical in figures 16 and 17, the evolution is clearly very different. As the two Q-balls are now in a repulsive phase the Q-matter gets distorted in a very different way. Rather than forming a single structure, as in figure 16c, the two individual Q-balls never actually coalesce because of the repulsive interaction, getting squashed separately and this distortion induces the fission of each. Thus, each Q-ball splits into two and the two pairs repel each other, producing four Q-balls in all, which are clearly visible in figure 17g. For this particular set of initial parameters the Q-ball pairs do not have enough energy to escape each others attraction and eventually recombine leaving two Q-balls which move off to infinity. By, for example, increasing the initial velocity it is found that the four Q-balls can be produced in such a way that they all separately move off to infinity without any subsequent recombination. If the initial velocity is small enough then the distortion is not large enough to induce fission and the two Q-balls eventually repel keeping their individual structure intact.
In figure 18 we investigate the same simulation as in figure 16 but with an initial relative phase $`\alpha =\pi /2.`$ In this case we expect that the distortion will also be accompanied by a charge transfer, and indeed this is what we find, as the first Q-ball loses charge to the second one and then each Q-ball undergoes fission to produce a total of two small Q-balls and two large Q-balls. The two small Q-balls move away from the interaction region at a faster rate than the large Q-balls, and they do not recombine. The two large Q-balls move away from the interaction region at only a very slow speed and in fact they eventually recombine after a time of around $`t=200`$ (the final plot in figure 18 is only at $`t=52`$).
As we discussed in the one-dimensional case, large Q-balls are more susceptible to fission than smaller Q-balls, so we expect that the scattering processes we have described above will vary depending on the charge of the initial Q-balls. For example, our reasoning predicts that the fission process in figure 16, which produced two Q-balls moving at right-angles to the initial line of approach, will be more difficult to reproduce for smaller charges. To test this we perform the same simulation, with $`v=0.4`$ again, but decrease the charge of the initial Q-balls by taking $`\omega =1.6`$ rather than $`\omega =1.5.`$ The resulting evolution is shown in figure 19 and confirms that now the distortion is not sufficient to liberate two Q-balls. The configuration oscillates for some time before settling down to a single larger Q-ball, after a small amount of charge has been dissipated through radiation. Fission can be produced for these smaller Q-balls by increasing the impact velocity, but this also has the result that some of the charge passes straight through the interaction region producing small Q-balls which continue to travel along the direction of approach. A collision at increased velocities is also a more violent process and more charge is lost to radiation in these circumstances. In figure 20 we display the evolution for the case where the velocity is increased to $`v=0.6`$, with all other parameters kept the same as in figure 19. Just visible in figure 20d are the four very small Q-balls which are produced by this collision together with a ring of charge carried away by the radiation generated. If the collision velocity is increased further then the two Q-balls have less time to interact and their momentum carries them through the collision process with no deflection. This is demonstrated in figure 21 where $`v=0.8`$ and no additional Q-balls are produced. In summary, as the charge is reduced an increased velocity is required in order for a sufficient deformation to be generated to produce fission, but this also results in more of the charge being carried straight through the collision. Thus for small charge there is a very limited window of velocities for which collisions of the form displayed in figure 16 may occur.
## 5 3D Q-balls
In the previous two sections we have built up a picture of the dynamics of Q-balls in one and two dimensions. In going from the extensive study in one dimension to two dimensions we have noted a number of subtle effects associated with the extra dimension. However, the basic processes involved are the same: attraction, repulsion and charge transfer. In this section we will apply the same numerical techniques to the case of three dimensions. To begin with we conducted an extensive study of the dynamics on grids containing $`100^3`$ points and have once again found that in many cases the dynamics are very similar to those in one-dimension. At the risk of labouring the point we found that for small Q-balls, if they were initially in-phase they coalesced, while if they were out-of-phase they repelled, and if they had any other phase they engaged in charge transfer. However, we did find some extremely complicated interactions which are related to those seen in the case of two dimensions. As was pointed out in the previous section when the Q-balls have a large charge, their interactions can have some interesting variants in two dimensions and it is these particular cases in three dimensions on which we will focus in this section.
In particular we have focussed on the analogues of figures 16, 17 and 18 in which complicated dynamical phenomena were identified. In each of these cases we have performed the analogous simulations on a grid of $`300^3`$ points with $`\mathrm{\Delta }x=0.3`$ and $`\mathrm{\Delta }t=0.03`$ in order to accurately simulate the complicated dynamical processes. In each of the three cases we start with 2 Q-balls each with $`\omega =1.5`$ at $`\pm (15,0,0)`$, Lorentz boosted toward each other with a velocity $`v=0.4`$. The results of the simulations are displayed in figures 22, 23 and 24.
The in-phase case (figure 22) has some marked similarities to the equivalent case in two dimensions (figure 16) if one looks at the two-dimensional slice through the centre of the Q-balls<sup>3</sup><sup>3</sup>3It should be noted that the two interactions are not equivalent even when the parameters are almost identical since the relationship between the charge $`Q`$ and the frequency $`\omega `$ is not the same in two and three dimensions.. However, the extra dimension has one remarkable effect: it allows for the production of a loop in the plane perpendicular to line joining the two incident Q-balls. This phenomena is the three dimensional analogue of the right-angled fission process described in the two dimensional case. But in three dimensions the fission takes place symmetrically in all directions in the plane while respecting the cylindrical symmetry of the initial configuration. The loop expands leaving some charge in the centre which later expands to create a second, much smaller loop. Later both the loops collapse and Q-balls emerge back along the incident direction.
The formation of the loop in this three dimensional simulation adds further weight to our earlier discussion of the two dimensional case where we pointed out that right-angled fission of two Q-balls was not of topological origin, nor was it even related to the topological interactions of, for example, vortices , monopoles and skyrmions . The reason being that in a topological interaction in three dimensions one would have expected there to have been a preferred direction (this is because in the case of topological solitons, for example skyrmions, the field configuration of a single soliton is not spherically symmetric, although the change in the field due to a spatial rotation can be undone by acting with a symmetry of the theory, which means quantities such as the energy density are spherically symmetric). But as we have already pointed out the formation of a loop is reliant on all directions being on an equal footing as far as the fission process is concerned, which of course is due to the fact that the field itself is spherically symmetric for a single Q-ball.
This explanation of the formation of a loop during the interaction of two large Q-balls in three dimensions is compatible with the results of the out-of-phase case (figure 23) and that of a general relative phase (figure 24). In both cases, the two dimensional slice through the interaction region is very similar to that of the equivalent two dimensional interactions (figures 17 and 18 respectively). In the out-of-phase interaction, two identical loops form which are then repelled back along the direction from which they came. As they move away they begin to collapse, the final outcome being a series of symmetrically placed Q-balls along the incident direction. The interaction for $`\alpha =\pi /2`$ is similar, except that, as expected, charge transfer takes place during the interaction and the two loops created have very different charge.
Just to finish this section off and by way of illustrating that this process is reasonably generic, we have performed an equivalent simulation to figure 22 with a much higher speed of incidence $`(v=0.8)`$. Due to the Lorentz contraction of the initial conditions, this requires a smaller value of $`\mathrm{\Delta }x=0.15`$ and consequently $`\mathrm{\Delta }t=0.015`$, and the results of this simulation are displayed in figure 25. We see the production of a big loop at the point of interaction plus two others which are repelled from the centre along the line of interaction. At the end of the simulation the loops are still expanding in size and are also getting close to the size of the discrete grid. It is an interesting question as to whether loops can be stable, and this question will be addressed in a separate publication .
## 6 Q-ball Anti-Q-ball Dynamics
In the preceding sections we have studied in detail 2-Q-ball interactions. The charge $`Q`$, however, can also be negative; this being achieved in the Q-ball solution by replacing $`\omega `$ with $`\omega `$ and these solutions are known as anti-Q-balls. In this section we will study the interactions of Q-ball/anti-Q-ball pairs in two and three dimensions.
Intuitively, one would expect slow soliton/anti-soliton interactions with equal and opposite charges to result in annihilation into radiation. However, only for a very small range of parameters does this annihilation take place in the case of Q-balls due to the complicated nature of the time-dependent interaction potential which we have highlighted in the case of 2-Q-ball interactions.
In general a Q-ball/anti-Q-ball interaction will result in either the two solitons bouncing back, or them passing through each other. In both cases, the charge is partially annihilated, but only for a very limited region of the interaction parameter space can it be thought of as being complete. This absence of the annihilation can be attributed to the concept of charge transfer which we have discussed in 2-Q-ball interactions<sup>4</sup><sup>4</sup>4Here, as in the case of two Q-ball interactions with different charges, the initial relative phase is a less important concept. But charge transfer can still take place since the difference between the two rotation speeds of the Q-balls is maximal. The main difference between the charge transfer process for 2-Q-ball interactions and the Q-ball/anti-Q-ball interactions under consideration here is the two solitons now have opposite charge and hence when charge is transfered it results in annihilation. However, we have showed that the charge transfer never takes place fully in 2-Q-ball interactions and hence annihilation also never takes place fully in a single interaction. When there is a lower bound on the charge of a Q-ball it is more likely that sufficient charge transfer can take place for complete annihilation, whereas in models where arbitrarily small Q-balls can exist complete annihilation is likely to be much more difficult.
The process of partial annihilation, via charge transfer, is illustrated in figure 26 where we have displayed the charge density at $`t=0,15,20,50`$ for the collision of a Q-ball and an anti-Q-ball in two dimensions. The solitons were initially at $`\pm (6,0)`$, with $`\omega _1=\omega _2=1.8`$ and were Lorentz boosted together with a velocity $`v=0.3`$. It is clear that the momentum of the Q-ball carries it through the interaction region and that there is some annihilation of the charge; the maximum charge density of the outgoing Q-ball being lower than that for the incoming one. As we have already discussed this is generically what takes place during a Q-ball/anti-Q-ball collision. A variant on this kind of interaction is that the Q-balls bounce back, again partially annihilating charge, which takes place at low incident velocities for this particular charge.
This picture of partial annihilation with bounce back at low velocities and the solitons passing through each other at high, suggests that there exists some critical incident velocity $`v_\mathrm{c}`$, a function of $`\omega `$ at which annihilation takes place, and indeed this is what we find<sup>5</sup><sup>5</sup>5In fact, there will exist a small range of velocities around $`v_\mathrm{c}`$ for which complete annihilation takes place.. Figure 27 illustrates this by displaying the charge density at $`t=0,25,50,100,150,250,300,350,400`$ for a Q-ball/anti-Q-ball collision, with the solitons initially positioned at $`\pm (6,0)`$, charges $`\omega _1=\omega _2=1.5`$, Lorentz boosted together with velocity $`v=0.3`$. These are the same parameters as in figure 26, except that the charge is much larger. It can be clearly seen that annihilation eventually takes place, but even in this case the mechanism is complicated, involving a number of oscillations of the system before it is achieved. The charge here is much higher than for the example above which had $`\omega =1.8`$, and in the Q-ball interactions we saw complicated fission processes for interactions involving solitons with this charge. This is also the case in Q-ball/anti-Q-ball interactions as is illustrated in figure 28, which uses the same parameters as in figure 27 except that the incident velocity is now much higher, $`v=0.6`$. The figures shown are at times $`t=0,10,15,25,30,35,40,45,50.`$ One can see, after some partial annihilation of charge, that the fission of the incident Q-balls takes place in the direction perpendicular to the line of incidence and that now there is an effective bounce back of the incident solitons with a reduced charge.
These processes are also prevalent in both one and three dimensions. By way of illustration we have also included two examples in three dimensions. In figure 29 the solitons initially at $`\pm (15,0,0)`$, are Lorentz boosted together with a velocity $`v=0.3`$ and have $`\omega _1=\omega _2=1.5`$. In this particular interaction there appears to be very little annihilation and the solitons effectively pass through each other. Figure 30 has the same initial configuration, except that the solitons are boosted together with an initial velocity of $`v=0.6`$. The subsequent interaction is complicated involving first the formation of two loops, which is the equivalent of the right-angled fission observed in two dimensions, and then what appears to be almost total annihilation once the loops collapse. Two small Q-balls are emitted along the line of incidence, along with much radiation.
## 7 Summary and conclusions
We have identified the key parameters in the interactions of Q-balls to be the relative phase, the incident velocity and the charge of Q-balls, with the resulting interactions being strongly dependent on these parameters. The generic interaction involves attraction if the relative phase $`\alpha =0`$ and repulsion if $`\alpha =\pi `$ when the two Q-balls have the same charge. If they have an initial relative phase other than $`\alpha =0`$ or $`\alpha =\pi `$, or if the charges of the Q-balls are different, the dynamics of the Q-balls results in charge transfer, a phenomena which is analogous to that observed in discrete breather systems (see, for example, ref. and references therein), although this behaviour often has to be induced by making the solitons come together via a Lorentz boost. With no Lorentz boost such breather systems naturally repel and this is also seen after the charge transfer process is complete. If the incident velocity is extremely high Q-balls can be made to pass through each other with very little interaction taking place.
This picture is almost independent of the number of dimensions if the charge of the Q-balls is small (for example, $`\omega 1.6`$, but if the charge is much larger fission can take place during the interaction process due to the compression of the charge. In one dimension this process results in the emission of Q-balls during the slow interaction of two large Q-balls. In higher dimensions complicated, but analogous, phenomena are observed in which fission takes place in the direction perpendicular to the line of incidence. This leads, under specialized circumstances to the right angled fission of Q-balls in two dimensions and the production of loops in three. These fission effects can also be coupled with those of attraction, repulsion and charge transfer to produce complicated compound phenomena.
Interestingly, the naive expectation that an Q-ball/anti-Q-ball should annhilate into radiation is modified by the complicated breather-type interactions which take place. Since the difference between the two charges is large this can be thought of a special case of the charge transfer process, leading to the phenomenon of partial charge annihilation, the case of complete annihilation being very special. Fission of the Q-ball and anti-Q-ball can also take place if the charge is high.
Our original motivation was to understand the microphysical interactions of Q-balls formed by the Affleck-Dine mechanism for baryogenesis within the framework of the MSSM. The potential that we have concentrated on in this paper is different to that expected in the MSSM, but as we have pointed out the main interaction processes that we have identified are related to the phase. Therefore, we expect our result to be qualitatively independent of potential, and hence our results have some bearing on this case. We have repeated several of the simulations described in this paper for a potential of type II and found the same qualitative results. The next step in this research is an analytic description of the dynamics in terms of a slow manifold approach , before their application to the problem of Q-ball formation in the Early Universe.
|
warning/0003/hep-ph0003044.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
Long time ago, Dolgov and Zakharov have introduced an alternative approach to the axial anomaly , based on the study of the triangular diagram, with two vector and one axial vertices, by means of dispersion relations. From this approach, the interpretation of the axial anomaly as an infrared phenomenon follows. The infrared aspect of the axial anomaly, arised in this paper, can be seen as complementary to the more familiar ultraviolet one, which emerges from the renormalization procedure. It allows to shed light upon the physical meaning of the chiral symmetry breaking as related to a non conservation of helicity.
In the absorbitive part of the triangular diagram is computed by making an unitary cut and results proportional to the product of two amplitudes. The first one is relative to the production of a fermion-antifermion pair by an axial source and the second one corresponds to the subsequent annihilation of the pair into two real photons. In both these processes there occur helicity flips, thus the chirality is not conserved in the zero mass limit. For what follows, it is important to examine in particular the second process contributing to the absorbitive part of the triangular diagram ().
By taking the $`m0`$ limit of the absorbitive part of the triangular diagram, Dolgov and Zakharov find a finite result :
$$\mathrm{Im}g_1(q^2)\pi \delta (q^2)\mathrm{as}m0,$$
(1)
with $`g_1(q^2)`$ an invariant scalar function and $`q^2`$ the momentum transfer. Eq. (1) shows that the $`m0`$ limit is not smooth .
This behaviour is connected with a non conservation of helicity, in the massless limit, a non conservation of chirality. Thus the axial anomaly emerges by studying the properties of the amplitude in the infrared region. The purpose of this work is to analyse the cancellation of mass singularities in polarized amplitudes. We will not discuss the physical implications of the zero fermion mass limit.
As stated in refs. , the finite result is a consequence of the singularity occurring in the fermion propagator as $`m0`$, that exactly cancels the suppression factor $`m^2`$ of the product of the amplitudes.
The connection between the anomaly and the non conservation of physical quantities has riceved a lot of attention in the literature. In a seminal work , Gribov has described the source of the anomalies as a collective motion of particles with arbitrarily large momenta in the vacuum. Related to this work is ref. . The Dolgov and Zakharov approach has been also considered in refs. .
We attempt here to interpret some simple processes, as the $`\pi ^+`$ and $`Z^0`$ polarized radiative decays, as a manifestation of the infrared nature of the axial anomaly . In particular, we calculate the relative decay rates, for the cases in which the outgoing leptons are in a definite helicity state and we examine in some detail the structure of the mass singularities and their cancellation. We study how the Kinoshita-Lee-Nauenberg (KLN) theorem applies to these cases and we consider the analogies and the differences with respect to the corresponding unpolarized decay rates.
## 2 Polarized decay of the pion
In the second process contributing to the absorbitive part of the triangular diagram a charged fermion with fixed helicity undergoes an helicity flip by emitting a photon . This is a forbidden process in the massless limit, but in this limit the absorbitive part doesn’t vanish.
We consider helicity changing reactions, studied for the first time by Lee and Nauenberg , by using the approach of Dolgov and Zakharov.
We have found in the literature other works dealing with this kind of processes. For example in ref. the authors consider the process
$$e^{}(p_{},\lambda )+A(p)e^{}(p_{}^{},\lambda ^{})+\gamma (k,\lambda _\gamma )+B(q_i),$$
(2)
where $`A`$ is a target, $`\gamma `$ is a bremsstrahlung photon, assumed almost collinear with respect to the direction of the incident electron and $`B`$ is a set of particles produced in the reaction.
For $`\lambda \lambda ^{}`$, the electron makes an helicity flip. The helicity flip cross section, computed to the leading order in the electron mass, is given by:
$$\frac{\mathrm{d}\sigma _{hf}}{\mathrm{d}x}=\sigma _0\left(s(1x)\right)\frac{\alpha }{2\pi }x,$$
(3)
where $`s=(p_{}+p_{}^{})^2`$, $`x=k_0/E`$, $`E`$ is the energy of the incoming electron and $`\sigma _0`$ is the cross section for the Born process
$$e^{}(p_{}k,\lambda )+A(p)e^{}(p_{}^{},\lambda )+B(q_i).$$
(4)
We see that the expression (3) does not vanish in the massless limit .
We now consider the leptonic decay of the pion. At the Born level, the total decay rate relative to the non radiative pion decay
$$\pi ^+l^++\nu _l,$$
where $`l^+`$ is an antilepton ($`e^+`$ or $`\mu ^+`$) and $`\nu _l`$ is the associated neutrino, is given by:
$$\mathrm{\Gamma }_0(\pi ^+l^+\nu _l)=\frac{G^2f_\pi ^2}{8\pi }V_{ud}^2\frac{m_l^2}{m_\pi ^3}(m_\pi ^2m_l^2)^2,$$
(5)
where $`G`$ is the Fermi coupling constant, $`f_\pi `$ is the pion decay constant, $`V_{ud}`$ is the CKM matrix element, $`m_l`$ and $`m_\pi `$ are the lepton and pion masses, respectively.
$`\mathrm{\Gamma }_0`$ is proportional to $`m_l^2`$; this factor is due to the fact, that, due to angular momentum conservation, the pion produces a left-handed lepton, while the structure of the weak coupling requires the $`l^+`$ to be right-handed for $`m_l=0`$.
The total decay rate for the process in which the lepton is polarized can be written as:
$`\mathrm{\Gamma }_0^{pol}(\pi ^+l^+\nu _l)={\displaystyle \frac{G^2f_\pi ^2}{16\pi }}V_{ud}^2{\displaystyle \frac{m_l^2}{m_\pi ^3}}(m_\pi ^2m_l^2)^2\left(1{\displaystyle \frac{𝐩_l𝐬_l}{|𝐩_l|}}\right),`$ (6)
here $`𝐩_l`$ and $`𝐬_𝐥`$ are the linear momentum and the spin vector of the lepton, respectively, giving $`𝐩_l𝐬_l/|𝐩_l|=\pm 1`$, for right-handed and left-handed lepton, respectively. Eq. (6) indicates that the lepton is mandatory left-handed, as requested by angular momentum conservation.
Let us consider the radiative dacay
$$\pi ^+l^++\nu _l+\gamma .$$
The relative amplitude can be written as the sum of two contributions
$$M(\pi ^+l^+\nu _l\gamma )=M_{IB}+M_{SD},$$
(7)
where $`M_{IB}`$ is the Inner Bremsstrahlung amplitude and $`M_{SD}`$ is the Structure Dependent one.
We calculate the probability that the lepton flips its helicity and becomes right-handed, by emitting a real photon. The relevant part, for the problem we are considering, is the Inner Bremsstrahlung contribution described by the diagrams of fig. 1. The $`IB_3`$ diagram gives the so called contact term, introduced to ensure gauge invariance (see for example ). For the moment, we consider only tree diagrams. At the first order in perturbation theory, we retain terms of all powers in the lepton mass. We will argue that the manifestation of the axial anomaly is strictly connected to these terms.
The Inner Bremsstrahlung amplitude is :
$`M_{IB}`$ $`=`$ $`{\displaystyle \underset{i=1}{\overset{3}{}}}IB_i`$ (8)
$`=`$ $`ie{\displaystyle \frac{G}{\sqrt{2}}}f_\pi V_{ud}m_l\overline{u}(p_\nu )(1+\gamma _5)\left[{\displaystyle \frac{pϵ}{pk}}{\displaystyle \frac{\overline{)}k\overline{)}ϵ+2p_lϵ}{2p_lk}}\right]v(p_l,s_{lR}),`$
where $`u`$ and $`v`$ are the Dirac spinors for the neutrino and the lepton, $`p`$ is the pion momentum, $`p_l`$ and $`s_{lR}`$ are the momentum and the polarization vector of the right-handed lepton, $`k`$ and $`ϵ`$ are the momentum and the polarization vector of the photon, respectively.
We see that $`M_{IB}`$ is proportional to the lepton mass $`m_l`$, thus the decay rate is proportional to $`m_l^2`$. As we have said above, this factor is a consequence of the structure of the weak current. We remove it by normalizing the radiative decay rate with respect to the non radiative one.
The differential Inner Bremsstrahlung contribution for right-handed lepton is:
$`{\displaystyle \frac{1}{\mathrm{\Gamma }_0}}{\displaystyle \frac{\mathrm{d}\mathrm{\Gamma }_{IB}^R}{\mathrm{d}y}}={\displaystyle \frac{\alpha }{4\pi }}{\displaystyle \frac{1}{(1r)^2}}{\displaystyle \frac{1}{A(1y+r)}}\{2A[1+y^22A+r(2A+r6)]`$
$`+\left[(A+2r)(1+r)^2+Ay(y4r)+2ry(1+y)y(1+y^2+5r^2)\right]\mathrm{ln}{\displaystyle \frac{y+A}{yA}}`$
$`+(1y+r)^2(y2rA)\mathrm{ln}{\displaystyle \frac{y+A2}{yA2}}\}.`$ (9)
The dimensionless variable $`y`$ is defined as
$$y=\frac{2E_l}{m_\pi }$$
and
$$r=\frac{m_l^2}{m_\pi ^2},A=\sqrt{y^24r}.$$
The physical region for $`y`$ is:
$$2\sqrt{r}y1+r.$$
(10)
We see that, in eq. (9), there is a term independent of the lepton mass, the one in the first square brackets. Owing to this term, the differential decay rate doesn’t vanish in the limit $`m_l0`$. We have:
$$\frac{1}{\mathrm{\Gamma }_0}\frac{\mathrm{d}\mathrm{\Gamma }_{IB}^R}{\mathrm{d}y}\frac{\alpha }{2\pi }(1y)\mathrm{as}r0.$$
(11)
This indicates that there occurs an helicity flip and thus a chirality non conservation in the limit of zero lepton mass. According to the interpretation given above, this term can be interpreted as connected to the axial anomaly. It corresponds to the anomalous term present in the divergence of the axial current.
Since the radiative process with the right-handed $`l^+`$ is forbidden in the limit $`m_l0`$ by the chiral invariance of the massless QED Lagrangian, the result different from zero, in this limit, indicates the occurrence of a cancellation mechanism, analogous to the one acting in the absorbitive part of the triangular diagram. Indeed, to obtain the decay rate in eq. (9), we have integrated over the emission angle terms containing the lepton propagator squared; this produces power collinear singularities, exactly cancelling the chiral suppression factor $`m_l^2`$ and thus producing the term related to the axial anomaly. The integration over the lepton propagator produces logarithmic collinear singularities.
## 3 Mass singularities
In this section we discuss the structure of mass singularities and their cancellation in the decay rate for the radiative pion and $`Z^0`$ decays and how the KLN theorem applies to these cases.
### 3.1 The pion case
It is useful to separate the cases of unpolarized, right-handed and left-handed outgoing lepton.
Let us consider the mass singularity cancellation for the familiar case of unpolarized radiative $`\pi ^+`$ decay to the first order in $`\alpha `$. The expression for the Inner Bremsstrahlung contribution, differential with respect to the lepton energy, is given by:
$`{\displaystyle \frac{1}{\mathrm{\Gamma }_0}}{\displaystyle \frac{\mathrm{d}\mathrm{\Gamma }_{IB}}{\mathrm{d}y}}`$ $`=`$ $`{\displaystyle \frac{\alpha }{4\pi }}{\displaystyle \frac{1}{(1r)^2}}{\displaystyle \frac{1}{(1y+r)}}\{4A(r1)+[(1+r)^2+y(y4r)]\mathrm{ln}{\displaystyle \frac{y+A}{yA}}`$ (12)
$``$ $`(1y+r)^2\mathrm{ln}{\displaystyle \frac{y+A2}{yA2}}\}.`$
One can easily see that the eq. (12) is divergent both in the collinear and in the infrared limits. The coefficients of the collinear logarithms don’t go to zero in the limit $`r0`$. There are also infrared divergences, since if we let $`y`$ reach its kinematical limit $`y^{MAX}=1+r`$, corresponding to the photon energy going to zero, the expression (12) diverges.
The decay rate is made free from mass singularities in the ordinary way: the divergences cancellation occurs in the total inclusive decay rate, when we add all the first order contributions to the perturbative expansion, i.e. those relative to real and virtual photon emission.
The diagrams describing the real photon emission were already given in fig. 1; the diagram for the virtual correction is drawn in fig. 2.
The expression for the lepton energy spectrum, including the Inner Bremsstrahlung contribution and the virtual photon one, calculated to the leading order in $`m_l/m_\pi `$, is given by :
$$\frac{1}{\mathrm{\Gamma }_0}\frac{\mathrm{d}\mathrm{\Gamma }}{\mathrm{d}y}=D(y,r)\left[1+\frac{\alpha }{\pi }K_l(y)\right].$$
(13)
$`D(y,r)`$ is the lepton distribution function, given, to the first order in $`\alpha `$, by:
$$D(y,r)=\delta (1y)+\left[\frac{\alpha }{\pi }(L1)+O(\alpha ^2)\right]P^{(1)}(y),$$
(14)
where $`L`$ is the logarithm
$$L=\mathrm{ln}\frac{m_\pi }{m_l},$$
(15)
diverging in the collinear limit; if the lepton is an electron, $`L5.6`$. $`P^{(1)}(y)`$ is the Gribov-Lipatov-Altarelli-Parisi kernel , which can be expressed in the form:
$$P^{(1)}(y)=\frac{1+y^2}{1y}\delta (1y)_0^1dz\frac{1+z^2}{1z}.$$
(16)
$`K_l(y)`$ is a finite term, free from infrared and collinear singularities, which has the expression:
$$K_l(y)=1y\frac{1}{2}(1y)\mathrm{ln}(1y)+\frac{1+y^2}{1y}\mathrm{ln}y.$$
(17)
The differential decay rate to order $`\alpha `$ therefore becomes:
$$\frac{1}{\mathrm{\Gamma }_0}\frac{\mathrm{d}\mathrm{\Gamma }}{\mathrm{d}y}=\delta (1y)+\frac{\alpha }{\pi }(L1)P^{(1)}(y)+\frac{\alpha }{\pi }K_l(y).$$
(18)
To calculate the inclusive decay rate, we have to integrate the expression in eq. (18) over $`y`$; to the leading order in the lepton mass, the physical region for y is $`0y1`$. The kernel $`P^{(n)}`$ has the property that:
$$_0^1dyP^{(n)}(y)=0;$$
(19)
thus, when we calculate the inclusive decay rate, the coefficient of the collinear logarithm vanishes and the resulting expression is finite in the zero mass limit.
Carring out the integration over $`y`$, we obtain the well known inclusive decay rate to order $`\alpha `$
$$\frac{\mathrm{\Gamma }}{\mathrm{\Gamma }_0}=1+\frac{\alpha }{\pi }[\frac{15}{8}\frac{\pi ^2}{3}].$$
(20)
As expected, the expression (20) is finite in the collinear limit and is also free from infrared divergences, because, as usual, the infrared divergences present in the soft photon contribution and in the virtual photon contribution have cancelled each other.
Let us now discuss the mass singularities in the case of the right-handed Inner Bremsstrahlung contribution, given in eq. (9). We have already seen in section 2 that $`\mathrm{d}\mathrm{\Gamma }_{IB}^R/\mathrm{d}y`$ is finite in the limit $`r0`$, i.e. it is free from collinear singularities. In this limit the coefficients of both the collinear logarithms vanish. Indeed, as we have discussed above, only the term related to the axial anomaly survives in the zero mass limit, that is the part of the decay rate indipendent of the lepton mass.
We observe that the right-handed Inner Bremsstrahlung contribution is free from infrared divergences as well. If we make the lepton energy $`y`$ reach its kinematical limit $`y^{MAX}`$, we obtain a finite result:
$`{\displaystyle \frac{1}{\mathrm{\Gamma }_0}}{\displaystyle \frac{\mathrm{d}\mathrm{\Gamma }_{IB}^R}{\mathrm{d}y}}0\mathrm{as}yy^{MAX}.`$ (21)
The result (21) shows that the emission of soft photons does not contribute to the radiative $`\pi ^+`$ decay with a right-handed lepton. Indeed, the soft photon contribution factorizes with respect to the Born decay rate, but this vanishes in the case of right-handed $`l^+`$ (see eq. (6)). Physically, eq.(21) is a consequence of the fact that soft photons don’t carry spin, thus they cannot contribute to the angular momentum balance; therefore the process with the right-handed lepton emitting a soft photon, is forbidden by angular momentum conservation.
For the same reason of angular momentum conservation, in the right-handed case also the virtual contribution is zero. The virtual correction is factorized with respect to the Born decay rate, but, as we have already seen, if the lepton is right-handed, this is identically zero.
In the right-handed case, the mass singularity cancellation occurs through a mechanism different from the one working in the unpolarized decay rate. The infrared and the collinear limits give, separately, a finite result. In particular, the coefficient of the collinear logarithms is the lepton mass, instead of the usual correction factor coming from the soft and the virtual photon contributions, as in eq. (18).
This particular mass cancellation mechanism is the consequence of the combination of two constraints: the angular momentum conservation in the pion vertex and the helicity flip in the photon-lepton vertex.
The situation is completely different if we consider the radiative process with the outgoing left-handed lepton, i.e. the process without helicity flip. The left-handed Inner Bremsstrahlung contribution is given by:
$`{\displaystyle \frac{1}{\mathrm{\Gamma }_0}}{\displaystyle \frac{\mathrm{d}\mathrm{\Gamma }_{IB}^L}{\mathrm{d}y}}={\displaystyle \frac{\alpha }{4\pi }}{\displaystyle \frac{1}{(1r)^2}}{\displaystyle \frac{1}{A(1y+r)}}\{2A[r(2Ar+6)y^212A]`$
$`+\left[(A2r)(1+r)^2+Ay(y4r)2ry(1+y)+y(5r^2+y^2+1)\right]\mathrm{ln}{\displaystyle \frac{y+A}{yA}}`$
$`+(1y+r)^2(2rAy)\mathrm{ln}{\displaystyle \frac{y+A2}{yA2}}\}`$ (22)
The expression (22) contains collinear singularities, since the coefficients of the collinear logarithms don’t vanish in the limit $`r0`$. It is also infrared divergent, as one can verify by taking the limit $`yy^{MAX}`$. In this case we don’t have the constraint constituted by the helicity flip in the photon-lepton vertex and in the pion vertex the angular momentum is conserved for soft and virtual photon emission. Thus, the mass singularity cancellation mechanism occurs in the ordinary way, as in the unpolarized case, i.e. in the total inclusive decay rate, obtained by adding all the order $`\alpha `$ contributions.
Let us show how the cancellation takes place. As we have already seen (eq. (6)), in the Born $`\pi ^+`$ decay the outgoing lepton is left-handed, due to angular momentum conservation. Thus the Born decay rates with unpolarized and left-handed $`l^+`$ coincide:
$$\mathrm{\Gamma }_0^L=\mathrm{\Gamma }_0.$$
(23)
Because of the factorization with respect to the Born decay rate, also the unpolarized and the left-handed virtual contributions are equal:
$$\mathrm{\Gamma }_v^L=\mathrm{\Gamma }_v.$$
(24)
Expressing the left-handed Inner Bremsstrahlung contribution in terms of the unpolarized and the right-handed ones, the total contribution to order $`\alpha `$ to the left-handed process is given by:
$$\mathrm{\Gamma }_{TOT}^L=(\mathrm{\Gamma }_0+\mathrm{\Gamma }_v+\mathrm{\Gamma }_{IB})\mathrm{\Gamma }_{IB}^R.$$
(25)
The expression (25) is finite both in the infrared and in the collinear limit, because the mass singularities present in the terms between brackets cancel each other, as we have seen (eq. (20)) and $`\mathrm{\Gamma }_{IB}^R`$ is free from mass singularities.
The presence of mass singularities is a conseguence of the fact that the states of a theory containing massless particles are highly degenerate. The KLN theorem states that the mass singularities disappear from the transition probability when we average it over the ensemble of degenerate states. It results that imposing on the outgoing lepton a polarization opposed to the one prescribed by the interaction taking place before the photon emission, implies a reduction of the degeneration subspace . Thus in this case we have a particular application of the KLN theorem.
### 3.2 The $`Z^0`$ case
Let us start the discussion on the $`Z^0`$ radiative decay by saying that it differs from the pion case. For radiative pion decay, due to the angular momentum conservation in the pion vertex, there is no room for a right-handed lepton. For such a channel, soft and virtual photon contribution give a contribution equal to zero. This result is valid independently of the lepton mass. Let us now consider a more general process, by loosing the value of the angular momentum of the decaying state. As an example, we study the radiative $`Z^0`$ decay in a lepton-antilepton $`(l^{}l^+)`$ pair, in which the lepton is in a definite helicity state.
The $`Z^0`$-leptons vertex is:
$$i\frac{M_Z}{\sqrt{2}}\left(\frac{G}{\sqrt{2}}\right)^{1/2}\gamma _\mu (g_vg_a\gamma _5),$$
with
$$g_v=14\mathrm{sin}\theta _{W}^{}{}_{}{}^{2}g_a=1$$
where $`\theta _W`$ is the Weinberg angle and $`M_Z`$ is the $`Z^0`$ mass.
If we set $`g_v=g_a=1`$, we require that in the limit of zero lepton mass, the $`Z^0`$ couples to a left-handed lepton. We calculate the decay rate for the process in which the outgoing lepton is right-handed.
At the Born level this is given by:
$`\mathrm{\Gamma }_0^R`$ $`=`$ $`{\displaystyle \frac{GM_Z^3}{48\sqrt{2}\pi }}\left\{\sqrt{14r}\left[(g_v^2+g_a^2)(1r)+3r(g_v^2g_a^2)\right]2g_vg_a(14r)\right\}.`$ (26)
If in eq. (26) we set $`g_v=g_a=1`$, $`\mathrm{\Gamma }_0^R`$ vanishes in the chiral limit, since there isn’t the term related to the axial anomaly.
Let us now study the decay process with the lepton emitting a real photon (see fig. 3)
and evaluate the probability that the outgoing lepton is right-handed. The electromagnetic interaction doesn’t couple states with different chirality, hence we expect the decay rate to vanish for $`m_l0`$.
The decay rate for the process described by the diagram of fig. 3, differential with respect to the lepton energy, is given by the expression:
$`{\displaystyle \frac{\mathrm{d}\mathrm{\Gamma }_\gamma ^R}{\mathrm{d}y}}={\displaystyle \frac{G\alpha M_Z^3}{96\sqrt{2}\pi ^2}}\{{\displaystyle \frac{(y2)(1y)^2}{4(1y+r)^2}}[(g_v^2+g_a^2)A+2g_vg_a(2ry)]`$
$`+`$ $`{\displaystyle \frac{(1y)}{2(1y+r)}}\left[(g_v^2+g_a^2)A(2ry)+2g_vg_a(y^22r)\right]`$
$`+`$ $`{\displaystyle \frac{2}{(y1)}}\left[2(g_v^22g_a^2)Ar+(g_v^2+g_a^2)A+2g_vg_a(4ry^2+y1)\right]`$
$`+`$ $`[g_v^2+g_a^2+2g_vg_a{\displaystyle \frac{1}{A}}(4r^2+r(yy^2+1)y)][\mathrm{ln}{\displaystyle \frac{y+A}{yA}}\mathrm{ln}{\displaystyle \frac{y+A2}{yA2}}]\}.`$
Here
$$r=\frac{m_l^2}{M_Z^2}$$
and $`y`$ is the usual dimensionless variable:
$$y=\frac{2E_1}{M_Z}$$
where $`E_1`$ is the lepton energy and
$$A=\sqrt{y^24r}.$$
The physical region for $`y`$ is
$$2\sqrt{r}y1.$$
(28)
The result obtained, as given by the emission of the photon by a single leg, is gauge dependent. To have a gauge independent amplitude, the contribution of the diagram b) of fig. 4 must be added . For the purpose of the polarized amplitude, however, the helicity flip contribution of the diagram b) of fig. 4 gives zero in the massless limit and is therefore negligible in our discussion.
From now on we set $`g_v=g_a=1`$, to have the condition of chirality conservation in the $`Z^0`$ vertex for $`m_l0`$. Taking this limit in eq. (LABEL:Rz), we see that $`\mathrm{d}\mathrm{\Gamma }_\gamma ^R/\mathrm{d}y`$ does not vanish:
$$\frac{\mathrm{d}\mathrm{\Gamma }_\gamma ^R}{\mathrm{d}y}\frac{G\alpha M_Z^3}{24\sqrt{2}\pi ^2}(1y)\mathrm{as}r0,g_v=g_a=1.$$
(29)
The result of this limit is the contribution related to the axial anomaly, which has the same form of the one found in the pion case.
Let us now discuss the mass singularity cancellation mechanism for the $`Z^0`$ radiative decay case. Eq. (29) shows that the collinear limit gives a finite result. Thus, we conclude that, as in the case of the radiative pion decay with right-handed lepton, the collinear and infrared limits are disconnected.
If we keep the lepton mass different from zero, the chirality is not fixed by the interaction occurring before the photon emission, even if we set $`g_v=g_a=1`$. Thus, for $`m_l0`$, the soft and virtual photon contributions are different from zero and diverge in the infrared limit. If we let the lepton energy reach its kinematical limit, $`y^{MAX}=1`$, we see that $`\mathrm{d}\mathrm{\Gamma }_\gamma ^R/\mathrm{d}y`$ diverges. We expect the infrared divergences to cancel, if we add all the first order contributions, given by the diagrams of fig. 3 and 4
and calculate the totally inclusive decay rate.
If we take the limit $`y1`$ in eq. (29), we obtain:
$$\frac{\mathrm{d}\mathrm{\Gamma }_\gamma ^R}{\mathrm{d}y}0\mathrm{as}m_l0,y1\mathrm{and}g_v=g_a=1.$$
(30)
Eq. (30) indicates that, in the massless limit, the soft photon contribution is zero. Indeed, it is factorized with respect to the Born decay rate, which, for $`r0`$ and $`g_v=g_a=1`$, vanishes. As we have discussed in section 1, the presence of the term related to the axial anomaly is directly connected to the emission of the photon, hence it vanishes in the limit in which the photon is not radiated.
The virtual photon contribution vanishes in the zero mass limit, as well. Indeed, it is factorized respect to the Born decay rate, which goes to zero as $`r0`$. The virtual correction factor can produce only a logarithmic collinear singularity, not a power-like one, needed for the cancellation of the chiral suppression.
In the massless limit, also the diagram with the photon emitted by $`l^+`$ doesn’t contribute, since, as the virtual contribution, cannot produces a power type collinear singularity.
Taking the infrared limit $`x0`$ in eq. (3), gives a finite result (indeed the cross section vanishes). This is a consequence of the fact that the cross section (3) has been calculated to the leading order in the lepton mass. From the eq. (LABEL:Rz), we see that the infrared divergent term is the one given by the term:
$$\frac{4r}{(y1)}\left[(g_v^22g_a^2)A+4g_vg_a\right]$$
i.e., it is proportional to the lepton mass. Performing the calculation, neglecting the mass terms as done in ref. , means imposing the chirality conservation law in the $`Z^0`$ vertex; thus the soft photon contribution is zero and the infrared divergences disappear. To the leading order in the lepton mass, one has (as in eq. 3) only the term related to the axial anomaly, which vanishes in the infrared limit.
## 4 Conclusions
We have seen that the decay rates for processes with a fermion changing helicity by emitting a massless vector boson don’t vanish in the chiral limit. We have interpreted this behaviour as related to the axial anomaly, as first suggested by Dolgov and Zakharov.
In the processes we have considered, the cancellation of the collinear singularities occurs through a mechanism different from the one for the unpolarized case. However, we have found a difference between the pion case and the more general case of the $`Z^0`$ decay. Due to the angular momentum conservation, the virtual and soft photon contributions are zero, even if the lepton mass is kept different from zero, giving finite infrared and collinear limits.
In the case of the radiative $`Z^0`$ decay, the decay rate is not finite in the infrared limit, since, for $`m_l0`$, the real soft photon contributions are different from zero. We observe that, in order to make the collinear limit finite, it is sufficient to sum over the set of degenerate states containing the fermion making the helicity flip accompanied by a hard collinear photon .
In these processes the collinear limit results disconnected from the infrared one, in the sense that the virtual photon contribution is not needed to render this limit finite.
We conclude that the collinear singularity cancellation mechanism is controlled by the anomalous breaking of the chiral symmetry. The axial anomaly implies that taking the collinear limit gives a finite result, independent of the fermion mass.
These results can be applied to interpret the mass singularity cancellation mechanism for any polarized process. The extension of this approach to the case of Quantum Cromodynamics for polarized processes is under study.
Acknowledgments
We wish to thank E. Kuraev and V. Fadin for valuable comments and S. Forte, J. Kodaira and L. Lipatov for useful discussions.
|
warning/0003/astro-ph0003209.html
|
ar5iv
|
text
|
# A Very Fast and Momentum-Conserving Tree Code
## 1 Introduction
The tree code (cf. Barnes & Hut 1986, hereafter B&H) has become an invaluable tool for the approximate but fast computation of the forces in studies of collisionless gravitational dynamics. It has been applied to a large variety of astrophysical problems. The gravitational potential generated by $`N`$ bodies of masses $`\mu _n`$ and at positions $`𝑿_n`$ is
$$\mathrm{\Phi }(𝑿)=\underset{n=1}{\overset{N}{}}\mu _ng(|𝑿𝑿_n|),$$
(1)
where $`g(r)`$ denotes the greens function, i.e. for un-softened gravity $`g(r)=G/r`$. The essence of the tree code is to approximate this sum over $`N`$ terms by replacing any partial sum over all bodies within a single cell which is well-separated from $`𝑿`$ by just one term. The inner structure of the cell is partly taken into account using its multipole moments. This method reduces the overall costs for the computation of all forces from $`𝒪(N^2)`$ to $`𝒪(N\mathrm{log}N)`$.
The tree code, however, does not exploit the fact that the force due to the contents of some cell is very similar at nearby positions (even though one may use the fact that nearby bodies tend to have very similar interaction lists, cf. Barnes 1990). Exploiting this is the idea of the fast multipole method (FMM) (Greengard & Rokhlin, 1987). The FMM employs a (usually) non-adaptive structure of hierarchical grids and considers only interactions between nodes on the same grid level according to their geometrical neighbourhood. The gravitational field due to some source cell and within some sink cell is approximated by a multipole expansion in spherical harmonics, the order of which is adapted to meet predefined accuracy limits. This method has been claimed to reduce the overall amount of operations to $`𝒪(N)`$, but the tables given by Cheng, Greengard & Rokhlin (1999) do not support this claim. Capuzzo-Dolcetta & Miocchi (1998) find that the FMM needs $`𝒪(N\mathrm{log}N)`$ operations, and is significantly slower for astrophysical applications than the tree code at comparable accuracy.
Instead of using a spherical multipole expansion of adaptive order, it is actually more efficient to use a Cartesian expansion of fixed order. Moreover, by preserving the symmetry of the gravitational interaction for mutual cell-cell interactions, one can (i) reduce the computational effort and (ii) obtain a code that satisfies Newton’s third law by construction and hence results in exact conservation of momentum, a property not shared by the traditional tree code.
## 2 Description of the Code
We start as the B&H tree code with a hierarchical tree of cubic cells. Each cell has up to eight sub-nodes corresponding to its octants. A node can be either a single body or another cell. The tree-building phase (cf. B&H) also includes the computation of the cells’ masses, centers $`𝒁`$ of mass, and quadrupole moments.
### 2.1 The Opening Criterion
In order to benefit from the symmetry of the gravitational interaction, the opening criterion, which decides whether or not two nodes are well-separated so that a direct mutual interaction is acceptable, must be symmetric, too. We employ an extension of the criterion used in the tree-code: nodes A and B are well-separated if
$$|𝒁_\mathrm{A}𝒁_\mathrm{B}|>(r_{\mathrm{max}\mathrm{A}}+r_{\mathrm{max}\mathrm{B}})/\theta ,$$
(2)
where the opening angle $`\theta `$ controls the accuracy of the code. $`r_{\mathrm{max}}`$ is the radius of a sphere centered on the node’s center of mass and encircling all bodies within it. Bodies naturally have $`r_{\mathrm{max}}0`$, i.e. two bodies are always well separated, while for the interaction between a body and a cell the criterion (2) reduces to that used in the tree code. Note that if one additionally to equation (2) requires $`r_{\mathrm{max}\mathrm{A}}=0`$, the standard tree code is re-covered, but the symmetry between A and B is broken.
There exist two upper limits for the radius $`r_{\mathrm{max}}`$. One is the distance $`b_{\mathrm{max}}`$ between the cell’s center of mass, $`𝒁`$, and its most distant corner (Salmon & Warren, 1994). The other is
$$\underset{\mathrm{sub}\mathrm{nodes}i}{\mathrm{max}}\left\{r_{\mathrm{max}i}+|𝒁_i𝒁|\right\}$$
(3)
(Benz et al., 1990). After computation of both these upper limits, we take the smaller one to be $`r_{\mathrm{max}}`$. For cells with only a few bodies like cell A in LABEL:example, the latter often gives values significantly smaller than $`b_{\mathrm{max}}`$, while for cells with many bodies, like cell B in LABEL:example, $`b_{\mathrm{max}}`$ is the tighter limit.
### 2.2 Approximating Gravity
Consider two bodies at $`𝑿`$ and $`𝒀`$ which reside in two well-separated cells A and B with centers of mass at, respectively, $`𝒁_\mathrm{A}`$ and $`𝒁_\mathrm{B}`$ and separation $`𝑹𝒁_\mathrm{A}𝒁_\mathrm{B}`$. We may re-write $`𝑿𝒀=𝑹+(𝒙𝒚)`$ with $`𝒙𝑿𝒁_\mathrm{A}`$ and $`𝒚𝒀𝒁_\mathrm{B}`$ being small in magnitude compared to $`𝑹`$ (because the cells are well-separated, cf. LABEL:example). The Taylor expansion of $`g(|𝑿𝒀|)`$ around $`𝑹`$ reads
$$g(|𝑿𝒀|)=\underset{p}{}\frac{1}{p!}\left[\left[(𝒙𝒚)\mathbf{}\right]^pg(|𝒓|)\right]_{𝒓=𝑹}.$$
(4)
Separating powers of $`x`$ from powers of $`y`$ in equation (4) and subsequently taking the mass weighted sum over cell B yields a Cartesian multipole expansion of the potential $`\mathrm{\Phi }_{\mathrm{B}\mathrm{A}}(𝑿)`$ at any position $`𝑿`$ within cell A and due to all bodies inside cell B (Warren & Salmon, 1995). Since $`𝒁_\mathrm{B}`$ was chosen to be the center of mass of cell B, its dipole vanishes. The highest-order multipole occurring in such an expansion may actually be omitted, since it only contributes a constant to the approximation for $`g`$ and does not affect the approximation for $`\mathbf{}g`$ and hence the force. The expression of third order thus reads (without the octopole; using Einstein’s sum convention)
$$\begin{array}{cccc}\hfill \mathrm{\Phi }_{\mathrm{B}\mathrm{A}}(𝑿)& M_\mathrm{B}\{\hfill & & 𝒟^{(0)}+\frac{1}{2}\stackrel{~}{Q}_{\mathrm{B}ij}𝒟_{ij}^{(2)}\hfill \\ & & +& x_i\left[𝒟_i^{(1)}+\frac{1}{2}\stackrel{~}{Q}_{\mathrm{B}jk}𝒟_{ijk}^{(3)}\right]\hfill \\ & & +& \frac{1}{2}x_ix_j𝒟_{ij}^{(2)}+\frac{1}{6}x_ix_jx_k𝒟_{ijk}^{(3)}\},\hfill \end{array}$$
(5)
where $`M_\mathrm{B}`$ and $`\stackrel{~}{Q}_\mathrm{B}`$ are the mass and specific quadrupole moment
$$\stackrel{~}{Q}_{\mathrm{B}ij}\frac{1}{M_\mathrm{B}}\underset{𝒚_n+𝒁_\mathrm{B}\mathrm{cell}\mathrm{B}}{}\mu _ny_{ni}y_{nj}$$
(6)
of cell B, while $`𝒟^{(n)}\mathbf{}^ng(r)|_{r=|𝑹|}`$, i.e.
$$\begin{array}{ccc}\hfill 𝒟^{(0)}& =& D^0,\hfill \\ \hfill 𝒟_i^{(1)}& =& R_iD^1,\hfill \\ \hfill 𝒟_{ij}^{(2)}& =& \delta _{ij}D^1(R)+R_iR_jD^2,\hfill \\ \hfill 𝒟_{ijk}^{(3)}& =& \left(\delta _{ij}R_k+\delta _{jk}R_i+\delta _{ki}R_j\right)D^2+R_iR_jR_kD^3\hfill \end{array}$$
(7)
with
$$D^n\left(\frac{1}{r}\frac{}{r}\right)^ng(r)|_{r=|𝑹|}.$$
(8)
The symmetry between $`𝒙`$ and $`𝒚`$ at every order of the Taylor expansion in equation (4) has two important consequences. First, if this expansion is used to compute both $`\mathrm{\Phi }_{\mathrm{B}\mathrm{A}}(𝑿)`$ and $`\mathrm{\Phi }_{\mathrm{A}\mathrm{B}}(𝒀)`$, Newton’s third law is satisfied by construction. Note, that our omission of the octopole term broke the symmetry only in the zeroth order and has no effect on the forces.
Second, the expressions for $`M_\mathrm{A}\mathrm{\Phi }_{\mathrm{B}\mathrm{A}}`$ and $`M_\mathrm{B}\mathrm{\Phi }_{\mathrm{A}\mathrm{B}}`$ are very similar: the expansion coefficients of second and third order differ only by mere signs, such that computing these coefficients for both Taylor series at one time is substantially faster than computing them at different times.
The transformation, or shifting, of the expansion center to some other position is trivial compared to the analogous procedure in the FMM (see Cheng et al.).
### 2.3 The Algorithm
The standard tree code computes the forces on each body by a recursive tree walk, which visits each node exactly once as a gravity sink, and thus exhibits an inherent asymmetry between sources and sinks. The new algorithm avoids this asymmetry.
First, in the interaction phase, the Taylor series coefficients are evaluated and accumulated in data fields associated with each node. This phase is based on the concept of mutual interactions (MIs), pairs of nodes, A and B, such that bodies in node A must receive forces from all bodies in node B, and vice versa. We start by the MI describing the root-root self-interaction, and process a given MI as follows. (1) A body self-interaction is ignored; (2) a cell self-interaction is split into the MIs between the sub-nodes<sup>1</sup><sup>1</sup>1 In a cubic oct-tree, these are at most 36 independent sub-MIs., and the process is continued on each of the new MIs; (3) a MI representing a well-separated pair of nodes is executed: the Taylor coefficients are computed and added to the nodes’ corresponding data fields; (4) finally, in any other case, the node with larger $`r_{\mathrm{max}}`$ is split, and up to eight new MIs are created and processed.
Secondly, in the collection phase, the Taylor coefficients are passed down the tree: the expansion center is shifted to the center of mass of the currently active cell and the coefficients are accumulated. The Taylor expansion is evaluated at the position of any body and the values for potential and acceleration are added to its data fields (which may already contain contributions accumulated during the interaction phase).
## 3 Performance Tests
We tested the new algorithm and compared it with the tree code in three typical astrophysical situations: (1) a spherical Plummer model, representing a rather homogeneous stellar system, (2) a spherical Hernquist-model (1990) galaxy, and (3) a group of five such galaxies with various masses and scale radii. We generated $`10^5`$ random initial positions from each of these cases, truncating the density at 1000 scale radii, and evaluated the exact mutual forces at all positions and the approximated forces due to the tree code (up to quadrupole order) and the new code for opening angles $`\theta `$ between 0.2 and 1. We used an optimally chosen softening with the biweight softening kernel (see Dehnen 2000), but the results are insensitive to these settings. Both approximate methods have been coded by the author<sup>2</sup><sup>2</sup>2 At the same $`\theta `$, the tree code is twice as fast as a code publicly available from J. Barnes, mainly because the new opening criterion leads to fewer interactions. However, even at the same number of interactions, the author’s code was about 30% faster. and use the same opening criterion (§2.1). In order to measure the accuracy of the approximated forces, we evaluated (cf. Capuzzo-Dolcetta & Miocchi 1998)
$$\epsilon _n=|a_na_n^{PP}|/a_n^{PP},$$
(9)
where $`a_n`$ denotes the magnitude of the acceleration of the $`n`$th body due to either of the approximate methods and $`a_n^{PP}`$ that of the exact computation.
Figure 2 plots the CPU time needed for the force approximation on a Pentium III/500Mhz PC versus the mean relative error and that at the 99 percentile. For the new code, the time consumption scales almost inversely with the error, while the tree code flattens off<sup>3</sup><sup>3</sup>3 This is, because of $`r_{\mathrm{max}}`$ is not proportional to the cell size, so that increasing $`\theta `$ from $`0.7`$ does not much decrease the number of interactions. at $`\theta 0.7`$. Evidently, at an acceptable level of accuracy, e.g. $`\epsilon _{99\%}=0.01`$, the new code is about four times faster than the tree code, even though it requires a smaller value of $`\theta `$ (0.5 as compared to 0.7 for the tree code).
For the test case of the Hernquist model, Figure 3 plots the CPU time consumption per body versus $`N`$ at fixed $`\theta =0.7`$ for the tree code and $`0.5`$ for the new code. The tree code shows the well-known $`\mathrm{log}N`$ scaling, while for $`N10^5`$ the new code requires only a constant amount of CPU time per body. This can be explained as follows. Arranging eight root cells to a new root box, increases $`N`$ to $`8N`$ and the number $`N_I`$ of interactions to $`8(N_I+N_+)`$, where $`8N_+`$ interactions are needed to compute the forces between the former root cells. Thus,
$$\frac{\mathrm{d}N_I}{\mathrm{d}N}\frac{N_I}{N}\frac{\mathrm{\Delta }\mathrm{log}N_I}{\mathrm{\Delta }\mathrm{log}N}=\frac{N_I+N_+/\mathrm{ln}8}{N}.$$
(10)
In the tree code, $`N_+N`$ yielding $`N_IN\mathrm{log}N`$. In the new code, $`N_+`$ may be estimated to contain two contributions, a constant term accounts for the interactions with distant nodes, and a term $`N^{2/3}`$ for those on the surface of the former root cells. Inserting this into equation (10) yields
$$N_INc_1N^{2/3}c_2$$
(11)
with constants $`c_1`$ and $`c_2`$ that depend on $`\theta `$. Thus, at large $`N`$, a linear relation is approached. Note that this argument differs from that given for the FMM by Greengard & Rokhlin (1987), who assumed that the resolution may remain fixed when increasing $`N`$.
## 4 Discussion
A new code for the approximate evaluation of gravitational forces has been presented, tested, and compared to the tree code. This new code is substantially faster than the tree code. Moreover, unlike the latter, it satisfies Newton’s third law by construction, such that any $`N`$-body code based on it will not introduce spurious net-accelerations. The new code is based on a Taylor expansion of the greens function in Cartesian coordinates and incorporates mutual cell-cell interactions. The simple algorithm is well suited for implementation on parallel computers: different mutual interactions (MIs) can be passed to different CPUs.
The scaling of the CPU time required for the mutual forces of a number $`N`$ of bodies becomes essentially linear at $`N10^5`$, so that with ever larger $`N`$ the new code is increasingly faster than the tree code, allowing for a substantial improvement in simulations employing large number of bodies. The only disadvantage is the increased requirement of memory compared to the standard tree code: 20 floating point numbers per cell are needed to hold the Taylor expansion coefficients. (By using a tree-walking algorithm instead of that given in §2.3, one can avoid this at the price of enhanced CPU time consumption.)
In spirit, the new code is similar to Greengard & Rokhlin’s (1987) fast multipole method, but is more efficient because it (i) uses a Cartesian instead of a spherical harmonic multipole expansion and (ii) fixes the order of the expansion while controlling the accuracy via the interaction condition, rather than fixing the interactions and adapting the expansion order to the accuracy.
A concern with codes based on cell-cell interactions is their performance in the presence of individual time steps. Clearly, when not all the forces are to be computed, such codes fare less favorably. However, when the forces for all bodies within some domain are desired, the new code is still a significant improvement over the tree code.
The new code has been written in C++ and will be electronically available from the author upon request.
The author thanks Tom Quinn for valuable discussions and Joshua Barnes, Lars Hernquist, Junichiro Makino, Andy Nelson, and Thorsten Naab for useful comments.
|
warning/0003/cond-mat0003288.html
|
ar5iv
|
text
|
# First-principles calculations of the self-trapped exciton in crystalline NaCl
## Abstract
The atomic and electronic structure of the lowest triplet state of the off-center (C$`_{\text{2v}}`$ symmetry) self-trapped exciton (STE) in crystalline NaCl is calculated using the local-spin-density (LSDA) approximation. In addition, the Franck-Condon broadening of the luminescence peak and the a$`{}_{1g}{}^{}\text{b}_{3u}`$ absorption peak are calculated and compared to experiment. LSDA accurately predicts transition energies if the initial and final states are both localized or delocalized, but 1 eV discrepancies with experiment occur if one state is localized and the other is delocalized.
Unlike a molecule, an extended system such as a solid can support both spatially localized and delocalized single particle states and excitations. Physical properties such as luminescence can differ dramatically depending on which type of state occurs. Deciding theoretically whether a localized or delocalized solution exists is a challenging problem . Here we examine NaCl, a classic example where electronic excitations self-localize by coupling to the lattice, creating local lattice distortions. Because the degree of localization will affect the Coulomb energies, approaches that incompletely cancel the self-interaction contribution to the exchange energy (e.g., the local density approximation) sometimes fail to predict the actual localized solution.
The ground state of alkali halides with one electron removed is the V<sub>K</sub> center : the resulting hole does not delocalize at the top of the valence band, but rather (symmetrically) attracts two Cl<sup>-</sup> ions into a tightly bound molecule , effectively becoming a Cl$`{}_{}{}^{}{}_{2}{}^{}`$ molecular ion. The local symmetry of this atomic configuration is D$`_{\text{2h}}`$. An excess electron in bulk NaCl forms a large mobile (Fröhlich) polaron , but in the presence of a self-trapped hole forms a self-trapped exciton (STE). It has been suggested that the self-trapped exciton state breaks symmetry and sits off-center with C$`_{\text{2v}}`$ symmetry.
Although there have been many theoretical studies of the self-trapped exciton and V<sub>K</sub> center problems, no density-functional calculations have been reported. In this letter we report local-spin-density approximation (LSDA) calculations of the STE and V<sub>K</sub> center. The advantage of LSDA calculations is that they provide one of the simplest tools capable of providing a realistic model of this competition. In both cases, our LSDA calculations give undistorted, delocalized solutions with lower energy than the self-trapped solution, contrary to experiment. For the V<sub>K</sub> center, no metastable local minimum trapped solution was found; however, for the (neutral) STE we find locally metastable solutions, with the on-center STE 0.14 eV higher in energy than the off-center STE solution, which in turn is higher by 0.20 eV than a free electron-hole pair. Even with this discrepancy in the total energy, the atomic positions for the STE solution (see Table I) are reasonable and agree well with Hartree-Fock second-order Møller-Plesset (MP2) perturbation theory . We focus on the properties of the local minimum solution for the off-center STE, which provides a test of the ability of density functional methods to treat the strong coupling of electronic and lattice degrees of freedom, and calculate the spectral properties of excited states.
To solve the LSDA equations we use a plane wave pseudopotential method with a spin-dependent exchange-correlation potential , and full structural relaxation in a supercell approach. For most calculations, we used a 32 atom supercell with translation vectors (2,0,0), (0,2,0) and (1,1,1), giving a nearest neighbor distance between STEs of 9.4 Å, and used four special k-points in the irreducible wedge for the Brillouin zone integrations. Tests varying the number of k-points and supercell size suggest that these parameters are adequate . One of the in $`xy`$ plane nearest Cl and two Na atoms are most displaced while the rest of the atoms move by a much smaller amount. Some calculated structural parameters for the STE are given in Table I.
The calculated atomic displacements from the ideal NaCl structure for the off-center STE are shown in Fig. 1. The b<sub>3u</sub> hole state in the triplet STE is localized on the Cl$`{}_{}{}^{}{}_{2}{}^{}`$ molecule (Fig. 1a), with nearly equal weight on the two Cl ions. The last spin up electron (a<sub>1g</sub>) is mostly localized on the (1/2,1/2,0) vacant halogen site (Fig. 1b), as in the case of the F-center. The formation of the Cl$`{}_{2}{}^{}{}_{}{}^{}`$ “molecule” in the STE is mainly due to the shift of a single Cl. This asymmetric shift can be rationalized by noting that the Madelung energy (with canonical charges of $`\pm `$1 for Na and Cl) of the D$`_{\text{2h}}`$ configuration is about 0.17 eV higher than the C$`_{\text{2v}}`$ one, i.e., the ionic Madelung terms that favor the rock salt structure in the first place favor keeping one of the Cl ions on a lattice site; in addition, the extra electron in the STE (compared to the $`V_K`$ center) can lower its energy by this distortion.
To obtain vibrational properties of the off-center STE, we made finite displacements from the equilibrium geometry. The Cl$`{}_{}{}^{}{}_{2}{}^{}`$ stretching mode $`\omega _{\text{str}}=242\text{cm}^1`$ is smaller than the experimental Raman frequency 361 cm<sup>-1</sup> . ¿From the force matrix associated with the Cl$`{}_{}{}^{}{}_{2}{}^{}`$ stretching mode, we found that only the two neighboring Na ions at ($`\frac{1}{2}`$,1,0) and (1,$`\frac{1}{2}`$,0) couple significantly. Unlike H center calculations , coupling to the Na atoms yields only a small shift in the frequency of the Cl$`{}_{}{}^{}{}_{2}{}^{}`$ mode to $`\omega _{\text{str}}=234\text{cm}^1`$. Since the LSDA places the STE too high in energy, it is not too surprising that the curvature of the STE local minimum is underestimated,
Figure 2 shows the computed density of states (DOS) of the perfect NaCl and the inverse participation ratio (IPR=$`\text{a}_0^3/8|\mathrm{\Psi }_i(\stackrel{}{r})|^4d^3r`$) for states of majority and minority spin for the off-center STE. (The IPR is a measure of the localization of a state.) The localized hole b<sub>3u</sub> and electron a<sub>1g</sub> states lie in a gap of about 6 eV between the conduction and valence bands of the perfect crystal, where as usual the LSDA underestimates the gap. Rather than using the single-particle eigenvalues, we obtain estimates of the excitation energies as the difference between total energies of different electronic configurations. To calculate the energy to create a free electron-hole pair (the gap energy), we occupy spin up states with one extra electron, while spin down states have one empty state. The energy difference between the two solutions (6.44 eV) for the same atomic positions should correspond to the free electron-hole pair (experimental value 7.96 eV ).
In the distorted STE solution, the energy of electron-hole pair recombination was found by comparing energies of the STE solution with the energy of NaCl having the same atomic displacements as that of the STE. This energy of 4.25 eV is roughly the same as the energy difference between a<sub>1g</sub> and b<sub>3u</sub> states given by the DOS, and compares to a value of 3.35 eV obtained from a luminescence experiment at 11 K. The distorted NaCl, with a lattice distortion energy of 2.5 eV relative to the ideal NaCl positions, is in a highly excited vibrational state. Thus, significant Franck-Condon effects in the spectral properties of the STE are expected. To qualitatively describe the luminescence, a ground state potential curve was calculated for the configuration coordinate $`\alpha `$. It was assumed that all atoms move back to the perfect crystal positions proportionally to their distortions in the STE solution. The result is shown in Fig. 3, along with a quadratic fit $`k_a\alpha ^2/2`$ with one adjustable parameter $`k_a`$. The effective one-dimensional Schrödinger equation for $`\mathrm{\Psi }(\alpha )`$ describes a harmonic oscillator with frequency $`\omega _a=\sqrt{k_a/I}=109\text{cm}^1`$. The moment of inertia $`I=M_n\delta \stackrel{}{R}_n^2`$ was chosen so that $`I\dot{\alpha }^2/2`$ equals the kinetic energy of the atoms with mass $`M_n`$, when they move from the initial displacements $`\delta \stackrel{}{R_n}`$.
The same type of the potential energy curve was calculated for the STE. The quadratic fit works only in the close vicinity of the exciton metastable minimum. The resulting frequency is $`\omega _b=123\text{cm}^1`$. The experimental temperature T=11 K justifies a zero temperature approximation (the STE initial state is the vibrational ground state). Since the luminescence peak position corresponds to the vibrational level $`n`$$``$170 of the electronic ground state, quasiclassical wavefunctions were used in the numerical integral evaluation. The sequence of vibrational sidebands should be replaced by a sequence of convolved densities of phonon states $`D(\omega )`$. We approximate this by a Gaussian, $`D(\omega )\mathrm{exp}(\omega ^2/2\gamma ^2)/\sqrt{2\pi }\gamma `$ with the width $`\gamma `$=43 cm<sup>-1</sup> chosen such that the first three moments coincide with the experimental phonon DOS. This gives a luminescence width of 0.43 eV, while the experimental width is 0.63 eV.
The optical response $`\sigma (\omega )`$ of the long-lived triplet STE also has been measured . The diagonal part of the optical conductivity tensor is:
$`\sigma _\alpha (\omega )={\displaystyle \frac{\pi e^2N}{m^2\omega \mathrm{\Omega }}}{\displaystyle \underset{\mathrm{}\mathrm{}^{}}{}}{\displaystyle \underset{\stackrel{}{k}}{}}(f_\mathrm{}\stackrel{}{k}f_\mathrm{}^{}\stackrel{}{k})`$ (1)
$`\mathrm{}\stackrel{}{k}|p_\alpha |\mathrm{}^{}\stackrel{}{k}^2\delta (\mathrm{}\omega E_\mathrm{}^{}\stackrel{}{k}+E_\mathrm{}\stackrel{}{k})`$ (2)
where $`f_\mathrm{}\stackrel{}{k}`$ is the occupancy of the state $`|\mathrm{}\stackrel{}{k}`$. The spin state index is included in the band index $`\mathrm{}`$. Integration over the zone has been performed using 26 k-points in the irreducible zone. In Fig. 4a, absolute optical conductivity curves are shown for the three polarizations, $`\stackrel{}{E}`$(1,-1,0), (0,0,1), and (1,1,0) (parallel to the the Cl$`{}_{}{}^{}{}_{2}{}^{}`$ molecular axis). Fig. 4b shows the average conductivity $`\sigma (\omega )=_i\sigma _i(\omega )/3`$, which is compared with experiment , rescaled so that the total weights under the both curves are the same.
The first peak in Fig. 4b is centered at 0.95 eV; the splitting between the peaks for the three different polarizations is not resolved in the calculations. Most of the weight in these peaks comes from transitions of the last localized spin-up a<sub>1g</sub> electron into empty conduction band delocalized states (see Fig. 2b).
The second peak centered at 3.58 eV is the a$`{}_{1g}{}^{}\text{b}_{3u}`$ transition for the spin down electron (see Fig. 2c). The energy difference between the ground state and electronically excited exciton state with the same atomic configuration turns out to be the same as the eigenvalue difference of b<sub>3u</sub> and a<sub>1g</sub> states of the ground STE. The excited exciton will lower its energy by moving atoms back to the undistorted positions of perfect NaCl. To apply the Franck-Condon principle, we repeat the same type of calculations for the excited exciton as we did for NaCl and the ground state STE (Fig. 3). When two Cl atoms move away from each other, the a<sub>1g</sub> empty state merges with the valence Cl 3$`p`$ band, which makes it very difficult to choose which state to depopulate during the iterations. Instead we used the results for the ground state STE to obtain the energy of the electronically excited state by adding eigenvalue difference $`\lambda _\mathrm{}1\lambda _\mathrm{}2`$ between two states for which dipole matrix element $`\mathrm{}1|p_3|\mathrm{}2^2`$ is the largest. Results are shown on Fig. 3 along with a quadratic fit ($`\omega _c`$=116 cm<sup>-1</sup>) which works for the entire range of the parameter $`\alpha `$. The delta-function of Eq. (2) corresponding to the a<sub>1g</sub>$``$b<sub>3u</sub> transition was replaced by a sequence of convolved Gaussian peaks.
The transition energy (3.58 eV) between the two localized states a<sub>1g</sub>$``$b<sub>3u</sub> (see Fig. 2c) agrees well with the experimental peak at 3.8 eV. But when initial and final states have different degrees of localization, errors of an eV in the luminescence (LSDA: 4.25 eV, expt. : 3.35 eV) and in the optical excitation of the bound electron into the conduction band (LSDA: 0.95 eV, expt. : 1.95, 2.13 and 2.00 eV for three different field polarizations) occur. A possible explanation for this error is that in LSDA the incomplete cancellation of the large repulsive self-interaction energy $`𝑑\stackrel{}{r}𝑑\stackrel{}{r^{}}n_i(\stackrel{}{r})n_i(\stackrel{}{r^{}})/|\stackrel{}{r}\stackrel{}{r^{}}|`$ of a localized state $`n_i=|\mathrm{\Psi }_i|^2`$ and the corresponding exchange term may destabilize a localized solution in favor of a delocalized one. Corrections such as self-interaction corrections (SIC) or LDA+U may reduce the error; in SIC, shifts in the energy of a state $`i`$ on the order of $`d^3rn_i^{4/3}(\stackrel{}{r})`$ are expected. If the degrees of localization of the two states are similar, however, then the LSDA transition energies are reasonable.
In summary, we have presented LSDA calculations for the STE in NaCl, including the coupling between the lattice and electronic states. The off-center STE is found to be more stable than the on-center STE, but both are metastable compared to free electron-hole pairs. Both luminescence and optical conductivity, including vibrational Franck-Condon effects, were also calculated. The hole state in the off-center STE is found to be rather evenly split between the two Cl atoms, but the electron state is localized to the vacant site left by the shifted Cl ion. The density functional description of electronic transitions between localized states, such as the a$`{}_{1g}{}^{}\text{b}_{3u}`$ absorption peak, agree well with experiment. For transitions between the localized and delocalized states, discrepancies of order 1 eV with experiment arise. Although the LSDA can capture many features of the STE states, when a localized solution competes with a delocalized solution, the incomplete cancellation of the self-interaction may destabilize the localized solution. Given the usefulness of the LSDA method for unraveling complex materials, it is important to test and develop approaches that can treat localized and delocalized states on the same footing.
###### Acknowledgements.
We thank M. L. Cohen, G. W. Fernando, P. M. Johnson, S.G. Louie, W. E. Pickett for discussions. This work was supported in part by NSF Grant No. DMR-9725037 and by DOE Grant No. DE-AC-02-98CH10886.
|
warning/0003/physics0003100.html
|
ar5iv
|
text
|
# Auditory sensitivity provided by self-tuned critical oscillations of hair cells
## I Generic aspects
Amplification and frequency filtering of a Hopf bifurcation. We discuss the behavior of a dynamical system which is controlled by a parameter $`C`$. Above a critical value, $`C>C_c`$, the system is stable; for $`C<C_c`$ it oscillates spontaneously. At the critical point (or Hopf bifurcation) $`C=C_c`$ the system shows remarkable response and amplification properties which do not depend on the physical mechanism at the basis of the bifurcation. These generic properties can be described as follows. Since we are interested in the response to a periodic stimulus with frequency $`\nu =\omega /2\pi `$, we express the hair-bundle deflection $`x(t)`$ by a Fourier series $`x(t)=x_ne^{in\omega t}`$ with complex amplitudes $`x_n=x_n^{}`$. In the vicinity of the bifurcation, the mode $`n=\pm 1`$ is dominant and the response to an externally-applied sinusoidal stimulus force $`f(t)=f_1e^{i\omega t}+f_1e^{i\omega t}`$ can be expressed in terms of a systematic expansion in $`x_1`$. Symmetry arguments (see Appendix A) imply that the first nonlinear term is cubic:
$$f_1=𝒜x_1+|x_1|^2x_1+\mathrm{}$$
(1)
where $`𝒜(\omega ,C)`$ and $`(\omega ,C)`$ are two complex functions. The bifurcation point is characterized by the fact that $`𝒜`$ vanishes for the critical frequency, $`𝒜(\omega _c,C_c)=0`$. For $`C<C_c`$ and no external force, the system oscillates with $`|x_1|^2\mathrm{\Delta }^2(C_cC)/C_c`$, where $`\mathrm{\Delta }`$ is a characteristic amplitude. For $`C=C_c`$ the response to a stimulus at the critical frequency has amplitude
$$|x_1|||^{1/3}|f_1|^{1/3}.$$
(2)
This represents an amplified response at the critical frequency with a gain
$$r=\frac{|x_1|}{|f_1|}|f_1|^{2/3}$$
(3)
that becomes arbitrarily large for small forces.
If the stimulus frequency differs from the critical frequency, the linear term in Eq. (1) is non-zero and can be expressed to first order as $`𝒜(\omega ,C_c)A_1(\omega \omega _c)`$. The dramatic amplification of weak signals, implied by Eq. (3), is maintained as long as this term does not exceed the cubic term in Eq. (1). If the frequency mismatch increases such that $`|\omega \omega _c||f_1|^{2/3}||^{1/3}/|A_1|`$, the response becomes linear
$$|x_1|\frac{|f_1|}{|(\omega \omega _c)A_1|}$$
(4)
i.e. the gain is independent of the strength of the stimulus.
Thus the Hopf resonance acts as a sharply tuned high-gain amplifier for weak stimuli, and as a low-gain filter for strong stimuli. This generic behavior is illustrated in Fig. 1 with data obtained by numerical simulation. Laser interferometry measurements of the motion of the basilar membrane, when the live cochlea is stimulated by pure tones, display strikingly similar features . In particular, the peak response as a function of force amplitude has been demonstrated to obey a power law $`|x_1||f_1|^{0.4\pm 0.2}`$ . This strongly suggests that the membrane is being driven by a dynamical system which is poised at a Hopf bifurcation.
Self-tuned critical oscillations. How does the system come to be so precisely balanced at the critical point?<sup>*</sup><sup>*</sup>*Self-tuning to a coexistence point has been discussed previously in certain dynamic first-order transitions . The control parameter must be tuned to $`CC_c`$, otherwise the nonlinear amplification is lost. Moreover, the value of $`C_c`$ differs for hair cells with different characteristic frequency. We propose a feedback mechanism which allows the dynamical system to operate automatically close to the bifurcation point, whatever its characteristic frequency. Without loss of generality, we assume that the control parameter decreases as long as the system does not oscillate. After some time, critical conditions are reached and spontaneous oscillations ensue. The onset of oscillations triggers an increase of the control parameter which tends to restore stability. Hence the system converges to an operating point close to the bifurcation point. To illustrate this general idea, we consider the following simple feedback which changes $`C`$ in response to deflections $`x`$:
$$\frac{1}{C}\frac{C}{t}=\frac{1}{\tau }\left(\frac{x^2}{\delta ^2}1\right)$$
(5)
where $`\delta `$ is a typical amplitude. If no external force is applied, this feedback, after a relaxation time $`\tau `$, tunes the control parameter to a value $`C_\delta `$ for which spontaneous oscillations with $`|x_1|\delta `$ occur. If $`\delta `$ is small compared to the characteristic amplitude $`\mathrm{\Delta }`$, this is on the oscillating side close to the bifurcation, $`|C_\delta C_c|/C_c(\delta /\mathrm{\Delta })^2`$. Two modes of signal detection are possible: (i) For transient stimuli short compared to $`\tau `$ the system operates at $`C_\delta `$. The amplitude $`|x_1|`$ shows the characteristic nonlinear response discussed above. (ii) For stimuli sustained over longer times, self-tuning maintains $`|x_1|\delta `$ constant for different stimulus amplitudes. This effect of the feedback represents a perfect adaptation mechanism. However, in the presence of noise, phase-locking of the response (to be discussed later) occurs as soon as an external stimulus is applied, and this can be detected.
## II Model
Mechanosensitivity and self-tuning mechanism provided by transduction channels. We demonstrate the general principles introduced above by devising a specific model for the amplification of acoustic stimuli by hair bundles in non-mammalian vertebrates. A schematic representation of a hair bundle is shown in Fig. 2. It consists of several stereocilia and a single kinocilium . Transduction of hair bundle deflection to a chemical signal occurs via channels located near the tip of each stereocilium. Tip links which connect neighboring stereocilia are believed to be the gating springs of the transduction channels. If the hair bundle is deflected, tension in the tip links triggers the opening of the channels. The subsequent influx of ions (principally $`\mathrm{K}^+`$, but also $`\mathrm{Ca}^{2+}`$) causes a corresponding change of the membrane potential of the hair cell which, in turn, generates a nervous signal.
The mechanosensor can be used for self-tuning, as well as signal detection. Many physiological processes are regulated by ionic concentrations, so it is natural to identify the $`\mathrm{Ca}^{2+}`$ concentration with the control parameter $`C`$. We assume that $`C`$ decreases if the transduction channels are closed, owing to the action of pumps in the cell membrane. When the hair bundle is deflected by $`x`$, the transduction channels open with probability $`P_o(x)`$. We therefore characterize the mechanosensor by the equation
$$\frac{C}{t}=\frac{C}{\tau }+J_oP_o(x)$$
(6)
where $`J_o`$ is the $`\mathrm{Ca}^{2+}`$ flux through open transduction channels. Note that this equation provides self-tuning. In this case it replaces the more simple but less realistic Eq. (5). For our numerical examples, we use a two-state model for the channels with
$$P_o(x)=\frac{1}{1+Ae^{x/\delta }}$$
(7)
where $`(1+A)^11`$ is the probability that a channel is open when the hair cell is quiescent and $`\delta `$ is the characteristic amplitude of motion to which the system is sensitive. For a sufficiently long relaxation time $`\tau `$, the slow variation $`C_0`$ of $`CC_0(t)+C_1e^{i\omega t}`$ can be separated from the small-amplitude oscillations, giving
$$_tC_0\frac{C_0}{\tau }+J_o\stackrel{~}{P}_o(|x_1|^2)$$
(8)
where $`\stackrel{~}{P}_o=_0^{1/\nu }𝑑tP_o(x_1e^{i\omega t}+x_1e^{i\omega t})`$ is the averaged probability of channel opening in the presence of oscillations, which increases monotonically with their amplitude $`x_1`$. For physically relevant parameter values, the system reaches a steady state close to the bifurcation point, independent of the initial value of $`C`$.
Oscillations generated by molecular motors. We still have to specify the nature and location of the oscillator within the hair bundle. It has been suggested that the transduction channels might be the source of the instability ; or, that myosin motors in the stereocilia might generate the force necessary to move the bundle . We propose a third possibility: the kinocilium could vibrate using its internal dynein motors. Recently, a simple physical mechanism has been proposed which allows motor proteins operating in collections to generate spontaneous oscillations by traversing a Hopf bifurcation . Typically, motors move along cytoskeletal filaments and elastic elements oppose this motion. In this case, two possibilities exist: The system either reaches a stable balance between opposing forces, or it oscillates around the balanced state. Three time scales characterize this behavior: The relaxation time $`\lambda /K`$ of passive relaxation, where $`\lambda `$ is the total friction and $`K`$ is the elastic modulus; and the times $`\omega _1^1`$ and $`\omega _2^1`$, where $`\omega _1`$ is the kinetic rate at which a motor detaches from a filament, and $`\omega _2`$ is the attachment rate. An explicit solution of a simple model is derived in Appendix B. We find that for an appropriate value of a control parameter $`C`$, which is related to the ratio $`\omega _1/\omega _2`$, a Hopf bifurcation occurs with critical frequency
$$\omega _c\left(\frac{K\alpha }{\lambda }\right)^{1/2},$$
(9)
which is the geometric mean of the passive relaxation rate $`K/\lambda `$ and the typical ATP hydrolysis rate $`\alpha =\omega _1+\omega _2`$. The above identification of $`C`$ with the $`\mathrm{Ca}^{2+}`$ concentration is consistent with the fact that $`\mathrm{Ca}^{2+}`$ regulates motor protein activity . The system oscillates if the elastic modulus does not exceed a maximal value $`K_{\mathrm{max}}k_0N`$, where $`k_0`$ is the crossbridge elasticity of a motor and $`N`$ is the total number of motors. The maximal frequency, obtained when $`K=K_{\mathrm{max}}`$, can be significantly higher than the ATP hydrolysis rate $`\alpha `$.
Characteristic frequency of a vibrating kinocilium. The kinocilium is a true cilium containing a cylindrical arrangement of microtubule doublets and dynein motors. Since cilia have the well-established tendency to beat and vibrate with frequencies from tens of Hz up to at least 1 kHz , the kinocilium is a natural candidate to be responsible for the Hopf bifurcation. A simple two-dimensional model can be used to discuss the main physical properties of a vibrating cilium near a Hopf bifurcation . In this model, motors induce the bending of a pair of elastic filaments separated by a distance $`a`$ (corresponding to the distance between neighboring microtubule doublets in the axoneme). An isolated kinocilium of length $`L`$ and bending rigidity $`\kappa `$, fixed at the basal end but free at its tip, will vibrate in a wave-like mode with wavelength $`\mathrm{\Lambda }4L`$. A typical displacement $`z`$ of the motors along the filaments leads to bending of the filament pair and a deflection of the tip by a distance $`xzL/a`$. The elastic bending energy is of order $`(\kappa /L^3)x^2(\kappa /a^2L)z^2`$. Therefore, the total elastic modulus experienced by the motors is given by
$$K\frac{\kappa }{La^2}.$$
(10)
The viscous energy dissipation per unit time due to motion of the kinocilium is of order $`\eta L(_tx)^2(\eta L^3/a^2)(_tz)^2`$. Therefore, the total friction experienced by the motors can be written as
$$\lambda \frac{\eta L^3}{a^2}+\lambda _0\rho L,$$
(11)
where $`\lambda _0`$ describes the dissipation within the kinocilium per motor and $`\rho =N/L`$ denotes the number of dynein motors per unit length along the axoneme. If internal friction can be neglected, i.e. if $`LL_0=(\lambda _0a^2\rho /\eta )^{1/2}`$, the frequency of a vibrating cilium at the bifurcation point is given by
$$\omega _c\left(\frac{\kappa \alpha }{\eta }\right)^{1/2}\frac{1}{L^2}.$$
(12)
Using typical values $`\lambda _010^9\mathrm{kg}/\mathrm{s}`$, $`a20\mathrm{nm}`$, $`\rho \mathrm{5\hspace{0.33em}10}^8\mathrm{m}^1`$ and $`\eta 10^3\mathrm{kg}/\mathrm{ms}`$, we find $`L_0200\mathrm{nm}`$, shorter than typical kinocilia. Using $`\alpha 10^3\mathrm{s}^1`$, $`k_010^3\mathrm{N}/\mathrm{m}`$ and $`\kappa \mathrm{4\hspace{0.33em}10}^{22}\mathrm{Nm}^2`$ (the bending rigidity of 20 microtubules), the frequency range between 100$`\mathrm{Hz}`$ and 10$`\mathrm{kHz}`$ can naturally be spanned by changing the length of the kinocilium between $`1\mu \mathrm{m}`$ and $`10\mu \mathrm{m}`$.
The above argument neglects the contribution of the stereocilia to elasticity and motion. The elastic response of stereocilia to hair bundle deflections has been measured . It can be well described by an angular spring at the base of each stereocilium which contributes an elastic energy per stereocilium of the order of $`k_s(x/L)^2`$, where $`k_s`$ is an angular elastic modulus. The kinocilium length $`L`$ in a hair bundle is approximately inversely proportional to the number $`N_s`$ of stereocilia , and we write $`N_sl_s/L`$, where $`l_s10^4\mathrm{m}`$ is the total length of stereocilia. With this assumption, we find an additional elastic modulus
$$K_s\frac{l_sk_s}{La^2}$$
(13)
contributed by the stereocilia to $`K`$. The contribution of stereocilia to the friction coefficient $`\lambda `$ can be estimated as
$$\lambda _s\eta l_sL^2/a^2.$$
(14)
The measured value of $`k_s`$ indicates that $`K_s`$ dominates the contribution to $`K`$ given by Eq. (10). Similarly, since $`l_s>L`$, $`\lambda _s`$ should dominate friction. In this case, we expect
$$\omega _c\left(\frac{k_s\alpha }{\eta L^3}\right)^{1/2}$$
(15)
and the range of frequencies is somewhat reduced.
If the stiffness of the ensemble of stereocilia greatly exceeds that of the kinocilium, a new situation arises. Since the kinocilium is attached to the stereocilia at its tip , movement of the tip is strongly reduced and the kinocilium preferentially vibrates in a mode with wavelength $`\mathrm{\Lambda }2L`$ for which the relation given by Eq. (12) is again valid.
Adaptation motors. Finally, we note that in order to obtain a robust self-tuning, the feedback mechanism must be sensitive only to the oscillation amplitude $`x_1`$ and not to a stationary displacement $`x_0`$; we assumed in Eq. (8) that $`\stackrel{~}{P}_o`$ is independent of $`x_0`$. The transduction mechanism of stereocilia and their tension-gated channels does indeed have this property: It is well known that an ATP-dependent adaptation mechanism exists which removes the dependence of the channel current on a constant displacement $`x_0`$. It is widely believed that this adaptation involves the motion of myosin motors, which maintains constant the steady-state tension in the tip links that control the transduction channels . Therefore, the stereociliar transduction mechanism has precisely the required properties to be used as a feedback signal for self-tuning the bifurcation.
## III Simulation
The above analysis does not include the effects of noise. Brownian motion is one source of fluctuations in the movement of the hair bundle . Another source of noise is caused by stochastic fluctuations in motor protein force as dynein molecules bind to, and detach from the microtubules in the kinocilium. In order to investigate the consequences of this randomness, we performed a Monte Carlo simulation of the two-state model described in , using a realistic number of motor proteins. The motors, when attached, experienced a potential of amplitude $`U`$ and period $`l`$ (the corresponding crossbridge elasticity is $`k_0=U/l^2`$). The attachment rate was constant and independent of position, $`\omega _2=\alpha `$. Detachment was localized to a region of width $`0.1l`$, centred on the potential minimum and the detachment rate $`\omega _1`$ was regulated by the $`\mathrm{Ca}^{2+}`$ concentration $`C`$. We chose $`\omega _1/\omega _2=(C_q/C)^3`$, where $`C_q`$ is the steady-state $`\mathrm{Ca}^{2+}`$ concentration in a quiescent hair cell; the precise functional dependence is unimportant as long as $`\omega _1/\omega _2`$ decreases monotonically with increasing $`C`$ in a fairly sensitive way. We simulated systems with $`N`$ = 1000-4000 motors, representing hair cells with different kinocilium length and stereociliary number.
Self-tuning and the characteristic frequency of spontaneous oscillations. The self-tuning of a hair cell to the vicinity of the bifurcation, where small-amplitude spontaneous oscillations occur, is demonstrated in Fig. 3. When a change of the internal $`\mathrm{Ca}^{2+}`$ concentration is imposed, the system is transiently perturbed; but after an interval of time of order $`\tau `$, it returns to the same steady state. Hair cells with different numbers of motors acquire different internal concentrations of $`\mathrm{Ca}^{2+}`$, in order to adjust the motor detachment rate in such a way that the system approaches the critical point. The spontaneous oscillations of three different hair bundles are shown in Fig. 4. Note that the characteristic frequency is approximately proportional to the inverse-square of the number of motors, as expected from Eq. (12), and that it can exceed the typical ATP cycle rate $`\alpha `$ when the total number of motors is small (short kinocilium). All three hair cells execute spontaneous oscillations with a similar amplitude, as expected from our arguments above. However, the noise introduces a significant new effect: The oscillations are irregular. The incoherence of the phase of the oscillation is evident in the Fourier transform of the displacement, which exhibits a broad peak centred on the characteristic frequency.
Dynamic response to a tone at the characteristic frequency. The response of a self-tuned hair bundle to a sinusoidal force with a frequency approximately equal to the bundle’s characteristic frequency is illustrated in Fig. 5. For weak stimuli,the amplitude of the oscillation does not increase with the amplitude $`|f_1|`$ of the applied force; this is because the small response to the stimulus is masked by the noisy, spontaneous motion. Instead, the phase of the hair-bundle oscillation becomes more regular; as it does so, a peak emerges from the Fourier spectrum at the driving frequency. The height of the peak grows approximately as $`|f_1|^{1/3}`$ for intermediate values of $`|f_1|`$, and approximately linearly for very weak stimuli, and also for very strong stimuli (for which the response is essentially passive). Thus the Fourier component of the hair-bundle displacement at the driving frequency responds to the stimulus in the generic manner discussed above.
Phase-locking of the nervous signal. The flow of ions through the transduction channels depolarizes the cell membrane which, in turn, opens voltage-gated channels at the base of the hair cell and generates a synaptic current . This sequence of events happens fast enough for variations at the synapse faithfully to reflect the hair-bundle motion at frequencies below 1 kHz. Information about the auditory stimulus is subsequently passed along the auditory nerve in the form of a spike train. Simplifying this transduction process, we assume that a spike is elicited whenever the transduction channel current $`J`$ passes a threshold value of $`0.5J_o`$. The resulting nervous response is shown in Fig. 6. In the absence of a stimulus, the self-tuned critical oscillations of the bundle cause the nerve to fire stochastically, at a low rate. When a weak sinusoidal stimulus is applied at the characteristic frequency, the firing rate does not increase above the spontaneous rate, but the spike train becomes detectably phase-locked to the stimulus. The degree of phase-locking increases rapidly as the amplitude of the stimulus increases, reflecting the growing regularity of the hair-bundle motion. It is only when the neural response is almost completely phase-locked that the firing rate begins to rise. Eventually, for strong stimuli, the spike rate saturates at the stimulus frequency. This behavior is strikingly similar to that which is observed experimentally. In particular, it is well known that the threshold for phase-locking in the auditory nerve fiber is 10-20 dB lower than the threshold at which the firing rate begins to rise .
## IV Discussion
The self-tuned critical oscillator which we have introduced as a system for signal detection has characteristic amplification properties which are generic and do not depend on the choice of a specific model. The oscillation frequency, however, depends on the physical mechanism involved. In our model of non-mammalian hair cells, the frequency is determined by the geometry of the hair bundle. A simple morphological gradient along the basilar membrane would endow the ear with the ability to analyze a wide range of frequencies. Experimentally, it is well established that the height of hair bundles progressively increases along the cochlea, and that concurrently the characteristic frequency of the hair cells declines . Our proposition that the kinocilium is likely to play an active role in non-mammalian vertebrate hair cells suggests experiments which study the motility of the kinocilium and its potential for generating oscillations. If the kinocilium is the only source of oscillations, its removal should suppress hair-bundle vibrations. If the self-tuning mechanism is removed instead (by cutting the tip links, for example) our model predicts that the kinocilium should exhibit stronger spontaneous oscillations. Careful control of the extracellular ion concentrations, such as $`\mathrm{Ca}^{2+}`$ would be essential in such experiments.
The simulations of our model reveal how a hair bundle can achieve its remarkable sensitivity to weak stimuli. By profiting from the periodicity of a sinusoidal input, and measuring phase-locking rather than the amplitude of response, the mechanosensor can detect forces considerably weaker than those exerted by a single molecular motor (if the bundle were a simple, passive structure, its response to such forces would be smaller than its Brownian motion). An important implication of this detection mechanism is that even though the hair cell selects a certain frequency, the signal must still be encoded by the interval between spikes elicited in the auditory nerve. Paradoxically, the stochastic noise caused by the motor proteins serves a useful purpose. It ensures that the self-tuned critical oscillations of the hair bundle are incoherent, so that the pattern of spontaneous firing in the nerve is irregular. Against this background, the regular response to a periodic stimulus can easily be detected.The benefits of noise have been discussed in a variety of situations such as those involving stochastic resonance .
Another beneficial feature of noise arises from the fact that weak stimuli do not increase the amplitude of oscillation above the spontaneous amplitude. Thus the $`\mathrm{Ca}^{2+}`$ concentration remains constant, the hair bundle stays in the critical regime, and active amplification can be sustained indefinitely. Stronger stimuli cause the system to drift away from the critical point, so that the degree of amplification diminishes over time. It is well known that both the perception of loudness and the firing rate of the auditory nerve decrease over a period of a few seconds, when a stimulus of moderate intensity is maintained. This phenomenon, which is usually referred to as ‘adaptation’, is consistent with our self-tuning mechanism. Following the sustained presentation of a loud stimulus, the spontaneous firing rate diminishes and the threshold to weak stimuli is augmented . Such ‘fatigue’ is also accounted for by our self-tuning mechanism.
Since self-tuning positions the system slightly on the oscillating side of the critical point, self-tuned criticality provides a natural explanation for otoacoustic emissions . In its normal working state, the inner ear would generate faint sounds with a broad range of frequencies. If the feedback mechanism were to fail in certain cells, the spontaneous oscillations could become large enough for distinct tones to be emitted.
A self-tuned Hopf bifurcation is ideal for sound detection because it provides sharp frequency selectivity and a nonlinear gain which compresses a wide range of stimulus intensities into a narrow range of response. We therefore believe that the concept applies to all vertebrate hearing systems, and potentially to other mechanoreceptor systems. Kinocilia are typically absent in mammalian cochlea, and we suggest that their force-generating role has been assumed by the outer hair cells. Self-tuning of these motile cells might be realized by a mechanism similar to that presented here, using transduction channels in their hair bundles. It could, however, work very differently; for example, it might involve feedback from the inner hair cells, via the efferent nervous system.
Tuning to the proximity of a critical point is likely to be a general strategy adopted by sensory systems. Simple molecular receptors , as well as the physiological sensors of higher organisms, can enhance their response to weak stimuli in this way. We propose that the physics of self-tuned criticality is the ‘central science of transducer physiology’ spoken of by Delbrück .
Acknowledgements: We thank P. Martin and C. Petit for useful discussions, and A.F. Huxley for referring us to the work of T. Gold. T. Duke is grateful for the hospitality of Institut Curie and the Niels Bohr Institute, and acknowledges the support of the Royal Society. After submission of our manuscript we learned that Eguiluz, Ospeck, Choe, Hudspeth and Magnasco have independently described the generic response of a system near a Hopf bifurcation; we thank them for communicating their unpublished results.
## A Generic behavior at a Hopf-bifurcation
### 1 Nonlinear relation between periodic stimulus and displacements
We are interested in the response $`x(t)`$ of a nonlinear system to a periodic stimulus force $`f(t)`$. If only one frequency $`\nu =\omega /2\pi `$ is present we use the Fourier expansions
$`f(t)`$ $`=`$ $`{\displaystyle \underset{n=\mathrm{}}{\overset{\mathrm{}}{}}}f_ne^{in\omega t}`$ (A1)
$`x(t)`$ $`=`$ $`{\displaystyle \underset{n=\mathrm{}}{\overset{\mathrm{}}{}}}x_ne^{in\omega t},`$ (A2)
where the complex coefficients $`x_n`$ and $`f_n`$ obey $`x_n=x_n^{}`$ and $`f_n=f_n^{}`$. This representation implies that we focus on the limit cycle solution and ignore all transient relaxation phenomena. We consider the class of systems for which the force at a given time depends in a nonlinear way on the history of the displacements $`x(t)`$ alone; as we will discuss in section D more complex cases do not change the basic properties. In this situation, the relation between $`x`$ and $`f`$ can be expressed as a systematic expansion of the force amplitudes $`f_n`$ in the amplitudes $`x_n`$:
$`f_k=F_{kl}^{(1)}x_l`$ $`+`$ $`F_{klm}^{(2)}x_lx_m`$ (A3)
$`+`$ $`F_{klmn}^{(3)}x_lx_mx_n+O(x^4),`$ (A4)
where the expansion coefficients $`F_{k,k_1,..,k_n}^{(n)}`$ are symmetric with respect to permutations of the indices $`k_1..k_n`$. The limit cycle solutions are invariant with respect to translations in time $`tt+\mathrm{\Delta }t`$. Under these transformations the amplitudes change as $`x_nx_ne^{in\omega \mathrm{\Delta }t}`$ and $`f_nf_ne^{in\omega \mathrm{\Delta }t}`$. Inspection of Eq. (A4) shows that the time translation symmetry allows only for those terms $`F_{k,k_1,..,k_n}^{(n)}x_{k_1}..x_{k_n}`$ for which $`k=k_1+\mathrm{}+k_n`$. For all other cases $`F_{k,k_1,..,k_n}^{(n)}`$ must vanish which significantly restricts the number of terms.
### 2 Hopf bifurcation
The nonlinear system exhibits spontaneous oscillations and a Hopf-bifurcation if nontrivial solutions to Eq. (A4) with $`x_n0`$ exist in the case where all $`f_k=0`$, i.e. if no stimulus force is applied. Without loss of generality, we consider here an instability of the mode $`x_1`$. In this case, the dominant terms allowed by symmetry read ($`f_k=0`$)
$`0`$ $``$ $`F_{11}^{(1)}x_1+2F_{1,2,1}^{(2)}x_1x_2`$ (A5)
$`+`$ $`6F_{1,1,1,1}^{(3)}x_1^2x_1+6F_{1,1,2,2}^{(3)}x_2x_2x_1`$ (A6)
$`0`$ $``$ $`F_{22}^{(2)}x_2+2F_{211}^{(2)}x_1^2.`$ (A7)
Eq. (A7) determines $`x_22(F_{211}^{(2)}/F_{22}^{(2)})x_1^2`$. Inserting this relation in Eq. (A6), we obtain to lowest order
$$0𝒜x_1+|x_1|^2x_1,$$
(A8)
where $`𝒜F_{11}^{(1)}`$ and $`3F_{1,1,1,1}^{(3)}4F_{211}^{(2)}F_{1,2,1}^{(2)}/F_{22}^{(2)}`$.
The coefficients $`𝒜(\omega ,C)`$ and $`(\omega ,C)`$ are complex and in general depend on frequency $`\omega `$ and a control parameter which we denote by $`C`$. A Hopf bifurcation occurs at a critical point $`C=C_c`$ at which $`𝒜`$ vanishes for a frequency $`\omega _c`$, i.e. $`𝒜(\omega _c,C_c)=0`$. This can be demonstrated as follows: A spontaneously oscillating solution satisfies
$$|x_1|^2=\frac{𝒜}{}$$
(A9)
Note, that such a solution can only exist if $`𝒜/`$ is real and negative. At the bifurcation point, $`𝒜=0`$ and $`𝒜/`$ is therefore real for $`\omega =\omega _c`$, however the corresponding amplitude $`|x_1|^2`$ vanishes. In the vicinity of this point we expect to find solutions with finite amplitude. We use the expansion
$$𝒜(\omega ,C)(\omega \omega _c)A_1+(CC_c)A_2$$
(A10)
where $`A_1`$ and $`A_2`$ are complex coefficients and we neglect higher order terms. Spontaneous oscillating solutions exist only if $`𝒜/`$ is real. This condition is satisfied for a particular frequency $`\omega =\omega _s`$ with
$$\omega _s=\omega _c+\frac{Im(A_2/)}{Im(A_1/)}(C_cC).$$
(A11)
The ratio $`𝒜/`$ at this frequency $`\omega _s`$ changes sign for $`C=C_c`$; here we assume without loss of generality that it is positive for $`C<C_c`$. In this case, the system oscillates spontaneously with an amplitude which according to Eq. (A9) behaves as $`|x_1|^2=\mathrm{\Delta }^2(C_cC)/C_c`$, where
$$\mathrm{\Delta }^2=C_c\left(Re(A_2/)Re(A_1/)\frac{Im(A_2/)}{Im(A_1/)}\right)$$
(A12)
is a typical amplitude. We have thus demonstrated that Eq. (A8) characterizes a Hopf-bifurcation if the complex coefficient $`𝒜`$ vanishes at a critical point $`C_c`$ for a critical frequency $`\omega _c`$.
### 3 Amplified response to sinusoidal stimuli
If a sinusoidal stimulus $`f(t)=f_1e^{i\omega t}+f_1e^{i\omega t}`$, for which all $`f_n`$ with $`n\pm 1`$ vanish, Eq. (A8) becomes
$$f_1𝒜x_1+|x_1|^2x_1.$$
(A13)
We consider a system that is tuned exactly to the bifurcation, $`C=C_c`$. In this situation spontaneous oscillations do not occur and $`𝒜=(\omega \omega _c)A_1`$. If the imposed frequency is equal to the critical frequency $`\omega =\omega _c`$, the coefficient $`𝒜`$ vanishes and we can solve Eq. (A13) for $`|x_1|`$ to find the nonlinear response
$$|x_1|||^{1/3}|f_1|^{1/3},$$
(A14)
as a function of the force amplitude $`|f_1|`$. This behavior represents an amplified response with a gain
$$r=\frac{|x_1|}{|f_1|}|f_1|^{2/3}$$
(A15)
that becomes arbitrarily large for small forces. If the frequency $`\omega `$ is different from $`\omega _c`$, this nonlinear response still holds as long as the linear term in Eq. (A13) is small compared to the cubic term and can be neglected. This is the case if $`|x_1|^2|𝒜/|=|\omega \omega _c||A_1/|`$. Therefore, the nonlinear regime characterized by Eq. (A14) holds for sufficiently large force amplitudes, $`|f_1||(\omega \omega _c)A_1|^{3/2}/||^{1/2}`$, or if the frequency is sufficiently close to the critical frequency, $`|\omega \omega _c||f_1|^{2/3}||^{1/3}/|A_1|`$.
If the frequency mismatch $`|\omega \omega _c|`$ becomes large, or if forces $`|f_1|`$ are small, a new regime occurs for which the linear term in (A13) dominates. In this regime, the response is linear,
$$|x_1|\frac{|f_1|}{|(\omega \omega _c)A_1|},$$
(A16)
and the gain is constant. This is a passive response if the stimulus frequency is too far from the critical frequency.
### 4 Additional remarks
The above derivation is based on an expansion (A4) in the displacements $`x_n`$. This excludes some nonlinearities in the force which can lead to additional nonlinear terms in Eq. (A13). The most general form of Eq. (A13) is
$`f_1𝒜x_1`$ $`+`$ $`|x_1|^2x_1+𝒞x_1|f_1|^2+𝒟x_1f_1^2`$ (A17)
$`+`$ $`|x_1|^2f_1+x_1^2f_1+𝒢|f_1|^2f_1.`$ (A18)
However, for small forces $`f_1`$ and small amplitudes $`x_1`$, the results derived above are not affected. The regime of nonlinear reponse $`|f_1||x_1|^3`$, as well as the linear response regime $`|f_1||x_1|`$ still exist. If $`|f_1||x_1|`$, the nonlinear terms in $`f_1`$ renormalize the third order term, which in this regime is negligable. If $`|f_1||x_1|^3`$, the nonlinear terms in $`f_1`$ are of even higher order and can be neglected.
## B Oscillations generated by molecular motors
### 1 Two state model
The two state model describes force-generation as a result of transitions between two states, a bound state and a detached state of a motor and its track filament. The interaction between a motor at position $`z`$ along the filament in states $`1`$ and $`2`$ is characterized by two periodic potentials $`W_1(z)=W_1(z+l)`$ and $`W_2(z)=W_2(z+l)`$ where $`l`$ is the period. We introduce the relative position $`\xi =z\mathrm{mod}l`$ with respect to the potential period. Detachment and attachment rates are denoted $`\omega _1(\xi )`$ and $`\omega _2(\xi )`$, respectively. Oscillations can occur in this model if a large number $`N`$ of motors move collectively against an external elastic element of modulus $`K`$.
We introduce the probability $`P_1(\xi )`$ and $`P_2(\xi )`$ of finding a motor bound at position $`\xi `$ in state $`1`$ or $`2`$, which satisfy the normalization condition
$$_0^l𝑑\xi (P_1+P_2)=1$$
(B1)
For a large number of motors collectively moving with the same velocity $`v`$ the dynamic equations read
$`_tP_1+v_\xi P_1`$ $`=`$ $`\omega _1P_1+\omega _2P_2`$ (B2)
$`_tP_2+v_\xi P_2`$ $`=`$ $`\omega _1P_1\omega _2P_2`$ (B3)
The velocity $`v`$ is determined by the force-balance condition
$$f=\lambda v+Kz+N_0^l𝑑\xi (P_1_\xi W_1+P_2_\xi W_2)$$
(B4)
where $`\lambda `$ is a friction coefficient describing the total friction and $`z`$ is the displacement of the motors, $`_tz=v`$. For an incommensurate arrangement of motors with respect to the track filament and a large number $`N`$ of motors, $`P_1(\xi )+P_2(\xi )=1/l`$ and the equations of motions simplify:
$$_tP+v_\xi P=(\omega _1+\omega _2)P+\omega _2/l,$$
(B5)
where we denote for simplicity $`P(\xi )=P_1(\xi )`$.
We discuss a simple choice for the potentials and transition rates for which the Hopf bifurcation is easy to determine analytically. We consider the potential
$$W_1(\xi )=U\mathrm{cos}(2\pi \xi /l)$$
(B6)
with amplitude $`U`$, and the potential $`W_2`$ to be constant. The transition rates are chosen to be periodic functions
$`\omega _1(\xi )`$ $`=`$ $`\beta \beta \mathrm{cos}(2\pi \xi /l)`$ (B7)
$`\omega _2(\xi )`$ $`=`$ $`\alpha \beta +\beta \mathrm{cos}(2\pi \xi /l)`$ (B8)
parameterized by two coefficients $`\alpha `$ and $`\beta `$. With this choice,
$$\omega _1(\xi )+\omega _2(\xi )=\alpha $$
(B9)
is constant and the fact that $`\omega _1`$ and $`\omega _2`$ are positive restricts $`\beta `$ to the interval $`0\beta \alpha /2`$.
### 2 Linear response function
In order to determine the linear coefficient $`𝒜`$ which determines the stability of the system, we look for small amplitude oscillations close to the resting state with $`v=0`$. We write
$`P`$ $``$ $`p_0+p_1e^{i\omega t}`$ (B10)
$`f`$ $``$ $`f_1e^{i\omega t}`$ (B11)
$`z`$ $``$ $`z_1e^{i\omega t}`$ (B12)
where $`p_0=\omega _2/\alpha l`$. To linear order in $`z_1`$, we find from Eq. (B5)
$$p_1=\frac{i\omega z_1}{i\omega +\alpha }_xp_0$$
(B13)
The corresponding force is given by
$$f_1𝒜z_1$$
(B14)
with
$$𝒜=i\omega \lambda +K+\chi ,$$
(B15)
where the active response $`\chi `$ of the motors is given by
$$\chi =N_0^l𝑑\xi \frac{i\omega }{i\omega +\alpha }_\xi p_0_\xi W_1$$
(B16)
For the choice of Eq. (B6) and (B8) the integral can be calculated and we obtain
$$𝒜(C,\omega )=i\omega \lambda +KNk_0C\frac{i\omega /\alpha +(\omega /\alpha )^2}{1+(\omega /\alpha )^2}.$$
(B17)
Here, we have introduced the dimensionless control parameter $`C2\pi ^2\beta /\alpha `$ with $`0<C<\pi ^2`$ and the cross-bridge elasticity $`k_0U/l^2`$ of the motors.
### 3 Hopf bifurcation
A Hopf bifurcation occurs if there is a pair of values $`(C,\omega )`$ for which $`𝒜`$ as given by Eq. (B17) vanishes. Such a point indeed exists. For the critical value
$$C_c=\frac{\lambda \alpha +K}{Nk_0}$$
(B18)
the bifurcation occurs for the critical frequency
$$\omega _c=\left(\frac{K\alpha }{\lambda }\right)^{1/2}$$
(B19)
The critical frequency is bounded by the fact that $`C_c<\pi ^2`$. The maximal frequency occurs for the maximal possible value of $`K`$
$$K_{\mathrm{max}}=Nk_0\pi ^2\lambda \alpha $$
(B20)
for which $`C_c=\pi ^2`$. This frequency is given by
$$\omega _{\mathrm{max}}=\alpha \left(N\pi ^2\frac{k_0}{\lambda \alpha }1\right)^{1/2}$$
(B21)
Note, that the maximal frequency can be significantly higher than the typical rate $`\alpha `$ of the chemical cycle.
|
warning/0003/astro-ph0003138.html
|
ar5iv
|
text
|
# The Enigmatic Radio Afterglow of GRB 991216
## 1 Introduction
The intense gamma-ray burst GRB 991216 was detected on 1999 December 16.67 UT by the Burst and Transient Experiment (BATSE) on board the Compton Gamma Ray Observatory satellite (Kippen, Preece & Giblin (1999)). Follow-up observations with the PCA instrument on the Rossi X-ray Timing Explorer (RXTE) satellite resulted in the detection of a previously uncataloged X-ray source, which was subsequently seen to fade by a factor of five, seven hours later (Takeshima et al. (1999)). Uglesich et al. (1999) identified a fading optical source, at a position consistent with the RXTE transient, and shortly thereafter the radio counterpart was discovered (Taylor & Berger (1999)).
Here we present radio measurements of this burst from 1 GHz to 350 GHz. While the emission from X-ray and optical afterglow was fairly typical (Halpern et al. (2000)), the radio afterglow of GRB 991216 was unusual in two respects. First, the onset of the decay began much earlier than that in most radio afterglows. Second, the temporal decay indices in the radio, optical and X-ray bands are markedly different from each other. We explore a number of possible explanations for these behaviors.
## 2 Observations
Very Large Array (VLA<sup>1</sup><sup>1</sup>1The NRAO is a facility of the National Science Foundation operated under cooperative agreement by Associated Universities, Inc. NRAO operates both the VLA and the VLBA.): A log of the observations and flux density measurements are summarized in Table 1. We used J0509+1011 (at 8.46 GHz and 4.86 GHz) and J0530+135 (at 1.43 GHz) for phase calibration. J0542+498 was used for flux calibration at all frequencies.
Very Long Baseline Array (VLBA<sup>6</sup>): A single 2-hr observation was carried out at 8.42 GHz and 2-bit samples of a 64 MHz bandwidth signal in one hand of polarization were recorded. The nearby ($`<1.1^{}`$) calibrator J0509+1011, a core-jet source, was observed every 3 minutes for delay, fringe rate and fringe phase calibration. The total flux density of the calibrator was found to be 9.5% less than was measured by the VLA on the same day. Given that the jet of J0509+1011 is likely to have some extended emission that is not detectable by the VLBA, it is likely that the absolute flux calibration of the VLBA is well within its nominal value of 5%.
The radio afterglow was detected at a position of (epoch J2000) $`\alpha =5^h9^m31.2983^s`$, $`\delta =+11^{}17^{}7.262^{\prime \prime }`$, with (conservative 1-$`\sigma `$ error of 0.001<sup>′′</sup> in each coordinate). The source is unresolved with a size of less than 0.001<sup>′′</sup>.
Ryle Telescope: Observations at 15 GHz with the Ryle Telescope at Cambridge (UK) were made by interleaving 15 minute scans of GRB 991216 with short scans of the phase calibrator J0509+1011. The flux density scale was tied to observations of 3C 48 and 3C 286.
Owens Valley Radio Observatory Interferometer (OVRO): The source was observed for a single 13 hr track in two continuum 1 GHz bands (central frequencies 98.481 GHz and 101.481 GHz) under good 3-mm weather conditions. Gain calibration used the quasar 0528$`+`$134, while observations of Uranus and 3C 454.3 provided the flux density calibration scale with an estimated uncertainty of $`20`$%. See Shepherd et al. (1998), for details of the calibration and imaging. No source was detected (see Table 1).
James Clark Maxwell Telescope (JCMT<sup>2</sup><sup>2</sup>2The JCMT is operated by The Joint Astronomy Centre on behalf of the Particle Physics and Astronomy Research Council of the UK, the Netherlands Organization for Scientific Research, and the National Research Council of Canada.): Observations in the 350 GHz band were made using the Sub-millimeter Common-User Bolometer Array (Holland et al. (1999)). The data were taken under good sky conditions on both nights. For flux calibration we used the source CRL618, and assumed its flux density to be 4.57 $`\pm `$ 0.21 Jy. The pointing was monitored and found to vary by less than 2″. See Kulkarni et al. (1999) for details of data reduction. The source was not detected at either epoch (see Table 1). At the position of GRB 991216 we derive an average flux of $`0.28\pm 1.1`$ mJy.
## 3 Results
In Figure 1 we display the 8.46 GHz light curve, as well as the X-ray and optical (R-band) light curves obtained from measurements reported in the GRB Coordinates Network (GCN)<sup>3</sup><sup>3</sup>3$`\mathrm{http}://\mathrm{lheawww}.\mathrm{gsfc}.\mathrm{nasa}.\mathrm{gov}/\mathrm{docs}/\mathrm{gamcosray}/\mathrm{legr}/\mathrm{bacodine}/\mathrm{gcn}\mathrm{\_}\mathrm{main}.\mathrm{html}.`$ and Halpern et al. (2000). A noise-weighted least squares fit of the form $`F_\nu t^{\alpha _\nu }`$ was made to each of these light curves. Using all the 8.46 GHz data, including the upper limits, we derive $`\alpha _r=0.82\pm 0.02`$ ($`\chi _r^2=26.5/15`$; here $`\chi _r^2`$ is the reduced $`\chi ^2`$).
A similar least squares fit of the optical and X-ray data over the first three days (Figure 1), yields $`\alpha _o=1.33\pm 0.01`$ ($`\chi _r^2=11/28`$) and $`\alpha _x=1.61\pm 0.06`$ ($`\chi _r^2=7.7/3`$). From a more extensive dataset, Halpern et al. (2000) fit $`\alpha _o=1.07_{0.08}^{+0.17}`$ over the same time range. We will use their value of $`\alpha _o`$ in the discussion to follow. The relatively large value of $`\chi _r^2`$ for fit to the X-ray data presumably reflects the uncertainties inherent in converting the counts measured by three different instruments (RXTE-ASM, RXTE-PCA and Chandra ACIS) into Jansky flux units. Using the RXTE-PCA data alone avoids this cross-calibration issue and yields $`\alpha _x=1.61\pm 0.05`$.
## 4 The Unusual Nature of the Radio Afterglow: The Failure of the Basic Afterglow Model
The radio afterglow from GRB 991216 is unusual on two counts. First, the radio afterglow in the centimeter band does not show the usual rise to a peak value $`f_m`$ (at epoch $`t_m`$) before undergoing a power law decay. The radio flux appears to decline continuously starting from the epoch of the first observation. Thus $`t_m<1.49`$ d as compared to the 10–100 d seen in other bursts (e.g. Frail, Waxman & Kulkarni (2000), Frail et al. (1999), Frail et al. (1999)). Second, the temporal decay indices ($`\alpha _\nu `$) in the radio, optical and X-ray bands are markedly different from each other. Proceeding from radio to higher frequencies, $`\alpha _\nu `$ steepens by $``$0.4 every four decades in frequency.
In contrast, the optical and X-ray afterglow appears to find a straightforward explanation in the standard afterglow model in which a jet geometry is invoked (Halpern et al. (2000)). Below we show that the radio observations cannot be reconciled with a standard jet (or sphere) afterglow model. We then explore possible modifications to the standard model.
The simplest afterglow model is one in which the broad-band afterglow emission arises from the forward shock of a relativistic blast wave propagating into a constant density medium (Sari, Piran & Narayan (1998)). It is assumed that the electrons in the forward shock region are accelerated to a power law distribution for $`\gamma _e>\gamma _m`$, $`dN/d\gamma _e\gamma _e^p`$; here $`\gamma _e`$ is the Lorentz factor of the electrons, $`p`$ is the power law index and $`\gamma _m`$ is the minimum Lorentz factor. Gyration of these electrons in strong post-shocked magnetic fields gives rise to broad-band afterglow emission. Two modifications to this picture are routinely considered. (1) An inhomogeneous circumburst medium (specifically, $`\rho (r)r^2`$; here $`\rho `$ is the density at distance $`r`$ from the source). Such a circumburst medium is expected should GRBs originate from massive stars (Chevalier & Li (1999)). (2) A jet-like geometry for the blast wave (Sari, Piran & Halpern (1999)). This modification is motivated by the propensity of jets in astrophysical sources.
Regardless of these modifications, the broad band spectrum is composed of three characteristic frequencies: $`\nu _a`$, the synchrotron self-absorption frequency; $`\nu _m`$, the frequency at which the emission peaks (and attributed to the electrons with Lorentz factor $`\gamma _m`$), and $`\nu _c`$, the cooling frequency. Electrons radiating photons with frequency $`>\nu _c`$ cool on timescales faster than the age of the blast wave. The evolution of these frequencies is determined by the dynamics of the blast wave. The usual ordering of these frequencies at epochs relevant to the discussion here is (going from low to high frequencies) $`\nu _a`$, $`\nu _m`$ and $`\nu _c`$.
For GRB 991216 the early radio decay implies that $`\nu _m`$ is already below the centimeter radio band at 1.49 days. The steepening of the afterglow emission from optical to X-ray can be explained by placing $`\nu _c`$ between the optical and X-ray bands. The expected steeping $`\mathrm{\Delta }\alpha `$ is 1/4 which is marginally consistent with $`\alpha _o\alpha _x=0.54_{0.10}^{+0.18}`$. However, even if we ignore this, we are simply unable to explain the decay in the radio band, since no additional steepening is expected between $`\nu _m`$ and $`\nu _c`$.
The standard afterglow model can be made to agree with the light curves by postulating an energy slope $`p`$ which gradually steepens with increasing electron energy $`\gamma _e`$. We use the spherical, constant density afterglow model (Sari, Piran & Narayan (1998)) to convert, in each band, the observed decay index to $`p`$ and obtain: $`p=2.09\pm 0.03`$ (radio), $`p=2.43_{0.11}^{+0.23}`$ (optical), and $`p=2.81\pm 0.08`$ (if $`\nu _c`$ is below the X-ray band) or $`p=3.15`$ (if $`\nu _c`$ above the X-ray band). We are justified in applying the spherical model for early times ($`t<t_J`$ 2–5 d) since the jet geometry is manifested only for $`t>t_J`$ (Halpern et al. (2000)).
Curvature is both observed and modeled in the synchrotron spectra of the non-relativistic shocks from supernova remnants which are accelerating cosmic rays (e.g. Baring et al. 1999). To date, models of ultra-relativistic shocks favor a universal value of $`p`$, independent of energy (Vietri (2000), Gallant et al. (2000)), but non-linear effects have yet to be treated.
Nonetheless, the invocation of curvature in the energy distribution of the electrons cannot explain the observed broad-band spectrum (Figure 2) of the afterglow on December 18 (corresponding to 1.33 days after the burst). A plausible fit to the entire data is obtained with $`\nu _a=1.3`$ GHz, $`\nu _m=270`$ GHz and $`\nu _c=7\times 10^{16}`$ Hz and $`f_m=3.4`$ mJy; this fit is displayed by the dashed line in Figure 2. As the blast wave slows down, $`\nu _m`$ moves to lower values while preserving $`f_m`$ and thus we expect the flux in the centimeter band to rise, whereas the observed flux falls. If the afterglow has a jet-like geometry then the radio afterglow is expected to rise until the epoch $`t_J`$, and subsequently decay very slowly ($`f_\nu t^{1/3}`$) until $`\nu _m`$ passes through the centimeter band, after which we expect to see a decline similar to that seen in the optical ($`f_\nu t^{2.2}`$) (Harrison et al. (1999)). As can be seen from Figure 2, the radio observations are grossly inconsistent with these expectations, particularly the decay is much faster than $`t^{1/3}`$.
To summarize, while the optical and X-ray observations can be accounted for by a jet model, the radio observations are inconsistent with the standard model. This forces us to consider afterglow models in which the radio emission (at least in bulk) arises from a source other than the usual forward shock.
## 5 A Forward and Reverse Shock Model
The most natural explanation for two components would be an early contribution from a reverse shock followed by a forward shock element at later times. This is the explanation invoked to account for the early (1-2 day) radio emission from the afterglow of GRB 990123 (Sari & Piran (1999), Kulkarni et al. (1999) but see Galama et al. (1999)). The two bursts share several common features: in both cases, a jet has been deduced with $`t_J`$ few days, both were quite bright at gamma-ray energies and finally both had a seemingly small value of $`t_m`$ (as measured in the centimeter band). However, in the case of GRB 990123, the peak flux of the forward shock was $`f_m<260\mu `$Jy (Kulkarni et al. (1999)) and the radio light curve was dominated by the reverse shock. In contrast, the forward shock for GRB 991216 appears to be quite strong. This difference then explains the seemingly different radio light curves.
At late times (i.e. timescales greater than the duration of the burst) the flux from the reverse shock is expected to fall as $`t^{1.8}`$ (Kobayashi & Sari (2000)). In contrast, the forward shock emission rises as $`t^{1/2}`$ for $`t<t_J`$ and then slowly decays, $`t^{1/3}`$ until the $`\nu _m`$ moves into the centimeter band. Since $`t_J`$ is known from optical observations (Halpern et al. (2000)), the remaining unknowns are the strength of the reverse and forward shock emission.
In this picture, the reverse shock dominates the radio emission for the first few days and the model fit consists of mainly fitting a power law with $`f_\nu t^{1.8}`$. We note that at day 1.5, the VLA 8.46 GHz flux and the Ryle 15 GHz flux are comparable. This suggests that the reverse shock is already optically thin at 8.46 GHz at this epoch – similar to the situation for GRB 990123 (Kulkarni et al. (1999)). We deduce the parameters of the forward shock by fitting the radio to optical spectrum around $`t_J`$=5 days to the forward shock model (the contribution of the reverse shock is expected to be negligible thanks to the steep decay and since $`t`$ is comparable to $`t_J`$, the spherical fireball model is still applicable); we find $`\nu _m1.4\times 10^{12}`$ Hz and $`f_m=1`$ mJy. As can be seen from Figure 3 this reverse-forward model provides a reasonable fit to the observations.
There are two predictions of this model. First, we expect $`\nu _m`$ to cross the centimeter band at $`t_b=t_J(\nu _m/8.46\mathrm{GHz})^{1/2}64`$ d. For $`t>t_b`$, we expect the radio flux to decline as steeply as the optical flux does for $`t>t_J`$. The low flux values as measured at the VLA around this epoch are in agreement with this model.
A second prediction is that for $`t<t_b`$, we expect, the spectrum to rise as $`\nu ^{1/3}`$ for $`\nu <8.46`$ GHz. Unfortunately, the data are too sparse to rigorously test this expectation. Nonetheless, we note that at day 17.44, the spectrum between 1.43 and 8.46 GHz can be described by a simple power-law with slope $`\beta _r=0.45`$, steeper than a $`\nu ^{1/3}`$ slope by 3.6$`\sigma `$. We consider this to be the weakest point of the model but do not consider it fatal since the quoted uncertainties include only instrumental errors and do not include external effects such as interstellar scintillation.
The strongest confirmation of this model would have been the detection of an optical flash, as in the case of GRB 990123 (Akerlof et al. (1999)). The strong radio emission from the reverse shock allows us to predict (by scaling from the optical and radio observations of GRB 990123, Sari & Piran (1999)) that the flash would have been 8th magnitude or even brighter. Unfortunately, this event occurred during daytime and therefore was not observed by existing prompt optical counterpart experiments (LOTIS – H. S. Park; ROSTE – C. Akerlof; pers. comm.)
We end this section by noting a worrisome and puzzling issue: we are unable to provide a consistent explanation for the near-IR, optical and X-ray observations with a standard fireball afterglow spectrum. As noted in Figure 2, there is a broad maximum around $`2\times 10^{14}`$ Hz – suggesting that this is the peak frequency ($`\nu _m`$) of the fireball. Fitting a template afterglow spectrum we obtain the following: $`\nu _{m1}=2.1\pm 0.6\times 10^{14}`$ Hz, $`f_{m1}=150\pm 10`$ $`\mu `$Jy and $`\nu _{c1}=2\times 10^{16}`$ Hz. We note that a similar broad peak in the near IR (and attributed to $`\nu _m3\times 10^{14}`$ Hz at $`\mathrm{\Delta }t=0.5`$ d) was observed for GRB 971214 (Ramaprakash et al. (1998)). However, if we evolve this $`\nu _m`$ back in time (with $`\nu _mt^{3.2}`$) we predict a rising R-band light curve, inconsistent with the observations (Figure 1). Moving $`\nu _m`$ to lower frequencies solves this problem but we are left with no explanation for the “near-IR” bump.
## 6 A Two-Component Forward Shock Model
We now consider a model in which much of the radio emission arises as the forward shock of an additional fireball (hereafter the second fireball). The principal attraction of the second fireball is that we no longer need to relate the radio decay rate to those at optical and radio frequencies. We clarify that the optical and X-ray observations are explained by the forward shock of the fireball (the first fireball) discussed in the previous section. As noted earlier, there is good evidence suggesting that the first fireball is a jet. Thus the second fireball must be a more isotropic fireball and move at a smaller Lorentz factor. Indeed, in some GRB models, the central engine is expected to inject two fireballs: a high $`\mathrm{\Gamma }`$ jet and a low $`\mathrm{\Gamma }`$ spherical wind.
A reasonable fit to the radio data of this second fireball (FS 2; see Figure 2) on day 1.33 is provided by $`f_{m2}1.2`$ mJy, $`\nu _{m2}`$=7 GHz, and $`\nu _{a2}`$=2 GHz. The location of the cooling frequency $`\nu _{c2}`$ is unconstrained. As a test, we evolved the afterglow spectrum forward in time. The model does an excellent job reproducing the declining flux density from 1.43 and 8.46 GHz at 17.44 days (an observation which the reverse-forward shock model fails to explain), but at day 60.40 it predicts a 1.43 GHz flux of $`100`$ $`\mu `$Jy, where only an upper limit of $`57\pm 44`$ $`\mu `$Jy is measured. Again we consider this 3-$`\sigma `$ discrepancy as a major, but not fatal, weakness of this model.
The three inferred parameters ($`\nu _{m2}`$, $`f_{m2}`$, $`\nu _{a2}`$) allow us to obtain the energy of the blast wave and the density of the ambient medium (Wijers & Galama (1999)): $`E_{52}10^2`$ erg and $`n10^4`$ cm<sup>-3</sup>; these values are relatively insensitive to the value of the unknown $`\nu _c`$ (which is however constrained to lie above the optical band). The large $`E`$ and small $`n`$ are primarily due to the small value of $`t_m`$.
If this interpretation is correct then we have uncovered the first example of a GRB exploding in a very low density medium – perhaps the halo of a host galaxy. The dynamics of explosions is governed by the ratio $`E/n`$, and as noted above, this ratio is perhaps $`10^5`$ larger than that typically derived in other afterglow. For this reason, both fireballs, the high $`\mathrm{\Gamma }`$ and the low $`\mathrm{\Gamma }`$ fireballs, would then be expanding at high Lorentz factors days after the burst. The $`\mathrm{\Gamma }`$ for the low $`\mathrm{\Gamma }`$ fireball would be an impressive 20 one day after the burst, and the fireball would have had a size of 100 microarcseconds three weeks after the burst – almost within reach of measurable with VLBI techniques (cf. Taylor et al. (1997)). The jet fireball would be expanding even faster in which case the opening angle of the jet is not 6 (Halpern et al. (2000)) but only 1.
To summarize, the radio afterglow of GRB 991216 is unusual and cannot be explained by the standard forward shock model. A conventional reverse-forward shock model or an exotic two-component forward shock model can account for the observations, but each has one major (but not necessarily fatal) weakness. Finally, we have no explanation for the near-IR bump seen on day 1.33. GRB 991216 shows that there may be yet new surprises in GRB afterglows.
DAF thanks Chris Fassnacht, Steve Myers, Lin Yan and Jim Ulvestad for generously giving up portions of their VLA time so that GRB 991216 could be observed. DAF thanks J. Halpern for making his paper available prior to publication and Y. Gallant and M. Vietri for useful discussions. We would like to thank S. Jogee for preparing the first observations of this burst at the Owens Valley Observatory and A. Sargent for generously allocating the time on short notice. Research at the Owens Valley Radio Observatory is supported by the National Science Foundation through NSF grant number AST 96-13717. SRK’s research is supported by grants from NSF and NASA. RS and TJG are supported by Sherman Fairchild Fellowships.
|
warning/0003/cond-mat0003368.html
|
ar5iv
|
text
|
# THE VISCOUS SLOWING DOWN OF SUPERCOOLED LIQUIDS AND THE GLASS TRANSITION: PHENOMENOLOGY, CONCEPTS, AND MODELS
## I INTRODUCTION
When cooling a liquid, usually under isobaric P=1atm conditions, one can often bypass crystallization, thereby obtaining a supercooled liquid that is metastable relative to the crystal. When the temperature is further lowered, the viscosity of the liquid, as well as the relaxation times associated with the primary ($`\alpha `$) relaxation of all kinds of structural, dielectric, macro- and micro-scopic observables, increase rapidly, until a temperature is reached at which the liquid can no longer flow and equilibrate in the time scale of the experiment. The system effectively appears as a rigid amorphous material and is then called a glass. Glass formation thus results from the strong viscous slowing down of a liquid with decreasing temperature<sup>*</sup><sup>*</sup>*Although not as widely used, there are other ways of generating glassy structures, such as vapor deposition, in situ polymerization or chemical reactions. , a slowing down that can be considered as a classical example of a “jamming process”. A characteristic of this process that is unanimously recognized, a unanimity rare in this otherwise quite open and controversial field, is that it is a dynamic effect. The so-called “glass transition” is not a bona fide thermodynamic phase transition, but represents a crossover below which a liquid falls out of equilibrium on the experimental time scale. The transition temperature, $`T_g`$ depends on this time scale, set either by the observation time (corresponding, for instance, to a relaxation time of $`10^2`$ or $`10^3`$ sec or a viscosity of $`10^{13}`$ Poise) and/or by the cooling rate (in a typical differential scanning calorimetry measurement, $`10`$ K per minute). The dependence on cooling rate is, however, weak, a difference of few K in $`T_g`$ for an order-of-magnitude change of the rate; this is so because on further lowering of the temperature the viscosity and $`\alpha `$-relaxation times rapidly become enormous and out of reach of any experimental technique.
In the following, we shall focus on the jamming process occurring in the supercooled liquid state. Both the crystalNote that for several liquids, such as m-fluoroaniline and dibutylphtalate at atmospheric pressure and atactic polymers, crystallization has never been observed., which is the stable phase below the melting point $`T_m`$ but can be ignored in discussing the glass transition, and the glassy state itself will be excluded from the present discussion. By appropriately eliminating the crystal (experimentally as well as theoretically), metastable supercooled liquids can be treated by equilibrium thermodynamics, statistical mechanics, and conventional linear-response formalisms. Glasses on the other hand, although mechanically stable, are out-of-equilibrium states; especially near $`T_g`$, they display nonlinear responses and relaxations known as aging or annealing, and their properties depend on their history of preparation. These phenomena will not be considered in this review.
## II SALIENT PHENOMENOLOGY
The distinctive feature of glass-forming liquids is the dramatic, continuous increase of viscosity and $`\alpha `$-relaxation times with decreasing temperature. This sort of jamming is observed in a large variety of substances: covalently bonded systems like $`SiO_2`$, hydrogen-bonded liquids, ionic mixtures, polymers, colloidal suspensions, molecular van der Waals liquids, etc. The emphasis will be placed on those liquids (the vast majority) that do not form 2- or 3-dimensional networks of strong bonds because they show the most striking behavior when passing from the high-temperature liquid phase to the deeply supercooled and very viscous regime.
### 1 Strong, super-Arrhenius T-dependence of viscosity and $`\alpha `$-relaxation times
The viscosity $`\eta `$ and $`\alpha `$-relaxation times can change by $`15`$ orders of magnitude for a mere decrease of temperature by a factor twoThe (shear) viscosity $`\eta `$ can be related to a time characteristic of $`\alpha `$-relaxation, the average shear stress relaxation time $`\tau _s`$, by $`\eta =G_{\mathrm{}}\tau _s`$, where $`G_{\mathrm{}}`$ is the infinite-frequency shear modulus; $`G_{\mathrm{}}`$ is typically of the order of $`10^{10}10^{11}`$ erg.cm<sup>-3</sup>, so that a viscosity of $`10^{13}`$ Poise roughly corresponds to a time of $`10^2`$ or $`10^3`$ sec.. Such a dramatic variation is conveniently represented on a logarithmic plot of $`\eta `$ or $`\tau _\alpha `$ versus $`1/T,`$ i.e., an Arrhenius plot: see Fig. 1a. A system like $`GeO_2`$, an example of a network-forming system, is characterized by an almost linear variation, which indicates an Arrhenius temperature dependence. For all other liquids on the figure there is a marked upward curvature, which represents a faster-than-Arrhenius, or super-Arrhenius, temperature dependence and is often described by an empirical Vogel-Fulcher-Tammann (VFT) formula (also called Williams-Landel-Ferry formula in the context of polymers studies),
$$\tau _\alpha =\tau _0\mathrm{exp}\left(D\frac{T}{TT_0}\right),$$
(1)
where $`\tau _0`$, $`D`$ and $`T_0<T_g`$ are adjustable parameters. On the basis of such Arrhenius plots, with the temperature scaled by $`T_g`$, Angell proposed the now standard classification of glass-forming liquids into strong (Arrhenius-like) and fragile (super-Arrhenius) systems; in Eq. (1), the smaller the value of $`D`$, the more fragile the liquid. There are, of course, alternative fitting formulas that have been used, some which do not imply a singularity at a nonzero temperature as does the expression in Eq. (1).
A different way of representing the phenomenon is to plot the effective activation free energy for $`\alpha `$-relaxation, $`E(T)`$, obtained from
$$\tau _\alpha =\tau _{\alpha ,\mathrm{}}\mathrm{exp}\left(\frac{E(T)}{k_BT}\right),$$
(2)
where $`k_B`$ is the Boltzmann constant and $`\tau _{\alpha ,\mathrm{}}`$ is a high-$`T`$ relaxation time, or from a similar equation for the viscosity. This is illustrated in Fig. 1b, where the temperature has been scaled for each liquid to a temperature T\* above which the dependence is roughly Arrhenius-like. Although the determination of $`T^{}`$ is subject to some uncertainty, the procedure emphasizes the crossover from Arrhenius-like to super-Arrhenius behavior that is typical of and quite distinct in most supercooled liquids. The appreciable size of the effective activation free energies $`E(T)`$, namely, $`40k_BT_g`$ at the glass transition, is indicative of thermally activated dynamics. Such a large effective activation free energy for weakly bonded fragile molecular liquids such as orthoterphenyl is an intriguing feature of the phenomenology. Another peculiar property of $`E(T)`$ for fragile systems is that it increases significantly between $`T^{}`$ and $`T_g`$ (a factor $`3`$, i.e., a factor of $`5`$ or $`6`$ in units of the thermal energy $`k_BT`$, for weakly bonded fragile liquids). Such a variation is not commonly encountered. For instance, in the field of critical phenomena, the slowing down of dynamics that occurs when approaching the critical point is usually characterized by a power law growth of the relaxation time; in terms of effective activation free energy, this corresponds to a logarithmic growth and it is slower than the variation described by the VFT formula, Eq. (1). Unusually strong slowing down, with exponentially growing times similar to Eq. (1), is found in some disordered systems like the random field Ising model and it is known as “activated dynamic scaling”.
### 2 Nonexponential relaxations
In an “ordinary” liquid above the melting point, relaxation functions are usually well described, after some transient time, by a simple exponential decay. Deviations are observed, but they are neither systematic nor very marked. The situation changes at lower temperatures, and the $`\alpha `$-relaxation is no longer characterized by an exponential decay. A better representation is provided by a “stretched exponential” (or Kohlrausch-Williams-Watts function),
$$f_\alpha (t)\mathrm{exp}\left[\left(\frac{t}{\tau _\alpha }\right)^\beta \right],$$
(3)
where $`\beta `$ is the stretching parameter; the smaller $`\beta `$ the more “stretched” the relaxation. Although not unambiguously established, the degree of departure from exponential behavior, or stretching, appears to increase (i.e., $`\beta `$ decreases) with decreasing temperature.
Alternatively, in frequency space, the spectrum of the imaginary part of the susceptibility, which is characterized by a peak at a frequency $`\omega _\alpha 1/\tau _\alpha `$, tends to be broader (when plotted as a function of $`log(\omega )`$ ) than the simple Lorentzian or Debye spectrum that is just the Fourier transform of the time-derivative of an exponential relaxation function (see Fig. 2). Fitting formulas related to Eq. (3), formulas like the Cole-Davidson function for frequency-dependent susceptibilities, $`(1i(\omega /\omega _\alpha ))^\beta ^{}`$, are used to fit the spectroscopic data, but similar trends are observed: the $`\alpha `$ peak, as observed for instance in the imaginary part of the dielectric susceptibility as a function of $`log(\omega )`$, broadens as the temperature is lowered towards $`T_g`$, which indicates increasing departure from Debye/exponential behavior. Except for network-forming systems, the stretching of the $`\alpha `$ relaxation is significant ($`\beta `$ is typically between $`0.3`$ and $`0.6`$ for fragile liquids at $`T_g`$). However, and this point may not have been given enough attention, the stretching, or broadening in frequency space, is relatively small when compared to the extremely rapid variation with temperature of the $`\alpha `$-relaxation time itself. This is to be contrasted for instance with the activated critical slowing discussed above. There, the power law growth of the activation free energy when the temperature is decreased toward the critical point comes with a more striking stretching of the relaxation function that occurs on a logarithmic scale: in this case, in place of a stretched exponential behavior as in Eq. (3), $`\mathrm{ln}(f(t))`$ goes as some power of $`(1/\mathrm{ln}(t))`$.
### 3 No marked changes in structural quantities
It is tempting to associate the huge increase in $`\alpha `$-relaxation times and viscosity with the growth of a structural correlation length. However, no such growth has been detected so far in supercooled liquids. Quite to the contrary, the variation of structure in liquids and glasses, as measured in neutron and X-ray diffraction experiments, appear rather bland (see Fig. 3). The ordinary, high-temperature liquid has only short-range order whose signature in the static structure factor $`S(Q)`$ is a broad peak (or a split peak for some molecular systems as illustrated in Fig. 3) at a wave vector $`Q`$ that roughly corresponds in real space to some typical mean distance between neighboring molecules. As the temperature is lowered and the supercooled regime is entered, there are small, continuous variations of the structure factor that mostly reflect the change in density (typically, a $`5\%`$ change between $`T_m`$ and $`T_g`$) and, possibly, some adjustments in the local arrangements of the molecules. There is no sign, however, of a significantly growing correlation length, nor of the appearance of a super-molecular length.
In network-forming and $`H`$-bonded systems, an additional “pre-peak” is sometimes detectable at wave vectors somewhat lower than that of the main peak, but it is attributed to specific effects induced by the strongly directional intermolecular bonds and not to a length scale that would correlate with the viscous slowing down.
In contrast to this lack of structural signature for the existence of an increasing super-molecular correlation length with decreasing temperature, there is significant evidence, as discussed below, that corresponding “dynamical” correlation length do exist.
### 4 Rapid entropy decrease and Kauzmann paradox
The absence of marked changes in the structure, at least at the level of two-particle density correlations, or of a strong increase in any directly measured static susceptibility is a puzzling feature of the jamming process associated with glass formation. The only static quantity that shows behavior that might be relevant is the entropy. Below the melting point, $`T_m`$, the heat capacity $`C_p(T)`$ of a supercooled liquid is larger than that of the corresponding crystal. (At $`T_g`$, the $`C_p`$ of the liquid drops to a value that is characteristic of the glass and is close to the $`C_p`$ of the crystal, but this is a consequence of the system no longer being properly equilibrated.) As a result of this “excess” heat capacity, the entropy difference between the liquid and the crystal decreases with temperature, typically by a factor of $`3`$ between $`T_m`$ and $`T_g`$ for fragile liquids. The effect is illustrated in Fig. 4 and leads to the famous Kauzmann paradox: if the entropy difference is extrapolated to temperatures below $`T_g`$, its extrapolated value vanishes at some nonzero temperature $`T_K`$, which results in the unpleasant feature that the entropy of the liquid becomes equal to that of the crystal (even more unpleasant: if the extrapolation is carried to still lower temperatures, the entropy of the liquid becomes negative, which violates the third law of thermodynamics). The paradox is that this extrapolated entropy crisis is avoided for a purely dynamic reason, the intervention of the glass transition: what would occur if one were able to keep the supercooled liquid equilibrated down to temperatures below $`T_g`$? There are certainly many ways to answer the question. The paradox could be resolved by the existence between $`T_g`$ and $`T_K`$ of an intrinsic limit of metastability of the liquid or of a second-order phase transition<sup>§</sup><sup>§</sup>§Note that a low-$`T`$ first-order transition does not resolve the paradox because it can be supercooled. (a speculation that gains additional credibility with the observation that the VFT temperature $`T_0`$ at which the extrapolated viscosity and $`\alpha `$-relaxation times diverge (see Eq. (1)) is often found close to $`T_K`$). Even more simply, one might find that the extrapolation breaks down above $`T_K`$ and that the entropy-difference curve levels off and goes smoothly to zero at zero $`K`$, in much the same way as it does in the Debye theory of crystals. These are of course all speculations, but it remains that the rapid decrease of the entropy of the supercooled liquid relative to that of the crystal represents an intriguing aspect of the phenomenology of fragile glass-formers.
### 5 Two-step relaxation and secondary processes
As we stressed before, the salient features related to glass formation concern the long-time (low-frequency) primary or $`\alpha `$ relaxation. As the $`\alpha `$-relaxation time increases with decreasing temperature, so too does the window between this time and typical microscopic, picosecond or sub-picosecond times. When the relaxation function is plotted against the logarithm of the time, one then observes what is sometimes called a “two-step relaxation”. An illustration is given in Fig. 5 by the dynamic structure factor of the fragile ionic glass-former $`Ca_{0.4}K_{0.6}(NO_3)_{1.4}`$ obtained by neutron techniques. At high temperature, the relaxation function is essentially a one-step process. However, as the liquid becomes more viscous, the relaxation proceeds in two steps separated by a plateau. Although the terminology is far from being universally accepted, the approach to the plateau from the short-time side is often referred to as $`\beta `$ or fast-$`\beta `$ relaxation. If one is to fit the long-time part by a stretched exponential (Eq. (3)), there is a large range of “mesoscopic” times that is not adequately described and that widens as the temperature is lowered. Power law functions of time are often used to reproduce the relaxation function in this mesoscopic range.
This two-step relaxation feature is common to all fragile liquids. In addition, there may also appear additional secondary processes, detected first by Johari and Goldstein in dielectric spectroscopy. Such secondary processes, whose presence and strength strongly vary from one liquid to another, have characteristic frequencies that are intermediate between those of the $`\alpha `$ and fast-$`\beta `$ relaxations. They are denoted Johari-Goldstein-$`\beta `$, slow-$`\beta `$, or simply $`\beta `$ processes. To make the description more complete, one should also mention the so-called “boson peak” that may be present on the high-frequency side ($`10^210^3Ghz`$) of light and neutron scattering (or absorption) spectra. Here we do not discuss either the slow-$`\beta `$ processes or the boson peak.
## III A SELECTION OF QUESTIONS
After this brief review of the salient aspects of the phenomenology of supercooled liquids as they get glassy, we discuss in more detail a number of questions, whose answers give justification or put constraints on the theoretical picture one can build to explain the viscous slowing down.
### 1 How universal is the behavior of glass-forming liquids ?
Universality is a key concept in physics and it has proven to be central in the field of critical phenomena. By the standards of critical phenomena studies, the observed behavior of glass-forming liquids is not universal, the main reason being that no singularity is detected experimentally (or approached asymptotically close), as stressed above. However, if one is willing to take a broader view of the notion of universality, one can find considerable generality or “universality” in the properties, those mostly associated with long-time and low-frequency phenomena, that characterize the approach to the glass transition. For instance, the super-Arrhenius $`T`$-dependence of the viscosity and $`\alpha `$-relaxation times and the nonexponential character of the relaxation function are observed for virtually all glass-formers, be they polymeric, $`H`$-bonded, ionic, van der Waals, etc., with the exception of a minority of strong network-forming systems; and, for a given liquid, these properties are found by a large variety of experimental techniques, such as dielectric relaxation, light and neutron scattering, NMR, viscosity measurements, specific heat spectroscopy, volume and enthalpy relaxation, optical probe methods.
The presence of an underlying ”universality” is supported by the fact that experimental data covering a wide range of temperatures and a great diversity of substances can be collapsed onto master curves with only a small number of species-dependent adjustable parameters. A good example is provided by the scaling plot of the frequency-dependent dielectric susceptibility proposed by Nagel and coworkers. As shown in Fig. 6, the data taken over a $`13`$ decade range of frequencies for many different liquids can be placed with good accuracy onto a single curve after scaling with only three parameters associated with the $`\alpha `$-peak position, width and intensity. The master-curve for the temperature dependence of the effective activation free energy for viscosity and $`\alpha `$ relaxation put forward by Kivelson et al. is another example. It is nonetheless fair to say that the fits resulting from these various scaling procedures are far from perfect, which leaves room for debate and conflicting interpretations. One can also ask the question whether the universality holds only up to the implied high-frequency cut-off of the susceptibility scaling curve of Nagel et al. (i.e., whether one should be focusing on slow behavior) or whether it extends higher in light of the fact that similarities have also been observed in the high-frequency susceptibilities.
### 2 Is the $`\alpha `$-relaxation homogeneous or heterogeneous?
The observation stressed above that the $`\alpha `$-relaxation is nonexponential in the supercooled liquid range and its representation by a stretched exponential as in Eq. (2) can be formally interpreted in terms of a superposition of exponentially decaying functions with a distribution of relaxation times; but, this per se does not guarantee that the dynamics be “heterogeneous”, in the sense that relaxation of the molecules differs from one environment to another with the environment life time being longer than the relaxation time. An alternative explanation can be offered within a “homogeneous” picture in which relaxation of the molecules everywhere in the liquid is intrinsically nonexponential.
In recent years, there has been mounting evidence that heterogeneities, sufficiently long-lived to be relevant to the $`\alpha `$-relaxation and to be at least partly responsible for its nonexponential feature, do exist in supercooled liquids. The heterogeneous character of the slow dynamics has been demonstrated in several experiments: multi-dimensional NMR, photobleaching probe rotation measurements, nonresonant dielectric hole burning. These techniques involve the selection of a sub-ensemble of molecules in the sample that is characterized by a fairly narrow distribution of relaxation times (and in general a relaxation slower than average) and the further monitoring of the gradual return to the equilibrium situation. Additional evidence of the spatially heterogeneous nature of the dynamics in fragile supercooled liquids is provided by the so-called breakdown of the Stokes-Einstein relation between the translational diffusion constant and the viscosity, and the concomitant “decoupling” between rotational and translational time scales: see Fig.7.
The size of the heterogeneities is not directly observable in the above mentioned experiments, but various estimates, obtained, e. g., from optical studies of the rotational relaxation of probes of varying size, NMR measurements, the study of excess light scattering, and the influence of a well-defined $`3`$-dimensional confinement lead to a typical length of several nanometers in different fragile liquids near $`T_g`$. One should recall that these signatures are all dynamical and that no signature at such a length scale has been detected so far in small-angle neutron and X-ray diffraction data. If the heterogeneous character of the $`\alpha `$-relaxation appear reasonably well established, at least for deeply supercooled fragile liquids, several points concerning the lifetime, the size, and the nature of the heterogeneities need still be settled.
### 3 Is density or temperature the dominant control variable?
The phenomenon of viscous slowing down and glass formation as it is studied most of the time (and described in the preceding sections) takes place under isobaric $`P=1atm`$ conditions. As a consequence, when the temperature is lowered, there is also an increase of the density of the liquid. This increase is small (a typical variation of $`5\%`$ between $`T_m`$ and $`T_g`$), but it could still have a major influence on the dynamics. Actually, there are theoretical models of jamming, such as those based on free volume concepts and hard sphere systems, that attribute the spectacular increase of viscosity and $`\alpha `$-relaxation times of fragile glass-formers (almost) entirely to the density changes. It is thus important to evaluate the role of density and temperature in driving the jamming process that leads to the glass transition at 1 atm.
Basic models and theories are usually formulated in terms of either density or temperature as control variable, but experiments are carried out with pressure and temperature as external control variables. The data must be converted, when enough experimental results are available, in order to analyze the influence of density at constant temperature and that of temperature at constant density, for a range of density and temperature that is characteristic of the phenomenon at 1 atm. Extant analyses are far from exhaustive. However, as illustrated in Fig. 8, the characteristic super-Arrhenius T-dependence of the viscosity, $`\eta `$, and $`\alpha `$-relaxation times, $`\tau _\alpha `$, appears predominantly due to the variation of temperature and not to that of density. This conclusion is confirmed by a comparative study of the contributions induced by density variations (at constant temperature) and by temperature variations (at constant density) to the rate of change of $`\eta `$ and $`\tau _\alpha `$ at constant (low) pressure in the viscous liquid regime of several molecular and polymeric glass-formers.
How general is the above result? Temperature appears to be the dominant variable controlling the viscous super-Arrhenius slowing down of supercooled liquids at low pressure, but this may not be the case at much higher pressure (although not much data are presently available to confirm this point), and it is most likely not true for describing the concentration-driven congestion of dynamics in colloidal suspensions. In the absence of a “super-universal” picture of the jamming associated with glass formation, we shall restrict ourselves, as we have implicitly done above, to the consideration of supercooled liquids at 1 atm.
### 4 What are the relevant characteristic temperatures?
There is no unbiased way of presenting the phenomenology of glass-forming liquids. Choices must be made about the emphasis put on the different aspects, about the best graphic representations, and about the way in which one analyzes experimental data. To make sense out of the wealth of observations and measurements, it is natural to look for characteristic temperatures about which to organize and scale the data. However, since no singularity is directly detected, the selection of one or several relevant temperatures is far from straightforward. The temperatures that can be easily determined experimentally are the boiling point, $`T_b`$, the melting point, $`T_m`$, and the glass transition temperature(s), $`T_g`$. Unfortunately, the former two are generally considered as irrelevant to the jamming phenomenon, and the latter has an operational rather than a fundamental nature (see the introduction).
Several other candidates have been suggested, that can be split into two groups. First, there are “extrapolation temperatures” below $`T_g`$, i.e., temperatures dynamically inaccessible to supercooled liquids, at which extrapolated behavior diverges or becomes singular. This is the case of the VFT temperature $`T_0`$ (see Eq. 1) and the Kauzmann temperature $`T_K`$ (see section II-4 and Fig. 4). In the second group are “crossover temperatures” above $`T_g`$ at which a new phenomenon seems to appear (decoupling of rotations and translations, emergence of a secondary $`\beta `$ process, etc.), a crossover of behavior or a change of $`\alpha `$-relaxation mechanism seems to take place (passage from Arrhenius to super-Arrhenius $`T`$-dependence, arrest of the relaxation mechanisms described by the mode-coupling theory, putative emergence of activated barrier crossing processes, etc.). A variety of such temperatures for the fragile glass-former $`OTP`$ are shown in Fig. 9: the location of the different characteristic temperatures is illustrated on an Arrhenius plot of the viscosity. It is interesting to note that most putative crossover behaviors occur in the region of strong curvature where $`\eta 110^2`$ Poise or $`\tau _\alpha 10^{10}10^8`$ sec, while, on the other hand, the temperatures obtained by extrapolation of data to low T lie fairly close to each other, some 40 K below $`T_g`$.
### 5 What can be learned from computer simulations?
Computer simulation studies, in particular those based on Molecular Dynamics algorithms, have proven extremely valuable in investigating the structure and dynamics of simple, ordinary liquids. Their contribution to the understanding of the glass transition is unfortunately limited, the main reason being the restricted range of lengths and times that are accessible to Molecular Dynamics simulations: typical simulations on atomistic models consider $`10^310^4`$ atoms and can follow relaxations for less than $`10^8`$ sec (when expressing the elementary time step in terms of parameters characteristic of simple liquids). As a result, the viscous, deeply supercooled regime of real glass-forming liquids, where strong super-Arrhenius behavior, heterogeneous dynamics, and other significant features associated with the jamming process develop, is out of reach, as is the laboratory glass transition that occurs on a time scale of $`10^2`$ or $`10^3`$ sec. Simple liquid models do form “glasses” on the observation, i.e., simulation, time with many of the attributes of the laboratory glass transition: abrupt change in the thermodynamic coefficients, dependence on the cooling rate, aging effects, etc However, these models are effectively high-temperature structures, since they correspond to liquid configurations that are kinetically arrested at a temperature at which the primary relaxation time is of the order only of nanoseconds. Even with systems specially designed to avoid crystallization, such as binary Lennard-Jones mixtures, the cooling rates to prepare glasses ($`10^8`$ to $`10^9`$ K/sec) are orders of magnitude higher than those commonly used in experiments.. However, these supercooled simulation liquids are not truly fragile (the $`T`$-dependence of the $`\alpha `$-relaxation time shows only small departure from Arrhenius behavior) and are not deeply supercooled so that one´s ability to extract insights into the deeply supercooled fragile liquids is questionable.
Computer simulations can be useful in studying the moderately supercooled liquid region, where one can observe the onset of viscous slowing down (see for instance Fig. 9). The major interest of such studies is that static and dynamic quantities that are not experimentally accessible can be investigated, such as multi-body (beyond two-particle) correlations involving a variety of variables and microscopic mechanisms for transport and relaxation. They also allow for testing in detail theoretical predictions made in the relevant window of times and lengths (e. g., those of the mode-coupling theory) and for analyzing properties associated with configurational or phase space (see below).
In addition to the much studied one- or two-component systems of spheres with spherically symmetric interaction potentials, the models investigated in computer simulations can be divided into two main groups: on one hand, more realistic microscopic models for molecular glass-formers that attempt to describe species-specific effects; on the other hand, more schematic systems, coarse-grained representations, lower-dimensional systems or toy-models, that bear less detailed resemblance with real glass-formers, but can be studied on much longer time scales and with bigger system sizes.
## IV THEORETICAL APPROACHES.
There is a large number of theories, models, or simply empirical formulae that attempt to reproduce pieces of the phenomenology of supercooled liquids. There are fewer approaches, however, that address the question of why and how the viscous slowing down leading to the glass transition, with its salient characteristics described in the preceding sections, occurs in liquids as they are cooled. In the following, we shall briefly review the main theoretical approaches, with an emphasis on the concepts and methods that may prove useful in other areas of physics where some sort of jamming process is also encounteredModels addressing more specific questions, such as the “coupling model” or the “continuous time random walk”, approach are discussed in the reviews cited in .. More specifically, we shall discuss phenomenological models based on free volume and configurational entropy, the description of a purely dynamic arrest resulting from mode-coupling approximations, ideas relying on the consideration of the topographic properties of the configurational space (energy and free-energy landscapes) or on the analogy with generalized spin glass models, and approaches centered on the concept of frustration.
### 1 Free volume
Free-volume models rest on the assumption that molecular transport in viscous fluids occurs only when voids having a volume large enough to accommodate a molecule form by the redistribution of some “free volume”, where this latter is loosely defined as some surplus volume that is not taken up by the molecules. In the standard presentation by Cohen and Turnbull, a molecule in a dense fluid is mostly confined to a cage formed by its nearest neighbors. The local free volume, $`v_f`$, is roughly that part of a cage space which exceeds that taken by a molecule. It is assumed that between two events contributing to molecular transport, a reshuffling of free volume among the cages occurs at no cost of energy. Assuming also that the local free volumes are statistically uncorrelated, one derives a probability distribution, $`P(v_f)`$, which is exponential,
$$P(v_f)\mathrm{exp}\left(\gamma \frac{v_f}{\overline{v_f}}\right),$$
(4)
where $`\overline{v_f}`$ is the average free volume per molecule and $`\gamma `$ is a constant of order $`1`$. Since the limiting mechanism for the diffusion of a molecule is the occurrence of a void, i.e., a local free volume $`v_f`$ larger than some critical value, $`v_0`$, that is approximately equal to the molecular volume, the diffusion constant $`D`$ is given by the probability of finding a free volume equal to $`v_0`$; this leads to an expression for $`D`$, and by extension for the viscosity $`\eta `$,
$$\eta 1/D\mathrm{exp}\left(\gamma \frac{v_0}{\overline{v_f}}\right),$$
(5)
which is similar to the formula first proposed by Doolittle.
In the Cohen-Turnbull formulation, the average free volume per molecule is given by $`\overline{v_f}=vv_0`$, where $`v=1/\rho `$ is the average total volume per molecule. The free-volume concept, in zeroth order, relies on a hard-sphere picture in which thermal activation plays no role. For application to real liquids, temperature enters through the fact that molecules, or molecular segments in the case of polymers, are not truly “hard” and that, consequently, the constant-pressure volume is temperature-dependent,
$$\overline{v_f(T)}\alpha _P(TT_0),$$
(6)
where $`\alpha _P`$ is the coefficient of isobaric expansivity and $`T_0`$ is the temperature at which all free volume is consumed, i.e., $`v=v_0`$. Inserting the above equation in the Doolittle formula, Eq. (4), gives the VFT expression, Eq. (1). An unanswered, but fundamental question associated with Eq. (6) is why the free volume should be consumed at a nonzero temperature, $`T_0`$? An extended version of the free-volume approach has been developed by Cohen and Grest, in which the cages or “cells” are divided into two groups, liquid-like and solid-like, and concepts from percolation theory are included to describe the dependence upon the fraction of liquid-like cells. (See also the model for molecular diffusivity in fluids of long rod molecules by Edwards and Vilgis.)
The main criticisms of the free volume models are (i) that the concept of free volume is ill-defined, which results in a variety of interpretations and difficulty in finding a proper operational procedure even for simple model systems, and (ii) that the pressure dependence of the viscosity (and $`\alpha `$-relaxation times) is not adequately reproduced. This latter feature has been emphasized in many studies, and it is a consequence of the observation made above (see II-3) that the viscous slowing down of glass-forming liquids at 1 atm and more generally at low pressure is primarily controlled by temperature and not by density or volume. Glass formation in supercooled liquids does not predominantly results from the drainage of free volume, but rather from thermally activated processes.
### 2 Mode-coupling approximations
The theory of glass-forming liquids that has had the highest visibility for more than a decade is the mode coupling theory. It predicts a dynamic arrest of the liquid structural relaxation without any significant change in the static properties. All structural quantities are assumed to behave smoothly and jamming results from a nonlinear feedback mechanism that affects the relaxation of the density fluctuations. Formally, the theory involves an analysis of a set of nonlinear integro-differential equations describing the evolution of pair correlation functions of wave-vector- and time-dependent fluctuations that characterize the liquid. These equations have the form of generalized Langevin equations, and they can be derived by using the Zwanzig-Mori projection-operator formalism. The equation for the quantity of prime interest in the theory, the (normalized) correlation function of the density fluctuations,
$$\varphi _Q(t)=\frac{<\rho _Q(t)\rho _Q^{}(0)>}{<|\rho _Q(0)|^2>},$$
(7)
where $`\rho _Q(t)=_j\mathrm{exp}(i\mathrm{𝐐𝐫}_j)`$ and $`𝐫_j`$ denotes the position of the $`j`$th particle, can be written as
$$\frac{d^2}{dt^2}\varphi _Q(t)+\mathrm{\Omega }_Q^2\varphi _Q(t)+_0^t𝑑t^{}m_Q(tt^{})\frac{d}{dt^{}}\varphi _Q(t^{})=0,$$
(8)
where $`\mathrm{\Omega }_Q`$ is a microscopic frequency obtainable from the static structure factor, $`S(Q)<|\rho _Q(0)|^2>`$, and $`m_Q(t)`$ is the time-dependent memory function that is formally related to the correlation function of a Q-dependent random force. The above equation being exact, the crux of the mode-coupling approach consists in formulating an approximate expression for $`m_Q(t)`$. The mode-coupling scheme has been implemented for liquids both in the frame of the kinetic theory of fluids and that of the fluctuating nonlinear hydrodynamics. It essentially boils down to approximating the memory function $`m_Q(t)`$ as the sum of a bare contribution coming from the fast relaxing variables and a mode-coupling contribution coming from the slowly decaying bilinear density modes,
$$m_Q(t)=\gamma \delta (t)+\underset{𝐐^{}}{}V_{\mathrm{𝐐𝐐}^{}}\varphi _𝐐^{}(t)\varphi _{|𝐐𝐐^{}|}(t),$$
(9)
where the vertices $`V_{\mathrm{𝐐𝐐}^{}}`$ can be expressed in terms of the static structure factor. The self-consistent solution of the resulting set of nonlinear equations predicts a slowing of the relaxation that is attributed, within a purely homogeneous picture (see II-2), to a cage effect and to the feedback mechanism above mentioned. This solution exhibits a dynamic arrest at a critical point, $`T_c`$, which represents a transition from an ergodic to a nonergodic state with no concomitant singularity in the thermodynamics and structure of the system. The main achievements of the mode-coupling approach are the predicted anomalous increase in relaxation time and the appearance of a two-step relaxation process with decreasing temperature, as indeed observed in real fragile glass-formers (compare Fig. 10 to Fig. 7) and in molecular dynamics simulations. Early on, however, it was realized, both from empirical fits to experimental data and from comparison to simulation data on model systems, that the dynamic arrest at $`T_c`$ did not describe the observed glass transition at $`T_g`$ nor the transition to an “ideal glass” at a temperature below $`T_g`$. This is illustrated in Fig. 11. Thus, the $`T_c`$ was interpreted as a temperature above $`T_g`$. The singularity at $`T_c`$ is avoided because of the breakdown of the simple mode-coupling approximation, Eq. 9, and the $`T_c`$ of what is called the “idealized” mode-coupling theory is taken as a crossover below which additional relaxation mechanisms, such as activated processes, presumably take over. Unfortunately, beyond some empirical introduction, activated processes are not theoretically described by mode-coupling approaches, and so the theory of $`\alpha `$ relaxation has not been extended to temperatures below $`T_c`$. To draw once again a parallel with critical phenomena (where a singularity occurs at $`T_c`$ in the structure and the thermodynamics of the system), mode-coupling approximations, as formulated for instance by Kawasaki, are known to describe quite well the standard critical slowing down, but not the activated dynamic scaling such as that observed in the random field Ising model (see section I-1). This failure is related to the underlying nature of the approximation that corresponds to a one-loop self-consistent resummation scheme in a perturbative treatment (see also below in III-4 the parallel with spin glass models).
Mode-coupling approaches can thus describe at best the dynamics of moderately supercooled liquids<sup>\**</sup><sup>\**</sup>\**It is possible that they are also applicable to the fast-$`\beta `$ relaxations even below $`T_c`$. (see Fig. 11). Because of the many detailed predictions it makes in this regime, the mode-coupling theory has stimulated the use and the development of experimental techniques, such as neutron and depolarized light scattering, and molecular dynamics simulations that are able to probe the early stage of the viscous slowing down; but, the very fact that the predicted dynamic singularity is not observed makes it difficult to reach any clear-cut conclusion about the quantitative adequacy of the theory, and this has led to much debate in recent years.
### 3 Configurational entropy and (free) energy landscape
The existence of a crossover temperature in the moderately supercooled liquid region where $`\alpha `$-relaxation times are of the order of $`10^9`$ sec (hence in the same region as the $`T_c`$ predicted by the mode-coupling theory) was advocated 30 years ago by Goldstein. Goldstein argued that below this crossover flow is dominated by potential energy barriers that are high compared to thermal energies and slow relaxation occurs as a result of thermally activated processes taking the system from one minimum of the potential energy hypersurface to another. The idea that molecular transport in viscous liquids approaching the glass transition could be best described by invoking motion of the representative state point of the system on the potential energy hypersurface had also been suggested by Gibbs. In his view, the slowing down of relaxations with decreasing temperature is related to a decrease of the number of available minima and to the increasing difficulty for the system to find such minima. The viscous slowing down would thus result from the decrease of some “configurational entropy” that is a measure of the number of accessible minima. These two concepts, potential energy hypersurface, also denoted “energy landscape”, and “configurational entropy”, have gained a renewed interest in recent years, boosted by the analogy with the situation encountered in several generalized spin glass models (see below).
The Adam-Gibbs approach represents a phenomenological attempt to relate the $`\alpha `$-relaxation time of a glass-forming liquid to the “configurational entropy”. In the picture, $`\alpha `$-relaxation takes place by increasingly cooperative rearrangements of groups of molecules. Any such group, called a “cooperatively rearranging region”, is assumed to relax independently of the others. It is a kind of long-lived heterogeneity. Molecular motion is activated and the effective activated free energy is equal to the typical energy barrier per molecule, which is taken as independent of temperature, times the number of molecules that are necessary to form a cooperatively rearranging region whose size permits a transition from one configuration to a new one independently of the environment. This latter number goes as the inverse of the configurational entropy per molecule, $`S_c(T)/N`$, where $`N`$ is the total number of molecules in the sample. Since $`S_c`$ decreases with decreasing temperature, the reasoning leads to an effective activation free energy that grows with decreasing temperature, i. e., to a super-Arrhenius behavior,
$$\tau _\alpha =\tau _0\mathrm{exp}\left(\frac{C}{TS_c(T)}\right),$$
(10)
where $`C`$ is proportional to $`N`$ times the typical energy barrier per molecule. If the configurational entropy vanishes at a nonzero temperature, an assumption somewhat analogous to that in Eq. (6) for the free volume model, but one that is inherent for instance in the Gibbs-di Marzio approximate mean field treatment of a lattice model of linear polymeric chains, then the $`\alpha `$-relaxation times diverge at this same nonzero temperature. In particular if the configurational entropy is identified as the entropy difference between the supercooled liquid and the crystal<sup>††</sup><sup>††</sup>††This phenomenological choice for the entropy of configuration has been criticized by Goldstein who showed for several glass-formers that only half of the entropy difference between the liquid and the crystal comes from strictly “configurational” sources; the remainder comes mostly from changes in vibrational anharmonicity or differences in the number of molecular groups able to engage in local motions. , the Adam-Gibbs theory allows one to correlate the extrapolated divergence of the $`\alpha `$-relaxation times with the Kauzmann paradox (see I-4): the Kauzmann temperature $`T_K`$ would then signal a singularity both in the dynamics and in the thermodynamics of a supercooled liquid<sup>‡‡</sup><sup>‡‡</sup>‡‡A recent careful, but conjectural analysis of dielectric relaxation data suggests that these data are consistent with the existence of a critical point, both structural and dynamical, at the approximate $`T_0`$ specified by the VTF expression in Eq. (1).. Note also that by using a hyperbolic temperature dependence to fit the experimental data on the heat capacity difference between the liquid and the crystal, $`\mathrm{\Delta }C_P(T)=K/T`$, and using this formula to extrapolate the configurational entropy down to the Kauzmann temperature, one converts Eq. (10) to a VFT formula,
$$\tau _\alpha =\tau _0\mathrm{exp}\left(\frac{CT_K}{K(TT_K)}\right),$$
(11)
with the VFT temperature $`T_0`$ equal to the Kauzmann temperature $`T_K`$. When comparing to experimental data, the configurational-entropy based expressions provide a good description at least over a restricted temperature range, but the resulting estimates for the critical number of molecules composing a cooperatively rearranging region is often found to be unphysically small (only a few molecules at $`T_g`$).
Building upon the early suggestion made by Goldstein, and others proposed that the apparent passage with decreasing temperature from flow dynamics described by a mode-coupling approach to activated dynamics such as pictured by the configurational-entropy theory of Adam and Gibbs could be rationalized by considering the physics of exploration of the energy landscape: see Fig. 12. The energy landscape is the potential energy in configurational space. It can be envisaged as an incredibly complex, multi-dimensional ($`3N`$ dimensions for a system of $`N`$ particles) set of hills, valleys, basins, saddle-points, and passage-ways around the hills. At constant volume and constant number of particles, this landscape is independent of temperature. However, the fraction of space that is statistically accessible to the representative state point of the system decreases with decreasing temperature, and the system becomes constrained to deeper and deeper wells. (Recall that below the melting point the deepest energy minima corresponding to the crystalline part of configurational space must be excluded when studying the supercooled liquid.) At low enough temperature, when the representative point of the supercooled liquid is mostly found in fairly deep and narrow wells, it seems reasonable to define a “configurational entropy” that is proportional to the logarithm of the number of minima that are accessible at a given temperature. The liquid configurations corresponding to these accessible minima have been called “inherent structures” and Stillinger and coworkers have devised a gradient-descent mapping procedure to find the inherent structures and study their properties in computer simulations. Interestingly, Stillinger has also shown, with fairly general arguments, that if one is to use the above defined notion of configurational entropy, an “ideal glass transition” of the type commonly associated with the Kauzmann paradox, i. e., one characterized by the vanishing of the configurational entropy at a nonzero temperature, cannot occur for systems of limited molecular weight and short-range interactions.
It may be more fruitful to investigate in place of the potential energy landscape a free-energy landscape. Such a landscape can only be defined if one is able to construct a free-energy functional by a suitable coarse-graining procedure, as can be done for instance in the case of mean-field spin glass models (see below). A free-energy landscape is temperature-dependent, and it is important to note that the “configurational entropy”, also called “complexity”, that one can define from the logarithm of the number of accessible free-energy minima differs from the “configurational entropy” computed from the potential energy landscape. In particular, the behavior of the complexity is not restricted by the Stillinger arguments given above.
The “landscape paradigm” is very appealing in rationalizing many observations on liquids and glasses and, more generally, in establishing a framework to describe qualitatively slow dynamics in complex systems that span a wide range of scientific fields. It has been used to motivate, in addition to the Adam-Gibbs theory and other phenomenological approaches like the soft-potential model, simple stochastic models of transport based on master equations. Nevertheless, it has not so far offered a way for elucidating the physical mechanism that is responsible for the distinctive features of the viscous slowing down of supercooled liquids.
### 4 Analogy with generalized spin glass models
If one takes seriously the observation that the extrapolated temperature dependence of both the viscosity (and $`\alpha `$-relaxation times) and the “configurational” entropy (taken as the difference of entropy between the liquid and the crystal) become divergent or singular at essentially the same temperature $`T_0T_K`$(see Fig. 9), one is naturally led to postulate the existence at this temperature of an underlying thermodynamic transition, usually referred to as the “ideal glass transition”. Looking for analogies with phase transitions in spin glasses is then appealing. However, the kind of dynamic activated scaling that would be required to describe the slowing down of relaxations when approaching the ideal glass transition (see I-1) is not found in the most studied Ising spin glasses. Kirkpatrick, Thirumalai, and Wolynes argued that generalized spin glass models, such as Potts glasses and random p-spin systems, would be better candidates. The random p-spin model, for instance, is defined by the following hamiltonian:
$$H=\underset{i_1<i_2<\mathrm{}<i_p}{}J_{i_1i_2\mathrm{}i_p}\sigma _{i_1}\sigma _{i_2}\mathrm{}\sigma _{i_p},$$
(12)
where the $`\sigma _i`$’s are Ising variables and the couplings $`J_{i_1i_2\mathrm{}i_p}`$’s are quenched independent random variables that can take positive and negative values according to a given probability distribution. The behavior of these systems, at least when solved in the mean-field limit where the interactions between spins have infinite range, bears many similarities with the theoretical description of glass-forming liquids outlined above. Indeed, mean-field Potts glasses (with a number of states strictly larger than 4) and mean-field p-spin models (with $`p3`$) have essentially the following characteristics:
1. At high temperature, the system is in a fully disordered (paramagnetic) state. At a temperature $`T_D`$, there appears an exponentially large number of metastable “glassy” states whose overall contribution to the partition function is equal to that of the paramagnetic minimum. The free energy and all other static equilibrium quantities are fully regular at $`T_D`$, but the dynamics have a singularity of the exact same type as that found in the mode-coupling theory of liquids. At $`T_D`$, the system is trapped in one of the metastable free-energy minima, and ergocity is broken.
2. Below $`T_D`$ , a peculiar situation occurs. The partition function has contributions from, both, the paramagnetic state and the exponentially large number of “glassy” (free-energy) minima, the logarithm of which defines the configurational entropy or complexity. This latter decreases as the temperature is lowered.
3. At a nonzero temperature $`T_s<T_D`$, the configurational entropy vanishes. The system undergoes a bona fide thermodynamic transition to a spin-glass phase. The transition has been termed “random first order” because it is second-order in the usual thermodynamic sense (with, e. g., no latent heat), but shows a discontinuous jump in the order parameter. (Technically, within the replica formalism, it corresponds to a one-step replica symmetry breaking with a discontinuous jump of the Edwards-Anderson order parameter.)
These mean-field systems are the simplest, analytically tractable models found so far that display a high-temperature mode-coupling dynamic singularity, a nontrivial free-energy landscape, and a low-temperature ideal (spin) glass transition with an “entropy crisis”<sup>\**</sup><sup>\**</sup>\**They are also aging phenomena, as discussed in Ref .. Analyzing them sheds light on the mode-coupling approximation, whose validity for fluid systems is otherwise hard to assess. The mode-coupling approximation becomes exact in the mean-field limit, because the barriers separating the metastable minima diverge (in the thermodynamic limit) at and below $`T_D`$ as a result of the assumed infinite range of the interactions. One expects that in a finite-range model, provided the same type of free-energy landscape is still encountered, barriers are large but finite and ergodicity is restored by thermally activated processes. Accordingly, the dynamic transition is smeared out, and the activated relaxation mechanisms that take over must be described in a nonperturbative way, as suggested for instance by Kirkpatrick, Thirumalai, and Wolynes in their dynamic scaling approach based on entropic droplets.
An advantage of an analogy between glass-forming liquids and generalized spin glasses<sup>\*†</sup><sup>\*†</sup>\*†See also the frustrated percolation model. is that the powerful tools that have been developed in the theory of spin-glass models to characterize the order parameter and the properties associated with the existence of a large number of metastable glassy states, among which the replica formalism, can be used mutatis mutandis tostudy liquids and glasses. However, to make the analogy really successful, one must still find a short-range model (even more convincing would be a model without quenched disorder) that actually displays activated dynamic scaling and a random first-order transition and make progress in describing the slow relaxation.
### 5 Intrinsic frustration without randomness
Spin glasses, and related systems like orientational glasses, vortex glasses and vulcanized matter, owe their fascinating behavior to two main ingredients: randomness, namely the presence of an externally imposed quenched disorder, and frustration, which expresses the impossibility of simultaneously minimizing all the interaction terms in the energy function of the system. Liquids and glasses (sometimes called ”structural glasses” to stress the difference with spin glasses) have no quenched randomness, but frustration has been suggested as a key feature to explain the phenomena associated with glass formation. Frustration in this context is attributed to a competition between a short-range tendency for the extension of a locally preferred order and global constraints that preclude the periodic tiling of the whole space with the local structure.
The best studied example of such an intrinsic frustration concerns single-component systems of spherical particles interacting with simple pair potentials. What is usually called ”geometric” or ”topological” frustration can be more easily understood by comparing the situations encountered in $`2`$ and $`3`$ dimensions (see Fig. 13). In $`2`$ dimensions, the arrangement of disks that is locally preferred, in the sense that it maximizes the density and minimizes the energy, is a hexagon of 6 disks around a central one, and this hexagonal structure can be extended to the whole space to form a triangular lattice. In $`3`$ dimensions, as was shown long ago by Frank, the locally preferred cluster of spheres is an icosahedron; however, the $`5`$-fold rotational symmetry characteristic of icosahedral order is incompatible with translational symmetry, and formation of a periodic icosahedral crystal is forbidden. Geometric or topological frustration is thus absent in the $`2`$-dimensional case but present in the $`3`$-dimensional case. A consequence of this, for instance, is that crystallization is continuous, or weakly first-order, in $`2`$ dimensions (with some subtleties related to ordering in $`2`$ dimensions ) whereas it is strongly first-order in $`3`$ dimensions and accompanied by the breaking of the local icosahedral structure to make the face-centered-cubic or hexagonal-close-packed order that allows to tile space periodically. (In contrast, aligned cubes in 3 dimensions have no frustration and undergo a continuous freezing transition to a crystalline state. ). The geometric frustration that affects spheres in $`3`$-dimensional Euclidean space can be relieved in curved space with a specially tuned curvature; the creation of topological defects (disclination lines) can then be viewed as the result of forcing the ideal icosahedral ordering into ”flat” space. This picture of sphere packing disrupted by frustration has been further developed in models for simple atomic systems and metallic glasses, and the slowing down of relaxations has been tentatively attributed to the topological constraints that hinder the kinetics of the entangled defect lines; however, the treatment remains only qualitative and incomplete.
A significant difficulty in applying the concept of geometric or topological frustration to supercooled liquids is that real fragile glass-formers are in general either mixtures or single-component systems of nonspherical molecules with a variety of shapes, all of which obscures the detailed mechanisms and constraints that are responsible for the frustration. Attempts have been made to get around this problem by proposing a more coarse-grained description of frustration<sup>\*‡</sup><sup>\*‡</sup>\*‡In addition, several “toy models”possessing frustration, but no quenched disorder have been studied by computer simulation: see for instance Ref...
In Stillinger’s ”tear and repair” mechanism for relaxation and shear flow and in the more recently introduced ”frustration-limited domain theory”, frustration is described as the source of a strain free energy that opposes the spatial extension of the locally preferred structure and grows super-extensively with system size. It results in the breaking up of the liquid into domains, whose size and growth with decreasing temperature are limited by frustration, the weaker the frustration the larger the domains. The super-Arrhenius temperature dependence of the viscosity and $`\alpha `$-relaxation times and the heterogeneous nature of the dynamics are attributed to these domains (see also Ref.). Progress has been made along these lines, by making use of a scaling approach based on the concept of avoided critical behaviorThis approach differs from both those based upon spin-glass analogies and those in which the slow kinetics are attributed to frustration-induced entangled defect lines in that these others scale about a low-temperature characteristic point signifying ultimate slowing down, whereas in the frustration-limited domain theory the scaling is carried out about a high-temperature characteristic point signifying the initiation of anomalously slow dynamics. For the same reason, it also differs from the domain (or cluster) picture that has been proposed on the basis of an analogy between a supercooled liquid approaching the glass transition and a mean-field model with purely repulsive interactions near its spinodal.. However, the putative order variable characterizing the locally preferred structure of the liquid has not yet been properly identified, and, as in the case of the generalized spin-glass models discussed above, one must still give convincing evidence that the $`3`$-dimensional statistical-mechanical frustrated models that have been suggested as minimal theoretical descriptions do show the expected activated dynamics.
## V CONCLUSION
The viscous slowing down of supercooled liquids that leads to glass formation can be considered as a classical and thoroughly studied example of a ”jamming process”. In this review, we have stressed the distinctive features characterizing the phenomenon: strong, super-Arrhenius temperature dependence of the viscosity and the $`\alpha `$-relaxation times, nonexponential and heterogeneous character of the $`\alpha `$ relaxation, absence of marked changes in the structural (static) quantities, rapid decrease of the liquid entropy relative to that of the crystal, appearance of a sequence of steps (or regimes) in the relaxation functions. These features are common to most glass-forming liquids (with the exception of systems forming $`2`$\- and $`3`$-dimensional networks of strong intermolecular bonds). We have also discussed the main theoretical approaches that have been proposed to describe the origin and the nature of the viscous slowing down and of the glass transition. We have emphasized the concepts, such as free volume, dynamic freezing and mode-coupling approximations, configurational entropy and (free) energy landscape, and frustration, that could be useful in other areas of physics where jamming processes are encountered.
|
warning/0003/cs0003079.html
|
ar5iv
|
text
|
# Differential Invariants under Gamma Correction
## 1 Introduction
Invariants are a popular concept in object recognition and image retrieval . They aim to provide descriptions that remain constant under certain geometric or radiometric transformations of the scene, thereby reducing the search space. They can be classified into global invariants, typically based either on a set of key points or on moments, and local invariants, typically based on derivatives of the image function which is assumed to be continuous and differentiable.
The geometric transformations of interest often include translation, rotation, and scaling, summarily referred to as similarity transformations. In a previous paper , building on work done by Schmid and Mohr , we have proposed differential invariants for those similarity transformations, plus linear brightness change. Here, we are looking at a non-linear brightness change known as gamma correction.
Gamma correction is a non-linear quantization of the brightness measurements performed by many cameras during the image formation process.<sup>1</sup><sup>1</sup>1 Historically, the parameter gamma was introduced to describe the non-linearity of photographic film. Today, its main use is to improve the output of cathode ray tube based monitors, but the gamma correction in display devices is of no concern to us here. The idea is to achieve better perceptual results by maintaining an approximately constant ratio between adjacent brightness levels, placing the quantization levels apart by the just noticeable difference. Incidentally, this non-linear quantization also precompensates for the non-linear mapping from voltage to brightness in electronic display devices .
Gamma correction can be expressed by the equation
$$I_\gamma =pI^\gamma $$
(1)
where $`I`$ is the input intensity, $`I_\gamma `$ is the output intensity, and $`p`$ is a normalization factor which is determined by the value of $`\gamma `$. For output devices, the NTSC standard specifies $`\gamma =2.22`$. For input devices like cameras, the parameter value is just inversed, resulting in a typical value of $`\gamma =1/2.22=0.45`$. The camera we used, the Sony 3 CCD color camera DXC 950, exhibited $`\gamma 0.6`$.<sup>2</sup><sup>2</sup>2 Martin reports the settings of $`\gamma =0.45,0.50,0.60`$ for the Kodak Megaplus XRC camera Fig. 1 shows the intensity mapping of 8-bit data for different values of $`\gamma `$.
It turns out that an invariant under gamma correction can be designed from first and second order derivatives. Additional invariance under scaling requires third order derivatives. Derivatives are by nature translationally invariant. Rotational invariance in 2-d is achieved by using rotationally symmetric operators.
## 2 The Invariants
The key idea for the design of the proposed invariants is to form suitable ratios of the derivatives of the image function such that the parameters describing the transformation of interest will cancel out. This idea has been used in to achieve invariance under linear brightness changes, and it can be adjusted to the context of gamma correction by – at least conceptually – considering the logarithm of the image function. For simplicity, we begin with 1-d image functions.
### 2.1 Invariance under Gamma Correction
Let $`f(x)`$ be the image function, i.e. the original signal, assumed to be continuous and differentiable, and $`f_\gamma (x)=pf(x)^\gamma `$ the corresponding gamma corrected function. Note that $`f(x)`$ is a special case of $`f_\gamma (x)`$ where $`\gamma =p=1`$. Taking the logarithm yields
$$\stackrel{~}{f}_\gamma (x)=\mathrm{ln}(pf(x)^\gamma )=\mathrm{ln}p+\gamma \mathrm{ln}f(x)$$
(2)
with the derivatives $`\stackrel{~}{f}_\gamma ^{}(x)=\gamma f^{}(x)/f(x)`$, and $`\stackrel{~}{f}_\gamma ^{\prime \prime }(x)=\gamma (f(x)f^{\prime \prime }(x)f^{}(x)^2)/f(x)^2`$. We can now define the invariant $`\mathrm{\Theta }_{12\gamma }`$ under gamma correction to be
| $`\mathrm{\Theta }_{12\gamma }(f(x))`$ | = $`\frac{\stackrel{~}{f}_\gamma ^{}\left(x\right)}{\stackrel{~}{f}_\gamma ^{\prime \prime }\left(x\right)}`$ |
| --- | --- |
| | = $`\frac{\gamma \frac{f^{}\left(x\right)}{f\left(x\right)}}{\gamma \frac{f\left(x\right)f^{\prime \prime }\left(x\right)f^{}\left(x\right)^2}{f\left(x\right)^2}}`$ |
| | = $`\frac{f\left(x\right)f^{}\left(x\right)}{f\left(x\right)f^{\prime \prime }\left(x\right)f^{}\left(x\right)^2}`$ |
(3)
The factor $`p`$ has been eliminated by taking derivatives, and $`\gamma `$ has canceled out. Furthermore, $`\mathrm{\Theta }_{12\gamma }`$ turns out to be completely specified in terms of the original image function and its derivatives, i.e. the logarithm actually doesn’t have to be computed. The notation $`\mathrm{\Theta }_{12\gamma }(f(x))`$ indicates that the invariant depends on the underlying image function $`f(x)`$ and location $`x`$ – the invariance holds under gamma correction, not under spatial changes of the image function.
A shortcoming of $`\mathrm{\Theta }_{12\gamma }`$ is that it is undefined where the denominator is zero. Therefore, we modify $`\mathrm{\Theta }_{12\gamma }`$ to be continuous everywhere:
| | 0 | if $`ff^{}=0ff^{\prime \prime }f_{}^{}{}_{}{}^{2}=0`$ |
| --- | --- | --- |
| $`\mathrm{\Theta }_{m12\gamma }=`$ | $`\frac{ff^{}}{ff^{\prime \prime }f_{}^{}{}_{}{}^{2}}`$ | if $`|ff^{}|<|ff^{\prime \prime }f_{}^{}{}_{}{}^{2}|`$ |
| | $`\frac{ff^{\prime \prime }f_{}^{}{}_{}{}^{2}}{ff^{}}`$ | else |
(4)
where, for notational convenience, we have dropped the variable $`x`$. The modification entails $`1\mathrm{\Theta }_{m12\gamma }1`$. Note that the modification is just a heuristic to deal with poles. If all derivatives are zero because the image function is constant, then differentials are certainly not the best way to represent the function.
### 2.2 Invariance under Gamma Correction and Scaling
If scaling is a transformation that has to be considered, then another parameter $`\alpha `$ describing the change of size has to be introduced. That is, scaling is modeled here as variable substitution : the scaled version of $`f(x)`$ is $`g(\alpha x)=g(u)`$. So we are looking at the function
$$\stackrel{~}{f}_\gamma (x)=\mathrm{ln}(pf(x)^\gamma )=\mathrm{ln}p+\gamma \mathrm{ln}g(\alpha x)=\stackrel{~}{g}_\gamma (u)$$
where the derivatives with respect to $`x`$ are $`\stackrel{~}{g}_\gamma ^{}(u)=\gamma \alpha g^{}(u)/g(u)`$, $`\stackrel{~}{g}_\gamma ^{\prime \prime }(u)=\gamma \alpha ^2(g(u)g^{\prime \prime }(u)g^{}(u)^2)/g(u)^2`$, and $`\stackrel{~}{g}_\gamma ^{\prime \prime \prime }(u)=\gamma \alpha ^3\left(g^{\prime \prime \prime }(u)/g(u)3g^{}(u)g^{\prime \prime }(u)/g(u)^2+2g^{}(u)^3/g(u)^3\right)`$. Now the invariant $`\mathrm{\Theta }_{123\gamma }(g(u))`$ is obtained by defining a suitable ratio of the derivatives such that both $`\gamma `$ and $`\alpha `$ cancel out:
| $`\mathrm{\Theta }_{123\gamma }(g(u))`$ | = $`\frac{\stackrel{~}{g}_\gamma ^{}\left(u\right)\stackrel{~}{g}_\gamma ^{\prime \prime \prime }\left(u\right)}{\stackrel{~}{g}_\gamma ^{\prime \prime }\left(u\right)^2}`$ |
| --- | --- |
| | = $`\frac{g^2g^{}g^{\prime \prime \prime }3gg_{}^{}{}_{}{}^{2}g^{\prime \prime }+2g_{}^{}{}_{}{}^{4}}{g^2g_{}^{\prime \prime }{}_{}{}^{2}2gg_{}^{}{}_{}{}^{2}g^{\prime \prime }+g_{}^{}{}_{}{}^{4}}`$ |
(5)
Analogously to eq. (4), we can define a modified invariant
| | 0 | if cond1 |
| --- | --- | --- |
| $`\mathrm{\Theta }_{m123\gamma }=`$ | $`\frac{g^2g^{}g^{\prime \prime \prime }3gg_{}^{}{}_{}{}^{2}g^{\prime \prime }+2g_{}^{}{}_{}{}^{4}}{g^2g_{}^{\prime \prime }{}_{}{}^{2}2gg_{}^{}{}_{}{}^{2}g^{\prime \prime }+g_{}^{}{}_{}{}^{4}}`$ | if cond2 |
| | $`\frac{g^2g_{}^{\prime \prime }{}_{}{}^{2}2gg_{}^{}{}_{}{}^{2}g^{\prime \prime }+g_{}^{}{}_{}{}^{4}}{g^2g^{}g^{\prime \prime \prime }3gg_{}^{}{}_{}{}^{2}g^{\prime \prime }+2g_{}^{}{}_{}{}^{4}}`$ | else |
(6)
where condition cond1 is $`g^2g^{}g^{\prime \prime \prime }3gg_{}^{}{}_{}{}^{2}g^{\prime \prime }+2g_{}^{}{}_{}{}^{4}=0`$ $``$ $`g^2g_{}^{\prime \prime }{}_{}{}^{2}2gg_{}^{}{}_{}{}^{2}g^{\prime \prime }+g_{}^{}{}_{}{}^{4}=0`$, and condition cond2 is $`|g^2g^{}g^{\prime \prime \prime }3gg_{}^{}{}_{}{}^{2}g^{\prime \prime }+2g_{}^{}{}_{}{}^{4}|`$ $`<`$ $`|g^2g_{}^{\prime \prime }{}_{}{}^{2}2gg_{}^{}{}_{}{}^{2}g^{\prime \prime }+g_{}^{}{}_{}{}^{4}|`$. Again, this modification entails $`1\mathrm{\Theta }_{m123\gamma }1`$.
### 2.3 An Analytical Example
It is a straightforward albeit cumbersome exercise to verify the invariants from eqs. (3) and (5) with an analytical, differentiable function. As an arbitrary example, we choose
$$f(x)=3x\mathrm{sin}(2\pi x)+30$$
The first three derivatives are $`f^{}(x)=3\mathrm{sin}(2\pi x)+6\pi x\mathrm{cos}(2\pi x)`$, $`f^{\prime \prime }(x)=12\pi \mathrm{cos}(2\pi x)12\pi ^2x\mathrm{sin}(2\pi x)`$, and $`f^{\prime \prime \prime }(x)=36\pi ^2\mathrm{sin}(2\pi x)24\pi ^3x\mathrm{cos}(2\pi x)`$. Then, according to eq. (3), $`\mathrm{\Theta }_{12\gamma }(f(x))=(3x\mathrm{sin}(2\pi x)+30)(3x\mathrm{sin}(2\pi x)+6\pi x\mathrm{cos}(2\pi x))/\left((3x\mathrm{sin}(2\pi x)+30)(12\pi \mathrm{cos}(2\pi x)12\pi ^2x\mathrm{sin}(2\pi x))(3\mathrm{sin}(2\pi x)+6\pi x\mathrm{cos}(2\pi x))^2\right)`$.
If we now replace $`f(x)`$ with a gamma corrected version, say $`f_{0.45}(x)=255^{10.45}3x\mathrm{sin}(2\pi x)+30)^{0.45}`$, the first derivative becomes $`f_{0.45}^{}(x)=255^{0.55}0.45(3\mathrm{sin}(2\pi x)+30)^{0.55}(3\mathrm{sin}(2\pi x)+6\pi x\mathrm{cos}(2\pi x))`$, the second derivative is $`f_{0.45}^{\prime \prime }(x)=255^{0.55}0.450.55(3\mathrm{sin}(2\pi x)+30)^{1.55}(3\mathrm{sin}(2\pi x)+6\pi x\mathrm{cos}(2\pi x))^2+255^{0.55}0.45(3x\mathrm{sin}(2\pi x)+30)^{0.55}(12\pi \mathrm{cos}(2\pi x)12\pi ^2x\mathrm{sin}(2\pi x))`$, and the third is $`f_{0.45}^{\prime \prime \prime }(x)=255^{0.55}0.45(3\mathrm{sin}(2\pi x)+30)^{0.55}\left(1.550.55(3\mathrm{sin}(2\pi x)+30)^2(3\mathrm{sin}(2\pi x)+6\pi x\mathrm{cos}(2\pi x))^3+(3)0.55(3\mathrm{sin}(2\pi x)+30)^1(3\mathrm{sin}(2\pi x)+6\pi x\mathrm{cos}(2\pi x))(12\pi \mathrm{cos}(2\pi x)12\pi ^2x\mathrm{sin}(2\pi x))+(36\pi ^2\mathrm{sin}(2\pi x)24\pi ^3x\mathrm{cos}(2\pi x))\right)`$. If we plug these derivatives into eq. (3), we obtain an expression for $`\mathrm{\Theta }_{12\gamma }(f_{0.45}(x))`$ which is identical to the one for $`\mathrm{\Theta }_{12\gamma }(f(x))`$ above. The algebraically inclined reader is encouraged to verify the invariant $`\mathrm{\Theta }_{123\gamma }`$ for the same function.
Fig. 2 shows the example function and its gamma corrected counterpart, together with their derivatives and the two modified invariants. As expected, the graphs of the invariants are the same on the right as on the left. Note that the invariants define a many-to-one mapping. That is, the mapping is not information preserving, and it is not possible to reconstruct the original image from its invariant representation.
### 2.4 Extension to 2-d
If $`\mathrm{\Theta }_{m12\gamma }`$ or $`\mathrm{\Theta }_{m123\gamma }`$ are to be computed on images, then eqs. (3) to (6) have to be generalized to two dimensions. This is to be done in a rotationally invariant way in order to achieve invariance under similarity transformations. The standard way is to use rotationally symmetric operators. For the first derivative, we have the well known gradient magnitude, defined as
$$(x,y)=\sqrt{I_x^2+I_y^2}:=I^{}$$
(7)
where $`I(x,y)`$ is the 2-d image function, and $`I_x`$, $`I_y`$ are partial derivatives along the x-axis and the y-axis. For the second order derivative, we can use the linear Laplacian
$$^2(x,y)=I_{xx}+I_{yy}:=I^{\prime \prime }$$
(8)
Horn also presents an alternative second order derivative operator, the quadratic variation<sup>3</sup><sup>3</sup>3Actually, unlike Horn, we have taken the square root.
$$\mathrm{QV}(x,y)=\sqrt{I_{xx}^2+2I_{xy}^2+I_{yy}^2}$$
(9)
Since the QV is not a linear operator and more expensive to compute, we use the Laplacian for our implementation. For the third order derivative, we can define, in close analogy with the quadratic variation, a cubic variation as
$$\mathrm{CV}(x,y)=\sqrt{I_{xxx}^2+3I_{xxy}^2+3I_{xyy}^2+I_{yyy}^2}:=I^{\prime \prime \prime }$$
(10)
The invariants from eqs. (3) to (6) remain valid in 2-d if we replace $`f^{}`$ with $`I^{}`$, $`f^{\prime \prime }`$ with $`I^{\prime \prime }`$, and $`f^{\prime \prime \prime }`$ with $`I^{\prime \prime \prime }`$. This can be verified by going through the same argument as for the 1-d functions. Recall that the critical observation in eq. (3) was that $`\gamma `$ cancels out, which is the case when all derivatives return a factor $`\gamma `$. But such is also the case with the rotationally symmetric operators mentioned above. For example, if we apply the gradient magnitude operator to $`\stackrel{~}{I}(x,y)`$, i.e. to the logarithm of a gamma corrected image function, we obtain
$$=\sqrt{\stackrel{~}{I}_x^2+\stackrel{~}{I}_y^2}=\sqrt{\left(\gamma \frac{I_x}{I}\right)^2+\left(\gamma \frac{I_y}{I}\right)^2}=\gamma \frac{\sqrt{I_x^2+I_y^2}}{I}$$
returning a factor $`\gamma `$, and analogously for $`^2`$, QV, and CV. A similar argument holds for eq. (5) where we have to show, in addition, that the first derivative returns a factor $`\alpha `$, the second derivative returns a factor $`\alpha ^2`$, and the third derivative returns a factor $`\alpha ^3`$, which is the case for our 2-d operators.
### 2.5 Differential Operators
While the derivatives of continuous, differentiable functions are uniquely defined, there are many ways to implement derivatives for sampled functions. We follow Schmid and Mohr , ter Haar Romeny , and many other researchers in employing the derivatives of the Gaussian function as filters to compute the derivatives of a sampled image function via convolution. This way, derivation is combined with smoothing. The 2-d zero mean Gaussian is defined as
$$G=\frac{1}{2\pi \sigma ^2}e^{\frac{x^2+y^2}{2\sigma ^2}}$$
(11)
The partial derivatives up to third order are $`G_x=x/\sigma ^2G`$, $`G_y=y/\sigma ^2G`$, $`G_{xx}=(x^2\sigma ^2)/\sigma ^4G`$, $`G_{xy}=xy/\sigma ^4G`$, $`G_{yy}=(y^2\sigma ^2)/\sigma ^4G`$, $`G_{xxx}=(3\sigma ^2xx^3)/\sigma ^6G`$, $`G_{xxy}=(\sigma ^2yx^2y)/\sigma ^6G`$, $`G_{xyy}=(\sigma ^2xxy^2)/\sigma ^6G`$, $`G_{yyy}=(3\sigma ^2yy^3)/\sigma ^6G`$. They are shown in fig. 3. We used the parameter setting $`\sigma =1.0`$ and kernel size $`7\times 7`$. With these kernels, eq. (3), for example, is implemented as
$$\mathrm{\Theta }_{12\gamma }=\frac{I\sqrt{\left(IG_x\right)^2+\left(IG_y\right)^2}}{I\left(IG_{xx}+IG_{yy}\right)\left(\left(IG_x\right)^2+\left(IG_y\right)^2\right)}$$
at each pixel $`(x,y)`$, where $``$ denotes convolution.
## 3 Experimental Data and Results
We evaluate the invariant $`\mathrm{\Theta }_{m12\gamma }`$ from eq. (4) in two different ways. First, we measure how much the invariant computed on an image without gamma correction is different from the invariant computed on the same image but with gamma correction. Theoretical, this difference should be zero, but in practice, it is not. Second, we compare template matching accuracy on intensity images, again without and with gamma correction, to the accuracy achievable if instead the invariant representation is used. We also examine whether the results can be improved by prefiltering.
### 3.1 Absolute and Relative Errors
A straightforward error measure is the absolute error,
$$\mathrm{\Delta }_{GC}(i,j)=|\mathrm{\Theta }_{GC}(i,j)\mathrm{\Theta }_{0GC}(i,j)|$$
(12)
where ”0GC” refers to the image without gamma correction, and GC stands for either ”SGC” if the gamma correction is done synthetically via eq. (1), or for ”CGC” if the gamma correction is done via the camera hardware. Like the invariant itself, the absolute error is computed at each pixel location $`(i,j)`$ of the image, except for the image boundaries where the derivatives and therefore the invariants cannot be computed reliably.
Fig. 4 shows an example image. The SGC image has been computed from the 0GC image, with $`\gamma =0.6`$. Note that the gamma correction is done after the quantization of the 0GC image, since we don’t have access to the 0GC image before quantization.
Fig. 5 shows the invariant representations of the image data from fig. 4 and the corresponding absolute errors. Since $`1\mathrm{\Theta }_{m12\gamma }1`$, we have $`0\mathrm{\Delta }_{GC}2`$. The dark points in fig. 5, (c) and (e), indicate areas of large errors. We observe two error sources:
* The invariant cannot be computed robustly in homogeneous regions. This is hardly surprising, given that it is based on differentials which are by definition only sensitive to spatial changes of the signal.
* There are outliers even in the SGC invariant representation, at points of very high contrast edges. They are a byproduct of the inherent smoothing when the derivatives are computed with differentials of the Gaussian. Note that the latter put a ceiling on the maximum gradient magnitude that is computable on 8-bit images.
In addition to computing the absolute error, we can also compute the relative error, in percent, as
$$\delta _{CGC}(i,j)=100\mathrm{\Delta }_{CGC}(i,j)/\mathrm{\Theta }_{0GC}(i,j)$$
(13)
Then we can define the set $`\mathrm{RP}_ϵ`$ of reliable points, relative to some error threshold $`ϵ`$, as
$$\mathrm{RP}_ϵ=\{(i,j)|\delta (i,j)ϵ\}$$
(14)
and $`\mathrm{PRP}_ϵ`$, the percentage of reliable points, as
$$\mathrm{PRP}_ϵ=100|\mathrm{RP}_ϵ|/n$$
(15)
where $`n`$ is the number of valid, i.e. non-boundary, pixels in the image. Fig. 6 shows, in the first row, the reliable points for three different values of the threshold $`ϵ`$. The second row shows the sets of reliable points for the same thresholds if we gently prefilter the 0GC and CGC images. The corresponding data for the ten test images from fig. 11 is summarized in table 1.
Derivatives are known to be sensitive to noise. Noise can be reduced by smoothing the original data before the invariants are computed. On the other hand, derivatives should be computed as locally as possible. With these conflicting goals to be considered, we experiment with gentle prefiltering, using a Gaussian filter of size $`\sigma _{pre}`$=1.0. The size of the Gaussian to compute the invariant $`\mathrm{\Theta }_{m12\gamma }`$ is set to $`\sigma _{der}`$=1.0. Note that $`\sigma _{pre}`$ and $`\sigma _{der}`$ can not be combined into just one Gaussian because of the non-linearity of the invariant.
With respect to the set of reliable points, we observe that after prefiltering, roughly half the points, on average, have a relative error of less than 20%. Gentle prefiltering consistently reduces both absolute and relative errors, but strong prefiltering does not.
### 3.2 Template Matching
Template matching is a frequently employed technique in computer vision. Here, we will examine how gamma correction affects the spatial accuracy of template matching, and whether that accuracy can be improved by using the invariant $`\mathrm{\Theta }_{m12\gamma }`$.
An overview of the testbed scenario is given in fig. 7. A small template of size $`6\times 8`$, representing the search pattern, is taken from a 0GC intensity image, i.e. without gamma correction. This query template is then correlated with the corresponding CGC intensity image, i.e. the same scene but with gamma correction switched on. If the correlation maximum occurs at exactly the location where the 0GC query template has been cut out, we call this a correct maximum correlation position, or CMCP.
The correlation function $`s(x,y)`$ employed here is based on a normalized mean squared difference $`c(x,y)`$ :
$`s`$ $`=\mathrm{max}(0,1c)`$
$`c`$ $`={\displaystyle \frac{_{i,j}((I(x+i,y+j)\overline{I})(T(i,j)\overline{T}))^2}{\sqrt{_{i,j}(I(x+i,y+j)\overline{I})^2_{i,j}(T(i,j)\overline{T})^2}}}`$
where $`I`$ is an image, $`T`$ is a template positioned at $`(x,y)`$, $`\overline{I}(x,y)`$ is the mean of the subimage of $`I`$ at $`(x,y)`$ of the same size as $`T`$, $`\overline{T}`$ is the mean of the template, and $`0s1`$. The template location problem then is to perform this correlation for the whole image and to determine whether the position of the correlation maximum occurs precisely at $`(x,y)`$.
Fig. 8 demonstrates the template location problem, on the left for an intensity image, and on the right for its invariant representation. The black box marks the position of the original template at (40,15), and the white box marks the position of the matched template, which is incorrectly located at (50,64) in the intensity image. On the right, the matched template (white) has overwritten the original template (black) at the same, correctly identified position. Fig. 9 visualizes the correlation function over the whole image. The white areas are regions of high correlation.
The example from figs. 8 and 9 deals with only one arbitrarily selected template. In order to systematically analyze the template location problem, we repeat the correlation process for all possible template locations. Then we can define the correlation accuracy CA as the percentage of correctly located templates,
$$\mathrm{CA}_{tn\times tm}=100|\mathrm{CMCP}_{tn\times tm}|/n$$
(16)
where $`tn\times tm`$ is the size of the template, $`\mathrm{CMCP}_{tn\times tm}`$ is the set of correct maximum correlation positions, and $`n`$, again, is the number of valid pixels. We compute the correlation accuracy both for unfiltered images and for gently prefiltered images, with $`\sigma _{pre}=1.0`$. Fig. 10 shows the binary correlation accuracy matrices for our example image. The CMCP set is shown in white, its complement and the boundaries in black. We observe a higher correlation accuracy for the invariant representation, which is improved by the prefiltering.
We have computed the correlation accuracy for all the images given in fig. 11. The results are shown in table 2 and visualized in fig. 12. We observe the following:
* The correlation accuracy CA is higher on the invariant representation than on the intensity images.
* The correlation accuracy is higher on the invariant representation with gentle prefiltering, $`\sigma _{pre}=1.0`$, than without prefiltering. We also observed a decrease in correlation accuracy if we increase the prefiltering well beyond $`\sigma _{pre}=1.0`$. By contrast, prefiltering seems to be always detrimental to the intensity images CA.
* The correlation accuracy shows a wide variation, roughly in the range 30%$`\mathrm{}`$90% for the unfiltered intensity images and 50%$`\mathrm{}`$100% for prefiltered invariant representations. Similarly, the gain in correlation accuracy ranges from close to zero up to 45%. For our test images, it turns out that the invariant representation is always superior, but that doesn’t necessarily have to be the case.
* The medians and means of the CAs over all test images confirm the gain in correlation accuracy for the invariant representation.
* The larger the template size, the higher the correlation accuracy, independent of the representation. A larger template size means more structure, and more discriminatory power.
## 4 Conclusion
We have proposed novel invariants that combine invariance under gamma correction with invariance under geometric transformations. In a general sense, the invariants can be seen as trading off derivatives for a power law parameter, which makes them interesting for applications beyond image processing. The error analysis of our implementation on real images has shown that, for sampled data, the invariants cannot be computed robustly everywhere. Nevertheless, the template matching application scenario has demonstrated that a performance gain is achievable by using the proposed invariant.
## 5 Acknowledgements
Bob Woodham suggested to the author to look into invariance under gamma correction. His meticulous comments on this work were much appreciated. Jochen Lang helped with the acquisition of image data through the ACME facility .
|
warning/0003/hep-ex0003008.html
|
ar5iv
|
text
|
# 1 The Primakoff 𝛾-pion Hybrid production process and kinematic variables (4-momenta): P1, P1′ = for initial/final pion, P2, P2′ = for initial/final target, k = for initial 𝛾, k′ = for final 𝜂.
Contributions to the International Workshop on Hadron Physics
with High Energy Muon and Hadron Beams at Fixed Target Experiments,
Workshop Chairmen S. Paul and F. Bradamante, Technical University Munich, Oct. 1999,
Tel Aviv U. Preprint TAUP-2605-99
Hybrid Meson Structure at COMPASS
Murray Moinester<sup>a</sup>, Suh Urk Chung<sup>b</sup>
a–School of Physics and Astronomy, R. and B. Sackler Faculty of Exact Sciences,
Tel Aviv University, 69978 Tel Aviv, Israel, e-mail: murraym$`\mathrm{@}`$tauphy.tau.ac.il
b–Department of Physics, Brookhaven National Laboratory,
Upton, NY 11973, USA, e-mail: suchung@qgs.phy.bnl.gov
Abstract
Objectives and Significance:
We describe a pion physics program attainable with the CERN COMPASS spectrometer, involving tracking detectors and an electromagnetic calorimeter. COMPASS can realize state-of-the-art pion beam hybrid meson and meson radiative transition studies. We review here the physics motivation for this program. We describe the beam, detector, trigger requirements, and hardware/software requirements for this program. The triggers for all this hybrid meson physics can be implemented for simultaneous data taking.
We will investigate hybrid meson production via pion-photon Primakoff and pion-Pomeron diffractive interactions. We will determine new properties of quark-antiquark-gluon hybrid mesons, using unique production methods, to improve our understanding of these exotic mesons.
Methodology:
The CERN COMPASS experiment uses 100-280 GeV beams ($`\mu `$, $`\pi `$), and magnetic spectrometers and calorimeters, to measure the complete kinematics of pion-photon and pion-Pomeron reactions. The COMPASS experiment is currently under construction, and scheduled to begin data runs in 2001. We carry out simulation studies to optimize the beam, detector, trigger, and hardware/software for achieving high statistics data with low systematic uncertainties in the hybrid meson component of this program. We will improve previous Primakoff Hybrid studies by three orders of magnitude. We implement special detectors and triggers for hybrid meson production reactions. We propose to prepare for these COMPASS pion beam hybrid studies by setting up with muon beam tests.
1. Description of Subject:
The COMPASS physics programs include studies of pion-photon Primakoff and pion-Pomeron diffractive interactions using 100-280 GeV/$`c`$ negative pion beams in dedicated data runs. Hybrid mesons can be studied in this way, and can provide significant tests of QCD predictions. COMPASS Primakoff planning studies were described in recent workshop proceedings , and at COMPASS collaboration meetings.
1.1 Hybrid Mesons
The hybrid ($`q\overline{q}g`$) mesons, along with glueballs ($`gg`$) are some of the most interesting consequences of the non-Abelian nature of QCD. Detection of these exotic states is a long-standing experimental puzzle. The most popular approach for hybrid searches is to look for the ‘oddballs’—mesons with quantum numbers not allowed for ordinary $`q\overline{q}`$ states. For Primakoff/diffractive production, the outgoing mesons preserve the charge of the incoming beam, i.e. $`I=1`$ for the resonances under study. Then, the ‘oddball’ mesons for $`J2`$ come in the following variety:
| $`I^G(J^{PC})`$ | ‘Oddball’ |
| --- | --- |
| $`1^+(0^{})`$ | $`\rho _0`$ |
| $`1^+(0^+)`$ | $`b_0`$ |
| $`1^{}(1^+)`$ | $`\pi _1`$ |
| $`1^+(2^+)`$ | $`b_2`$ |
Barnes and Isgur first discussed hybrid meson properties, and more recently in the flux-tube model . In the flux-tube model, the mass of the lightest gluonic hybrid is predicted be around 1.9 GeV, with the quantum numbers of $`J^{PC}=1^+`$. Close and Page predict that such a gluonic hybrid should decay into the following channels:
| $`b_1\pi `$ | $`f_1\pi `$ | $`\rho \pi `$ | $`\eta \pi `$ | $`\eta ^{}\pi `$ |
| --- | --- | --- | --- | --- |
| 170 | 60 | $`520`$ | $`010`$ | $`010`$ |
where the numbers refer to the partial widths in MeV. According to them, its total width must be larger than 235-270 MeV, since the $`s+\overline{s}`$ decay modes were not included. Recent updates on hybrid meson structure are given in Refs.
From more than a decade of experimental efforts at IHEP , CERN , KEK , and BNL , several hybrid candidates have been identified. More recently, new information came from the BNL E852 experiment , which studied the $`\pi ^{}p`$ interaction at 18 GeV/$`c`$. They reported two $`J^{PC}=1^+`$ resonant signals at masses of 1.4 and 1.6 GeV in $`\eta \pi ^{}`$ and $`\eta \pi ^0`$ systems, as well as in $`\pi ^+\pi ^{}\pi ^{}`$, $`\pi ^{}\pi ^0\pi ^0`$, $`\eta ^{}\pi ^{}`$ and $`f_1(1285)\pi ^{}`$. Also, a VES group has published analyses of $`\eta \pi ^{}`$, $`\eta ^{}\pi ^{}`$, $`f_1(1285)\pi ^{}`$, $`b_1(1235)\pi ^{}`$ and $`\rho \pi ^{}`$ systems produced in $`\pi ^{}\mathrm{Be}`$ interactions at 37 GeV/$`c`$. VES sees the $`J^{PC}=1^+`$ wave clearly in all channels, and they report an indication of a resonance at 1.6 GeV. It is striking that the VES phase motion of the $`1^+`$ wave in $`\eta \pi ^{}`$ shows a rise at 1.4 GeV, identical to that of the BNL E852 data. The most recent information on the 1.4-GeV state comes from two analyses by the Crystal Barrel collaboration on the $`\overline{p}p`$ and $`\overline{p}n`$ annihilations at rest into $`\pi \pi \eta `$ ,. Their observed masses and widths are consistent with those of BNL E852. VES reports that the ratio of $`\eta ^{}\pi `$ to $`\eta \pi `$ $`P`$-waves at 1.4 GeV is low, while that at 1.6 GeV is high. This is considered as evidence that the hybrid nature of the exotic wave at 1.6 GeV is gluonic; i.e., its constituents are $`q+\overline{q}+\mathrm{gluon}`$, where $`q`$ stands for light nonstrange quarks.
The partial-wave analysis (PWA) of systems such as $`\eta \pi `$ or $`\eta ^{}\pi `$ in the mass region below 2 GeV requires care and experience. This is so because (1) this region is dominated by the strong $`2^+`$ ‘background’ ($`a_2`$ resonance), and (2) that the PWA may give ambiguous results for the weaker $`1^+`$ wave. For Primakoff production, the hybrid production cross section may increase relative to the $`a_2`$ state, considering the estimated radiative widths. These are $`\mathrm{\Gamma }(a_2\pi \gamma )=300`$ keV, and $`\mathrm{\Gamma }(\pi _1\pi \gamma )90540`$ keV, as discussed in Section 2. Therefore, the PWA uncertainties for the $`1^+`$ wave will be different and may even improve. The problem generally is that the PWA of the $`\eta \pi `$ system must take into account $`S`$-, $`P`$\- and $`D`$-waves, and the number of observables is not sufficient to solve all equations unambiguously. The strength and phase ambiguities as a function of mass of different partial wave solutions are discussed in ref. . However, in certain experimental situations, the ambiguous solutions have relatively little impact on a particular exotic wave under study—such a situation seems to be the case with the BNL E852 data. In addition, it has been shown that certain other assumptions, e.g., the rank-one condition, can be removed in a systematic study in which the mass-dependence of each partial wave is introduced explicitly into the analysis.
The masses and widths of $`\pi _1(1400)`$ meson in the decay channel $`\pi \eta `$ are summarized in Table I.
| Table I: Parameters for $`\pi _1(1400)\eta \pi `$ | | |
| --- | --- | --- |
| Expt. | Mass(MeV) | Width(MeV) |
| KEK | $`1323.1\pm 4.6`$ | $`143.2\pm 12.5`$ |
| BNL(’94) | $`1370\pm 16\genfrac{}{}{0pt}{}{\text{ }+\text{ }50}{\text{ }\text{ }30}`$ | $`385\pm 40\genfrac{}{}{0pt}{}{\text{ }+\text{ }65}{\text{ }\text{ }105}`$ |
| BNL(’95) | $`1359\genfrac{}{}{0pt}{}{\text{ }+\text{ }16+10}{\text{ }\text{ }1424}`$ | $`314\genfrac{}{}{0pt}{}{\text{ }+\text{ }31+9}{\text{ }\text{ }2966}`$ |
| CB | $`1400\pm 20\pm 20`$ | $`310\pm 50\genfrac{}{}{0pt}{}{\text{ }+\text{ }50}{\text{ }\text{ }30}`$ |
| CB | $`1360\pm 25`$ | $`220\pm 90`$ |
At a recent Workshop on Hadron Spectroscopy , the VES collaboration presented the results of a coupled-channel analysis of the $`\pi _1(1600)`$ meson in the channels $`\rho \pi `$, $`\eta ^{}\pi `$ and $`b_1(1235)\pi `$. Their results are consistent with the BNL results, as seen in Table II.
| Table II: Parameters for $`\pi _1(1600)`$ Decay | | | |
| --- | --- | --- | --- |
| Expt. | Mass(MeV) | Width(MeV) | Decay |
| BNL | $`1593\pm 8\genfrac{}{}{0pt}{}{\text{ }+\text{ }20}{\text{ }\text{ }47}`$ | $`168\pm 20\genfrac{}{}{0pt}{}{\text{ }+\text{ }150}{\text{ }\text{ }12}`$ | $`\rho \pi `$ |
| BNL | $`1596\pm 8`$ | $`387\pm 23`$ | $`\eta ^{}\pi `$ |
| VES | $`1610\pm 20`$ | $`290\pm 30`$ | $`\rho \pi `$,$`\eta ^{}\pi `$,$`b_1\pi `$ |
For both BNL E852 and VES data, it is not known what Regge exchanges are responsible for the production of the $`J^{PC}=1^+`$ exotic states at 1.4 and 1.6 GeV, Both the $`a_2(1320)`$ and the exotic waves are produced via natural-parity exchanges which include the Pomeron. If Pomeron exchange is indeed responsible the production, then diffractive production in COMPASS can provide an additional handle with which to tackle the study of exotic waves.
One can succinctly summarize the situation as follows: a production of the wave $`I^G(J^{PC})=1^{}(1^+)`$ is dependent on the strength of the $`\pi \rho `$ decay modes in the case of the Primakoff production, whereas in diffractive production the relative strengths depend on the supposed decay modes $`\pi _1(1400)\pi `$ and $`\pi _1(1600)\pi `$ of the tensor glueball (2<sup>++</sup>), since the Pomeron is thought to be on the Regge trajectory corresponding to the tensor glueball with a presumed mass around 2 GeV. Corresponding to the glueball decay $`G(2^{++})\pi ^+Hybrid`$, one expects diffractive production via $`\pi ^{}G(2^{++})Hybrid`$. This is an additional strong advantage of the COMPASS hybrid meson study. We can look forward to two complementary production modes of exotic mesons, increasing our chance for achieving a decisive advance on our understanding of the meson constituents. Finally, the E852 collaboration finds preliminary evidence of a third exotic meson at around 1.9 GeV. The search for this state as well as others can continue, with exciting new results anticipated. In summary, COMPASS can move into a forefront of hadron spectroscopy, by studying Primakoff and diffractive production of nonstrange light-quark hybrid mesons in the 1.4-2.5 GeV mass region, including all the hybrid candidates from previous studies.
1.2 Radiative Transitions
Radiative decay widths of mesons and baryons are powerful tools for understanding the structure of elementary particles and for constructing dynamical theories of hadronic systems. Straightforward predictions for radiative widths make possible the direct comparison of experiment and theory. The small value of branching ratios of radiative decays makes them difficult to measure directly, because of the large background decay $`\pi ^0`$s from strong decays. Studying the inverse reaction $`\gamma +\pi ^{}M^{}`$ provides a relatively clean method for the determination of the radiative widths. Very good tracking resolution is needed (and available through silicon strip detectors) to measure initial and final state momenta, and to thus exhibit the Primakoff signal at small momentum transfers, where the electromagnetic processes dominate over the strong interaction.
In COMPASS, we will study radiative transitions of incident mesons to higher excited states. We will obtain new data () for radiative transitions leading from the pion to a<sub>1</sub>(1260), a<sub>2</sub>(1320), and $`\rho `$ mesons. The previous Coulomb field measurements of the a<sub>1</sub>(1260) $`\pi \gamma `$ width () is $`0.64\pm 0.25`$ MeV; of the a<sub>2</sub>(1320) $`\pi \gamma `$ width is 0.30 $`\pm `$ 0.06 MeV; and of the $`\rho \pi \gamma `$ width () is 60 to 81 keV. We will obtain independent and significantly higher precision data and statistics for these and higher resonances. This will be valuable in order to allow more meaningful comparisons with theoretical predictions, and as a normalization of the Hybrid meson studies.
1.3 Experimental Requirements
We consider the beam, target, detector, and trigger requirements for hybrid meson production and detection with minimum background contamination.
1.3.1 Monte Carlo Simulations
We carry out Monte Carlo simulations with HYBRID, an event generator adapted to COMPASS for Hybrid meson physics studies. For hybrid mesons, simulations were carried out for the FNAL SELEX apparatus, a low rate forward spectrometer, otherwise very similar to COMPASS. SELEX however did not obtain quality hybrid physics data. We base our initial planning on the previous FNAL Primakoff experiments and on available SELEX simulations, while pursuing further simulations for COMPASS .
1.3.2 Beam Requirements
A beam Cherenkov detector (CEDARS) far upstream of the target provides $`\pi /K/p`$ identification. We will take data with both positive and negative beams. We may use a 280 GeV beam, the highest energy available at COMPASS, because Hybrid meson Primakoff production cross sections increase with increasing energy, and because the calorimeter acceptances are higher at the highest energy. Beam rates lower than the 40 MHz COMPASS design rate are planned for initial setup studies, in which many of the COMPASS systems (DAQ, detectors, etc.) must be implemented. We base our count rate estimates on beam intensities of 10 MHz ($`20\times 10^7`$ particles in a 2 second spill with a total cycle time of 14.8 seconds).
1.3.3 Target and Target Detectors
We mainly veto target break-up events by positioning veto scintillators around the target. We also veto target break-up events by selecting multiplicity 1 or 3 events in downstream hodoscopes H1 and H2 (Fig. 2) at the trigger level, and by selecting low-t events in the off-line analysis. Before and after the target, charged particles are tracked by high resolution tracking detectors. We achieve good angular resolution for the final state charged particles by minimizing the multiple scattering in the targets and detectors. The Primakoff targets will be Pb of 1% interaction length = 2 g/cm<sup>2</sup> =0.30 radiation length, and other targets such as Cu of similar and also smaller radiation length. The beam and outgoing pion multiple Coulomb scattering in the target gives an rms angular resolution of about 40 $`\mu `$rad.
1.3.4 The Magnetic Spectrometer and the t-Resolution
The incoming beam momentum is measured with upstream SPS detectors. The final state pion and $`\gamma `$ momenta are measured with good resolution in downstream COMPASS magnetic spectrometers and in the photon calorimeter, respectively. Via the measurement of incident and final state momenta, we obtain a precise determination of the square of the four momentum transfer $`t`$ to the target nucleus. The small transverse momentum kick p<sub>T</sub> to the target and $`t`$ are related by $`t=p_T^2`$. We aim for a p<sub>T</sub> resolution or about 10 MeV/c, corresponding to resolution $`\mathrm{\Delta }t=2\times 10^4`$ GeV<sup>2</sup> for production of a 2 GeV Hybrid. This goal is based on the need to minimize contributions to the Coulomb Primakoff data from diffractive production. A peak in the $`t`$ distribution at low $`t`$ provides the main signature of the Primakoff process, and the means to separate Primakoff from diffractive scattering. The Pb diffractive data for example falls as $`\mathrm{exp}(t/0.0025)`$ with t expressed in GeV<sup>2</sup>. Our t-resolution goal is then about a factor of 10 smaller than the slope in t observed for diffractive data on a Pb target. This goal is clearly achievable, as one may see from the t distributions measured at a low statistics but high resolution experiment for $`\pi ^{}\pi ^{}\pi ^0`$ and $`\pi ^{}\pi ^{}\gamma `$ Primakoff scattering at 200 GeV at FNAL. The t distribution of the $`\pi ^{}\pi ^{}\gamma `$ data agrees well with the Primakoff formalism out to t = $`10^3`$ GeV<sup>2</sup>, which indicates that the data are indeed dominated by Coulomb production.
1.3.5 The Photon Calorimeter ECAL2
The momentum kicks of the two COMPASS magnets are set $`\mathrm{𝑎𝑑𝑑𝑖𝑡𝑖𝑣𝑒}`$ for maximum deflection of the beam from the zero degree (neutral ray) line. This maximizes the distance from the zero degree line to the beam hole in ECAL2 (located about 30 meters from target), and attains an acceptable distance of at least 10 cm between the zero degree line and the deflected beam position at ECAL2 for the proposed 280 GeV beam energy. The hole size and position must be optimized to minimize the number of pions hitting ECAL2 blocks at the hole perimeter. The Primakoff $`\gamma `$’s are centered around the zero degree line, and a good $`\gamma `$ measurement requires clean signals from 9 blocks, centered on the hit block. The ADC electronic readouts for these blocks have been designed and are being built.
As can be seen in Fig. 1, COMPASS needs to also detect $`\eta `$s for the hybrid study. The two $`\gamma `$s from $`\eta `$ decay have half-opening angles $`\theta _{\gamma \gamma }^h`$ for the symmetric decays of $`\theta _{\gamma \gamma }^h=m/E_\eta `$, where m is the mass ($`\eta `$) and E<sub>η</sub> is the $`\eta `$ energy. Opening angles are somewhat larger for asymmetric decays. In order to catch about 50% of the decays, it is necessary to subtend a cone with double that angle, i.e. $`\pm 2m/E_\eta `$, neglecting the angular spread of the original $`\eta `$s around the beam direction. Consider an ECAL2 $`\gamma `$ detector with a circular active area with 2 m diameter. Consider the $`\pi \eta `$ channel. For an ECAL2 of 1 m radius at 30 m from the target, $`\eta `$s above E<sub>η</sub>=33 GeV are therefore accepted. At half this energy, the acceptance practically vanishes. The acceptance of course depends on the Hybrid mass, mostly between 1.4 and 2.5 GeV for the planned COMPASS study. Detailed Monte Carlo studies are in progress for the different possibly Hybrid decay modes, for a range of assumed masses. One also must consider the efficiency that the two $`\gamma `$s are sufficiently separated, to be able to get a position and energy measurement for each of them. For the $`\pi `$f<sub>1</sub> channel, with for example $`f_1\pi \pi \eta `$, the $`\eta `$s will have low energy, and therefore large gamma angles. To maintain good acceptance for low energy $`\eta `$s, the ECAL2 diameter should be 2 m or more.
Primakoff physics requires a very good position and energy resolution of photon calorimeters. The ECAL2 blocks will have their gains well matched, and their analog signals will be electronically summed and discriminated to provide a trigger signal on minimal energy deposit. For the precise monitoring of the energy calibration of the photon calorimeters, COMPASS will use a dedicated laser system .
1.3.6 The Primakoff Trigger
We design the COMPASS Primakoff/Diffractive hybrid meson trigger to enhance the acceptance and statistics, and also to yield a trigger rate closer to the natural rate given by the low hybrid cross sections. The trigger should suppress the beam rate by a factor 10<sup>3</sup>-10<sup>4</sup> or better, while also achieving high acceptance. The resulting rate will be significantly lower than the maximum of 10<sup>5</sup> per second DAQ limit in COMPASS . We veto target break-up events via veto scintillators around the target. For hybrid meson physics, the trigger uses the characteristic hybrid decay pattern: one or three charged hadrons with gamma hits, or three charged hadrons and no gamma hits. The hybrid trigger for the $`\pi \eta `$ hybrid decay channel (charged particle multiplicity =1) is based on a determination of the pion energy loss (via its characteristic angular deflection), correlated in downstream scintillator hodoscopes stations (H1 versus H2) with the aid of a fast matrix chip, as shown in Fig. 2. The chip was designed and already tested for the analogous COMPASS energy-loss trigger for the muon beam runs.
We will also test alternative and/or complementary trigger concepts. We have already successfully carried out trigger tests at the CERN test channels, and made reports at COMPASS collaboration meetings. For example, the non-interacting beam may be detected and vetoed by the Beam Kill veto trigger detectors BK1/BK2, which follow the pion trajectory, as shown in Fig. 2. This Beam Kill idea must be tested. It requires very fine segmentation for the BK1 and BK2 detectors to be able to accept a 40 MHz beam rate, and it requires very high efficiency since detector inefficiencies in killing the beam can lead to artificially high trigger rates. We will test the energy loss trigger described above, with and without adding the beam kill trigger capability. Depending on the results, we will make a decision on implementing the BK detectors. Even if finally not implemented for the data taking, tests with the beam killer detectors will be valuable for setting up the energy loss trigger.
We foresee a three-level Primakoff trigger scheme: T0 = beam definition, T1 = event topology, T2 = online software filter. T0 is a fast logical signal that includes upstream CEDARS Cherenkov detectors for beam PID. T1 is a downstream coincidence between scintillation H1-H2 hodoscope signals, with charged particle multiplicity 1 ($`\pi ^{}`$) or 3 (2 negative, one positive) conditions. T2 is an intelligent software filter, placed in the DAQ stream after the event builder, which counts the number of reconstructed track segments upstream and downstream from the target, and also sets cuts on event quality characteristics (goodness of kink or vertex reconstruction, cuts on the t-distribution, etc.).
As an example, for M<sub>HY</sub> near 1.5 GeV, HY $`\pi ^{}\eta `$ events can have 40-235 GeV pions at angles larger than 0.5 mrad, close to the non-interacting 280 GeV beam, and two forward photons at angles less than 30 mrad. Pions with energy lower than 40 GeV are blocked by the magnet yokes. The kinematic variables for the HY $`\pi ^{}\eta `$ Primakoff process are shown in Fig. 1. A virtual photon from the Coulomb field of the target nucleus interacts with the pion, a Hybrid meson is produced and decays to $`\pi ^{}\eta `$ at small forward angles in the laboratory frame, while the target nucleus (in the ground state) recoils coherently with a small transverse kick p<sub>T</sub>. The peak at small target $`p_T`$ used to identify the Primakoff process is measured offline using the beam and vertex silicon detectors. For diffractive processes, the beam pion interacts with an exchanged Pomeron.
The T1 trigger scheme (for photon-pion reactions with one charged pion and two photons in the final state) for an assumed 280 GeV pion beam is shown in Fig. 2. BK1 and BK2 are small scintillator hodoscopes; while H1 and H2 are larger scintillator hodoscopes with larger segmentation, all in the beam bend plane. The beam passes through holes in H1 and H2. An anticoincidence in the “beam” region of BK1 and BK2 can be used to veto non-interacting beam pions. According to Monte Carlo simulation, a 30 GeV energy loss condition achieves beam suppression with 99.8% efficiency, while maintaining 100% efficiency for all Primakoff scattered pions with momenta at least 30 GeV/c lower than the beam momentum. For a trigger that accepts Primakoff scattered pions with momenta at least 45 GeV/c lower than the beam momentum, beam suppression is of course yet better. For the case of one charged pion in the final state, we require an ECAL2 $`\gamma `$ signal above 45 GeV in coincidence with a final-state pion in the energy range $`(40235\mathrm{GeV})`$. The threshold for the ECAL2 $`\gamma `$ signal is set to match the kinematic region of Primakoff scattered pions that satisfy the 45 GeV energy loss trigger condition. The BK1 and BK2 hodoscope sizes are optimized for beam pions. We measure the energy loss of the $`\pi ^{}`$ via characteristic angular deflections, correlated with hits in the H1 versus H2 hodoscopes. A fast matrix chip is used for this purpose, as developed for the standard muon energy-loss trigger in planned COMPASS studies of gluon polarization in the proton. A coincidence based on H1/H2 hodoscope correlations, with multiplicity=1 condition, where the H1-H2 line projects back to the target beam spot, is used to trigger on the Primakoff decay pion in the $`\pi ^{}\eta `$ channel.
This trigger configuration for Primakoff scattering strongly suppresses backgrounds associated with both non-interacting beam particles and those involving nuclear interaction of pions in the target and in COMPASS apparatus. The Primakoff trigger design suppresses the beam rate by up to a factor 10<sup>3</sup>-10<sup>4</sup>, achieving high acceptance efficiency for events versus the important kinematic variables. The diffractive cross sections leading to the $`\pi ^{}\pi ^0\pi ^0`$ and $`\pi ^{}\pi ^{}\pi ^+`$ final states are large. Since we aim to also observe diffractive production, we may need to prescale the triggers for such strong channels.
We will study the acceptance of this trigger in COMPASS for the $`\pi ^{}\eta `$ hybrid meson decay mode using our MC code HYBRID , which generates Primakoff pion-photon hybrid meson production interactions, with realistic beam phase space. We will also study the $`\pi f_1`$ and other decay mode acceptances with HYBRID. The T1 trigger scheme for the $`\pi f_1`$ decay channel is based on part or all of the following: a multiplicity = 3 condition, condition of one positive and 2 negative tracks, condition that one can find 3 H1-H2 track lines that all point back to the small beam spot on target, condition that the total energy associated with the tracks and $`\gamma `$ energy is of order 280 GeV. How many of these conditions we include depends on how low a trigger rate we may easily achieve with minimum bias. For the case of three charged particles and also $`\gamma `$s in the final state, the trigger rate is lower than the beam rate by a lower factor than for $`\pi \eta `$, but still lower than the COMPASS DAQ limit. In addition, the H1 and H2 hodoscope may veto charged particles with larger angles than expected for all the hybrid decay channels, and also events with multiplicity higher than 3.
Minimum material (radiation and interaction lengths) in COMPASS will give a clean (low background) trigger. This is so since $`\gamma `$s may arrive at ECAL2 with minimum interaction losses, while producing minimum background interactions. That is, it is best to use a minimally instrumented COMPASS apparatus for Primakoff and Diffractive hybrid meson structure studies.
2. Objectives and Expected Significance
COMPASS can contribute significantly to the further investigation of hybrids by studying Primakoff and Diffractive production of $`I^G(J^{PC})=1^{}(1^+)`$$`\pi _1`$’—or more generally $`I^G(J^{PC})=1^+(0^+)`$$`b_0`$’ or $`I^G(J^{PC})=1^+(2^+)`$$`b_2`$’—hybrids. The possibilities for Primakoff production of the $`\pi _1`$ with energetic pion beams, and detection via different decay channels, were discussed previously in Refs. , and Monte Carlo simulations for this physics are in progress for COMPASS . The experiment will be run with the COMPASS spectrometer, consisting of the spectrometer magnets, the central tracking detectors, the ECAL2 calorimeter, and a relatively simple trigger. The COMPASS Primakoff trigger will allow observation of the $`\pi _1`$ via the $`\eta \pi ^{}`$ decay mode. With a relative $`P`$-wave (L=1), the $`\eta \pi ^{}`$ system has $`J^{PC}=1^+`$. The other decay channels of $`\pi _1`$ may be studied simultaneously in COMPASS by relatively simple particle multiplicity triggers (three charged particles in final state, etc.).
We make rough estimates of the statistics attainable for hybrid production in the COMPASS experiment. Monte Carlo simulations in progress will refine these estimates. We assume a 125-1250$`\mu `$b Hybrid meson production cross section per Pb nucleus (near 1.5 GeV mass). This estimate is based on two considerations. First, a straightforward application of VDM with $`\rho \gamma `$ coupling g$`{}_{\rho \gamma }{}^{2}/\pi `$=2.5, gives a width of $`\mathrm{\Gamma }(\pi _1\pi \gamma `$) = 75-750 keV for a 1.5 GeV Hybrid, assuming $`\mathrm{\Gamma }(\pi _1\pi \rho `$)= 10-100 MeV, a range corresponding to 3.3-33% of the claimed 1.5 GeV hybrid width. Integrating the Primakoff Hybrid production differential cross section for a 280 GeV pion beam with this $`\mathrm{\Gamma }(\pi _1\pi \gamma `$) width gives 125-1250 $`\mu `$b. Second, a FNAL E272 measurement indicated (but with high uncertainty) that $`\mathrm{\Gamma }(\pi _1\pi \gamma )\times BR(\pi _1\pi f_1)250`$ keV for a 1.6 GeV Hybrid candidate. This would be consistent with the above maximum VDM $`\mathrm{\Gamma }(\pi _1\pi \gamma `$) estimate for BR($`\pi _1\pi f_1`$) = 33%. With a total $`\pi `$ inelastic cross section per Pb nucleus of 0.8 barn, the Primakoff Hybrid production event rate R (events per interaction) is then R= 1.6-16 $`\times `$ 10<sup>-4</sup>.
In four months of running, we obtain 1.4 $`\times `$ 10<sup>13</sup> beam pions. With a 1% interaction length target, we obtain 1.4 $`\times `$ 10<sup>11</sup> interactions. Therefore, one obtains 2.2-22$`\times `$10<sup>7</sup> Hybrid Primakoff events at 100% efficiency. We assume now a 50% accelerator operation efficiency. We also estimate a global 10% average detection efficiency over all decay channels for tracking, $`\gamma `$ detection, $`\eta `$ acceptance and identification, trigger acceptance, global geometric acceptance, and event reconstruction efficiency. All these effects give a global efficiency of 5%. Therefore, we may expect to observe a total of 1.1-11$`\times `$10<sup>6</sup> Hybrid decays in all decay channels. For example, following the Close and Page predictions, we may expect 24% in $`\pi f_1`$, 2-8% in $`\pi \rho `$, 67% in b$`{}_{1}{}^{}\pi `$, 0-4% in $`\eta \pi `$, 0-4% in $`\eta ^{}\pi `$, etc.
For 2, 2.5, 3.0 GeV mass Hybrids, the number of useful events decreases by factors of 6, 25, and 100, respectively. But even in these cases, assuming again a global 5% efficiency, that represents very interesting potential samples of 1.8-18$`\times `$10<sup>5</sup>, 4.4-44$`\times `$10<sup>4</sup>, and 1.1-11$`\times `$10<sup>4</sup> Hybrid meson detected events, with masses 2, 2.5, and 3 GeV respectively.
COMPASS can study hybrid meson candidates near 1.4, 1.6, 1.9 GeV produced by the Primakoff and Diffractive processes. COMPASS should also be sensitive to pionic hybrids in the 2-3 GeV mass range. We may obtain superior statistics for hybrid states if they exist, and via a different production mechanism, without possible complication by hadronic final state interactions. We may also get important data on the different decay modes for this state. The observation of these/other hybrids in different decay modes and in a different experiment would constitute the next important step following the evidence so far reported.
COMPASS provides a unique opportunity to investigate QCD hybrid exotics, via diffractive and Primakoff production. Taking into account the very high beam intensity, fast data acquisition, high acceptance and good resolution of the COMPASS setup, one can expect from COMPASS the highest statistics and a ‘systematics-free’ data sample that includes many tests to control possible systematic errors. Intercomparisons between COMPASS and other experiments with complementary methodologies should allow fast progress on understanding hybrid meson structure, and on fixing the systematic uncertainties.
3. Plan of Operation:
COMPASS studies hybrid meson structure via the scattering of high energy pions from “photon” and “Pomeron” targets. We use 100-280 GeV beams ($`\mu `$, $`\pi `$) and magnetic spectrometers and calorimeters to measure the complete kinematics of pion-photon and pion-Pomeron hybrid production reactions. Initial COMPASS set up runs are scheduled for Summer 2000. COMPASS will then study “proton spin” using a muon beam to measure the gluon polarization, and Hybrid meson structure using a pion beam. These programs will benefit from high statistics, excellent beam focusing and momentum analysis, and dedicated low background runs. We will analyze pion and muon test run data, we will carry out simulation studies, and we will plan and analyze dedicated COMPASS runs to maximize quality results.
More accurate physics and trigger simulations are required using the new C-programmed COMPASS GEANT package. This code should include the updated experimental setup, trigger and the DAQ schemes, accurate magnetic field mapping, and event reconstruction. We work on this package, incorporating a Primakoff hybrid meson structure event generator. We develop COMPASS event reconstruction algorithms, and test them on GEANT simulated events.
For the COMPASS Hybrid meson structure effort, we need to plan, construct and implement all hardware and software for the trigger. We will prepare the COMPASS pion-photon Primakoff trigger system by the following phases: (1) continue to investigate the hodoscope matrix energy loss and multiplicity 3 triggers, mentioned above, via MC simulations, (2) refine our MC trigger simulations using COMGEANT, (3) construct the trigger hardware, including an upgrade of the existing CEDARS Cherenkov beam PID detector, scintillation hodoscopes, fast signal summing circuitry, mechanical supports, etc., (5) install the system at CERN, (6) set up the trigger detectors and electronics in the COMPASS muon beam, (7) take preliminary data with muons, writing event reconstruction algorithms, and checking trigger performance, (8) use it for running the COMPASS hybrid meson structure experiment with pion beams.
4. Acknowledgments:
This work was supported by the Israel Science Foundation founded by the Israel Academy Sciences and Humanities, Jerusalem, Israel; and the US Department of Energy. Thanks are due to F. Bradamante, S. Paul, L. Landsberg, T. Ferbel, D. Casey, G. Mallot, H.-W. Siebert, V. Steiner, D. von Harrach, A. Bravar, M. Faessler, A. Singovski, J. Pochodzalla, A. Olchevski, S. Prakhov, Y. Khokhlov, H. Willutzki, and T. Walcher for valuable discussions.
|
warning/0003/cs0003042.html
|
ar5iv
|
text
|
# Fages’ Theorem and Answer Set Programming
## Introduction
This note is about the relationship between the completion semantics (?) and the answer set (“stable model”) semantics (?) for logic programs with negation as failure. The study of this relationship is important in connection with the emergence of answer set programming (???). Whenever the two semantics are equivalent, answer sets can be computed by a satisfiability solver, and the use of “answer set solvers” such as smodels<sup>1</sup><sup>1</sup>1http://saturn.hut.fi/pub/smodels/ . and dlv<sup>2</sup><sup>2</sup>2http://www.dbai.tuwien.ac.at/proj/dlv/ . is unnecessary.
Consider a finite propositional (or grounded) program $`\mathrm{\Pi }`$ without classical negation, and a set $`X`$ of atoms. If $`X`$ is an answer set for $`\mathrm{\Pi }`$ then $`X`$, viewed as a truth assignment, satisfies the completion of $`\mathrm{\Pi }`$. The converse, generally, is not true. For instance, the completion of
$$pp$$
(1)
is $`pp`$. This formula has two models $`\mathrm{}`$, $`\{p\}`$; the first is an answer set for (1), but the second is not. François Fages \[?\] defined a syntactic condition on logic programs that implies the equivalence between the two semantics—“positive-order-consistency,” also called “tightness” (?). What he requires is the existence of a function $`\lambda `$ from atoms to nonnegative integers (or, more generally, ordinals) such that, for every rule
$$A_0A_1,\mathrm{},A_m,\text{not}A_{m+1},\mathrm{},\text{not}A_n$$
in $`\mathrm{\Pi }`$,
$$\lambda (A_1),\mathrm{},\lambda (A_m)<\lambda (A_0).$$
It is clear, for instance, that program (1) is not tight. Fages proved that, for a tight program, every model of its completion is an answer set. Thus, for tight programs, the completion semantics and the answer set semantics are equivalent.
Our generalization of Fages’ theorem allows us to draw similar conclusions for some programs that are not tight. Here is one such program:
$$\begin{array}{c}p\text{not}q,\hfill \\ q\text{not}p,\hfill \\ pp,r.\hfill \end{array}$$
(2)
It is not tight. Nevertheless, each of the two models $`\{p\}`$, $`\{q\}`$ of its completion
$$\begin{array}{c}p\neg q(pr),\hfill \\ q\neg p,\hfill \\ r\hfill \end{array}$$
is an answer set for (2).
The idea of this generalization is to make function $`\lambda `$ partial. Instead of tight programs, we will consider programs that are “tight on a set of literals.”
First we relate answer sets to a model-theoretic counterpart of completion introduced in (?), called supportedness. This allows us to make the theorem applicable to programs with both negation as failure and classical negation, and to programs with infinitely many rules.<sup>3</sup><sup>3</sup>3The familiar definition of completion (see Appendix) is applicable to finite programs only, unless we allow infinite disjunctions in completion formulas. Then a corollary about completion is derived, and applied to a logic programming representation of the blocks world due to Ilkka Niemelä. We show how the satisfiability solver sato (?) can be used to find answer sets for that representation, and compare the performance of smodels and sato on several benchmarks.
## Generalized Fages’ Theorem
We define a rule to be an expression of the form
$$\text{Head}L_1,\mathrm{},L_m,\text{not}L_{m+1},\mathrm{},\text{not}L_n$$
(3)
$`(nm0)`$ where each $`L_i`$ is a literal (propositional atom possibly preceded by classical negation $`\neg `$), and Head is a literal or the symbol $``$. A rule (3) is called a fact if $`n=0`$, and a constraint if $`\text{Head}=`$. A program is a set of rules. The familiar definitions of answer sets, closed sets and supported sets for a program, as well as the definition of the completion of a program, are reproduced in the appendix.
Instead of “level mappings” used by Fages, we consider here partial level mappings—partial functions from literals to ordinals. A program $`\mathrm{\Pi }`$ is tight on a set $`X`$ of literals if there exists a partial level mapping $`\lambda `$ with the domain $`X`$ such that, for every rule (3) in $`\mathrm{\Pi }`$, if $`\text{Head},L_1,\mathrm{},L_mX`$ then
$$\lambda (L_1),\mathrm{},\lambda (L_m)<\lambda (\text{Head}).$$
(For the constraints in $`\mathrm{\Pi }`$ this condition holds trivially, because the head of a constraint is not a literal and thus cannot belong to $`X`$.)
###### Theorem
For any program $`\mathrm{\Pi }`$ and any consistent set $`X`$ of literals such that $`\mathrm{\Pi }`$ is tight on $`X`$, $`X`$ is an answer set for $`\mathrm{\Pi }`$ iff $`X`$ is closed under and supported by $`\mathrm{\Pi }`$.
The proof below is almost unchanged from the proof of Fages’ theorem given in (?, Section 7.4).
###### Lemma
For any program $`\mathrm{\Pi }`$ without negation as failure and any consistent set $`X`$ of literals such that $`\mathrm{\Pi }`$ is tight on $`X`$, if $`X`$ is closed under and supported by $`\mathrm{\Pi }`$, then $`X`$ is an answer set for $`\mathrm{\Pi }`$.
Proof: We need to show that $`X`$ is minimal among the sets closed under $`\mathrm{\Pi }`$. Assume that it is not. Let $`Y`$ be a proper subset of $`X`$ that is closed under $`\mathrm{\Pi }`$, and let $`\lambda `$ be a partial level mapping establishing that $`\mathrm{\Pi }`$ is tight on $`X`$. Take a literal $`LXY`$ such that $`\lambda (L)`$ is minimal. Since $`X`$ is supported by $`\mathrm{\Pi }`$, there is a rule
$$LL_1,\mathrm{},L_m$$
in $`\mathrm{\Pi }`$ such that $`L_1,\mathrm{},L_mX`$. By the choice of $`\lambda `$,
$$\lambda (L_1),\mathrm{},\lambda (L_m)<\lambda (L).$$
By the choice of $`L`$, we can conclude that
$$L_1,\mathrm{},L_mY.$$
Consequently $`Y`$ is not closed under $`\mathrm{\Pi }`$, contrary to the choice of $`Y`$.
Proof of the Theorem: Left-to-right, the proof is straightforward. Right-to-left: assume that $`X`$ is closed under and supported by $`\mathrm{\Pi }`$. Then $`X`$ is closed under and supported by $`\mathrm{\Pi }^X`$. Since $`\mathrm{\Pi }`$ is tight on $`X`$, so is $`\mathrm{\Pi }^X`$. Hence, by the lemma, $`X`$ is an answer set for $`\mathrm{\Pi }^X`$, and consequently an answer set for $`\mathrm{\Pi }`$.
In the special case when $`\mathrm{\Pi }`$ is a finite program without classical negation, a set of atoms satisfies the completion of $`\mathrm{\Pi }`$ iff it is closed under and supported by $`\mathrm{\Pi }`$. We conclude:
###### Corollary 1
For any finite program $`\mathrm{\Pi }`$ without classical negation and any set $`X`$ of atoms such that $`\mathrm{\Pi }`$ is tight on $`X`$, $`X`$ is an answer set for $`\mathrm{\Pi }`$ iff $`X`$ satisfies the completion of $`\mathrm{\Pi }`$.
For instance, program (2) is tight on the model $`\{p\}`$ of its completion: take $`\lambda (p)=0`$. By Corollary 1, it follows that $`\{p\}`$ is an answer set for (2). In a similar way, the theorem shows that $`\{q\}`$ is an answer set also.
By $`\text{pos}(\mathrm{\Pi })`$ we denote the set of all literals that occur without negation as failure at least once in the body of a rule of $`\mathrm{\Pi }`$.
###### Corollary 2
For any program $`\mathrm{\Pi }`$ and any consistent set $`X`$ of literals disjoint from $`\text{pos}(\mathrm{\Pi })`$, $`X`$ is an answer set for $`\mathrm{\Pi }`$ iff $`X`$ is closed under and supported by $`\mathrm{\Pi }`$.
###### Corollary 3
For any finite program $`\mathrm{\Pi }`$ without classical negation and any set $`X`$ of atoms disjoint from $`pos(\mathrm{\Pi })`$, $`X`$ is an answer set for $`\mathrm{\Pi }`$ iff $`X`$ satisfies the completion of $`\mathrm{\Pi }`$.
To derive Corollary 2 from the theorem, and Corollary 3 from Corollary 1, take $`\lambda (L)=0`$ for every $`LX`$.
Consider, for instance, the program
$$\begin{array}{c}p\text{not}q,\hfill \\ q\text{not}p,\hfill \\ rr,\hfill \\ pr.\hfill \end{array}$$
(4)
The completion of (4) is
$$\begin{array}{c}p\neg qr,\hfill \\ q\neg p,\hfill \\ rr.\hfill \end{array}$$
The models of these formulas are $`\{p\}`$, $`\{q\}`$ and $`\{p,r\}`$. The only literal occurring in the bodies of the rules of (4) without negation as failure is $`r`$. In accordance with Corollary 3, the models of the completion that do not contain $`r`$—sets $`\{p\}`$ and $`\{q\}`$—are answer sets for (4).
## Planning in the Blocks World
As a more interesting example, consider a logic programming encoding of the blocks world due to Ilkka Niemelä. The main part of the encoding consists of the following schematic rules:
```
goal :- time(T), goal(T).
:- not goal.
goal(T2) :- nextstate(T2,T1), goal(T1).
moveop(X,Y,T):-
time(T), block(X), object(Y), X != Y,
on_something(X,T), available(Y,T),
not covered(X,T), not covered(Y,T),Ψ
not blocked_move(X,Y,T).
on(X,Y,T2) :-
block(X), object(Y), nextstate(T2,T1),
moveop(X,Y,T1).
on_something(X,T) :-
block(X), object(Z), time(T), on(X,Z,T).
available(table,T) :- time(T).
available(X,T) :-
block(X), time(T), on_something(X,T).
covered(X,T) :-
block(Z), block(X), time(T), on(Z,X,T).
on(X,Y,T2) :-
nextstate(T2,T1), block(X), object(Y),
on(X,Y,T1), not moving(X,T1).
moving(X,T) :- time(T), block(X), object(Y),
moveop(X,Y,T).
blocked_move(X,Y,T):-
block(X), object(Y), time(T), goal(T).
blocked_move(X,Y,T) :-
time(T), block(X), object(Y),
not moveop(X,Y,T).
blocked_move(X,Y,T) :-
block(X), object(Y), object(Z), time(T),
moveop(X,Z,T), Y != Z.
blocked_move(X,Y,T) :-
block(X), object(Y), time(T), moving(Y,T).
blocked_move(X,Y,T) :-
block(X), block(Y), block(Z), time(T),
moveop(Z,Y,T), X != Z.
:- block(X), time(T), moveop(X,table,T),
on(X,table,T).
:- nextstate(T2,T1), block(X), object(Y),
moveop(X,Y,T1), moveop(X,table,T2).
nextstate(Y,X) :- time(X), time(Y),
Y = X + 1.
object(table).
object(X) :- block(X).
```
To solve a planning problem, we combine these rules with
1. a set of facts defining time/1 as an initial segment of nonnegative integers, for instance
```
time(0). time(1). time(2).
```
2. a set of facts defining block/1, such as
```
block(a). block(b). block(c).
```
3. a set of facts encoding the initial state, such as
```
on(a,b,0). on(b,table,0).
```
4. a rule that encodes the goal, such as
```
goal(T) :- time(T), on(a,b,T), on(b,c,T).
```
The union is given as input to the “intelligent grounding” program lparse, and the result of grounding is passed on to smodels (?, Section 7). The answer sets for the program correspond to valid plans.
Concurrently executed actions are allowed in this formalization as long as their effects are not in conflict, so that they can be arbitrarily interleaved.
The schematic rules above contain the variables T, T1, T2, X, Y, Z that range over the object constants occurring in the program, that is, over the nonnegative integers that occur in the definition of time/1, the names of blocks a, b,$`\mathrm{}`$ that occur in the definition of block/1, and the object constant table.
The expressions in the bodies of the schematic rules that contain = and != restrict the constants that are substituted for the variables in the process of grounding. For instance, we understand the schematic rule
```
nextstate(Y,X) :- time(X), time(Y),
Y = X + 1.
```
as an abbreviation for the set of all ground instances of
```
nextstate(Y,X) :- time(X), time(Y).
```
in which X and Y are instantiated by a pair of consecutive integers. The schematic rule
```
blocked_move(X,Y,T) :-
block(X), object(Y), object(Z), time(T),
moveop(X,Z,T), Y != Z.
```
stands for the set of all ground instances of
```
blocked_move(X,Y,T) :-
block(X), object(Y), object(Z), time(T),
moveop(X,Z,T).
```
in which Y and Z are instantiated by different object constants.
According to this understanding of variables and “built-in predicates,” Niemelä’s schematic program, including rules (i)–(iv), is an abbreviation for a finite program BW in the sense defined above.
In the proposition below we assume that schematic rule (iv) has the form
```
goal(T) :- time(T), ...
```
where the dots stand for a list of schematic atoms with the predicate symbol on and the last argument T.
###### Proposition
Program BW is tight on each of the models of its completion.
###### Lemma
For any atom of the form nextstate(Y,X) that belongs to a model of the completion of program BW, $`𝚈=𝚇+1`$.
Proof: The completion of BW contains the formula
$$\mathrm{𝚗𝚎𝚡𝚝𝚜𝚝𝚊𝚝𝚎}(𝚈,𝚇)\mathrm{𝚏𝚊𝚕𝚜𝚎}$$
for all Y, X such that $`𝚈𝚇+1`$.
Proof of the Proposition. Let $`X`$ be an answer set for BW. By $`T_{max}`$ we denote the largest argument of time/1 in its definition (i). Consider the partial level mapping $`\lambda `$ with domain $`X`$ defined as follows:
$$\begin{array}{c}\lambda (\mathrm{𝚝𝚒𝚖𝚎}(𝚃))=0,\hfill \\ \lambda (\mathrm{𝚋𝚕𝚘𝚌𝚔}(𝚇))=0,\hfill \\ \lambda (\mathrm{𝚘𝚋𝚓𝚎𝚌𝚝}(𝚇))=1,\hfill \\ \lambda (\mathrm{𝚗𝚎𝚡𝚝𝚜𝚝𝚊𝚝𝚎}(𝚈,𝚇))=1,\hfill \\ \lambda (\mathrm{𝚌𝚘𝚟𝚎𝚛𝚎𝚍}(𝚇,𝚃))=4𝚃+3,\hfill \\ \lambda (\mathrm{𝚘𝚗}\mathrm{\_}\mathrm{𝚜𝚘𝚖𝚎𝚝𝚑𝚒𝚗𝚐}(𝚇,𝚃))=4𝚃+3,\hfill \\ \lambda (\mathrm{𝚊𝚟𝚊𝚒𝚕𝚊𝚋𝚕𝚎}(𝚇,𝚃))=4𝚃+4,\hfill \\ \lambda (\mathrm{𝚖𝚘𝚟𝚎𝚘𝚙}(𝚇,𝚈,𝚃))=4𝚃+5,\hfill \\ \lambda (\mathrm{𝚘𝚗}(𝚇,𝚈,𝚃))=4𝚃+2,\hfill \\ \lambda (\mathrm{𝚖𝚘𝚟𝚒𝚗𝚐}(𝚇,𝚃))=4𝚃+6,\hfill \\ \lambda (\mathrm{𝚐𝚘𝚊𝚕}(𝚃))=4𝚃+3,\hfill \\ \lambda (\mathrm{𝚋𝚕𝚘𝚌𝚔𝚎𝚍}\mathrm{\_}\mathrm{𝚖𝚘𝚟𝚎}(𝚇,𝚈,𝚃))=4𝚃+7,\hfill \\ \lambda (\mathrm{𝚐𝚘𝚊𝚕})=4T_{max}+4.\hfill \end{array}$$
This level mapping satisfies the inequality from the definition of a tight program for every rule of BW; the lemma above allows us to verify this assertion for the rules containing nextstate in the body.
According to Corollary 1, we can conclude that the answer sets for program BW can be equivalently characterized as the models of the completion of BW.
## Answer Set Programming <br>with CCALC and SATO
The equivalence of the completion semantics to the answer set semantics for program BW shows that it is not necessary to use an answer set solver, such as smodels, to compute answer sets for BW; a satisfiability solver can be used instead. Planning by giving the completion of BW as input to a satisfiability solver is a form of answer set programming and, at the same time, a special case of satisfiability planning (?).
The Causal Calculator, or ccalc<sup>4</sup><sup>4</sup>4http://www.cs.utexas.edu/users/tag/ccalc/ ., is a system that is capable, among other things, of grounding and completing a schematic logic program, clausifying the completion, and calling a satisfiability solver (for instance, sato) to find a model. We have conducted a series of experiments aimed at comparing the run times of sato, when its input is generated from BW by ccalc, with the run times of smodels, when its input is generated from BW by lparse.
Because the built-in arithmetic of ccalc is somewhat different from that of lparse, we had to modify BW slightly. Our ccalc input file uses variables of sorts object, block and time instead of the unary predicates with these names. The rules of BW that contain those predicates in their bodies are modified accordingly. For instance, rule
```
on_something(X,T) :-
block(X), object(Z), time(T), on(X,Z,T).
```
turns into
```
on_something(B1,T) :- on(B1,O2,T).
```
The macro expansion facility of ccalc expands
```
nextstate(T2,T1)
```
into the expression
```
T2 is T1 + 1
```
that contains Prolog’s built-in is.
Figure 1 shows the run times of smodels (Version 2.24 and sato (Version 3.1.2) in seconds, measured using the Unix time command, on the benchmarks from (?, Section 9, Table 3). For each problem, one of the two entries corresponds to the largest number of steps for which the problem is not solvable, and the other to the smallest number of steps for which a solution exists. The experiments were performed on an UltraSPARC with 124 MB main memory and a 167 MHz CPU.
The numbers in Figure 1 are “search times”—the grounding and completion times are not included. The computation involved in grounding and completion does not depend on the initial state or the goal of the planning problem and, in this sense, can be viewed as “preprocessing.” lparse performs grounding more efficiently than ccalc, partly because the former is written in C$`++`$ and the latter in Prolog. The last benchmark in Figure 1 was grounded by lparse (Version 0.99.41) in 16 seconds; ccalc (Version 1.23) spent 50 seconds in grounding and about the same amount of time forming the completion. But the size of the grounded program is approximately the same in both cases: lparse generated 191621 rules containing 13422 atoms, and ccalc generated 200661 rules containing 13410 atoms.
## Discussion
Fages’ theorem, and its generalization proved in this note, allow us to compute answer sets for some programs by completing them and then calling a satisfiability solver. We showed that this method can be applied, for instance, to the representation of the blocks world proposed in (?). This example shows that satisfiability solvers may serve as useful computational tools in answer set programming.
There are cases, on the other hand, when the completion method is not applicable. Consider computing Hamiltonian cycles in a directed graph (?). We combine the rules
```
in(U,V) :- edge(U,V), not out(U,V).
out(U,V) :- edge(U,V), not in(U,V).
:- in(U,V), in(U,W), V != W.
:- in(U,W), in(V,W), U != V.
reachable(V) :- in(v0,V).
reachable(V) :- reachable(U), in(U,V).
:- vertex(U), not reachable(U).
```
with a set of facts defining the vertices and edges of the graph; v0 is assumed to be one of the vertices. The answer sets for the resulting program correspond to the Hamiltonian cycles. Generally, the completion of the program has models different from its answer sets. Take, for instance, the graph consisting of two disjoint loops:
```
vertex(v0). vertex(v1).
edge(v0,v0). edge(v1,v1).
```
This graph has no Hamiltonian cycles, and, accordingly, the corresponding program has no answer sets. But the set
```
vertex(v0), vertex(v1), edge(v0,v0),
edge(v1,v1), in(v0,v0), in(v1,v1),
reachable(v0), reachable(v1)
```
is a model of the program’s completion.
## Acknowledgements
We are grateful to Victor Marek, Emilio Remolina, Mirek Truszczyński and Hudson Turner, and to the anonymous referees, for comments and criticisms. This work was partially supported by National Science Foundation under grant IIS-9732744.
## Appendix: Definitions
The notion of an answer set is defined first for programs whose rules do not contain negation as failure. Let $`\mathrm{\Pi }`$ be such a program, and let $`X`$ be a consistent set of literals. We say that $`X`$ is closed under $`\mathrm{\Pi }`$ if, for every rule
$$\text{Head}\text{Body}$$
in $`\mathrm{\Pi }`$, $`\text{Head}X`$ whenever $`\text{Body}X`$. (For a constraint, this condition means that the body is not contained in $`X`$.) We say that $`X`$ is an answer set for $`\mathrm{\Pi }`$ if $`X`$ is minimal among the sets closed under $`\mathrm{\Pi }`$ w.r.t. set inclusion. It is clear that a program without negation as failure can have at most one answer set.
To extend this definition to arbitrary programs, take any program $`\mathrm{\Pi }`$, and let $`X`$ be a consistent set of literals. The reduct $`\mathrm{\Pi }^X`$ of $`\mathrm{\Pi }`$ relative to $`X`$ is the set of rules
$$\text{Head}L_1,\mathrm{},L_m$$
for all rules (3) in $`\mathrm{\Pi }`$ such that $`L_{m+1},\mathrm{},L_nX`$. Thus $`\mathrm{\Pi }^X`$ is a program without negation as failure. We say that $`X`$ is an answer set for $`\mathrm{\Pi }`$ if $`X`$ is an answer set for $`\mathrm{\Pi }^X`$.
A set $`X`$ of literals is closed under $`\mathrm{\Pi }`$ if, for every rule (3) in $`\mathrm{\Pi }`$, $`\text{Head}X`$ whenever $`L_1,\mathrm{},L_mX`$ and $`L_{m+1},\mathrm{},L_nX`$. We say that $`X`$ is supported by $`\mathrm{\Pi }`$ if, for every $`LX`$, there is a rule (3) in $`\mathrm{\Pi }`$ such that $`\text{Head}=L`$, $`L_1,\mathrm{},L_mX`$ and $`L_{m+1},\mathrm{},L_nX`$.
Let $`\mathrm{\Pi }`$ be a finite program without classical negation. If $`H`$ is an atom or the symbol $``$, by $`\text{Comp}(\mathrm{\Pi },H)`$ we denote the formula
$$H(A_1\mathrm{}A_m\neg A_{m+1}\mathrm{}\neg A_n)$$
where the disjunction extends over all rules
$$HA_1,\mathrm{},A_m,\text{not}A_{m+1},\mathrm{},\text{not}A_n$$
in $`\mathrm{\Pi }`$ with the head $`H`$. The completion of $`\mathrm{\Pi }`$ is set of formulas $`\text{Comp}(\mathrm{\Pi },H)`$ for all $`H`$.
|
warning/0003/astro-ph0003398.html
|
ar5iv
|
text
|
# Spherical Scalar Field Halo in Galaxies
## Abstract
We study a spherically symmetric fluctuation of scalar dark matter in the cosmos and show that it could be the dark matter in galaxies, provided that the scalar field has an exponential potential whose overall sign is negative and whose exponent is constrained observationally by the rotation velocities of galaxies. The local space-time of the fluctuation contains a three dimensional space-like hypersurface with surplus of angle.
The existence of dark matter in the Universe has been firmly established by astronomical observations at very different length-scales, ranging from single galaxies, to clusters of galaxies, up to cosmological scale (see for example ). A large fraction of the mass needed to produce the observed dynamical effects in all these very different systems is not seen. At the galactic scale, the problem is clearly posed: The measurements of rotation curves (tangential velocities of objects) in spiral galaxies show that the coplanar orbital motion of gas in the outer parts of these galaxies keeps a more or less constant velocity up to several luminous radii , forming a radii independent curve in the outer parts of the rotational curves profile; a motion which does not correspond to the one due to the observed matter distribution, hence there must be present some type of dark matter causing the observed motion. The flat profile of the rotational curves is maybe the main feature observed in many galaxies. It is believed that the dark matter in galaxies has an almost spherical distribution which decays like $`1/r^2`$. With this distribution of some kind of matter it is possible to fit the rotational curves of galaxies quite well . Nevertheless, the main question of the dark matter problem remains; which is the nature of the dark matter in galaxies? The problem is not easy to solve, it is not sufficient to find out an exotic particle which could exist in galaxies in the low energy regime of some theory. It is necessary to show as well, that this particle (baryonic or exotic) distributes in a very similar manner in all these galaxies, and finally, to give some reason for its existence in galaxies.
In previous works it has been explored, with considerable success, the possibility that scalar fields could be the dark matter in spiral galaxies by assuming that the scalar dark matter distributes as an axially symmetric halo . The idea of these works is to explore whether a scalar field can fluctuate along the history of the Universe and thus forming concentrations of scalar field density. If, for example, the scalar field evolves with a scalar field potential $`V(\mathrm{\Phi })\mathrm{\Phi }^2`$, the evolution of this scalar field will be similar to the evolution of a perfect fluid with equation of state $`p=0`$, $`i.e.`$, it would evolve as cold dark matter . However, it is not clear whether a spherical scalar field fluctuation can serve as dark matter in galaxies. In this letter we show that this could be the case. We assume that the halo of a galaxy is a spherical fluctuation of cosmological scalar dark matter and study the consequences for the space-time background at this scale, in order to restrict the state equation corresponding to the dark matter inside the fluctuation. We start from the general spherically symmetric line element and find out the conditions on the metric in order that the test particles in the galaxy possess a flat rotation curve in the region where the scalar field (the dark matter) dominates. Finally we show that a spherical fluctuation of the scalar field could be the dark matter in galaxies.
Assuming thus that the dark matter is scalar, we start with the energy momentum tensor $`T_{\mu \nu }=\mathrm{\Phi }_{,\mu }\mathrm{\Phi }_{,\nu }1/2g_{\mu \nu }\mathrm{\Phi }^{,\sigma }\mathrm{\Phi }_{,\sigma }g_{\mu \nu }V(\mathrm{\Phi })`$, being $`\mathrm{\Phi }`$ the scalar field and $`V(\mathrm{\Phi })`$ the scalar potential. The Klein-Gordon and Einstein equations respectively are:
$`\mathrm{\Phi }_{;\mu }^{;\mu }{\displaystyle \frac{dV}{d\mathrm{\Phi }}}=0`$
$`R_{\mu \nu }`$ $`=`$ $`\kappa _0[\mathrm{\Phi }_{,\mu }\mathrm{\Phi }_{,\nu }+g_{\mu \nu }V(\mathrm{\Phi })],`$
where $`R_{\mu \nu }`$ is the Ricci tensor, $`\sqrt{g}`$ the determinant of the metric, $`\kappa _0=8\pi G`$ and a semicolon stands for covariant derivative according to the background space-time; $`\mu ,\nu =0,1,2,3`$.
Assuming that the halo has spherical symmetry and that dragging effects on stars and dust are inappreciable, i.e. the space-time is static, the following line element is the appropriate
$$ds^2=B(r)dt^2+A(r)dr^2+r^2d\theta ^2+r^2\mathrm{sin}^2\theta d\phi ^2$$
(1)
where $`A`$ and $`B`$ are arbitrary functions of the coordinate $`r`$. Following the analysis made for axisymmetric stationary space-times , we consider the Lagrangian for a test particle travelling on the space time described by (1) which is
$$2=B\dot{t}^2+A\dot{r}^2+r^2\dot{\theta }^2+r^2\mathrm{sin}^2\theta \dot{\phi }^2$$
(2)
where a dot means derivative with respect to the proper time. From (2) the generalized momenta read:
$`p_t`$ $`=`$ $`E=B\dot{t}`$ (3)
$`p_r`$ $`=`$ $`A\dot{r}`$ (4)
$`p_\theta `$ $`=`$ $`L_\theta =r^2\dot{\theta }`$ (5)
$`p_\phi `$ $`=`$ $`L_\phi =r^2\mathrm{sin}^2\theta \dot{\phi }`$ (6)
being $`E`$ the total energy of a test particle and $`L_i`$ the component of its angular momentum. It can be defined the Hamiltonian $`=p^\mu \dot{q_\mu }`$ and after rescalling the proper time for the lagrangian to equal $`1/2`$ for time-like geodesics, the geodesic equation for material particles (stars and dust) arises
$$\dot{r}^2+\frac{1}{A}[1+\frac{L_T^2}{r^2}\frac{E^2}{B}]=0$$
(7)
being $`L_T^2=L_\theta ^2+\frac{L_\phi ^2}{\mathrm{sin}^2\theta }`$ the first integral corresponding to the squared total angular momentum. We are interested in circular and stable motion of test particles, therefore the following conditions must be satisfied
i) $`\dot{r}=0`$, circular trajectories
ii)$`\frac{V(r)}{r}=0`$, extreme ones
iii)$`\frac{^2V(r)}{r^2}|_{extr}>0`$, and stable.
being $`V(r)=\left[1+L_T^2/r^2E^2/B\right]/A`$. Following it is found that the tangential velocity of the test particle is
$$v^{tangential}=v^\phi =\sqrt{\frac{rB^{}}{2B}}$$
(8)
where means derivative with respect to $`r`$. It is easy to show that if flat rotation curves are required, it arises the following flat curve condition from (8), that is $`B=B_0r^l`$ with $`l=2(v^\phi )^2`$. With the flat curve condition, metric (1) becomes
$$ds^2=B_0r^ldt^2+A(r)dr^2+r^2d\theta ^2+r^2\mathrm{sin}^2\theta d\phi ^2$$
(9)
This result is not surprising. Remember that the Newtonian potential $`\psi `$ is defined as $`g_{00}=exp(2\psi )=12\psi \mathrm{}`$. On the other side, the observed rotational curve profile in the dark matter dominated region is such that the rotational velocity $`v^\phi `$ of the stars is constant, the force is then given by $`F=(v^\phi )^2/r`$, which respective Newtonian potential is $`\psi =(v^\phi )^2\mathrm{ln}(r)`$. If we now read the Newtonian potential from the metric (9), we just obtain the same result. Metric (9 ) is then the metric of the general relativistic version of a matter distribution, which test particles move in constant rotational curves. Function $`A`$ will be determined by the kind of substance we are supposing the dark matter is made of. Assuming the flat curve condition in the scalar dark matter hypothesis, we are in the position to write down the set of field equations. Using (9), the Klein Gordon equation reads
$$\mathrm{\Phi }^{\prime \prime }+\frac{1}{2r}\left[l+4\frac{A^{}}{A}r\right]\mathrm{\Phi }^{}\frac{1}{4}A\frac{dV(\mathrm{\Phi })}{d\mathrm{\Phi }}=0$$
(10)
and the Einstein equations are
$`{\displaystyle \frac{A(l+1)}{r^2}}`$ $`=`$ $`\kappa _0\left[{\displaystyle \frac{1}{2}}\mathrm{\Phi }^{}{}_{}{}^{2}AV(\mathrm{\Phi })\right]`$ (11)
$`{\displaystyle \frac{1}{4r^2}}\left[l^2{\displaystyle \frac{A^{}}{A}}r\left(l+2\right)\right]`$ $`=`$ $`\kappa _0\left[{\displaystyle \frac{1}{2}}\mathrm{\Phi }^{}{}_{}{}^{2}+AV(\mathrm{\Phi })\right]`$ (12)
$`{\displaystyle \frac{1}{r^2}}\left[1A{\displaystyle \frac{A^{}}{A}}r\right]`$ $`=`$ $`\kappa _0\left[{\displaystyle \frac{1}{2}}\mathrm{\Phi }^{}{}_{}{}^{2}+AV(\mathrm{\Phi })\right]`$ (13)
In order to solve equations (11-13), observe that the combination of the previous equations $`[(2l)`$(eq. 11)$`4`$ (eq. 12)$`+(2+l)`$(eq. 13)$`]`$ implies
$$V=\frac{l}{\kappa _0(2l)}\frac{1}{r^2}$$
(14)
This is a very important result, namely the scalar potential goes always as $`1/r^2`$ for a spherically symmetric metric with the flat curve condition. It is remarkable that this behavior of the stress tensor coincides with the expected behavior of the energy density of the dark matter in a galaxy. We can go further and solve the field equations, the general solution of equations (11-13) is
$`A(r)=\left(4l^2\right)/\left(4+C\left(4l^2\right)r^{\left(l+2\right)}\right),`$
being $`C`$ an integration constant and we can thus integrate the function $`\mathrm{\Phi }.`$ Nevertheless, in this letter we consider the most simple solution of the field equations with $`C=0`$. Observe that for this particular solution the stress tensor goes like $`1/r^2`$. The energy momentum tensor is made essentially of two parts. One is the scalar potential and the other one contains products of the derivatives of the scalar field, both going as $`1/r^2`$ . Furthermore, as $`(\mathrm{\Phi }_{,r})^21/r^2`$, this means that $`\mathrm{\Phi }\mathrm{ln}(r)`$, implying that the scalar potential is exponential $`V\mathrm{exp}(2\alpha \mathrm{\Phi })`$ such as has been found useful for structure formation scenarios and scaling solutions with a primordial scalar field in the cosmological context including quintessential scenarios . Thus, the particular solution for the system (10 -13) that we are considering is
$`A`$ $`=`$ $`{\displaystyle \frac{4l^2}{4}},`$ (15)
$`\mathrm{\Phi }`$ $`=`$ $`\sqrt{{\displaystyle \frac{l}{\kappa _0}}}\mathrm{ln}(r)+\mathrm{\Phi }_0,`$ (16)
$`V(\mathrm{\Phi })`$ $`=`$ $`{\displaystyle \frac{l}{2l}}\mathrm{exp}[2\sqrt{{\displaystyle \frac{\kappa _0}{l}}}(\mathrm{\Phi }\mathrm{\Phi }_0)].`$ (17)
where (15) and (16) approach asymptotically ($`r\mathrm{}`$) the case with $`m=2,n=2l/(2l)`$ in the general study of the global properties of spherically symmetric solutions in dimensionally reduced space-times . Function $`A`$ corresponds to an exact solution of the Einstein equations of a spherically symmetric space-time, in which the matter contents is a scalar field with an exponential potential. Let us perform the rescalling $`r^24r^2/\left(4l^2\right)`$. In this case the three dimensional space corresponds to a surplus of angle (analogous to the deficit of angle) one; the metric reads
$$ds^2=B_0r^ldt^2+dr^2+\frac{4}{4l^2}r^2\left[d\theta ^2+\mathrm{sin}^2\theta d\phi ^2\right]$$
(18)
for which the two dimensional hypersurface area is $`4\pi r^2\times 4/(4l^2)=4\pi r^2/(1(v^\phi )^4)`$. Observe that if the rotational velocity of the test particles were the speed of light $`v^\phi 1`$, this area would grow very fast. Nevertheless, for a typical galaxy, the rotational velocities are $`v^\phi 10^3`$ ($`300km/s`$), in this case the rate of the difference of this hypersurface area and a flat one is $`(v^\phi )^4/(1(v^\phi )^4)10^{12}`$, which is too small to be measured, but sufficient to give the right behavior of the motion of stars in a galaxy.
Let us consider the components of the scalar field as those of a perfect fluid, it is found that the components of the stress-energy tensor have the following form
$`\rho `$ $`=`$ $`T^0{}_{0}{}^{}={\displaystyle \frac{l^2}{(4l^2)}}{\displaystyle \frac{1}{\kappa _0r^2}}`$ (19)
$`P`$ $`=`$ $`T^r{}_{r}{}^{}={\displaystyle \frac{l(l+4)}{(4l^2)}}{\displaystyle \frac{1}{\kappa _0r^2}}`$ (20)
while the angular pressures are $`P_\theta =P_\phi =\rho `$. The analysis of an axially symmetric perfect fluid in general is given in , where a similar result was found (see also ).
The effective density (19) depends on the velocities of the stars in the galaxy, $`\rho =(v^\phi )^4/(1(v^\phi )^4)\times 1/(\kappa _0r^2)`$ which for the typical velocities in a galaxy is $`\rho 10^{12}\times 1/(\kappa _0r^2)`$, while the effective radial pressure is $`P=(v^\phi )^2((v^\phi )^2+2)/(1(v^\phi )^2)\times 1/(\kappa _0r^2)10^6\times 1/(\kappa _0r^2)`$, $`i.e.`$, six orders of magnitude greater than the scalar field density. This is the reason why it is not possible to understand a galaxy with Newtonian dynamics. Newton theory is the limit of the Einstein theory for weak fields, small velocities but also for small pressures (in comparison with densities). A galaxy fulfills the first two conditions, but it has pressures six orders of magnitude bigger than the dark matter density, which is the dominating density in a galaxy. This effective pressure is the responsible for the behavior of the flat rotation curves in the dark matter dominated part of the galaxies.
Metric (18) is not asymptotically flat, it could not be so. An asymptotically flat metric behaves necessarily like a Newtonian potential provoking that the velocity profile somewhere decays, which is not the observed case in galaxies. Nevertheless, the energy density in the halo of the galaxy decays as
$$\rho \frac{10^{12}}{\kappa _0r^2}=\frac{10^{12}H_0^2}{3r^2}\rho _{crit}$$
(21)
where $`H_0^1=\sqrt{3}/h10^6Kpc`$ is the Hubble parameter and $`\rho _{crit}`$ is the critical density of the Universe. This means that after a relative small distance $`r_{crit}\sqrt{3/h^2}3Kpc`$ the effective density of the halo is similar as the critical density of the Universe. One expects, of course, that the matter density around a galaxy is smaller than the critical density , say $`\rho _{around}0.06\rho _{crit}`$, then $`r_{crit}14Kpc`$. Observe also that metric (18) has an almost flat three dimensional space-like hypersurface. The difference between a flat three dimensional hypersurface area and the three dimensional hypersurface area of metric (18) is $`10^{12}`$ , this is the reason why the space-time of a galaxies seems to be so flat. We think that these results show that it is possible that the scalar field could be the missing matter (the dark matter) of galaxies and maybe of the Universe.
Possibly the greatest problem with the present model is the physical origin of the exponential potential (17). Firstly, its sign is necessarily opposite to that of the exponential potentials that have been considered in quintessence cosmologies . Secondly, although exponential scalar potentials with an overall negative sign do arise from dimensional reduction of higher-dimensional gravity with the extra dimensions forming a compact Einstein space of dimension $`n2`$ , such models also constrain the parameter $`l=2n/(n+2)`$ to take values $`l1`$, which are inconsistent with its interpretation as $`l=2\left(v^\phi \right)^2`$ for velocities $`v^\phi `$ of the order of magnitude of the rotation velocity of galaxies. Similar considerations apply to the exponent of a single exponential potential obtained by the dimensional reduction of a theory with a higher–dimensional cosmological constant and Ricci–flat internal space . Nonetheless, exponential potentials can arise in a variety of ways in stringy gravity, possibly via symmetry breaking or similar mechanisms, and so we are hopeful that a natural origin can be found for potentials of the type considered here.
###### Acknowledgements.
We want to thank L. Arturo Ureña-López, Michael Reisenberger, Daniel Sudarsky and Ulises Nucamendi for many helpful discussions. We also want to express our acknowledgment to the relativity group in Jena for its kind hospitality. This work is also partially supported by CONACyT México, by grants 94890 (Guzmán) and by the DGAPA-UNAM IN121298 (Núñez) and by a cooperations grant DFG-CONACyT.
|
warning/0003/cond-mat0003257.html
|
ar5iv
|
text
|
# The relaxation dynamics of a simple glass former confined in a pore
## Abstract
We use molecular dynamics computer simulations to investigate the relaxation dynamics of a binary Lennard-Jones liquid confined in a narrow pore. We find that the average dynamics is strongly influenced by the confinement in that time correlation functions are much more stretched than in the bulk. By investigating the dynamics of the particles as a function of their distance from the wall, we can show that this stretching is due to a strong dependence of the relaxation time on this distance, i.e. that the dynamics is spatially very heterogeneous. In particular we find that the typical relaxation time of the particles close to the wall is orders of magnitude larger than the one of particles in the center of the pore.
Despite the remarkable progress that has been made in recent years in our understanding of the dynamics of supercooled liquids there are still situations in which the nature of this dynamics is very unclear. One example is the behavior of supercooled liquids in restrictive geometries, such as in thin films or in narrow pores . The motivation to study the dynamics of fluids in such geometries lies in their importance as catalysts, flow through porous materials, or capillaries in biological systems, to name a few. Apart from their relevance in various applications, this dynamics is also of considerable interest for basic science since it might help us to gain a better understanding of the dynamics of supercooled liquids and the glass transition in the bulk, a subject which is, despite the progress mentioned above, still a field of very active research. In particular this type of study might be used to investigate the existence of a diverging length scale in supercooled systems. E.g. investigations of systems that show a conventional ordering phenomena with a growing characteristic length scale, i.e. phase transitions, have shown that, by perturbing the system by means of a wall and by monitoring over which distance this perturbation vanishes, the size of the length scale can be determined. Thus it should be possible to use this approach also in the case of supercooled liquids .
Since at low temperatures the modes leading to the relaxation of the system are generally assumed to be very cooperative, one might argue that the confinement will supress such modes and therefore lead to a dynamics which is slower than the one of the bulk. On the other hand one could also think that the interaction between the liquid and the wall of the host material leads to an acceleration of the dynamics, at least close to the walls. In order to decide what is actually happening many experimentalists started to study such systems. Using new nano- and meso scale materials with adjustable pore size, such as Vycor glass and zeolites, and various techniques, such as differential scanning calorimetry, dielectric relaxation measurements, and solvation dynamics, the relaxation dynamics was investigated as a function of pore size, type of liquid and host material, temperature, etc. . Surprisingly the obtained results do not give a clear answer at all because for certain materials the glass transition temperature $`T_g`$ in the confined system is found to be higher than the one of the bulk system , whereas in other systems it seems to be lower . Sometimes it is also observed that the confinement does not affect the value of $`T_g`$ at all . However, one has to bear in mind that one of the main problems of these experiments is the proper interpretation of the observed results. Although the topology of the pores can presently be controlled reasonably well , the structure of a single pore is never determined exactly. Furthermore it is experimentally not really possible to distinguish between the influence of the fluid-wall interaction, which can be the dominant effect , on the one hand from pure confinement effects on the other hand. Therefore it has so far not been possible to obtain a clear picture of the relaxation dynamics in confined systems.
Also computer simulations have been done in order to gain some insight into the dynamics of liquids in restrictive geometries . Since in simulations it is easy to control the details of the geometry (size, topology) and also the interaction between confined fluid and host material, this method is well suited to investigate this dynamics. In addition it is possible to measure directly all quantities of interest, since at any time the positions and velocities of the particles are known, an advantage of this method which has been made use of extensively in simulations of the bulk . In the past most simulations with restrictive geometries have been done for planar geometries , whereas curved geometries, i.e. tubes or cavities, have been considered only seldom . In the present paper we therefore report the results of a large scale molecular dynamics computer simulation in which we investigated the dynamics of a simple glass former in a narrow pore. In particular we will focus on the dependence of the relaxation dynamics on temperature and the distance from the confining wall.
The system considered is a binary mixture of 80% A and 20% B particles that have the same mass $`m`$ and which interact via a Lennard-Jones potential of the form $`V_{\alpha \beta }(r)=4ϵ_{\alpha \beta }[(\sigma _{\alpha \beta }/r)^{12}(\sigma _{\alpha \beta }/r)^6]`$, with $`\alpha ,\beta \{\mathrm{A},\mathrm{B}\}`$ and cut-off radii $`r_{\alpha \beta }^\mathrm{C}`$=$`2.5\sigma _{\alpha \beta }`$. The parameters $`ϵ_{\alpha \beta }`$ and $`\sigma _{\alpha \beta }`$ are $`ϵ_{\mathrm{AA}}=1.0`$, $`\sigma _{\mathrm{AA}}=1.0`$, $`ϵ_{\mathrm{AB}}=1.5`$, $`\sigma _{\mathrm{AB}}=0.8`$, $`ϵ_{\mathrm{BB}}=0.5`$, and $`\sigma _{\mathrm{BB}}=0.88`$. In the following all results will be given in reduced units, i.e. length in units of $`\sigma _{\mathrm{AA}}`$, energy in units of $`ϵ_{\mathrm{AA}}`$ and time in units of $`(m\sigma _{\mathrm{AA}}^2/48ϵ_{\mathrm{AA}})^{1/2}`$. For Argon these units correspond to a length of 3.4Å, an energy of 120K$`k_B`$ and a time of $`310^{13}`$s. The fluid is confined in a cylindrical tube with radius $`\rho _\mathrm{T}=5.0`$ and below we will compare its properties with the ones in the bulk, using results that have been obtained earlier . In agreement with other computer simulations of confined systems , we find that a smooth wall leads to a (in our case strong) layering of the confined liquid, i.e. its structure is very different from the one in the bulk. Since our goal is to investigate the effect of the confinement onto the dynamical properties and not the effect due to a change of structure, a significant change of the structural properties is not acceptable. In order to avoid this latter possibility we chose the wall of the pore to have a liquid-like structure similar to the one of the confined liquid. For this purpose, we equilibrated a large bulk system at a temperature at which it is slightly supercooled, $`T=0.8`$, and extracted a cylinder with radius $`\rho _\mathrm{T}+r_{\mathrm{AA}}^\mathrm{C}`$. Those particles that had a distance $`\rho `$ from the center of the axis of the cylinder larger than $`\rho _\mathrm{T}`$ were subsequently frozen at their position and constituted the wall particles. The particles for which $`\rho <\rho _\mathrm{T}`$ constituted the fluid and were allowed to continue to move according to Newton’s laws using the same Lennard-Jones potential as interaction. With this choice of the wall the static properties of the confined system (density profile, radial distribution function) remain essentially unchanged , and therefore the changes in the dynamic properties are only due to the confinement.
The time evolution of the system was calculated by solving the equations of motion with the velocity form of the Verlet algorithm with a time step of 0.01 at high ($`T1.0`$) and 0.02 at low ($`T0.8`$) temperatures. To improve the statistics of the results we simulated between 8 and 16 independent systems, each containing 1905 fluid particles and about 2300 wall particles. The length of the tube was 20.137 which resulted in an average particle density of 1.2, the same value as used in the earlier simulations of the bulk . The temperatures investigated were $`T=5.0`$, 2.0, 1.0, 0.8, 0.7, 0.6, and 0.55. The equilibration was done by coupling the liquid periodically to a stochastic heat bath for a time which was sufficiently long to equilibrate the system. All data presented here were produced during a micro-canonical run at constant energy and volume. During this run we observed that the particles of the fluid are able to enter the wall to some extent. Therefore we defined a “penetration radius” $`\rho _\mathrm{p}`$ as the distance beyond which it is very unlikely to find a particle of the fluid, i.e. the probability is less than $`10^4`$. We found that at low temperatures, $`T0.7`$, the value of $`\rho _\mathrm{p}`$ is essentially independent of temperature and that for the A and B particles it is $`5.5\pm 0.2`$ and $`6.1\pm 0.2`$, respectively. Below we will see that this penetration radius, a static quantity, is also relevant for the dynamics of the system.
A quantity which in the past has been found to be useful to characterize the dynamics of liquids in the bulk is the so-called self intermediate scattering function $`F_\mathrm{s}(𝐪,t)`$ for wave-vector $`𝐪`$, which is related to a density fluctuation $`\rho _\mathrm{s}(𝐪,t)=\mathrm{exp}(i𝐪𝐫_j(t))`$ by $`F_\mathrm{s}(𝐪,t)=\rho _\mathrm{s}(𝐪,t)\rho _\mathrm{s}(𝐪,0)`$. (Here $`𝐫_j(t)`$ is the position of particle $`j`$ and $`.`$ is the canonical average.) In Fig. 1 we show the time dependence of $`F_\mathrm{s}(𝐪,t)`$ for all temperatures investigated. The wave-vector is along the axis of the tube and has magnitude $`q=7.18`$, which is the location of the main peak in the static structure factor $`S(q)`$ . As it was the case in the bulk , the correlation functions show that with decreasing temperature the dynamics of the system slows down rapidly and relaxes at low temperatures in two steps, an effect that is related to the cage-effect. The details of this slowing down, as well as the shape of the correlation function is, however, very different from the bulk behavior. To see this we have included in the figure also three bulk curves for $`T=5.0`$, 0.55, and 0.466 (bold dashed lines). We recognize that, whereas the typical relaxation time of the function is basically the same at $`T=5.0`$, a decrease of the temperature leads to a slowing down of the dynamics which is much more pronounced in the restricted system than in the bulk system (see the curves for $`T=0.55`$). In addition to this also the shape of the curves depends on the geometry in that even at the high temperature $`T=5.0`$ the two curves differ since the one for the tube shows a small tail at long times ($`t4`$) that is not present in the bulk curve. If the temperature is lowered this tail becomes more pronounced and starts at higher and higher values of $`F_\mathrm{s}(q,t)`$, whereas no such tail is observed in the bulk curves even at the lowest temperatures. Thus the curve for the tube shows a much stronger stretching than the one for the bulk, in agreement with the experimental findings of Refs. .
In order to find the reason for this different dynamical behavior we consider a generalization of the self intermediate scattering function by defining the function $`F_\mathrm{s}(𝐪,\rho ,t)`$ as
$$F_\mathrm{s}(𝐪,\rho ,t)=\mathrm{exp}[i𝐪(𝐫_j(t)𝐫_j(0))]\delta (𝐫_j(0)\rho ).$$
(1)
Thus this correlation function considers only particles that at time zero had a distance $`\rho `$ from the central axis of the tube. By investigating the $`\rho `$ dependence of this correlator it becomes hence possible to understand which particles are moving relatively fast and which ones relatively slow. In Fig. 2 we show the time dependence of $`F_\mathrm{s}(𝐪,\rho ,t)`$ at a low temperature. The direction of the wave-vector is the same as in Fig. 1, i.e. parallel to the axis of the tube. The value $`\rho _i`$ for the distance $`\rho `$ is chosen such that the values $`(\rho _\mathrm{p}\rho _i)^1`$ are spaced equidistantly, where $`\rho _\mathrm{p}`$ is the penetration radius introduced above. (The reason for this choice will become clear below.) Thus the leftmost curve corresponds to $`\rho =0.5`$ and the rightmost one to $`\rho =4.96`$. From the figure we recognize that the dynamics of the particles close to the center of the tube is similar to the one of the bulk (shown in the figure as bold dashed line). However, for particles that are close to the wall the dynamics is slowed down by several orders of magnitude. We also note that although the time scale of the relaxation function depends strongly on $`\rho `$, its shape, which could, e.g., be characterized by the value of the Kohlrausch parameter $`\beta `$, seems to be independent of it within the accuracy of our data and that this shape is similar to the one found in the bulk. Since the relaxation curve of the whole system is the weighted sum over the curves for the different $`\rho `$, the resulting correlator is much more stretched than the one of the bulk, in agreement with the observation from Fig. 1.
In order to investigate this $`\rho `$-dependence of the correlators in more detail we define the $`\alpha `$-relaxation time $`\tau (q,\rho ,T)`$ by the time it takes the correlation function to decay to $`e^1`$ of its initial value (see the horizontal dashed line in Fig. 2). Such a definition is reasonable since the shape of the correlation function is essentially independent of $`\rho `$. In Fig. 3 we plot this $`\rho `$-dependence for the different temperatures (curves with open symbols) . For the sake of comparison we have included in the figure also the relaxation times of the system in the bulk (filled symbols). From this graph we see that at high temperatures the relaxation time is independent of $`\rho `$, if $`\rho `$ is not too large, and is the same as in the bulk. Only upon approach to the wall, $`\rho 4.0`$, $`\tau (q,\rho ,T)`$ starts to increase significantly. With decreasing temperature this crossover distance becomes smaller until at the lowest temperatures the whole system is affected by the slowing down due to the wall. Thus at these temperatures the dynamics of all the particles is slower than the one of the bulk, which can be recognized from the fact that an extrapolation of the curves towards $`\rho =0`$ intercepts the vertical line at $`\rho _\mathrm{p}^1`$ above the value for the bulk.
From the figure we see that the relaxation times seem to show an apparent divergence upon approaching the wall. We have found that at low temperatures this strong increase can be described well by the functional form $`\tau (q,\rho ,T)\mathrm{exp}[\mathrm{\Delta }_q/(\rho _\mathrm{p}\rho )]`$, where $`\rho _\mathrm{p}`$ is the penetration radius introduced above, i.e. it is not a fit parameter. To demonstrate the quality of this fit we plot in Fig. 4 the relaxation time in a logarithmic way as a function of $`(\rho _\mathrm{p}\rho )^1`$ and we see that at low temperatures the resulting fit is indeed very good (bold lines). With increasing temperature we see the expected deviations at small $`\rho `$, since there the relaxation times have to approach the bulk values, i.e. they have to become independent of $`\rho `$. Nevertheless, even at intermediate and moderately high temperatures the fit describes the data well for large values of $`\rho `$. Note that the value of the parameter $`\mathrm{\Delta }_q=7.8\pm 0.2`$ is independent of $`T`$, at least within the accuracy of our data. Thus close to the wall the only effect of a decrease in temperature is an increase of the prefactor of $`\mathrm{exp}[\mathrm{\Delta }_q/(\rho _\mathrm{p}\rho )]`$. We have found that this temperature dependence is independent of the wave-vector $`q`$ or the type of particle considered (A or B). At low temperatures we can therefore write the whole temperature and $`\rho `$-dependence of the relaxation time as
$$\tau (q,\rho ,T)=C(q)f(T)\mathrm{exp}[\mathrm{\Delta }_q/(\rho _\mathrm{p}\rho )],$$
(2)
where the function $`f(T)`$ depends strongly on temperature. We have found that this dependence is compatible with an Arrhenius law, in agreement with the experimental findings of Ref. , but also with a power-law of the form $`(TT_c)^\gamma `$, a functional form suggested by mode-coupling theory . The value of the critical temperature $`T_c`$ is around $`0.39\pm 0.02`$, i.e. is significantly less than the one found in the bulk which is 0.435 . Also the value for the exponent, $`\gamma =4.7\pm 0.5`$, is much larger than the bulk value 2.35 . Whether or not the dynamics of the particles in the confined geometry can be described by mode-coupling theory as well as it is the case in the bulk is the focus of current investigations.
Before we conclude we briefly comment to what extent these results can be generalized to other systems. It might be expected, e.g., that the strong slowing down of the dynamics close to the wall is related to the fact that the particles of the wall do not move at all, i.e. that they are at zero temperature. However, we have repeated the present calculations also for a system in which the particles in the wall behaved like interacting Einstein oscillators and found that the dynamics is qualitatively the same, i.e. also in this case a strong slowing down close to the wall is found. Furthermore we expect that the same type of slowing down is observed also in a slit geometry, since the fact that the wall is curved is probably irrelevant.
In summary we can say that by using a wall which does not affect the static properties of the system we have been able to study the effect of the confinement onto the dynamics of a simple liquid in a very clean way. We find that the presence of the wall leads to a strong slowing down of this dynamics and that the time correlation functions are much stronger stretched in the confined system than in the bulk. By measuring the dynamics of the particles as a function of their distance from the wall, we find that the relaxation times show a very strong increase when this distance is small. Therefore we are able to identify the reason why the relaxation of the whole system is so stretched. Note that this increase is a smooth function of the distance from the wall, i.e. we do not see any evidence that the layer of particles closest to the wall is decoupling dynamically from the rest of the liquid, as it has been suggested before . In the future it will be of interest to see whether this type of simulation is also useful to identify a growing length scale in supercooled liquids. Work in this direction is currently done by Parisi and coworkers .
Acknowledgement: We gratefully acknowledge the financial support by DFG under SFB 262 and the NIC in Jülich.
|
warning/0003/gr-qc0003091.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
In order to rescue the generalized second law (GSL) of thermodynamics, Bekenstein conjectured, some time ago, that there exists an upper bound on the entropy of any neutral object of energy $`E`$ and maximal radius $`R`$ in the form $`S2\pi ER/\mathrm{}`$ . This derivation was criticized by Unruh, Wald and Pelath \[2-4\] for neglecting the effects of buoyancy in acceleration radiation. They concluded that no additional assumption for upper entropy bound is necessary to maintain the GSL. However their criticism was refuted by Bekenstein who showed that the buoyancy can really be negligible and does not spoil the entropy bound derivation . Bekenstein’s entropy bound has received independent support \[7-10\].
Recently, extending the derivation of an upper bound on the entropy to any charged object, Bekenstein and Mayo , Hod and Linet have shown that Bekenstein’s original entropy bound can be improved. A tighter entropy bound for the nonrotating object of mass $`\mu `$, radius $`R`$ and charge $`e`$ is required. It has been shown that such a bound is $`S(2\pi R/\mathrm{})(E^2e^2/2R)`$. This result agrees to an earlier finding by Zaslavskii in another context. However, the fact that this entropy bound for a charged system is necessary to uphold the GSL has been challenged as well .
A tighter bound on entropy for objects with angular momentum has also been derived recently . Refering to Hojman and Hojman’s integrals of motion for a neutral object with spin $`s`$ moving on a Kerr black hole background, Hod obtains the entropy bound $`S{\displaystyle \frac{2\pi ER}{\mathrm{}}}(1s^2/E^2R^2)^{1/2}`$. He claimed that this bound is universal and independent of the black hole parameters which were used to derive it. In order to examine this argument, in this paper we will study the entropy bound for a spinning object falling into Anti de Sitter (AdS) black holes including (3+1)-dimensional Kerr-AdS black holes and (2+1)-dimensional BTZ black holes. We will derive geodesic equations of the motion of the spinning objects around Kerr-AdS black holes and BTZ black holes, respectively. Based upon these geodesic equations, we will show that the entropy bound for a rotating system depends neither on the black hole parameters, nor on spacetimes dimensions. It is a universal result.
## 2 Entropy bound from the Kerr-AdS black holes
Recently the study of the Kerr-AdS black hole model has been undertaken and shown to be important in many aspects\[18-21\]. The metric is
$`\mathrm{d}s^2`$ $`=`$ $`{\displaystyle \frac{\mathrm{\Delta }_r}{\rho ^2}}[\mathrm{d}t{\displaystyle \frac{a}{\mathrm{\Sigma }}}\mathrm{sin}^2\theta \mathrm{d}\varphi ]^2+{\displaystyle \frac{\rho ^2}{\mathrm{\Delta }_r}}\mathrm{d}r^2+{\displaystyle \frac{\rho ^2}{\mathrm{\Delta }_\theta }}\mathrm{d}\theta ^2`$
$`+{\displaystyle \frac{\mathrm{sin}^2\theta \mathrm{\Delta }_\theta }{\rho ^2}}[a\mathrm{d}t{\displaystyle \frac{(r^2+a^2)}{\mathrm{\Sigma }}}\mathrm{d}\varphi ]^2`$
where
$`\rho ^2`$ $`=`$ $`r^2+a^2\mathrm{cos}^2\theta `$ (2)
$`\mathrm{\Delta }_r`$ $`=`$ $`(r^2+a^2)(1+l^2r^2)2Mr`$
$`\mathrm{\Delta }_\theta `$ $`=`$ $`1l^2a^2\mathrm{cos}^2\theta `$
$`\mathrm{\Sigma }`$ $`=`$ $`1l^2a^2`$
The parameter $`M`$ is the mass, $`a`$ the angular momentum per unit mass, and $`l^2=\mathrm{\Lambda }/3`$, where $`\mathrm{\Lambda }`$ is the (negative) cosmological constant. The black hole solution is valid for $`a^2<l^2`$ \[18-21\]. For $`l=0`$, eq.(1) goes back to the metric for Kerr black hole. There are four roots of the polynomial $`\mathrm{\Delta }_r`$, the largest root $`r_+`$ corresponds to the event horizon, the other positive root $`r_{}`$ is the Cauchy horizon and another two roots $`r_1,r_2`$ are negative and satisfy $`r_1+r_2=(r_++r_{}),r_1r_2={\displaystyle \frac{a^2}{l^2r_+r_{}}}`$. The mass $`M`$, and the angular momentum per unit mass $`a`$, can both be expressed in terms of $`r_+,r_{}`$ and $`l`$ as
$`M`$ $`=`$ $`{\displaystyle \frac{(1+l^2r_{}^2)(r_++r_{})(1+l^2r_+^2)}{2(1l^2r_+r_{})}},`$ (3)
$`a`$ $`=`$ $`\sqrt{{\displaystyle \frac{r_+r_{}(1+l^2r_{}^2+l^2r_+r_{}+l^2r_+^2)}{1l^2r_+r_{}}}}.`$ (4)
The above equations require $`l^2<1/r_+r_{}`$ to ensure real and positive values of $`a`$ and $`M`$, respectively. For the extreme black hole case $`r_+`$ and $`r_{}`$ degenerate and $`M=M_e`$, where $`M_e`$ is the critical mass parameter given in .
We consider a spinning object of rest mass $`m`$, intrinsic spin $`s`$ and proper cylindrical radius $`R`$, which is descending into the Kerr-AdS black hole. Following , the constants of motion associated with the $`t`$ and $`\varphi `$ variables are
$`E`$ $`=`$ $`\pi _tg_{t\varphi ,r}\pi _t{\displaystyle \frac{s\mathrm{\Sigma }}{2rm}}+g_{tt,r}\pi _\varphi {\displaystyle \frac{s\mathrm{\Sigma }}{2rm}}`$ (5)
$`J`$ $`=`$ $`\pi _\varphi g_{\varphi t,r}\pi _\varphi {\displaystyle \frac{s\mathrm{\Sigma }}{2rm}}+g_{\varphi \varphi ,r}\pi _t{\displaystyle \frac{s\mathrm{\Sigma }}{2rm}}`$ (6)
where
$`\pi _t=g_{tt}\dot{t}+g_{t\varphi }\dot{\varphi }`$ (7)
$`\pi _\varphi =g_{t\varphi }\dot{t}+g_{\varphi \varphi }\dot{\varphi }`$ (8)
For simplicity we just consider the equatorial motions of the object. The quadratic equation for the conserved energy $`E`$ of the body is
$$\stackrel{~}{\alpha }E^22\stackrel{~}{\beta }E+\stackrel{~}{\gamma }=0$$
(9)
where the very long expressions for $`\stackrel{~}{\alpha },\stackrel{~}{\beta },\stackrel{~}{\gamma }`$ are given in the appendix. It is worth noting that taking $`l0`$, they reproduce the expressions given in .
In the spirit of the analysis of Bekenstein and of Hod, we neglect the buoyancy contribution, for simplicity. Suppose the gradual approach to the black hole must stop when the proper distance from the body’s center of mass to the black hole horizon equals $`R`$, the body’s radius
$$_{r_+}^{r_++\delta (R)}(g_{rr})^{1/2}𝑑r=R,$$
(10)
with $`g_{rr}={\displaystyle \frac{r^2}{\mathrm{\Delta }_r}}`$ (in equatorial plane) and $`\mathrm{\Delta }_r=l^2(rr_+)(rr_{})(rr_1)(rr_2)`$. One can get from integrating Eq.(10)
$$\delta (R)=\frac{l^2(r_+r_{})(r_+r_1)(r_+r_2)R^2}{4r_+^2}.$$
(11)
Considering the relation between $`r_1,r_2`$ and $`r_+,r_{}`$ together with eqs.(3,4), it is not difficult to find that when $`l0`$, eq.(11) reproduces eq.(6) of .
Using the test particle approximation $`s/(mr_+)1,Rr_+`$ together with the condition $`l^2<1/(r_+r_{})`$, we can solve eq.(9) for $`E`$ to first order in the small quantities at the point of capture $`r=r_++\delta (R)`$,
$$E=u+vs+R\sqrt{w}$$
(12)
where
$`u`$ $`=`$ $`{\displaystyle \frac{aJ(a^2l^21)(2Ma^2l^2r_+l^2r_+^3)}{2a^2Ma^2r_++a^4l^2r_+r_+^3+a^2l^2r_+^3}}`$ (13)
$`v`$ $`=`$ $`{\displaystyle \frac{J(a^2l^21)}{mr_+(2a^2M+a^2r_+a^4l^2r_++r_+^3a^2l^2r_+^3)^2}}`$
$`\times (2a^4M6a^2M^2r_++3a^2Mr_+^2+5a^4l^2Mr_+^2a^2r_+^3+2a^4l^2r_+^3`$
$`a^6l^4r_+^3+3Mr_+^4+3a^2l^2Mr_+^4r_+^5+2a^2l^2r_+^5a^4l^4r_+^5)`$
$`w`$ $`=`$ $`{\displaystyle \frac{(r_+r_{})^2(1l^2r_{}^22l^2r_{}r_+3l^2r_+^2+l^4r_{}^2r_+^2+2l^4r_{}r_+^3)^2}{4(1+l^2r_{}^2)^4r_+^2(r_{}+r_+)^4}}`$
$`\times (J^2+m^2r_{}^2+2l^2m^2r_{}^4+l^4m^2r_{}^64J^2l^2r_{}r_++2m^2r_{}r_+2J^2l^4r_{}^3r_+`$
$`+4l^2m^2r_{}^3r_++2l^4m^2r_{}^5r_++m^2r_+^2+2J^2l^4r_{}^2r_+^2+2l^2m^2r_{}^2r_+^2+4J^2l^6r_{}^4r_+^2`$
$`+l^4m^2r_{}^4r_+^2+J^2l^8r_{}^6r_+^22J^2l^4r_{}r_+^3+4J^2l^6r_{}^3r_+^3+2J^2l^8r_{}^5r_+^3+4J^2l^6r_{}^2r_+^4`$
$`+3J^2l^8r_{}^4r_+^4+2J^2l^8r_{}^3r_+^5+J^2l^8r_{}^2r_+^6).`$
Eq.(12) reduces to (7) of after taking the limit $`l0`$ and substitution of Eqs.(3,4).
After the assimilation of the spinning body, the change of the black hole mass and angular momentum are $`dM=E`$ and $`dL=J`$, respectively. Taking cognizance of eq.(12) and using the first-law of black hole thermodynamics,
$$dM=\frac{\kappa }{8\pi }dA+\mathrm{\Omega }dL$$
(16)
where $`\kappa ={\displaystyle \frac{r_+(1+a^2l^2+3l^2r_+^2a^2/r_+^2)}{2(r_+^2+a^2)}}`$ and $`\mathrm{\Omega }={\displaystyle \frac{a(1l^2a^2)}{r_+^2+a^2}}`$ are the surface gravity and rotational angular frequency of the black hole respectively, we find
$$dA=\frac{8\pi }{\kappa }(u+vs+R\sqrt{w}\mathrm{\Omega }J)$$
(17)
Substituting eqs.(3,4), it is easy to see that $`u\mathrm{\Omega }J=0`$. Carefully choosing the total angular momentum of the body at the critical value
$$J=J^{}=\sqrt{\frac{m^2(1+l^2r_{}^2)^2(r_++r_{})^2s^2}{(1+2l^2r_+r_{}+l^4r_{}^3r_++l^4r_{}^2r_+^2+l^4r_{}r_+^3)^2(m^2R^2s^2)}},$$
the minimum value of the increase in the black hole surface area caused by an assimilation of a spinning body with given parameters $`m,s,R`$ is
$$dA_{min}=8\pi \sqrt{m^2R^2s^2}$$
(18)
The minimum exists only for $`smR`$.
By virtue of the GSL, we derived an upper bound to the entropy $`S`$ of an arbitrary system of proper energy $`E`$, intrinsic angular momentum $`s`$ and proper radius $`R`$ falling into the Kerr-AdS black hole
$$S2\pi \sqrt{(RE)^2s^2}$$
(19)
It is evident that the entropy $`S`$ of the rotating system should be bounded and this bound is more stringent than the original Bekenstein bound . It is worth noting that although we used a different black hole model to derive the entropy bound, the final result is the same as in and independent of the black hole parameters which are used to derive it.
## 3 Entropy bound from the BTZ black hole
The entropy bound for the rotating system derived from the (3+1)-dimensional Kerr-AdS black hole is the same as that from Kerr black hole , which shows that this bound does not depend on the black hole parameters. Now it is of interest to ask the question whether this bound only exists for (3+1)-dimensional cases and whether it will be changed if it is derived from a different dimensional black hole. In this section we will concentrate our attention on the (2+1)-dimensional BTZ black holes . The metric of this black hole reads
$$\mathrm{d}s^2=N^2\mathrm{d}t^2+N^2\mathrm{d}r^2+r^2(N^\varphi \mathrm{d}t+\mathrm{d}\varphi )^2$$
(20)
where the squared lapse $`N^2(r)`$ and the angular shift $`N^\varphi (r)`$ are
$`N^2`$ $`=`$ $`M+{\displaystyle \frac{r^2}{l^2}}+{\displaystyle \frac{J^2}{4r^2}}`$ (21)
$`N^\varphi `$ $`=`$ $`{\displaystyle \frac{J}{2r^2}}`$
where $`\mathrm{}<t<+\mathrm{},0<r<\mathrm{}`$, and $`0\varphi 2\pi `$. $`M,J`$ appearing in (22) are two constants of integration which can be interpreted as the black hole mass and angular momentum. The lapse function $`N^2(r)`$ vanishes when
$$r_\pm =l\{\frac{M}{2}(1\pm [1(\frac{J}{Ml})^2]^{1/2})\}^{1/2}$$
(22)
Here $`r_+`$ is the event horizon and $`r_{}`$ is the Cauchy horizon for $`M>0`$ and $`|J|<Ml`$. In the extreme case $`|J|=Ml`$, $`r_+`$ and $`r_{}`$ degenerate.
We proceed to devise a gedanken experiment in which a spinning object of mass $`m`$, intrinsic spin $`s`$ and proper cylindrical radius $`R`$ is decending into the BTZ black hole. Using constants of motion associated with the $`t`$ and $`\varphi `$ definition
$`E`$ $`=`$ $`\pi _tg_{t\varphi ,r}\pi _t{\displaystyle \frac{s}{2mr}}+g_{tt,r}\pi _\varphi {\displaystyle \frac{s}{2mr}}`$ (23)
$`L`$ $`=`$ $`\pi _\varphi g_{\varphi t,r}\pi _\varphi {\displaystyle \frac{s}{2mr}}+g_{\varphi \varphi ,r}\pi _t{\displaystyle \frac{s}{2mr}}`$ (24)
We obtained the quadratic equation for the conserved energy $`E`$ of the spinning body in the form of (9), here
$`\alpha `$ $`=`$ $`r^2Js/m+(Mr^2/l^2)s^2/m^2`$ (25)
$`\beta `$ $`=`$ $`JL/2LMs/m+JLs^2/(2l^2m^2)`$ (26)
$`\gamma `$ $`=`$ $`L^2MJ^2m^2/4L^2r^2/l^2+Mm^2r^2m^2r^4/l^2JL^2s/(l^2m)`$
$`+[J^2/2l^22Mr^2/l^2+L^2r^2/(l^4m^2)+2r^4/l^4]s^2(J^2/4Mr^2+r^2/l^2)s^4/(l^4m^2)`$
Taking $`s0`$ and adopting re-scalings given in , the quadratic equation reproduces the special case given in .
Neglecting the buoyancy contribution, we suppose the gradual approach of the spinning object to the black hole must stop when the proper distance from the body’s center of mass to the black hole horizon equals $`R`$. Considering $`N^2={\displaystyle \frac{(r^2r_+^2)(r^2r_{}^2)}{l^2r^2}}`$, from Eq.(10), we have
$$\delta (R)=\frac{(r_+^2r_{}^2)R^2}{2r_+l^2}.$$
(28)
Using conditions of approximation for test particle $`s/(mr_+)1,Rr_+`$, we can get the energy expression in the same form as (12), where $`u,v,w`$ here are
$`u`$ $`=`$ $`{\displaystyle \frac{JL}{2r_+^2}}`$ (29)
$`v`$ $`=`$ $`{\displaystyle \frac{J^2L}{2mr_+^4}}{\displaystyle \frac{LM}{mr_+^2}}`$ (30)
$`w`$ $`=`$ $`{\displaystyle \frac{(r_+r_{})^2(r_++r_{})^2(L^2+m^2r_+^2)}{l^4r_+^4}}.`$ (31)
After the infall of the spinning object, the change of the black hole mass and angular momentum are $`dM=E`$ and $`dL=L`$, respectively. Employing the first-law of black hole thermodynamics Eq. (16), where the surface gravity and angular velocity here are $`\kappa ={\displaystyle \frac{\sqrt{M^2J^2/l^2}}{r_+}}={\displaystyle \frac{r_+^2r_{}^2}{l^2r_+}},\mathrm{\Omega }={\displaystyle \frac{J}{2r_+^2}}`$. When the total angular momentum attains the critical value
$$L^{}=\frac{mr_+s}{\sqrt{m^2R^2s^2}},$$
(32)
there is a minimum increase in the black hole surface area caused by an assimilation of the spinning body in the form
$$dA_{min}=8\pi \sqrt{m^2R^2s^2}$$
(33)
This minimum exists only for $`smR`$.
Arguing from the GSL, we derive an upper bound to the entropy $`S`$ of an arbitrary system of proper energy $`E`$, intrinsic angular momentum $`s`$ and proper radius $`R`$ from the BTZ black hole
$$S2\pi \sqrt{(RE)^2s^2}$$
(34)
This result is the same as obtained from the (3+1)-dimensional cases.
## 4 Conclusions and discussions
Using the method proposed by Hojman and Hojman , we have derived the geodesic equations for the spinning object moving around the (3+1)-dimensional Kerr-AdS black holes and (2+1)-dimensional BTZ black holes, respectively. These geodesic equations are general compared to those obtained in the models without cosmological constant in for Kerr black hole case and without considering the spinning of the object for BTZ black hole . Based upon these geodesic equations, we derived the entropy bound for the rotating system to maintain the GSL. These results coincide with that obtained from Kerr black hole , which supports Hod’s argument that the entropy bound of the rotating sysytem is independent of the black hole parameters. Besides it is worth noting that our result from the three-dimensional BTZ black holes indicates that this entropy bound is also independent of the dimensions of spacetimes. Therefore the entropy bound for the rotating system is universal.
However at the first sight, the universality of the entropy bound for the rotating system cannot be extended to a charged system. Because at least the electric potential will change the form when we study it for the (2+1)-dimensional case. The exact dependence of the entropy bound on the system’s parameters for the charged system still needs further exploration.
ACKNOWLEDGEMENT: This work was partically supported by Fundacão de Amparo à Pesquisa do Estado de São Paulo (FAPESP) and Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPQ). B. Wang would also like to acknowledge the support given by Shanghai Science and Technology Commission.
## 5 Appendix
Here we give the expressions of $`\stackrel{~}{\alpha },\stackrel{~}{\beta },\stackrel{~}{\gamma }`$ in eq.(9) in the main text.
$`\stackrel{~}{\alpha }`$ $`=`$ $`\alpha +k_1l^2+k_2l^4`$ (35)
$`\stackrel{~}{\beta }`$ $`=`$ $`\beta +p_1l^2+p_2l^4+p_3l^6`$
$`\stackrel{~}{\gamma }`$ $`=`$ $`\gamma +q_1l^2+q_2l^4+q_3l^6+q_4l^8`$
where
$`k_1`$ $`=`$ $`a^4a^2r^2+(a^2+2a^2M/rr^2)s^2/m^2`$ (36)
$`k_2`$ $`=`$ $`a^2r^2s^2/m^2`$
$`p_1`$ $`=`$ $`a^3J2a^3JM/raJr^2+(2a^2J+2a^4JM/r^3+3a^2JM/r)s/m`$ (37)
$`+(aJa^3JM^2/r^4a^3JM/r^3+2aJM/r)s^2/m^2`$
$`p_2`$ $`=`$ $`a^5J+a^3Jr^2+a^4Js/m+(a^3J2a^3JM/r+aJr^2)s^2/m^2`$
$`p_3`$ $`=`$ $`a^3Jr^2s^2/m^2`$
$`q_1`$ $`=`$ $`a^2J^24a^2J^2M/rJ^2r^2a^2m^2r^2m^2r^4+(2aJ^2+4a^3J^2M/r^3)s/m`$
$`+[2a^22a^2J^2M^2/(m^2r^4)+2a^2M/r+2J^2M/(m^2r)2Mr+2r^2]s^2`$
$`+[a^2M^2/r^42a^2M/r^3+3M^2/r^22M/r]s^4/m^2`$
$`q_2`$ $`=`$ $`a^4J^2+2a^4J^2M/r+2a^2J^2r^2+(4a^3J^22a^5J^2M/r^3)s/m`$
$`+[a^4J^2M^2/(m^2r^4)4a^2J^2M/(m^2r)+2a^2r^2+J^2r^2/m^2+2r^4]s^2`$
$`+(a^22a^2M/rr^2)s^4/m^2`$
$`q_3`$ $`=`$ $`a^6J^2a^4J^2r^22a^5J^2s/m`$
$`+(2a^4J^2M/r2a^2J^2r^2)s^2/m^2+(a^2r^2r^4)s^4/m^2`$
$`q_4`$ $`=`$ $`a^4J^2r^2s^2/m^2`$
$`\alpha ,\beta `$ have the same expressions given in . There is a sign mistake for the expression $`\gamma `$ in , where the last term should be negative. The correct formula is
$$\gamma =e^2\varphi ^2k_1+2e\varphi (jeh)k_3(jeh)^2k_2\delta ^2\mathrm{\Delta }M^2$$
so that the requirement
$$\beta ^2\alpha \gamma =\mathrm{\Delta }\delta ^2[(jeh)^2k_1M^2]$$
can be satisfied.
|
warning/0003/hep-th0003197.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
Soliton solutions are a central theme in the study of supersymmetric theories. In particular domain wall solitons in five-dimensions have received substantial attention because of their potential to admit chiral fermion zero modes. In this paper we wish to describe some observations about chiral fermion zero modes of domain walls viewed as stable finite energy solutions which interpolate between two vacua of the theory. Indeed we will argue, at least in a wide class of supersymmetric theories in odd dimensions, that no smooth domain walls exist.
The analysis given below was initiated by the question as to whether or not a Randall-Sundrum scenario can be extended to a smooth domain wall in a supergravity theory. This question has several motivations. It was pointed out in that such an embedding would solve the fine-tuning problem associated with matching the domain wall tension and bulk cosmological constant needed in . Indeed without supersymmetry one is led to question the general stability of a domain wall . In addition, with a smooth domain wall solution one can improve upon the thin wall approximation in and provide a complete non-linear analysis of the Randall-Sundrum scenario . Finally there is widespread belief that supersymmetry and supergravity are relevant phenomenologically and in this context it is natural to embed our universe in a higher dimensional theory containing supergravity. Certainly from a theoretical point of view one would like to place such a “brane-world” in the context of supergravity and ultimately string theory. The difficulty in obtaining a smooth Randall-Sundrum domain wall in five-dimensional supergravity has been discussed recently and a no-go theorem can be proven in various cases <sup>1</sup><sup>1</sup>1Recently the original but discontinuous Randall-Sundrum domain wall has been embedded into a supergravity .. In this paper we argue against the existence of smooth domain walls interpolating between supersymmetric vacua on rather general grounds in a wide class of odd-dimensional supergravities (although not all, e.g. see ).
Supersymmetric domain wall spacetimes have also received interest recently due to their role in the AdS/CFT correspondence. In particular the domain wall central charge has been identified with the c-function of a four-dimensional field theory . From this perspective the absence of smooth Randall-Sundrum domain walls in a particular supergravity is interpreted as the statement that the (monotonic) c-function
$$C(r)=\frac{C_0}{|W(r)|^{D2}},$$
(1)
is bounded along the renormalisation group flow (i.e. that $`W(r)`$ does not pass through zero).
The obstruction to finding supergravity domain walls in five dimensions seems to be obtaining solutions where the real superpotential $`W`$ changes sign . In supergravity theories $`W`$ appears in the mass terms for the fermions. Domain walls that interpolate between regions in which $`W`$ changes sign connect regions with positive fermion mass to those with negative fermion mass. In even dimensions the sign of a fermion mass term has no physical significance; it may be reversed by multiplying the fermion field by $`\mathrm{\Gamma }^{D+1}`$. By contrast, in odd dimensions, $`\mathrm{\Gamma }^{D+1}=\pm 1`$, and a fermion mass term breaks parity. Furthermore in odd dimensions there are two inequivalent irreducible representations of the Clifford algebra labelled by the sign of $`\mathrm{\Gamma }^{D+1}`$. Theories with different signs for the $`\mathrm{\Gamma }`$-matrices are rather like different superselection sectors and one would not expect that these two sectors could be realised in single connected spacetime. However a change in sign of all fermion masses may be effected by a change in sign of the $`\mathrm{\Gamma }`$-matrices. Thus a domain wall in an odd-dimensional theory in which $`W`$ changes sign looks as if it connects two distinct superselection sectors and one might doubt that this is physically sensible. Perhaps this is the reason why domain walls with $`W`$ changing sign have not been found in supergravity. A similar reservation was raised in .
This also raises the question of whether domain walls coupled to fermions in which mass terms change sign can exist in flat space theories, supersymmetric or not. As is well-known such domain walls admit localised zero energy fermion modes which are chiral. As long as the worldvolume theory is not anomalous, this would seem to lead to no contradiction. Of course if it were anomalous, there would have to be some in-flow to balance the anomaly and such domain walls might be incompatible with being supersymmetric, i.e. BPS.
Thus it seems that the key to understanding the absence of supersymmetric domain wall solutions lies in understanding the Goldstino fermions. Therefore the rest of this paper is organised as follows. In section two we shall discuss non-gravitational domain walls. We find that even if four-dimensional supersymmetric domain walls exist their Goldstinos are non-chiral and hence non-anomalous. Finally in section three we discuss domain walls in supergravity. There we find that the Goldstino fermion modes would diverge if $`W`$ changed sign.
## 2 Non-Gravitational Domain Walls
### 2.1 Fermion Zero-Modes
Let us consider a general theory which includes a fermionic field $`\psi `$. Around any vacuum of this theory we may consider the fluctuations of the fermion which we assume satisfy the Dirac equation (we use a “mostly” plus metric in $`D`$ spacetime dimensions, $`m,n=0,1,2,\mathrm{},D1`$)
$$\mathrm{\Gamma }^m_m\psi +M\psi =0.$$
(2)
As is well known this equation admits both positive and negative energy solutions $`\psi ^{(\pm )}`$. In particular, particles at rest have one-particle wave functions given by “plane-wave” solutions $`\psi ^{(\pm )}=e^{i|M|t}\eta _\pm `$ where $`\eta _\pm `$ is a constant spinor and $`i\mathrm{\Gamma }^0\eta _\pm =\pm sign(M)\eta _\pm `$. The resolution of this “energy crisis” in the quantum theory is to simply assert that in a given vacuum all the negative energy states $`\psi ^{()}`$ are filled.
Now imagine that there is a domain solution associated with a scalar field $`\varphi (r)`$ which interpolates between two vacua, where $`r`$ is the coordinate transverse to the domain wall. As a consequence the fermion mass $`M(r)`$ becomes dependent upon $`r`$. We must now look for solutions of the form
$$\psi =e^{ip_\mu x^\mu }\chi (r),$$
(3)
where $`\mu =0,1,2,\mathrm{},D2`$. There are two cases to consider. In the first case $`\mathrm{\Gamma }^\mu p_\mu \chi =0`$ so that $`p_\mu p^\mu =0`$. The solution then takes the form
$$\chi (r)=\chi _\pm e^{ip_\mu x^\mu }exp\left(_0^rM(r^{})𝑑r^{}\right),$$
(4)
where $`\mathrm{\Gamma }^r\chi _\pm =\pm \chi _\pm `$. If the spacetime dimension is odd, $`\mathrm{\Gamma }^0\mathrm{\Gamma }^1\mathrm{}\mathrm{\Gamma }^r=\pm 1`$ and hence $`\mathrm{\Gamma }^r`$ determines the chirality of the fermions with respect to the wall. If $`M(r)`$ changes sign then one chiral mode in (4) will be normalisable and the other non-normalisable. This implies that a massless chiral fermion is bound to the domain wall. If $`M(r)`$ does not change sign then neither mode is normalisable. We will not be interested in this situation in this paper.
In the case that $`\mathrm{\Gamma }^\mu p_\mu \chi 0`$ we may, without loss of generality, take $`p_\mu =(E,0,\mathrm{},0)`$. Using the $`\mathrm{\Gamma }`$-matrix algebra we have $`i\mathrm{\Gamma }^0\chi _\pm =\chi _{}`$ leading to the following system of equations
$`(_r+M)\chi _+`$ $`=`$ $`E\chi _{},`$
$`(_rM)\chi _{}`$ $`=`$ $`E\chi _+.`$
Elimination leads to a second order differential equation in which $`\chi _\pm `$ decouple
$$\left(_r^2+V_\pm (r)\right)\chi _\pm =E^2\chi _\pm ,V_\pm (r)=M^2(r)_rM(r).$$
(6)
For an explicit example we let $`M(r)=m\mathrm{tanh}(mr)`$. We can solve exactly these equations since $`V_+=m^2`$ and hence
$$\chi _+=Ae^{i\sqrt{E^2m^2}r}+Be^{i\sqrt{E^2m^2}r},$$
(7)
where $`A`$ and $`B`$ are arbitrary constants. These represent plane wave solutions for $`E^2m^2`$ but are non-normalisable if $`E^2<m^2`$. We may then obtain
$$\chi _{}=E^1(_r+m\mathrm{tanh}(mr))\chi _+.$$
(8)
This completes the spectrum of fermion modes in the domain wall background. In summary, if $`M(r)`$ changes sign then there is a single chiral fermion with zero energy localised on the wall. In addition there are modes with non-vanishing energy. Note that we may perform a boost to transform these modes into their rest frame where $`E=m`$ so that the fermion wave function is
$$\psi =e^{imt}(\chi _+^0+\mathrm{tanh}(mr)\chi _{}^0),$$
(9)
which smoothly interpolates between the constant eigenstates $`\eta _\pm ^0=\chi _+^0\pm \chi _{}^0`$ of $`i\mathrm{\Gamma }^0`$ that appear in the Dirac sea in each vacuum.
This example shows that one can have domain walls in odd dimensions which interpolate between two vacua with opposite signs for the fermion mass. Thus if the domain wall has a $`𝐙_2`$ symmetry then globally parity is a good symmetry, even though it is broken in each vacua. We also see that fermions with positive energy can travel freely between the two vacua. In particular there is no mixing of positive and negative frequencies which would indicate a quantum instability.
### 2.2 Chiral Goldstinos
Let us now consider supersymmetric domain walls in five dimensions where the fermion zero modes arise as Goldstinos. First we note that the mass term in (2) comes from a term of the form $`M(\varphi )\overline{\psi }\psi `$ in the Lagrangian. For a supersymmetric theory in five dimensions, the scalar $`\varphi `$ belongs in a vector multiplet, tensor multiplet or a hyper multiplet. For a vector multiplet a coupling of the form $`\varphi \overline{\psi }\psi `$ might seem natural, since this comes from the dimensional reduction of a covariant derivative term for $`\psi `$ in six dimensions. However in this case there can be no potential for $`\varphi `$ by gauge invariance and hence no domain wall. In fact in general, if a five-dimensional multiplet with eight supercharges comes via compactification from six dimensions, then it must come from a chiral multiplet. Therefore there are no fermion mass terms possible (unless we consider theories with sixteen supercharges in which case there are no scalar potentials possible).
In fact regardless of which kind of multiplet $`\varphi `$ belongs in five dimensions, when solving the domain wall equations we effectively compactify the system to the two dimensions $`t`$ and $`r`$. The action will then have $`(4,4)`$ supersymmetry and this constrains the potential to take the form $`V=g_{ij}k^ik^j`$ where $`g_{ij}(\varphi )`$ is a hyper-Kähler metric appearing in the scalar kinetic term and $`k^i(\varphi )`$ is a tri-holomorphic Killing vector. The Yukawa term is $`_ik_j(\varphi )\overline{\psi }^i\mathrm{\Gamma }^{0r}\psi ^j`$ . Let us again write $`\mathrm{\Gamma }^r\psi _\pm ^i=\pm \psi _\pm ^i`$, $`i\mathrm{\Gamma }^0\psi _\pm ^i=\psi _{}^i`$ and $`\psi ^i=e^{iEt}\chi ^i`$. In this case the equations of motion for the fermions become
$`_r\chi _+^iM_j^i\chi _{}^j`$ $`=`$ $`E\chi _{}^i,`$
$`_r\chi _{}^iM_j^i\chi _+^j`$ $`=`$ $`E\chi _+^i,`$
where $`M_j^i=^ik_j`$, $`_r\psi _\pm ^i=_r\psi _\pm ^i+\mathrm{\Gamma }_{jk}^i_r\varphi ^j\psi _\pm ^k`$ and we have ignored a term cubic in the fermions involving the curvature of the metric $`g_{ij}`$.
The system (LABEL:eqtwo) is quite different to (LABEL:coupledeq) because the left hand side contains both $`\chi _{}^i`$ and $`\chi _+^i`$. In particular if $`E=0`$, $`\chi _+^i`$ and $`\chi _{}^i`$ do not decouple. Moreover if $`E=0`$, $`\chi _+^i`$ and $`\chi _{}^i`$ satisfy the same equation and hence the fermion zero modes come in pairs containing both chiralities. Note that this result no longer holds in lower dimensions since the resulting two-dimensional transverse theories need only have $`(1,1)`$ or $`(2,2)`$ supersymmetry and equations of motion of the form $`(\text{LABEL:coupledeq})`$ are possible . We also observe that this form of the Yukawa term does not break parity.
## 3 Domain Walls in Supergravity
We now wish to discuss domain wall solitons in supergravity. First let us review some basic features of supergravity domain walls. We assume that the bosonic action takes the form
$$S=d^Dx\sqrt{g}\left(R\gamma _{AB}(\varphi )_m\varphi ^A^m\varphi ^BV(\varphi )\right),$$
(11)
where $`\varphi ^A`$, $`A=1,2,3,\mathrm{},N`$ are scalar modes and we assume that metric $`\gamma _{AB}`$ appearing in their kinetic term is positive definite. We further assume that (11) is the consistent truncation of a supergravity theory which is invariant under supersymmetry transformations of the form
$`\delta \psi _m`$ $`=`$ $`(_mϵ+W\mathrm{\Gamma }_mϵ+_mW_2ϵ),`$
$`\delta \lambda _A`$ $`=`$ $`({\displaystyle \frac{1}{2}}\gamma _{AB}\mathrm{\Gamma }^m_m\varphi ^B+W_{3A})ϵ.`$
Here $`W,W_2`$ and $`W_{3A}`$ are functions of the scalars $`\varphi ^A`$ which we will avoid specifying in order to keep our argument as general as possible. In fact we can remove the term in (LABEL:susy) involving $`W_2`$ by performing the field redefinitions $`ϵe^{W_2}ϵ`$, $`\psi _me^{W_2}\psi _m`$ and $`\lambda _Ae^{W_2}\lambda _A`$. Therefore, without loss of generality, we set $`W_2=0`$. We have also assumed that any internal indices on the spinors $`ϵ`$ may be ignored. This form for the supersymmetry transformation is quite general for $`N=2`$ supergravity in five dimensions but does not include all extended supergravities (e.g. see ). We will also ignore any higher order fermion terms since it is clear that their inclusion would not affect our discussion.
Let us now look for a supersymmetric domain wall. Without loss of generality we may choose the spacetime to have the metric
$$ds^2=dr^2+e^{2A(r)}\eta _{\mu \nu }dx^\mu dx^\nu ,$$
(13)
where $`\mu ,\nu =0,1,2,\mathrm{},D2`$ and the scalars depend only on $`r`$. The requirement that some supersymmetry is preserved gives rise to the Bogomoln’yi equations
$`A^{}=2W,`$
$`\varphi _{}^{A}{}_{}{}^{}=\pm 2\gamma ^{AB}W_{3B},`$
(14)
where a prime denotes differentiation with respect to $`r`$. The preserved supersymmetries (i.e. Killing Spinors) for these domain walls are
$$ϵ=e^{\frac{1}{2}A}ϵ_\pm ,$$
(15)
where $`\mathrm{\Gamma }_{\underset{¯}{r}}ϵ_\pm =\pm ϵ_\pm `$ and an underlined index refers to the tangent frame.
It is instructive to consider supersymmetric vacua of this theory. Here we set all the scalars to constants $`\varphi ^A=\varphi _0^A`$. Clearly this can only occur at the “critical” points where $`W_{3A}(\varphi _0^A)=V/\varphi ^A=0`$. The spacetime (13) is now just pure AdS space with $`A=2W(\varphi _0^A)r`$. In this case there are additional Killing Spinors given by
$$ϵ=\left(e^{\frac{1}{2}A}2W(\varphi _0^A)e^{\frac{1}{2}A}x^\nu \mathrm{\Gamma }_{\underset{¯}{\nu }}\right)ϵ_{}.$$
(16)
There may also be non-supersymmetric vacua where $`V/\varphi ^A=0`$ but $`W_{3A}0`$. However we will have little so say about these cases.
In a Randall-Sundrum domain wall $`A(r)|r|`$ as $`r\pm \mathrm{}`$ . Thus asymptotically $`g_{00}=e^{2A}`$ falls off exponentially and gravity is is localised to the domain wall. This will be the case for a domain wall of the theory (11) if $`W`$ changes sign between the two vacua. For example in the original proposal there are no scalars $`\varphi ^A`$ or fermions $`\lambda _A`$ and $`VW^2`$ is constant. The domain wall is obtained by simply choosing the sign of $`W`$ to be positive on one side and negative on the other, i.e. $`W(r)`$ is discontinuous. Note that from the point of view of the supergravity equations this domain wall is equivalent to keeping $`W`$ fixed everywhere but choosing one representation for the $`\mathrm{\Gamma }`$-matrices on one side and the opposite representation (obtained by $`\mathrm{\Gamma }^m\mathrm{\Gamma }^m`$) on the other. Given the comments in the introduction it is natural to be concerned that this is unphysical.
In supergravity theories there is a standard argument for the stability of BPS backgrounds. We will briefly review it here and note that it is insensitive to a change in sign of $`W`$. We construct a “Nester” tensor
$$N^{mn}=\overline{ϵ}\mathrm{\Gamma }^{mnp}\delta \psi _p,$$
(17)
with $`\delta \psi _m`$ given in (LABEL:susy). Such a tensor has the property that, on shell,
$$_mN^{mn}=\overline{\delta \psi _m}\mathrm{\Gamma }^{mnp}\delta \psi _p+\gamma ^{AB}\overline{\delta \lambda _A}\mathrm{\Gamma }^n\delta \lambda _B.$$
(18)
So in particular $`_mN^{m0}`$ is negative definite (provided that we impose the Witten condition $`\mathrm{\Gamma }^m\delta \psi _m=0`$ ) and vanishes if and only if some supersymmetry is preserved. In our case this case the requirement that $`N^{mn}`$ satisfies (18) implies
$`W_{3A}`$ $`=`$ $`(D2){\displaystyle \frac{W}{\varphi ^A}},`$
$`V`$ $`=`$ $`4(D2)^2\left[\gamma ^{AB}{\displaystyle \frac{W}{\varphi ^A}}{\displaystyle \frac{W}{\varphi ^B}}\left({\displaystyle \frac{D1}{D2}}\right)W^2\right].`$
The Nester tensor can be used to provide a bound on the tension of an arbitrary domain wall in terms of a central charge of the supersymmetry algebra which in turn provides a non-perturbative proof of the stability of the solution. Following we integrate $`N^{mn}`$ over a spacelike boundary which encloses the domain wall
$$\frac{1}{2}𝑑\mathrm{\Sigma }_{mn}N^{mn}=𝑑\mathrm{\Sigma }_{0r}N^{0r}=𝑑\mathrm{\Sigma }_0_mN^{m0}0.$$
(20)
On the other hand we can directly evaluate the surface integral
$$𝑑\mathrm{\Sigma }_{0r}N^{0r}=\sigma |W(r=\mathrm{})W(r=\mathrm{})|,$$
(21)
where $`\sigma `$ is the tension of the domain wall and we have assumed that the domain wall interpolates smoothly between two AdS vacua. Combining these two equations we learn that $`\sigma |W(r=\mathrm{})W(r=\mathrm{})|`$ for all domain walls with equality if and only if some supersymmetry is preserved.
Note that this proof does not actually require that the action (11) admit a supersymmetric completion. The proof of stability merely requires that the identities (LABEL:restrictions) hold and that there are solutions to the supersymmetry Killing spinor equations (LABEL:susy). In particular it places no restriction on the function $`W`$ and hence any domain wall satisfying (14) will be stable in the purely bosonic theory . On the other hand we will shortly see that some choices of the function $`W(\varphi )`$ can never appear in a consistent supergravity because one could not consistently couple the theory to fermions.
To begin our discussion of the fermions we first obtain their equations of motion by constructing the most general form and then imposing the condition that their variation under supersymmetry vanishes when the scalars are on-shell. After a lengthy calculation we find
$`\mathrm{\Gamma }^m_m\lambda _A`$ $`+`$ $`M_A^B\lambda _B(D2)\gamma ^{CD}{\displaystyle \frac{\gamma _{BD}}{\varphi ^A}}{\displaystyle \frac{W}{\varphi ^C}}\lambda ^B{\displaystyle \frac{1}{2}}{\displaystyle \frac{\gamma _{BD}}{\varphi ^A}}\mathrm{\Gamma }^m_m\varphi ^B\lambda ^D`$ (22)
$`+`$ $`{\displaystyle \frac{1}{2}}\gamma _{AB}\mathrm{\Gamma }^m\mathrm{\Gamma }^n_n\varphi ^B\psi _m(D2){\displaystyle \frac{W}{\varphi ^A}}\mathrm{\Gamma }^m\psi _m=0,`$
$`\mathrm{\Gamma }^{mnp}_n\psi _p`$ $``$ $`(D2)W\mathrm{\Gamma }^{mn}\psi _n+(D2){\displaystyle \frac{W}{\varphi ^A}}\mathrm{\Gamma }^m\lambda ^A`$
$`+`$ $`{\displaystyle \frac{1}{2}}(g^{mn}\mathrm{\Gamma }^{mn})_n\varphi ^A\lambda _A=0,`$
(23)
where
$$M_A^B=2(D2)\frac{W}{\varphi ^A\varphi ^C}\gamma ^{BC}(D2)W\delta _A^B.$$
(24)
Therefore, in a supersymmetric AdS vacuum, we may set $`\psi _m=0`$ and obtain the equation of motion
$$\mathrm{\Gamma }^m_m\lambda _A+M_A^B\lambda _B=0.$$
(25)
Consider now a stable domain wall, i.e. one that satisfies (14). In particular since half of the supersymmetries are broken, one expects that a finite tension domain wall has massless Goldstino $`\lambda _A`$ modes bound to it. Therefore we should look for a solutions to the fermion equations in a background given by (14) which are invariant under the Poincare symmetry of the domain wall, i.e. with $`_\mu =\psi _\mu =0`$. It is important to note that the two fermion equations (22) and (23) do not decouple in this case and we must have $`\psi _r0`$. Specifically we find from the $`m=r`$ component of the $`\psi _m`$ equation (23) that $`\mathrm{\Gamma }^r\lambda _A=\lambda _A`$. For $`mr`$ the $`\psi _m`$ equation implies that
$$\psi _r=\frac{1}{W}\frac{W}{\varphi ^A}\lambda ^A.$$
(26)
Thus the fermion zero modes are chiral, as expected from the chiral form of the broken supersymmetries. Substituting (26) into the $`\lambda _A`$ equation (22) and using (14) yields the equation
$$\pm _r\lambda _A=W\lambda _A+2(D2)\left(\frac{^2W}{\varphi ^A\varphi ^B}\frac{1}{W}\frac{W}{\varphi ^A}\frac{W}{\varphi ^B}\right)\lambda ^B.$$
(27)
Thus we obtain the wavefunctions for the chiral Goldstino modes
$`\lambda _A`$ $`=`$ $`{\displaystyle \frac{1}{W}}{\displaystyle \frac{W}{\varphi ^A}}e^{\frac{1}{2}A}ϵ_{},`$
$`\psi _r`$ $`=`$ $`{\displaystyle \frac{1}{W^2}}\gamma ^{AB}{\displaystyle \frac{W}{\varphi ^A}}{\displaystyle \frac{W}{\varphi ^B}}e^{\frac{1}{2}A}ϵ_{},`$
where $`ϵ_{}`$ is a constant spinor satisfying $`\mathrm{\Gamma }^rϵ_{}=ϵ_{}`$. From (LABEL:Goldstinos) it is clear that if $`W(r)`$ passes through zero (e.g. if $`W`$ changes sign) the Goldstino modes will diverge on the domain wall (or more precisely where $`W=0`$) and will not be normalisable. From supersymmetry we expect that a smooth finite tension domain wall should have Goldstinos. We therefore conclude that $`W`$ can not change sign (by passing through zero) in a supergravity theory. In particular there are no smooth domain walls of the Randall-Sundrum type.
To be more explicit consider a single scalar and suppose that near the point where $`W`$ changes sign we may write $`W(\varphi \varphi _0)^\gamma `$ with $`\gamma 1`$ so that $`\varphi _0`$ is not a critical point. We then find that
$$\varphi \varphi _0(rr_0)^{\frac{1}{2\gamma }},A(rr_0)^{\frac{2}{2\gamma }},\lambda (rr_0)^{\frac{1}{2\gamma }}.$$
(29)
Thus the Goldstinos diverge where $`W`$ changes sign. Note that the metric and Killing spinors are bounded near $`r=r_0`$ but they will have a cusp singularity for $`\gamma <0`$ (i.e. if $`W`$ diverges). In addition the norm $`𝑑r\sqrt{g}\overline{\lambda }\lambda `$ will be convergent at $`r=0`$ only for $`\gamma <0`$.
The argument just given depends crucially on the form of the supersymmetry transformation rules (LABEL:susy). In four dimensions, for example, other possibilities arise and our results on the divergence of the Goldstino modes will not necessarily apply. Indeed four-dimensional supersymmetric supergravity domain walls do exist .
To illustrate the above points we may consider a case with just one scalar and a superpotential of the form
$$W(\varphi )=\alpha \left(\varphi \frac{1}{3}\beta ^2\varphi ^3\right),$$
(30)
where $`\alpha `$ and $`\beta `$ are constants and $`\gamma _{AB}=\delta _{AB}`$. The critical points occur at $`\varphi _0=\pm \beta ^1`$ where $`W=\pm 2\alpha /3`$ and indeed one can find smooth supersymmetric domain walls . The stability of these domain walls in the bosonic theory follows from the equations (20) and (21). However we see that this superpotential can never be consistently embedded in a supergravity because $`W`$ changes sign between the two critical points.
## 4 Conclusion
In this paper we have discussed the existence of domain walls in supersymmetric odd-dimensional theories. In the case of global supersymmetry we argued that no supersymmetric domain walls exist with purely chiral Goldstino modes. In the case of supergravities in odd dimensions we argued that the superpotential $`W`$ cannot change sign because if it did the Goldstino modes would diverge.
## 5 Acknowledgments
This work was initiated during a visit by G.W.G. to l’ Ecole Normale Supérieure. He would like to thank members of the L.P.T. for their kind hospitality. He would also like to thank Renata Kallosh, Andrei Linde and Shoichi Ichinose for helpful discussions. N.D.L. was supported by the EU grant ERBFMRX-CT96-0012 and would like to thank Eugène Cremmer for discussions. We would also like to Juan Maldacena for helpful comments on original version.
|
warning/0003/hep-ph0003032.html
|
ar5iv
|
text
|
# The QCD heavy-quark potential to order 𝑣²: one loop matching conditions
## I Introduction
For processes involving a non-relativistic heavy quark and antiquark, it is useful to combine the $`\alpha _s`$ expansion of QCD with an expansion in powers of the relative velocity $`v`$. The scattering of a heavy quark and antiquark, $`Q(𝐩)+\overline{Q}(𝐩)Q(𝐩^{})+\overline{Q}(𝐩^{})`$, can be described using a potential $`V`$, which has an expansion in powers of the velocity $`v`$ and $`\alpha _s`$,
$`V(𝐩,𝐩^{})={\displaystyle \underset{n=2}{\overset{\mathrm{}}{}}}V^{(n)},V^{(n)}={\displaystyle \underset{j=1}{\overset{\mathrm{}}{}}}V^{(n,j)},`$ (1)
$`\text{where}V^{(n)}v^n,V^{(n,j)}v^n\alpha _s^j.`$ (2)
An important complication of non-relativistic scattering is the presence of two low-energy scales, $`mv`$ and $`mv^2`$, which correspond to the momentum and energy of the heavy quark, respectively. In this paper, we will assume that $`mv^2\mathrm{\Lambda }_{\mathrm{QCD}}`$, and ignore any non-perturbative effects.
The first term in Eq. (1), $`V^{(2,1)}`$, is the static Coulomb potential at tree-level,
$`V^{(2,1)}=C{\displaystyle \frac{4\pi \alpha _s}{\left|𝐩𝐩^{}\right|^2}},`$ (3)
where the color factor $`C`$ depends on the relative color state of the incident quark and antiquark. For $`j>1`$, $`V^{(2,j)}`$ are perturbative corrections to the Coulomb potential which are known to two loops .
The static potential for a color singlet $`Q\overline{Q}`$ pair is often defined in terms of the expectation value of a rectangular Wilson loop of width $`R`$ and length $`T`$:
$`V^{(2)}(R)`$ $`=`$ $`\underset{T\mathrm{}}{lim}{\displaystyle \frac{1}{iT}}\mathrm{ln}\mathrm{Tr}P\mathrm{exp}\left(ig{\displaystyle 𝑑x_\mu A_a^\mu T^a}\right).`$ (4)
The Feynman diagrams corresponding to this expectation value build up the exponential of the static potential. As a result, in computing the static potential $`V^{(2,j)}`$ at order $`j`$, iterations of lower order terms in the potential, $`V^{(2,n)},n<j`$, are subtracted. At three loops Appelquist, Dine and Muzinich have shown that infrared divergences are encountered in $`V^{(2,4)}(R)`$ which are not canceled by simply subtracting iterations of the lower order potentials. Thus, using the definition in Eq. (4) at higher orders in perturbation theory, or generalizing this approach to subleading terms in the $`v`$ expansion, becomes cumbersome as the set of subtractions become more complicated, and perturbative subtractions are insufficient to render the potential in Eq. (4) well-defined.
A convenient framework for investigating the expansion in Eq. (1) is the effective field theory for non-relativistic QCD (NRQCD), formulated with a power counting in $`v`$ . In the effective field theory, it is more convenient to define the potential as the Wilson coefficient of a four-quark operator
$`_p={\displaystyle \underset{𝐩,p^{}}{}}V_{\alpha \beta \lambda \tau }(𝐩,𝐩^{})\mu ^{2ϵ}\left[\psi _{𝐩^{}}^{}{}_{\alpha }{}^{}(x)\psi _{𝐩}^{}{}_{\beta }{}^{}(x)\chi _{𝐩^{}}^{}{}_{\lambda }{}^{}(x)\chi _𝐩{}_{\tau }{}^{}(x)\right].`$ (5)
In Eq. (14), the fields $`\psi _{}`$ and $`\chi _{}`$ annihilate a quark and an antiquark, respectively. The fields are labelled by a momentum $`𝐩`$, and a greek index for their color and spin. The operator in Eq. (5) is local on the scale $`x1/mv^2`$, but non-local on the scale $`𝐩mv`$. The potential $`V`$ is computed as a matching coefficient at the scale $`\mu =m`$ between QCD and an effective theory for non-relativistic QCD valid below the scale $`m`$. The effective theory is constructed to have the same infrared structure as perturbative QCD for the two heavy quark system. Therefore, defining the potential as a matching coefficient provides an infrared safe definition. For instance, Ref. shows how the three-loop matching potential $`V`$ is infrared safe, despite the divergence in the QCD potential of Appelquist, Dine and Muzinich.
Although several different formulations of the effective theory for non-relativistic QCD are currently in use , certain universal features have emerged. The on-shell degrees of freedom in the effective theory include quarks with energy $`Emv^2`$ and momentum $`pmv`$, soft gluons with $`Epmv`$, and ultrasoft gluons with $`Epmv^2`$. The soft and ultrasoft modes are distinct, for instance a consistent power counting in $`v`$ demands that the ultrasoft gluon interactions are multipole expanded, while soft gluon interactions are not. The soft gluons are essential to correctly reproduce the beta function in the effective theory , and run between the scales $`m`$ and $`mv`$. Other massless on-shell fields, such as light quarks, will have ultrasoft and soft components too. There are also important off-shell field components, such as the exchange of gluons with $`Emv^2`$ and $`pmv`$ that build up the potential. Soft heavy quarks with $`Epmv`$ are also off-shell relative to the heavy quark states of interest. These off-shell field components can be integrated out of the Lagrangian in the effective theory. Doing this leaves a Lagrangian that is non-local at the scale $`mv`$, but local at $`mv^2`$. This procedure, which treats the potential components as four quark operators, was first seriously investigated in Ref. , and the resulting effective theory is referred to as pNRQCD. In Ref. it was proposed that the matching onto effective theories should take place in two stages: at $`\mu =m`$ one matches QCD onto NRQCD as originally defined in Ref. , and then matches NRQCD onto pNRQCD at the scale $`\mu =mv`$.
The matching of four-quark operators at $`m`$ was considered in Ref. , following the proposal in Ref. that the matching procedure should be similar to that in HQET. However, in general this procedure seems to be slightly problematic. First note that it is necessary to include the kinetic term for the potential quarks immediately below the scale $`m`$. For instance, this is necessary to reproduce the threshold value of the two-loop anomalous dimension for the heavy quark production current in the effective theory . With the kinetic energy term included at leading order, the consistency of the $`v`$ power counting forces us to have both ultrasoft and soft modes right below $`m`$ (along with the multipole expansion etc.). If we note that the kinetic energy term is necessary to correctly reproduce the Coulombic infrared divergences in the effective theory, then it might seem obvious that it must be included in order to have the correct infrared structure. At one-loop, matching exactly at threshold in dimensional regularization seemed to provide a method of avoiding this , but this procedure fails at two-loops.
Since we must include the kinetic energy term for all $`\mu <m`$ it seems quite natural to immediately consider matching at $`\mu =m`$ onto an effective theory where the off-shell potential gluons and soft quarks have been integrated out. Such a formulation was proposed in Ref. and will be considered in this paper. It will be referred to as vNRQCD. In this theory it is more appropriate to consider the running from $`m`$ to the scales $`mv`$ and $`mv^2`$ in a single step. This is accomplished by using the velocity renormalization group . Once the vNRQCD effective theory has been run down to the low scale the soft degrees of freedom have served their purpose and may be integrated out. In this final effective theory no additional running needs to be considered.
In this paper we compute the one-loop matching for the $`Q\overline{Q}`$ and $`QQ`$ potentials between QCD and vNRQCD up to corrections of order $`v^2`$. Ours results are compared to the matching calculation of Pineda and Soto for the four-quark operators and to the threshold expansion . The main difference between the the non-relativistic theories in Refs. and is the way in which large logarithms of $`v`$ are summed in the effective theory. There is also some difference in the matching corrections; some contributions in vNRQCD that arise at the scale $`\mu =m`$ are instead computed at $`\mu =mv`$ in pNRQCD. We will discuss the differences between the two approaches in the text of the paper.
The full QCD calculation of the $`Q\overline{Q}`$ scattering to order $`\alpha _s^2v^0`$ in the non-relativistic expansion has been done before . In calculating the potentials, Ref. performs an additional expansion, assuming
$`{\displaystyle \frac{(𝐩^{}+𝐩)^2}{(𝐩^{}𝐩)^2}}1.`$ (6)
In the usual $`v`$ power counting, $`𝐩`$ and $`𝐩^{}`$ are both of order $`v`$, so the ratio in Eq. (6) is of order unity, and cannot be treated as small. For this reason, in section II we redo the QCD calculation keeping the full dependence on this ratio for both the color singlet and octet channels. For the order $`v^0`$ spin-independent potential, terms proportional to $`(𝐩+𝐩^{})^2`$ which were dropped in previous calculations are necessary to match infrared divergences between the full and effective theories.
In section III we discuss the order $`v^0`$ potential of the form
$`{\displaystyle \frac{(𝐩_{}^{}{}_{}{}^{2}𝐩^2)^2}{m^2(𝐩^{}𝐩)^4}}.`$ (7)
For free quark states this potential vanishes on-shell by energy conservation, $`𝐩_{}^{}{}_{}{}^{2}=𝐩^2`$, and need not be included in the potential if one uses on-shell matching . The potential in Eq. (7) gives a non-zero contribution in loops or in matrix elements with external Coulomb states, and is often included in the quark potential. For instance, the usual Breit Hamiltonian includes a potential of the form in Eq. (7). A potential that is only non-zero off-shell can be gauge dependent, and including the potential of Eq. (7) induces gauge dependence in the coefficients of on-shell potentials. Using an off-shell potential gives correct results if all calculations are performed in the same gauge. We will use an on-shell basis for the potential where the term in Eq. (7) does not occur. The matching coefficients for the on-shell potential are gauge independent, since scattering amplitudes are measurable quantities. The difference between the on-shell matching potential and the Breit Hamiltonian is compensated for by a corresponding change in the $`1/|𝐩^{}𝐩|`$ potential, as discussed in more detail in section III (see also Refs. ).
In section IV we extend all of our results for the quark-antiquark potential to the quark-quark potential and in section V we give the QED limit of our results. Section VI gives our conclusions.
## II Matching the potential to $`𝒪`$$`(v^2)`$
The vNRQCD effective Lagrangian has the form
$`=_u+_s+_p.`$ (8)
The ultrasoft Lagrangian $`_u`$ involves the fields $`\psi _𝐩`$ which annihilate a quark, $`\chi _𝐩`$ which annihilate an antiquark, and $`A^\mu `$ which annihilate and create ultrasoft gluons. The potential Lagrangian $`_p`$ contains operators with four or more quark fields including the quark-antiquark potential. Finally the soft Lagrangian $`_s`$ contains all terms that involve soft fields which have energy and momenta of order $`mv`$. Heavy quarks with soft energy and momenta are off-shell and are therefore integrated out, so they do not appear explicitly in $`_s`$. The terms we need in the ultrasoft Lagrangian are
$`_u`$ $`=`$ $`{\displaystyle \frac{1}{4}}F^{\mu \nu }F_{\mu \nu }+{\displaystyle \underset{𝐩}{}}\psi _𝐩^{}\left\{iD^0{\displaystyle \frac{\left(𝐩i𝐃\right)^2}{2m}}+{\displaystyle \frac{𝐩^4}{8m^3}}\right\}\psi _𝐩`$ (10)
$`+\chi _𝐩^{}\left\{iD^0{\displaystyle \frac{\left(𝐩i𝐃\right)^2}{2m}}+{\displaystyle \frac{𝐩^4}{8m^3}}\right\}\chi _𝐩.`$
The covariant derivative on $`\psi _𝐩`$ and $`\chi _𝐩`$ contain the color matrices $`T^A`$ and $`\overline{T}^A`$ for the $`\mathrm{𝟑}`$ and $`\overline{\mathrm{𝟑}}`$ representations, respectively. Here $`D^\mu `$ involves only the ultrasoft gluon fields: $`D^\mu =^\mu +ig\mu _U^ϵA^\mu =(D^0,𝐃)`$, so $`D^0=^0+ig\mu _U^ϵA^0`$, $`𝐃=ig\mu _U^ϵ𝐀`$. The factors of the ultrasoft scale parameter $`\mu _U^ϵ`$ are included to make $`g=g(\mu _U)`$ dimensionless in $`d=42ϵ`$ dimensions. The ultrasoft gluon field $`A^\mu `$ is order $`v^2`$ and has dimension $`1ϵ`$, so $`A^\mu (mv^2)^{1ϵ}`$. Consistency of the $`v`$ power counting for the covariant derivative in $`d`$ dimensions then requires $`\mu _U=m\nu ^2`$ where $`\nu v`$. This reproduces the dependence of $`\mu _U`$ on the subtraction velocity $`\nu `$ given in Ref. . The terms we need in the soft Lagrangian are
$`_s`$ $`=`$ $`{\displaystyle \underset{q}{}}\left\{\left|q^\mu A_q^\nu q^\nu A_q^\mu \right|^2+\overline{\phi }_qq/\phi _q+\overline{c}_qq^2c_q\right\}`$ (13)
$`g^2\mu _S^{2ϵ}{\displaystyle \underset{𝐩,p^{},q,q^{}}{}}\{{\displaystyle \frac{1}{2}}\psi _{𝐩^{}}^{}{}_{}{}^{}[A_q^{}^\mu ,A_q^\nu ]U_{\mu \nu }^{(\sigma )}\psi _𝐩+{\displaystyle \frac{1}{2}}\psi _{𝐩^{}}^{}{}_{}{}^{}\{A_q^{}^\mu ,A_q^\nu \}W_{\mu \nu }^{(\sigma )}\psi _𝐩`$
$`+\psi _{𝐩^{}}^{}{}_{}{}^{}[\overline{c}_q^{},c_q]Y^{(\sigma )}\psi _𝐩+(\psi _{𝐩^{}}^{}{}_{}{}^{}T^BZ_\mu ^{(\sigma )}\psi _𝐩)(\overline{\phi }_q^{}\gamma ^\mu T^B\phi _q)\}+(\psi \chi ,T\overline{T}),`$
where $`A_q`$, $`c_q`$, and $`\phi _q`$ are soft gluons, ghosts, and massless quarks. The functions $`U`$, $`W`$, $`Y`$, and $`Z`$ are given in Appendix A. After integrating out the soft quarks the Lagrangian $`_s`$ is no longer manifestly gauge invariant with respect to gauge transformations at the scale $`mv`$. Therefore, determining the dependence of the soft scale parameter $`\mu _S`$ on $`\nu `$ may seem more difficult than the ultrasoft case. However, prior to integrating out the soft quarks the combination $`g\mu _S^ϵA_q`$ is from a covariant derivative, and $`A_q(mv)^{1ϵ}`$, yielding $`\mu _Sm\nu `$ in agreement with Ref. . In Eq. (13), $`g=g(\mu _S)`$. In general it is important to realize that the $`v`$ scaling of $`\mu _U`$ and $`\mu _S`$ are different. If the matching calculation is performed at the scaleIt is not necessary to match exactly at $`m`$. If the matching scale is $`\mu =\mu _hm`$, then one still sets $`\mu _U=\mu _S=\mu _h`$ and $`\nu =1`$, and factors of $`\mathrm{ln}(\mu _h/m)`$ appear in the matching coefficients. For convenience we choose $`\mu _h=m`$ in this paper. $`m`$, then for this computation $`\mu =\mu _S=\mu _U=m`$ and $`\nu =1`$ (where the usual QCD scale parameter is denoted by $`\mu `$). Therefore, for the matching at $`m`$ it is not essential to distinguish between $`\mu _S`$ and $`\mu _U`$.
The potential interaction relevant for our calculation is
$`_p={\displaystyle \underset{𝐩,p^{}}{}}V_{\alpha \beta \lambda \tau }(𝐩,𝐩^{})\mu _S^{2ϵ}\psi _{𝐩^{}}^{}{}_{\alpha }{}^{}\psi _{𝐩}^{}{}_{\beta }{}^{}\chi _{𝐩^{}}^{}{}_{\lambda }{}^{}\chi _𝐩{}_{\tau }{}^{}.`$ (14)
The factor of $`\mu _S^{2ϵ}`$ is included so that in $`d=42ϵ`$ dimensions the potential $`V`$ has dimension $`2`$. $`\alpha ,\beta ,\lambda ,\tau `$ denote color and spin indices. We will use the basis in which the potential is written as a linear combination of $`11`$ and $`T^a\overline{T}^a`$ in color space. One can convert to the color singlet and octet potential using the linear transformation
$`\left[\begin{array}{c}V_{\mathrm{singlet}}\\ V_{\mathrm{octet}}\end{array}\right]=\left[\begin{array}{ccc}1& & C_F\\ 1& & \frac{1}{2}C_AC_F\end{array}\right]\left[\begin{array}{c}V_{11}\\ V_{T\overline{T}}\end{array}\right],`$ (21)
where $`C_F=(N_c^21)/(2N_c)`$ and $`C_A=N_c`$. We will also need the invariants $`C_1=(N_c^21)/(4N_c^2)`$ and $`C_d=N_c4/N_c`$. These arise in the identities
$`T^AT^B\overline{T}^A\overline{T}^B`$ $`=`$ $`{\displaystyle \frac{1}{4}}(C_A+C_d)T^A\overline{T}^A+C_111,`$ (22)
$`T^AT^B\overline{T}^B\overline{T}^A`$ $`=`$ $`{\displaystyle \frac{1}{4}}(C_AC_d)T^A\overline{T}^A+C_111.`$ (23)
Written as a matrix, the order $`v^2`$ Coulomb potential is
$`V^{(2)}`$ $`=`$ $`(T^A\overline{T}^A){\displaystyle \frac{𝒱_c^{(T)}}{𝐤^2}}+(11){\displaystyle \frac{𝒱_c^{(1)}}{𝐤^2}},`$ (24)
where $`𝐤=𝐩^{}𝐩`$, and the coefficients $`𝒱_c^{(T)}`$ and $`𝒱_c^{(1)}`$ have an expansion in $`\alpha _s`$.
The order $`v^0`$ potential includes
$`V^{(0)}`$ $`=`$ $`(T^A\overline{T}^A)\left[{\displaystyle \frac{𝒱_2^{(T)}}{m^2}}+{\displaystyle \frac{𝒱_r^{(T)}(𝐩^2+p^2)}{2m^2𝐤^2}}+{\displaystyle \frac{𝒱_s^{(T)}}{m^2}}𝐒^2+{\displaystyle \frac{𝒱_\mathrm{\Lambda }^{(T)}}{m^2}}\mathrm{\Lambda }(𝐩^{},p)+{\displaystyle \frac{𝒱_t^{(T)}}{m^2}}𝐓(𝐤)\right]`$ (26)
$`+(11)\left[{\displaystyle \frac{𝒱_2^{(1)}}{m^2}}+{\displaystyle \frac{𝒱_s^{(1)}}{m^2}}𝐒^2\right],`$
where
$`𝐒`$ $`=`$ $`{\displaystyle \frac{𝝈_1+𝝈_2}{2}},\mathrm{\Lambda }(𝐩^{},p)=i{\displaystyle \frac{𝐒(p^{}\times p)}{𝐤^2}},𝐓(𝐤)=𝝈_1𝝈_2{\displaystyle \frac{3𝐤𝝈_1𝐤𝝈_2}{𝐤^2}},`$ (27)
and $`𝝈_1/2`$ and $`𝝈_2/2`$ are the spin-operators on the quark and anti-quark. Note that on-shell $`𝐩^{}{}_{}{}^{\mathrm{𝟐}}=𝐩^\mathrm{𝟐}`$, but we have written $`𝐩^2+p^2`$ in Eq. (26) so that the Lagrangian $`_p`$ is hermitian.
The tree level diagram in Fig. 1a generates terms of $`𝒪(v^{2k}\alpha _s),k1`$, in the QCD potential. Matching at $`\mu =m`$, $`\nu =1`$ gives
$`𝒱_c^{(T)}`$ $`=`$ $`4\pi \alpha _s(m),𝒱_c^{(1)}=0,𝒱_r^{(T)}=4\pi \alpha _s(m),`$ (28)
$`𝒱_s^{(T)}`$ $`=`$ $`{\displaystyle \frac{4\pi \alpha _s(m)}{3}},𝒱_\mathrm{\Lambda }^{(T)}=6\pi \alpha _s(m),𝒱_t^{(T)}={\displaystyle \frac{\pi \alpha _s(m)}{3}},`$ (29)
$`𝒱_s^{(1)}`$ $`=`$ $`0,𝒱_2^{(T)}=0,𝒱_2^{(1)}=0.`$ (30)
The annihilation diagram in Fig. 1b generates terms of order $`\alpha _sv^{2k},k0`$ in the potential. Using Fierz identities and charge conjugation, these operators can be transformed into the basis in Eq. (26) and give additional contributions to the matching. Only $`𝒱_s^{(T,1)}`$ receive non-zero annihilation contributions at tree level:
$`𝒱_{s,a}^{(T)}={\displaystyle \frac{1}{N_c}}\pi \alpha _s(m),𝒱_{s,a}^{(1)}={\displaystyle \frac{(N_c^21)}{2N_c^2}}\pi \alpha _s(m).`$ (31)
The complete tree level matching is given by adding the terms in Eqs. (28) and (31). We have found it convenient to distinguish the annihilation contributions by including an additional subscript $`a`$ on the coefficients they generate. The leading-log values of the order $`v^0`$ potentials in Eq. (26) were calculated in Refs. , but are not needed here. Nonzero values for $`𝒱_2^{(T,1)}`$ are generated in the renormalization group flow below the scale $`m`$ , as well as by the one-loop matching as we will see below.
At one-loop the matching onto QCD gives order $`1/v`$ terms of the form
$`V^{(1)}`$ $`=`$ $`{\displaystyle \frac{\pi ^2}{m|𝐤|}}\left[𝒱_k^{(T)}(T^A\overline{T}^A)+𝒱_k^{(1)}(11)\right],`$ (32)
where the coefficients $`𝒱_k^{(T,1)}`$ are dimensionless. In $`d=42ϵ`$ dimensions the one-loop matching produces a potential with the dependence $`\mu ^{2ϵ}/|𝐤|^{1+2ϵ}`$. We have chosen to define $`V^{(1)}`$ by taking the $`d4`$ limit, which differs from the definition of this operator used in Ref. .
### A The QCD Calculation
To perform the potential matching calculation, we consider the on-shell $`Q\overline{Q}`$ scattering amplitude in QCD and in vNRQCD to order $`\alpha _s^2v^0`$. We will use Feynman gauge, regulate infrared divergences with a finite gluon mass<sup>§</sup><sup>§</sup>§Using a finite gluon mass is dangerous in the presence of diagrams with the non-abelian gluon vertices. All such diagrams we require here are IR finite in Feynman gauge. $`\lambda `$, and renormalize ultraviolet divergences with dimensional regularization and the $`\overline{\mathrm{MS}}`$ scheme. Since the calculation is performed on-shell the resulting matching coefficients will be gauge independent.
We begin by considering the QCD diagrams. The most complicated diagram is the QCD box diagram which includes contributions of order $`1/v^3`$, $`1/v^2`$, $`1/v`$ and $`v^0`$, as well as higher order terms which we do not need in this paper.
$`=`$ $`{\displaystyle \frac{i\alpha _s^2}{k^2}}(T^AT^B\overline{T}^A\overline{T}^B)[{\displaystyle \frac{2im\pi }{p}}\mathrm{ln}\left({\displaystyle \frac{k^2}{\lambda ^2}}\right)4\mathrm{ln}\left({\displaystyle \frac{k^2}{\lambda ^2}}\right)`$ (33)
$`+`$ $`{\displaystyle \frac{\pi ^2k}{mp(2p+k)}}\left\{6p^2+{\displaystyle \frac{5pk}{2}}+{\displaystyle \frac{k^2}{4}}3k^2\mathrm{\Lambda }\left(pk+{\displaystyle \frac{5k^2}{6}}\right)𝐒^2{\displaystyle \frac{k^2𝐓}{12}}{\displaystyle \frac{k𝐑}{4(2p+k)}}\right\}`$ (34)
$`+`$ $`{\displaystyle \frac{6i\pi k^2\mathrm{\Lambda }}{mp(4p^2k^2)}}\left\{k^2\mathrm{ln}\left({\displaystyle \frac{2p}{\lambda }}\right)4p^2\mathrm{ln}\left({\displaystyle \frac{k}{\lambda }}\right)\right\}`$ (35)
$`+`$ $`{\displaystyle \frac{i\pi k^2𝐒^2}{3mp(4p^2k^2)}}\left\{k^24p^2+(5k^212p^2)\mathrm{ln}\left({\displaystyle \frac{2p}{\lambda }}\right)(4p^2+k^2)\mathrm{ln}\left({\displaystyle \frac{k}{\lambda }}\right)\right\}`$ (36)
$`+`$ $`{\displaystyle \frac{i\pi k^2𝐓}{6mp(4p^2k^2)}}\left\{4p^2k^2+k^2\mathrm{ln}\left({\displaystyle \frac{2p}{\lambda }}\right)+(8p^2+k^2)\mathrm{ln}\left({\displaystyle \frac{k}{\lambda }}\right)\right\}`$ (37)
$``$ $`{\displaystyle \frac{i\pi k^2𝐑}{2mp(4p^2k^2)^2}}\left\{k^24p^2+(4p^2+k^2)\mathrm{ln}\left({\displaystyle \frac{2p}{k}}\right)\right\}`$ (38)
$`+`$ $`{\displaystyle \frac{i\pi }{2mp(4p^2k^2)}}\{4p^2k^2k^4k^2(4p^2+k^2)\mathrm{ln}\left({\displaystyle \frac{2p}{\lambda }}\right)`$ (39)
$`+`$ $`(80p^416p^2k^2+k^4)\mathrm{ln}\left({\displaystyle \frac{k}{\lambda }}\right)\}{\displaystyle \frac{6k^2}{m^2}}+{\displaystyle \frac{8k^2𝐒^\mathrm{𝟐}}{3m^2}}{\displaystyle \frac{k^2𝐓}{3m^2}}`$ (40)
$`+`$ $`\mathrm{ln}\left({\displaystyle \frac{\lambda }{k}}\right)\{{\displaystyle \frac{56p^2}{3m^2}}{\displaystyle \frac{12k^2\mathrm{\Lambda }}{m^2}}{\displaystyle \frac{8k^2𝐒^2}{3m^2}}{\displaystyle \frac{2k^2𝐓}{3m^2}}\}+\mathrm{ln}\left({\displaystyle \frac{k}{m}}\right)\{{\displaystyle \frac{4k^2}{m^2}}{\displaystyle \frac{2k^2𝐒^2}{m^2}}\}],`$ (41)
where $`k=|𝐤|`$, $`p=|𝐩|=|𝐩^{}|`$ and
$$𝐑=(𝐩+p^{})𝝈_1(𝐩+p^{})𝝈_2.$$
(42)
The real part of the order $`1/v`$ amplitude agrees with Ref. in the limit $`pk/2`$. We have kept the full $`p`$ dependence since taking this limit is not justified by the power counting. We have also kept imaginary terms generated by the cut amplitude to emphasize how these terms are correctly reproduced in the effective theory. The real part of the spin dependent order $`v^0`$ amplitude agree with the result in Ref. . The real part of the spin independent order $`v^0`$ amplitude agrees with Ref. , except for the order $`v^0`$ $`\mathrm{ln}(k)`$ and $`\mathrm{ln}(\lambda )`$ dependence. The difference is due to the condition $`|𝐩^{}+𝐩|k`$ which was imposed in Refs. , but which violates the $`v`$ power counting. (For the crossed box given below in Eq. (42a), the order $`v^0`$ $`\mathrm{ln}(\lambda )`$ and $`\mathrm{ln}(k)`$ also differs from Ref. .) The remaining direct scattering diagrams are less complicated since they have no cuts:
$`=`$ $`{\displaystyle \frac{i\alpha _s^2}{k^2}}(T^AT^B\overline{T}^B\overline{T}^A)[4\mathrm{ln}\left({\displaystyle \frac{k^2}{\lambda ^2}}\right){\displaystyle \frac{\pi ^2k}{m}}`$ (46)
$`+\mathrm{ln}\left({\displaystyle \frac{\lambda }{k}}\right)\left\{{\displaystyle \frac{8k^2}{3m^2}}{\displaystyle \frac{56p^2}{3m^2}}+{\displaystyle \frac{12k^2\mathrm{\Lambda }}{m^2}}+{\displaystyle \frac{8k^2𝐒^2}{3m^2}}+{\displaystyle \frac{2k^2𝐓}{3m^2}}\right\}`$
$`+\mathrm{ln}\left({\displaystyle \frac{k}{m}}\right)\{{\displaystyle \frac{6k^2}{m^2}}+{\displaystyle \frac{2k^2𝐒^2}{m^2}}\}+{\displaystyle \frac{2k^2}{m^2}}{\displaystyle \frac{2k^2𝐒^2}{3m^2}}+{\displaystyle \frac{k^2𝐓}{3m^2}}],`$
$`\text{}+\text{}`$ $`=`$ $`{\displaystyle \frac{i\alpha _s^2}{k^2}}C_A(T^A\overline{T}^A)[3\mathrm{ln}\left({\displaystyle \frac{\mu ^2}{m^2}}\right)+4+{\displaystyle \frac{\pi ^2k}{2m}}`$ (50)
$`+\mathrm{ln}\left({\displaystyle \frac{k}{m}}\right)\left\{{\displaystyle \frac{4k^2}{m^2}}{\displaystyle \frac{4k^2𝐒^2}{3m^2}}{\displaystyle \frac{k^2𝐓}{3m^2}}{\displaystyle \frac{4k^2\mathrm{\Lambda }}{m^2}}\right\}`$
$`+\mathrm{ln}\left({\displaystyle \frac{\mu }{m}}\right)\left\{{\displaystyle \frac{6p^2}{m^2}}{\displaystyle \frac{9k^2\mathrm{\Lambda }}{m^2}}{\displaystyle \frac{2k^2𝐒^2}{m^2}}{\displaystyle \frac{k^2𝐓}{2m^2}}\right\}`$
$`+{\displaystyle \frac{4p^2}{m^2}}{\displaystyle \frac{8k^2\mathrm{\Lambda }}{m^2}}{\displaystyle \frac{2k^2𝐒^\mathrm{𝟐}}{m^2}}{\displaystyle \frac{k^2𝐓}{2m^2}}],`$
$`\text{}+\text{}`$ $`=`$ $`{\displaystyle \frac{i\alpha _s^2}{k^2}}(2C_FC_A)(T^A\overline{T}^A)[\{2\mathrm{ln}\left({\displaystyle \frac{\lambda ^2}{m^2}}\right)+\mathrm{ln}\left({\displaystyle \frac{\mu ^2}{m^2}}\right)+4\}M^0`$ (52)
$`+{\displaystyle \frac{k^2}{m^2}}{\displaystyle \frac{2k^2𝐒^\mathrm{𝟐}}{3m^2}}{\displaystyle \frac{k^2𝐓}{6m^2}}{\displaystyle \frac{2k^2\mathrm{\Lambda }}{m^2}}+{\displaystyle \frac{4k^2}{3m^2}}\mathrm{ln}\left({\displaystyle \frac{\lambda }{m}}\right)],`$
$`\text{}+\text{perms}`$ $`=`$ $`{\displaystyle \frac{i\alpha _s^2}{k^2}}\mathrm{\hspace{0.17em}2}C_F(T^A\overline{T}^A)\left\{2\mathrm{ln}\left({\displaystyle \frac{\lambda ^2}{m^2}}\right)+\mathrm{ln}\left({\displaystyle \frac{\mu ^2}{m^2}}\right)+4\right\}M^0,`$ (53)
$`\text{}+\text{}`$ $`=`$ $`{\displaystyle \frac{i\alpha _s^2}{k^2}}C_A(T^A\overline{T}^A)\left\{{\displaystyle \frac{5}{3}}\mathrm{ln}\left({\displaystyle \frac{\mu ^2}{k^2}}\right)+{\displaystyle \frac{31}{9}}\right\}M^0,`$ (54)
and the light quark loops for $`n_f`$ flavors gives:
$`=`$ $`{\displaystyle \frac{i\alpha _s^2}{k^2}}n_fT_F(T^A\overline{T}^A)\left\{{\displaystyle \frac{4}{3}}\mathrm{ln}\left({\displaystyle \frac{\mu ^2}{k^2}}\right)+{\displaystyle \frac{20}{9}}\right\}M^0,`$ (57)
while the heavy quark fermion loop gives:
$`=`$ $`{\displaystyle \frac{i\alpha _s^2}{k^2}}T_F(T^A\overline{T}^A)\left\{{\displaystyle \frac{4}{3}}\mathrm{ln}\left({\displaystyle \frac{\mu ^2}{m^2}}\right)M^0{\displaystyle \frac{4k^2}{15m^2}}\right\}.`$ (60)
In Eq. (42) the matrix element
$`M^0=1+{\displaystyle \frac{p^2}{m^2}}{\displaystyle \frac{k^2𝐒^2}{3m^2}}{\displaystyle \frac{3k^2\mathrm{\Lambda }}{2m^2}}{\displaystyle \frac{k^2𝐓}{12m^2}}.`$ (61)
Note that we disagree with Ref. on the order $`v^0`$ spin independent part of Eq. (42b). Ref. has an additional non-logarithmic $`1/m^2`$ term. The difference arises because we find a different value for the order $`k^2`$ term in the non-Abelian $`F_1`$ form-factor. Eqs. (42c) through (42g) agree with Ref. .
The diagrams in Eq. (33) and (42) have contributions from several different scales. In particular, in the language of the threshold expansion, the hard regime gives $`\mathrm{ln}(\mu /m)`$’s, the soft regime gives $`\mathrm{ln}(\mu /k)`$’s, and the ultrasoft regime gives $`\mathrm{ln}(\mu /\lambda )`$’s. There are also $`i\mathrm{ln}(\lambda /p)`$ and $`i\mathrm{ln}(\lambda /k)`$ terms from the Coulomb divergence in the potential regime. In addition to logarithms, all regimes can give constant factors. In the effective theory, terms from the hard regime are absorbed into matching coefficients such as the four-quark potential operators and the remaining terms correspond to graphs involving modes in the effective theory. These graphs are discussed in more detail below.
Next consider the one-loop annihilation diagrams in QCD. Since the intermediate gluons are hard we expect these graphs to include a factor of $`1/m^2`$, thus giving hard order $`v^0`$ contributions to the potential. However, the graph in Eq. (II Ac) also has terms enhanced by a factor of $`m/p`$, which are order $`1/v`$ contributions.
$`\text{}+\text{}`$ $`=`$ $`{\displaystyle \frac{i\alpha _s^2}{m^2}}[\{({\displaystyle \frac{C_1}{N_c}}+{\displaystyle \frac{C_d(N_c^21)}{8N_c^2}})(11)+({\displaystyle \frac{C_d}{4N_c}}2C_1)(T^A\overline{T}^A)\}`$ (63)
$`\times `$ $`(𝐒^\mathrm{𝟐}2)(i\pi +22\mathrm{ln}2)C_A\left\{{\displaystyle \frac{(N_c^21)}{2N_c^2}}(11)+{\displaystyle \frac{1}{N_c}}(T^A\overline{T}^A)\right\}`$ (64)
$`\times `$ $`𝐒^\mathrm{𝟐}\{{\displaystyle \frac{i\pi }{12}}+{\displaystyle \frac{1}{6}}{\displaystyle \frac{\mathrm{ln}2}{6}}\mathrm{ln}\left({\displaystyle \frac{\lambda }{m}}\right)\}],`$ (65)
$`\text{}+\text{}`$ $`=`$ $`{\displaystyle \frac{i\alpha _s^2}{m^2}}C_A\left\{(T^A\overline{T}^A){\displaystyle \frac{1}{N_c}}+(11){\displaystyle \frac{(N_c^21)}{2N_c}}\right\}`$ (66)
$`\times `$ $`𝐒^\mathrm{𝟐}\left\{{\displaystyle \frac{3}{2}}\mathrm{ln}\left({\displaystyle \frac{\mu }{m}}\right)+{\displaystyle \frac{4}{3}}+{\displaystyle \frac{2\mathrm{ln}2}{3}}{\displaystyle \frac{i\pi }{3}}\right\},`$ (67)
$`\text{}+\text{}`$ $`=`$ $`{\displaystyle \frac{i\alpha _s^2}{4m^2}}(2C_FC_A)\left\{(T^A\overline{T}^A){\displaystyle \frac{1}{N_c}}+(11){\displaystyle \frac{(N_c^21)}{2N_c}}\right\}`$ (68)
$`\times `$ $`𝐒^\mathrm{𝟐}\left\{{\displaystyle \frac{\pi ^2m}{p}}+{\displaystyle \frac{2i\pi m}{p}}\mathrm{ln}\left({\displaystyle \frac{2p}{\lambda }}\right)4+2\mathrm{ln}\left({\displaystyle \frac{\mu }{m}}\right)+4\mathrm{ln}\left({\displaystyle \frac{\lambda }{m}}\right)\right\},`$ (69)
$`\text{}+\text{perms}`$ $`=`$ $`{\displaystyle \frac{i\alpha _s^2}{m^2}}C_F\left\{(T^A\overline{T}^A){\displaystyle \frac{1}{N_c}}+(11){\displaystyle \frac{(N_c^21)}{2N_c}}\right\}`$ (70)
$`\times `$ $`𝐒^\mathrm{𝟐}\left\{\mathrm{ln}\left({\displaystyle \frac{\lambda ^2}{m^2}}\right)+\mathrm{ln}\left({\displaystyle \frac{\mu }{m}}\right)+2\right\},`$ (71)
$`\text{}+\text{}`$ $`=`$ $`{\displaystyle \frac{i\alpha _s^2}{m^2}}{\displaystyle \frac{5C_A}{12}}\left\{(T^A\overline{T}^A){\displaystyle \frac{1}{N_c}}+(11){\displaystyle \frac{(N_c^21)}{2N_c}}\right\}`$ (72)
$`\times `$ $`𝐒^\mathrm{𝟐}\left\{\mathrm{ln}\left({\displaystyle \frac{\mu ^2}{m^2}}\right)2\mathrm{ln}2+{\displaystyle \frac{31}{15}}+i\pi \right\},`$ (73)
and the light quark loop for $`n_f`$ flavors gives:
$`=`$ $`{\displaystyle \frac{i\alpha _s^2}{m^2}}{\displaystyle \frac{n_fT_F}{3}}\left\{(T^A\overline{T}^A){\displaystyle \frac{1}{N_c}}+(11){\displaystyle \frac{(N_c^21)}{2N_c}}\right\}`$ (75)
$`\times `$ $`𝐒^\mathrm{𝟐}\left\{\mathrm{ln}\left({\displaystyle \frac{\mu ^2}{m^2}}\right)2\mathrm{ln}2+{\displaystyle \frac{5}{3}}+i\pi \right\},`$ (76)
while the heavy quark loop gives:
$`=`$ $`{\displaystyle \frac{i\alpha _s^2}{m^2}}{\displaystyle \frac{T_F}{3}}\left\{(T^A\overline{T}^A){\displaystyle \frac{1}{N_c}}+(11){\displaystyle \frac{(N_c^21)}{2N_c}}\right\}`$ (78)
$`\times `$ $`𝐒^\mathrm{𝟐}\left\{\mathrm{ln}\left({\displaystyle \frac{\mu ^2}{m^2}}\right)+{\displaystyle \frac{8}{3}}\right\}.`$ (79)
In the limit $`pk/2`$, the results in Eq. (II A) agree with Ref. , except for Fig. (II Ac) and differences that can be accounted for by the fact that we are using $`\overline{\mathrm{MS}}`$ rather than an on-shell subtraction scheme. For Fig. (II Ac), Ref. has a $`\pi ^2/k`$ term which for us is $`\pi ^2/(2p)`$. This sign for the $`\pi ^2/p`$ term is in agreement with Ref. . The imaginary parts of these one-loop annihilation amplitudes also agree with Ref. . Note that in Eqs. (33), (42), and (II A), $`\alpha _s=\alpha _s(\mu )`$.
### B The Effective Theory Calculation
The effective theory contains potential, soft and ultrasoft loops. We have organized the terms by their order in the velocity expansion. The order in $`v`$ can be determined using the $`v`$ power counting formulaHere the power of $`v`$ is given for the amputated diagram, so unlike Eq. (40) in Ref. the factors of $`v`$ associated with external lines are not included. in Eq. (40) of Ref. . A loop graph with two insertions of the Coulomb potential contributes to a $`1/v^3`$ potential, and with an insertion of one Coulomb and one $`V^{(0)}`$ potential contributes to a $`1/v`$ potential. A soft loop with two vertices of order $`\sigma `$ and $`\sigma ^{}`$ contributes to the $`v^{\sigma +\sigma ^{}2}`$ potential. Graphs involving the exchange of an ultrasoft gluon begin to contribute to the potential at order $`v^0`$, etc.
#### 1 Order $`1/v^3`$
In the effective theory taking two insertions of the tree level Coulomb potential in a loop gives the only diagram that is order $`\alpha _s^2/v^3`$:
$`=`$ $`i{\displaystyle \frac{[𝒱_c^{(T)}]^2}{16\pi ^2}}(T^AT^B\overline{T}^A\overline{T}^B){\displaystyle \frac{2im\pi }{k^2p}}\mathrm{ln}\left({\displaystyle \frac{k^2}{\lambda ^2}}\right).`$ (80)
Taking $`\mu =m`$ and $`\nu =1`$ and using the tree level value of $`V_c^{(T)}`$, this graph exactly reproduces the order $`\alpha _s^2/v^3`$ “Coulomb singularity” term in the QCD box diagram in Eq. (33). Thus, there is no matching correction at this order.
#### 2 Order $`1/v^2`$
At order $`\alpha _s^2/v^2`$ in the effective theory, the only non-zero diagram involves the exchange of soft gluons, ghosts and quarks:
$`=`$ $`{\displaystyle \frac{i\alpha _s^2(\mu _S)}{k^2}}(T^A\overline{T}^A)\left[{\displaystyle \frac{4T_Fn_f11C_A}{3}}\left\{{\displaystyle \frac{1}{\widehat{ϵ}}}+\mathrm{ln}\left({\displaystyle \frac{\mu _S^2}{k^2}}\right)\right\}+{\displaystyle \frac{20T_Fn_f31C_A}{9}}\right],`$ (81)
where $`1/\widehat{ϵ}=1/ϵ\gamma +\mathrm{ln}(4\pi )`$ is the combination subtracted in $`\overline{\mathrm{MS}}`$. The divergence in this graph is responsible for the one-loop running of $`𝒱_c^{(T)}`$. At one-loop, to order $`v^0`$ the effective theory counterterms (and running of the potential) were computed in Ref. and from now on the $`1/\widehat{ϵ}`$ dependence will be dropped. The sum of order $`\alpha _s^2/v^2`$ terms from the QCD diagrams in Eqs. (33) and (42) is
$`{\displaystyle \frac{i\alpha _s^2(\mu )}{k^2}}(T^A\overline{T}^A)\left[{\displaystyle \frac{4T_F}{3}}\mathrm{ln}\left({\displaystyle \frac{\mu ^2}{m^2}}\right)+{\displaystyle \frac{4T_Fn_f11C_A}{3}}\mathrm{ln}\left({\displaystyle \frac{\mu ^2}{k^2}}\right)+{\displaystyle \frac{20T_Fn_f31C_A}{9}}\right].`$ (82)
At the scale $`\mu =\mu _S=m`$ the $`\overline{\mathrm{MS}}`$ values of Eq. (81) and Eq. (82) are identical, so there is no one-loop matching correction to $`𝒱_c`$. The difference in $`\mathrm{ln}(\mu )`$’s in Eqs. (82) and (81) is the change in $`\beta `$-function from $`n_f+1`$ to $`n_f`$ flavors. We would expect a new contribution to $`𝒱_c`$ only if the QCD graphs have a contribution from the hard regime where energy and momenta are order $`m`$. In Eqs. (42b), (42c), and (42d), the factors of $`3\mathrm{ln}(m^2)+4`$ are from the hard regime, but they sum to zero. For the $`\mathrm{ln}(m)`$ terms this is a consequence of the Ward identity derived in Ref. . The constant factor $`(20T_Fn_f31C_A)/9`$ vanishes in the matching condition at $`m`$, and is carried to scales below $`m`$ by the soft modes.
Note that if at the scale $`\nu =v_kk/m`$ we integrate out the soft modes, then the Coulomb potential for the theory below this scale, $`\overline{𝒱}_c`$, will obtain an additional contribution from this second stage of matching: $`\overline{𝒱}_c^{(T)}(v_k)=𝒱_c^{(T)}(v_k)(20T_Fn_f31C_A)\alpha _s^2(mv_k)/9`$. In this expression $`𝒱_c^{(T)}(v_k)`$ is the value of $`𝒱_c^{(T)}`$ obtained from running this coefficient from $`\nu =1`$ to $`\nu =v_k`$ using its two-loop anomalous dimension. This reproduces the constant factor that is typically associated with the Coulomb potential at next-to-leading order (see for instance, Ref. ). Similarly, one can obtain the additional terms from the matching contributions for the $`v^0`$ potentials at $`mv`$ from the value of the soft loops given in Eq. (102).
#### 3 Order $`1/v`$
The possible order $`\alpha _s^2/v`$ diagrams with soft gluons vanish explicitly, so the only order $`\alpha _s^2/v`$ diagrams in the effective theory involve two iterations of the potential. There are two diagrams with insertions of $`𝒱_c^{(T)}`$:
$`\text{}+\mathrm{\Delta }E\text{}`$ $`=`$ $`{\displaystyle \frac{i[𝒱_c^{(T)}]^2}{4m}}T^AT^B\overline{T}^A\overline{T}^B\left(I_F+2p^2I_0\right),`$ (83)
where the integrals $`I_0`$ and $`I_F`$ are given in Appendix B. The dependence on $`𝒱_c^{(1)}`$ is not needed since tree level matching gives $`𝒱_c^{(1)}=0`$. In the first diagram in Eq. (83), the cross denotes an insertion of the $`𝐩^4/m^3`$ operator from Eq. (10). The second diagram is nominally order $`\alpha _s^2/v^3`$; however it depends on the heavy quark energy $`E=𝐩^2/(2m)𝐩^4/(8m^3)+\mathrm{}`$. When the energy is expanded in terms of momenta, the graph includes a contribution of order $`\alpha _s^2/v`$ which we indicate by the pre-factor $`\mathrm{\Delta }E`$ in Eq. (83). Each of the diagrams in Eq. (83) has an IR divergence that is not regulated by $`\lambda `$, but the IR divergence in the sum of the integrands for the two diagrams is regulated.
Additional $`\alpha _s^2/v`$ diagrams are generated by including one insertion of the Coulomb potential and one order $`v^0`$ potential:
$`\text{}+\text{}`$ $`=`$ $`{\displaystyle \frac{i𝒱_c^{(T)}𝒱_r^{(T)}}{m}}T^AT^B\overline{T}^A\overline{T}^B\left(I_F+2p^2I_0\right),`$ (85)
$`\text{}+\text{}`$ $`=`$ $`{\displaystyle \frac{i𝒱_c^{(T)}𝒱_s^{(T)}}{m}}T^AT^B\overline{T}^A\overline{T}^B\left(2I_P𝐒^2\right),`$ (86)
$`\text{}+\text{}`$ $`=`$ $`{\displaystyle \frac{i𝒱_c^{(T)}𝒱_\mathrm{\Lambda }^{(T)}}{m}}T^AT^B\overline{T}^A\overline{T}^B\left(2k^2\mathrm{\Lambda }(𝐩^{},𝐩)I_A\right),`$ (87)
$`\text{}+\text{}`$ $`=`$ $`{\displaystyle \frac{i𝒱_c^{(T)}𝒱_t^{(T)}}{2m}}T^AT^B\overline{T}^A\overline{T}^B[𝐤\sigma _1k\sigma _2(12I_C+3I_0)`$ (88)
$`4`$ $`\sigma _1\sigma _2(I_P3I_B)+𝐑(12I_D12I_A+3I_0)],`$ (89)
$`\text{}+\text{}`$ $`=`$ $`{\displaystyle \frac{i𝒱_c^{(T)}𝒱_{s,a}^{(T)}}{m}}T^AT^B\overline{T}^A\overline{T}^B\left(𝐒^\mathrm{𝟐}2I_P\right)`$ (91)
$`+{\displaystyle \frac{i𝒱_c^{(T)}𝒱_{s,a}^{(1)}}{m}}T^A\overline{T}^A\left(𝐒^\mathrm{𝟐}2I_P\right).`$
The last two diagrams involve insertions of terms in the potential generated by the tree level annihilation diagram. In Eqs. (83) and (II B 3) the dependence on $`11`$ potentials whose coefficients vanish at tree level are not shown. Thus, both the $`11`$ and $`T\overline{T}`$ contributions are only shown in Eq. (II B 3e). The new integrals $`I_P`$, $`I_A`$, $`I_B`$, $`I_C`$, and $`I_D`$ that appear in Eq. (II B 3) are given in Appendix B, and $`𝐑`$ is defined in Eq. (42).
The infrared divergences in the $`1/v`$ amplitude are due to the Coulomb singularity. The divergences in the full theory are reproduced by the potential loops in Eqs. (83) and (II B 3). The imaginary terms in the amplitude from the QCD box graph exactly agree with those in the effective theory, and do not contribute in the matching coefficients, as expected. Subtracting the effective theory graphs in Eqs. (83) and (II B 3) from the order $`\alpha _s^2/v`$ terms in Eq. (33), (42)a,b and (II A)c gives the matching result for the $`V^{(1)}`$ potential at $`\mu =m`$, $`\nu =1`$:
$`𝒱_k^{(T)}`$ $`=`$ $`\alpha _s^2(m)\left({\displaystyle \frac{7C_A}{8}}{\displaystyle \frac{C_d}{8}}\right),𝒱_k^{(1)}=\alpha _s^2(m){\displaystyle \frac{C_1}{2}}.`$ (92)
For the color singlet channel this gives the matching coefficient $`𝒱_k^{(s)}=\alpha _s^2(m)(C_F^2/2C_FC_A)`$ in agreement with Ref. . Note that the real part of the $`1/v`$ amplitudes in the full and effective theory have a complicated dependence on $`p`$ and $`k`$, but the momentum dependence of the matching for the direct potentials in Eq. (92) is only of the form $`1/\left|𝐤\right|`$. For the annihilation graphs the $`1/p`$ terms cancel between the full and effective theories. There would have been a $`1/p`$ matching coefficient for the annihilation graphs if the results of Ref. had been used for the full theory graph.
The matching coefficients in Eq. (92) correspond to a contribution to the $`1/v`$ potential at the scale $`\mu =m`$, which does not come from the hard part of any graph. This is in apparent contradiction with the threshold expansion. It also appears to disagree with Refs. and , where the $`1/|𝐤|`$ four-quark potential operator is said to only arise at the scale $`mv`$. However, in Refs. and a formulation of NRQCD is being used in which off-shell potential field components have not been integrated out for scales $`mv<\mu <m`$, but instead are considered to be dynamical fields in the Lagrangain (see Refs. ). In the full theory the $`1/|𝐤|`$ terms come from three types of graphs, shown in Eq. (33) and Eqs. (42a,b). At the scale $`m`$ there are now effective theory graphs analogous to those in Eq. (42b) but with two $`A^0`$ potential gluons and one $`𝐀^i`$ potential gluon, which reproduce the $`1/|𝐤|`$ term. For the box and crossed box in Feynman gauge the $`1/|𝐤|`$ terms are reproduced by contributions that can be associated with the potential momentum regime. Thus, in Refs. and the $`1/|𝐤|`$ potential in Eq. (92) effectively exists at the scale $`m`$. In our approach off-shell potential gluons and soft quarks are integrated out, so the matching for the $`1/|𝐤|`$ potential is not simply given by the hard part of the QCD diagrams.
#### 4 Order $`v^0`$
For the order $`\alpha _s^2v^0`$ matching we will consider the direct and annihilation diagrams separately. The sum of the order $`\alpha _s^2v^0`$ terms in the direct scattering QCD diagrams in Eqs. (33) and (42) is
$`{\displaystyle \frac{i\alpha _s^2(\mu )}{m^2}}`$ $`(11)C_1\left[42𝐒^\mathrm{𝟐}+2\mathrm{ln}\left({\displaystyle \frac{k}{m}}\right){\displaystyle \frac{8}{3}}\mathrm{ln}\left({\displaystyle \frac{\lambda }{k}}\right)\right]`$ (93)
$`{\displaystyle \frac{i\alpha _s^2(\mu )}{m^2}}`$ $`(T^A\overline{T}^A)[{\displaystyle \frac{p^2}{k^2}}\{{\displaystyle \frac{31C_A}{9}}+{\displaystyle \frac{16C_A}{3}}\mathrm{ln}\left({\displaystyle \frac{\lambda }{\mu }}\right)+{\displaystyle \frac{38C_A}{3}}\mathrm{ln}\left({\displaystyle \frac{\mu }{k}}\right)\}`$ (101)
$`+\{2C_F3C_AC_d+({\displaystyle \frac{8C_F}{3}}+{\displaystyle \frac{2C_d}{3}}2C_A)\mathrm{ln}\left({\displaystyle \frac{\lambda }{\mu }}\right)`$
$`+({\displaystyle \frac{7C_d}{6}}{\displaystyle \frac{43C_A}{6}})\mathrm{ln}\left({\displaystyle \frac{\mu }{k}}\right)+({\displaystyle \frac{8C_F}{3}}+{\displaystyle \frac{31C_A}{6}}{\displaystyle \frac{C_d}{2}})\mathrm{ln}\left({\displaystyle \frac{\mu }{m}}\right)\}`$
$`+\mathrm{\Lambda }\left\{{\displaystyle \frac{31C_A}{6}}4C_F7C_A\mathrm{ln}\left({\displaystyle \frac{\mu }{k}}\right)4C_A\mathrm{ln}\left({\displaystyle \frac{\mu }{m}}\right)\right\}`$
$`+𝐒^\mathrm{𝟐}\left\{{\displaystyle \frac{C_d}{2}}{\displaystyle \frac{17C_A}{54}}{\displaystyle \frac{4C_F}{3}}{\displaystyle \frac{C_A}{9}}\mathrm{ln}\left({\displaystyle \frac{\mu }{k}}\right){\displaystyle \frac{7C_A}{3}}\mathrm{ln}\left({\displaystyle \frac{\mu }{m}}\right)\right\}`$
$`+𝐓\left\{{\displaystyle \frac{49C_A}{108}}{\displaystyle \frac{C_F}{3}}{\displaystyle \frac{5C_A}{18}}\mathrm{ln}\left({\displaystyle \frac{\mu }{k}}\right){\displaystyle \frac{C_A}{3}}\mathrm{ln}\left({\displaystyle \frac{\mu }{m}}\right)\right\}`$
$`n_fT_F\left\{{\displaystyle \frac{4}{3}}\mathrm{ln}\left({\displaystyle \frac{\mu ^2}{k^2}}\right)+{\displaystyle \frac{20}{9}}\right\}\left\{{\displaystyle \frac{p^2}{k^2}}{\displaystyle \frac{𝐒^2}{3}}{\displaystyle \frac{3\mathrm{\Lambda }}{2}}{\displaystyle \frac{𝐓}{12}}\right\}`$
$`{\displaystyle \frac{4T_F}{3}}\mathrm{ln}\left({\displaystyle \frac{\mu ^2}{m^2}}\right)\{{\displaystyle \frac{p^2}{k^2}}{\displaystyle \frac{𝐒^2}{3}}{\displaystyle \frac{3\mathrm{\Lambda }}{2}}{\displaystyle \frac{𝐓}{12}}\}+{\displaystyle \frac{4T_F}{15}}].`$
The effective theory contribution from order $`\alpha _s^2v^0`$ diagrams with soft gluons, ghosts or quarks is:
$`=`$ $`{\displaystyle \frac{i\alpha _s^2(\mu _S)}{m^2}}(11)\left[{\displaystyle \frac{14C_1}{3}}\mathrm{ln}\left({\displaystyle \frac{\mu _S}{k}}\right){\displaystyle \frac{C_1}{3}}\right]`$ (102)
$`+`$ $`{\displaystyle \frac{i\alpha _s^2(\mu _S)}{m^2}}(T^A\overline{T}^A)[C_A\mathrm{ln}\left({\displaystyle \frac{\mu _S}{k}}\right)\{{\displaystyle \frac{43}{6}}{\displaystyle \frac{38p^2}{3k^2}}+7\mathrm{\Lambda }+{\displaystyle \frac{𝐒^2}{9}}+{\displaystyle \frac{5𝐓}{18}}\}`$ (105)
$`{\displaystyle \frac{7C_d}{6}}\mathrm{ln}\left({\displaystyle \frac{\mu _S}{k}}\right)+{\displaystyle \frac{C_d}{12}}+{\displaystyle \frac{7C_A}{3}}{\displaystyle \frac{31C_Ap^2}{9k^2}}+{\displaystyle \frac{7C_A\mathrm{\Lambda }}{6}}{\displaystyle \frac{14C_A𝐒^\mathrm{𝟐}}{27}}+{\displaystyle \frac{13C_A𝐓}{108}}`$
$`+n_fT_F\{{\displaystyle \frac{4}{3}}\mathrm{ln}\left({\displaystyle \frac{\mu _S^2}{k^2}}\right)+{\displaystyle \frac{20}{9}}\}\{{\displaystyle \frac{p^2}{k^2}}{\displaystyle \frac{𝐒^2}{3}}{\displaystyle \frac{3\mathrm{\Lambda }}{2}}{\displaystyle \frac{𝐓}{12}}\}].`$
There are also non-zero order $`\alpha _s^2v^0`$ diagrams with an ultrasoft gluon and one insertion of the Coulomb potential. These diagrams were calculated in Ref. :
$`\text{}+\mathrm{}`$ $`=`$ $`{\displaystyle \frac{8i}{3m^2}}\alpha _s(\mu _S)\alpha _s(\mu _U)\mathrm{ln}\left({\displaystyle \frac{\mu _U}{\lambda }}\right)[C_1(11)`$ (107)
$`+(T\overline{T})\{C_F+{\displaystyle \frac{C_d}{4}}{\displaystyle \frac{3C_A}{4}}+{\displaystyle \frac{2p^2C_A}{k^2}}\}].`$
In Eq. (107) the ellipses denote all possible diagrams that are generated by attaching the ultrasoft gluon to two fermion legs. Subtracting the sum of Eq. (102) and (107) from Eq. (93) and setting $`\mu =\mu _S=\mu _U=m`$ gives the one-loop matching contribution for the order $`v^0`$ direct potentials. At order $`v^0`$ we find that all the infrared divergences in QCD are matched by infrared divergences from the effective theory graphs with ultrasoft gluons. All $`\mathrm{ln}(k)`$’s in the full theory are matched by $`\mathrm{ln}(k)`$’s in Eq. (102). The contributions to the matching coefficients at one-loop are:
(113)
The total order $`\alpha _s^2v^0`$ matching coefficients are given by adding the tree level values in Eq. (28) to the results in Eq. (113), and are summarized in Table I at the end of the paper.
Next consider the matching of the order $`\alpha _s^2v^0`$ annihilation contributions. Since there are no corresponding diagrams in the effective theory, the sum of the $`\alpha _s^2v^0`$ terms in Eq. (II A) directly give the matching coefficients. This sum is infrared finite. Matching at $`\mu =m`$, $`\nu =1`$ we find that the one-loop annihilation contributions to the potential coefficients are:
$`𝒱_{s,a}^{(T)}`$ $`=`$ $`\left({\displaystyle \frac{C_d}{4N_c}}2C_1\right)(i\pi +22\mathrm{ln}2)\alpha _s^2(m)`$ (115)
$`+{\displaystyle \frac{1}{N_c}}\left\{{\displaystyle \frac{109C_A}{36}}4C_F+{\displaystyle \frac{n_fT_F}{3}}\left(2\mathrm{ln}2{\displaystyle \frac{5}{3}}i\pi \right){\displaystyle \frac{8T_F}{9}}\right\}\alpha _s^2(m),`$
$`𝒱_{s,a}^{(1)}`$ $`=`$ $`\left\{{\displaystyle \frac{C_1}{N_c}}+{\displaystyle \frac{C_d(N_c^21)}{8N_c^2}}\right\}(i\pi +22\mathrm{ln}2)\alpha _s^2(m)`$ (117)
$`+{\displaystyle \frac{(N_c^21)}{2N_c^2}}\left\{{\displaystyle \frac{109C_A}{36}}4C_F+{\displaystyle \frac{n_fT_F}{3}}\left(2\mathrm{ln}2{\displaystyle \frac{5}{3}}i\pi \right){\displaystyle \frac{8T_F}{9}}\right\}\alpha _s^2(m),`$
$`𝒱_{2,a}^{(T)}`$ $`=`$ $`2\left({\displaystyle \frac{C_d}{4N_c}}2C_1\right)(i\pi +22\mathrm{ln}2)\alpha _s^2(m),`$ (118)
$`𝒱_{2,a}^{(1)}`$ $`=`$ $`2\left\{{\displaystyle \frac{C_1}{N_c}}+{\displaystyle \frac{C_d(N_c^21)}{8N_c^2}}\right\}(i\pi +22\mathrm{ln}2)\alpha _s^2(m).`$ (119)
The imaginary terms in these potentials contribute to the cross section for annihilation of a color octet heavy quark and anti-quark into light hadrons, and agree with the results of Ref. . The total annihilation contribution to the order $`\alpha _s^2v^0`$ matching coefficients are given by adding the tree level results in Eq. (31) to the results in Eq. (115), and are summarized in Table I.
If a different matching scale, $`\mu _h`$, had been used then the coefficients in Eqs. (113) and (115) would also depend on $`\mathrm{ln}(\mu _h/m)`$. Since the prediction for observables is independent of $`\mu _h`$ the most convenient choice, $`\mu _h=m`$, has been adopted.
In the threshold expansion, the full QCD diagram is the sum of hard, soft, ultrasoft and potential graphs. The soft, ultrasoft and potential graphs are the graphs in the effective theory up to renormalization effects such as the dependence on the scales $`\mu _S`$ and $`\mu _U`$. The matching conditions in Eq. (113) and (115) can also be computed from the hard part of the full theory graphs, and we have verified that this gives the same result as in Eqs. (113) and (115). The matching onto four-quark operators was first considered by Pineda and Soto in the context of pNRQCD . In their approach, the hard parts of the vertex correction and wavefunction renormalization are included in matching coefficients in the single heavy quark sector of the Lagrangian. The direct matching for four-quark operators is given by the hard part of the box and crossed-box. The hard part of our box and crossed-box agree with Ref. , except for the finite part of the $`\sigma _1\sigma _2`$ terms, which causes our $`𝒱_2^{(T)}`$ and $`𝒱_s^{(T)}`$ to disagree with theirs. This is related to the treatment of epsilon tensors and the non-relativistic reduction of matrix elements of spin operators in $`d`$ dimensions. We have chosen to take the lowest order term in the matrix element of $`\gamma ^{[\alpha }\gamma ^\sigma \gamma ^{\beta ]}\gamma _{[\alpha }\gamma _\tau \gamma _{\beta ]}`$ to be $`ϵ^{\alpha \beta \sigma k}ϵ_{\alpha \beta \tau k^{}}\sigma _1^k\sigma _2^k^{}`$, and have used the ’t Hooft-Veltmann scheme for the epsilon tensors so that $`ϵ_{\mu \nu \alpha \beta }`$ and $`ϵ^{ijk}`$ are only non-zero when the indices are in four and three dimensions respectively. This convention was used in both the soft loop integrals as well as when cross checking our result by computing the hard part of the box diagrams using the threshold expansion. This scheme dependence is related to the issue of evanescent operators . The annihilation results in Eq. (115) agree completely with Ref. .
Note that in a leading-log expansion the two-loop anomalous dimension is needed at the same time as the one-loop matching results. In the color-singlet channel the $`1/v^2`$ two-loop anomalous dimensions is known , since the running of the Coulomb potential is still given by the QCD $`\beta `$-function at this order. The result for the two-loop anomalous dimension for $`V^{(1)}`$ will be presented in another publication.
## III Terms in the potential which vanish on-shell.
For the scattering $`Q(p_1^0,𝐩)+\overline{Q}(p_2^0,𝐩)Q(p_3^0,𝐩^{})+\overline{Q}(p_4^0,𝐩^{})`$ consider adding
$`V^{(0)}`$ $`=`$ $`\left[𝒱_{\mathrm{\Delta }1}^{(T)}(T^A\overline{T}^A)+𝒱_{\mathrm{\Delta }1}^{(1)}(11)\right]{\displaystyle \frac{(p_3^0p_1^0)^2}{𝐤^4}}`$ (120)
$`+`$ $`\left[𝒱_{\mathrm{\Delta }2}^{(T)}(T^A\overline{T}^A)+𝒱_{\mathrm{\Delta }2}^{(1)}(11)\right]{\displaystyle \frac{(𝐩^2p^2)^2}{4m^2𝐤^4}}`$ (121)
to the potential in Eq. (26). Here $`𝐤=𝐩^{}𝐩`$ and by energy conservation $`p_3^0p_1^0=p_2^0p_4^0`$. On-shell the potentials in Eq. (120) vanish, since $`p_1^0=p_3^0`$, $`p_4^0=p_2^0`$, and $`𝐩^\mathrm{𝟐}=𝐩_{}^{}{}_{}{}^{2}`$. However, if we work off-shell, they are valid terms and in fact show up in many calculations that make use of time-ordered perturbation theory.
Matching to the tree level diagrams in Fig. 1 with $`p_1^0p_3^0`$ gives a contribution to the potentials in Eq. (120). In Feynman gauge one finds
$`𝒱_{\mathrm{\Delta }1}^{(T)}`$ $`=`$ $`4\pi \alpha _s(m),𝒱_{\mathrm{\Delta }1}^{(1)}=0,𝒱_{\mathrm{\Delta }2}^{(T)}=0,𝒱_{\mathrm{\Delta }2}^{(1)}=0,`$ (122)
while in Coulomb gauge one gets a different answer
$`𝒱_{\mathrm{\Delta }1}^{(T)}=0,𝒱_{\mathrm{\Delta }1}^{(1)}=0,𝒱_{\mathrm{\Delta }2}^{(T)}`$ $`=`$ $`4\pi \alpha _s(m),𝒱_{\mathrm{\Delta }2}^{(1)}=0.`$ (123)
Unlike the on-shell potentials discussed in the previous section, the matching conditions for off-shell potentials are gauge dependent. Using the expressions for the soft vertices in Feynman gauge in Appendix A, the one loop anomalous dimensions are
$`\nu {\displaystyle \frac{}{\nu }}𝒱_{\mathrm{\Delta }2}^{(T)}=2\beta _0\alpha _s(m\nu )^2,\nu {\displaystyle \frac{}{\nu }}𝒱_{\mathrm{\Delta }1}^{(T)}=\nu {\displaystyle \frac{}{\nu }}𝒱_{\mathrm{\Delta }1}^{(1)}=\nu {\displaystyle \frac{}{\nu }}𝒱_{\mathrm{\Delta }2}^{(1)}=0.`$ (124)
Including the potential in Eq. (120) modifies the matching condition for the $`1/|𝐤|`$ potentials and also makes this matching gauge dependent. This is because in the effective theory there are now two new order $`\alpha _s^2/v`$ diagrams:
(125)
$`\text{}+\text{}`$ $`=`$ $`{\displaystyle \frac{i𝒱_c^{(T)}𝒱_\mathrm{\Delta }^{(T)}}{32mk}}T^AT^B\overline{T}^A\overline{T}^B+{\displaystyle \frac{i𝒱_c^{(T)}𝒱_\mathrm{\Delta }^{(1)}}{32mk}}T^A\overline{T}^A+\mathrm{}.`$ (126)
In Eq. (125) the box labeled by $`𝒱_\mathrm{\Delta }`$ denotes an insertion of both operators in Eq. (120), the ellipses denote operators that vanish by the equations of motion (proportional to $`p_1^0𝐩^\mathrm{𝟐}/2m`$, $`p_3^0𝐩^{}{}_{}{}^{\mathrm{𝟐}}/2m`$ etc.), and we have defined
$`𝒱_\mathrm{\Delta }^{(T)}=𝒱_{\mathrm{\Delta }1}^{(T)}+𝒱_{\mathrm{\Delta }2}^{(T)},𝒱_\mathrm{\Delta }^{(1)}=𝒱_{\mathrm{\Delta }1}^{(1)}+𝒱_{\mathrm{\Delta }2}^{(1)}.`$ (128)
In Feynman gauge the matching conditions in Eq. (92) now become
$`𝒱_{kF}^{(T)}`$ $`=`$ $`\alpha _s^2(m)\left({\displaystyle \frac{3C_A}{4}}{\displaystyle \frac{C_d}{4}}\right),𝒱_{kF}^{(1)}=\alpha _s^2(m)C_1,`$ (129)
while in Coulomb gauge
$`𝒱_{kC}^{(T)}`$ $`=`$ $`C_A\alpha _s^2(m),𝒱_{kC}^{(1)}=0.`$ (130)
The $`1/|𝐤|`$ potential is often referred to as the non-abelian potential. From Eqs. (129) and (130), we see that the $`1/|𝐤|`$ potential only vanishes in QED if the potential is taken to include off-shell components and Coulomb gauge is used.
The result in Eq. (125) shows that in the off-shell matching potential one can make the replacements
$`V_\mathrm{\Delta }^{(T)}`$ $``$ $`V_\mathrm{\Delta }^{(T)}+\zeta ^{(T)},`$ (131)
$`V_\mathrm{\Delta }^{(1)}`$ $``$ $`V_\mathrm{\Delta }^{(1)}+\zeta ^{(1)},`$ (132)
$`V_k^{(T)}`$ $``$ $`V_k^{(T)}+{\displaystyle \frac{1}{32\pi ^2}}V_c^{(T)}\left[\zeta ^{(1)}{\displaystyle \frac{1}{4}}(C_A+C_d)\zeta ^{(T)}\right],`$ (133)
$`V_k^{(1)}`$ $``$ $`V_k^{(1)}+{\displaystyle \frac{1}{32\pi ^2}}V_c^{(T)}C_1\zeta ^{(T)},`$ (134)
for arbitrary $`\zeta ^{(T,1)}`$. For instance, at the matching scale taking $`\zeta ^{(1)}=0`$, $`\zeta ^{(T)}=8\pi \alpha _s(m)`$ effectively transforms the Feynman gauge result in Eqs. (122) and (129) into the Coulomb gauge result in Eqs. (123) and (130). Furthermore, taking $`\zeta ^{(1)}=0`$ and $`\zeta ^{(T)}=4\pi \alpha _s(m)`$ transforms the off-shell Coulomb gauge result into the on-shell result in section II. These transformations convert terms in the potential that are order $`\alpha _sv^0`$ to order $`\alpha _s^2/v`$. Similar transformations for the position space color singlet potentials have been pointed out previously in Refs. .
## IV The Quark-Quark Potential
The quark-quark potentials in the effective theory can be defined in the same way as the quark-antiquark potentials. The only difference is that the $`V^{(T)}`$ terms are now the coefficient of the $`T^AT^A`$ tensor, rather than the $`T^A\overline{T}^A`$ tensor. The computation of the quark-quark potential is almost identical to that of the quark-antiquark potential. The result can be obtained from the quark-antiquark potential by omitting the annihilation terms, and making the substitutions $`C_dC_d`$ and $`\overline{T}^AT^A`$. The change in sign of $`C_d`$ arises from the identities in Eq. (22). Replacing $`\overline{T}`$ by $`T`$ in these equations changes the sign of the $`C_d`$ terms.
The potential in the symmetric and antisymmetric $`QQ`$ color channels (the $`\mathrm{𝟔}`$ and $`\overline{\mathrm{𝟑}}`$ for $`SU(3)`$) are given by
$`\left[\begin{array}{c}V_{\mathrm{symmetric}}\\ V_{\mathrm{antisymmetric}}\end{array}\right]=\left[\begin{array}{ccc}1& & \frac{N_c1}{2N_c}\\ 1& & \frac{N_c+1}{2N_c}\end{array}\right]\left[\begin{array}{c}V_{11}\\ V_{TT}\end{array}\right].`$ (141)
The spin-$`1`$ $`QQ`$ combination is spin-symmetric and the spin-$`0`$ $`QQ`$ combination is spin-antisymmetric. For identical fermions in the initial state, for the symmetric spin-color states we must antisymmetrize the potential in the momenta, $`V=V(𝐩,𝐩^{})V(𝐩,𝐩^{})`$, and for the antisymmetric spin-color states we must symmetrize the potential in the momenta, $`V=V(𝐩,𝐩^{})+V(𝐩,𝐩^{})`$. This corresponds to including the crossed diagrams in the computation of the potentials.
## V QED
It is straightforward to obtain the QED potential from our results. For oppositely charged particles of charge $`\pm Q`$, the QED direct potential is given by $`Q^2(V_{11}V_{T\overline{T}})`$, where $`V_{11}`$ and $`V_{T\overline{T}}`$ are given by our QCD results with $`C_1=C_F=T_F=1`$, $`C_A=C_d=0`$, and $`\alpha _s\alpha `$. The on-shell $`1/|𝐤|`$ potential does not vanish in this limit. In the results for the annihilation potentials at the scale $`m`$ in Eqs. (31) and (115), explicit factors of $`N_c`$ were included in the Fierz transformation, so it is simplest to just separately list the QED limit of these results:
$`𝒱_{s,a}^{(T)}`$ $`=`$ $`(i\pi +22\mathrm{ln}2)\alpha ^2(m)Q^2,`$ (142)
$`𝒱_{s,a}^{(1)}`$ $`=`$ $`\pi \alpha (m)+\left\{{\displaystyle \frac{44}{9}}+{\displaystyle \frac{n_f}{3}}\left(2\mathrm{ln}2{\displaystyle \frac{5}{3}}i\pi \right)\right\}\alpha ^2(m)Q^2,`$ (143)
$`𝒱_{2,a}^{(T)}`$ $`=`$ $`2(i\pi +22\mathrm{ln}2)\alpha ^2(m)Q^2,`$ (144)
$`𝒱_{2,a}^{(1)}`$ $`=`$ $`0.`$ (145)
For $`e^+e^{}`$ we have $`n_f=0`$ and the terms in our $`v^0`$ potentials agree with Ref. .
For identical particles with charge $`Q`$, there is only the direct potential contribution, which is given by $`Q^2(V_{11}+V_{TT})`$, with $`C_1=C_F=T_F=1`$, $`C_A=C_d=0`$, and $`\alpha _s\alpha `$ as above. As discussed in Section IV, including the crossed diagrams gives a final potential that is symmetric or antisymmetric in the momenta depending on the symmetry of the spin state.
## VI Conclusion
We have computed the $`Q\overline{Q}`$ and $`QQ`$ scattering amplitudes in QCD to order $`v^2`$, and compared our results with previous calculations. We have also computed the scattering graphs in vNRQCD, and computed the matching condition at $`\mu =m`$ between the two theories. The matching potential was computed using on-shell matching, omitting terms which vanish by the equations of motion. The result was compared with approaches that include terms in the potential that vanish on-shell. The one-loop matching coefficients are summarized in Table I, and can be combined with the two-loop running to give the potential at next to leading log order. The computation of the heavy quark production current at next to leading logarithmic order uses these results and will be discussed in a subsequent publication .
We would like to thank A. Hoang and J. Soto for discussions. This work was supported in part by the Department of Energy under grant DOE-FG03-97ER40546.
## A Coefficients for the soft Lagrangian
The coefficient functions for the soft Lagrangian in Eq. (13) in Feynman gauge are:
$`U_{00}^{(0)}`$ $`=`$ $`{\displaystyle \frac{1}{q^0}},U_{0i}^{(0)}={\displaystyle \frac{(2p^{}2pq)^i}{(𝐩^{}𝐩)^2}},U_{i0}^{(0)}={\displaystyle \frac{(pp^{}q)^i}{(𝐩^{}𝐩)^2}},U_{ij}^{(0)}={\displaystyle \frac{2q^0\delta ^{ij}}{(𝐩^{}𝐩)^2}},`$ (A1)
$`U_{00}^{(1)}`$ $`=`$ $`{\displaystyle \frac{(p^{}+p)q}{2m(q^0)^2}}{\displaystyle \frac{(p^{}+p)q}{m(𝐩^{}𝐩)^2}}{\displaystyle \frac{ic_F𝝈[𝐪\times (pp^{})]}{m(𝐩^{}𝐩)^2}}+{\displaystyle \frac{(𝐩_{}^{}{}_{}{}^{2}𝐩^2)}{2m(𝐩^{}𝐩)^2}},`$ (A2)
$`U_{0i}^{(1)}`$ $`=`$ $`{\displaystyle \frac{(p+p^{})^i}{2mq^0}}+{\displaystyle \frac{ic_F(𝐪\times 𝝈)^i}{2mq^0}}+{\displaystyle \frac{q^0(p+p^{})^i}{2m(𝐩^{}𝐩)^2}}+{\displaystyle \frac{ic_Fq^0[(pp^{})\times 𝝈]^i}{2m(𝐩^{}𝐩)^2}},`$ (A3)
$`U_{i0}^{(1)}`$ $`=`$ $`{\displaystyle \frac{(p+p^{})^i}{2mq^0}}{\displaystyle \frac{ic_F[(pp^{}+q)\times 𝝈]^i}{2mq^0}}+{\displaystyle \frac{q^0(p+p^{})^i}{2m(𝐩^{}𝐩)^2}}+{\displaystyle \frac{ic_Fq^0[(pp^{})\times 𝝈]^i}{2m(𝐩^{}𝐩)^2}},`$ (A4)
$`U_{ij}^{(1)}`$ $`=`$ $`{\displaystyle \frac{ic_Fϵ^{ijk}𝝈^k}{2m}}+[2\delta ^{ij}𝐪^m+\delta ^{im}(2p^{}2pq)^j+\delta ^{jm}(pp^{}q)^i]`$ (A6)
$`\times \left[{\displaystyle \frac{(p+p^{})^m+ic_Fϵ^{mkl}(pp^{})^k𝝈^l}{2m(𝐩^{}p)^2}}\right],`$
$`U_{00}^{(2)}`$ $`=`$ $`{\displaystyle \frac{c_D(𝐩^{}𝐩)^2}{8m^2q^0}}+{\displaystyle \frac{c_Si𝝈(𝐩^{}\times p)}{4m^2q^0}}+{\displaystyle \frac{(𝐩𝐪)^2+(𝐩^{}𝐪)^2}{2m^2(q^0)^3}}+{\displaystyle \frac{(2c_D)(pp^{})q}{4m^2q^0}}`$ (A8)
$`+{\displaystyle \frac{(1c_D)𝐪^2}{4m^2q^0}},`$
$`U_{0i}^{(2)}`$ $`=`$ $`{\displaystyle \frac{[𝐩q(2p+q)^i+𝐩^{}q(2p^{}q)^i]}{4m^2(q^0)^2}}+{\displaystyle \frac{ic_F[𝐪\times 𝝈]^i(p+p^{})q}{4m^2(q^0)^2}}`$ (A11)
$`+{\displaystyle \frac{(c_D1)(𝐩p^{}+q)^i}{4m^2}}+{\displaystyle \frac{(2p^{}2pq)^i}{4m^2}}\left[{\displaystyle \frac{c_D}{2}}{\displaystyle \frac{c_Si𝝈(𝐩^{}\times p)}{(𝐩^{}𝐩)^2}}\right]`$
$`+{\displaystyle \frac{(𝐩_{}^{}{}_{}{}^{2}𝐩^2)}{4m^2(𝐩^{}𝐩)^2}}\left[(𝐩+𝐩^{})^i+c_Fi𝝈(𝐩^{}\times p)\right]{\displaystyle \frac{(2p^{}2pq)^i(𝐩_{}^{}{}_{}{}^{2}𝐩^2)^2}{4m^2(𝐩^{}𝐩)^4}},`$
$`U_{i0}^{(2)}`$ $`=`$ $`{\displaystyle \frac{[𝐩q(p+p^{}+q)^i+𝐩^{}q(p+p^{}q)^i]}{4m^2(q^0)^2}}{\displaystyle \frac{ic_F[(pp^{}+q)\times 𝝈]^i(p+p^{})q}{4m^2(q^0)^2}}`$ (A14)
$`+{\displaystyle \frac{(c_D1)𝐪^i}{4m^2}}+{\displaystyle \frac{(pp^{}q)^i}{4m^2}}\left[{\displaystyle \frac{c_D}{2}}{\displaystyle \frac{c_Si𝝈(𝐩^{}\times p)}{(𝐩^{}𝐩)^2}}\right]`$
$`{\displaystyle \frac{(𝐩_{}^{}{}_{}{}^{2}𝐩^2)}{2m^2(𝐩^{}𝐩)^2}}\left[(𝐩+𝐩^{})^i+c_Fi𝝈(𝐩^{}\times p)\right]{\displaystyle \frac{(pp^{}q)^i(𝐩_{}^{}{}_{}{}^{2}𝐩^2)^2}{4m^2(𝐩^{}𝐩)^4}},`$
$`U_{ij}^{(2)}`$ $`=`$ $`{\displaystyle \frac{(p+p^{})^i(p+p^{})^j}{4m^2q^0}}+{\displaystyle \frac{c_F^2(pp^{})q\delta ^{ij}}{4m^2q^0}}+{\displaystyle \frac{ic_F(p+p^{})^j[(pp^{})\times 𝝈]^i}{4m^2q^0}}`$ (A17)
$`{\displaystyle \frac{ic_Fϵ^{ijk}𝐪^k𝝈(p+p^{})}{4m^2q^0}}+{\displaystyle \frac{ic_Fϵ^{ijk}𝝈^k(p+p^{})q}{4m^2q^0}}+{\displaystyle \frac{(1c_F^2)𝐪^i(𝐩p^{}+q)^j}{4m^2q^0}}`$
$`+{\displaystyle \frac{c_F^2𝐪^2\delta ^{ij}}{4m^2q^0}}{\displaystyle \frac{i\delta ^{ij}q^0c_Si𝝈(𝐩^{}\times p)}{2m^2(𝐩^{}𝐩)^2}}{\displaystyle \frac{2q^0\delta ^{ij}(𝐩_{}^{}{}_{}{}^{2}𝐩^2)^2}{4m^2(𝐩^{}𝐩)^4}},`$
$`W_{\mu \nu }^{(0)}`$ $`=`$ $`0,`$ (A18)
$`W_{00}^{(1)}`$ $`=`$ $`{\displaystyle \frac{1}{2m}}+{\displaystyle \frac{(pp^{})q}{2m(q^0)^2}},W_{0i}^{(1)}={\displaystyle \frac{(pp^{}+q)^i}{2mq^0}},W_{i0}^{(1)}={\displaystyle \frac{𝐪^i}{2mq^0}},W_{ij}^{(1)}={\displaystyle \frac{\delta ^{ij}}{2m}},`$ (A19)
$`Y^{(0)}`$ $`=`$ $`{\displaystyle \frac{q^0}{(𝐩^{}𝐩)^2}},Y^{(1)}={\displaystyle \frac{𝐪(p+p^{})+ic_F𝝈[q\times (pp^{})]}{2m(𝐩^{}𝐩)^2}},`$ (A20)
$`Y^{(2)}`$ $`=`$ $`{\displaystyle \frac{c_Dq^0}{8m^2}}{\displaystyle \frac{c_Si𝝈(𝐩^{}\times p)q^0}{4m^2(𝐩^{}𝐩)^2}},`$ (A21)
$`Z_0^{(0)}`$ $`=`$ $`{\displaystyle \frac{1}{(𝐩^{}𝐩)^2}},Z_i^{(0)}=0,Z_0^{(1)}=0,Z_i^{(1)}={\displaystyle \frac{(𝐩+p^{})^iic_F[(𝐩p^{})\times 𝝈]^i}{2m(𝐩^{}𝐩)^2}},`$ (A22)
$`Z_0^{(2)}`$ $`=`$ $`{\displaystyle \frac{1}{4m^2}}\left[{\displaystyle \frac{c_D}{2}}{\displaystyle \frac{c_Si𝝈(𝐩^{}\times p)}{(𝐩^{}𝐩)^2}}\right],Z_i^{(2)}=0.`$ (A23)
These expressions differ from those in Ref. by terms proportional to $`𝐩_{}^{}{}_{}{}^{2}𝐩^2`$, and the values in Eq. (A1) are the complete on-shell expressions. The $`𝐩_{}^{}{}_{}{}^{2}𝐩^2`$ terms were not needed in calculating the one-loop running of the on-shell order $`v^0`$ potentials in Ref. . In section III, these terms were used to compute the running of the off-shell potential in Eq. (124), and this result depends on the fact that we used the on-shell soft Lagrangian. The coefficients in Eq. (A1) can be written in a manifestly Hermitian form, however we have instead used $`𝐪^{}=𝐪+𝐩𝐩^{}`$ and $`q^{}{}_{}{}^{0}=q^0+p^0p^{}^0`$ to eliminate $`q^{}`$ since this form is more convenient for calculations. Reparameterization invariance can be used to eliminate $`c_S`$ by the relation $`c_S=2c_F1`$. The running of $`c_D`$ and $`c_F`$ are given in Refs. .
## B Integrals
The effective theory integrals that appear in Eqs. (83) and (II B 3) are
$`I_0`$ $`=`$ $`{\displaystyle d^3q\frac{1}{[(𝐪p)^2+\lambda ^2][(𝐪p^{})^2+\lambda ^2][𝐪^\mathrm{𝟐}𝐩^2iϵ]}}={\displaystyle \frac{i}{8\pi |𝐩|𝐤^2}}\mathrm{ln}\left({\displaystyle \frac{\lambda ^2}{𝐤^2}}\right),`$ (B1)
$`I_F`$ $`=`$ $`{\displaystyle d^3q\frac{1}{(𝐪p)^2(𝐪p^{})^2}}={\displaystyle \frac{1}{8|𝐤|}},`$ (B2)
$`I_P`$ $`=`$ $`{\displaystyle d^3q\frac{1}{[(𝐪p^{})^2+\lambda ^2][𝐪^\mathrm{𝟐}𝐩^2iϵ]}}={\displaystyle \frac{1}{16|𝐩|}}+{\displaystyle \frac{i}{8\pi |𝐩|}}\mathrm{ln}\left({\displaystyle \frac{2|𝐩|}{\lambda }}\right),`$ (B3)
and
$`{\displaystyle d^3q\frac{𝐪^i}{[(𝐪p)^2+\lambda ^2][(𝐪p^{})^2+\lambda ^2][𝐪^\mathrm{𝟐}𝐩^2iϵ]}}=(𝐩^{}+𝐩)^iI_A,`$ (B4)
$`{\displaystyle d^3q\frac{𝐪^i𝐪^j}{[(𝐪p)^2+\lambda ^2][(𝐪p^{})^2+\lambda ^2][𝐪^\mathrm{𝟐}𝐩^2iϵ]}}=\delta ^{ij}I_B+(𝐩^{}𝐩)^i(𝐩^{}𝐩)^jI_C`$ (B5)
$`+(𝐩^{}+𝐩)^i(𝐩^{}+𝐩)^jI_D,`$ (B6)
where $`𝐤=𝐩^{}𝐩`$ and
$`I_A`$ $`=`$ $`{\displaystyle \frac{(2|𝐩||𝐤|𝐤^2)\pi 2i𝐤^2\mathrm{ln}\left(\frac{2|𝐩|}{\lambda }\right)4i𝐩^2\mathrm{ln}\left(\frac{\lambda ^2}{𝐤^2}\right)}{16\pi |𝐩|𝐤^2(4𝐩^2𝐤^2)}},`$ (B7)
$`I_B`$ $`=`$ $`{\displaystyle \frac{(2|𝐩||𝐤|)\pi +2i|𝐩|\mathrm{ln}\left(\frac{4𝐩^2}{𝐤^2}\right)}{16\pi (4𝐩^2𝐤^2)}},`$ (B8)
$`I_C`$ $`=`$ $`{\displaystyle \frac{(2|𝐩||𝐤|𝐤^2)\pi 2i(4𝐩^2𝐤^2)2i𝐤^2\mathrm{ln}\left(\frac{2|𝐩|}{\lambda }\right)4i𝐩^2\mathrm{ln}\left(\frac{\lambda ^2}{𝐤^2}\right)}{32\pi |𝐩|𝐤^2(4𝐩^2𝐤^2)}},`$ (B9)
$`I_D`$ $`=`$ $`{\displaystyle \frac{1}{32\pi |𝐩|𝐤^2(4𝐩^2𝐤^2)^2}}[(2|𝐩||𝐤|)^2(4|𝐩|+|𝐤|)|𝐤|\pi +2i𝐤^2(4𝐩^2𝐤^2)`$ (B11)
$`2i𝐤^2(12𝐩^2𝐤^2)\mathrm{ln}\left({\displaystyle \frac{2|𝐩|}{\lambda }}\right)4i𝐩^2(4𝐩^2+𝐤^2)\mathrm{ln}\left({\displaystyle \frac{\lambda ^2}{𝐤^2}}\right)].`$
|
warning/0003/nucl-ex0003004.html
|
ar5iv
|
text
|
# Isospin dependence of isobaric ratio Y(3H)/Y(3He) and its relation to temperature
## Abstract
A dependence of the isobaric ratio Y(<sup>3</sup>H)/Y(<sup>3</sup>He) on the N/Z ratio of the reconstructed quasiprojectile for the reaction of <sup>28</sup>Si beam with <sup>112,124</sup>Sn targets at two different projectile energies of 30 and 50 MeV/nucleon is presented. We demonstrate a linear dependence of the observable ln(Y(<sup>3</sup>H)/Y(<sup>3</sup>He)) on the N/Z ratio of the quasiprojectile and show the dependence of the slope on the reconstructed excitation energy of the quasiprojectile. We relate this slope dependence at a given excitation energy to the temperature of the fragmenting system. Using the model assumptions of the statistical multifragmentation model, a method of temperature determination is proposed. A caloric curve is constructed and compared to the result of the double isotope ratio method for the same set of the data and to the results of other studies.
, , , , , , , and
thanks: Phone: (979)-845-1411, fax: (979)-845-1899, e-mail: veselsky@comp.tamu.eduthanks: On leave of absence from Institute of Physics of SASc, Bratislava, Slovakiathanks: Present address: Brookhaven National Laboratory, Brookhaven, New York, USAthanks: Present address: Mallinckrodt Institute of Radiology, St. Louis, Missouri, USAthanks: Present address: Microcal Software Inc, One Roundhouse Plaza, Northampton, MA 01060, USA
Multifragmentation of highly excited nuclei has been subject of significant interest for many years. Since the nuclei are two-component systems consisting of protons and neutrons, the influence of isospin on the disassembly of hot nuclei is a question of general interest. Experimental studies have demonstrated the influence of the isospin of projectile and target on the isotopic yields of fragments . Theoretical framework for fragmentation of asymmetric systems have been developed and a possible phase transition into an isospin-symmetric liquid and an isospin-asymmetric gas phase have been suggested. In our recent work we presented the characteristics of the quasiprojectiles with $`Z=1215`$, reconstructed from the fully isotopically resolved fragments with $`Z_f5`$, detected in the forward angles using the FAUST multidetector array at Cyclotron Institute of Texas A&M University. The distributions of the quasiprojectile velocity, mass, charge and excitation energy were obtained in the reaction of a <sup>28</sup>Si beam with <sup>112,124</sup>Sn targets at projectile energies 30 and 50 MeV/nucleon. We demonstrated that these experimental observables of the quasiprojectile can be reproduced satisfactorily by hybrid simulations, using a code implementing the model of deep inelastic transfer for a calculation of the properties of excited quasiprojectiles and the SMM code for a description of multifragmentation of excited quasiprojectiles. Further details can be found in . The level of an agreement obtained implies that the influence of the non-statistical effects caused by violent collisions of the projectile and the target on the properties of detected quasiprojectiles can be practically excluded for the selected events. The obtained set of data allows further study of thermal quasiprojectile multifragmentation, especially the study of the influence of the quasiprojectile isospin on the properties of fragmenting system. This data is of specific interest because the isospin of the system which actually undergoes multifragmentation is known with good precision. The first part of the study was presented in our recent work where multiplicities and N/Z ratios of light charged particles ( LCPs ) and intermediate mass fragments ( IMFs ) were studied. Different trends of multiplicities and an inhomogeneous distribution of isospin between light charged particles and intermediate mass fragments were observed. The multiplicity of LCPs increases rapidly with proton excess, while the multiplicity of IMFs increases slowly with increasing neutron number. For quasiprojectiles with large proton excess an inhomogeneous distribution of isospin into proton-rich LCPs and more symmetric IMFs occurs. Furthermore, dependences of the isobaric ratio Y(<sup>3</sup>H)/Y(<sup>3</sup>He) on the N/Z ratio ( as a variable closely connected to isospin ) and on the excitation energy of the reconstructed quasiprojectiles were presented in . In this work we will present a detailed study of the trends of the isobaric ratio Y(<sup>3</sup>H)/Y(<sup>3</sup>He). The relationship of Y(<sup>3</sup>H)/Y(<sup>3</sup>He) to thermodynamical observables will be discussed.
The dependences of the isobaric ratio Y(<sup>3</sup>H)/Y(<sup>3</sup>He) on the N/Z ratio of the quasiprojectile ( $`N/Z_{QP}`$ ) are shown in Fig. 1 for the reaction of a <sup>28</sup>Si beam with <sup>112,124</sup>Sn targets at projectile energies of 30 and 50 MeV/nucleon. The quasiprojectiles with $`Z`$ = 12 - 15, reconstructed from the fully isotopically resolved charged fragments with $`Z_f5`$, are selected ( for details of the experiment see ). Various symbols represent the experimental data for different targets at each beam energy. The data within each set form nearly straight lines thus implying an exponential dependence of the isobaric ratio Y(<sup>3</sup>H)/Y(<sup>3</sup>He) on $`N/Z_{QP}`$. For a given projectile energy, the Y(<sup>3</sup>H)/Y(<sup>3</sup>He) dependence on $`N/Z_{QP}`$ is practically identical for the two different targets <sup>112,124</sup>Sn. This shows that the isotopic trends are determined by the isospin of the excited system which fragments and the trends obtained by comparison of the data for different projectile-target systems describe only the overall trends. A detailed discussion of this issue is in our previous work . For the two different projectile energies 30 and 50 MeV/nucleon the Y(<sup>3</sup>H)/Y(<sup>3</sup>He) dependence exhibits a different slope. In the above mentioned work , the distributions of the apparent excitation energy ( $`E_{app}^{}`$ ) are practically identical for the different targets at a given projectile energy and shift to the increasing values of $`E_{app}^{}`$ with an increasing beam energy. The mean values of $`E_{app}^{}`$ are 101.2 and 102.3 MeV for <sup>112</sup>Sn and <sup>124</sup>Sn targets at the projectile energy 30 MeV/nucleon and 142.8 MeV for both targets at the projectile energy 50 MeV/nucleon. Such a similarity may imply that the slope of the dependence of isobaric ratio Y(<sup>3</sup>H)/Y(<sup>3</sup>He) on $`N/Z_{QP}`$ correlates with the mean excitation energy of the quasiprojectile. Preliminary indications of this dependence have been reported in ref. . This assumption may be examined by extracting the isobaric ratio Y(<sup>3</sup>H)/Y(<sup>3</sup>He) for different bins of the apparent excitation energy of the quasiprojectile.
Fig. 2 shows the dependences of the isobaric ratio Y(<sup>3</sup>H)/Y(<sup>3</sup>He) on $`N/Z_{QP}`$ for the nine bins of the apparent excitation energy per mass unit of the quasiprojectile ( $`ϵ_{app}`$ ). The data for both targets and projectile energies were combined to increase the statistics. This is possible because for a given excitation energy bin, the Y(<sup>3</sup>H)/Y(<sup>3</sup>He) dependences agree within the statistical deviations. The squares represent the experimental data and the lines are the fits obtained using the MINUIT minimization package . As one can see on Fig. 2, linearity is the overall feature of the logarithmic plots of the Y(<sup>3</sup>H)/Y(<sup>3</sup>He) ratio in all excitation energy bins and is especially significant in the excitation energy bins with a high statistics. The slopes are steepest at low excitation energies and become flatter with increasing excitation energy.
Since the experimental dependences of the isobaric ratio Y(<sup>3</sup>H)/Y(<sup>3</sup>He) on $`N/Z_{QP}`$ correlate with $`ϵ_{app}`$ and are linear on a logarithmic scale ( and thus depend exponentially on $`N/Z_{QP}`$ ), it would be of interest to relate the slope to the thermodynamical observables of the quasiprojectile multifragmentation. As a starting point we chose the statistical multifragmentation model ( SMM ) which was employed in the simulation successfully describing our data . In the macrocanonical approach, the SMM gives the mean number of the emitted fragments of a given $`N_k`$ and $`Z_k`$ as
$$<M_{N_kZ_k}>=g_{N_kZ_k}\frac{V_f}{\lambda _T^3}A^{3/2}\mathrm{exp}(\frac{F_{N_kZ_k}(T,V)\mu _nN_k\mu _pZ_k}{T})$$
(1)
where $`T`$ is the freeze-out temperature, $`V`$ is the freeze-out volume, $`g_{N_kZ_k}`$ is the degeneracy of the fragment ground state, $`V_f`$ is the free volume available for translational motion of the fragment, $`\lambda _T`$ is the thermal wavelength of the nucleon, $`F_{N_kZ_k}(T,V)`$ is the internal free energy of the fragment and $`\mu _n`$ and $`\mu _p`$ are the chemical potentials for neutrons and protons. The assumption of chemical equilibrium is not incorporated into SMM and the chemical potentials $`\mu _n`$ and $`\mu _p`$ are used as a Lagrange multipliers ensuring the conservation of the mean neutron and proton number. When considering the isospin dependence of the experimental Y(<sup>3</sup>H)/Y(<sup>3</sup>He) ratio for a given excitation energy bin, the exponential dependence suggests a possibility to determine a single temperature for each bin as a simplest possible interpretation. The isobaric ratio Y(<sup>3</sup>H)/Y(<sup>3</sup>He) then can be determined as
$$\frac{Y(^3H)}{Y(^3He)}=\frac{<M_{21}>}{<M_{12}>}=\mathrm{exp}(\frac{F_{21}(T,V)F_{12}(T,V)\mu _n+\mu _p}{T})$$
(2)
where we used the fact that $`V_f`$ and $`\lambda _T`$ are the same for the whole partition and that the ground state spins of both fragments are equal.
We approximate the $`\mu _n\mu _p`$ by the difference in the neutron ( $`S_n`$ ) and proton ( $`S_p`$ ) ground state separation energy. According to the model of a nuclear Fermi gas at zero temperature, $`S_n`$ and $`S_p`$ can be identified with the chemical potentials for protons and neutrons in the ground state. When introducing finite temperature, the Fermi gas model predicts only a moderate change of the chemical potential for temperatures well below the Fermi energy. Thus the approximation of chemical potentials by separation energies can be still considered valid for highly excited thermally equilibrated nuclei and determines the correct chemical potentials prior to multifragmentation. During multifragmentation the system expands which creates two different trends. When considering the system as Fermi gas expanding to the freeze-out density ( typically 0.4$`\rho _0`$ for the present case according to the formula used in the SMM ), the values of the Fermi energy decrease by more than 40 %. The same behavior can be expected also for $`\mu _n\mu _p`$. On the other hand, chemical potentials of nucleons are influenced by the process of multifragmentation. As reported in our previous work , the IMFs are more isospin-symmetric than light charged particles. Within statistical multifragmentation model, the isobaric ratio Y(<sup>3</sup>H)/Y(<sup>3</sup>He) can be identified with the ratio of mean densities of free neutrons and protons for a given fragment partition ( as one can see after dividing equation 1 by freeze-out volume ). As follows from Figs. 1,2, such ratios differ dramatically from the starting values and one can expect also the values of $`\mu _n\mu _p`$ significantly higher than those of an expanded Fermi gas. When assuming that both effects have similar magnitude, the approximation used can be considered as a reasonable estimate of the actual values. For the full set of data used in Fig.2, $`\mu _n\mu _p`$ can be approximated as
$$\mu _n\mu _pS_pS_n=(43.44\pm 0.97)+(38.57\pm 0.94)N/Z_{QP}\text{ [MeV]. }$$
(3)
Experimental mass excesses have been used for the evaluation. The dependence of $`S_nS_p`$ on the N/Z ratio of the quasiprojectile can be considered linear with good precision. The difference in the internal free energy of the fragments <sup>3</sup>H and <sup>3</sup>He, according to the evaluation formulas given in , is of the order of a few hundred keV on a broad range of $`N/Z_{QP}`$ and changes much more slowly than $`\mu _n\mu _p`$. In this paper we neglect its weak dependence on $`N/Z_{QP}`$ and treat it as a constant independent on $`N/Z_{QP}`$. Within our approximation, the resulting expression for the isobaric ratio Y(<sup>3</sup>H)/Y(<sup>3</sup>He) will be
$$\mathrm{ln}(Y(^3\text{H})/Y(^3\text{He}))=\mathrm{ln}(K(T))+(\mu _n\mu _p)/T$$
(4)
where $`K(T)`$ is a proportionality factor dependent on the temperature but independent of the N/Z ratio of the fragmenting system. According to the evaluation formulas for the free energy given in , one may expect the values of proportionality factor $`K(T)`$ to be close to unity.
Since both ln(Y(<sup>3</sup>H)/Y(<sup>3</sup>He)) and $`\mu _n\mu _p`$ depend linearly on $`N/Z_{QP}`$ ( as may be seen from Fig. 2 and equation 3 ) and since we choose the value of $`\mathrm{ln}(K(T))`$ to be a constant, independent of $`N/Z_{QP}`$, the expressions on both sides of equation 4 are in fact first order polynomials of $`N/Z_{QP}`$. Then the solution is obvious, constants and coefficients of first order ( slopes ) should be equal on both sides of the equation. This allows a determination not only of the temperature ( as a ratio of the slopes ) but also of ln($`K(T)`$) from the comparison of the zero order coefficients ( constants ) using the extracted temperature. The resulting values of $`T`$ and $`\mathrm{ln}(K(T))`$ are given in Fig. 3 and Fig. 4 for the different quasiprojectile excitation energy bins. The numerical values of $`T`$ and $`\mathrm{ln}(K(T))`$ are given in the Table 1 along with the parameters of the linear fits of ln(Y(<sup>3</sup>H)/Y(<sup>3</sup>He)) given in Fig. 2. The dependence of the temperature on the excitation energy ( caloric curve ) given in Fig. 3 ( solid squares ) is compared to the experimental caloric curve obtained for the same set of the data by the double isotope ratio method for the thermometer d,t/<sup>3</sup>He,<sup>4</sup>He ( the dashed line indicates the double isotope ratio temperature and the solid lines indicate the statistical errors ). The formula
$$T_{\mathrm{d},\mathrm{t}/^3\mathrm{He},^4\mathrm{He}}=14.31\mathrm{MeV}/\mathrm{ln}(1.59\frac{Y({}_{}{}^{2}\mathrm{H})/Y({}_{}{}^{3}\mathrm{H})}{Y({}_{}{}^{3}\mathrm{He})/Y({}_{}{}^{4}\mathrm{He})})$$
(5)
was used for the calculations. The agreement between the two plots is reasonable. Both methods give the value of the temperature between 5 and 7 MeV for the apparent excitation energies above 5 MeV/nucleon. Similar values of the temperature are reported by other experimental studies . The values of the temperature predicted by the SMM are also in the same range. The values of $`\mathrm{ln}(K(T))`$ are between 0 and 1 and suggest the values of $`K(T)`$ between 1 and 3, which are in good agreement with the values of the SMM. The dependence of $`\mathrm{ln}(K(T))`$ on the temperature is quite weak, which is also consistent with the expectations. Large relative errors of the $`\mathrm{ln}(K(T))`$ values are caused by the subtraction of numbers of similar magnitude. The agreement between temperatures determined by two different methods shows that the assumptions made for $`\mu _n\mu _p`$ reflect the physical trends which take place in the freeze-out configuration.
Since the neutron emission from the quasiprojectile is not included in the data presented here, we tried to estimate its influence. Using a backtracing procedure described in , we estimated the initial values of the excitation energy of a hot quasiprojectile at the stage where it also includes the neutrons that are later emitted. The mean value of the excitation energy taken away by a neutron emission ranges from 0.5 MeV/nucleon for the highest bin of $`ϵ_{app}`$ to 1.5 MeV/nucleon for the lowest bin. This applies to both methods of the temperature determination presented here and implies that the events represent a still broader range of the initial excitation energies of the quasiprojectile. Since the multiplicity of the charged fragments decreases with decreasing $`ϵ_{app}`$, the decrease of the determined temperatures in the lowest bins of $`ϵ_{app}`$ may be a signature of the onset of a low energy deexcitation mode where several neutrons and light charged particles are emitted prior to the breakup of the partially cooled residue into two massive fragments ( such a mass distribution is observed in the channels with 3 and 4 charged fragments ). The temperatures determined thus become mean values representing the range of temperatures at which hydrogen and helium isotopes are emitted during the deexcitation cascade. The use of <sup>3</sup>He for the evaluation of the isobaric ratio Y(<sup>3</sup>H)/Y(<sup>3</sup>He) could be affected by the so-called <sup>3</sup>He-puzzle ( see ), nevertheless it appears that the <sup>3</sup>He-anomaly is not relevant for peripheral collisions induced by light projectile nuclei.
We presented the dependence of the isobaric ratio Y(<sup>3</sup>H)/Y(<sup>3</sup>He) on $`N/Z_{QP}`$ for the reaction of a <sup>28</sup>Si beam on <sup>112,124</sup>Sn targets at two different projectile energies 30 and 50 MeV/nucleon. We demonstrated a linear dependence of ln(Y(<sup>3</sup>H)/Y(<sup>3</sup>He)) on $`N/Z_{QP}`$ and the dependence of the slope on the excitation energy of the quasiprojectile. Using the model assumptions of the statistical multifragmentation model, we demonstrated the possibility of relating the slope of the dependence of ln(Y(<sup>3</sup>H)/Y(<sup>3</sup>He)) on $`N/Z_{QP}`$ to the temperature at a given excitation energy. We constructed a caloric curve and demonstrated its equivalence to the results of the double isotope ratio method. Our results show that the temperatures extracted from the isospin dependence of Y(<sup>3</sup>H)/Y(<sup>3</sup>He) are consistent with the temperature of the freeze-out configuration where thermal equilibration occurs. The approximation used for the evaluation of the difference of neutron and proton chemical potentials is consistent with significant changes of densities of neutrons and protons during multifragmentation.
The authors wish to thank the Cyclotron Institute staff for the excellent beam quality. This work was supported in part by the NSF through grant No. PHY-9457376, the Robert A. Welch Foundation through grant No. A-1266, and the Department of Energy through grant No. DE-FG03-93ER40773. M. V. was partially supported through grant VEGA-2/5121/98.
|
warning/0003/cond-mat0003328.html
|
ar5iv
|
text
|
# Multifractal behavior of linear polymers in disordered media
## I Introduction
The question of how linear polymers behave in a disordered medium has attracted much attention in recent years. The problem is not only interesting from a theoretical point of view, but may also be relevant for understanding transport properties of polymeric chains in porous media, such as in enhanced oil recovery, gel electrophoresis, gel permeation chromatography, etc. . In this context, it is useful to learn about the static or conformational properties of linear chains, modelled by self-avoiding walks (SAWs), in the presence of quenched disorder, e.g. how the surrounding structural disorder influences their spatial configuration. As a quite general model of a random medium, percolation may be considered as the paradigm for a broad class of disordered systems and has therefore been mostly used so far.
We are interested in how the statistical behavior of SAWs on percolation clusters at criticality ($`p=p_\mathrm{c}`$) differs from their behavior on regular lattices. While the values of the exponents for SAWs on regular lattices are well established , there is no complete agreement about their values on percolation clusters at $`p_\mathrm{c}`$ . Here, we study (i) the so-called effective coordination number of the cluster, where contradicting results have been reported using different numerical techniques. Next, we consider (ii) the enhancement exponent $`\gamma `$ and (iii) the exponents $`\nu _r`$ and $`\nu _{\mathrm{}}`$ characterising the end-to-end distance of SAWs in the $`r`$\- and $`\mathrm{}`$-space metrics. Finally, we determine (iv) the values of the critical exponents describing the corresponding structural distribution functions.
We concentrate on SAWs on percolation clusters at $`p_\mathrm{c}`$ in two and three dimensions. In the literature, two distinct methods have been used for evaluating SAWs: Exact enumeration (EE) and Monte Carlo (MC) simulation. In the EE technique, all SAW configurations on a given cluster are taken into account, but only relatively short chains can be evaluated. In a MC simulation, longer chains can be studied, but inherently the ensemble of configurations remains incomplete. Here, we use the EE technique in combination with an appropriate finite-size scaling procedure to determine the relevant exponents. Since ‘infinitely’ long chains can only exist on the backbone of the cluster, where dangling ends are absent on all length scales, we study the SAWs directly on the backbone. This enables us to generate longer chains on a given cluster and to average over a larger set of different cluster configurations.
Specifically, we enumerate all possible SAW configurations of $`N`$ steps for a single backbone and study different moments of the total number of SAWs and of their end-to-end distance by averaging over many different backbone configurations. Our analysis shows that the critical exponents $`\nu _r`$ and $`\nu _{\mathrm{}}`$ do not depend on the order $`q`$ of the moments, while the enhancement exponents and the effective coordination numbers do depend on $`q`$, leading to multifractal behavior. In particular we find that the first moment of the effective coordination number $`\mu _1`$ satisfies $`\mu _1=p_\mathrm{c}\mu `$, where $`\mu `$ is the effective coordination number of the underlying regular lattice, resolving previous controversies. The mean structural distribution functions for the end-to-end distance after $`N`$ steps, both in Euclidean and topological space, are obtained numerically, supporting the expected scaling forms .
The paper is organized as follows: In Section II, we briefly review the main relevant properties of SAWs on regular lattices to illustrate the different numerical procedures employed in this work. In Section III, we present results for the total number and the mean end-to-end distance of SAWs on the backbone of the incipient percolation cluster. The corresponding distribution functions of the end-to-end distance and their scaling behavior, in Euclidean and topological space, are also discussed. Finally, in Section IV, we summarize our main results.
## II SAWs on regular lattices reviewed
In this section we illustrate the different numerical techniques we use in the following by briefly reviewing the main results for SAWs on regular lattices. The main idea is to show that our finite-size scaling employed in the later sections enables us to obtain quite accurate estimates for the critical exponents based on EE results for relatively short chains. Here, we consider the case $`d=2`$, which is particularly suitable since many results are known exactly.
### A Total number of SAW configurations $`C_N`$
The total number $`C_N`$ of SAW configurations of $`N`$ steps behaves as
$$C_N=A\mu ^NN^{\gamma 1}$$
(1)
where $`\mu `$ is the effective coordination number of the lattice, $`\gamma `$ is the universal enhancement exponent, and $`A`$ is a constant. To determine $`\mu `$, $`\gamma `$, and $`A`$, we choose to study the behavior of the quantity
$$\frac{\mathrm{ln}C_N}{N}=\frac{\mathrm{ln}A}{N}+\mathrm{ln}\mu +(\gamma 1)\frac{\mathrm{ln}N}{N}$$
(2)
as a function of $`N`$. Fig. 1 shows for the square lattice, that the values for $`\mu `$ and $`\gamma `$ obtained by fitting the EE data using Eq. (2) agree well with the accepted values reported in literature (see Table I).
### B Mean end-to-end distance and structural distribution function
The root mean square end-to-end distance of SAWs of $`N`$ steps, $`\overline{r}(N)[\overline{r^2}(N)]^{1/2}`$, averaged over all possible SAW configurations behaves as
$$\overline{r}(N)N^{\nu _\mathrm{F}}$$
(3)
with the universal exponent $`\nu _\mathrm{F}=3/4`$ in $`d=2`$ as suggested by Flory . In Fig. 2, we show values for $`\overline{r}(N)`$ versus $`N`$ obtained by EE technique . The asymptotic value for $`\nu _\mathrm{F}`$ (see also Table I) is obtained using successive slopes as shown in the inset of Fig. 2 and is in excellent agreement with the theoretical prediction.
A more detailed information about the spatial structure of SAWs is given by the distribution function $`P(r,N)`$, where $`P(r,N)\mathrm{d}r`$ is the probability that after $`N`$ steps the end-to-end distance of a chain is between $`r`$ and $`r+\mathrm{d}r`$. This quantity obeys the scaling form
$$P(r,N)\frac{1}{r}f(r/N^{\nu _\mathrm{F}})$$
(4)
and is normalized according to $`_0^{\mathrm{}}drP(r,N)=1`$. The analytic form of the scaling function $`f(x)`$ is known asymptotically:
$$f(x)\{\begin{array}{cc}x^{g_1+d},\hfill & x1\hfill \\ x^{g_2+d}\mathrm{exp}\left(cx^\delta \right),\hfill & x1\hfill \end{array},$$
(5)
where $`g_1=(\gamma 1)/\nu _\mathrm{F}`$ , $`g_2=\delta (d[\nu _\mathrm{F}1/2][\gamma 1])`$ and $`\delta =1/(1\nu _\mathrm{F})`$ . Values for these exponents are summarized in Table I. We have verified these predictions by enumerating all SAW configurations for $`N=23`$ and $`24`$ and calculating the corresponding distributions $`P(r,N)`$, from which we have extracted the different exponents (see Fig. 3). We show that a more accurate determination of the exponent $`g_2`$ compared to a simple fit using Eq. (5) can be obtained by employing a specific numerical procedure described in Appendix A (see inset of Fig. 3). The obtained values are in agreement with the theoretical predictions (see Table I).
## III SAWs on the backbone of the incipient percolation cluster
Next, we consider SAWs on the incipient percolation cluster by generating all SAW configurations directly on the backbone of the cluster. We obtain the backbone of a given cluster grown by the Leath algorithm by randomly choosing one of the sites of the last grown cluster shell (e.g. site $`A`$ in Fig. 4) and determining the backbone between site $`A`$ and the seed of the cluster (site $`S`$ in Fig. 4) by the burning procedure described in . The SAWs start at the seed $`S`$ of the cluster. To avoid finite size effects, the chemical distance between both endpoints $`S`$ and $`A`$ of the backbone is chosen at least 20 times larger than the chemical length of the SAWs. The large ratio between both chemical lengths is needed, since close to the endpoint $`A`$, the backbone has a quasi-linear structure, which would falsify the results for the SAWs. The straightforward idea to use all sites on the last grown shell as endpoints for the backbone does not help, but introduces boundary effects in the opposite direction, since in this case the backbone coincides with the cluster near the end points; cf. .
We analyse the results for SAWs on the incipient percolation cluster by applying analogous numerical procedures on the data as described above for SAWs on regular lattices. In contrast to the case of regular lattices, on a percolation cluster two different metrics can be defined, the Euclidean metric and the topological metric. On average, the chemical distance $`\mathrm{}`$ between two backbone sites separated by the Euclidean distance $`r`$ increases with $`r`$ as
$$\mathrm{}r^{d_{\mathrm{min}}}$$
(6)
where $`d_{\mathrm{min}}=1.1306\pm 0.0003`$ in $`d=2`$ and $`d_{\mathrm{min}}=1.374\pm 0.004`$ in $`d=3`$ . Thus Eq. (6) yields the scaling relation between the two metrics, which will be used in what follows. Numerically it is found, that data obtained in $`\mathrm{}`$-space show less fluctuations (cf. e.g. ). Therefore more accurate estimates for many characteristic quantities (such as critical exponents) in $`r`$-space can be determined by studying the corresponding quantity in $`\mathrm{}`$-space and transforming it to $`r`$-space. For example, the fractal dimension of the backbone in $`\mathrm{}`$-space is $`d_{\mathrm{}}^\mathrm{B}=1.45\pm 0.01`$ in $`d=2`$ and $`d_{\mathrm{}}^\mathrm{B}=1.36\pm 0.02`$ in $`d=3`$. Using Eq. (6), this leads to the values $`d_\mathrm{f}^\mathrm{B}=d_{\mathrm{}}^\mathrm{B}d_{\mathrm{min}}=1.64\pm 0.02`$ and $`d_\mathrm{f}^\mathrm{B}=1.87\pm 0.03`$ in $`r`$-space, respectively .
### A Total number of SAW configurations: Multifractality
Due to the disordered structure of the clusters, the total number $`C_{N,\mathrm{B}}`$ of SAW configurations that are generated on a single backbone, with the seed $`S`$ of the cluster as starting point, fluctuates strongly among different backbone configurations. To characterize these fluctuations, we study the moments $`C_{N,\mathrm{B}}^q`$. A similar study on percolation clusters at criticality has been performed for ‘ideal’ chains, i.e. chains which can intersect themselves. This model leads to a non-trivial dependence on $`q`$ .
In generalizing Eq. (1), we make the ansatz
$$C_{N,\mathrm{B}}^q^{1/q}=A_q\mu _q^NN^{\gamma _q1},$$
(7)
where $`\mu _q`$ are the generalized effective coordination numbers of the backbone and $`\gamma _q`$ the generalized enhancement exponents. Results for different values of $`q`$ are shown in Figs. 5(a) and 5(b) for the square and the simple cubic lattice, respectively, employing the numerical procedure described in Section IIA. The values for $`\mu _q`$ and $`\gamma _q`$ are displayed in Fig. 6 for $`d=2`$, clearly revealing a dependence on $`q`$, reminiscent of a multifractal behavior. For large negative values of $`q`$, backbone configurations with a small number of SAW configurations $`C_{N,\mathrm{B}}`$ are singled out in the averaging procedure. We find that $`\mu _q1`$ and $`\gamma _q1`$ for $`q\mathrm{}`$, pointing to rare configurations of backbones with an almost linear shape. On the contrary, for large values of $`q`$ the averaging procedure emphasizes backbone configurations with a large number of SAW configurations $`C_{N,\mathrm{B}}`$. Since these backbones are the most compact ones, $`\mu _q`$ and $`\gamma _q`$ are strongly enlarged. Fig. 6 seems to suggest that the structure of the most compact backbones differs distinctively from the structure of a regular square lattice, as $`lim_q\mathrm{}\mu _q1.9`$, which is well below the value for $`\mu `$ on the regular square lattice, and $`lim_q\mathrm{}\gamma _q1.7`$ is well above the value for $`\gamma `$ on the regular square lattice.
These results resolve earlier controversies regarding the values for both $`\mu `$ and $`\gamma `$ for percolation obtained from MC simulations and by EE techniques. For the square lattice, for example, the values $`\mu _{\mathrm{perc}}(\mathrm{EE})=1.53\pm 0.05`$ and $`\gamma _{\mathrm{perc}}(\mathrm{EE})=1.33\pm 0.02`$ have been obtained from exact enumerations calculations, while from MC simulations the values $`\mu _{\mathrm{perc}}(\mathrm{MC})=1.459\pm 0.003`$ and $`\gamma _{\mathrm{perc}}(\mathrm{MC})=1.31\pm 0.03`$ were determined. We find $`\mu _1=1.565\pm 0.005`$, $`\gamma _1=1.34\pm 0.05`$ and $`\mu _0=1.456\pm 0.005`$, $`\gamma _0=1.26\pm 0.05`$, corresponding to the EE and MC results, respectively. This can be understood by noting that EE calculations yield by definition the whole ensemble (the so-called ‘annealed’ average), corresponding to the case $`q=1`$, i.e. the normal arithmetic average. In contrast, MC simulations intrinsicly sample only a small subset of all possible configurations, omitting rare configurations, yielding ‘typical’ subsets of the ensemble (the so-called ‘quenched’ average). This quenched average is usually described by a logarithmic average, i.e. $`C_{N,\mathrm{B}}_{\mathrm{typ}}\mathrm{exp}\mathrm{ln}C_{N,\mathrm{B}}`$, and is equivalent to the limit $`q0`$ of Eq. (7), i.e. $`lim_{q0}C_{N,\mathrm{B}}^q^{1/q}=\mathrm{exp}\mathrm{ln}C_{N,\mathrm{B}}`$. Indeed, our results are in excellent agreement, in both $`d=2`$ and $`d=3`$, with the relation
$$\mu _1=p_\mathrm{c}\mu $$
(8)
where $`\mu `$ is the effective coordination number of the underlying regular lattice, $`p_\mathrm{c}=0.5927460`$ for the square lattice and $`p_\mathrm{c}=0.311605`$ for the simple cubic lattice . This relation, which was originally suggested in the form $`\mu _{\mathrm{perc}}=p_\mathrm{c}\mu `$ , could not be confirmed earlier on because of the different values obtained for $`\mu _{\mathrm{perc}}`$.
Because of the possible existence of rare events playing a dominant role in the average procedure, we have performed a detailed analysis of our numerical data to confirm that we have considered a sufficiently large set of cluster configurations (cf. Appendix B).
### B Mean end-to-end distances and structural distribution functions
Next, we study the scaling behavior of the distribution functions for the end-to-end distance, $`P_\mathrm{B}(\mathrm{},N)`$ and $`P_\mathrm{B}(r,N)`$, averaged over many backbone configurations, where $`P_\mathrm{B}(\mathrm{},N)\mathrm{d}\mathrm{}`$ is the probability that after $`N`$ steps the chemical end-to-end distance of a chain on a single backbone is between $`\mathrm{}`$ and $`\mathrm{}+\mathrm{d}\mathrm{}`$, and $`P_\mathrm{B}(r,N)\mathrm{d}r`$ is the analogous quantity in $`r`$-space. These distribution functions are expected to obey scaling forms similar to the one valid on regular lattices, Eq. (4), with the corresponding scaling exponents . The mean chemical end-to-end distance $`\overline{\mathrm{}}(N)`$ and the root mean square Euclidean end-to-end distance $`\overline{r}(N)\left[\overline{r^2}(N)\right]^{1/2}`$ scale with $`N`$ as
$$\overline{\mathrm{}}(N)N^\nu _{\mathrm{}}$$
(9)
and
$$\overline{r}(N)N^{\nu _r},$$
(10)
respectively. The first average is performed over all SAW configurations on a single backbone, the second average is carried out over many backbone configurations. Following Eq. (6), the exponents $`\nu _{\mathrm{}}`$ and $`\nu _r`$ are related to each other by $`\nu _r=\nu _{\mathrm{}}/d_{\mathrm{min}}`$. The numerical results for $`\nu _{\mathrm{}}`$ and $`\nu _r`$ obtained by the successive slopes technique discussed in Section II B for regular lattices are reported in Table II. As an example Fig. 7 shows the determination of $`\nu _{\mathrm{}}`$ in $`d=3`$.
Accordingly, the scaling variable in chemical space is $`\mathrm{}/N^\nu _{\mathrm{}}`$ and the mean structural distribution function, averaged over many backbone configurations, has the form
$$P_\mathrm{B}(\mathrm{},N)\frac{1}{\mathrm{}}f(\mathrm{}/N^\nu _{\mathrm{}}),$$
(11)
with the scaling function
$$f_{\mathrm{}}(x)\{\begin{array}{cc}x^{g_1^{\mathrm{}}+d_{\mathrm{}}^\mathrm{B}},\hfill & x1\hfill \\ x^{g_2^{\mathrm{}}+d_{\mathrm{}}^\mathrm{B}}\mathrm{exp}\left(c_{d,\mathrm{}}x^\delta _{\mathrm{}}\right),\hfill & x1\hfill \end{array}.$$
(12)
Equivalently, in $`r`$-space, the scaling variable is $`r/N^{\nu _r}`$, and one has
$$P_\mathrm{B}(r,N)\frac{1}{r}f(r/N^{\nu _r})$$
(13)
with
$$f_r(x)\{\begin{array}{cc}x^{g_1^r+d_\mathrm{f}^\mathrm{B}},\hfill & x1\hfill \\ x^{g_2^r+d_\mathrm{f}^\mathrm{B}}\mathrm{exp}\left(c_{d,r}x^{\delta _r}\right),\hfill & x1\hfill \end{array}.$$
(14)
Both distribution functions are normalized according to $`_0^{\mathrm{}}d\mathrm{}P_\mathrm{B}(\mathrm{},N)=1`$ and $`_0^{\mathrm{}}drP_\mathrm{B}(r,N)=1`$.
The numerical results for the distribution functions in $`d=2`$ and $`d=3`$ are shown in Figs. 8 and 9, respectively, in both $`\mathrm{}`$\- and $`r`$-space. The values for the exponents $`\nu _{\mathrm{}}`$ and $`\nu _r=\nu _{\mathrm{}}/d_{\mathrm{min}}`$ reported in Table II are used in the scaling variables. For the determination of the exponents $`g_1^{\mathrm{}}`$, $`g_2^{\mathrm{}}`$, $`g_1^r`$, and $`g_2^r`$ according to Eqs. (12) and (14), we use the previously reported values of the fractal dimensions $`d_{\mathrm{}}^\mathrm{B}`$ and $`d_\mathrm{f}^\mathrm{B}`$ . The exponents $`g_1^{\mathrm{}}`$ and $`g_1^r`$ can be estimated directly from the slope of $`f_{\mathrm{}}`$ and $`f_r`$ in the double logarithmic plots. Since $`g_1^{\mathrm{}}`$ and $`g_1^r`$ are related by $`g_1^r=g_1^{\mathrm{}}d_{\mathrm{min}}`$ , a more precise estimate for $`g_1^r`$ can be derived from the estimate for $`g_1^{\mathrm{}}`$. The determination of $`g_2^{\mathrm{}}`$ and $`g_2^r`$ is more difficult, since both exponents occur in the non-dominant part and are masked by the exponential. Therefore it requires the use of the slightly more involved numerical procedure discussed in Appendix A (see the insets of Figs. 8 and 9 for $`d=2`$ and $`d=3`$, respectively). The numerical results we obtain for $`g_1^{\mathrm{}}`$, $`g_2^{\mathrm{}}`$, $`g_1^r`$ and $`g_2^r`$ are reported in Table II. Regarding the exponential factors, our results for the exponents $`\delta _{\mathrm{}}`$ and $`\delta _r`$ are consistent, within the present accuracy, with the expressions $`\delta _{\mathrm{}}=1/(1\nu _{\mathrm{}})`$ and $`\delta _r=1/(1\nu _r)`$, respectively.
As discussed in Section II B, for regular lattices the exponents $`g_1`$, $`\nu _\mathrm{F}`$ and $`\gamma `$ are related by the des Cloizeaux relation $`g_1=(\gamma 1)/\nu _\mathrm{F}`$. Therefore, it is legitimate to ask if a similar ‘generalized des Cloizeaux’ relation holds also for SAWs in percolation. Since the enhancement exponent $`\gamma _q`$ depends on $`q`$, it is necessary to find out whether the exponents $`\nu _{\mathrm{}}`$ and $`g_1^{\mathrm{}}`$ as well as $`\nu _r`$ and $`g_1^r`$ depend on $`q`$. To this end we generalize the averages $`\overline{\mathrm{}}(N)`$ and $`\overline{r}(N)`$ to $`\overline{\mathrm{}}^q(N)^{1/q}N^{\nu _{\mathrm{}}^{(q)}}`$ and $`\overline{r}^q(N)^{1/q}N^{\nu _r^{(q)}}`$. Since this is equivalent of studying the quantities $`\left[\mathrm{}^qP_\mathrm{B}(\mathrm{},N)d\mathrm{}\right]^{1/q}`$ and $`\left[r^qP_\mathrm{B}(r,N)dr\right]^{1/q}`$, respectively, and $`P_\mathrm{B}(\mathrm{},N)`$ and $`P_\mathrm{B}(r,N)`$ scale with $`\mathrm{}/N^\nu _{\mathrm{}}`$ and $`r/N^{\nu _r}`$, also $`\overline{\mathrm{}}^q(N)^{1/q}`$ and $`\overline{r}^q(N)^{1/q}`$ must scale with $`\mathrm{}/N^\nu _{\mathrm{}}`$ and $`r/N^{\nu _r}`$, respectively. Therefore $`\nu _{\mathrm{}}^{(q)}=\nu _{\mathrm{}}`$ and $`\nu _r^{(q)}=\nu _r`$ for all $`q`$, which we confirmed numerically. Furthermore we have numerically verified that the exponents $`g_1^{\mathrm{}}`$ and $`g_1^r`$ (as well as $`g_2^{\mathrm{}}`$, $`g_2^r`$, $`\delta _{\mathrm{}}`$ and $`\delta _r`$) are independent of $`q`$. Regarding the ‘generalized des Cloizeaux’ relation, our numerical results suggest that in $`d=2`$ the relations $`g_1^{\mathrm{}}=(\gamma _{q=1}1)/\nu _{\mathrm{}}`$ and $`g_1^r=(\gamma _{q=1}1)/\nu _r`$ hold, where the, to some extent arbitrary, choice of $`\gamma _{q=1}`$ is motivated by the fact that $`q=1`$ describes the annealed case of the whole SAW ensemble. However in $`d=3`$ these relations are not satisfied by the present numerical results.
## IV Concluding remarks
We have studied structural properties of SAWs on the backbone of the incipient percolation cluster in both $`d=2`$ and $`d=3`$, applying exact enumeration techniques. Our results suggest that SAWs display multifractal behavior, caused by the underlying multiplicative process yielding an infinite hierarchy of generalized coordination numbers $`\mu _q`$ and enhancement exponents $`\gamma _q`$ describing the moments $`C_{N,\mathrm{B}}^q`$ of the total number of SAWs of length $`N`$. The present results resolve previous inconsistencies regarding the suggested relation $`\mu _{\mathrm{perc}}=p_\mathrm{c}\mu `$, where $`p_\mathrm{c}`$ is the percolation threshold of a specific regular lattice, and $`\mu `$ and $`\mu _{\mathrm{perc}}`$ are the corresponding effective coordination numbers of SAWs for the ordered case and on the incipient percolation cluster, respectively. We find that this relation is accurately obeyed on the square and simple cubic lattice by identifying $`\mu _{\mathrm{perc}}=\mu _1`$.
## Acknowledgements
We thank A. Blumen, S. Edwards and P. Grassberger for very useful discussions. Financial support from the German-Israeli Foundation (GIF), the Minerva Center for Mesoscopics, Fractals, and Neural Networks and the Deutsche Forschungsgemeinschaft is gratefully acknowledged. M.P. gratefully acknowledges the Alexander-von-Humboldt foundation (Feodor-Lynen program) for financial support.
## A Improved procedure for the determination of $`g_2`$
The procedure used for extracting the exponents $`g_2`$, $`g_2^{\mathrm{}}`$, and $`g_2^r`$, describing the scaling form of the structural distribution functions, is an improved version of the procedure by Wittkop et al. (cf. ) and is illustrated here for the case of regular lattices. The distribution function Eq. (4) can be written as
$$P(r,N)=\frac{\mathrm{\Omega }B}{r}f(r/N^{\nu _\mathrm{F}})$$
(A1)
with $`\mathrm{\Omega }=2\pi `$ in $`d=2`$ and $`\mathrm{\Omega }=4\pi `$ in $`d=3`$ and the scaling function $`f(x)`$ defined as
$$f(x)=\{\begin{array}{cc}x^{g_1+d},\hfill & x<z\hfill \\ x^{g_2+d}\mathrm{exp}\left(bx^\delta \right),\hfill & x>z\hfill \end{array},$$
(A2)
where $`\delta =1/(1\nu _\mathrm{F})`$. The actual value of the crossover $`z`$ is determined from the numerical results. The constants $`B`$ and $`b`$ can be obtained from the normalization condition
$$_0^{\mathrm{}}P(r,N)dr=1$$
(A3)
and from the second moment
$$_0^{\mathrm{}}r^2P(r,N)dr=\overline{r^2}(N)N^{2\nu _\mathrm{F}}.$$
(A4)
Upon integration of Eqs. (A3) and (A4), one gets the exact relations
$$B=\frac{1}{\mathrm{\Omega }}\left[\frac{1}{\delta b^{(g_2+d)/\delta }}\mathrm{\Gamma }(\frac{g_2+d}{\delta },bz^\delta )+\frac{z^{g_1+d}}{g_1+d}\right]^1,$$
(A5)
where $`\mathrm{\Gamma }(u,z)`$ is the incomplete Gamma function, and
$$\mathrm{\Omega }B\left\{\frac{1}{\delta b^{(g_2+d+2)/\delta }}\mathrm{\Gamma }(\frac{g_2+d+2}{\delta },bz^\delta )+\frac{z^{g_1+d+2}}{g_1+d+2}\right\}=1.$$
(A6)
Thus, by plotting the distribution function in the case $`x>z`$ as $`yb^{(g_2+d)/\delta }(\mathrm{\Omega }B)^1rP(r,N)\mathrm{exp}\left[\left(b^{1/\delta }r/N^{\nu _\mathrm{F}}\right)^\delta \right]`$ versus $`b^{1/\delta }r/N^{\nu _\mathrm{F}}`$ in a double logarithmic plot, the exponent $`g_2`$ can be read off from the relation $`yx^{g_2+d}`$ and adjusted until the above relations Eqs. (A5) and (A6) are satisfied. This method yields much more accurate results than by directly fitting the distribution function itself. The accuracy of the result can be assessed by plotting $`y\mathrm{ln}\left[b^{(g_2+d)/\delta }(\mathrm{\Omega }B)^1rP(r,N)\left(b^{1/\delta }r/N^{\nu _\mathrm{F}}\right)^{(g_2+d)}\right]=b\left(r/N^{\nu _\mathrm{F}}\right)^\delta `$ versus $`b^{1/\delta }r/N^{\nu _\mathrm{F}}`$ in a double logarithmic plot, from which the exponent $`\delta `$ can be determined and compared with the expected value $`\delta =1/(1\nu _\mathrm{F})`$. The procedure can be extended straightforwardly to study the distribution functions $`P_\mathrm{B}(\mathrm{},N)`$ and $`P_\mathrm{B}(r,N)`$ of SAWs on the backbone of critical percolation clusters.
## B Generalized averaging procedure
To obtain an estimate of whether the ensemble $``$ of backbone configurations considered is sufficiently large in order to get convergent results, we analyse the data by a generalized averaging procedure as follows: The total ensemble $``$ containing $`n_{\mathrm{tot}}`$ backbone configurations is divided into subsets $`_i`$ containing $`n_{\mathrm{eff}}`$ configurations each. The generalized average is then defined as
$$C_{N,\mathrm{B}}_{n_{\mathrm{eff}}}^{(q)}=\left(\frac{1}{n_{\mathrm{eff}}}\underset{i=1}{\overset{n_{\mathrm{eff}}}{}}(C_{N,\mathrm{B}})_i^q\right)^{1/q}.$$
(B1)
The obtained results $`C_{N,\mathrm{B}}_{n_{\mathrm{eff}}}^{(q)}`$ depend sensitively on the different backbone configurations and display strong fluctuations, indicating that the system is not self-averaging. In order to smooth out these fluctuations, a second average is performed. This second step is a logarithmic average over the $`n_{\mathrm{tot}}/n_{\mathrm{eff}}`$ subsets
$$C_{N,\mathrm{B}}(q,n_{\mathrm{eff}})=\mathrm{exp}\mathrm{ln}C_{N,\mathrm{B}}_{n_{\mathrm{eff}}}^{(q)}=A_{q,n_{\mathrm{eff}}}\mu _{q,n_{\mathrm{eff}}}^NN^{\gamma _{q,n_{\mathrm{eff}}}1}.$$
(B2)
In Eq. (B2), the limiting case $`n_{\mathrm{eff}}=1`$ corresponds to the limit $`q0`$, while the usual average (cf. Eq. (7)) is recovered when $`n_{\mathrm{eff}}=n_{\mathrm{tot}}`$. The results for the coordination numbers $`\mu _{q,n_{\mathrm{eff}}}`$ and enhancement exponents $`\gamma _{q,n_{\mathrm{eff}}}`$ are shown in Figs. 10(a) (for $`q=1`$) and 10(b) (for $`q=2`$). A dependence of these two values on $`n_{\mathrm{eff}}`$ indicates that the given ensemble is too small to obtain the asymptotic values. If, on the contrary, the ensemble of backbone configurations is sufficiently large, then $`\mu _{q,n_{\mathrm{eff}}}`$ and $`\gamma _{q,n_{\mathrm{eff}}}`$ no longer depend on $`n_{\mathrm{eff}}`$. For $`q=1`$, this seems to be the case when $`n_{\mathrm{eff}}10^3`$, and for $`q=2`$ when $`n_{\mathrm{eff}}10^4`$.
|
warning/0003/gr-qc0003049.html
|
ar5iv
|
text
|
# Scalar field cosmology in three-dimensions.
## I Introduction.
Many recent works have focused attention on the issue of general relativity in $`2+1`$-dimensions . Most of them deal with two important issues: black holes and cosmology.
The main motivation for the most recent works in black hole physics came from the important discovery of the BTZ black hole . They investigate different classical and quantum properties of the BTZ and other black hole solutions , .
On the other hand, the simplicity of the theory in three-dimensions, when compared with its version in four-dimensions, is the main motivation for the works in cosmology. Most of them deal with quantum aspects of the theory, in particular the attempt to derive a wave-function for the Universe , , .
In the present work we would like to study an analytical solution to the Einstein’s equations in $`2+1`$-dimensions. The space-time is dynamical and has a line symmetry. The matter content is represented by a minimally coupled, massless, scalar field. We shall see that this solution gives rise, depending on the value of certain parameters, to three different space-times. The first one is flat space-time. Then, we have a big bang model with a negative curvature scalar and a real scalar field. The last case is a big bang model with event horizons where the curvature scalar vanishes and the scalar field changes from real to purely imaginary.
In Sec. II, we start by writing down the appropriate Einstein’s equations. A solution ($`S`$) of this set of equations has already been found, in different circumstances Ref. . We shall use $`S`$ suitably adapted to our problem.
$`S`$ is composed of few functions of the relevant coordinates and free parameters ($`p_i`$). In Sec. III, we impose conditions upon $`S`$ such that it may be interpreted as a physically acceptable solution. These conditions restrict the domains of the $`p_i`$’s, and divide $`S`$ in three distinct types of space-times, depending on the values of the $`p_i`$’s. The space-times are the ones mentioned above and we study them in great details.
Finally, in Sec. IV we summarize the main points and results of the paper.
## II Einstein-scalar equations.
We shall start by writing down the ansatz for the $`2+1`$-dimensional, dynamical, line symmetric, space-time metric as,
$$ds^2=\mathrm{\hspace{0.17em}2}e^{(N(u,v)+U(u,v)/2)}dudv+e^{2U(u,v)}dy^2.$$
(1)
where $`N(u,v)`$ and $`U(u,v)`$ are two arbitrary functions to be determined by the field equations, $`(u,v)`$ is a pair of null coordinates varying in the range $`(\mathrm{},\mathrm{})`$, and $`y`$ is a coordinate taking values over a real line: $`(\mathrm{},\mathrm{})`$.
The scalar field $`\phi `$ will be a function only of the two null coordinates and the expression for its stress-energy tensor $`T_{\alpha \beta }`$ is given by ,
$$T_{\alpha \beta }=\phi ,_\alpha \phi ,_\beta \frac{1}{2}g_{\alpha \beta }\phi ,_\lambda \phi ^{,_\lambda }.$$
(2)
where , denotes partial differentiation.
Now, combining Eqs. (1) and (2) we may obtain the Einstein’s equations, which in the units of Ref. take the following form,
$$2U,_{uu}(U,_u)^2+\mathrm{\hspace{0.17em}2}N,_uU,_u=\mathrm{\hspace{0.17em}2}(\phi ,_u)^2,$$
(3)
$$2U,_{vv}(U,_v)^2+\mathrm{\hspace{0.17em}2}N,_vU,_v=\mathrm{\hspace{0.17em}2}(\phi ,_v)^2,$$
(4)
$$2N,_{uv}+U,_{uv}=\mathrm{\hspace{0.17em}2}\phi ,_u\phi ,_v,$$
(5)
$$(e^U),_{uv}=\mathrm{\hspace{0.17em}0}.$$
(6)
The equation of motion for the scalar field, in these coordinates, is
$$2\phi ,_{uv}U,_u\phi ,_vU,_v\phi ,_u=\mathrm{\hspace{0.17em}0},$$
(7)
Comparing the above set of non-linear, second-order, coupled, partial, differential equations (3)-(7) with the one derived in Ref. , called type $`A`$, we notice that they are identical, although the situation treated here is different from the one treated there. Therefore, we shall take as the solution of the system above Eqs. (3)-(7), the type $`A`$ space-times, obtained in Ref. . They will be introduced and studied in great details in the next section.
## III Solutions.
From our discussion in the previous section, we suitably adapt the type $`A`$ space-times of Ref. to the present situation, and write the following solution to the system Eqs. (3)-(7),
$$\mathrm{exp}(U(u,v))t^2(u,v)=(1u^{2\eta }v^{2\mu })^2,$$
(8)
$`\mathrm{exp}(N(u,v))W(u,v)=`$ $`t^{2\delta }\left[{\displaystyle \frac{u^\eta v^\mu (1v^{2\mu })^{1/2}(1u^{2\eta })^{1/2}}{u^\eta v^\mu +(1v^{2\mu })^{1/2}(1u^{2\eta })^{1/2}}}\right]^{2\gamma }\left[{\displaystyle \frac{u^\mu (1v^{2\mu })^{1/2}}{u^\eta +(1v^{2\mu })^{1/2}}}\right]^{4a\delta _+}`$ (10)
$`\times \left[{\displaystyle \frac{v^\mu (1u^{2\eta })^{1/2}}{v^\mu +(1u^{2\eta })^{1/2}}}\right]^{4a\delta _{}}(1v^{2\mu })^{\delta _+^2}(1u^{2\eta })^{\delta _{}^2},`$
$`\phi (u,v)=`$ $`2a\mathrm{ln}t^2+\rho _+\mathrm{ln}\left[{\displaystyle \frac{1u^\eta (1v^{2\mu })^{1/2}v^\mu (1u^{2\eta })^{1/2}}{1+u^\eta (1v^{2\mu })^{1/2}+v^\mu (1u^{2\eta })^{1/2}}}\right]`$ (12)
$`+\rho _{}\mathrm{ln}\left[{\displaystyle \frac{1u^\eta (1v^{2\mu })^{1/2}+v^\mu (1u^{2\eta })^{1/2}}{1+u^\eta (1v^{2\mu })^{1/2}v^\mu (1u^{2\eta })^{1/2}}}\right]`$
where $`\mu `$ and $`\eta `$ $``$ $`\mathrm{}`$ and they are greater than or equal to $`1/2`$, $`\delta =2a^2+(31/\eta 1/\mu )/4`$, $`\gamma ^2=(11/2\eta )(11/2\mu )`$, $`\delta _+^2=11/2\eta `$, $`\delta _{}^2=11/2\mu `$, $`\rho _\pm =(1/2)(\delta _+\pm \delta _{})`$, and $`a`$ is a constant, real, number.
In terms of the new quantities $`t(u,v)`$ Eq. (8), and $`W(u,v)`$ Eq. (10), the line element Eq. (1) becomes,
$$ds^2=2W(u,v)t^{1/2}(u,v)dudv+t^2(u,v)dy^2,$$
(13)
One may notice from Eqs. (8)-(12), that for different values of $`\eta `$, $`\mu `$ and $`a`$, one has different space-times. It is also important to notice that the choice of the letter $`t`$ in Eq. (8), was not casual. We shall be able to interpret it as a time coordinate.
Observing Eq. (13), we notice that these space-times have a singularity at $`t=0`$. It is a physical singularity as can be seen directly from the curvature scalar $`R`$, and also from Ref. .
In order to show this result we start writing down the Ricci tensor that, in the present case, has the following expression ,
$$R_{\alpha \beta }=\phi ,_\alpha \phi ,_\beta .$$
(14)
From it, we may compute $`R`$ straightforwardly with the aid of Eqs. (8)-(13), finding,
$$R=\frac{8\mu \eta v^{2\mu 1}u^{2\eta 1}(4a+\delta _++\delta _{})^2}{Wt^{5/2}}.$$
(15)
Finally, taking the limit $`t0`$ in $`R`$ Eq. (15), we find that this quantity diverges at $`t=0`$. There is no other physical singularity for these space-times because $`R`$ is well defined outside $`t=0`$. We can also learn that taking the limit $`t\mathrm{}`$ of $`R`$ Eq. (15), this quantity goes to zero.
The space-times above will only be physically acceptable if $`t`$ Eq. (8) is a real, positive function and $`W`$ Eq. (10) is a real function. Therefore, only few distinct sets of values of $`\mu `$, $`\eta `$ and $`a`$ will be allowed. Each of them giving rise to a different type of space-time. It is important to note that since $`t`$ and $`W`$ are functions of ($`u`$,$`v`$), the range of these coordinates shall also be restricted.
Here we shall loose the condition that the scalar field $`\phi `$ Eq. (12), be real. We shall permit it to be purely imaginary in some space-time regions, which means that in those regions $`\phi `$ will be an example of the so-called ‘exotic matter’ .
### A Space-times.
We start now the determination of the allowed values of $`\mu `$, $`\eta `$ and $`a`$ by imposing that $`t`$ be a real, positive, function. In order to simplify our study we shall demand that $`2\mu `$ and $`2\eta `$ be integers, greater than or equal to $`1`$.
Observing the definition of $`t`$ Eq. (8), we notice that if we permit $`2\mu `$ or $`2\eta `$ to be even, for positive $`t`$, we would have very limited domains for the coordinates $`u`$ and $`v`$, respectively. Since we would like to have the biggest possible domains for these coordinates, we shall restrict our attention to odd, integer, values of $`2\mu `$ and $`2\eta `$. The positiveness of $`t`$ will also select the physically accessible space-time volume.
From the expression of W Eq. (10), we learn that the three distinct functions of $`u`$ and $`v`$, inside brackets, respectively with exponents $`2\gamma `$, $`4a\delta _+`$ and $`4a\delta _{}`$, may be negative even for positive $`t`$. Since, in order to interpret $`t`$ as a time coordinate, $`W`$ has to be positive (see Eq. (16) below), we shall eliminate these terms. The best way to accomplish this is by setting theirs exponents to zero.
With the aid of the formulae for $`\gamma `$, $`\delta _+`$ and $`\delta _{}`$, given just below Eq. (12), we understand that there are three distinct manners to set the above exponents to zero: (a) by choosing $`a=0`$ and $`2\mu =2\eta =1`$, (b) by choosing $`2\mu =2\eta =1`$, and finally (c) by choosing $`a=0`$ and either $`2\mu `$ or $`2\eta `$ equal to $`1`$. As we shall see below, these possibilities will produce the distinct sets of space-times associated to our solution.
Now, we are in the position to identify $`t`$ as a time coordinate. For positive $`W`$ and $`t`$, we may compute the norm of the quadri-vector normal to surfaces of constant $`t`$. Which gives,
$$8\eta \mu u^{2\eta 1}v^{2\mu 1}\frac{1}{Wt^{1/2}},$$
(16)
which is always negative for $`2\eta `$ and $`2\mu `$, odd, integer, numbers.
Gathering together all the conditions obtained above, we may group the solutions satisfying these conditions in three distinct sets.
#### 1 Flat space-time.
The first is empty, flat space-time, obtained for $`2\mu =2\eta =1`$ and $`a=0`$. It has the following line element, from Eqs. (8), (10) and (13),
$$ds^2=2dudv+t^2dy^2,$$
(17)
where, $`t=1uv`$.
For positive $`t`$, the coordinates $`u`$ and $`v`$ will vary in the range ($`\mathrm{}`$ , $`\mathrm{}`$), but under the condition,
$$u+v<\mathrm{\hspace{0.17em}1}.$$
(18)
The surface $`t=0`$ in this case is not a physical singularity, it is just a coordinate singularity.
#### 2 Big bang cosmology without horizons.
The solutions belonging to this set have $`2\mu =2\eta =1`$ and $`a0`$. Introducing these values in Eqs. (8), (10) and (13), we may write the following line element,
$$ds^2=2t^{4a^2}dudv+t^2dy^2,$$
(19)
where $`t=1uv`$, $`u`$, $`v`$ $``$ ($`\mathrm{},\mathrm{}`$) and satisfy Eq. (18).
The scalar field Eq. (12) is now,
$$\phi =\mathrm{\hspace{0.17em}2}a\mathrm{ln}t^2.$$
(20)
In the present case, we may see from Eq. (20) that the scalar field is always real, therefore the stress-energy Eq. (2) is always positive.
The scalar of curvature Eq. (15), is written
$$R=\frac{32a^2}{t^{(4a^2+2)}}.$$
(21)
From this expression is easy to see that $`t=0`$ is still a physical singularity for these space-times. The scalar field Eq. (20), is also singular there. Therefore, we may interpret this singularity as a big bang, for this space-time.
The dynamical nature of this space-time may be better appreciated if we re-write the line element Eq. (19) in terms of $`t`$ and $`x=uv`$,
$$ds^2=\frac{t^{4a^2}}{2}(dt^2+dx^2)+t^2dy^2,$$
(22)
where $`\mathrm{}<x<\mathrm{}`$.
#### 3 Big bang cosmology with event horizons.
The last set of space-times is determined when we set $`a=0`$ and either $`2\mu `$ or $`2\eta `$ equal to $`1`$. As a matter of definition, and without loosing the generality, let us choose $`2\eta =1`$ and $`2\mu =2n+1`$, where $`n`$ is a positive integer.
With the aid of Eqs. (8), (10) and (13), the line element of these space-times is,
$$ds^2=2\left(\frac{t}{1u}\right)^{(\frac{2n}{2n+1})}dudv+t^2dy^2,$$
(23)
where $`t=1uv^{2n+1}`$, $`u`$, $`v`$ $``$ ($`\mathrm{},\mathrm{}`$) and satisfy the condition $`u+v^{2n+1}<1`$.
Observing Eq. (23), we identify besides the singularity at $`t=0`$, another one at $`u=1`$. The second singularity is not a physical one as can be seen directly from $`R`$ Eq. (15), which for the present situation is,
$$R=\frac{4nv^{2n}(1u)^{(\frac{2n}{2n+1})}}{t^{(\frac{6n+2}{2n+1})}}.$$
(24)
Indeed, $`u=1`$ is an event horizon as we shall demonstrate below.
Since $`u=1`$ is just a coordinate singularity, there are new coordinates which let Eq. (23) regular at this event. On the other hand, in order to better understand the physical effect of the horizon and the dynamical nature of the space-time, it is more convenient to re-write the line element Eq. (23) in terms of, $`t`$ and $`x=\mathrm{\hspace{0.17em}1}+uv^{2n+1}`$. Which gives,
$$ds^2=\frac{1}{4n+2}\left(\frac{4t}{x^2t^2}\right)^{\left(\frac{2n}{2n+1}\right)}(dt^2+dx^2)+t^2dy^2.$$
(25)
Here, it is clear that we have not only the $`u=0`$ horizon, which in the new coordinates is the surface $`t=x`$, but another one at $`t=x`$. In the old coordinates this is the surface $`v=0`$.
The basic property of event horizons, is the fact that they isolate certain space-time regions from another ones . This can be demonstrated for the surfaces $`t=\pm x`$, in the following way if we restrict our attention to the $`(t,x)`$ plane.
Let us start by calling sector I, the region in the past of the horizons ($`0<t<\pm x`$, $`\mathrm{}<x<\mathrm{}`$). We also introduce the sector II, as the region to the future of the horizons ($`\pm x<t<\mathrm{}`$, $`\mathrm{}<x<\mathrm{}`$). From Eq. (25) we can see that null rays describe $`45^{}`$ or $`135^{}`$ straight lines in the ($`t,x`$) plane. Therefore, not even the light will be able to return from sector II to sector I, once it has entered it. Although $`t`$ does not change the role of time with $`x`$ after crossing the horizons, as we shall see few important facts take place there.
The scalar field Eq. (12), may be obtained for the space-times being studied, in the new set of coordinates.
$$\phi (t,x)=\frac{1}{2}\sqrt{\frac{2n}{2n+1}}\mathrm{ln}\left[\frac{(x+(x^2t^2)^{1/2}}{(x(x^2t^2)^{1/2}}\right].$$
(26)
If one inspects the expression of $`\phi (t,x)`$ Eq. (26), one notices that for $`t=0`$ it diverges. Therefore, also for the present space-times we may interpret $`t=0`$ as a big bang singularity.
In sector I, $`\phi (t,x)`$ Eq. (26) is a real function. On the other hand, in sector II the scalar field is purely imaginary. It means that in sector II, the stress-energy Eq. (2) may be negative and the scalar field will be an example of the so-called ‘exotic matter’.
Observing $`R`$ Eq. (24) we see that it vanishes in both horizons but has the same sign in sectors I and II.
Finally, If one were interested in studying quantum field theory in sector II, the resulting theory would be unitary. This is the case because there is no singularities there and, from Eq. (16), the space-times under investigation possess a global timelike Killing vector field .
## IV Conclusions.
In the present work we have studied an analytical solution to the Einstein’s equations in $`2+1`$-dimensions. The space-time was dynamical and had a line symmetry. The matter content was represented by a minimally coupled, massless, scalar field.
The Einstein’s equations for this system were identical to another one already solved in the literature. We have used the known solution ($`S`$), and studied it.
We have imposed certain conditions upon $`S`$ such that it could be interpreted as a physically acceptable solution. These conditions restricted the domains of few free parameters ($`p_i`$) of $`S`$, and divided $`S`$ in three distinct types of space-times, depending on the values of the $`p_i`$’s. The first one was flat space-time. Then, we had a big bang model with a negative curvature scalar and a real scalar field. The last case was a big bang model with event horizons where the curvature scalar vanishes and the scalar field changes from real to purely imaginary.
###### Acknowledgements.
I am grateful to A. Wang for suggestive discussions in the course of this work. I would like also to thank I. D. Soares for helpful discussions and FAPEMIG for the invaluable financial support.
|
warning/0003/astro-ph0003012.html
|
ar5iv
|
text
|
# Stellar Populations in the Large Magellanic Cloud from 2MASS
## 1 Introduction
Large and homogeneous data sets of near-infrared photometry for the entire LMC, a by-product of large-scale infrared sky surveys such as 2MASS Skrutskie et al. (1997) and DENIS Epchtein et al. (1997), have become available to astronomical community only recently. Cioni et al. (2000) introduced the DENIS Point Source Catalog towards the Magellanic Clouds. Here, we present the LMC data from The Two Micron All Sky Survey (2MASS).
The 2MASS has observed the entirety of the Large Magellanic Cloud and much of these data are included in the recent second incremental data release. Empirically, the photometry has signal-to-noise (SNR) ratio 10 at J, H, K<sub>s</sub> magnitudes of 16.3, 15.3, 14.7, respectively, slightly better than the nominal survey limit. At these limits, we can observe all of the thermally-pulsing asymptotic giant branch (AGB) populations and part way down the red giant branch (RGB). The red clump, representing helium burning giants, lies $`2`$ mag below the sensitivity limit of these data. The extinction in near-infrared (NIR) is small throughout the LMC and negligible on average everywhere but the inner degree of arc. Together with the high quality of 2MASS photometry ($`\sigma _m0.03^m`$), overall zero-point stability (better than $`0.01^m`$) and with reliable identification of LMC populations, the survey is ideal for studies of spatial structure of the LMC or its evolution (see the companion paper, Weinberg & Nikolaev 2000; hereafter WN).
We describe data selection, cross-correlation with a few well-known populations, and comparison of the 2MASS color system in §2. We present a morphological analysis the color-magnitude diagram in §3. In particular, we make identifications of the Galactic and LMC populations corresponding to all features in the color-magnitude diagram, correlate these with spatial distribution, and estimate the density of LMC populations alone. The LMC giant branch is well determined and we separately identify the AGB, first-ascent red giant branch tip (TRGB), and the carbon star sequence. The luminosity function of RGB and AGB populations is derived (§4) and compared with the galactic giant-branch luminosity function. We explore the feasibility of determining the spatial dependence of metallicity using giant-branch morphology. Finally, §5 summarizes our results and discusses the implications and opportunity for future study.
## 2 Data
Our LMC field is $`250`$ sq. degrees and covers the range from $`4^h00^m`$ to $`6^h56^m`$ in right ascension and from $`77^{}`$ to $`61^{}`$ in declination (coordinates in J2000.0). The initial sample of $`7,092,894`$ sources is drawn from the Working Survey Data Base and includes possible artifacts, such as filter glints and diffraction spikes from nearby bright stars, source confusion, and detection upper-limits. The known contaminants and flux statistics are well-characterized and identified during processing Cutri et al. (1999). Eliminating artifacts and requiring detections at all bands with $`\sigma _m0.11^m`$ (SNR $`10`$) leaves $`1,246,304`$ stars. Unlike the released catalog, these data contain multiple apparitions sources because of scan overlaps. We identify the multiple entries based on 1) spatial proximity ($`|\mathrm{\Delta }r|2^{\prime \prime }`$), and 2) matching $`K_s`$-photometry ($`|\mathrm{\Delta }r|5\sigma `$). Note that this procedure differs from that of the 2MASS catalog release (see Cutri et al. 1999). Our final sample contains $`823,037`$ sources.
The spatial distribution of these stars is shown in Figure 1. The figure shows major structural components of the LMC, the bar and disk, immersed in the field of Galactic foreground. The gradient of the foreground sources (the direction to the Galactic center is indicated by an arrow) distorts the isopleths of the outer LMC disk.
The source density near the optical center of the bar (at $`\alpha =5^h24^m`$, $`\delta _{J2000}=69^{}44^{}`$) exceeds $`3.6\times 10^4`$ deg<sup>-2</sup>. The expected mean separation among sources at this density, $`\zeta =1/\sqrt{\rho }20^{\prime \prime }`$, is much greater than 2MASS resolution and therefore confusion should not be significant (see also Wood 1994). As a separate check, we have examined the source counts as the function of magnitude in dense central fields and in more sparse fields in the outer LMC. The distribution of source counts as a function of magnitude have similar shapes in both dense and sparse regions, consistent with low confusion.
We have cross-correlated our sources with the database of long-period variables (LPV) from Hughes & Wood (1990). Their sample includes 376 large-amplitude Mira-like variables and 224 smaller amlitude semi-regular variables. We find 370 ($`98.4\%`$) and 224 2MASS counterparts, respectively<sup>1</sup><sup>1</sup>1The search radius extends to $`5^{\prime \prime }`$ from the listed positions of sources.. Three of the ‘missing’ large-amplitude variables are present in the raw 2MASS data, but are degraded by artifacts (two diffraction spikes and one blend). Of the remaining three, two are matched by stars of the appropriate magnitudes within $`7^{\prime \prime }`$ radius, and only one does not have any match to within $`20^{\prime \prime }`$ radius. All 134 Wolf-Rayet stars in the LMC Breysacher (1999) have been observed by 2MASS van Dyk et al. (1999).
### 2.1 $`K_s`$ vs. $`K`$ Photometry
The 2MASS photometric system is similar to CIT/CTIO system Elias et al. (1982), except that it uses $`K_s`$ band ($`2.002.32\mu \mathrm{m}`$) rather than $`K`$ band. The $`K_s`$ (‘K-short’) bandpass is described by Persson et al. (1998). It was designed to reduce the ground-based thermal background. The transmissivity curve for the filter is given in Persson et al., who also compared $`K_{CIT}`$ with $`K_s`$ photometry for a set of solar-type stars and red standards (see their Tables 2 and 3, respectively). Based on their data, the difference $`KK_s`$ shows no significant systematic trend in the color range $`0<JK<3`$. The strongest trends follow from the presence of CO-band absorption, which affects the $`K`$ filter more than the $`K_s`$ filter. The absolute value of the difference $`|KK_s|`$ is less than $`0.05^m`$ and we will assume $`K_s=K_{CIT}`$ in comparing CIT/CTIO system-based stellar sequences with 2MASS data.
### 2.2 Interstellar Reddening
One of the advantages of 2MASS as compared to optical surveys is low interstellar reddening, since extinction at $`2\mu `$m is approximately 10 times smaller than in $`V`$. The values of interstellar reddening $`E_{BV}`$ found in the literature<sup>2</sup><sup>2</sup>2foreground plus LMC internal reddening fall in the range between $`0.08`$ Mateo & Hodge (1987) and $`0.20`$ Harris et al. (1997). The distribution of reddening from Harris et al. (1997) has non-Gaussian tail to high values. Greve et al. (1990) have reported values as high as $`E_{BV}=1.1`$, found from their investigation of dust in emission nebulae in the LMC. Bessell (1991) has summarized reddening determinations from photometry, stellar polarization and HI column densities to derive foreground and intrinsic mean reddening in the Clouds. He obtained typical LMC internal reddening of $`0.06`$ (with substantial variations), and foreground reddening in the range $`0.040.09`$. Galactic foreground reddening can be surprisingly large in the outer regions of the LMC: Walker (1990) reported $`E_{BV}=0.18\pm 0.02`$ at NGC 1841, about $`15^{}`$ from the optical center of the LMC. On the other hand, in the cluster GLC 0435-59 (Reticulum), $`11^{}`$ from the LMC center, the reddening is only $`E_{BV}=0.03`$ Walker (1992). Zaritsky (1999) has indicated that reddening for F and G stars in the LMC is much less than that for OB stars and derived the average $`<E_{BV}>=0.03`$ for late-type stars in the disk.
In the present study, the data are not dereddened. Rather, each diagram shows the direction and magnitude of the reddening vector for a specified value of $`E_{BV}`$. The reddening vector is based on relations from Koornneef (1982):
$$A_K=0.189E_{BV};E_{JK}=0.651E_{BV},$$
for $`R=A_V/E_{BV}=3.1`$. Information about the reddening may be obtained directly from 2MASS data from the analysis of the color-color diagram (see below).
## 3 Analysis of the Color-Magnitude Diagram
Figure 2 shows the color-color ($`JH`$ vs. $`HK_s`$) diagram of the LMC for $`823,037`$ 2MASS sources selected in §2. The diagram shows relatively few distinct features. The most prominent among them the extended ‘arm’ of the thermally-pulsing AGB stars (TP-AGB) in the upper right corner. Typical colors of LMC M giants are in the range $`0.2\text{ }<\text{ }HK_s\text{ }<\text{ }0.3`$, $`JH\text{ }>\text{ }0.8`$ Frogel & Blanco (1990). Most stars on the extended arm, with colors redder than $`JK_s=1.6`$ are carbon-rich stars Hughes & Wood (1990). Of those with the extremely red colors, $`JK_s>2.0`$, many probably possess dusty circumstellar envelopes. The sample contains $`2000`$ such sources. Their locus is consistent with the track following the reddening vector for $`HK_s>1.0^m`$. Figure 2 also shows fiducial color tracks for both giants and dwarfs from Wainscoat et al. (1992; hereafter W92). The two sequences overlap near $`HK_s=0.15`$, $`JH=0.5`$, since NIR colors of late G — early M type dwarfs are the same as colors of late F — early K type giants.
Because of the small number of features and the general compactness of the color-color diagram, its usefulness in discriminating the major populations is limited, especially in the overlap region, $`0.5<JK_s<0.8`$. However, some LMC populations which occupy distinct regions in the diagram can still be identified based of their location in the color-color plot (see Figure 2). In particular, the color-color diagram is quite useful in identifying candidate sources with large infrared excess, such as young protostars, cocoon stars, or obscured AGB carbon stars.
The color-color diagram may be used to determine the reddening distribution. The giant population forms a tight branch in near-infrared colors. The reddening direction nearly coincides with $`JK_s`$ and therefore the distribution in $`JK`$ of a sample from narrow color interval in $`JH`$ along the giant branch provides a sensitive diagnostic for reddening by dust. We considered a sample of sources in the range $`0.78<JH<0.85`$. The resulting shift of the peak is $`\mathrm{\Delta }(JK_s)<0.03`$ ($`\mathrm{\Delta }(JK_s)<0.06`$) outside (inside) of the central region, suggesting only minor reddening on average on scales larger than 0.1 square degrees.
The color-magnitude diagram (CMD) presented in Figure 3 reveals a wide variety of details. Our goal is reliable identification criteria for LMC stellar populations based on their positions in the diagram. The CMD is hand-shaped with vertically-stretched ‘fingers’ (e.g., at $`JK_s`$ colors of 0.4, 0.6, 1.1) due to varying distance modulus for both Galactic and Magellanic sources. We have identified 12 regions shown in Figure 3 that highlight distinct features of the CMD. The regions are marked A through L and enclose $`99.7\%`$ of the $`823,037`$ sources in the field. To identify stellar populations in each region, we use a combination of several techniques. The Galactic foreground contribution is modeled by a synthetic CMD based on the tabulated near-infrared model of W92. The LMC populations are identified based on the infrared photometry of known populations found in the literature. In addition, we do a preliminary isochrone analysis, where we match the features of the CMD with Girardi et al. (2000) isochrones to derive the ages of populations and draw evolutionary connections among the CMD regions. The details of our population matching procedure are given in §3.1. In addition, we use the spatial density distributions of sources in each region to better discriminate local and LMC populations. The spatial distributions for each region are shown in Figure 4.
In each panel of the figure, we plot source density contour levels of 15, 30, 60, 120, 240, 350, 480, 960, 1920, 3840, 7680, and 15000 deg<sup>-2</sup>. The contour levels are selected both to show the underlying density profiles with maximum details and facilitate the comparison among different population densities. However, due to strong variations in relative density in the CMD, not all contour levels in this sequence are displayed: in panels E and I the lowest density contour corresponds to 60 deg<sup>-2</sup>, in panels B and C the contours start from 240 deg<sup>-2</sup>, and in very dense Region D the lowest contour level is 960 deg<sup>-2</sup>.
### 3.1 Identifying Stellar Populations
The initial analysis of the CMD regions has two parts: 1) use of the spatial density distribution to estimate the location (Galaxy or LMC); and 2) use of the theoretical colors/isochrones to derive the properties of the population, such as age, approximate spectral class and distance modulus. Here we describe the procedure of identifying the populations, before examining each region of CMD in detail.
The CMD of the LMC field (Fig. 3) contains both Galactic and LMC populations. To quantify Galactic foreground, we use near-infrared model of W92, based on 8-25 micron point source counts. Galactic model of W92 has five structural components: exponential disk, bulge, stellar halo, spiral arms and molecular ring. The main contribution to the Galactic source density toward the LMC ($`l=280.5`$; $`b=32.9^{}`$) is the exponential disk. The source density due to other Galactic structural components combined does not exceed $`0.0005\%`$ in any region of the CMD. The luminosity function in W92 model is represented by a sum of stellar classes, allowing independent estimate of the contribution of each class to the CMD. Each class of source is assumed to have a Gaussian distribution,
$$N(M)\mathrm{exp}\left[\frac{(MM_\lambda )^2}{2\sigma ^2}\right].$$
To model the Galaxy, we use the first 33 classes from Table 2 of W92 (Galactic dwarfs, giants and supergiants). The remainder (AGBs, planetary nebulae, etc.) are expected to give only a small contribution to source density and thus do not affect the CMD. In the Galactic model, we use reddening parameters from Rieke & Lebofsky (1985). Dust is assumed to follow a double exponential distribution, with the radial scale length of the disk and the scale height of 100 pc. We reduced the magnitude dispersion $`\sigma `$ of each stellar class by ten to accurately represent the CMD. Reducing $`\sigma `$ results in a spiky differential luminosity function but this does not affect our application and the cumulative luminosity function remains well-approximated. Given the granularity of the model, the agreement between original W92 luminosity function and ours is acceptable. Our synthetic ‘foreground’ CMD is shown in Figure 5, along with the observed CMD of a Galactic field <sup>3</sup><sup>3</sup>3To define our Galactic field, we combined three small fields, each $`0.6`$ sq. deg. These fields were selected at the boundary of our LMC field where contamination from LMC source is minimal. At the time, other contiguous fields were not available.. The agreement between the model and observed CMD is good, with few easily explainable discrepancies. For example, the extension of the CMD at $`JK_s\text{ }>\text{ }1`$, $`K_s1314`$ is due to the population of field galaxies Jarrett (1998).
Conclusions about the LMC populations in the CMD are made based on isochrone fitting or on empirical matching of populations found in the literature to features of the CMD. We use theoretical isochrones from Girardi et al. (2000). These isochrones supersede Bertelli’s set Bertelli et al. (1994), and use updated opacities and equations of state. The isochrones follow the evolution of low- and intermediate-mass stars ($`0.15M_{}<M<7M_{}`$) from the main sequence up to the tip of the RGB or the start of the thermally-pulsing AGB. Distinct LMC populations are identified by matching morphological features of the CMD with colors of known populations from the literature. In particular, we use Cepheid colors from Madore & Freedman (1991), early M supergiant color sequence from Elias et al. (1985), and data on long-period variables from Hughes & Wood (1990). This matching is purely qualitative and is only used as a supplement. The fiducial colors of Galactic giants and supergiants from W92 model are unsuitable for LMC, since LMC has a lower metallicity compared to the Milky Way. Moved to the LMC distance, $`\mu =18.5`$, the W92 giant branch provides a poor fit to the observed RGB (see Figure 7 below).
### 3.2 Region A: Blue Supergiants, O Dwarfs
These blue-colored sources are readily identified as early type Population I stars in the LMC. This group of stars is the evidence of recent ($`<30`$ Myr) star formation. Plotting the theoretical evolutionary tracks in the CMD (Figure 6) confirms that the region is populated by blue supergiants and brightest dwarfs (ZAMS). Only the hottest and most massive dwarfs of types O3—O6 can be seen in the LMC at $`\mu =18.5`$. All other main sequence (MS) populations are too faint and fall below the limit at SNR $`=10`$ imposed for this work. The supergiant population in the region are core helium-burning stars with masses $`4\text{ }<\text{ }M\text{ }<\text{ }9M_{}`$. These stars spend most of their post-MS lifetimes as blue or red supergiants Maeder & Meynet (1989) looping between Regions A and H of the CMD (cf. Figure 6b). Region A encompasses stars which are at the blue tips of their blue loops. While crossing Regions B and C, these stars enter the instability strip and become Cepheids (see §§ 3.3, 3.4).
The spatial distribution of these objects also clearly indicates an LMC population. The distribution is rather clumpy with several richest OB associations outlining the location of spiral arms and brightest and largest HII regions (e.g., 30 Dor). The density concentrations in Fig. 4A are consistent with the well-known superassociations and Shapley’s Constellations Martin et al. (1976); van den Bergh (1981). These youngest populations do not trace the bar of the LMC, in agreement with de Vaucouleurs & Freeman (1973). Quantitative analysis of the distribution Weinberg & Nikolaev (2000) puts the centroid of the population at $`\alpha =5^h23^m`$, $`\delta =68^{}48^{}`$, about $`1^{}`$ north of the optical center of the bar.
The Galactic population of early dwarfs is readily seen in the CMD directly above Region A, blueward of Region B. The apparent magnitudes of these stars suggest a distance modulus between 5 and 10 ($`r0.11`$ kpc). In addition, this area of the CMD may also contain contribution from field blue stragglers and blue horizontal branch stars.
### 3.3 Region B: Galactic Disk F—K Dwarfs, LMC Supergiants
Region B is a vertically stretched band in the CMD with $`JK_s=0.20.5`$. This color cut isolates the main sequence turnoff of the halo ($`JK_s0.3`$) and the disk ($`JK_s0.4`$). The spatial density distribution increases toward NE corner of the field (Fig. 4B), i.e. toward the Galactic center, and indicates a predominantly Galactic population. The vertical extent in the CMD indicates a wide range of distance moduli for these stars. Based on relative population abundances in our synthetic Galactic CMD, we conclude that these sources are disk dwarfs of spectral classes in the range from late F to early K. These stars account for $`90\%`$ of the foreground source density in the region. Their position in the CMD (see Figure 6a) suggests that the dwarfs have distance moduli $`\mu =310`$ ($`r0.041.0`$ kpc). Galactic giants in the region are of types F—G, but their contribution to foreground source density is insignificant, smaller than $`5\%`$.
The distorted shape of the central isopleth in Figure 4B suggests presence of the LMC population in this region. The isopleth outlines the structure similar to the one seen in the central regions in Figure 4A. Note the overdensity near $`\alpha =5^h40^m`$, $`\delta =69^{}`$ and $`\alpha =5^h35^m`$, $`\delta =67^{}30^{}`$, marking positions of superassociations IV and V, respectively Martin et al. (1976). Based on the colors, the LMC component is comprised of young blue and yellow supergiants, corresponding to spectral types A—G. This population includes luminous blue variables and short period Cepheids ($`P\text{ }<\text{ }50^d`$). Figure 7b shows the Cepheid sequence based on PL relations for LMC Cepheids Madore & Freedman (1991). Figure 6b also shows the colors of supergiants from W92 (Table 2). In addition, Region B contains the majority ($`\text{ }>\text{ }80\%`$) of the known LMC Wolf-Rayet stars. Most LMC Wolf-Rayet stars have infrared colors in the range $`0<JK_s<0.5`$. Their numbers, however, are not significant to produce an observable effect in the CMD density.
### 3.4 Region C: Disk K Dwarfs and K Giants, Young Supergiants in the LMC Bar
Similar to B, Region C is stretched along the magnitude axis, indicating that the CMD feature is formed by sources at a range of distances. The colors of this population are in a tight range, $`\mathrm{\Delta }(JK_s)0.3`$. Our synthetic Galactic CMD suggests that most ($`70\%`$) of the observed density in this region is produced by disk K dwarfs at $`\mu <9`$ ($`r\text{ }<\text{ }600`$ pc). Disk K giants are also present in this region. Most of them have $`\mu 613`$ ($`r0.24`$ kpc). They contribute $`20\%`$ of the foreground density. The inspection of isochrones in Figure 6a suggests that Galactic giants in this region are in the evolutionary phase of red clump/horizontal branch stars. The intrinsic brightness and color of these stars ($`M_K=1.4\pm 0.1`$, $`JK_s=0.6\pm 0.1`$) make them natural candidates in this region. Because of the narrow magnitude range of the clump, the source distribution in Region C along the magnitude axis could help constrain the structure of the Galactic disk.
The LMC population in Region C (seen in Figure 4C) is slightly older than youngest supergiants in Regions A, B. The central isopleths of the figure outline the bar of the LMC and show no overdensity at the positions of superassociations, seen in previous two regions. Most of the LMC sources in Region C have $`K_s>10.5`$. The similarity in the shapes of central isopleths between Fig. 4C and 4I suggests they are lower mass young supergiants with ages $`300500`$ Myr, evolving into Region I (Figure 6a). These stars trace the bar of the Cloud Grebel & Brandner (1998). Some contribution from more massive supergiants, including longer-period Cepheids ($`P\text{ }<\text{ }100^d`$) may also be present.
### 3.5 Region D: Disk G—M Dwarfs and LMC RGB and Early AGB Stars
Region D is the most heavily populated area of the CMD: it includes more than a half of all sources in the sample. Because of its position in the CMD and the large color range it spans ($`0.25<JK_s<1.2`$), this region is also the most inhomogeneous. The spatial distribution, shown in Figure 4D shows both the foreground and LMC populations (note the distorting effect of Galactic populations on outer LMC isopleths). The observed CMD is distinctly bimodal in this region, with the red half populated by RGB and early AGB stars in the LMC, and the blue half populated mostly by G—M dwarfs in the Galaxy. The AGB stars in the red half of the region ($`JK_s>0.7`$) are in their ‘early-AGB’ (E-AGB) phase, during which the energy is produced in the thick helium shell and outer hydrogen shell is extinguished. These stars have recently passed the base of the AGB, the so-called ‘AGB-bump’ at $`K_s16`$ that marks the transition from core to shell helium burning Castellani et al. (1991). The AGB-bump was first observed by Hardy et al. (1984) in their CMD of the LMC bar. At only $`1`$ mag brighter than the horizontal branch, this feature is not visible in the CMD in Figure 3 although it is present in deeper data (Figure 10). Empirically, most stars at the E-AGB are M type (oxygen-rich).
Only foreground stars are a minor contributor to the red half of Region D. (cf. Figure 5). Galactic dwarfs in this region have $`\mu 811`$ ($`r0.41.6`$ kpc). Populations contributing insignificantly to the source density in this region include young supergiants, Cepheids, intermediate mass red stars in the vertically extended red clump (VRC; see §3.10).
### 3.6 Region E: Upper RGB and Tip of the RGB
Region E covers the upper RGB and includes the tip of the RGB (see §4). Most of these stars are on the first-ascent red giant branch; they have degenerate helium cores and hydrogen burning shells. The majority of these stars have ages anywhere between $`1`$ and $`15`$ Gyr old. The tip of the RGB is defined the helium flash, the ignition of the degenerate helium core in old (low-mass) stars (Renzini & Fusi Pecci 1989). Stars at the TRGB ignite helium in their cores and evolve rapidly to the horizontal branch. The region also contains a significant fraction of AGB stars in transition from E-AGB to TP-AGB, the stage at which the outer hydrogen shell is re-ignited Iben & Renzini (1983). During thermal pulses, the star begins alternating between hydrogen and helium shell burning. The transition from E-AGB to TP-AGB is theoretically predicted to occur near the TRGB. While on the TP-AGB, these stars may also experience the shorter-term atmospheric pulsations that lead to Mira-type variability. Analysis of MACHO data Alves et al. (1998); Wood (1999) suggests that essentially all stars brighter and redder than the TRGB are variable. Most of the E-AGB stars in this region are M stars. Extrapolated to brighter magnitudes, the sequence of oxygen-rich AGB stars extends to Regions F and G (§§3.7, 3.8).
Stars in this region of the CMD carry the most weight in our analysis of the RGB+AGB luminosity function (see §4). Their spatial distribution is relatively smooth, showing strong disk and bar components. Note the absence of significant foreground population in Figure 4E: the outer contours are elliptical in shape. A small fraction of foreground sources in this region is due to disk M dwarfs. Their density is steadily increasing toward fainter magnitudes (cf. Figure 5).
### 3.7 Region F: O-Rich AGBs
Region F contains primarily oxygen-rich AGB stars of intermediate age ($`\text{ }>\text{ }1`$ Gyr) that are the descendants of stars in Region E (note the similarity between Fig. 4E and 4F). These are E-AGB and TP-AGB stars. The outer CMD isopleth in this region (Figure 3) is distorted and extends into Region J and indicates the presence of carbon stars in this region. During thermal pulses, the outer convective envelope may reach into the region where He has been transformed into C and bring carbon-enriched material to the surface. This dredge-up process leads to an increase in C/O ratio, and M stars in Regions F and G may become carbon stars. In the CMD, carbon stars form a ‘branch’ with redder colors, $`JK_s\text{ }>\text{ }1.4`$ (see §3.11). Some fraction of Region F stars are LPVs (see Figure 8) and reddened supergiants. Figures 4F and G do not show isopleths due to a Galactic population. The Galactic component in these regions are negligible because it is both too bright and too red for disk M dwarfs and too faint for Galactic AGB stars.
### 3.8 Region G: AGB Stars
Region G contains the most massive stars with degenerate C/O cores. This is a population of young AGB, post core-helium burning stars with initial masses between about 5-8 solar masses. These are too short-lived to become carbon stars ($`0.11`$ Gyr old), but not massive enough to become red supergiants. Similar to Region F, this region also includes LPVs (Figure 8). The period-color relation for oxygen-rich Miras derived from Feast et al. (1989) rather closely traces the young AGB branch of the CMD outlined by Regions F and G Weinberg & Nikolaev (2000). Region G lies at bright enough apparent magnitudes that the foreground density of M dwarfs is low, even in comparison to the relatively small number of O-rich luminous AGB stars in the LMC.
### 3.9 Region H: LMC K—M Supergiants, Galactic M Dwarfs, K—M Giants
Panel H of Figure 4 reveals an LMC population. The spatial distribution is similar to the distribution of young OB stars (Region A), suggesting that these objects are also relatively young. At $`\mu =18.5`$, these stars are too bright to be normal M giants. Based on their near infrared colors, we identify the Region H sources as supergiants of M type. They trace the spiral structure of the LMC and do not show significant overdensity in the bar of the Cloud, consistent with a young population. The masses of these stars are believed to be $`29`$ solar masses Bertelli et al. (1985). In the evolutionary sequence, these stars are descendants of stars in Region A (note the similarity between corresponding panels in Figure 4). These stars are also the high mass extension of the VRC.
In Figure 8, we plot the observed colors of M1—M4 supergiants from the sample of Elias et al. (1985). The colors of their sample occupy a portion of Region H, supporting our identification. The Galactic foreground consists of roughly equal contributions from disk K—M giants and M dwarfs, but their overall contribution to the source density in the region is only a few percent. This is confirmed by the absence of the Galactic isopleths in Figure 4H.
### 3.10 Region I: LMC Intermediate-Mass Red Supergiants, Galactic K—M Dwarfs
Region I is located at the center of the CMD, at $`0.7JK_s1.0`$. The observed overdensity in the CMD is associated with the vertically extended red clump (Zaritsky & Lin 1997). This feature consists of intermediate mass stars and is the low-mass extension of the red supergiants (Region H). The VRC extends upward from the red clump at $`K_s17`$ and becomes visible in the CMD near $`K_s=13.5`$. At the this point, the redward slope of the RGB is sufficient to distinguish the VRC.
The spatial distribution is dominated by the bar and shows traces of the spiral structure. We conclude that this LMC population is young, with the age $`\text{ }<\text{ }500`$ Myr. The major LMC contributors to the source density in this region are K and M supergiants. This is supported by the overall similarity of LMC isopleths in Figures 4I and 4H, and also the fact that Region I is at the extension of Region H to fainter magnitudes and lower masses. Figure 8 shows K and M type Miras and SR variables in the sample of Hughes & Wood (1990). A significant fraction of their variables falls in this CMD region suggesting that some of these 2MASS stars also are variables.
The distribution in Figure 4I also reveals foreground populations. The Galactic foreground consists of M and late K dwarfs. Galactic giants contribute less than 5% of the foreground density. The dwarfs are located in the disk, with distance moduli $`\mu =58`$ ($`r0.10.4`$ kpc; Figure 6). The contribution from the Milky Way halo is smaller than 0.0005% by number for this and for all other regions of the CMD.
### 3.11 Region J: Carbon Stars in the LMC
At $`JK_s\text{ }>\text{ }1.4`$, Region J sources are primarily carbon-rich TP-AGB stars. These stars are descendants of oxygen-rich TP-AGBs in Regions F and G. Their outer layers are enriched in C through convection from stellar interior. As mentioned in §3.6, most of these stars are long-period variables. The variability cannot be determined based on single epoch 2MASS data, but the well-defined sequence motivates a follow-up campaign. Figure 8 shows the sample of C-rich LPVs from Hughes & Wood (1990) overplotted on the 2MASS CMD. The contamination by M-type LPVs is small. The spatial distribution of C stars in the field is similar to the distribution of their precursors (Figs. 4F). The distribution is rather smooth and shows a loop of stellar material<sup>4</sup><sup>4</sup>4The feature to the SE of the bar in Figure 4J, near $`\alpha 5^h50^m`$, $`\delta 73^{}`$ represents a hole in the disk., which has been described by Westerlund (1964). The loop is the extension of the main northern spiral arm circling the main body of the system and returning toward the bar after a nearly complete turn.
Sources in this region of the CMD offer the best opportunity to study the three-dimensional structure of the LMC for two reasons. First, the spatial coverage of the Cloud achieved by 2MASS is total and allows to probe the entire LMC. Second, as long-period variables, Region J stars are potentially good standard candles, since their intrinsic luminosity can be characterized based on their period or color. Given the selection efficiency<sup>5</sup><sup>5</sup>5owing to their extremely red colors, these stars are uncontaminated by other populations and easily quantifiable intrinsic brightness through the period-luminosity-color relation (e.g., Feast et al. 1989), these stars are excellent probes of the LMC structure along the line of sight. Preliminary results (WN) indicate that the width of the intrinsic brightness distribution is smaller than $`\sigma _M=0.2`$ magnitudes in a narrow color range, $`\mathrm{\Delta }(JK_s)0.1`$. At this accuracy, these standard candles can resolve features in the LMC at $`\mathrm{\Delta }r4.5`$ kpc. 2MASS detected approximately $`10^4`$ potential carbon LPVs and these are sufficient to attain a reasonable confidence level in the inferred spatial structure. In Weinberg & Nikolaev (2000), we present our study of the three-dimensional structure of the LMC.
### 3.12 Region K: Dusty AGBs
An extension of Region J, Region K contains extremely red objects. We identify them with obscured AGB carbon-rich stars. Their large $`JK_s`$ colors are due to dusty circumstellar envelopes ($`E_{BV}\text{ }>\text{ }1`$). The latter is confirmed by the appearance of their spatial and CMD distributions: (1) Figure 4K shows traces of the spiral structure outlined by these sources; and (2) the distribution in the CMD spreads from the end of Region J in the direction of reddening vector. Matching with existing near infrared photometry of obscured AGB stars in the LMC Zijlstra et al. (1996); van Loon et al. (1998) shows that most of these sources are indeed in this region of the CMD. Other extremely red populations could also be found here, e.g. ‘cocoon’ stars Reid (1991), or OH/IR stars Wood et al. (1992); van Loon et al. (1998). In addition, two of the known LMC protostars, N159-P1 and N159-P2 Jones et al. (1986) also fall in this region.
### 3.13 Region L: Reddened LMC M Giants, Galactic M Dwarfs and 2MASS Galaxies
Stellar sources in Region L are reddened M giants in the LMC and a small number of reddened Galactic M dwarfs. However, a significant number of sources in this region are background galaxies. The predicted CMD density in Region L due to Galactic stars is too low even after the decrease in the photometric quality near the flux limit has been taken into account; the decreasing signal-to-noise ratio causes the apparent widening of the contour levels (see Figure 5). According to Jarrett (1998), more than $`90\%`$ of 2MASS galaxies have colors redder than $`JK_s=1`$.
The spatial distribution of sources (Figure 4L) shows some overdensity toward the center of the LMC. The densest part of the diagram corresponds to the position of 30 Doradus complex. Traces of spiral structure of the LMC are also visible. Based on their colors, these sources are heavily obscured RGB stars in the LMC: they lie in the direction of the reddening vector from the RGB. The inferred reddening for these sources, $`E_{BV}0.5`$, is consistent with the extended tail of the LMC reddening distribution Harris et al. (1997). Region L also includes contribution from massive ($`>10M_{}`$) protostars and ultra-compact H II regions, see Figure 8.
A population of dwarfs is implied by the outer isopleths of Figure 4L that show the increase in the direction of Galactic center. These are local M dwarfs in the disk of the Milky Way, with $`\mu 58`$ ($`r100400`$ pc).
## 4 Luminosity Function of LMC RGB and AGB Populations
We derive the LMC giant branch luminosity function (LF) from the color-magnitude diagram after subtracting Galactic foreground. Since we did not have the access to Galactic data at the time, the foreground contribution was estimated from three small fields located at the edges of our LMC field. We then scaled the resulting foreground CMD to the entire LMC field by using the estimate for the number of Galactic sources from our synthetic model. Figure 9a shows the field CMD for LMC populations only, after subtracting Galactic foreground. The expected Galactic source counts is $`4\times 10^5`$ or about $`50\%`$. The uncertainties in the observed CMD and in the Galactic model result in negative density regions (dotted contours) in Figure 9. The average negative density is $`10^{3.2}`$ mag<sup>-2</sup>.
The luminosity function of the LMC giants is obtained by projecting the color-magnitude diagram perpendicular to the giant branch ridge line. The function is normalized to unity. Numerical values for the RGB luminosity function are given in Table 1 and shown in the inset to Figure 9. A strong feature of the LF is a significant excess at $`K_s12.5`$, due to the TRGB. From the analysis of the derivative of the apparent luminosity function, we derive the position of the TRGB at $`K_s=12.3\pm 0.1`$. Brightward of the TRGB, the number of RGB stars drops off. The increase in the number density at the faint end, $`K_s\text{ }>\text{ }14`$, is due to the increased contribution from Galactic M dwarfs (cf. Fig.5). At the bright end of the magnitude range, $`11<K_s<12`$, the luminosity function is nearly constant. It is also well above the expected number from extrapolated RGB counts. The fraction of RGB stars is small in this magnitude range and most of the stars contributing to the LF are on the AGB. As discussed above (§3.7), these stars tend to be oxygen-rich, but carbon-rich AGBs (and LPVs) are also present.
We select two $`2^{}\times 1^{}`$ fields, one near the optical center of the bar, and the other one at $`\alpha =93^{}`$, $`\delta =67.5^{}`$ (J2000.0), near the outer loop, to compare the observed M giants luminosity functions in distinct LMC environments. The outer field is probing the LMC’s outer loop delineated by the evolved stars (see §3.11). For each of the fields, we subtract the estimated Galactic foreground density scaled to the surface area of the fields. Similar to Fig. 9a, inaccuracies in both observations and models produce negative density regions. However, the average negative density in these regions is small compared to the giant branch, only $`10^{0.4}`$ mag<sup>-2</sup> for bar region and $`10^{0.5}`$ mag<sup>-2</sup> for loop field. Comparing the two CMDs qualitatively, we note that contribution of young OB stars and supergiants, at $`JK_s\text{ }<\text{ }0.2`$, appears stronger in the central regions of the LMC. Even though the luminosity functions in the respective fields appear different, careful analysis shows that the difference is superficial: $`\chi ^2/`$d.o.f.$`0.2`$, which gives no motivation to entertain any difference in the parent populations. The bar LF is similar to the luminosity function of the entire field (Figure 9a), which is not surprising because the source density in the LMC field is dominated by the LMC bar. The bar luminosity function also has a pronounced TRGB at $`K_s12.3`$ and excess number density due to AGBs at $`11<K_s<12`$. The sharp increase due to Galactic M stars, seen in Figure 9a at $`K_s\text{ }>\text{ }14`$, has disappeared in the bar field, because we have boosted the ratio of LMC/Galactic counts by narrowing down the field to the area of greatest LMC density. The off-bar LF shows only a mild increase in the source counts at the location of TRGB, but has the same, roughly constant profile at $`K_s<12`$, due to the AGB population, visible in the other two luminosity functions. To quantify both LFs, we present their numerical values in Table 1. The luminosity functions are given in relative units, normalized to unity. The table also gives the source counts for the LMC giant branch. For the entire LMC field we present the total counts per magnitude bin, and for the two smaller fields we give stellar density (counts mag<sup>-1</sup> deg<sup>-2</sup>).
We fit theoretical isochrones Girardi et al. (2000) to each giant branch in Figure 9 to test for differences in metallicity between central and outer parts of the LMC. We chose $`20`$ equally-spaced grid points in $`K_s`$ magnitude between $`14.3`$ and $`12.3`$ and compute the peak in the distribution in $`JK_s`$ at these fixed $`K_s`$ points. The difference between an isochrone (model) and the RGB (data) is characterized by the cost function (mean integrated square error):
$$f=\underset{j}{}[(JK_s)_{j,RGB}(JK_s)_{j,iso}]^2+w[K_s^{TRGB}12.3]^2,$$
(1)
where the second term is weighted measure of the match between theoretical magnitude at helium flash and observed TRGB. The weight $`w`$ is an adjustable parameter on the order of unity. The cost function (1) is minimized on a grid of parameter values, where the free parameters are the log-age $`\tau `$, metallicity $`Z`$, distance modulus $`\mu `$ and average reddening $`E_{BV}`$. The best fit isochrones are as follows: $`(\tau ,Z,\mu ,E_{BV})=(9.8\pm 0.3,0.004_{0.001}^{+0.002},18.45\pm 0.11,0.21\pm 0.07)`$ for the central field and $`(\tau ,Z,\mu ,E_{BV})=(9.8\pm 0.4,0.004_{0.001}^{+0.002},18.50\pm 0.13,0.13\pm 0.09)`$ for the outer field (all errors statistical). These results imply an age range for RGB populations from $`3`$ to $`13`$ Gyr with an average of $`6`$ Gyr. The slope degeneracy of the isochrones in the RGB makes specific tests of star formation history difficult. In particular, our preliminary RGB isochrone analysis cannot distinguish between a continuous and single/multiple burst star formation history of the Cloud prior to approximately $`4`$ Gyr ago. Overall, our results do not indicate a radial metallicity gradient and provide only marginal evidence for larger reddening in central fields. The absence of strong metallicity gradients in the LMC is in agreement with results of Olszewski et al. (1991), who found no evidence for abundance gradient for cluster system. Constant C/M star ratio across the face of the LMC Westerlund (1997) and Cepheid abundances Harris et al. (1983) also support this result. Our results imply the range of abundances for field populations $`0.8<`$\[Fe/H\]$`<0.5`$, which is in good agreement with the mean abundance \[Fe/H\]$`=0.58\pm 0.05`$ (systematic) $`\pm 0.30`$ (statistical) for the inner LMC disk Cole (1999). Our results agree with the disk abundance \[Fe/H\]$`=0.7`$ Cowley & Hartwick (1982), and with results of Bica et al. (1998), who derived the range $`1.1<`$\[Fe/H\]$`<0.4`$ with the average $`<[`$Fe/H$`]>=0.61\pm 0.11`$ from fields in the outer disk of the LMC.
The adopted detection threshold (SNR $`=10`$; §2) leads to the effective completeness limit of $`K_s14.3`$ in our data (cf. Figure 3). With this flux level, only the upper RGB is visible, leaving out both the AGB-bump at $`K_s16`$ and the red clump at $`K_s17`$. To resolve the giant branch down to $`K_s1617`$, we use 2MASS engineering data, which includes six LMC scans, positioned as shown in Figure 10. Each of the ‘deep’ scans has six times the standard exposure.
The color-magnitude diagram of the deep data is shown in Figure 11. Total number of sources in the diagram is $`87,093`$, of which $`69,878`$ ($`80\%`$) are in the three bar scans. Because of the bar dominance in deep data, the CMDs of the entire deep sample and the bar scans (Figs. 11a and 11b, respectively) are similar. The increased sensitivity reaches the AGB-bump at $`JK_s=0.7`$, $`K_s=15.8`$, but still shy of the red giant clump. As with the main dataset, we quantify the deep RGB population by fitting isochrones. The resulting best fit parameters are $`Z=0.004_{0.001}^{+0.002}`$, $`\tau =9.7\pm 0.3`$, $`\mu =18.50\pm 0.12`$, $`E_{BV}=0.19\pm 0.08`$. The uncertainties here are statistical errors, derived from the shape of the $`\chi ^2`$ surface near minimum. The range of ages for RGB populations inferred from the deep data is similar to that derived for the main data set: from $`3`$ to $`10`$ Gyr, with the average of $`5`$ Gyr. The other parameters are also consistent with the values derived from the regular 2MASS data. These estimates are in good agreement with recent results in the literature, e.g., the average reddening $`E_{BV}=0.20`$ Harris et al. (1997); the LMC distance from Key Project, $`\mu =18.5`$ Mould et al. (1999); the LMC distance from TRGB, $`\mu =18.59\pm 0.09`$ Sakai et al. (1999).
## 5 Summary
In this paper, we analyzed the near-infrared CMD of the Large Magellanic Cloud and identified the major stellar populations. The populations are identified based on isochrone fitting and matching the theoretical CMD colors of known populations to the observed CMD source density. Tables 2 and 3 summarize the contents of the CMD regions.
The main points of this preliminary analysis of 2MASS data are the following:
* The quantity and the quality of 2MASS data allow unprecedented look at the entire LMC. 2MASS has produced a rich sample of LMC sources, a few million stars, with the photometric accuracy of 3-4%. $`JHK`$ 2MASS photometry is potentially useful for studying the star-formation history of the Cloud. Cross-correlating 2MASS database with existing catalogs will provide homogeneous and accurate IR photometry of supergiants (Sanduleak catalog), Wolf-Rayet stars (Breysacher catalog), Cepheids (OGLE and EROS datasets), LBVs and LPVs;
* The color-color diagram is generally ill-suited to distinguish between giant (III) and dwarf (V) populations, especially in the color range $`0.5JK_s0.8`$. Nevertheless, the diagram may be useful in identifying some candidate LMC objects with infrared excess, such as obscured AGB stars, B\[e\] stars, or LMC protostars. In addition, the distribution of $`JK_s`$ colors of a population in a narrow $`JH`$ color range is a sensitive reddening test;
* Major populations can be identified based on the comparison of observed CMD features with theoretical positions of known populations. Isochrone overplotting provides tentative age and metallicity estimates. We identify a substantial LMC population of AGBs ($`\text{ }>\text{ }10^4`$ sources), and obscured AGBs ($`2000`$ sources);
* The luminosity function of the LMC giants is determined and tabulated. We find the RGB tip at $`K_s=12.3\pm 0.1`$. Our preliminary analysis of luminosity functions in two test fields suggest that luminosity function is the same in the bar and the outer regions of the Cloud;
* Fitting isochrones to the location of the giant branch (including TRGB) gives metal abundances consistent between fields. In particular, we derive average metallicity $`Z=0.004_{0.001}^{+0.002}`$ for our fields. Analysis of deep data gives the same average metallicity. Our results confirm the absence of strong radial metallicity gradient in the field populations of the LMC;
* The estimates of the distance modulus obtained from our isochrone fits to the RGB, are consistent with each other and the most recent results in the literature. The average reddening is marginally different between the bar and the outer field, the $`E_{BV}`$ for the bar field being greater. Distance modulus and reddening estimates from the analysis of deep data produces similar values;
* The ages of dominant RGB populations fall in the range from $`3`$ to $`13`$ Gyr, with the average age $`6`$ Gyr. Isochrone fits to the deep data produce similar age interval, from $`3`$ to $`10`$ Gyr, with the average age $`5`$ Gyr. A more detailed isochrone analysis is required to draw conclusions about the history of star formation in the LMC prior to $`34`$ Gyr ago;
* Carbon-rich long-period variables are noted as potential standard candles. Due to their significant numbers and narrow luminosity range (which may be parametrized through a period-luminosity or luminosity-color relations), these stars are ideally suited for studying the structure of the LMC along the line of sight.
## Acknowledgements
It is a pleasure to thank Andrew Cole for his insightful comments and numerous suggestions which lead to improvement of the paper. We are grateful to Michael Skrutskie for his careful reading of the manuscript. We also thank Shashi Kanbur for discussions regarding pulsating variables and Peter Wood for advice on long-period variables. This publication makes use of data products from the Two Micron All Sky Survey, which is a joint project of the University of Massachusetts and the Infrared Processing and Analysis Center, funded by the National Aeronautics and Space Administration and the National Science Foundation.
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.