id
stringlengths 30
36
| source
stringclasses 1
value | format
stringclasses 1
value | text
stringlengths 5
878k
|
|---|---|---|---|
no-problem/9904/astro-ph9904161.html
|
ar5iv
|
text
|
# A spectroscopically identified galaxy of probable redshift 𝑧=6.68
<sup>1</sup>Department of Physics and Astronomy, State University of New York at Stony Brook, Stony Brook, NY 11794–3800, USA
The detection and identification of distant galaxies is a prominent goal of observational cosmology because distant galaxies are seen as they were in the distant past and hence probe early galaxy formation, due to the cosmologically significant light travel time. We have sought to identify distant galaxies in very deep spectroscopy by combining a new spectrum extraction technique with photometric and spectroscopic analysis techniques. Here we report the identification of a galaxy of redshift $`z=6.68`$, which is the most distant object ever identified. The spectrum of the galaxy is characterized by an abrupt discontinuity at wavelength $`\lambda 9300`$ Å, which we interpret as the $`\mathrm{Ly}\alpha `$ decrement (produced by intervening Hydrogen absorption), and by an emission line at wavelength $`\lambda 9334`$, which we interpret as $`\mathrm{Ly}\alpha `$. The galaxy is relatively bright, and the ultraviolet luminosity density contributed by the galaxy alone is almost ten times the value measured at $`z3`$.
At near-infrared wavelengths, where background sky light is the dominant source of noise, the Hubble Space Telescope (HST) using the Space Telescope Imaging Spectrograph (STIS) is more sensitive than the Keck telescope because (1) the sky is considerably fainter from space and (2) considerably higher spatial resolution is attained from space, thus admitting far less contaminating sky light. To exploit the unique sensitivity of STIS at near-infrared wavelengths, the Space Telescope Science Institute and the STIS instrument team at the Goddard Space Flight Center initiated the STIS Parallel Survey, in which deep STIS observations are obtained in parallel with other observations<sup>1</sup>. We selected for analysis very deep observations acquired in slitless spectroscopy mode (which records the dispersed light of an entire field of view), because these observations are best suited for identifying distant galaxies.
The very deep observations were obtained by HST using STIS from 23 through 26 December, 1997 toward a region of sky flanking the Hubble Deep Field. Most observations consisted of a pair of images: a direct image taken using no filter and a dispersed image taken using the G750L grating. Additional observations consisted of only a direct image. The integration time of the direct images totaled 4.5 h over 82 exposures, and the integration time of the dispersed images totaled 13.5 h over 60 exposures.
We summed the direct and dispersed images using conventional image processing techniques. First, we reduced the images using standard pipeline software and applied corrections for flat-field variations and illumination pattern. Next, we registered each pair of images to a common origin, using pointing offsets determined from positions of bright stars in the direct images. Next, we measured the noise characteristics of the images, which are set by the background sky level and the readout noise. (The background sky level varies by as much as a factor of two between individual exposures.) Finally, we summed the direct images to form a summed direct image and summed the dispersed images to form a summed dispersed image, using optimal weights determined from the measured noise characteristics and rejecting deviant pixel values caused by cosmic ray events and hot pixels. The spatial resolution of the summed direct and dispersed images is $`\mathrm{FWHM}0.08`$ arcsec, the $`1\sigma `$ single pixel detection threshold of the summed direct image is $`26.2`$ mag arcsec<sup>-2</sup>, and the $`1\sigma `$ single pixel detection threshold of the summed dispersed image at $`\lambda 9800`$ Å is $`5.5\times 10^{18}`$ erg s<sup>-1</sup> cm<sup>-2</sup> Å<sup>-1</sup> arcsec<sup>-1</sup>.
We extracted one-dimensional spectra from the summed dispersed image using a new spectrum extraction technique. The extraction is made especially difficult because the image of the field is covered by light of faint galaxies, which when dispersed overlaps in a complicated way. To overcome this difficulty, we used the summed direct image to determine not only the exact locations but also the exact two-dimensional spatial profiles of the spectra on the summed dispersed image. These spatial profiles are crucial because (1) they provide the “weights” needed to optimally extract the spectra, and (2) they provide the models needed to deblend the overlapping spectra and determine the background sky level. First, we identified objects in the summed direct image, using the SExtractor program of Bertin & Arnouts<sup>2</sup>. Roughly 250 objects were identified in the summed direct image. Next, we modeled each pixel of the summed dispersed image as a linear sum of contributions from (1) relevant portions of all overlapping neighboring objects and (2) background sky. The model parameters included roughly 250,000 “object” parameters (from roughly 1000 spectral pixels each of roughly 250 objects) and 5000 “sky” parameters (from roughly 1000 fourth-order polynomials). Finally, we minimized $`\chi ^2`$ between the model and the data with respect to the model parameters to form one-dimensional spectra. (In practice, we imposed a condition of smoothness on the spectral variation of the sky in order to reduce the number of sky parameters. First, we smoothed the model of the sky by 10 pixels in the spectral direction, and we subtracted this model from the data. Next, we performed the extraction again, but this time with sky subtraction turned off. In this way we reduced the effective number of sky parameters from roughly 5000 to more like 500. We found by experimentation that this number of parameters is required to represent the spatial and spectral variations of the sky.)
We measured photometric redshifts from the one-dimensional spectra using a variation of the photometric redshift technique described previously by Lanzetta, Yahil, & Fernández-Soto<sup>3,4</sup> and Fernández-Soto, Lanzetta, & Yahil<sup>5</sup>.
The results of the analysis are photometric redshift measurements of roughly 250 objects identified in the summed direct image. Here we focus on one particular galaxy identified by the analysis at a very high redshift, which we designate as galaxy A. The J2000 coordinates of galaxy A are $`\alpha =12:36:27.3`$ and $`\delta =+62:17:55.9`$. Interpretation of the spectrum of galaxy A is especially straightforward because this galaxy does not overlap any other object in the dispersion direction. This means that in this case much of the machinery of the extraction technique described above (involving deblending overlapping spectra) does not actually come into play, although the technique is required in order to accurately determine the background sky level.
Figure 1 shows the direct image of galaxy A and its neighbors in the top panel. The dispersed image of galaxy A and its neighbors and a map of the detected emission lines are shown in the next two panels. The dispersed image of galaxy A with best-fit model spatial profiles of all objects but galaxy A subtracted is shown in the next panel. The one-dimensional spectrum of galaxy A, a redshifted spectrum of a moderate-redshift ($`z=4.421`$) galaxy<sup>6</sup>, and the one-dimensional spectrum of galaxy A cast into 325 Å bins together with the best-fit spectrophotometric template are shown in the next three panels. Figure 2 shows the redshift likelihood function of galaxy A. The spectrum of galaxy A is characterized by (1) an abrupt discontinuity at wavelength $`\lambda 9300`$ Å, with detectable continuum emission at wavelengths $`\lambda =93009950`$ Å and an absence of detectable continuum emission at wavelengths $`\lambda _{}^<\mathrm{\hspace{0.25em}9300}`$ Å, and (2) an emission line at wavelength $`\lambda =9337\pm 6`$ Å. The redshift likelihood function of galaxy A is characterized by a maximum at redshift $`z=6.84`$ and a local maximum at redshift $`z=6.66`$, with the likelihood values of the primary and secondary maxima statistically indistinguishable. The energy flux of the emission line is $`2.6\pm 0.5\times 10^{17}`$ erg s<sup>-1</sup> cm<sup>-2</sup>. We interpret the abrupt discontinuity as the $`\mathrm{Ly}\alpha `$ decrement and the emission line as $`\mathrm{Ly}\alpha `$, in which case the redshift of the galaxy determined from the emission line is $`z=6.68\pm 0.005`$.
Several lines of evidence support the validity of the spectrum extraction and the redshift identification:
First, the integrated energy flux measured from the direct image is in excellent agreement with the integrated energy flux measured from the dispersed image. The clear magnitude of galaxy A measured from the direct image is $`AB(\mathrm{clear})=27.67\pm 0.09`$, and the clear magnitude of galaxy A measured from the dispersed image is $`AB(\mathrm{clear})=27.72\pm 0.29`$ (which is obtained by integrating the product of the spectrum and the unfiltered spectrograph and system throughput). This demonstrates that the continuum emission detected at wavelengths $`\lambda 93009950`$ Å—which is crucial for establishing the redshift of galaxy A—is of exactly the amount required to explain the direct image. The clear magnitude of the emission line alone of galaxy A measured from the dispersed image is $`AB(\mathrm{clear})=29.15\pm 0.24`$. This demonstrates that the emission line alone cannot explain the direct image; rather, additional continuum emission must be present, namely the continuum emission detected at wavelengths $`\lambda 93009950`$ Å. It is clearly unlikely that the analysis could have missed significant continuum emission at wavelengths $`\lambda <9300`$ Å—where the spectrograph and system throughput is relatively high—and instead detected spurious emission (of exactly the amount required to explain the direct image) at wavelengths $`\lambda >9300`$ Å.
Next, various arguments demonstrate the reality of other faint emission lines visible in the dispersed image. For example, we applied the SExtractor program to the smoothed dispersed image to objectively identify emission lines, setting the detection threshold such that nothing was detected in the negative of the image. The resulting segmentation map is shown in the third panel of Figure 1. Every emission line detected in the dispersed image by the SExtractor program can be attributed to a galaxy detected in the direct image. Specifically, the bottom-most emission lines arise in a very faint galaxy, designated galaxy B in Figure 1. The spectrum of galaxy B is relatively uncomplicated, because the galaxy just barely overlaps other galaxies above and below in the dispersion direction. Only very weak continuum emission is detected from the galaxy, but at least three emission lines are clearly evident in the dispersed image: an emission line to the right and below the emission line of galaxy A, another emission line in the same row just over 1/2 of the way from the left-hand edge of Figure 1, and another emission line in the same row off the right-hand edge of Figure 1. The identifications of the three emission lines are (left to right) Mg II $`\lambda 2800`$, He II $`\lambda 3203`$, and \[O II\] $`\lambda 3727`$ at a redshift of $`z=1.213`$, which is substantiated by comparison with the ultraviolet spectrum of a starburst galaxy of Kinney et al.<sup>7</sup> Furthermore, every emission line visible in Figure 1 (including another emission line below and to the right of the emission line of galaxy A and a “complex” of emission lines above and to the right of the emission line of galaxy A) is robust against image processing and cosmic ray rejection methods and is exactly coincident with the spatial profile of a galaxy detected in the direct image, as is evident from the bottom image of Figure 1.
Next, the abrupt discontinuity is consistent with interpretation as the $`\mathrm{Ly}\alpha `$ decrement but inconsistent with the interpretation as other spectral breaks commonly observed in galaxy spectra. The average energy flux density measured from the dispersed image at wavelengths $`\lambda =72509250`$ Å is $`f_\nu (8250)=0.01\pm 0.04`$ and at wavelengths $`\lambda =94009950`$ Å is $`f_\nu (9675)=0.67\pm 0.22`$ $`\mu `$Jy, which implies a decrement $`1f_\nu (8250)/f_\nu (9675)=1.01\pm 0.06`$. The corresponding $`3\sigma `$ lower limit to the break amplitude is 5.9, which exceeds by a significant factor the largest measured break amplitudes $`2.6`$ of early-type galaxies and $`3`$ of main-sequence stars (see Spinrad et al.<sup>8</sup>).
Next, the statistically insignificant but suggestive “dip” in the spectrum of galaxy A at wavelengths between 9300 and 9500 Å matches qualitatively the expectation for stellar continuum radiation viewed through interstellar neutral Hydrogen, which is expected to imprint a damped $`\mathrm{Ly}\alpha `$ absorption feature. Specifically, the redshifted spectrum of the moderate-redshift galaxy shows a corresponding dip, as is evident from comparison of the middle and top one-dimensional spectra of Figure 1. A similar feature is also present in the composite spectrum of 12 high-redshift ($`z3`$) galaxies presented by Lowenthal et al.<sup>9</sup>
Finally, the redshift determined from the emission line is in excellent agreement with the redshift determined from the redshift likelihood function. Although difficult to quantify, the detection of an emission line at exactly the wavelength predicted by the photometric redshift measurement lends a posteriori support to the redshift identification. In summary, the spectrum of galaxy A matches exactly the spectrum expected of a galaxy of redshift approaching $`z=7`$ but is unlike the spectra expected of lower-redshift galaxies.
The most striking property of galaxy A is that it is relatively bright at wavelengths longward of $`\mathrm{Ly}\alpha `$. Although galaxy A is exceedingly faint in any ordinary optical bandpass (due to very strong absorption by the $`\mathrm{Ly}\alpha `$ forest), its continuum energy flux density at observed-frame wavelength $`\lambda 9800`$ Å or rest-frame wavelength $`\lambda 1300`$ Å is $`f_\nu 1`$ $`\mu `$Jy, which is comparable to the continuum energy flux densities at similar rest-frame wavelengths of galaxies identified by Steidel and collaborators<sup>10</sup> at redshifts $`z3`$. This corresponds to an unobscured star formation rate of $`17(147)h^2`$ $`M_{}`$ yr<sup>-1</sup> for $`q_0=0.5(0.0)`$, using the relationship between ultraviolet luminosity and star formation rate with a Salpeter initial mass function of Madau et al.<sup>11</sup>. Taking the redshift range searched for very high redshift galaxies to extend from $`z=6`$ to 7, the ultraviolet luminosity density contributed by galaxy A alone is almost ten times the value measured at $`z3`$. If the galaxy is typical of the very high redshift galaxy population, then apparently (1) very high redshift galaxies are luminous at rest-frame ultraviolet wavelengths and (2) the unobscured cosmic star formation rate may be substantially larger at $`z7`$ than at lower redshifts.
Received .
1. Gardner et al. The STIS Parallel Survey: Introduction and First Results. Astrophys. J. 492, L99–L102 (1998).
2. Bertin, E. & Arnouts, S. SExtractor: Software For Source Extraction. Astr. Astrophys. Suppl. 117, 393–404 (1996).
3. Lanzetta, K. M., Yahil, A., & Fernández-Soto, A. Star-forming galaxies at very high redshifts. Nature 381, 759–763 (1996).
4. Lanzetta, K. M., Fernández-Soto, A., & Yahil, A. Photometric Redshifts of Galaxies in the Hubble Deep Field in “The Hubble Deep Field, Proceedings of the Space Telescope Science Institute 1997 May Symposium,” ed. M. Livio, S. M. Fall, & P. Madau (Cambridge: Cambridge University Press) in the press (1998).
5. Fernández-Soto, A., Lanzetta, K. M., & Yahil, A. A new catalog of photometric redshifts in the Hubble Deep Field. Astrophys. J., in the press (1998).
6. Steidel, C. C., Adelberger, K. L., Giavalisco, M., Dickinson, M., & Pettini, M Lyman Break Galaxies at $`z>4`$ and the Evolution of the UV Luminosity Density at High Redshift. Astrophys. J., submitted (astro-ph/9811399).
7. Kinney, A. L., Calzetti, D., Bohlin, R. C., McQuade, K., Storchi-Bergmann, T., & Schmitt, H. R. Template Ultraviolet to Near-Infrared Spectra of Star-Forming Galaxies and Their Application to K-Corrections. Astrophys. J. 467, 38–60 (1998).
8. Spinrad, H. Stern, D., Bunker, A., Dey, A., Lanzetta, K., Yahil, A., Pascarelle, S., & Fernández-Soto, A. A Sub–galactic Pair at $`z=5.35`$. Astr. J., 116, 2617–2623 (1998).
9. Lowenthal, J. D., Koo, D. C., Guzmán, R., Gallego, J., Phillips, A. C., Faber, S. M., Vogt, N. P., Illingworth, G. D., & Gronwall, C. Keck Spectroscopy of Redshift $`z`$ Approximately 3 Galaxies in the Hubble Deep Field. Astrophys. J. 481, 673-688 (1997).
10. Steidel, C., Giavalisco, M., Pettini, M., Dickinson, M., & Adelberger, K. Spectroscopic Confirmation of a Population of Normal Star-forming Galaxies at Redshifts $`z>3`$. Astrophys. J. 462, L17–L21 (1996).
11. Madau, P., Pozzetti, L., & Dickinson, M. The Star Formation History of Field Galaxies. Astrophys. J. 498, 106–116 (1998).
###### Acknowledgements.
ACKNOWLEDGEMENTS. We are grateful to H. Spinrad and to B. Woodgate, B. Hill, and the rest of the STIS instrument team for important discussions. This research was supported by NASA and NSF. CORRESPONDENCE should be addressed to H.-W.C. (email: hchen@sbastr.ess.sunysb.edu).
|
no-problem/9904/hep-th9904118.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
Recent developments in string theory are understandings of the correspondence between the brane dynamics and the physics of gauge theories in various dimensions. Hanany and Witten have first pointed out that $`N=4`$ supersymmetric gauge theory in three-dimensions is realized as worldvolume effective theory on D3-branes stretched between two NS5-branes and the “mirror symmetry” can be understand as a consequence of $`SL(2,𝐙)`$ duality of Type IIB superstring theory.
Their construction is extended to different numbers of supersymmetries and gauge theory in various dimensions (for a review see ). The advantage of this approach is that we can geometrically analyze the structure of the moduli spaces of vacua as the configuration space of branes. The geometrical construction of moduli space make easy to classify possible supersymmetric vacuum.
From the point of view of brane solutions in Type IIB superstring theory, the three-dimensional Hanany-Witten type configuration is generalized to the case including rotated 5-branes at arbitrary angles . It contain the interesting 5-brane solution preserving $`3/16`$ of supersymmetry. This means that in three-dimensions we have $`N=3`$ supersymmetric theory. One of the realization of the three-dimensional $`N=3`$ supersymmetry is obtained by adding the Chern-Simons term . So the brane configuration with $`3/16`$ supersymmetry can be associated with three-dimensional Chern-Simons theory.
Moreover, the systematic classification of possible Hanany-Witten type configuration with various supersymmetries has been studied in ref. . Since the Type IIB brane configuration with $`3/16`$ supersymmetry contains a $`(p,q)`$5-brane which is a bound state of NS5 and D5-brane, it is reasonable to consider that the replacement of one of the NS5-branes with the $`(p,q)`$5-brane is related to adding the Chern-Simons interaction. Under this identification, it is found that the vacuum structure of Maxwell Chern-Simons theory is consistent with the brane picture and a “mirror symmetry” is similarly realized by Type IIB S-duality. This “mirror symmetry” in supersymmetric Chern-Simons theory is recently confirmed by using a generalized Fourier transformation in ref. .
In this paper, we investigate the correspondence between the $`(p,q)`$5-brane configuration introduced by ref. and Yang-Mills (Maxwell) Chern-Simons theories in detail. More specifically, we construct a explicit Lagrangian for the supersymmetric Chern-Simons theories and identify vevs of emergent fields with moduli parameters of Type IIB branes. This identification gives us the knowledge to understand the complicated dynamics of three-dimensional massive gauge theories and the dynamics of Type IIB branes.
This paper is organized as follows: In section 2 we briefly review the construction of the brane configurations with $`(p.q)`$5-brane. These configurations in Type IIB theory are reduced from M5-brane configurations in M-theory with various supersymmetries. We give all possible Hanany-Witten type configuration with $`N2`$ supersymmetry. In section 3, we see how to describe the Coulomb branch by vevs of adjoint scalars in supersymmetric Chern-Simons theory. The number of the adjoint scalar fields corresponds to the number of freely moving directions of D3-branes. Finally, we consider the vacuum of Maxwell Chern-Simons theory with matter superfields. In this model, there are two distinct vacua. One of vacua, which is called as a symmetric phase, exists on in the Maxwell Chern-Simons Higgs system. We can find the brane configuration of the symmetric phase. We also identify the topological vortex in the asymmetric phase as a bound state of strings.
## 2 Supersymmetric Configuration with $`(p,q)`$5-brane
In this section we briefly review the construction of the brane configuration with $`(p,q)`$5-brane.
The brane configurations in Type IIB string theory which are given by Hanany and Witten describe three dimensional supersymmetric gauge theories. In this configuration D3-branes are suspended between two parallel NS5-branes and the $`3+1`$ dimensional worldvolume of D3-brane is compactified on a finite line segment. We denote the extended directions of two NS5-branes and $`N_c`$ D3-branes as the following symbols:
$`2\times \mathrm{NS5}`$ $`:`$ $`(012345),`$
$`N_c\times \mathrm{D3}`$ $`:`$ $`(012|6|),`$
where the numbers in the parenthesis are the worldvolume directions and the vertical lines on both side of $`6`$ mean that the worldvolume of D3-branes are restricted to finite interval in the $`x^6`$-direction. The “left” NS5-brane is located at $`x^6=L/2`$ and the “right” is at $`x^6=L/2`$.
The low-energy effective theory on $`N_c`$ parallel D3-branes is $`3+1`$ dimensional $`SU(N_c)`$ supersymmetric Yang-Mills theory in the limit that string scale $`l_s`$ goes to zero, but this compactification on the line reduces the D3-brane worldvolume theory to $`2+1`$ dimensions. Upon this Kaluza-Klein reduction on the line segment we find the gauge coupling constant $`g`$ of this worldvolume theory is given by
$$\frac{1}{g^2}=\frac{L}{g_s},$$
(2.1)
where $`g_s`$ is a string coupling constant of Type IIB theory and $`L`$ is a length of the line segment, namely, a distance between two NS5-branes in the $`x^6`$-direction.
In this configuration the presence of these two kinds of branes preserves 8 of 32 supercharges of Type IIB string theory since each kind of brane preserves half of supercharges. Therefore we have $`N=4`$ supersymmetry in $`2+1`$ dimensions.
To add matter in the fundamental representation of the gauge group we can introduce $`N_f`$ D5-branes, which have the worldvolume:
$`N_f\times \mathrm{D5}`$ $`:`$ $`(012789),`$
without breaking any supersymmetries. The matter hypermultiplets come from strings stretched between the $`N_c`$ suspended D3-branes and the $`N_f`$ D5-branes. So $`N_f`$ represents the number of flavors.
In addition to these types of branes there also exists the $`(p,q)`$5-brane which is a bound state of NS5- and D5-branes. One of the extension of the above setup is replacement of the right NS5-brane in the Hanany-Witten configuration with this $`(p,q)`$5-brane. However, if we simply substitute the $`(p,q)`$5-brane for the right NS5-brane and the left NS5-brane and the right $`(p,q)`$5-brane is parallel then all supersymmetry is broken. To preserve some of the supercharges we must rotate the $`(p,q)`$5-brane by some angles.
In order to count the residual supercharges we now lift the above Type IIB configuration to M5- and M2-brane configurations in M-theory. T-duality in the $`x^2`$-direction maps the NS5-brane and the D5-brane into the NS5-brane and the D4-brane in Type IIA theory, respectively. By lifting to M-theory these objects become M5-branes. The NS5-brane corresponds to the M5-brane wrapping on the $`x^2`$-direction and the D5-brane corresponds to the M5-brane wrapping on the 11th-direction $`x^{10}`$. So two cycles of the torus $`T^2`$ with coordinates $`(x^2,x^{10})`$ are related to NSNS-charge and RR-charge in Type IIB theory. Similarly, $`(p,q)`$5-brane in Type IIB theory is also unified as a M5-brane which is obliquely wrapping on the torus $`T^2`$ at some angle $`\theta `$. This angle and charges of the $`(p,q)`$5-brane are related each other as follows
$$\mathrm{tan}\theta =\frac{2\pi R_{10}p}{2\pi R_2q}=g_s\frac{p}{q},$$
(2.2)
where $`R_2`$ and $`R_{10}`$ are radii of the torus $`T^2`$.
Using the above procedure D3-branes in Type IIB theory become M2-branes stretched between two M5-branes in M-theory. In this situation the right M5-brane tilted by the angle $`\theta `$ can be rotated by more three angles in the configuration space of M-theory. So, we have a general configuration in M-theory:
$`\mathrm{M5}`$ $`:`$ $`(012345)`$
$`N_c\times \mathrm{M2}`$ $`:`$ $`(01|6|)`$
$`\mathrm{M5}^{}`$ $`:`$ $`(01\left[{\displaystyle \genfrac{}{}{0pt}{}{2}{10}}\right]_\theta \left[{\displaystyle \genfrac{}{}{0pt}{}{3}{7}}\right]_\psi \left[{\displaystyle \genfrac{}{}{0pt}{}{4}{8}}\right]_\phi \left[{\displaystyle \genfrac{}{}{0pt}{}{5}{9}}\right]_\rho ),`$
where the symbol $`\left[\genfrac{}{}{0pt}{}{\mu }{\nu }\right]_\theta `$ means that the extended direction of the $`\mathrm{M5}^{}`$-brane are rotated by the angle $`\theta `$ in the $`(x^\mu ,x^\nu )`$-space. If the angle $`\theta `$ is zero $`\mathrm{M5}^{}`$-brane become a rotated NS5-brane in Type IIB configuration space. On the other hand, the case of $`\theta =\pi /2`$ corresponds to a D5-brane.
The presence of these branes imposes the constrains on the 11-dimensional Killing spinors $`ϵ`$
$`\mathrm{M5}`$ $`:`$ $`\mathrm{\Gamma }_{012345}ϵ=ϵ,`$ (2.3)
$`\mathrm{M2}`$ $`:`$ $`\mathrm{\Gamma }_{016}ϵ=ϵ,`$ (2.4)
$`\mathrm{M5}^{}`$ $`:`$ $`R\mathrm{\Gamma }_{012345}R^1ϵ=ϵ,`$ (2.5)
where $`R`$ is the rotation matrix in the spinor representation and give by
$$R=\mathrm{exp}\left\{\frac{\theta }{2}\mathrm{\Gamma }_{2,10}+\frac{\psi }{2}\mathrm{\Gamma }_{37}+\frac{\phi }{2}\mathrm{\Gamma }_{48}+\frac{\rho }{2}\mathrm{\Gamma }_{59}\right\}.$$
The numbers of the remaining supersymmetry are obtained by solving eqs. (2.3)-(2.5) simultaneously. All possible solutions are completely classified in ref. by setting some relations between four angles and we have $`1/16`$, $`1/8`$, $`3/16`$ and $`1/4`$ of the original 32 supercharges. These fractions of supersymmetry correspond to $`N=1,2,3,4`$ supersymmetries, respectively, in $`2+1`$ dimensions.
In present paper we treat only the configuration for $`N2`$ supersymmetry. This is because quantum corrections of higher supersymmetric is suppressed rather than the $`N=1`$ case and it is not so complicated to find the correspondence to the branelogy.
$`N2`$ supersymmetric configurations are divided into the following two cases. One is the NS5 configuration with $`N=2,4`$ supersymmetry where the right 5-brane is the NS5-brane. The other is the configuration including $`(p,q)`$5-brane with $`N=2,3`$.
We first consider the NS5 configuration. For $`N=4`$ configuration we must set all angles to zero, $`\theta =\psi =\phi =\rho =0`$. The right $`\mathrm{M5}^{}`$-brane is a NS5-brane in Type IIB theory, which is parallel to the left NS5-brane. This is nothing but the Hanany-Witten configuration. As we mentioned above, this configuration describes $`N=4`$ $`SU(N_c)`$ supersymmetric Yang-Mills theory. This theory includes three real scalar fields in adjoint representation. Vevs of these adjoint scalars and scalars dual to the gauge fields parametrize the Coulomb branch. On the other hand, D3-branes in the brane configuration can freely move along the $`(x^3,x^4,x^5)`$-space. These positions of D3-branes on NS5-brane correspond to the vevs of real adjoint scalars. In addition the $`SO(3)`$ rotation group in the $`(x^3,x^4,x^5)`$ can be identified with a part of the $`R`$-symmetry $`SU(2)_V`$ of $`N=4`$ supersymmetry algebra in three dimensions. The other part of the $`R`$-symmetry $`SU(2)_H`$ corresponds to rotations in the $`(x^7,x^8,x^9)`$-space.
The $`N=2`$ NS5 configuration is obtained by a rotation of the right NS5-brane. In this rotation we must set same values on two pairs of angles. Since $`\theta `$ is zero in NS5 configuration one of our choice for angles is $`\theta =\rho =0`$ and $`\phi =\psi 0`$. The special case of $`\phi =\psi =\pi /2`$ is investigated in ref. and this brane configuration space well describes the moduli space of vacua of $`N=2`$ supersymmetric Yang-Mills theory. The Coulomb branch of this theory is parametrized by vevs of one real massless adjoint scalar and dual gauge fields. In this configuration D3-branes can not move in the $`(x^3,x^4)`$-space any longer since two NS5-branes are completely twisted in that space. Then the vev of real adjoint scalar corresponds to the positions of D3-branes in $`x^5`$-direction. For the general value of $`\psi `$ this rotation angle is related to a mass of two real adjoint scalars , which originally belong to the $`N=4`$ vector multiplet. Therefore we can identify the rotation of NS5-brane with the supplement of the mass term of two adjoint scalars into $`N=4`$ theory. Actually, this brane rotation breaks the $`SU(2)_V\times SU(2)_H`$ rotation symmetry to $`SO(2)U(1)_R`$, which coincide with the $`N=2`$ $`R`$-symmetry group.
Let us next mention about the configuration with $`(p,q)`$5-brane. This configuration is realized by setting $`\theta 0`$.
A maximal supersymmetric configuration with $`(p,q)`$5-brane is the case of $`\theta =\psi =\phi =\rho `$. This configuration preserves $`N=3`$ supersymmetry in $`2+1`$ dimensions. Since the all rotation angles are same except for the signature, the rotation symmetry $`SU(2)_V\times SU(2)_H`$ in the $`(x^3,x^4,x^5)`$\- and $`(x^7,x^8,x^9)`$-spaces breaks to a diagonal $`SU(2)_D`$. This rotational symmetry $`SU(2)_D`$ can be considered as the $`R`$-symmetry of the three-dimensional $`N=3`$ supersymmetry algebra.
$`N=2`$ configuration is given by setting $`\rho =\theta \phi =\psi `$, that is, two different sets of same angles. If we take $`\theta =\rho =0`$, then we have the $`N=2`$ NS5 configuration as mentioned above. Similarly, we can choose as $`\phi =\psi =0`$. In this case the right 5-brane is $`(p,q)`$5-brane but still $`N=2`$ supersymmetric.
It is field theoretically shown that the three-dimensional $`N=3`$ maximal supersymmetric system can be constructed by adding Chern-Simons terms . So one expect that the worldvolume effective theory on this $`2+1`$-dimensional intersection of rotated 5-branes is gauge theories with Chern-Simons terms. In fact, since there exists a non-trivial vev of axion (RR-scalar) field $`C_0`$ around $`(p,q)`$5-brane in Type IIB theory, the coupling $`C_0FF`$ on the D3-brane induces the Chern-Simons term with coupling $`\kappa =p/q`$ .
In the following section, we investigate the correspondence between these supersymmetric configurations with the $`(p,q)`$5-brane and the moduli space of vacua of Chern-Simons theories.
## 3 Coulomb Branches of The Non-Abelian Theories
### 3.1 $`N=3`$ supersymmetric Yang-Mills Chern-Simons theory
We first begin with the $`N=3`$ configuration. The configuration in Type IIB theory is written as follows in the notation of the previous section:
$`\mathrm{NS5}`$ $`:`$ $`(012345)`$
$`N_c\times \mathrm{D3}`$ $`:`$ $`(012|6|)`$
$`(p,q)5`$ $`:`$ $`(01\left[{\displaystyle \genfrac{}{}{0pt}{}{3}{7}}\right]_\theta \left[{\displaystyle \genfrac{}{}{0pt}{}{4}{8}}\right]_\theta \left[{\displaystyle \genfrac{}{}{0pt}{}{5}{9}}\right]_\theta ),`$
where $`\theta =\mathrm{tan}^1\left(g_s\frac{p}{q}\right)`$. There originally exists $`SU(N_c)`$ Yang-Mills gauge theory on the D3-branes. As we have explained, the right $`(p,q)`$5-brane provides the Chern-Simons term into the Yang-Mills theory. So we have the Yang-Mills Chern-Simons theory as the effective theory on this configuration. The coupling constant of Chern-Simons term is given by $`\kappa =p/q`$. However, the coupling constant of the non-Abelian Chern-Simons theory should be a integer in order to preserve gauge invariance under a large transformation. Therefore we set the charges of $`(p,q)`$5-brane as $`(p,q)=(n,1)`$.
$`N=3`$ supersymmetric field theory in three dimensions has the $`SU(2)_D`$ $`R`$-symmetry. The vector multiplet contains one spin 1 massive vector field $`A_\mu `$, three spin $`1/2`$ spinors $`\lambda _i`$, three spin 0 real adjoint fields $`X_i`$ and one spin $`1/2`$ spinor $`\chi `$. Three spinors and adjoint scalars form a triplet under $`SU(2)_D`$. The bosonic part of the Lagrangian of this $`N=3`$ supersymmetric Yang-Mills Chern-Simons theory is given by
$`_\mathrm{B}`$ $`=`$ $`{\displaystyle \frac{1}{g^2}}\mathrm{Tr}\left\{{\displaystyle \frac{1}{2}}F_{\mu \nu }^{}{}_{}{}^{2}+\left(_\mu X_i\right)^2+{\displaystyle \frac{1}{2}}[X_i,X_j]^2\left({\displaystyle \frac{\kappa g^2}{4\pi }}\right)^2X_{i}^{}{}_{}{}^{2}\right\}`$ (3.2)
$`+{\displaystyle \frac{\kappa }{4\pi }}\mathrm{Tr}\left\{ϵ^{\mu \nu \rho }\left(A_\mu _\nu A_\rho +{\displaystyle \frac{2}{3}}iA_\mu A_\nu A_\rho \right){\displaystyle \frac{i}{3}}ϵ^{ijk}X_i[X_j,X_k]\right\}.`$
Due to the Chern-Simons term the vector field topologically gains mass $`M=\left|\frac{\kappa g^2}{4\pi }\right|`$ which is the same mass as the adjoint scalars have. Note that in the limit of $`\kappa 0`$, which corresponds to the limit $`\theta 0`$ in the brane configuration, one obtains the $`N=4`$ supersymmetric pure Yang-Mills Lagrangian.
The Coulomb branch of supersymmetric gauge theory is described by solutions of the following flatness condition for the adjoint scalar fields
$$M^2X_i\frac{M}{2}iϵ_{ijk}[X_j,X_k][X_j,[X_i,X_j]]=0.$$
(3.3)
For real $`X_i`$, the general solution of this condition is $`X_i=0`$. So the Coulomb branch of this theory is completely lifted because of the mass term coming from the Chern-Simons interaction.
From the point of view of the brane configuration, it is easy to understand this Coulomb branch moduli as the following. In the $`N=4`$ brane configuration, $`N_c`$ D3-branes can freely moved on the $`(x^3,x^4,x^5)`$-space since two NS5-branes are parallel. So we can identify positions of D3-branes in the $`(x^3,x^4,x^5)`$-space with vevs of the three real adjoint scalar fields, which describe a part of the Coulomb branch.
However, in the $`N=3`$ configuration, since the left NS5-brane and the right $`(p,q)`$5-brane is completely twisted in $`(x^3,x^4,x^5)`$\- and $`(x^7,x^8,x^9)`$-space, D3-branes can not move anywhere. This means that the corresponding vevs of the adjoint scalars must be at origin, namely, $`X_i=0`$, which coincide with the solution of the flatness condition (3.3).
The mass of the adjoint scalars can be written in terms of the rotation angle of the $`(p,q)`$5-brane as
$$\left|\frac{\kappa g^2}{4\pi }\right|=\frac{1}{4\pi L}|\mathrm{tan}\theta |,$$
(3.4)
where we use the relation (2.1) and (2.2). This relation also agree with the results in ref. .
In the $`\theta 0`$ limit the mass of the adjoint scalar fields vanishes and at the same time the Chern-Simons interaction is dropped. In the brane configuration the right $`(p,q)`$5-brane become NS5-brane and to be parallel to the left NS5. So we have $`N=4`$ supersymmetric pure Yang-Mills theory. On the other hand, in the limit $`\theta \pi /2`$ the mass of the vector multiplet goes to infinity and all gauge degrees of freedom are decoupled. In this limit the right $`(p,q)`$5-brane become D5-brane which extends in the $`(x^0,x^1,x^2,x^7,x^8,x^9)`$-space. This configuration also recover the $`N=4`$ supersymmetry again and appears in the Higgs branch of the Hanany-Witten configuration .
### 3.2 Breaking to $`N=2`$
We next consider the supersymmetry breaking of $`N=3`$ theory down to $`N=2`$. In the brane language this is done by making two of the angles different from the others. Namely, we have the configuration
$`\mathrm{NS5}`$ $`:`$ $`(012345)`$
$`N_c\times \mathrm{D3}`$ $`:`$ $`(012|6|)`$ (3.5)
$`(p,q)5`$ $`:`$ $`\left(012\left[{\displaystyle \genfrac{}{}{0pt}{}{3}{7}}\right]_\psi \left[{\displaystyle \genfrac{}{}{0pt}{}{4}{8}}\right]_\psi \left[{\displaystyle \genfrac{}{}{0pt}{}{5}{9}}\right]_\theta \right).`$
As the result, the corresponding mass of the adjoint scalars become different.
We can field theoretically understand this supersymmetry breaking. Originally $`N=3`$ vector multiplet contains the fields $`(A_\mu ,\lambda _i,X_i,\chi )`$. If we give the different mass $`\mu `$ to the $`X_1,X_2`$ and $`\lambda _3,\chi `$, these fields form the $`N=2`$ neutral chiral supermultiplet $`\mathrm{\Phi }`$. The rests of the fields $`(A_\mu ,\lambda _a,X_3)`$ are contents of the $`N=2`$ vector multiplet, which still have the mass $`\left|\frac{\kappa g^2}{4\pi }\right|`$.
In this $`N=2`$ theory three adjoint scalars are massive as long as $`\psi 0`$. So the Coulomb branch of this theory is still lifted and there is no moduli. This corresponds to the twisting configuration of the NS5-brane and the rotated $`(p,q)`$5-brane. However, if we set $`\psi =0`$ the NS5-brane and $`(p,q)`$5-brane become parallel in the $`(x^3,x^4)`$-space. Therefore D3-branes can now freely move in the $`(x^3,x^4)`$-space. In this configuration, two adjoint scalars $`X_1`$ and $`X_2`$ become massless and the Coulomb branch appears.
## 4 Broken and Unbroken Phases of Abelian Theory
In this section we consider the Higgs branches of the $`N=2`$ supersymmetric Chern-Simons theory. We first briefly review the description of the Higgs branch in the Hanany-Witten configuration. The inclusion of the matter hypermultiplets corresponds to the addition of the D5-branes, which extend in the $`(x^0,x^1,x^2,x^7,x^8,x^9)`$-space, between two NS5-branes. The matter hypermultiplets come from the open string stretched between D5-branes and D3-branes<sup>1</sup><sup>1</sup>1There is a couple of chiral fields with opposite chirality with respect to orientation of strings.. For simplicity, we consider only Abelian theory with one matter hypermultiplet, which includes two complex scalar fields $`(q,\stackrel{~}{q})`$ from now on.
If we denote the distance between the D5-brane and the D3-brane in the $`(x^3,x^4,x^5)`$-space as $`d`$, the mass of the hypermultiplet is given by
$$m=T_{\mathrm{F1}}d=\frac{d}{2\pi l_{s}^{}{}_{}{}^{2}},$$
(4.1)
where $`T_{\mathrm{F1}}`$ is the string tension and $`l_s`$ is the string length. In the decoupling limit of massive excitation of strings $`l_s0`$, the quantity of $`m`$ must be fixed.
The Higgs branch of this theory emanates from the point where the hypermultiplet is massless, namely, $`d=0`$ in the brane configuration. When $`d=0`$, the D5-brane can separate the D3-brane into two parts. As the result, two NS5-branes can shift each other’s positions in the $`(x^7,x^8,x^9)`$-space. We denote this difference as $`\stackrel{~}{d}`$.
The gauge symmetry in this branch is completely broken since the D3-brane is divided into two parts by the D5-brane and all gauge degrees of freedom are decoupled. In the broken phase of the Abelian gauge theory, there exists a vortex state, which is an electric-magnetic dual object of the fundamental matter. Since the electric-magnetic duality of the brane effective theory is achieved by the Type IIB S-duality, the stringy counterpart of the vortex state is a D-string stretched between two separated parts of D3-brane. The mass of this state is
$$\stackrel{~}{m}=T_{\mathrm{D1}}\stackrel{~}{d}=\frac{\stackrel{~}{d}}{2\pi \stackrel{~}{l_s}^2},$$
(4.2)
where $`\stackrel{~}{l_s}=g_{s}^{}{}_{}{}^{1/2}l_s`$ is a string length of the dual theory. Filed theoretically, the mass of the vortex state is given by the Fayet-Iliopoulos (FI) parameter $`\zeta `$. If we define as
$$\zeta =\frac{\stackrel{~}{m}}{4\pi }=\frac{\stackrel{~}{d}}{8\pi ^2\stackrel{~}{l_s}^2},$$
(4.3)
the FI parameter is proportional to the difference of two NS5-brane positions in the $`x^9`$-direction.
Next, we consider a similar situation in the $`N=2`$ brane configuration with $`(p,q)`$5-brane. We add only one set of massless matter superfields which plays a role of Higgs field and consider the Abelian case, that is, the Maxwell Chern-Simons Higgs system. Moreover, we set the angles $`\psi `$ to $`\pi /2`$ in the $`N=2`$ brane configuration. So the brane configuration is as follows
$`\mathrm{NS5}`$ $`:`$ $`(012345)`$
$`\mathrm{D3}`$ $`:`$ $`(012|6|)`$
$`\mathrm{D5}`$ $`:`$ $`(012789)`$
$`(p,q)5`$ $`:`$ $`\left(01278\left[{\displaystyle \genfrac{}{}{0pt}{}{5}{9}}\right]_\theta \right).`$
The two adjoint scalars $`X_1`$ and $`X_2`$ are decoupled from the theory and so we can easily pay attention to the true nature of the Higgs phase of the Abelian theory.
The squark Higgs fields $`q,\stackrel{~}{q}`$ in the matter superfields couple to a non-dynamical D-field in the off-shell vector multiplet. If we integrate out the auxiliary D-field we have a bosonic scalar potential
$$V=\frac{g^2}{2}\left(|q|^2|\stackrel{~}{q}|^2\zeta +\frac{\kappa }{4\pi }X_3\right)^2X_{3}^{}{}_{}{}^{2}\left(|q|^2+|\stackrel{~}{q}|^2\right),$$
(4.4)
where $`\zeta `$ is the FI parameter.
The potential of the aforementioned Maxwell Higgs theory is obtained by setting $`\kappa =X_3=0`$. In this case there is only the asymmetric phase $`|q|^2|\stackrel{~}{q}|^2=\zeta `$. However, the potential (4.4) admits two distinct vacua, an asymmetric phase and a symmetric phase. We examine the correspondence between each branch and the brane configuration from now on.
asymmetric phase
The asymmetric phase is described by $`|q|^2|\stackrel{~}{q}|^2=\zeta `$ and $`X_3=0`$. This phase is the same as the asymmetric phase in the previous case. So the corresponding configuration is also very similar to the Maxwell Higgs one. Namely, the D5-brane divide the D3-brane into two segments. One of the segments of the D3-brane stretch between the left NS5-brane and the D5-brane and the other is between the D5-brane and the right $`(p,q)`$5-brane. In this phase the relative position of the 5-branes along the $`x^9`$-direction can differ because of this split of the D3-brane and the gauge symmetry on the D3-brane is broken simultaneously. We denote the difference of 5-branes in the $`x^9`$-direction as $`\stackrel{~}{d}`$. (See Fig.1.)
There also exists the topological vortex state in this broken phase of the Maxwell Chern-Simons Higgs system. However, in the $`(p,q)`$5-brane configuration we can not simply identify this vortex state with the D-string stretch between two fragments of the D3-brane. From the SUGRA solution of the $`(p,q)`$5-brane, there is a non-trivial vev of the axion field like as
$$\chi (x^6,x^7,x^8,x^9)=\frac{1}{g_s}\frac{\mathrm{sin}\theta \mathrm{cos}\theta (H^{1/3}H^{2/3})}{H^{1/3}\mathrm{sin}^2\theta +H^{2/3}\mathrm{cos}^2\theta },$$
(4.5)
where $`H=1+\frac{l_{s}^{}{}_{}{}^{2}}{r^2}`$ is a harmonic function on the $`(x^6,x^7,x^8,x^9)`$-space and $`r`$ is a distance from the $`(p,q)`$5-brane. We now assume that the distance $`L`$ between two 5-branes is satisfied $`Ll_s`$<sup>2</sup><sup>2</sup>2This limit is connected with the pure Chern-Simons limit ($`\kappa `$ is fixed and $`g\mathrm{}`$).. Since the D5-brane sits between two 5-branes, the distance from $`(p,q)`$5-brane $`r`$ is also satisfied $`rl_s`$. In this limit, that is, the axion field take a constant value near the D5-brane
$$\chi =\frac{1}{g_s}\mathrm{tan}\theta =\kappa .$$
(4.6)
In this background, D-string must appear as a dyon with electric charge proportional to $`\chi `$ by the Witten effect. In fact, since the tension of $`(a,b)`$-string in this background is given by
$$T_{(a,b)}=\frac{1}{2\pi l_{s}^{}{}_{}{}^{2}}\sqrt{\left(\frac{b}{g_s}\right)^2+\left(a+b\chi \right)^2},$$
(4.7)
the D-string, i.e. $`(0,1)`$-string, is not lightest state. A lighter state is a $`(\kappa ,1)`$-string with the tension
$$T_{(\kappa ,1)}=\frac{1}{2\pi l_{s}^{}{}_{}{}^{2}g_s},$$
(4.8)
which is the same tension as the D-string in the Maxwell theory configuration. Therefore the vortex state near the $`(p,q)`$5-brane configuration is not only magnetically charged but also electrically charged.
Field theoretically, the Chern-Simons interaction gives the relation between electric charge $`Q`$ and magnetic flux $`\mathrm{\Phi }`$
$$Q=\kappa \mathrm{\Phi }$$
(4.9)
from the Gauss law constraint. So the vortex solution carry both magnetic charge and electric charge in the ratio of 1 to $`\kappa `$. This consequence exactly agree with the $`(\kappa ,1)`$-string state around the $`(p,q)`$5-brane background.
symmetric phase
For the symmetric phase the vevs of the scaler fields are
$`q=\stackrel{~}{q}=0,`$ (4.10)
$`X_3=4\pi \zeta /\kappa .`$ (4.11)
This phase does not exist in the Maxwell Higgs system. And also there is no corresponding configuration in the $`N=4`$ Hanany-Witten setup. However, in our case there exists another possible vacuum configuration for $`\zeta 0`$ due to the rotation of the 5-brane. If we want to set $`\zeta 0`$ we must shift the positions of the 5-branes each other along the $`x^9`$-direction. In general, the D3-brane suspended between the 5-branes prevents this motion. One of settlements is the division of the D3-brane into two parts as we mentioned before. Another solution is that the D3-brane slide up along the $`x^5`$-direction and is attached again on the intersecting point of the NS5-brane and the rotated $`(p,q)`$5-brane in the $`(x^5,x^9)`$-plane. (See Fig. 2.)
In this symmetric phase configuration, the position of the D3-brane in the $`x^5`$ coordinate is $`x^5=\stackrel{~}{d}/\mathrm{tan}\theta `$ where $`\stackrel{~}{d}`$ is a gap of the 5-brane positions. The gap $`\stackrel{~}{d}`$ corresponds to the FI parameter $`\zeta `$ as in eq. (4.3) and the $`x^5`$ position of the D3-brane corresponds to the vev of the adjoint scalar $`X_3`$. So we find that the adjoint scalar vev is given by
$`X_3`$ $`=`$ $`{\displaystyle \frac{\stackrel{~}{d}}{2\pi l_{s}^{}{}_{}{}^{2}\mathrm{tan}\theta }}`$ (4.12)
$`=`$ $`{\displaystyle \frac{4\pi \zeta }{p/q}}.`$ (4.13)
This result is exactly agree with (4.11).
Moreover, the distance between the D3 and D5-brane is $`\stackrel{~}{d}/\mathrm{tan}\theta `$. This means that the mass of the matter superfields which is coming from the fundamental string between D3 and D5-brane is $`4\pi \zeta /\kappa `$. Correspondingly, when the $`X_3`$ field has the vev (4.11) the scalar potential (4.4) gives the mass $`4\pi \zeta /\kappa `$ to the $`q`$ and $`\stackrel{~}{q}`$. This is also consistent with the brane picture.
Finally, we would like to comment on vortex states in this phase. In this symmetric phase, it is known that there exists a non-topological vortex . However, since charges of the non-topological vortex contain a continuous parameter, it seems to be difficult to find the corresponding state in string theory.
## Acknowledgments
I would like to thank T. Kitao and N. Ohta for arguments during the early stage of this work. I also grateful to B.-H. Lee for useful discussions and comments at Niseko Winter School.
|
no-problem/9904/astro-ph9904399.html
|
ar5iv
|
text
|
# In search for natural wormholes
## I Introduction
Wormholes are nontrivial topological configurations of spacetime that can be represented by solutions of Einstein field equations with stress-energy tensor fields that somewhere violate the so-called average null energy condition (see Ref. for a detailed discussion). Although microscopic violations of the energy conditions are well known (e.g. the Casimir effect), it is far from clear whether stable, macroscopic wormholes can naturally exist in the Universe. One of the ways in which one may obtain violations to the energy conditions is via a scalar fields coupled to gravity (see for instance and references therein).
Wormhole formation at a late cosmic time requires Lorentzian topology change in space, something that appears to be more than problematic to most physicists because it implies causality violations . However, if wormholes are created altogether with spacetime and not formed by astrophysical processes, one could expect a cosmological population of these objects without the uncomfortable predictions of topology change theorems.
In a couple of recent papers we have discussed the observable effects that could arise from an intergalactic population of natural wormholes . Since wormhole’s mouths could have a total negative mass, they should exert a repulsive gravitational force that can provide very peculiar microlensing events when acting upon the light of compact, background sources . Extragalactic wormholes with absolute masses of $``$ 1 M would produce very compact Einstein rings, in such a way that just small, ultraluminous sources like the $`\gamma `$-ray emitting core of quasars (typical size $`10^{14}10^{15}`$ cm) might result gravitationally magnified. We have shown in Ref. that the lightcurve signature of wormhole microlensing events of this sort very much resembles some kinds of gamma ray bursts (GRBs).
When a negative mass lens crosses the line of sight to a distant quasar, dragging the caustic pattern along with it, two bursting $`\gamma `$-ray events will appear in the observer’s frame: the first one is the specular image of a fast-rise-exponential-decay (FRED) burst, whereas the second, after a period of stillness that can last several years, is a pure FRED event. In our previous study , we have used the available database of GRB observations gathered by the BATSE instrument, part of the Compton satellite, to set an upper limit to the amount of negative mass (under the form of compact objects of astrophysical size) in the Universe. Such limit results as low as $`|\rho |2\times 10^{33}`$ g cm<sup>-3</sup> with the most optimistic assumptions.
In the present paper we give a step further and embark on the first detailed search for individual wormhole signatures in astronomical databases. GRBs produced by natural wormholes can be differentiated from those originated in fireballs because of two very definite properties: 1) they repeat, and 2) one of the repeating bursts has an anti-FRED time profile, something that cannot be the result of an explosive event (the companion burst must display a FRED-like lightcurve).
We have quantitatively analysed a subsample of the GRBs included in the BATSE 3B cataloge with the aim of identify events that could be unequivocally attributed to wormhole lensing. In what follows we present the results we have obtained.
## II Data analysis
We have analysed a sample of 631 bursts from BATSE 3B catalog whose global symmetry properties were already discussed by Link & Epstein , and Romero et al. . This sample contains both faint and bright bursts, spanning 200-fold range in peak flux. PREB + DISC data tapes at 64 ms time resolution, with four energy channels, were used in the analysis.
Since the variety of burst profiles is huge and simple visual inspection can be misleading, we have used the skewness function $`𝒜`$ introduced by Link & Epstein in order to separate those GRBs with anti-FRED profiles. The skewness is basically defined as the third moment of the individual burst time profile and can be directly computed from the observational data as in Ref. . Negative values of $`𝒜`$ correspond to events with slower rising than decaying timescales, thus showing a peculiar asymmetric burst (PAB).
In a first step, we estimated $`𝒜`$ for all GRBs in the sample at different background cutoff levels. Just 91 out of 631 bursts present $`𝒜<`$0 at any background. As discussed in Ref. , most of these events can be explained within the standard fireball model of GRBs . Just anti-FRED–like single peaked events remain inconsistent with the explosive hypothesis. There are 26 of these GRBs in our sample (4.1 %).
Since wormhole lensing not just provides bursts with $`𝒜<`$0 but also a repetition with specular signature, we have searched for time-space clustering in the sample. We have found that 15 out of 26 candidates (about 60 %) present companions within error boxes at less than $`4^o`$ (the average positional uncertainty in BATSE catalog). We have estimated the statistical significance of this level of positional coincidences through numerical simulations of random sets of 26 events against a background distribution of 605 GRBs. After 1500 simulations we established that the chance associations expected in the subsample are $`13.3\pm 2.5`$, i.e. there is no need to claim for repetition to explain the observed coincidences at error boxes of $`4^o`$. However, if positional coincidences separated by less than $`1^o`$ are considered, we find that 3 out of 26 events present companions. According to a new set of simulations, these results can be attributed to chance only at a $`2\sigma `$ level. Despite the sample is too scarce to draw any conclusion, it is worth mentioning that when a similar study is carried out with the 91 bursts with $`𝒜<`$0 it is found that there are just 4 positional coincidences at less than $`1^o`$, at $`1\sigma `$ confidence level. This could imply that the apparent excess is exclusively associatted to single peaked events, as expected from the microlensing model.
In order to detect whether there is some suitable wormhole candidate behind the above mentioned statistical analyses, we turned to the individual single peaked events with $`𝒜<`$0 in a finer search.
## III Results
In Table 1 we list by trigger number all single peaked bursts with $`𝒜<`$0 in our sample. In column 2 we indicate the trigger number of any companion burst in the entire BATSE database within a circle of $`4^o`$ in radius. Columns 3 and 4 display the temporal and angular distances of pairs of events. A negative value of $`\mathrm{\Delta }T`$ means that the anti-FRED event followed to its companion; such bursts can be eliminated as wormhole candidates, at least over the timescales under consideration here. The final column in the table lists the sign of $`𝒜`$, when defined, for companion bursts that belong to our subsample. Bursts with no entries in this column where not analyzed in the present study, circunscribed to the previously defined set of 631 GRBs.
We shall consider as candidates for wormhole microlensing just events with companions that present $`𝒜>`$0 at all levels of background (notice that this is a very restrictive criterion, and eliminates the event mentioned in Ref. ). This left us with only 4 candidates: #254, #444, #1924, and #2201. A further step can be made now by detecting active galactic nuclei (AGNs) within the error boxes of the bursts. These AGNs would constitute the potential background sources of gamma rays.
In Table 2 we list pairs of GRBs along with the AGNs (namely compact QSOs) within the BATSE field. We also indicate the morphological type of the bursts with $`𝒜>`$0. As can be appreciated from this table, three pairs of events present quasars in their fields: #254, #444, and #1924. These are the stronger candidates for wormhole microlensing events in our sample of 631 bursts. None of them, however, can be considered as a certain identification because the profiles of the second bursts in each pair are not exact FREDs despite presenting $`𝒜>`$0 at all levels. These bursts have profiles with some substructure which is not present in the first event of the pair. Although such a fine substructure could be an effect of the different light propagation paths (the light can be exposed during its travel to lensing effects by ordinary matter that might result in a distortion of the original profile ) or even an artifact due to the different orientation of the spacecraft at the detection times, we think that the evidence is not strong enough to claim for an indisputable identification.
In the case of the pair {#2201, #2679}, both bursts are single peaked and present the correct symmetry in their profiles. By other hand, there are no cataloged AGNs in the corresponding sky field. This would not be an insurmountable problem for wormhole microlensing because even very weak and normally undetected QSOs can be enhanced by caustic crossing in such a way as to appear as a bright source during a few seconds . However, in this particular case, the flux ratio of both bursts (which should be similar as it comes from the same source) is too far from unity as to make a case for the lensing argument.
The best candidate in the whole sample is the pair {#254, #2477}. They present correct symmetry properties and unity flux ratio. The event #2477 has a substructure that make of it not a perfect FRED, but this, as it was mentioned, could stem from propagation effects. Two AGNs, with redshifts of 0.15 and 1.52, are present within the positional uncertainties.<sup>*</sup><sup>*</sup>*It should also be pointed out the FRED–anti-FRED pair, #688 and #2788, which is located at the same position in the sky (within 4<sup>o</sup>) together with 156 QSOs. These two bursts might be produced by two different microlensing phenomena with timescales that span out the BATSE operation time.
The greatest difficulty at present time is the huge positional uncertainty, something that will be significantly improved in ten years time. The results of our search, although not conclusive, are sufficiently suggestive as to encourage new studies over larger samples and with the more accurate detectors of the forthcoming generation of gamma ray satellites.
## IV Final comment
Our search for natural wormholes through microlensing was sensitive to timescales up to 3.5 years. Repetition of anti-FRED events over longer scales can not be ruled out. Tegmark et al. have made a repetition study on the entire BATSE sample concluding that a repetition level of $`5`$ % with timescales of a few years is compatible with the current data. Our results show that, if repetition is associated to wormhole microlensing alone, it could reach, at most, a level $`4\%`$ over timescales larger than 3.5 years. At shorter scales, wormhole-induced repetitions are constrained at a level $`<0.2`$ % (assuming {#254, #2477} as the sole possible candidate).
Since microlensing timescale increases with larger masses of the lenses, the absence of clear detections in our search might be saying that wormholes, if there exist at all, have a mass distribution peaking far beyond the few tenths of solar masses required to produce typical microlensing events with timescales of a few years. If we recall that a negative mass of the size of Jupiter is necessary to keep open a wormhole throat of about 2 m in diameter , this result could be kindly greeted by optimistic interstellar-travel afficionados looking for larger spacetime tunnels.
## Acknowledgements
We acknowledge Bennett Link for earlier discussions on burst temporal profiles. The present research has made use of the NASA/IPAC Extragalactic Database, which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. This work has been partially supported by Argentine agencies CONICET (G.E.R., D.F.T.) (also, through research grant PIP 0430/98 and ANPCT, G.E.R.). L.A.A. thanks FOMEC Program for additional support.
|
no-problem/9904/gr-qc9904060.html
|
ar5iv
|
text
|
# (2+1)-Gravity and the embedding its dynamiycal symmetry and para-supersymmetry into 𝑆𝑂(4,𝑐) group
## 1 INTRODUCTION
Due to vanishing of Weyl tensor in $`(2+1)`$ dimensions, the Riemann curvature tensor can be identified with zero in matter-void regions of spacetime. Consequently, spacetime is flat in local vacua. Addition of cosmological constant term in the absence of matter leads to solutions with constant curvature, where the sign of cosmological constant determines the sign of scalar curvature . In the last few years many interesting problems have been investigated in $`(2+1)`$ dimensional gravity, such as the spacetime metric of multi-point particle with or without spin . The Einstein-Maxwell equations in $`(2+1)`$ dimensions have already been treated . Indeed, in many physical situations in $`(3+1)`$ dimensions there is no structure along one of spatial dimension like an infinite cosmic string, where the theory becomes $`(2+1)`$ dimensional. There are also some interesting works concerning the quantum mechanics of a point mass particle in the presence of a very heavy particle in $`(2+1)`$ dimensions, both relativistically and nonrelativistically .
We introduce the Einstein-Maxwell action in the presence of matter together with cosmological term in $`(2+1)`$ spacetime dimension. Then, we choose the solutions that correspond to a spacetime with a spatial part of locally constant curvature surface with deficit of angle at location of a very heavy point mass, and magnetic field of a magnetic monopole. For positive cosmological term we have Minkowskian and Euclidean spacetimes with spatial part of locally constant curvature. For vanishing cosmological term we have a spacetime with locally flat spatial part. Over these spacetimes, in absence of angular deficit, the quantum Hamiltonian associated with a free test particle leads to solvable systems with degeneracy group $`GL(2,c)`$, where their eigen-states can be obtained algebraically, too. We will show that quantum solvable models will be obtained by restricting the Casimir of $`SO(4,c)`$ group to the Casimir of $`SL(2,c)`$ group. These models possess simultaneously the degeneracy group $`SL(2,c)`$ and shape invariance symmetry, where both symmetries are kind of realization of para-supersymmetry of arbitrary order.
## 2 SOME SPECIAL EXACT SOLUTIONS OF EINSTEIN EQUATIONS
In $`(2+1)`$ dimensions the Einstein-Hilbert action of gravity coupled to matter and electromagnetic field, together with the cosmological term can be written as
$$S=d^3x\sqrt{g}\{\frac{1}{4\pi G}(R+2\mathrm{\Lambda })+\frac{1}{4}F_{\mu \nu }F^{\mu \nu }_M\},$$
(2.1)
where $`_{}`$ is the matter Lagrangian. We have rescaled G by a factor of $`4`$. Variation of the action (2.1) leads to Einstein-Maxwell equations in $`(2+1)`$ dimensions
$$R_{\mu \nu }\frac{1}{2}g_{\mu \nu }R=2\pi GT_{\mu \nu }^{eff}$$
(2.2a)
$$_\mu (\sqrt{g}F^{\mu \nu })=0$$
(2.2b)
with $`T_{\mu \nu }^{eff}=T_{\mu \nu }^{(M)}+T_{\mu \nu }^{(EM)}+\frac{\mathrm{\Lambda }}{2\pi G}g_{\mu \nu }`$. The energy-momentum tensor of matter $`T_{\mu \nu }^{(M)}`$ and electromagnetic $`T_{\mu \nu }^{(EM)}`$ are respectively
$$T_{\mu \nu }^{(M)}=\frac{2}{\sqrt{g}}\frac{\delta (\sqrt{g}_M)}{\delta g^{\mu \nu }},$$
(2.3a)
$$T_{\mu \nu }^{(EM)}=g^{\alpha \beta }F_{\mu \alpha }F_{\nu \beta }+\frac{1}{4}g_{\mu \nu }F_{\lambda \sigma }F^{\lambda \sigma }.$$
(2.3b)
We assume that our spacetime is described by the axial symmetric static metric, that is $`(_tg_{\mu \nu }=0,g_{_{0i}}=0)`$ :
$$ds_{(3)}^2=N^2(𝐫)dt^2\rho (𝐫)(dr^2+r^2d\varphi ^2),$$
(2.4)
where $`0r<\mathrm{}`$ and $`0\varphi <2\pi `$ are the usual polar coordinates. The non-zero components of electromagnetic field tensor are
$$F_{0r}=E(r),F_{r\varphi }=B(r)andF_{0\varphi }=0$$
with E and B as electric and magnetic fields, respectively. After, changing the coordinates as $`x^1=r\mathrm{cos}\varphi `$ and $`x^2=r\mathrm{sin}\varphi `$, the electromagnetic energy-momentum tensor (2.3b) takes the following form
$$T_{\alpha \beta }^{(EM)}=\left(\begin{array}{ccc}\frac{1}{2\rho }(\frac{N^2}{\rho r^2}B^2+E^2)& \frac{x^2}{\rho r^2}EB& \frac{x^1}{\rho r^2}EB\\ \frac{x^2}{\rho r^2}EB& \frac{1}{2\rho r^2}B^2+\frac{(x^2)^2(x^1)^2}{2r^2N^2}E^2& \frac{x^1x^2}{N^2r^2}E^2\\ \frac{x^1}{\rho r^2}EB& \frac{x^1x^2}{N^2r^2}E^2& \frac{1}{2\rho r^2}B^2\frac{(x^2)^2(x^1)^2}{2r^2N^2}E^2\end{array}\right).$$
(2.5)
In this article we consider the matter as a point mass located at the origin $`(x^1=x^2=0)`$ with the only nonvanishing component of $`T_{\mu \nu }^{(M)}`$ as $`T_0^{(M)0}=\frac{1}{\sqrt{{}_{}{}^{(2)}g}}M\delta (x^1)\delta (x^2)`$, where $`{}_{}{}^{(2)}g`$ is the determinant of the spatial part of the metric. We remind that the Ricci scalar for metric (2.4) is
$$R=\frac{2}{\rho N}^2N+\frac{1}{\rho ^2}^2\rho \frac{1}{\rho ^3}(\stackrel{}{}\rho )^2,$$
(2.6)
where $`^2`$ and $`\stackrel{}{}`$ are the usual Euclidean 2-dimensional Laplacian and gradient operators, respectively. For a spatially conformal metric with $`\rho (𝐫)=\rho ___0e^{2GM\mathrm{ln}r+\chi }`$, the singular term on the right hand side of Eqs. (2.2a) disappears. Thus, the equations can be written in the following singularity free form
$`EB=0`$
$`^2\chi +2\mathrm{\Lambda }\rho ___0r^{2GM}e^\chi +{\displaystyle \frac{2\pi G}{\rho ___0}}r^{2GM2}e^\chi B^2+{\displaystyle \frac{2\pi G}{N^2}}E^2=0`$
$`_1_2N{\displaystyle \frac{1}{2}}(_1N_2\chi +_1\chi _2N)+{\displaystyle \frac{GM}{r^2}}(x^2_1N+x^1_2N)2\pi G{\displaystyle \frac{x^1x^2}{Nr^2}}E^2=0`$
$`_2^2N+\mathrm{\Lambda }\rho ___0Nr^{2GM}e^\chi +{\displaystyle \frac{1}{2}}(_1\chi _1N_2\chi _2N)GMr^2(x^1_1Nx^2_2N)`$
$`{\displaystyle \frac{\pi G}{\rho ___0}}Nr^{2GM2}e^\chi B^2\pi Gr^2{\displaystyle \frac{(x^2)^2(x^1)^2}{N}}E^2=0`$
$`_1^2N+\mathrm{\Lambda }\rho ___0Nr^{2GM}e^\chi +{\displaystyle \frac{1}{2}}(_2\chi _2N_1\chi _1N)GMr^2(x^2_2Nx^1_1N)`$
$`{\displaystyle \frac{\pi G}{\rho ___0}}Nr^{2GM2}e^\chi B^2\pi Gr^2{\displaystyle \frac{(x^1)^2(x^2)^2}{N}}E^2=0.`$ (2.7)
Letting $`E=0`$ in Eqs. (2.7) and using the following ansatz for the magnetic field $`B`$
$$B^2=h\frac{\mathrm{\Lambda }\rho ___0^2}{\pi G}r^{24GM}e^{2\chi }$$
where $`h`$ is a constant parameter. Then, with a change of variable $`u=r^{1GM}`$, the Eqs. (2.7) can be written as
$$\frac{1}{u}\frac{d}{du}(u\frac{d}{du})\chi +\frac{2(1+h)\mathrm{\Lambda }\rho ___0}{(1GM)^2}e^\chi =0$$
(2.8a)
$$\frac{1}{u}\frac{dN}{du}=\frac{c}{1GM}e^\chi $$
(2.8b)
$$\frac{1}{u}\frac{d}{du}(u\frac{d}{du})N+\frac{2(1h)\mathrm{\Lambda }\rho ___0}{(1GM)^2}Ne^\chi =0,$$
(2.8c)
where $`c`$ is a constant of integration. Choosing the following ansatz as a solution for the Eq. (2.8a)
$$e^\chi =\frac{1}{(1+\frac{(1+h)\mathrm{\Lambda }\rho ___0}{4(1GM)^2}u^2)^2},$$
(2.9)
we make use of (2.9) to solve the Eq. (2.8b) as
$$N=c\frac{\frac{2(1GM)}{(1+h)\mathrm{\Lambda }\rho ___0}}{1+\frac{(1+h)\mathrm{\Lambda }\rho ___0}{4(1GM)^2}u^2}+d,$$
where $`d`$ is another constant of integration. Next, having inserted the results just obtained for $`\chi `$ and $`N`$ in Eq. (2.8c), we get the following equations for constants $`d`$ and $`c`$
$`{\displaystyle \frac{(1h^2)\mathrm{\Lambda }\rho ___0}{1GM}}d(1+h)c=0`$
$`{\displaystyle \frac{(1h)\mathrm{\Lambda }\rho ___0}{1GM}}d{\displaystyle \frac{13h}{1+h}}c=0.`$ (2.10)
The nontrivial solutions of Eqs. (2.10) exist only for $`h=0`$ and $`1`$. If $`h=0`$, the magnetic field vanishes and Maxwell Eqs. (2.2b) are satisfied. Then, (2.4) becomes the metric of spacetime in the presence of a point mass located at the origin, together with the cosmological term which has already been studied in Refs. .
But, for $`h=1`$ we have $`N=1`$ and
$$B^2=\frac{\mathrm{\Lambda }\rho ___0^2}{\pi G}\frac{r^{24GM}}{(1+\frac{\mathrm{\Lambda }\rho ___0}{2(1GM)^2}r^{22GM})^4},$$
(2.11)
together with the following metric of spacetime
$$ds_{(3)}^2=dt^2\frac{\rho ___0r^{2GM}}{(1+\frac{\mathrm{\Lambda }\rho ___0}{2(1GM)^2}r^{22GM})^2}(dr^2+r^2d\varphi ^2).$$
(2.12)
It turns out that the magnetic field (2.11) and the metric (2.12) satisfy Einstein-Maxwell Eqs. (2.2). Using the formula (2.6) we calculate the scalar curvature of metric (2.12)
$$R=4\mathrm{\Lambda }4\pi GMr^{2GM}(1+\frac{\mathrm{\Lambda }\rho ___0}{2(1GM)^2}r^{22GM})^2\delta ^2(𝐫).$$
Hence, except for a delta singularity at the origin, the spacetime has a constant curvature. We will discuss the solutions of Einstein equations in $`(2+1)`$-dimensional spacetime corresponding to a point mass located at origin in the presence of cosmological constant and magnetic field in the next section. In the zero limit of cosmological constant, the magnetic field (2.11) vanishes and the metric (2.12) changes to the metric of a point mass located at origin with angular deficit . Note that, we have considered here only the most simple solution of Liouville equation (2.8a), while one can take some less trivial solutions using the Backlund transformations.
## 3 EMBEDDING OF DEGENERACY AND SHAPE INVARIANCE OF QUANTUM STATES IN $`SO(4,c)`$ GROUP
We investigate the solutions (2.12) of Einstein equation in $`(2+1)`$ dimensions with nonnegative cosmological constant $`\mathrm{\Lambda }`$, point mass $`M`$ and magnetic field (2.11) as sources of energy-momentum tensor. Now, we introduce the parameter $`\gamma `$ which only takes the values $`0`$, $`1`$ and $`i=\sqrt{1}`$ and redefine the cosmological constant $`\mathrm{\Lambda }`$ as $`\mathrm{\Lambda }=\gamma ^2\lambda `$. In this article, $`\lambda `$ and $`\rho ___0`$ are arbitrary positive (negative) nonvanishing constants for $`\gamma =0`$ and $`1`$ ($`i`$). With change of variables
$`{\displaystyle \frac{r^{1GM}}{1GM}}=\sqrt{{\displaystyle \frac{2}{\lambda \rho ___0}}}{\displaystyle \frac{\mathrm{tan}\frac{\gamma \theta }{2}}{\gamma }}`$
$`\mathrm{\Phi }:=(1GM)\varphi ,`$ (3.1)
the metric (2.12) takes the form
$$ds_{(3)}^2=dt^2\frac{1}{2\lambda }(d\theta ^2+\frac{\mathrm{sin}^2\gamma \theta }{\gamma ^2}d\mathrm{\Phi }^2).$$
(3.2)
It is obvious that for $`\gamma =i`$ the spacetime is described by an Euclidean metric, while for $`\gamma =0`$ and $`1`$ the spacetime is described by Minkowskian metrics. Using the general coordinate transformation, the magnetic field can be written as (3.3) in terms of the new coordinates $`\theta `$ and $`\mathrm{\Phi }`$
$$=q\gamma \mathrm{sin}\gamma \theta ,$$
(3.3)
where
$`q=\{\begin{array}{cc}\frac{1}{2\sqrt{\pi G\lambda }}\hfill & \text{if }\gamma =0\text{ , }1\hfill \\ \frac{1}{2\sqrt{\pi G|\lambda |}}\hfill & \text{if }\gamma =i.\hfill \end{array}`$ (3.6)
The magnetic potential one-form $`A`$ corresponding to magnetic field (3.3) is $`A=q(1\mathrm{cos}\gamma \theta )(\frac{i\gamma }{\mathrm{sin}\gamma \theta }d\theta +d\mathrm{\Phi })`$. Obviously, in terms of the new coordinates $`\theta `$ and $`\mathrm{\Phi }`$, the choice of $`\gamma =0`$ leads to a Minkowskian metric with flat spatial part and angular deficit in its metric. Here, $`\theta `$ is its radial coordinate and the magnetic field $``$ is zero. The choice of $`\gamma =1`$ leads to Minkowskian metric of $`(2+1)`$-dimensional spacetime with local constant curvature together with angular deficit and magnetic field $`\frac{1}{2\sqrt{\pi G\lambda }}\mathrm{sin}\theta `$. Finally, for $`\gamma =i`$, we get Euclidean metric of $`(2+1)`$-dimensional spacetime with local constant curvature and deficit of angle and magnetic field $`\frac{1}{2\sqrt{\pi G|\lambda |}}\mathrm{sinh}\theta `$.
In the presence of the magnetic field (3.3), the quantum states of a point mass with mass $``$, which is negligible compared to the mass $`M`$ of a point source located at the origin, i.e. $`M`$, can be described in terms of the energy spectrum of the Hamiltonian $`H=\frac{1}{2}D_i^AD^{Ai}`$ with $`D_i^A=_iiA_i`$ as covariant derivative. Using the metric (3.2) and the given connection one-form, the Hamiltonian can be written as
$$H=\frac{\lambda }{}\{\frac{^2}{\theta ^2}+\gamma (\frac{12q}{\mathrm{tan}\gamma \theta }+\frac{2q}{\mathrm{sin}\gamma \theta })\frac{}{\theta }+\frac{\gamma ^2}{\mathrm{sin}^2\gamma \theta }\frac{^2}{\mathrm{\Phi }^2}+2iq\gamma ^2(\frac{1}{\mathrm{tan}^2\gamma \theta \mathrm{cos}\gamma \theta }\frac{1}{\mathrm{sin}^2\gamma \theta })\frac{}{\mathrm{\Phi }}+q\gamma ^2\},$$
(3.7)
with the spectrum given by
$$E(n)=\frac{\lambda \gamma ^2}{}[n(n+1)q(2n+1)],$$
(3.8)
where in absence of angular deficit $`n`$ is a non-negative integer . Writting the eigenstates of the Hamitonian (3.4) in the following form
$$\mathrm{\Psi }_{n,m}(\theta ,\mathrm{\Phi })=(e^{i\mathrm{\Phi }}\frac{\mathrm{tan}\frac{\gamma \theta }{2}}{\mathrm{sin}\gamma \theta })^qe^{i\frac{m}{1GM}\mathrm{\Phi }}(1\mathrm{cos}\gamma \theta )^{\frac{1}{2}|\frac{m}{1GM}+q|}(1+\mathrm{cos}\gamma \theta )^{\frac{1}{2}|\frac{m}{1GM}q|}F(\mathrm{cos}\gamma \theta )$$
and choosing the change of variable $`z=\frac{1\mathrm{cos}\gamma \theta }{2}`$, the eigenvalue equation turns in to hypergeometric differential equation
$`z(1z){\displaystyle \frac{d^2}{dz^2}}F+[1+|{\displaystyle \frac{m}{1GM}}+q|(|{\displaystyle \frac{m}{1GM}}q|+|{\displaystyle \frac{m}{1GM}}+q|+2)z]{\displaystyle \frac{d}{dz}}F`$
$`+[nq{\displaystyle \frac{1}{2}}(|{\displaystyle \frac{m}{1GM}}q|+|{\displaystyle \frac{m}{1GM}}+q|)][nq+{\displaystyle \frac{1}{2}}(|{\displaystyle \frac{m}{1GM}}q|+|{\displaystyle \frac{m}{1GM}}+q|)+1]F=0.`$
The wave equations for relativistic and non-relativistic quantum states are
$$H\mathrm{\Psi }_{nonrel}(\theta ,\mathrm{\Phi })=E(n)\mathrm{\Psi }_{nonrel}(\theta ,\mathrm{\Phi })$$
(3.9a)
$$H\mathrm{\Psi }_{rel}(\theta ,\mathrm{\Phi })=\frac{E_{rel}^2^2}{2}\mathrm{\Psi }_{rel}(\theta ,\mathrm{\Phi }).$$
(3.6b)
For $`\gamma =0`$ we define $`n\gamma =k`$, which for $`n`$ very large $`k`$ is an arbitrarily finite constant, and we get $`E(n)\frac{k^2}{2}`$. Then, with assumption $`E_{rel}>`$, solutions of Eqs. (3.6) are
$$\mathrm{\Psi }_{_{nonrel}}(\theta ,\mathrm{\Phi })=e^{i\frac{m}{1GM}\mathrm{\Phi }}J_{\frac{|m|}{1GM}}(k\theta )$$
(3.7a)
$$\mathrm{\Psi }_{_{rel}}(\theta ,\mathrm{\Phi })=e^{i\frac{m}{1GM}\mathrm{\Phi }}J_{\frac{|m|}{1GM}}(\frac{1}{2\lambda }\sqrt{E_{rel}^2^2}\theta ).$$
(3.7b)
It is straitforward to see that by defining $`\gamma =i\delta `$ and taking the zero limit of $`\delta `$ as $`n\delta =k`$ together with $`E_{rel}<`$, we get the solutions of Eqs. (3.6) in terms of modified Bessel functions
$$\mathrm{\Psi }_{_{nonrel}}(\theta ,\mathrm{\Phi })=e^{i\frac{m}{1GM}\mathrm{\Phi }}K_{\frac{|m|}{1GM}}(k\theta )$$
(3.8a)
$$\mathrm{\Psi }_{_{rel}}(\theta ,\mathrm{\Phi })=e^{i\frac{m}{1GM}\mathrm{\Phi }}K_{\frac{|m|}{1GM}}(\frac{1}{2\lambda }\sqrt{^2E_{rel}^2}\theta ).$$
(3.8b)
Therefore, in the presence of a heavy particle $`M`$ which is located at origin we get scattering and bound states of a point particle with mass $``$, similar to Eqs. (3.7) and (3.8). These results are in agreement with reference .
So far the domain of $`\mathrm{\Phi }`$ was $`[0,2\pi (1GM)]`$, but, from now on, for simplicity, we ignore the presence of this angular deficit in the rest of the article. In general, in order to obtain the eigen-spectrum algebraically, we introduce generators of $`gl(2,c)`$ Lie algebra as
$`[L_+,L_{}]=2\gamma ^2L_32q\gamma ^2I`$
$`[L_3,L_\pm ]=\pm L_\pm `$
$`[𝐋,I]=0.`$ (3.9)
Note that the algebra (3.9) becomes $`iso(2)u(1)`$ algebra for $`\gamma =0`$, $`u(2)`$ Lie algebra for $`\gamma =1`$, and $`u(1,1)`$ Lie algebra for $`\gamma =i`$. The raising $`L__+`$ and lowering $`L_{_{}}`$ operators have the following coordinate representations
$`L_+`$ $`=`$ $`e^{i\varphi }({\displaystyle \frac{}{\theta }}+{\displaystyle \frac{i\gamma }{\mathrm{tan}\gamma \theta }}{\displaystyle \frac{}{\varphi }})L_3={\displaystyle \frac{1}{i}}{\displaystyle \frac{}{\varphi }}`$
$`L_{}`$ $`=`$ $`e^{i\varphi }({\displaystyle \frac{}{\theta }}+{\displaystyle \frac{i\gamma }{\mathrm{tan}\gamma \theta }}{\displaystyle \frac{}{\varphi }}+2q{\displaystyle \frac{\gamma }{\mathrm{tan}\gamma \theta }}2q{\displaystyle \frac{\gamma }{\mathrm{sin}\gamma \theta }})I=1.`$ (3.10)
It is easy to show that the Hamiltonian (3.4) without angular deficit is the Casimir operator of $`gl(2,c)`$ Lie algebra, that is
$`H={\displaystyle \frac{2\lambda }{M}}[{\displaystyle \frac{1}{4}}L__+L_{_{}}+{\displaystyle \frac{1}{4}}L_{_{}}L__++{\displaystyle \frac{1}{2}}\gamma ^2L__3^2q\gamma ^2L__3].`$ (3.11)
For $`\gamma 0`$, its highest weight can be obtained as follows
$$\mathrm{\Psi }_{n,n}(\theta ,\varphi )=e^{in\varphi }(\frac{1}{\gamma }\mathrm{sin}\frac{\gamma \theta }{2})^n(\mathrm{cos}\frac{\gamma \theta }{2})^n,$$
where $`n`$ is a nonnegative integer of integration constant. The other states can be obtained by applying the lowering operator $`L_{}`$ over $`\mathrm{\Psi }_{n,n}(\theta ,\varphi )`$ repeatedly. That is,
$`\mathrm{\Psi }_{_{nonrel}}(\theta ,\varphi )`$ $`=`$ $`\mathrm{\Psi }_{n,m}(\theta ,\varphi )=(L_{})^{nm}\mathrm{\Psi }_{n,n}(\theta ,\varphi )`$
$`=`$ $`(2q2n)_{nm}e^{im\varphi }({\displaystyle \frac{\mathrm{sin}\frac{\gamma \theta }{2}}{\gamma }})^m(\mathrm{cos}{\displaystyle \frac{\gamma \theta }{2}})^{2nm}F(mn,2qn,2q2n,{\displaystyle \frac{1}{\mathrm{cos}^2\frac{\gamma \theta }{2}}}).`$
Hence, using the usual properties of hypergeometric function , up to a constant coefficient $`(1)^{n2q}(2q2n)_{_{nm}}`$, in the limit $`n\mathrm{}`$ and $`\gamma 0`$ such that $`n\gamma =k=`$finite, the solution of algebraic method can be written as
$`\mathrm{\Psi }__m(\theta ,\varphi )`$ $`=`$ $`\underset{n\mathrm{}}{lim}k^me^{im\varphi }\mathrm{cos}^{4q}({\displaystyle \frac{k\theta }{2n}})n^m\mathrm{tan}^m({\displaystyle \frac{k\theta }{2n}})F(2qn,n+1,m+1,\mathrm{sin}^2({\displaystyle \frac{k\theta }{2n}}))`$
$`=`$ $`k^me^{im\varphi }\mathrm{\Gamma }(m+1)J_m(k\theta ),`$
which is the same as special case $`M=0`$ in Eq. (3.7a). It is obvious that with the definition $`\gamma =i\delta `$, we will again get the solution (3.8a) without angular deficit in the zero limit of $`\delta `$.
For the left and right invariant generators of $`SL(2,c)`$ group manifold with $`sl(2,c)`$ Lie algebra
$`J__\pm ^{(L)}=e^{\pm i\varphi }(\pm {\displaystyle \frac{}{\theta }}+i{\displaystyle \frac{\gamma }{\mathrm{tan}\gamma \theta }}{\displaystyle \frac{}{\varphi }}i{\displaystyle \frac{\gamma }{\mathrm{sin}\gamma \theta }}{\displaystyle \frac{}{\psi }})`$
$`J__3^{(L)}=i{\displaystyle \frac{}{\varphi }},`$ (3.12)
$`J__\pm ^{(R)}=e^{\pm i\psi }(\pm {\displaystyle \frac{}{\theta }}i{\displaystyle \frac{\gamma }{\mathrm{sin}\gamma \theta }}{\displaystyle \frac{}{\varphi }}+i{\displaystyle \frac{\gamma }{\mathrm{tan}\gamma \theta }}{\displaystyle \frac{}{\psi }})`$
$`J__3^{(R)}=i{\displaystyle \frac{}{\psi }},`$ (3.13)
where $`\theta `$, $`\varphi `$ and $`\psi `$ are complex variables at present, the Casimir operators are equal with each other, i.e. $`H_{_{sl(2,c)}}=H^{(L)}=H^{(R)}`$, such that
$`H^{(L,R)}={\displaystyle \frac{1}{4}}J__+^{(L,R)}J_{_{}}^{(L,R)}+{\displaystyle \frac{1}{4}}J_{_{}}^{(L,R)}J__+^{(L,R)}+{\displaystyle \frac{1}{2}}\gamma ^2J__3^{(L,R)^2}.`$ (3.14)
Since $`so(4,c)=sl(2,c)sl(2,c)`$, therefore the Casimir operator of $`so(4,c)`$, can be given in terms of the $`sl(2,c)`$ Casimir operators as $`H_{_{so(4,c)}}=H^{(L)}+H^{(R)}`$, but note that, choosing the coordinates of left and right generators the same will be equivalent to restricting the space of $`so(4,c)`$ Lie algebra to one of the subspaces $`sl(2,c)`$. Now we restrict ourselves to real values of $`\theta `$, $`\varphi `$ and $`\psi `$ variables such that $`0\varphi <2\pi `$ and $`0\psi <4\pi `$ for all values of $`\gamma `$, while we choose $`0\theta <\pi `$ for $`\gamma =1`$ and $`0\theta <\mathrm{}`$ for other cases. Due to this restriction, the $`sl(2,c)`$ algebra reduces to its different real forms as follows: in the case of $`\gamma =1`$ the $`sl(2,c)`$ Lie algebra reduces to $`su(2)`$, while for $`\gamma =i`$ it reduces to $`su(1,1)`$ Lie algebra and finally for $`\gamma =0`$ it reduces to $`iso(2)`$ Lie algebra . We should remind that for $`\gamma =1`$, $`\gamma =i`$ and $`\gamma =0`$ the direct sum of left and right invariant generators becomes $`so(4)=su(2)su(2)`$, $`so(2,2)=su(1,1)su(1,1)`$ and $`iso(2)iso(2)`$ Lie algebra respectively, which are different real forms of $`so(4,c)`$.
If we define $`l:=nq`$, the eigenvalue equation for Casimir operator $`H_{_{sl(2,c)}}`$ will be
$$H_{_{sl(2,c)}}\mathrm{\Psi }_{_{l,m,q}}(\theta ,\varphi ,\psi )=\frac{1}{2}\gamma ^2l(l+1)\mathrm{\Psi }_{_{l,m,q}}(\theta ,\varphi ,\psi )$$
(3.15)
with $`\mathrm{\Psi }_{_{l,m,q}}(\theta ,\varphi ,\psi )=e^{im\varphi iq\psi }P_{_{m,q}}^l(\mathrm{cos}\gamma \theta ).`$ After transfer the $`\psi `$ dependence factor $`e^{iq\psi }`$ to the left hand side of equation (3.15), the reduced Casimir operator with degeneracy group $`SL(2,c)`$, is
$`H_{_{sl(2,c)}}(q)={\displaystyle \frac{1}{4}}J__+^{(L)}(q)J_{_{}}^{(L)}(q)+{\displaystyle \frac{1}{4}}J_{_{}}^{(L)}(q)J__+^{(L)}(q)+{\displaystyle \frac{1}{2}}\gamma ^2J__3^{(L)^2}(q),`$ (3.16)
where
$`J__\pm ^{(L)}(q)=e^{\pm i\varphi }(\pm {\displaystyle \frac{}{\theta }}+{\displaystyle \frac{i\gamma }{\mathrm{tan}\gamma \theta }}{\displaystyle \frac{}{\varphi }}{\displaystyle \frac{q\gamma }{\mathrm{sin}\gamma \theta }})`$
$`J__3^{(L)}(q)=i{\displaystyle \frac{}{\varphi }}.`$
The above reduced $`sl(2,c)`$ generators are related to (3.10) generators of $`gl(2,c)`$ through a similarity transformation together with a change of basis, such that, one can obtain the eigenstates of $`gl(2,c)`$ Casimir operator. Thus, the eigenvalue equation (3.15) reduces to
$$H_{_{sl(2,c)}}(q)\mathrm{\Psi }_{_{l,m,q}}(\theta ,\varphi )=\frac{1}{2}\gamma ^2l(l+1)\mathrm{\Psi }_{_{l,m,q}}(\theta ,\varphi ),$$
(3.17)
with $`\mathrm{\Psi }_{_{l,m,q}}(\theta ,\varphi )=e^{im\varphi }P_{_{m,q}}^l(\mathrm{cos}\gamma \theta )`$. From there we can introduce the eigenfunction
$`\mathrm{\Psi }_{n,m}(\theta ,\varphi )=(2{\displaystyle \frac{\mathrm{tan}\frac{\gamma \theta }{2}}{\mathrm{sin}\gamma \theta }}e^{i\varphi })^qe^{im\varphi }P_{_{m,q}}^l(\mathrm{cos}\gamma \theta ),`$
for Hamiltonian (3.11) with the same eigenvalue (3.15). Then, we see that the Dirac’s quantization of magnetic charge follows very naturally from the $`SL(2,c)`$ representation, that is integrity of $`q`$. Also, the $`so(4,c)`$ Lie algebra is the dynamical group of the quantum dynamics of a point mass particle in the presence of a constant magnetic field over two dimensional surface with global constant curvature with $`gl(2,c)`$ sub-algebra as its degeneracy group.
Transferring the function $`e^{iq\psi }`$ to the left of the Casimir operator $`H_{_{sl(2,c)}}`$ in the eigenvalue equation (3.15), the Casimir operator $`H_{_{sl(2,c)}}(q)`$ can be written as
$$H_{_{sl(2,c)}}(q)=\frac{1}{2}J__+^{(R)}(q+1)J_{_{}}^{(R)}(q)+\frac{1}{2}q(q+1)\gamma ^2$$
(3.18a)
$$=\frac{1}{2}J_{_{}}^{(R)}(q1)J__+^{(R)}(q)+\frac{1}{2}q(q1)\gamma ^2$$
(3.18b)
with
$`J__\pm ^{(R)}(q)=\pm {\displaystyle \frac{}{\theta }}i{\displaystyle \frac{\gamma }{\mathrm{sin}\gamma \theta }}{\displaystyle \frac{}{\varphi }}+{\displaystyle \frac{q\gamma }{\mathrm{tan}\gamma \theta }}.`$
Using the Eqs. (3.18) in the eigenvalue equation (3.17) we obtain
$`J__+^{(R)}(q)J_{_{}}^{(R)}(q1)\mathrm{\Psi }_{_{l,m,q1}}(\theta ,\varphi )=E__q\mathrm{\Psi }_{_{l,m,q1}}(\theta ,\varphi )`$
$`J_{_{}}^{(R)}(q1)J__+^{(R)}(q)\mathrm{\Psi }_{_{l,m,q}}(\theta ,\varphi )=E__q\mathrm{\Psi }_{_{l,m,q}}(\theta ,\varphi ),`$
where $`E__q`$ is defined as
$`E__q=\gamma ^2[l(l+1)q(q1)].`$
The restriction of $`so(4,c)`$ Lie algebra to $`sl(2,c)`$, together with $`H_{_{sl(2,c)}}`$ reduction, lead to eigenvalue equation (3.17) which can be obtained from Hamiltonian (3.11) via similarity transformation. The Hamiltonian $`H_{_{sl(2,c)}}(q)`$, on the one hand, is quadratic Casimir operator of $`sl(2,c)`$ Lie algebra with $`J__\pm ^{(L)}(q)`$ and $`J__3^{(L)}(q)`$ generators which, as a result, possesses $`gl(2,c)`$ degeneracy symmetry. On the other hand, writting the Hamiltonian in terms of $`J__\pm ^{(R)}(q)`$ will possess the shape invariance symmetry. Both properties are related to the realization of para-supersymmetry of arbitrary order .
ACKNOWLEDGEMENT
We wish to thank Dr. S.K.A. Seyed Yagoobi for carefully reading the article and for his constructive comments.
|
no-problem/9904/nucl-ex9904012.html
|
ar5iv
|
text
|
# Observation of Thermodynamical Properties in the 162Dy, 166Er and 172Yb Nuclei
## Abstract
The density of accessible levels in the (<sup>3</sup>He,$`\alpha \gamma `$) reaction has been extracted for the <sup>162</sup>Dy, <sup>166</sup>Er and <sup>172</sup>Yb nuclei. The nuclear temperature is measured as a function of excitation energy in the region of 0 – 6 MeV. The temperature curves reveal structures indicating new degrees of freedom. The heat capacity of the nuclear system is discussed within the framework of a canonical ensemble.
A challenging goal in nuclear physics is to trace thermodynamical quantities as functions of excitation energy. These quantities depend on statistical properties in the nuclear many body system and may reveal phase transitions. Unfortunately, it is difficult to investigate these aspects - both experimentally and theoretically.
The density of levels as a function of excitation energy is the starting point to extract quantities like entropy, temperature and heat capacity. In the pioneer work of Bethe the level density was described within the Fermi gas model using a partition function for the grand-canonical ensemble. This picture of the nucleus as a gas of non-interacting fermions confined to the nuclear volume has later been modified. The phenomenological back-shifted Fermi gas model is a popular extension of the model, which simulates shell and pair correlation effects. This model works well at excitation energies where the level density is high, typically above the neutron binding energy.
In addition to reveal statistical properties of nuclear matter, knowledge of the level density is important in nuclear astrophysics. The level density is essential for the understanding of the nucleosynthesis in stars, where thousands of cross-sections have to be included in the calculations . In parallel with these applications new theoretical approaches are emerging. Recent microscopic model calculations include pair correlations as well as shell effects. With the recent shell model Monte Carlo method one is able to estimate level densities in heavy nuclei up to high excitation energies.
The theoretical progress has so far not been followed by new experimental data. The straightforward way to determine level densities is by counting discrete levels. However, this technique is restricted to light nuclei and/or low excitation energies where the experimental resolution is high enough to resolve individual lines in the spectra. In rare earth nuclei, the estimates obtained by counting levels are only valid up to $``$2 MeV of excitation energy. A very useful experimental quantity is the average level spacing observed in slow neutron resonance capture. From these spacings rather accurate level densities for a certain spin value can be determined at the neutron binding energy region. In addition, the level density can be extracted from the shape of continuum particle spectra. However, if pre-equilibrium particle emission takes place, this procedure may be doubtful since high-energy particles reveal high temperatures and, thus, too low level densities.
Recently, the Oslo group has presented a new way of extracting level densities from measured $`\gamma `$-ray spectra . The main advantage of this method is that the nuclear system is very likely thermalized prior to the $`\gamma `$-ray emission. In addition, the method allows the simultaneous extraction of level density and the $`\gamma `$-strength function over a wide energy region. In this letter we report for the first time on experimentally deduced temperatures and heat capacities of rare earth nuclei in the 0 – 6 MeV excitation energy region.
The experiments were carried out with 45 MeV <sup>3</sup>He-projectiles at the Oslo Cyclotron Laboratory (OCL). The experimental data are obtained with the CACTUS multidetector array using the (<sup>3</sup>He,$`\alpha \gamma `$) reaction on <sup>163</sup>Dy, <sup>167</sup>Er and <sup>173</sup>Yb self-supporting targets. The charged ejectiles were detected with eight particle telescopes placed at an angle of 45 relative to the beam direction. An array of 28 NaI $`\gamma `$-ray detectors with a total efficiency of $``$15% surrounded the target and particle detectors.
The assumptions behind the method of data analysis and techniques are described in Refs. and only a few comments pertinent to the present work are made here. The experimental level density is deduced from $`\gamma `$-ray spectra recorded at a number of initial excitation energies $`E`$, determined by the measured $`\alpha `$-energy. These data are the basis for making the first-generation (or primary) $`\gamma `$-ray matrix, which is factorized according to the Brink-Axel hypothesis as
$$P(E,E_\gamma )=\sigma (E_\gamma )\rho (EE_\gamma ).$$
(1)
From this expression the $`\gamma `$-ray energy dependent function $`\sigma `$ as well as the level density $`\rho `$ is deduced by an iteration procedure , using the same 0th-order trial functions as described in Ref. , and the average slopes and the absolute density depend on this starting point. In the following we concentrate only on the level density and its fine structure, which is assumed to be independent of particular $`\gamma `$-ray decay routes.
The extracted level densities for the <sup>162</sup>Dy, <sup>166</sup>Er and <sup>172</sup>Yb nuclei are shown as data points in Fig. 1. The data for <sup>162</sup>Dy and <sup>172</sup>Yb deviate slightly from the results previously published . In the present work the extraction procedure has been refined by omitting data with $`E_\gamma <`$ 1 MeV, where the first-generation $`\gamma `$-spectra exhibit methodical uncertainties. By excluding these data points, the error bars for the resulting level density are reduced.
The level density $`\rho (E)`$ is proportional to the number of states accessible to the nuclear system at excitation energy $`E`$. Thus, the entropy in the microcanonical ensemble is given by
$$S(E)=S_0+\mathrm{ln}\rho (E),$$
(2)
where, for convenience, the Boltzmann constant is set to unity ($`k_B`$ =1). The normalization constant $`S_0`$ is not important in this discussion, since it vanishes in the evaluation of the temperature
$$T(E)=\frac{1}{(S/E)_V}.$$
(3)
Small statistical deviations in the entropy $`S`$ may give rise to large contributions in the temperature $`T`$. In order to reduce this sensitivity, the differentiation of $`S`$ is performed by a least square fit of a straight line to five adjacent data points at a time. The slope of the straight line is taken as the differential of $`S`$ at that energy. Thus, an effective smoothing of about 0.5 MeV is performed through this procedure. Since the energy particle resolution is around 0.3 MeV, this differentiation procedure will not significantly reduce the potential experimental information.
The temperatures deduced are shown as discrete data points in Fig. 2. The data reveal several broad structures in the 1 – 5 MeV region, which are not explained in a Fermi gas description. The most pronounced bumps are located at 1.8 MeV and 3.2 MeV in <sup>162</sup>Dy, at 2.5 MeV and 3.7 MeV in <sup>166</sup>Er and at 1.8 MeV and 3.0 MeV in <sup>172</sup>Yb. These structures are interpreted as being the breaking of nucleon pairs and, at higher energies, the possible quenching of pair correlations.
The extraction of specific heat capacity is given by
$$C_V(E)=\frac{1}{(T/E)_V},$$
(4)
which effectively means that the entropy $`S(E)`$ has to be differentiated twice. From the scattered data points of Fig. 2, one immediately sees that this will not give meaningful information for $`C_V`$ with the present experimental statistics.
At this point one could introduce a strong smoothing of the temperature curve in order to extract $`C_V`$. However, it seems rather incidental how this should be done. One way to proceed is to introduce the canonical ensemble in the description of the nuclear system. Since the temperature enters as a fixed parameter in this formalism, the statistical uncertainty of $`T`$ does not introduce any additional fluctuations. The cost of this nice feature is that one has to make an average over a wide excitation energy region.
The partition function in the canonical ensemble
$$Z(T)=\underset{n=0}{\overset{\mathrm{}}{}}\rho (E_n)e^{E_n/T}$$
(5)
is determined by the multiplicity of states at a certain energy. Experimentally, the multiplicity corresponds to the level density of accessible states, $`\rho (E_n)`$, in the present nuclear reaction at energy bin $`E_n`$. The widths of the energy bins are 120, 120 and 96 keV for the <sup>162</sup>Dy, <sup>166</sup>Er and <sup>172</sup>Yb nuclei, respectively.
The mathematical justification of Eq. (5) is that the sum is performed from zero to infinity. The experimental information on $`\rho `$ covers the excitation region 0 – 7 MeV, only. For the three nuclei investigated, the proton and neutron binding energies are $``$8 MeV. For excitation energies above this binding energy it is reasonable to assume Fermi gas properties, since single particles may be excited into a region with a very high level density (continuum). Therefore, due to lack of experimental data above 7 MeV, the level density is extrapolated from 7 MeV to higher energies by the Fermi gas model expression
$$\rho _{FG}(E)=CE^{5/4}e^{2\sqrt{aE}},$$
(6)
where $`C`$ is a normalization factor and $`a`$ is the level density parameter. The best fits of $`\rho _{FG}`$ to data in the 3.5 – 6.0 MeV excitation region give a level density parameter of $`a`$ =17.8, 19.0, and 18.7 MeV<sup>-1</sup> for the <sup>162</sup>Dy, <sup>166</sup>Er and <sup>172</sup>Y nuclei, respectively. These values are in agreement with data from slow neutron resonances and the semiempirical formula for $`a`$ with values in between $`A/10`$ and $`A/8`$ MeV<sup>-1</sup>. The theoretical level density functions are displayed as solid lines in Fig. 1 and are drawn from 3.5 MeV in order to visualize the fits to data.
In order to determine how far the sum of Eq. (5) has to be performed, the probability density function
$$p(E)=\frac{\rho _{FG}(E)e^{E/T}}{\rho _{FG}(E)e^{E/T}𝑑E}$$
(7)
is shown in Fig. 3 for three typical temperatures. At low temperatures, say $`T`$ = 0.3 MeV, the nucleons are scattered no higher than $`E`$ =5 MeV. However, this upper limit increases very rapidly for higher temperatures. For the typical temperatures studied in this letter, $`E`$ =30 MeV has been tested and found to be a sufficiently high upper limit for the summing in Eq. (5).
The excitation energy in the canonical ensemble is given by the thermal average
$$<E(T)>=Z^1\underset{n=0}{\overset{\mathrm{}}{}}E_n\rho (E_n)e^{E_n/T},$$
(8)
where $`\rho `$ is the level density (shown in Fig. 1) composed of an experimental and a theoretical part for the excitation regions below and above $``$7 MeV, respectively.
The smoothing effect implied by the canonical ensemble can be investigated by calculating the standard deviation $`\sigma _E`$ for the thermal average of the energy
$$\sigma _E=\sqrt{<E^2><E>^2},$$
(9)
giving e.g. $`\sigma _E`$ = 3 MeV at $`E`$ =7 MeV. This again shows that the energy in the canonical ensemble is strongly smoothed for a given temperature. Thus, one cannot expect to find abrupt changes in the thermodynamical quantities.
From Eq. (8) the temperature in the canonical ensemble can be studied as a function of $`<E>`$. In Fig. 2 the canonical ensemble gives a temperature dependence (solid lines) that coincides well with the average values found in the microcanonical ensemble (data points).
This gratifying behaviour of the canonical temperature encourages us to use the canonical ensemble to estimate the heat capacity $`C_V`$ as well. The heat capacity can be deduced by simply calculating the increase in the thermal average of the energy $`<E>`$ with respect to $`T`$
$$C_V(T)=\frac{<E>}{T}.$$
(10)
The deduced heat capacities for the <sup>162</sup>Dy, <sup>166</sup>Er and <sup>172</sup>Yb nuclei as functions of $`<E(T)>`$ are shown in Fig. 4. All nuclei display similar dependencies, reflecting that the thermal averages of the excitation energies smear out structures seen in the experimental level densities. The heat capacities are mainly following the theoretical values obtained by using Eq. (6) with the proper values of $`a`$ deduced from the fits of Fig. 1. For comparison, we also show the simplified expression $`C_V=2\sqrt{aE}`$ assuming $`\rho \mathrm{exp}(2\sqrt{aE})`$ with $`a=18.5`$ MeV<sup>-1</sup>. The canonical heat capacity shows no traces of the fine structures found in the level density below $``$4 MeV of excitation energies, as also was the case for the canonical temperature.
In conclusion, for the first time temperature and heat capacity based on $`\gamma `$-ray spectra have been extracted for rare earth nuclei. The extracted temperature curves deduced for the microcanonical ensemble reveal structures of unknown origin. The structures in the 1 – 5 MeV excitation region are interpreted as the breaking of nucleon pairs and quenching of the pair correlations. The heat capacity could only be extracted by the use of the canonical ensemble. These semiexperimental values, which are not expected to show any fine structures, are in agreement with the Fermi gas predictions. It would be very interesting to see if more realistic theoretical calculations can describe the fine structures observed in the microcanonical ensemble.
The authors are grateful to E.A. Olsen and J. Wikne for providing the excellent experimental conditions. We wish to acknowledge the support from the Norwegian Research Council (NFR).
|
no-problem/9904/cond-mat9904220.html
|
ar5iv
|
text
|
# Ground state parameters, finite-size scaling, and low-temperature properties of the two-dimensional 𝑆=1/2 XY model
## I Introduction
Studies of effective continuum field theories have resulted in detailed predictions for the low-energy physics of quantum spin systems in two dimensions. In order to make use of these predictions for a given model Hamiltonian, the ground state parameters appearing in the Lagrangian formulation have to be determined. Spin-wave theory can in some cases give good estimates, but in general some numerical method has to be employed in order to obtain accurate results. Since the theories can also predict the finite-size scaling behavior of various physical quantities, numerical results for a series of lattice sizes can be used to extract the ground state parameters. Such calculations are also important for testing theoretical predictions.
With todays computers, Lanczos and related exact diagonalization methods can be used for square lattices with up to $`6\times 6`$ spins. This relatively small maximum size, and the small number of different lattices available, can make the finite-size scaling procedures problematic if sub-leading corrections are significant. It is therefore important to consider also alternative methods that can reach larger lattice sizes.
Here we discuss quantum Monte Carlo (QMC) results for the $`S=1/2`$ XY model, defined by the Hamiltonian
$$H=J\underset{i,j}{}[S_i^xS_j^x+S_i^yS_j^y],$$
(1)
where $`S_i^x`$ and $`S_i^y`$ are the $`x`$\- and $`y`$-components of a spin-1/2 operator at site $`i`$, and $`i,j`$ denotes a pair of nearest-neighbor sites on a square lattice. Numerical studies of this model have a long history. Exact finite-lattice calculations gave the first indications that the O(2) symmetry is spontaneously broken at $`T=0`$ in the thermodynamic limit. This was later proved rigorously. QMC simulations have resulted in a quite precise value of the ordered moment and have also shown that there is a Kosterliz-Thouless (KT) transition at a temperature $`T_{\mathrm{KT}}0.343J`$. For $`TT_{\mathrm{KT}}`$ the system acquires quasi-long-range order, i.e., power law decay of the spin-spin correlation function and a non-zero spin stiffness constant. In accordance with the Mermin-Wagner theorem, true long-range order develops only at $`T=0`$.
Our motivation for carrying out calculations for the XY model to even higher precision is to test predictions of effective Lagrangian theories. In particular, the chiral perturbation calculations by Hasenfratz and Niedermayer have resulted in detailed predictions for the finite-size and finite-temperature corrections of various quantities, in some cases beyond leading order. The finite-size and finite-temperature scaling behavior of the O(3) symmetric Heisenberg model has been the topic of numerous studies, and the agreement with the predictions has been confirmed to high precision. The predictions for the O(2) symmetric XY model have not yet been tested exhaustively, however. A recent finite-size scaling study of exact energies for systems with up to $`6\times 6`$ spins has quite convincingly demonstrated agreement with the leading finite-size behavior. Here we consider also several other physical observables. This allows us to carry out a number of independent tests of the consistency of the scaling predictions. With access to larger system sizes, we can also improve considerably on the accuracy of the extrapolated ground state parameters and finite-size corrections. Our data are sufficiently accurate for addressing also sub-leading corrections. In addition to calculations in the ground state, we have also studied systems at finite temperature, on lattices sufficiently large to enable extraction of the leading finite-temperature corrections in the thermodynamic limit.
We have used a numerically exact finite-temperature QMC method based on “stochastic series expansion”, i.e., importance sampling of diagonal matrix elements of the power series expansion of $`\mathrm{exp}(\beta H)`$, where $`\beta =1/T`$ is the inverse temperature. The ground state can be obtained by choosing a sufficiently large $`\beta `$ value. This method has previously been used to study the finite-size scaling behavior and the ground state parameters of the square-lattice Heisenberg model. Recently, a new way of sampling the terms of the expansion was proposed (the “operator-loop” algorithm), which enables calculations to even higher precision. We have used this improved method to study the ground state of the XY model on lattices with $`N=L\times L`$ spins with $`L`$ up to $`16`$. We have carried out calculations at finite temperature for systems with $`L`$ up to $`64`$.
In Sec. II we discuss the physical quantities that we have calculated. The numerical ground state data and the results for finite-size corrections and extrapolations to infinite size are presented in Sec. III. In Sec. IV we discuss calculations at $`T>0`$. Sec. V concludes with a brief summary.
## II Calculated quantities
The QMC algorithm can be implemented in two different ways; with the spin quantization either along the $`z`$ or $`x`$ axis. The internal energy, $`E=H/N`$, can be easily calculated in both cases. In the $`z`$-representation, the spin stiffness constant $`\rho `$ can also be evaluated. It is defined as the second derivative of the internal energy with respect to a twist $`\varphi `$ under which, for all nearest-neighbor spin pairs in either the $`x`$ or $`y`$ lattice direction, the spin-spin interaction is modified according to
$`S_i^xS_j^x`$ $`+S_i^yS_j^y(S_i^xS_j^x+S_i^yS_j^y)\mathrm{cos}(\varphi )`$ (3)
$`S_i^xS_j^y\mathrm{sin}(\varphi )+S_i^yS_j^x\mathrm{sin}(\varphi ).`$
In analogy with the superfluid density of a boson system, one can show that
$$\rho =\frac{^2E(\varphi )}{^2\varphi }=\frac{W_x^2+W_y^2}{2N\beta },$$
(4)
where $`E(\varphi )`$ is the internal energy per spin and $`W_x`$ and $`W_y`$ are the “winding numbers”, i.e., the net spin currents across the periodic boundaries in the $`x`$ and $`y`$ directions that characterize configurations in simulations carried out in the $`z`$-representation. We also calculate the spin susceptibility, given by
$$\chi =\frac{\beta }{N^2}\left(\underset{i=1}{\overset{N}{}}S_i^z\right)^2.$$
(5)
In the $`z`$-representation, it is not convenient to calculate spin-spin correlations involving the $`x`$ or $`y`$ spin components. We therefore also use an algorithm implemented in the $`x`$ representation, and there calculate the squared magnetization $`M_x^2`$;
$$M_x^2=\frac{1}{N^2}\left(\underset{i=1}{\overset{N}{}}S_i^x\right)^2.$$
(6)
Since the O(2) symmetry is not broken on a finite lattice, the magnetization $`m`$ per spin in the thermodynamic limit is given by the infinite-size extrapolated value of
$$m=\sqrt{2M_x^2}.$$
(7)
We use periodic boundary condition and include each spin pair only once in Eq. (1). For technical details on the simulation algorithm we refer to previously published work.
## III Ground state results
In order to obtain the ground state, we have carried out simulations at inverse temperatures as high as $`\beta =1024`$ for lattices with $`L=416`$. In the $`z`$-representation simulations, we have calculated the ground state energy to within relative statistical errors lower than $`10^5`$. The relative accuracy of the stiffness is on the order of $`10^3`$. Simulations in the $`x`$-representation result in statistical errors for the energy roughly twice as large as in the $`z`$-representation (for simulations of comparable length). The squared magnetization is the least accurate quantity, with statistical errors of roughly $`0.3\%`$ for the largest system size considered in this case ($`L=12`$). In Table I we list the results used in the finite-size analysis presented below. The results for the energy for $`L=4`$ and $`6`$ are in perfect agreement with previous exact diagonalizations. The magnetization for $`L=4`$ also agrees with the exact diagonalization result. However, for $`L=6`$ there is an $`5\sigma `$ deviation from the result presented in Ref. . We do not know the reason for this disagreement. The QMC method has previously been shown to give perfect agreement with isotropic Heisenberg results for both $`L=4`$ and $`L=6`$, and therefore a failure for (only) the XY-model magnetization for $`L=6`$ would be surprising.
For a model with O(2) symmetry, chiral perturbation theory predicts the size dependence of the energy as
$$E_0(L)=E_0+\frac{e_3}{L^3}+\frac{e_5}{L^5}+\mathrm{},$$
(8)
with no $`O(1/L^4)`$ term. The leading-order correction is given by
$$e_3=\gamma c/2,$$
(9)
where the constant $`\gamma =1.437745`$ and $`c`$ is the spin-wave velocity. In order to obtain a good fit to our $`L=416`$ data, the $`O(1/L^5)`$ term has to be included. The fit then has a chi-squared value per degree of freedom of $`0.7`$. Using instead an $`O(1/L^4)`$ term gives chi-squared $`6`$. We can therefore conclude that our results support the prediction that $`e_4=0`$. In Figure 1 we show the data for $`L8`$ along with the best fit. The extrapolation $`L\mathrm{}`$ gives $`E_0=0.548824(2)`$. The finite-size correction constants are $`e_3=0.807(2)`$ and $`e_5=1.07(3)`$. Using Eq. (9) we obtain the spin-wave velocity $`c=1.123(2)`$. These results are in agreement with the previous extrapolations using exact diagonalization data for $`L=2,3,4,5,6`$, but the statistical errors are considerably smaller. The energy is also in perfect agreement with a previous Green’s function Monte Carlo calculation, which gave $`E_0=0.54883(1)`$.
We are not aware of any predictions for the size dependence of the spin stiffness of the XY model. For the Heisenberg model, the leading correction is $`O(1/L)`$. In contrast, our XY data can be very well fit with only an $`O(1/L^3)`$ term;
$$\rho (L)=\rho +\frac{r_3}{L^3}+\mathrm{},$$
(10)
with the infinite-size value $`\rho =0.2696(2)`$ and the cubic correction $`r_3=0.47(1)`$. This fit is shown in Figure 2.
The stiffness result is in perfect agreement with the value obtained using exact diagonalization and finite-size scaling of the ground state energy in higher magnetization sectors. The energy per spin in the sector with magnetization $`j=_iS_i^z`$ should scale as
$$E_jE_0=\frac{j^2c^2}{2\rho L^4}+O(j^4,1/L^6).$$
(11)
We have also carried out some QMC calculations in the $`j=1`$ sector. Figure 3 shows results for $`L=8,10`$, and $`12`$, along with the exact result for $`L=6`$. Using also the exact $`L=4`$ result, a fit to Eq. (11), including an $`O(1/L^6)`$ term, gives $`c^2/\rho =4.70(4)`$. This value agrees with the above separate estimates of $`c`$ and $`\rho `$. The consistency between the results, obtained in two different ways, clearly gives very strong support to the Lagrangian theory.
The susceptibility can be obtained from $`c`$ and $`\rho `$ using the standard hydrodynamic relation
$$\chi =\rho /c^2,$$
(12)
which with our values of $`c`$ and $`\rho `$ gives $`\chi =0.2138(8)`$.
For the magnetization, we assume a finite-size scaling
$$M_x^2(L)=M_x^2+\frac{a_1}{L}+\frac{a_2}{L^2}+\mathrm{}.$$
(13)
A fit to this form, including the quadratic term, is shown in Figure 4 and gives $`M_x^2=0.0956(6)`$ and $`a_1=0.1212(5)`$. According to Eq. (7), the magnetization is thus $`m=0.437(2)`$, in good agreement with a previous Green’s function Monte Carlo result \[$`m=0.441(5)`$\]. For the Heisenberg model, the linear correction factor is related to ground state parameters according to $`a_1=\alpha M_x^2/(c\chi )`$, where $`\alpha =0.62075`$. The $`a_1`$ obtained here is instead consistent to within a few percent with $`a_1=\alpha M_x^2/(2c\chi )`$. As in the scaling of the energy, the leading size correction is hence proportional to the number of gapless modes in the symmetry-broken system.
## IV Finite-temperature results
We now discuss calculations aimed at extracting the leading finite-temperature corrections to the ground state. Previously, extensive simulations were carried out in order to study the KT transition, and a critical temperature $`T_{\mathrm{KT}}0.343J`$ was found. Here we focus on the susceptibility and the internal energy density at lower temperatures.
Figure 5 shows the susceptibility, defined in Eq. (5), calculated for lattices with $`L=32`$ and $`64`$ for temperatures $`T/J0.05`$. Within statistical errors there are no differences between the data for the two system sizes, and hence the results represent the thermodynamic limit. Chiral perturbation theory predicts a temperature-independent susceptibility for the XY model. Within error bars, our data is temperature independent for $`T/J0.15`$, with the constant value estimated at $`\chi =0.2096(2)`$. This result is lower (by about five standard deviations, or $`2`$%) than the value extracted in Sec III from the ratio $`\chi =\rho /c^2`$. The spin-wave velocity was obtained from the leading finite-size correction to the ground state energy. It is possible that this correction is affected by the presence of sub-leading corrections that are not completely taken into account by only including up to $`O(1/L^5)`$ terms in Eq. (8). Using the data shown in Fig. 5 is a more direct way of extracting the $`T=0`$ susceptibility, and we therefore consider it more reliable. An improved estimate of $`c`$ is then $`c=\rho /\chi =1.134(2)`$.
For a finite lattice, the gap between the $`j=0`$ ground state and the finite-magnetization states implies an exponential decay of $`\chi `$ to zero as $`T0`$. Chiral perturbation theory predicts that the form of this decay for $`TL^2\rho /c^21`$ is given by
$$\chi =\frac{1}{TL^2}\mathrm{exp}[c^2/(2\rho TL^2)].$$
(14)
In Figure 6 we show QMC data for $`L=416`$ along with this prediction, where we have used $`c^2/\rho =4.77`$, corresponding to our best estimates of $`c`$ and $`\rho `$. The agreement is not perfect, but satisfactory considering that there are no adjustable parameters and that there should also be corrections to Eq. (14). For $`L=4`$, we have also calculated $`\chi `$ using exact diagonalization down to much lower temperatures. As shown in Figure 7, there is a very good agreement with the predicted form over a sizable low-temperature range. Note, however, that the asymptotic $`T0`$ decay is always purely exponential, without the $`1/T`$ factor in Eq. (14), as only the lowest $`j=\pm 1`$ states contribute.
Chiral perturbation theory predicts that the low-temperature form of the internal energy is given by
$$E(T)=E(0)+\frac{\zeta (3)}{\pi c^2}T^3+O(T^5),$$
(15)
where $`\zeta (3)=1.20206`$. In Figure 8 we show results for $`T/J0.05`$ calculated for $`L=64`$ lattices. We also show a comparison between $`L=16,32`$, and $`64`$, which suggests that $`L=64`$ gives the thermodynamic limit within error bars. A fit to the $`L=64`$ data with $`O(T^3)`$ and $`O(T^5)`$ terms, and including also the $`T=0`$ energy extracted in Sec. III, gives the cubic correction $`\zeta (3)/\pi c^2=0.284(5)`$. This corresponds to $`c=1.16(1)`$, which is higher than the value $`1.134(2)`$ extracted above using the estimates of $`\rho `$ and $`\chi `$ by about 2.5 error bars. Again, this small deviation may indicate some influence of higher-order corrections in the energy scaling.
## V Summary
In summary, we have presented extensive QMC calculations for the two-dimensional $`S=1/2`$ XY model. We have carried out finite-size and finite-temperature scaling of several physical quantities. The results are consistent with the predictions of effective Lagrangian theory to within $`12`$%.
The best estimates of the ground state parameters resulting from our calculations are
$`E_0`$ $`=`$ $`0.548824(2),`$ (16)
$`\rho `$ $`=`$ $`0.2696(2),`$ (17)
$`m`$ $`=`$ $`0.437(2),`$ (18)
$`\chi `$ $`=`$ $`0.2096(2),`$ (19)
$`c`$ $`=`$ $`1.134(2).`$ (20)
The ground state energy, $`E_0`$, the spin stiffness, $`\rho `$, and the square of the magnetization, $`m^2`$, were all calculated directly in the ground state for systems of linear size $`L=416`$ and extrapolated to infinite size. The susceptibility $`\chi `$ perpendicular to the XY spin-plane was calculated at finite temperature and extrapolated to $`T=0`$. Only the spin-wave velocity $`c`$ was obtained by a more indirect procedure, using the relation $`c^2=\rho /\chi `$. The results are in good agreement with previous exact diagonalization and QMC work, but the precision is considerably improved over other estimates. Our results for both the energy and the magnetization are in remarkably good agreement with a series expansion calculation. The magnetization is also in excellent agreement with second order spin-wave theory. The energy obtained in spin-wave theory is only $`0.2`$% higher than the numerical result obtained here. It would be interesting to carry out high-order spin-wave calculations also for the other quantities discussed here.
## VI Acknowledgments
We would like to thank Jaan Oitmaa for discussions and comments on the manuscript. A. W. S. would like to thank the School of Physics at the University of New South Wales for hospitality and financial support during a visit. Support from the NSF under Grant No. DMR-9712765 is also acknowledged.
|
no-problem/9904/astro-ph9904121.html
|
ar5iv
|
text
|
# EXPLOSIONS DURING GALAXY FORMATION. SCALE-FREE SIMULATIONS
## 1 INTRODUCTION
Galaxy formation by gravitational condensation out of the expanding cosmological background universe is affected by the complex interplay of gravitational dynamics of collisionless dark matter and the gas dynamics of the baryon-electron fluid, including the feedback on the latter which results when energy is released by the stars and AGN’s that form. Results are presented here of 3D numerical gas dynamical simulations of the effect of this energy release, utilizing our anisotropic Adaptive Smoothed Particle Hydrodynamics (“ASPH”) method. Current computational limitations make it virtually impossible for numerical simulations of the initial value problem involving a realistic initial spectrum of small-amplitude, Gaussian random noise primordial density fluctuations to resolve the full range of length and mass scales necessary to form galaxies and the stars within them at the same time. We choose, instead, to focus here on an idealized model of structure formation which we can hope to resolve more accurately, thereby elucidating some of the most important aspects of the problem which may be relevant, as well, to more realistic initial conditions. We are particularly interested in studying the effects of nonspherical geometry and continuous cosmological infall on the problem of explosion-driven blow-out from galaxies during their formation.
It is well-known that structure formation from Gaussian random noise proceeds in a highly anisotropic way, favoring pancake and filament formation over the formation of quasi-spherical objects. Our previous work (,) has demonstrated, however, that a cosmological pancake, modeled as the nonlinear outcome of a single plane-wave density fluctuation, is subject to a linear gravitational instability which results in the formation of filaments and lumps in the central plane of the pancake. The lumps of collisionless dark matter that form in this way are quasi-spherical and develop a universal density profile which is reminiscent of the universal profiles found from N-body simulations of 3D Gaussian noise density fluctuations in hierarchical clustering models like the CDM model. As such, they provide an ideal test-bed for exploring the gas dynamics of structure formation and feedback effects without the troublesome complexity of Gaussian random noise initial conditions. The pancake problem, moreover, is completely scale-free, once all lengths are expressed in units of the pancake wavelength $`\lambda _p`$, time is expressed in terms of the cosmic scale factor $`a`$, divided by the scale factor $`a_c`$ at which the unperturbed pancake collapses to form density caustics in the dark matter and shocks in the gas, and the energy release is expressed in units of the total energy contained in a comoving cube of side $`\lambda _p`$. As such, each simulation of galaxy formation via 3D pancake gravitational instability serves to represent the generic behavior independent of the particular mass or collapse epoch of the object which forms. In order to preserve this universality and scale-free character of the problem, we will neglect radiative cooling in these first calculations. Once the key features of this scale-free problem are delineated, we will later consider the scale-dependent effects of radiative cooling and photoheating. Among the results we seek to quantify in this way are:
* The amount of energy release required to blow the gas out of a dark matter halo.
* The efficiency for ejecting the fraction of gas which is initially responsible for receiving the energy release (and, in the case of supernova explosions, the metallicity associated with the SN ejecta).
* The distinction between “blowout,” in which the energy release results in the escape of some energy and gas into the surrounding IGM but leaves the bulk of the gas in the object unaffected, and “blowaway,” in which most or all of the gas is ejected from the dark matter potential well.
* The energy release rate required to shock-heat the entire IGM by the overlapping effect of energy release from neighboring objects.
## 2 PANCAKE INSTABILITY AND FRAGMENTATION AS A TEST-BED MODEL FOR GALAXY FORMATION
### 2.1 Model and Initial Conditions
We consider an Einstein-de Sitter universe (density parameter $`\mathrm{\Omega }_0=1`$, cosmological constant $`\lambda _0=0`$) with $`\mathrm{\Omega }_B=0.03`$ and $`\mathrm{\Omega }_X=0.97`$ (where $`\mathrm{\Omega }_B`$ and $`\mathrm{\Omega }_X`$ are the contributions of baryons and dark matter to $`\mathrm{\Omega }_0`$, respectively). The initial conditions correspond to the growing mode of a single sinusoidal plane-wave density fluctuation of wavelength $`\lambda _p`$ and dimensionless wavevector $`𝐤=\widehat{𝐱}`$ (length unit = $`\lambda _p`$). We adjust the initial amplitude $`\delta _i`$ such that a density caustic forms in the collisionless dark matter component at scale factor $`a=a_c`$.
We perturb this system by adding to the initial conditions two transverse, plane-wave density fluctuations with equal wavelength $`\lambda _s=\lambda _p`$, wavenumbers $`𝐤_s`$ pointing along the orthogonal vectors $`\widehat{𝐲}`$ and $`\widehat{𝐳}`$, and amplitude $`ϵ_y\delta _i`$ and $`ϵ_z\delta _i`$, respectively, where $`ϵ_y1`$ and $`ϵ_z1`$. We use the notation $`S_{1,ϵ_y,ϵ_z}`$ to designate a pancake perturbed by two such transverse perturbation modes. All results presented here refer to the case $`S_{1,0.2,0.2}`$ unless otherwise noted. The presence of the two perturbation modes will result in the formation of two perpendicular filaments in the plane of the pancake, with a dense, quasi-spherical cluster at the intersection of the filaments.
### 2.2 Self-Similar Profiles for Dark Matter Halos
In 3D, our previous $`\mathrm{P}^3\mathrm{M}`$ simulations of collisionless matter involving $`64^3`$ particles and $`128^3`$ grid cells in a comoving cubic box of side equal to $`\lambda _p`$, with gravitational softening parameter $`\eta =0.3`$ grid spacings demonstrated that the quasi-spherical lumps that form as one of the generic outcomes of pancake gravitational instability in 3D, evolve self-similar density profiles of universal shape (,). The universal profile is well-approximated as a power-law $`\rho r^{2.75}`$ over a large range of density, which flattens somewhat at small radii, a shape which is independent of the details of the initial perturbations to the pancake. This self-similar profile is similar to the universal profile found to fit the results of 3D N-body simulations of the collisionless dark matter in a CDM model . This suggests that this 3D instability of cosmological pancakes which leads generically to the formation of such quasi-spherical dark matter halos may be used as an alternative to the details of the CDM model with its Gaussian random noise initial density fluctuations as a test-bed in which to study halo and galaxy formation further.
To illustrate the universal density profile associated with this pancake instability, we show in Figure 1 results from and for one particular case, $`S_{1,0.2,0.2}`$, for various values of the expansion factor, as well as a summary of results for several different perturbation modes, at $`a/a_c=7`$.
## 3 THE EFFECT OF EXPLOSIVE ENERGY RELEASE ON GALAXY FORMATION: BLOW-AWAY AMIDST CONTINUOUS INFALL
Galaxy formation which leads to star formation can be affected by feedback when stars evolve to the point of supernova (SN) explosions and the resulting shock-heating and outward acceleration of interstellar and intergalactic gas. Previous attempts to model this effect have typically been along one of three lines, that which adopts a smooth initial gas distribution in a galaxy-like, fixed dark matter gravitational potential well (e.g. ), that which considers a single, isolated, but evolving, density fluctuation (i.e. without merging, infall or the effects of external tidal forces) (e.g. ), and that in which the galaxy forms by condensation out of Gaussian random noise primordial density fluctuations such as in the CDM model (e.g. ,,, ,,). In the first case, the computational ability to resolve shocks which propagate away from the sites of explosive energy release is generally greater, while the last is perhaps more realistic in terms of the initial and boundary conditions, but the resolving of shocks is still generally quite poor.
We compromise here between these two limits by using the pancake instability problem as the model of galaxy formation in which to explore the feedback effect of the explosive release of energy by SNe inside a protogalaxy. This affords some of the benefit of greater ease of the first approach mentioned above in resolving the explosion-driven shocks which are the crucial element in blowing gas out and away. It also provides a self-consistent cosmological origin and boundary condition for a protogalaxy or cluster, including the important effect of anisotropic gravitational collapse and continuous infall.
Sharing the initial conditions described above for the formation of a dark matter halo via 3D pancake instability mode $`S_{1,0.2,0.2}`$ is an additional component of baryon-electron gas. We model the explosive release of energy due to SNe in terms of a single impulsive explosion which may represent a starburst or the collective effect of multiple SNe. We initiate the explosion by waiting until the first gas particles at the center of our dark matter halo reach a density contrast relative to the average background density, $`\rho /\rho `$, exceeding $`10^3`$, at which point we suddenly multiply the thermal energy of these particles by a factor $`\chi `$ and share some of the explosion energy smoothly amongst their nearest neighbor particles via ASPH kernel smoothing as well. This occurs at $`a/a_c=2.06`$. While we have performed a series of simulations for different values of $`\chi `$, we shall report here only the results for two limiting cases: $`\chi =0`$ (no explosion) and $`\chi =10^3`$ (blowaway regime). Simulations end at an expansion factor $`a/a_c=3`$. (Note: The proper, numerical prescription for depositing the energy of explosions in the interstellar gas surrounding the explosion site is a complicated question, dependent as it is on explosion details which are unresolvable by current numerical treatments. Some recent discussion of this question is contained in and , including the question of the fraction of the energy of a given SN explosion which ends up as kinetic energy of the SN remnant rather than as thermal energy. However, since the simulations described here are adiabatic and neglect radiative cooling, it is self-consistent for us to deposit the entire explosion energy initially as thermal energy.)
All our gas dynamical simulations are based upon the new 3D version of our ASPH method (,,), coupled to a $`\mathrm{P}^3\mathrm{M}`$ gravity solver. The simulations reported here use $`32^3`$ gas particles, $`32^3`$ dark matter particles (with unequal particle masses, $`m_{\mathrm{dark}}/m_{\mathrm{gas}}=\mathrm{\Omega }_X/\mathrm{\Omega }_B=32.3`$), and a $`\mathrm{P}^3\mathrm{M}`$ grid of $`64^3`$ cells with softening length $`\eta =0.3`$ grid spacings.
## 4 RESULTS
In the absence of explosion, the simulation produces two orthogonal filaments within the pancake central plane at whose intersection is located a denser, quasi-spherical ball of gas which sits in the gravitational potential well of a dense, quasi-spherical dark matter halo like that in Figure 1. In the case with explosion, we found that away from the central object, the pancake and the filaments within it are hardly affected. However, gas has been blown out of the center and some exterior gas which was infalling along directions perpendicular to the pancake plane has been swept back out, as well, some as far as to the outer edge of the box.
In Figure 2, we show a shaded contour plot of the gas temperature at $`a/a_c=3`$, in a plane perpendicular to the pancake and intersecting the center of the central cluster. The explosion is confined by the gas in the plane of the pancake (seen edge-on on this figure) outside the central object, with the hottest gas at the very center. The filaments are hardly affected, however, nor is the shocked pancake gas far from the central object. This edge-on view of the pancake reveals that a major blow-out has occurred in which the explosion, led by an outer shock, has propagated all the way to the edge of the box, half-way to the nearest neighbor pancake, and collided there with the explosion shock expanding away from the neighboring pancakes’ central object and toward the pancake in this box. The temperature plot reveals multiple shocks interior to the explosion, especially along the symmetry axis of the blow-out. Although the central object in which the explosion took place was quasi-spherical, the existence of the pancake plane and of the filaments which intersect at the location of the central dark halo ensure that a highly anisotropic explosion results and serves to channel the energy and mass ejection outward along the symmetry axis.
Velocity arrows for the gas particles at $`a/a_c=3`$ are displayed in Figures 3 and 4. Figure 3 shows a thin slice of the computational volume which contains the pancake central plane (i.e. a top view looking down on the pancake central plane). Figure 4 shows a slice perpendicular to the pancake (that is, the same plane as in Figure 2). These show that outflow is restricted to the symmetry axis, while infall continues within the pancake plane and especially inward along the filaments.
The effect of the explosion in blowing gas away is illustrated by Figure 5, in which different particle groups are distinguished according to their fate with and without the explosion. In the top panel, the solid dots show the particles which were the original recipients of the explosion energy and, by implication, the metal-enriched SN ejecta, which were previously located at the very center of the central dark halo at $`a/a_c=2.06`$. Gas initially outside this core region which was within the halo defined according to a mean overdensity $`\rho /\rho 200`$ but which did not receive the initial explosion energy or ejecta directly is shown as open dots. All of the gas originally inside the halo when the explosion occurred has been blown away by $`a/a_c=3.0`$. The lower panel of Figure 5 shows all those gas particles which were found to be inside the central halo with mean overdensity $`\rho /\rho 200`$ at $`a/a_c=3`$ in the case with no explosion but which are not inside this overdensity at $`a/a_c=3.0`$ in the case with the explosion.
The effect of the explosion on the build-up of the gas mass of the “galaxy” by continuous infall is illustrated by the plot in Figure 6 of the collapsed gas fraction in the box, defined as the gas of overdensity $`\rho /\rho 200`$, for the cases with and without the explosion. With no explosion, the central mass grows continuously from $`a/a_c=2`$ to $`a/a_c=3`$ to encompass 10% of all the mass in the box. With the explosion, however, all the gas in the central halo is blown away shortly after the explosion occurs at $`a/a_c=2.06`$, but by $`a/a_c=2.7`$, unobstructed infall within the pancake plane and along the filaments starts to resupply the central halo with gas at a significant rate.
## 5 SUMMARY
* We find that blow-out and blow-away are generically anisotropic events which channel energy and mass loss outward preferentially along the symmetry axis of the local pancake and away from the intersections of filaments in the pancake plane.
* This means that metal ejection from dwarf galaxies at high redshift due to explosive energy release is less likely to pollute the local filaments and pancake in which the dwarf galaxies reside and more likely to channel the metals away from those denser regions.
* Despite the complete blow-away of gas initially in the dark matter potential well of the “galaxy” by the large explosion simulated here, continuous infall is not completely halted in the directions away from the preferred direction of blow-out, so infall partially replenishes the gas which is blown-away.
## Acknowledgments
This work was supported by NASA Grants NAG5-2785, NAG5-7363, and NAG5-7821, NSF Grants PHY-9800725 and ASC-9504046, and the High Performance Computing Facility, University of Texas.
## References
|
no-problem/9904/solv-int9904013.html
|
ar5iv
|
text
|
# SOLITONS IN A 3d INTEGRABLE MODEL
## 1 Introduction
Recently there was proved the exact integrability of one discrete $`3d`$ model, classical as well as quantum . This model was formulated originally in terms of the pure time evolution as the map of the dynamical variables from time $`t`$ to time $`t+1`$. Namely, for $`t`$ fixed the system of the dynamical variables are the system of
$$[\text{u}_{\alpha ,a,b},\text{w}_{\alpha ,a,b}],\alpha =1,2,3,a,bZ_M,$$
(1)
where $`M`$ is the finite spatial size of the system, and the Poisson brackets in the classical case are
$$\{\text{u}_{\alpha ,a,b},\text{w}_{\alpha ^{},a^{},b^{}}\}=\delta _{\alpha ,\alpha ^{}}\delta _{a,a^{}}\delta _{b,b^{}}\text{u}_{\alpha ,a,b}\text{w}_{\alpha ,a,b}.$$
(2)
The evolution was formulated as the explicit form of functions $`f,g`$,
$$[\text{u},\text{w}]=[\text{u}(t),\text{w}(t)][\text{u}(t+1),\text{w}(t+1)]=[f(\text{u},\text{w}),g(\text{u},\text{w})],$$
(3)
such that the Poisson brackets are conserved by this map. The existence of the complete set of involutive integrals of motion was proved in .
The map $`tt+1`$ may be considered as a sort of the Hamiltonian equations of motion for this classical discrete model. In this paper we will derive their Lagrangian counterpart. Following , one may guess à priori that the Lagrangian variable is Hirota’s tau function and it might obey Hirota’s discrete bilinear equations . But we will show that this is not true in general, although equations we will derive resemble Hirota’s ones.
In this note we recall the reader the Hamiltonian form of the equations of motion first, then introduce the analogous of the tau function and derive the Lagrangian form of the equations of motion, and finally give the multi-soliton solution of them.
## 2 Hamiltonian equations of motion
In this paper we investigate the equations of motion, so we need no the evolution operator and may choose more appropriate coordinate system. Actually, the coordinates we will use here are the light cone frame with respect to previous $`t,a,b`$. Let $`\text{e}_1,\text{e}_2,\text{e}_3`$ be three translations making
$$\begin{array}{ccc}\text{e}_1& :& (t,a,b)(t+1,a,b),\hfill \\ & & \\ \text{e}_2& :& (t,a,b)(t+1,a+1,b),\hfill \\ & & \\ \text{e}_3& :& (t,a,b)(t+1,a,b+1).\hfill \end{array}$$
(4)
$`\text{e}_1,\text{e}_2,\text{e}_3`$ are the elementary orthonormal shifts of the three dimensional cubic lattice. So for any site of the cubic lattice p given, the following eight sites form the elementary cube of the lattice:
$$\left(\begin{array}{cccc}\text{p},& \text{p}+\text{e}_2+\text{e}_3,& \text{p}+\text{e}_1+\text{e}_3,& \text{p}+\text{e}_1+\text{e}_2,\\ & & & \\ \text{p}+\text{e}_1+\text{e}_2+\text{e}_3,& \text{p}+\text{e}_1,& \text{p}+\text{e}_2,& \text{p}+\text{e}_3.\end{array}\right)$$
(5)
Equations of motion for $`\text{u}_{\alpha ,\text{p}}`$, $`\text{w}_{\alpha ,\text{p}}`$ may be extracted from the form of $`f,g`$ in the definition of the evolution in , and in our coordinates look like
(i)
$$\{\begin{array}{ccc}\text{w}_{1,\text{p}+\text{e}_1}& =& \frac{\text{w}_{1,\text{p}}\text{w}_{2,\text{p}}+\text{u}_{3,\text{p}}\text{w}_{2,\text{p}}+\kappa _3\text{u}_{3,\text{p}}\text{w}_{3,\text{p}}}{\text{w}_{3,\text{p}}},\hfill \\ & & \\ \text{u}_{1,\text{p}+\text{e}_1}& =& \frac{\kappa _2\text{u}_{1,\text{p}}\text{u}_{2,\text{p}}\text{w}_{2,\text{p}}}{\kappa _1\text{u}_{1,\text{p}}\text{w}_{2,\text{p}}+\kappa _3\text{u}_{2,\text{p}}\text{w}_{3,\text{p}}+\kappa _1\kappa _3\text{u}_{1,\text{p}}\text{w}_{3,\text{p}}},\hfill \end{array}$$
(6)
(ii)
$$\{\begin{array}{ccc}\text{w}_{2,\text{p}+\text{e}_2}& =& \frac{\text{w}_{1,\text{p}}\text{w}_{2,\text{p}}\text{w}_{3,\text{p}}}{\text{w}_{1,\text{p}}\text{w}_{2,\text{p}}+\text{u}_{3,\text{p}}\text{w}_{2,\text{p}}+\kappa _3\text{u}_{3,\text{p}}\text{w}_{3,\text{p}}},\hfill \\ & & \\ \text{u}_{2,\text{p}+\text{e}_2}& =& \frac{\text{u}_{1,\text{p}}\text{u}_{2,\text{p}}\text{u}_{3,\text{p}}}{\text{u}_{2,\text{p}}\text{u}_{3,\text{p}}+\text{u}_{2,\text{p}}\text{w}_{1,\text{p}}+\kappa _1\text{u}_{1,\text{p}}\text{w}_{1,\text{p}}},\hfill \end{array}$$
(7)
(iii)
$$\{\begin{array}{ccc}\text{w}_{3,\text{p}+\text{e}_3}& =& \frac{\kappa _2\text{u}_{2,\text{p}}\text{w}_{2,\text{p}}\text{w}_{3,\text{p}}}{\kappa _1\text{u}_{1,\text{p}}\text{w}_{2,\text{p}}+\kappa _3\text{u}_{2,\text{p}}\text{w}_{3,\text{p}}+\kappa _1\kappa _3\text{u}_{1,\text{p}}\text{w}_{3,\text{p}}},\hfill \\ & & \\ \text{u}_{3,\text{p}+\text{e}_3}& =& \frac{\text{u}_{2,\text{p}}\text{u}_{3,\text{p}}+\text{u}_{2,\text{p}}\text{w}_{1,\text{p}}+\kappa _1\text{u}_{1,\text{p}}\text{w}_{1,\text{p}}}{\text{u}_{1,\text{p}}}.\hfill \end{array}$$
(8)
Describe in a couple of words the idea of the derivation of equations (6-8). These relations may be obtained as the zero curvature condition of the following system. Let $`\phi _\text{p}`$ be an auxiliary variable assigned to the sites of the cubic lattice. Consider three orthogonal plaquettes of the cube endowed by the following relations
$$\begin{array}{ccc}0=f_{1,\text{p}}& \stackrel{def}{=}& \phi _\text{p}\text{w}_{1,\text{p}}\phi _{\text{p}+\text{e}_2+\text{e}_3}\text{u}_{1,\text{p}}+\phi _{\text{p}+\text{e}_2}+\phi _{\text{p}+\text{e}_3}\kappa _1\text{u}_{1,\text{p}}\text{w}_{1,\text{p}},\\ & & \\ 0=f_{2,\text{p}}& \stackrel{def}{=}& \phi _\text{p}+\phi _{\text{p}+\text{e}_1+\text{e}_3}\kappa _2\text{u}_{2,\text{p}}\text{w}_{2,\text{p}}+\phi _{\text{p}+\text{e}_1}\text{w}_{2,\text{p}}\phi _{\text{p}+\text{e}_3}\text{u}_{2,\text{p}},\\ & & \\ 0=f_{3,\text{p}}& \stackrel{def}{=}& \phi _\text{p}\text{u}_{3,\text{p}}+\phi _{\text{p}+\text{e}_1+\text{e}_2}\text{w}_{3,\text{p}}+\phi _{\text{p}+\text{e}_2}+\phi _{\text{p}+\text{e}_1}\kappa _3\text{u}_{3,\text{p}}\text{w}_{3,\text{p}}.\end{array}$$
(9)
Easy to see that the number of linear equations is three times greater then the number of $`\phi _\text{p}`$. This means that the coefficients of the system of linear equations must obey extra conditions. As an example consider a cube. Six plaquettes give six linear relations for eight corner $`\phi _\text{p}`$:
$$f_{\alpha ,\text{p}}=f_{\alpha ,\text{p}+\text{e}_\alpha }=\mathrm{\hspace{0.33em}0},\alpha =\mathrm{\hspace{0.33em}1},2,3.$$
(10)
Demand that between these six relations there are only four linearly independent, so that the cube relations fix only four of eight corner linear variables $`\phi _\text{p}`$ (e.g. we just may express $`\phi _{\text{p}+\text{e}_1+\text{e}_2}`$, $`\phi _{\text{p}+\text{e}_2+\text{e}_3}`$, $`\phi _{\text{p}+\text{e}_1+\text{e}_3}`$ and $`\phi _{\text{p}+\text{e}_1+\text{e}_2+\text{e}_3}`$ via independent $`\phi _\text{p}`$, $`\phi _{\text{p}+\text{e}_1}`$, $`\phi _{\text{p}+\text{e}_2}`$ and $`\phi _{\text{p}+\text{e}_3}`$ and nothing more). This demand gives equations of motion (6-8) immediately.
## 3 Lagrangian equations of motion
Three simple relations between the equations of motion are to be mentioned:
$$\{\begin{array}{ccc}\text{w}_{1,\text{p}}\text{w}_{2,\text{p}}& =& \text{w}_{1,\text{p}+\text{e}_1}\text{w}_{2,\text{p}+\text{e}_2},\\ & & \\ \text{u}_{2,\text{p}}\text{u}_{3,\text{p}}& =& \text{u}_{2,\text{p}+\text{e}_2}\text{u}_{3,\text{p}+\text{e}_3},\\ & & \\ \text{u}_{1,\text{p}}/\text{w}_{3,\text{p}}& =& \text{u}_{1,\text{p}+\text{e}_1}/\text{w}_{3,\text{p}+\text{e}_3}.\end{array}$$
(11)
Following , a parametrization of $`\text{u},\text{w}`$ in terms of tau functions must turn relations (11) into tautologies. Thus without lost of generality (we deal with the infinite system, avoiding hence all the boundary problems)
$$\{\begin{array}{cc}\text{w}_{1,\text{p}}=\frac{𝒯_{3,\text{p}+\text{e}_2}}{𝒯_{3,\text{p}}},\hfill & \text{u}_{1,\text{p}}=\frac{𝒯_{2,\text{p}}}{𝒯_{2,\text{p}+\text{e}_3}},\hfill \\ & \\ \text{w}_{2,\text{p}}=\frac{𝒯_{3,\text{p}}}{𝒯_{3,\text{p}+\text{e}_1}},\hfill & \text{u}_{2,\text{p}}=\frac{𝒯_{1,\text{p}}}{𝒯_{1,\text{p}+\text{e}_3}},\hfill \\ & \\ \text{w}_{3,\text{p}}=\frac{𝒯_{2,\text{p}}}{𝒯_{2,\text{p}+\text{e}_1}},\hfill & \text{u}_{3,\text{p}}=\frac{𝒯_{1,\text{p}+\text{e}_2}}{𝒯_{1,\text{p}}}.\hfill \end{array}$$
(12)
As usual, $`𝒯_{\alpha ,\text{p}}`$ may contain a quadratic and linear pre-exponent,
$$𝒯_{\alpha ,\text{p}}=\text{e}^{\frac{1}{2}(\text{p},Q_\alpha ,\text{p})+(\text{q}_\alpha ,\text{p})}\tau _{\alpha ,\text{p}},$$
(13)
where $`\text{q}_\alpha `$ and diagonals of $`Q_\alpha `$ are arbitrary, but all non-diagonal elements of $`Q_\alpha `$ must coincide (in the natural basis $`\text{e}_\alpha `$).
Substituting now the parametrization (12) to (6-8) and taking (13) into account, we obtain Three Trilinear relations for Three $`\tau `$ functions:
$$\{\begin{array}{ccc}r_1\tau _{1,\text{p}+\text{e}_2+\text{e}_3}\tau _{2,\text{p}}\tau _{3,\text{p}}& =& \tau _{1,\text{p}}\tau _{2,\text{p}+\text{e}_3}\tau _{3,\text{p}+\text{e}_2}\\ & & \\ & +& s_2\tau _{1,\text{p}+\text{e}_2}\tau _{2,\text{p}+\text{e}_3}\tau _{3,\text{p}}\\ & & \\ & +& s_3^1\tau _{1,\text{p}+\text{e}_3}\tau _{2,\text{p}}\tau _{3,\text{p}+\text{e}_2},\end{array}$$
(14)
$$\{\begin{array}{ccc}r_2\tau _{1,\text{p}}\tau _{2,\text{p}+\text{e}_1+\text{e}_3}\tau _{3,\text{p}}& =& \tau _{1,\text{p}+\text{e}_3}\tau _{2,\text{p}}\tau _{3,\text{p}+\text{e}_1}\hfill \\ & & \\ & +& s_3\tau _{1,\text{p}}\tau _{2,\text{p}+\text{e}_3}\tau _{3,\text{p}+\text{e}_1}\hfill \\ & & \\ & +& s_1^1\tau _{1,\text{p}+\text{e}_3}\tau _{2,\text{p}+\text{e}_1}\tau _{3,\text{p}},\hfill \end{array}$$
(15)
$$\{\begin{array}{ccc}r_3\tau _{1,\text{p}}\tau _{2,\text{p}}\tau _{3,\text{p}+\text{e}_1+\text{e}_2}& =& \tau _{1,\text{p}+\text{e}_2}\tau _{2,\text{p}+\text{e}_1}\tau _{3,\text{p}}\hfill \\ & & \\ & +& s_1\tau _{1,\text{p}+\text{e}_2}\tau _{2,\text{p}}\tau _{3,\text{p}+\text{e}_1}\hfill \\ & & \\ & +& s_2^1\tau _{1,\text{p}}\tau _{2,\text{p}+\text{e}_1}\tau _{3,\text{p}+\text{e}_2}.\hfill \end{array}$$
(16)
Here we gather all $`\kappa `$-s and all the pre-exponents into arbitrary $`s_\alpha ,r_\alpha `$, $`\alpha =1,2,3`$. One may change them in any appropriate way.
Because of the possibility to introduce the Poisson brackets (2) in the system of $`\text{u}_{\alpha ,\text{p}},\text{w}_{\alpha ,\text{p}}`$, the original equations of motion (6-8) may be regarded as a kind of the Hamiltonian equations of motion. Contrary to that, the tautological parametrization of (11) resembles the introduction of a velocity instead of a momentum, so that the number of variables decreases twice, so the trilinear relations (14-16) are nothing but the Lagrangian form of the equations of motion.
(14-16) may be reduced to Hirota’s bilinear discrete equation in the limit
$$\kappa _1<<\kappa _2=\kappa _3<<\mathrm{\hspace{0.33em}1}.$$
(17)
## 4 Solitons
Now describe the simple solitons of the trilinear relations (14-16).
Let $`(\alpha ,\beta ,\gamma )`$ be any cyclic permutation of $`(1,2,3)`$. Fix $`r_\alpha `$ via
$$r_\alpha =\mathrm{\hspace{0.33em}1}+s_\beta +\frac{1}{s_\gamma }.$$
(18)
Then let the exponent
$$W=\text{e}^{i(\text{k},\text{p})}=\text{e}^{ik_1p_1+ik_2p_2+ik_3p_3}$$
(19)
for $`s_\alpha `$ given, is defined by
$$\text{e}^{ik_\alpha }=\frac{\lambda _\alpha ((\lambda _\alpha \lambda _\gamma )+s_\alpha (\lambda _\alpha \lambda _\beta ))}{\lambda _\beta (\lambda _\alpha \lambda _\gamma )+s_\alpha \lambda _\gamma (\lambda _\alpha \lambda _\beta )}.$$
(20)
This parametrization of the dispersion relation we will exhibit as $`W=W(\lambda )`$. For the dispersion curve $`W=W(\lambda )`$ given, let for the shortness
$$W_\alpha ^{(k)}=\lambda _\alpha ^{(k)}W(\lambda ^{(k)}).$$
(21)
The phase shift $`D`$ is defined by
$$D(\lambda ,\lambda ^{})=\frac{d(\lambda ,\lambda ^{})d(\lambda ^1,\lambda ^1)}{d(\lambda ^1,\lambda ^{})d(\lambda ,\lambda ^1)},$$
(22)
where
$$d(\lambda ,\lambda ^{})=det\left(\begin{array}{ccc}1& 1& 1\\ & & \\ \lambda _1& \lambda _2& \lambda _3\\ & & \\ \lambda _1^{}& \lambda _2^{}& \lambda _3^{}\end{array}\right),$$
(23)
and $`\lambda ^1`$ stands for $`\{\lambda _1^1,\lambda _2^1,\lambda _3^1\}`$. For the shortness let
$$D^{(k,l)}=D(\lambda ^{(k)},\lambda ^{(l)}).$$
(24)
One soliton solution is $`\tau _{\alpha ,\text{p}}^{(1)}=\mathrm{\hspace{0.33em}1}W_\alpha `$. Two solitons are
$$\tau _{\alpha ,\text{p}}^{(2)}=\mathrm{\hspace{0.33em}1}W_\alpha ^{(1)}W_\alpha ^{(2)}+D^{(1,2)}W_\alpha ^{(1)}W_\alpha ^{(2)}.$$
(25)
Three solitons are
$$\begin{array}{ccc}\tau _{\alpha ,\text{p}}^{(3)}& =& 1W_\alpha ^{(1)}W_\alpha ^{(2)}W_\alpha ^{(3)}\hfill \\ & & \\ & & +D^{(1,2)}W_\alpha ^{(1)}W_\alpha ^{(2)}+D^{(1,3)}W_\alpha ^{(1)}W_\alpha ^{(3)}+D^{(2,3)}W_\alpha ^{(2)}W_\alpha ^{(3)}\hfill \\ & & \\ & & D^{(1,2)}D^{(1,3)}D^{(2,3)}W_\alpha ^{(1)}W_\alpha ^{(2)}W_\alpha ^{(3)}.\hfill \end{array}$$
(26)
And so on, common expression is usual. Remarkable is that the dispersion relation for the exponents $`\text{e}^{ik_\alpha }`$ may be parametrized by their origin $`\lambda _\alpha `$, (20), and parameters $`s_\alpha `$ are defined by an arbitrary linear pre-exponent.
## 5 Discussion
System (14-16) contains a set of bilinear relations as its necessary conditions. Looking for a solution of trilinear relations in a form of holomorphic functions, we get the following set of necessary relations
$$\mathrm{\Lambda }_{\alpha ,\text{p}+\text{e}_\alpha }\tau _{\alpha ,\text{p}}=\tau _{\beta ,\text{p}+\text{e}_\alpha }\tau _{\gamma ,\text{p}}+s_\alpha \tau _{\beta ,\text{p}}\tau _{\gamma ,\text{p}+\text{e}_\alpha },$$
(27)
where as before $`(\alpha ,\beta ,\gamma )`$ are any cyclic permutation of $`(1,2,3)`$, and auxiliary functions $`\mathrm{\Lambda }_{\alpha ,\text{p}}`$ are supposed to be holomorphic. Thus (14-15) have a host of solutions similar to that of Hirota’s equation. Namely, identifying (27) with a set of Fay’s or Plücker’s relations, we may obtain elliptic or determinant solution of (14-16). Probably, the exception is a Bethe – ansatz – type solution of (27), because in this case (14-16) do not follow from (27) tautologically. Because of all these do not contain anything so surprising as the solitons, we do not give any explicit formula here. Mention only two aspects.
First one is the symmetry group of (14-16). For Hirota’s equation all the symmetries with respect to permutations and reflections of the space directions are obvious. In the case of (14-16) the Cube group is less trivial because of it acts on the parameters $`s_\alpha ,r_\alpha `$. As an example give the realisation of the complete reflection $`P`$. For a solution $`\tau _{\alpha ,\text{p}}`$ of the trilinear equations with some $`s_\alpha ,r_\alpha `$ given let
$$\overline{\tau }_{\alpha ,\text{p}}=\tau _{\alpha ,\text{p}\text{e}_\alpha },\overline{s}_\alpha =\frac{r_\beta }{r_\gamma }\frac{s_\alpha }{s_\beta s_\gamma },\overline{r}_\alpha =\frac{s_\beta }{s_\gamma }\frac{r_\beta r_\gamma }{r_\alpha }.$$
(28)
Then the set of “overlined” objects obey the same set of the trilinear relations (14-16) but with reflected $`\text{e}_\alpha \text{e}_\alpha `$ for all $`\alpha `$.
Mention also the possibility to construct the complete solution of original finite evolution model in terms of the elliptic functions on a finite genus curve. In such case one may consider a non-homogeneous and non-equidistant version of the trilinear relations with a boundary conditions taken into account. The number of the parameters (moduli, initial divisor and homogeneousless data) will coincide with the number of variables of the initial state. This will be the subject of a separate paper.
Acknowledgments. I would like to thank Rinat Kashaev for wery useful explanaions and comments. The work was supported by the RFBR grant No. 98-01-00070.
|
no-problem/9904/hep-ph9904460.html
|
ar5iv
|
text
|
# Theoretical Developments in Inclusive 𝐵 Decays *footnote **footnote *Invited talk at the American Physical Society, Division of Particles and Fields Conference, Jan. 5–9, 1999, Los Angeles, CA.
## I Introduction
In the near future, a large part of the high energy experimental program will be devoted to testing the Cabibbo–Kobayashi–Maskawa (CKM) picture of quark mixing and $`CP`$ violation by directly measuring the sides and (some) angles of the unitarity triangle. If the value of $`\mathrm{sin}(2\beta )`$, the $`CP`$ asymmetry in $`BJ/\psi K_S`$, is not too far from the CDF central value , then searching for new physics at the $`B`$ factories will require a combination of several precision measurements. Particularly important are $`|V_{ub}|`$ and $`|V_{td}|`$, which are the least precisely known elements of the CKM matrix. The latter will be measured hopefully in the upcoming run of the Tevatron from the ratio of $`B_s`$ and $`B_d`$ mixing, and will not be discussed here. This talk is motivated by trying to understand what the chances are to
* …reduce (conservative) error in $`|V_{cb}|`$ below $`5\%`$? Although sometimes a smaller error is quoted already (e.g., by the Particle Data Group), it is hard to bound model independently the possible quark-hadron duality violation in the inclusive, and the size of $`1/m_{c,b}^2`$ corrections in the exclusive determination.
* …determine $`|V_{ub}|`$ with less than $`10\%`$ error? The inclusive measurements require significant cuts on the available phase space; the exclusive measurements require knowledge of the form factors.
* …reduce the theoretical uncertainties in the $`BX_s\gamma `$ photon spectrum? The effect of the experimental cut on the photon energy needs to be better understood, and there are subtleties in the OPE beyond leading order.
The theoretical reliablility of inclusive measurements can be competitive with the exclusive ones (or even better in some cases). For example, for the determination of $`|V_{cb}|`$ model dependence enters at the same order of $`\mathrm{\Lambda }_{\mathrm{QCD}}^2/m^2`$ corrections from both the inclusive semileptonic $`\overline{B}X_ce\overline{\nu }`$ width and the $`\overline{B}D^{}e\overline{\nu }`$ rate near zero recoil.
Inclusive $`B`$ decay rates can be computed model independently in a series in $`\mathrm{\Lambda }_{\mathrm{QCD}}/m_b`$, using an operator product expansion (OPE) . The $`m_b\mathrm{}`$ limit is given by $`b`$ quark decay. For most quantities of interest, this result is known including the order $`\alpha _s`$ and the dominant part of the order $`\alpha _s^2`$ corrections. Observables which do not depend on the four-momentum of the hadronic final state (e.g., total decay rate and lepton spectra) receive no correction at order $`\mathrm{\Lambda }_{\mathrm{QCD}}/m_b`$ when written in terms of $`m_b`$, whereas differential rates with respect to hadronic variables (e.g., hadronic energy and invariant mass spectra) also depend on $`\overline{\mathrm{\Lambda }}/m_b`$, where $`\overline{\mathrm{\Lambda }}`$ is the $`m_Bm_b`$ mass difference in the $`m_b\mathrm{}`$ limit. At order $`\mathrm{\Lambda }_{\mathrm{QCD}}^2/m_b^2`$, the corrections are parameterized by two hadronic matrix elements, usually denoted by $`\lambda _1`$ and $`\lambda _2`$. The value $`\lambda _20.12\mathrm{GeV}^2`$ is known from the $`B^{}B`$ mass splitting.
For inclusive $`bq`$ decay, corrections to the $`m_b\mathrm{}`$ limit are expected to be under control in parts of phase space where several hadronic final states are allowed to contribute with invariant masses satisfying $`m_{X_q}^2m_q^2+(\mathrm{few}\mathrm{times})\mathrm{\Lambda }_{\mathrm{QCD}}m_b`$.However, it is not true that several such states are required to contribute; e.g., $`\overline{B}X_ce\overline{\nu }`$ decay in the small velocity limit can be computed reliably, even though it is saturated by $`D`$ and $`D^{}`$ only . Such observables which “average” sufficiently over different hadronic final states may be predicted reliably. (We are just beginning to learn quantitatively what sufficient averaging is.) The major uncertainty in these predictions is from the values of the quark masses and $`\lambda _1`$, or equivalently, the values of $`\overline{\mathrm{\Lambda }}`$ and $`\lambda _1`$. These quantities can be extracted from heavy meson decay spectra, which is the subject of a large part of this talk.
An important theoretical subtlety is related to the fact that $`\overline{\mathrm{\Lambda }}`$ cannot be defined unambiguously beyond perturbation theory , and its value extracted from data using theoretical expressions valid to different orders in the $`\alpha _s`$ may vary by order $`\mathrm{\Lambda }_{\mathrm{QCD}}`$. However, these ambiguities cancel when one relates consistently physical observables to one another, i.e., the formulae used to determine $`\overline{\mathrm{\Lambda }}`$ and $`\lambda _1`$ contain the same orders in the $`\alpha _s`$ perturbation series as those in which the so extracted values are applied to make predictions.
## II The HQET parameters $`\overline{\mathrm{\Lambda }}`$ and $`\lambda _1`$
The shape of the lepton energy or hadronic invariant mass spectrum in semileptonic $`\overline{B}X_c\mathrm{}\overline{\nu }`$ decay, and the photon spectrum in $`\overline{B}X_s\gamma `$ can be used to measure the heavy quark effective theory (HQET) parameters $`\overline{\mathrm{\Lambda }}`$ and $`\lambda _1`$. Testing our understanding of the $`\overline{B}X_ce\overline{\nu }`$ spectra is important to assess the reliability of the inclusive determination of $`|V_{cb}|`$, and especially that of $`|V_{ub}|`$. Understanding the photon spectrum in $`\overline{B}X_s\gamma `$ is important to evaluate how precisely the total rate can be predicted in the presence of an experimental cut on the photon energy . Studying these distributions is also useful to establish the limitations of models which were built to fit the lepton energy spectrum in semileptonic decay.
The theoretical predictions are known to order $`\alpha _s^2\beta _0`$, where $`\beta _0=112n_f/3`$ is the coefficient of the one-loop $`\beta `$-function. This part of the order $`\alpha _s^2`$ piece usually provides a reliable estimate of the full order $`\alpha _s^2`$ correction, and it is straightforward to compute using the method of Smith and Voloshin . In some cases the full $`\alpha _s^2`$ correction is known. The order $`\mathrm{\Lambda }_{\mathrm{QCD}}^3/m_b^3`$ terms have also been studied , and are used to estimate the uncertainties.
The OPE for semileptonic (or radiative) $`B`$ decay does not reproduce the physical spectrum point-by-point in regions of phase space where the hadronic invariant mass of the final state is restricted. For example, the maximum electron energy for a particular hadronic final state $`X`$ is $`E_e^{(\mathrm{max})}=(m_B^2m_X^2)/2m_B`$, so comparison with experimental data near the endpoint can only be made after sufficient smearing, or after integrating over a large enough region. The minimal size of this region was estimated to be around $`300500`$MeV . The hadron mass spectrum cannot be predicted point-by-point without additional assumptions, but moments of it are calculable. In general, higher moments of a distribution or moments over smaller regions are less reliable than lower moments or moments over larger regions. The strategy is to find observables which are sensitive to $`\overline{\mathrm{\Lambda }}`$ and $`\lambda _1`$, but the deviations from $`b`$ quark decay are small, so that the contributions from operators with dimension greater than 5 are not too important.
### A $`\overline{B}X_ce\overline{\nu }`$ decay spectra
Last year the CLEO Collaboration measured the first two moments of the hadronic invariant mass-squared ($`s_H`$) distribution, $`s_H\overline{m}_D^2`$ and $`(s_H\overline{m}_D^2)^2`$, subject to the constraint $`E_e>1.5`$GeV . Here $`\overline{m}_D=(m_D+3m_D^{})/4`$. Each of these measurements give an allowed band on the $`\overline{\mathrm{\Lambda }}\lambda _1`$ plane. These bands are almost perpendicular, so they give the fairly small intersection region shown in Fig. 1. The central values at order $`\alpha _s`$ are
$$\overline{\mathrm{\Lambda }}=(0.33\pm 0.08)\mathrm{GeV},\lambda _1=(0.13\pm 0.06)\mathrm{GeV}^2.$$
(1)
The unknown order $`\mathrm{\Lambda }_{\mathrm{QCD}}^3/m_b^3`$ corrections not included in this result introduce a large uncertainty, especially for the second moment. As a result, the allowed range is much longer in the direction of the first moment band than perpendicular to it; see Fig. 2.
Similar information on $`\overline{\mathrm{\Lambda }}`$ and $`\lambda _1`$ can be obtained from the lepton energy spectrum, $`d\mathrm{\Gamma }/dE_e`$. It has been measured both by demanding only one charged lepton tag, and using a double tagged data sample where the charge of a high momentum lepton determines whether the other lepton in the event comes directly from semileptonic $`B`$ decay (primary) or from the semileptonic decay of a $`B`$ decay product charmed hadron (secondary). The single tagged data has smaller statistical errors, but it is significantly contaminated by secondary leptons below about $`1.5`$GeV. Therefore, Ref. considered the observables
FIG. 1. Bands in the $`\overline{\mathrm{\Lambda }}\lambda _1`$ plane defined by the measured first and second moments of the hadronic mass-squared and lepton energy distributions. Note that $`M_X^2s_H`$. (From Ref. .)
FIG. 2. Estimates of the theoretical uncertainty in $`\overline{\mathrm{\Lambda }}`$ and $`\lambda _1`$ due to unknown $`1/m_b^3`$ terms from the shape of the electron spectrum (solid ellipse), from hadronic mass moments (long dashed ellipse), and the $`\overline{B}X_s\gamma `$ spectrum (short dashed ellipse; see Sec. IIB). Only the relative size and orientation of the ellipses are meaningful, not their position. (From Refs. and .)
$$R_1=\frac{{\displaystyle _{1.5\mathrm{GeV}}}E_{\mathrm{}}{\displaystyle \frac{d\mathrm{\Gamma }}{dE_{\mathrm{}}}}𝑑E_{\mathrm{}}}{{\displaystyle _{1.5\mathrm{GeV}}}{\displaystyle \frac{d\mathrm{\Gamma }}{dE_{\mathrm{}}}}𝑑E_{\mathrm{}}},R_2=\frac{{\displaystyle _{1.7\mathrm{GeV}}}{\displaystyle \frac{d\mathrm{\Gamma }}{dE_{\mathrm{}}}}𝑑E_{\mathrm{}}}{{\displaystyle _{1.5\mathrm{GeV}}}{\displaystyle \frac{d\mathrm{\Gamma }}{dE_{\mathrm{}}}}𝑑E_{\mathrm{}}}.$$
(2)
Using the CLEO data , the central values $`\overline{\mathrm{\Lambda }}=0.39`$GeV and $`\lambda _1=0.19\mathrm{GeV}^2`$ were obtained , which is in good agreement with Eq. (1).
However, following Ref. , CLEO determined the first two moments of the spectrum, $`E_e`$ and $`(E_eE_e)^2`$, without any restriction on $`E_e`$, using the double tagged data and an extrapolation to $`E_e<0.6`$GeV. The result of this analysis is also plotted in Fig. 1, and yields quite improbable values for $`\overline{\mathrm{\Lambda }}`$ and $`\lambda _1`$. The extrapolation to $`E_e<0.6`$GeV introduces unnecessary model dependence, so this result may be less reliable than the hadronic invariant mass analysis. The extracted values of $`\overline{\mathrm{\Lambda }}`$ and $`\lambda _1`$ are very sensitive to small systematic effects; and it seems problematic for the exclusive models used for the extrapolation to simultaneously reproduce the inclusive lepton spectrum and the $`B`$ semileptonic branching fraction .
There are a number of points to emphasize regarding these results:
1. Taking $`\overline{\mathrm{\Lambda }}`$ from Eq. (1) gives a determination of $`|V_{cb}|`$ from semileptonic $`B`$ width with $`3\%`$ uncertainty.
2. The inclusive $`|V_{cb}|`$ seems to be slightly larger than the exclusive (maybe not “significantly” but “consistently”). Can the theoretical uncertainties be reduced by combining the exclusive and inclusive determinations of $`|V_{cb}|`$?
3. Since the bands from the lepton energy spectrum and the first moment of the hadronic mass-squared spectrum are almost parallel, an independent constraint on the $`\overline{\mathrm{\Lambda }}\lambda _1`$ plane is needed. (See Section II B.)
4. Models do not seem to do well for $`\overline{B}D^{}e\overline{\nu }`$ and $`\overline{B}D^{()}\pi e\overline{\nu }`$ — relying on them may be dangerous. Composition of semileptonic $`B`$ decay seems to be not really understood yet. (See next subsection.)
#### 1 Semileptonic $`B`$ decays to excited charmed mesons
In the heavy quark symmetry limit , the spin and parity of the light degrees of freedom in a heavy meson are conserved. In the charm sector, the ground state is the $`(D,D^{})`$ doublet of heavy quark spin symmetry with spin-parity of the light degrees of freedom $`s_{\mathrm{}}^\pi _{\mathrm{}}=\frac{1}{2}^{}`$. The four lightest excited states, sometimes referred to as $`D^{}`$, are the $`(D_0^{},D_1^{})`$ doublet with $`s_{\mathrm{}}^\pi _{\mathrm{}}=\frac{1}{2}^+`$ and the $`(D_1,D_2^{})`$ doublet with $`s_{\mathrm{}}^\pi _{\mathrm{}}=\frac{3}{2}^{}`$. The $`D_1`$ and $`D_2^{}`$ have been observed with masses near $`2.42`$ and $`2.46`$GeV, respectively, and width around $`20`$MeV. States in the $`s_l^{\pi _l}=\frac{1}{2}^+`$ doublet can decay into $`D^{()}\pi `$ in an $`s`$-wave, and so they are expected to be much broader than the $`D_1`$ and $`D_2^{}`$ which can only decay in a $`d`$-wave. (An $`s`$-wave decay for the $`D_1`$ is forbidden by heavy quark spin symmetry .) The first observation of the $`D_1^{}`$ state with mass and width about $`2.46`$GeV and $`290`$MeV, respectively, was reported at this Conference .
$`\overline{B}D_1e\overline{\nu }`$ and $`\overline{B}D_2^{}e\overline{\nu }`$ account for sizable fractions of semileptonic $`B`$ decays, and are probably the only three-body semileptonic $`B`$ decays (other than $`\overline{B}D^{()}e\overline{\nu }`$) whose differential decay distributions will be precisely measured. ALEPH and CLEO measured recently, with some assumptions, $`(\overline{B}D_1e\overline{\nu }_e)=(6.0\pm 1.1)\times 10^3`$ , while $`\overline{B}D_2^{}e\overline{\nu }_e`$ has not been seen yet. Heavy quark symmetry implies that in the $`m_Q\mathrm{}`$ limit ($`Q=c,b`$), matrix elements of the weak currents between a $`B`$ meson and an excited charmed meson vanish at zero recoil. However, in some cases at order $`\mathrm{\Lambda }_{\mathrm{QCD}}/m_Q`$ these matrix elements are not zero. Since most of the phase space for semileptonic $`B`$ decay to excited charmed mesons is near zero recoil, $`1<vv^{}1.3`$, $`\mathrm{\Lambda }_{\mathrm{QCD}}/m_Q`$ corrections can be very important.
The matrix elements of the weak currents between $`B`$ mesons and $`D_1`$ or $`D_2^{}`$ mesons are conventionally parameterized in terms of a set of eight form factors $`f_i`$ and $`k_i`$ . At zero recoil only $`f_{V_1}`$, defined by
$$D_1(v^{},ϵ)|\overline{c}\gamma ^\mu b|B(v)=\sqrt{m_{D_1}m_B}[f_{V_1}ϵ^\mu +(f_{V_2}v^\mu +f_{V_3}v^\mu )(ϵ^{}v)],$$
(3)
can contribute to the matrix elements. In the $`m_Q\mathrm{}`$ limit, $`f_i`$ and $`k_i`$ are given in terms of a single Isgur-Wise function, $`\tau (w)`$ . Heavy quark symmetry does not fix $`\tau (1)`$, since $`f_{V_1}=(1w^2)\tau (w)+𝒪(1/m_Q)`$, and so $`f_{V_1}(1)=0`$ in the infinite mass limit independent of the value of $`\tau (1)`$.
At order $`1/m_Q`$ several new Isgur-Wise functions occur, together with the parameter $`\overline{\mathrm{\Lambda }}^{}`$, which is the analog of $`\overline{\mathrm{\Lambda }}`$ for the $`(D_1,D_2^{})`$ doublet. $`\overline{\mathrm{\Lambda }}^{}\overline{\mathrm{\Lambda }}0.39`$GeV follows from the measured meson masses . At order $`1/m_Q`$, $`f_{V_1}(1)`$ is no longer zero, but it can be written in terms of $`\overline{\mathrm{\Lambda }}^{}\overline{\mathrm{\Lambda }}`$ and the Isgur-Wise function $`\tau (w)`$ evaluated at zero recoil ,
$$\sqrt{6}f_{V_1}(1)=\frac{4(\overline{\mathrm{\Lambda }}^{}\overline{\mathrm{\Lambda }})}{m_c}\tau (1).$$
(4)
This relation means that at zero recoil heavy quark symmetry gives some model independent information about the $`1/m_Q`$ corrections, similar to Luke’s theorem for the decay into the ground state $`(D,D^{})`$ doublet.
Since the allowed kinematic range for $`\overline{B}D_1e\overline{\nu }_e`$ and $`\overline{B}D_2^{}e\overline{\nu }_e`$ decay are fairly small ($`1<w1.3`$), and there are some constraints on the $`1/m_Q`$ corrections at zero recoil, it is useful to consider the decay rates expanded in powers of $`w1`$. The general structure of the expansion of $`d\mathrm{\Gamma }/dw`$ is elucidated schematically below,
$`{\displaystyle \frac{d\mathrm{\Gamma }_{D_1}^{(\lambda =0)}}{dw}}`$ $``$ $`\sqrt{w^21}\left[(0+0\epsilon +\epsilon ^2+\mathrm{})+(w1)(0+\epsilon +\mathrm{})+(w1)^2(1+\epsilon +\mathrm{})+\mathrm{}\right],`$ (5)
$`{\displaystyle \frac{d\mathrm{\Gamma }_{D_1}^{(|\lambda |=1)}}{dw}}`$ $``$ $`\sqrt{w^21}\left[(0+0\epsilon +\epsilon ^2+\mathrm{})+(w1)(1+\epsilon +\mathrm{})+(w1)^2(1+\epsilon +\mathrm{})+\mathrm{}\right],`$ (6)
$`{\displaystyle \frac{d\mathrm{\Gamma }_{D_2^{}}^{(|\lambda |=0,1)}}{dw}}`$ $``$ $`(w^21)^{3/2}\left[(1+\epsilon +\mathrm{})+(w1)(1+\epsilon +\mathrm{})+\mathrm{}\right].`$ (7)
Here $`\lambda `$ is the helicity of the $`D_1`$ or $`D_2^{}`$, and $`\epsilon ^n`$ denotes a term of order $`(\mathrm{\Lambda }_{\mathrm{QCD}}/m_Q)^n`$. The zeros in Eq. (5) are consequences of heavy quark symmetry.
Table I summarizes some of the more important predictions. The first row ($`B_{\mathrm{}}`$) shows the infinite mass limit. Approximations $`B_1`$ and $`B_2`$ are two different ways of treating unknown $`1/m_Q`$ corrections, and may give an indication of the uncertainty at this order. One of the most interesting predictions is that the $`BD_1`$ decay rate should be larger than $`BD_2^{}`$, contrary to the infinite mass limit, but in agreement with the data. A number of other predictions for different helicity amplitudes, factorization of $`\overline{B}(D_1,D_2^{})\pi e\overline{\nu }`$ decay, sum rules, etc., are presented in Ref. (see also ). The main points are:
1. At zero recoil, order $`1/m_Q`$ contributions to semileptonic $`BD_1,D_2^{}`$ decays (any excited charmed meson with $`+`$ parity) are determined by the $`m_Q\mathrm{}`$ Isgur-Wise function and known hadron mass splittings.
2. Decay spectra can be predicted near zero recoil, including the $`1/m_Q`$ corrections, with reasonable assumptions.
3. Test heavy quark symmetry for $`B`$ decays to excited charmed mesons, where the $`1/m_Q`$ terms are sometimes the leading contributions. Large $`1/m_Q`$ corrections to some predictions can be checked against data.
4. Better understanding of inclusive$`=`$exclusive in semileptonic $`B`$ decay. If semileptonic $`B`$ decays into the four $`D^{}`$ states indeed account for less than 2% of the $`B`$ width, then about another 2% of $`B`$ decays must be semileptonic decays to higher mass excitations or nonresonant channels.
5. Constrain the validity of models constructed to fit the decays to the ground state final state (most predict, for example, larger $`BD_2^{}`$ than $`BD_1`$ semileptonic rate).
### B $`\overline{B}X_s\gamma `$ photon spectrum
Comparison of the measured weak radiative $`\overline{B}X_s\gamma `$ decay rate with theory is an important test of the standard model. In contrast to the decay rate itself, the shape of the photon spectrum is not expected to be sensitive to new physics, but it can nevertheless provide important information. First of all, studying the photon spectrum is important for understanding how precisely the total rate can be predicted in the presence of an experimental cut on the photon energy , which is needed for a model independent interpretation of the resulting decay rate. Secondly, moments of the photon spectrum may be used to measure the HQET parameters $`\overline{\mathrm{\Lambda }}`$ and $`\lambda _1`$ .
An important observable is
$$\overline{(1x_B)}|_{x_B>1\delta }=\frac{{\displaystyle _{1\delta }^1}𝑑x_B(1x_B){\displaystyle \frac{d\mathrm{\Gamma }}{dx_B}}}{{\displaystyle _{1\delta }^1}𝑑x_B{\displaystyle \frac{d\mathrm{\Gamma }}{dx_B}}},$$
(8)
where $`x_B=2E_\gamma /m_B`$ is the rescaled photon energy. The parameter $`\delta =12E_\gamma ^{\mathrm{min}}/m_B`$ corresponds to the experimental lower cut on $`E_\gamma `$, and it has to satisfy $`\delta >\mathrm{\Lambda }_{\mathrm{QCD}}/m_B`$; otherwise nonperturbative effects are not under control. It is straightforward to show that
$$\overline{(1x_B)}|_{x_B>1\delta }=\frac{\overline{\mathrm{\Lambda }}}{m_B}+\left(1\frac{\overline{\mathrm{\Lambda }}}{m_B}\right)_{1\delta }^1𝑑x_b(1x_b)\frac{1}{\mathrm{\Gamma }_0}\frac{d\mathrm{\Gamma }}{dx_b}\frac{\overline{\mathrm{\Lambda }}}{m_B}\delta (1\delta )\frac{1}{\mathrm{\Gamma }_0}\frac{d\mathrm{\Gamma }}{dx_b}|_{x_b=1\delta }+\mathrm{},$$
(9)
where $`x_b=2E_\gamma /m_b`$, and $`d\mathrm{\Gamma }/dx_b`$ is the photon spectrum in $`b`$ quark decay. It has recently been computed away from the endpoint ($`x_b=1`$) to order $`\alpha _s^2\beta _0`$ . $`\mathrm{\Gamma }_0=G_F^2|V_{tb}V_{ts}^{}|^2\alpha _{\mathrm{em}}C_7^2m_b^5/32\pi ^4`$ is the contribution of the tree level matrix element of $`O_7(\overline{s}_L\sigma _{\mu \nu }F^{\mu \nu }b_R)`$ to the $`\overline{B}X_s\gamma `$ decay rate. All terms but the first one on the right-hand-side of Eq. (9) have perturbative expansions which begin at order $`\alpha _s`$. The ellipses denote contributions of order $`(\mathrm{\Lambda }_{\mathrm{QCD}}/m_B)^3`$, $`\alpha _s(\mathrm{\Lambda }_{\mathrm{QCD}}/m_B)^2`$, and $`\alpha _s^2`$ terms not enhanced by $`\beta _0`$, but it does not contain contributions of order $`(\mathrm{\Lambda }_{\mathrm{QCD}}/m_B)^2`$ or additional terms of order $`\alpha _s(\mathrm{\Lambda }_{\mathrm{QCD}}/m_B)`$. Terms in the operator product expansion proportional to $`\lambda _{1,2}/m_b^2`$ enter precisely in the form so that they are absorbed in $`m_B`$ in Eq. (9.
A determination of $`\overline{\mathrm{\Lambda }}`$ is straightforward using Eq. (9). The left hand side is directly measurable, while the quantities entering on the right-hand-side are presented in Ref. . The CLEO data in the region $`E_\gamma >2.1`$GeV yields the central values $`\overline{\mathrm{\Lambda }}_{\alpha _s^2\beta _0}270`$MeV and $`\overline{\mathrm{\Lambda }}_{\alpha _s}390`$MeV. Uncertainties due to the unknown order $`\mathrm{\Lambda }_{\mathrm{QCD}}^3/m_b^3`$ terms in the OPE give the short dashed ellipse in Fig. 2 , whose major axis is roughly perpendicular to those from $`\overline{B}X_ce\overline{\nu }`$. This is why it is important to determine $`\overline{\mathrm{\Lambda }}`$ and $`\lambda _1`$ from both analyses. The potentially most serious uncertainty is from both nonperturbative and perturbative terms that are singular as $`x_b1`$ and sum into a shape function that modifies the spectrum near the endpoint. For sufficiently large $`\delta `$ these effects are not important. They have been estimated in Refs. using phenomenological models. Whether these effects are small in a certain range of $`\delta `$ can be tested experimentally by checking if the extracted value of $`\overline{\mathrm{\Lambda }}`$ is independent of $`\delta `$. This would also improve our confidence that the total decay rate in the region $`x_B>1\delta `$ can be predicted model independently.
There have been other important developments for the $`\overline{B}X_s\gamma `$ decay rate. The next-to-leading order computation of the total rate has been completed , reducing the theoretical uncertainties significantly. It was realized by Voloshin that beyond leading order there are terms in the OPE for the decay rate which are suppressed only by $`\mathrm{\Lambda }_{\mathrm{QCD}}^2/m_c^2`$ instead of $`\mathrm{\Lambda }_{\mathrm{QCD}}^2/m_b^2`$ . In fact, there is a series of contributions of the form $`(\mathrm{\Lambda }_{\mathrm{QCD}}^2/m_c^2)(\mathrm{\Lambda }_{\mathrm{QCD}}m_b/m_c^2)^n`$, which gives for $`n=0`$ a calculable correction of $`\delta \mathrm{\Gamma }/\mathrm{\Gamma }=(C_2/9C_7)(\lambda _2/m_c^2)2.5\%`$. It is also known that there are uncalculable contributions suppressed by $`\alpha _s`$ but not by $`\mathrm{\Lambda }_{\mathrm{QCD}}/m_b`$ from photon coupling to light quarks, for which there is no OPE . While no correction larger than a few percent has been identified, a better understanding of the nonperturbative contributions would be desirable once the four-quark operators are included.
## III $`|V_{ub}|`$ from the $`\overline{B}X_ue\overline{\nu }`$ hadron mass spectrum
The traditional method for extracting $`|V_{ub}|`$ involves a study of the electron energy spectrum in inclusive semileptonic $`B`$ decay. Electrons with energies in the endpoint region $`m_B/2>E_e>(m_B^2m_D^2)/2m_B`$ (in the $`B`$ rest frame, and neglecting the pion mass) must arise from $`bu`$ transition. Since the size of this region is only about $`300`$MeV, at the present time it is not known how to make a model independent prediction for the spectrum in this region. Another possibility for extracting $`|V_{ub}|`$ is based on reconstructing the neutrino momentum. The idea is to reconstruct $`p_{\overline{\nu }}`$ and infer the invariant mass-squared of the hadronic final state, $`s_H=(p_Bq)^2`$, where $`q=p_e+p_{\overline{\nu }}`$. Semileptonic $`B`$ decays satisfying $`s_H<m_D^2`$ must come from $`bu`$ transition . The first analyses of LEP data utilizing this idea have been performed recently .
Both the invariant mass region, $`s_H<m_D^2`$, and the electron endpoint region, $`m_B/2>E_e>(m_B^2m_D^2)/2m_B`$, receive contributions from hadronic final states with invariant masses between $`m_\pi `$ and $`m_D`$. However, for the electron endpoint region the contribution of states with masses nearer to $`m_D`$ is strongly suppressed kinematically. In fact, in the ISGW model the electron endpoint region is dominated by the $`\pi `$ and the $`\rho `$ with higher mass states making a small contribution, and this region includes only of order 10% of the $`\overline{B}X_ue\overline{\nu }`$ rate. The situation is very different for the low invariant mass region, $`s_H<m_D^2`$. Now all states with invariant masses up to $`m_D`$ contribute without any preferential weighting towards the lowest mass ones. In this case the ISGW model suggests the $`\pi `$ and the $`\rho `$ comprise only about a quarter of the $`B`$ semileptonic decays to states with $`s_H<m_D`$, and only of order 10% of the $`\overline{B}X_ue\overline{\nu }`$ rate is excluded from this region. Consequently, it is much more likely that the first few terms in the OPE provide an accurate description of $`B`$ semileptonic decay in the region $`s_H<m_D^2`$ than in the region $`E_e>(m_B^2m_D^2)/2m_B`$. Combining the invariant mass constraint with a modest cut on the electron energy will not destroy this conclusion.
Let us first consider the contribution of dimension three operators in the OPE to the hadronic mass-squared spectrum. This is equivalent to $`b`$ quark decay, and implies a result for $`d\mathrm{\Gamma }/dE_0ds_0`$ (where $`E_0=p_b(p_bq)/m_b`$ and $`s_0=(p_bq)^2`$ are the energy and invariant mass of the strongly interacting partons arising from the $`b`$ quark decay) that is straightforward to calculate to order $`\alpha _s^2\beta _0`$ . Even at this leading order in the OPE there are important nonperturbative effects that come from the difference between $`m_b`$ and $`m_B`$. The most significant effect comes from $`\overline{\mathrm{\Lambda }}`$, and including only it (i.e., neglecting $`\lambda _{1,2}`$), the hadronic invariant mass $`s_H`$ is related to $`s_0`$ and $`E_0`$ by
$$s_H=s_0+2\overline{\mathrm{\Lambda }}E_0+\overline{\mathrm{\Lambda }}^2+\mathrm{}.$$
(10)
Changing variables from $`(s_0,E_0)`$ to $`(s_H,E_0)`$ and integrating $`E_0`$ over the range
$$\sqrt{s_H}\overline{\mathrm{\Lambda }}<E_0<\frac{1}{2m_B}(s_H2\overline{\mathrm{\Lambda }}m_B+m_B^2),$$
(11)
gives $`d\mathrm{\Gamma }/ds_H`$, where $`\overline{\mathrm{\Lambda }}^2<s_H<m_B^2`$. Feynman diagrams with only a $`u`$ quark in the final state contribute at $`s_0=0`$, which corresponds to the region $`\overline{\mathrm{\Lambda }}^2<s_H<\overline{\mathrm{\Lambda }}m_B`$.
Although $`d\mathrm{\Gamma }/ds_H`$ is integrable in perturbation theory, it has a double logarithmic singularity at $`s_H=\overline{\mathrm{\Lambda }}m_B`$. At higher orders in perturbation theory, increasing powers of $`\alpha _s\mathrm{ln}^2[(s_H\overline{\mathrm{\Lambda }}m_B)/m_B^2]`$ appear in the invariant mass spectrum. Therefore, $`d\mathrm{\Gamma }/ds_H`$ in the vicinity of $`s_H=\overline{\mathrm{\Lambda }}m_B`$ is hard to predict reliably even in perturbation theory. (In the region $`s_H\overline{\mathrm{\Lambda }}m_B`$ nonperturbative effects are also important.) The behavior of the spectrum near $`s_H=\overline{\mathrm{\Lambda }}m_B`$ becomes less important for observables that average over larger regions of the spectrum, such as $`d\mathrm{\Gamma }/ds_H`$ integrated over $`s_H<\mathrm{\Delta }^2`$, with $`\mathrm{\Delta }^2`$ significantly greater than $`\overline{\mathrm{\Lambda }}m_B`$. Fig. 3 shows the fraction of $`\overline{B}X_ue\overline{\nu }`$ events in the region $`s_H<\mathrm{\Delta }^2`$ as a function of $`\mathrm{\Delta }^2`$ for three different values of $`\overline{\mathrm{\Lambda }}`$. For a certain value of $`\overline{\mathrm{\Lambda }}`$, Fig. 3 together with Eq. (15) can be used to extract $`|V_{ub}|`$ from data, up to the nonperturbative effects discussed next.
In the low mass region, $`s_H\overline{\mathrm{\Lambda }}m_B`$, nonperturbative corrections from higher dimension operators in the OPE are very important. Just as in the case of the electron endpoint region in semileptonic $`B`$ decay or the photon energy endpoint region in radiative $`B`$ decay, the most singular terms can be identified and summed into a shape function. These shape functions depend on the same infinite set of matrix elements. Since $`\overline{\mathrm{\Lambda }}m_b2\mathrm{GeV}^2`$ is not too far from $`m_D^2`$, it is necessary to estimate the influence of the nonperturbative effects on the fraction of $`B`$ decays with $`s_H<m_D^2`$. It is difficult to estimate this model independently, but upper bounds can be derived on the fraction of $`\overline{B}X_ue\overline{\nu }`$ events with $`s_H>\mathrm{\Delta }^2`$ assuming that the shape function is positive . In the ACCMM model with reasonable parameters, the shape function causes a small (i.e., $`4\%`$ with $`\overline{\mathrm{\Lambda }}=0.4`$GeV, and perturbative QCD corrections neglected) fraction of the events to have $`s_H>m_D^2`$ . This suggests that sensitivity to unknown higher dimension operators in the OPE will probably not give rise to a large uncertainty in $`|V_{ub}|`$ if it is determined from the hadronic invariant mass spectrum in the region $`s_H<m_D^2`$. If experimental resolution forces one to consider a significantly smaller region, then the sensitivity to higher dimension operators increases rapidly.
In summary, to extract $`|V_{ub}|`$ from $`d\mathrm{\Gamma }/ds_H`$ with small theoretical uncertainty, one needs to:
1. move the experimental cut $`\mathrm{\Delta }`$ as close to $`m_D`$ as possible;
2. determine $`\overline{\mathrm{\Lambda }}`$ (at order $`\alpha _s^2`$) with $`50`$MeV uncertainty.
Then a determination of $`|V_{ub}|`$ with $`10\%`$ theoretical uncertainty seems feasible.
## IV Upsilon expansion
The main uncertainties in the theoretical predictions for inclusive $`B`$ decay rates, e.g., $`\overline{B}X_ue\overline{\nu }`$, arise from the $`m_b^5`$ dependence on the $`b`$ quark mass and the bad behavior of the series of perturbative corrections when it is written in terms of the pole mass. In fact, only the product of these quantities is unambiguous, but perturbative multi-loop calculations are most comfortably done in terms of the pole mass. Of course, one would like to eliminate the quark mass altogether from the predictions in favor of a physical observable. Here we present a new method of eliminating $`m_b`$ in terms of the $`\mathrm{{\rm Y}}(1S)`$ meson mass (instead of $`m_B`$ and $`\overline{\mathrm{\Lambda }}`$ discussed in Sec. II).
Let us consider the inclusive $`\overline{B}X_ue\overline{\nu }`$ decay rate . At the scale $`\mu =m_b`$,
$$\mathrm{\Gamma }(BX_ue\overline{\nu })=\frac{G_F^2|V_{ub}|^2}{192\pi ^3}m_b^5\left[12.41\frac{\alpha _s}{\pi }ϵ3.22\frac{\alpha _s^2}{\pi ^2}\beta _0ϵ^25.18\frac{\alpha _s^3}{\pi ^3}\beta _0^2ϵ^3\mathrm{}\frac{9\lambda _2\lambda _1}{2m_b^2}+\mathrm{}\right].$$
(12)
The variable $`ϵ1`$ denotes the order in the modified expansion. The complete order $`\alpha _s^2`$ calculation was done recently , and the result is about 90% of the $`\alpha _s^2\beta _0`$ part. In comparison, the expansion of the $`\mathrm{{\rm Y}}(1S)`$ mass in terms of $`m_b`$ has a different structure,
$$\frac{m_\mathrm{{\rm Y}}}{2m_b}=1\frac{(\alpha _sC_F)^2}{8}\left\{1ϵ+\frac{\alpha _s}{\pi }\left[\left(\mathrm{}+\frac{11}{6}\right)\beta _04\right]ϵ^2+\left(\frac{\alpha _s\beta _0}{2\pi }\right)^2\left(3\mathrm{}^2+9\mathrm{}+2\zeta (3)+\frac{\pi ^2}{6}+\frac{77}{12}\right)ϵ^3+\mathrm{}\right\},$$
(13)
where $`\mathrm{}=\mathrm{ln}[\mu /(m_b\alpha _sC_F)]`$ and $`C_F=4/3`$. In this expansion we assigned to each term one less power of $`ϵ`$ than the power of $`\alpha _s`$, because as we will sketch below, this is the consistent way of combining Eqs. (12) and (13). It is also convenient to choose the same renormalization scale $`\mu `$. The prescription of counting $`[\alpha _s(m_b)]^n`$ in $`B`$ decay rates as order $`ϵ^n`$, and $`[\alpha _s(m_b)]^n`$ in $`m_\mathrm{{\rm Y}}`$ as order $`ϵ^{n1}`$ is called the upsilon expansion. Note that it combines different orders in the $`\alpha _s`$ perturbation series in Eqs. (12) and (13).
The theoretical consistency of the upsilon expansion was shown at large orders for the terms containing the highest possible power of $`\beta _0`$, and to order $`ϵ^2`$ including non-Abelian contributions. An explicit calculation using the Borel transform of the static quark potential shows that the coefficient of the order $`\alpha _s^{n+2}`$ term in Eq. (13) of the form $`(\mathrm{}^n+\mathrm{}^{n1}+\mathrm{}+1)`$ exponentiates to give $`\mathrm{exp}(\mathrm{})=\mu /(m_b\alpha _sC_F)`$, and corrects the mismatch of the power of $`\alpha _s`$ between the two series. This is also needed for the cancellation of the renormalon ambiguities in the energy levels as given by $`2m_b`$ plus the potential and kinetic energies . The infrared sensitivity of Feynman diagrams can be studied by introducing a fictitious infrared cutoff $`\lambda `$. The infrared sensitive terms are nonanalytic in $`\lambda ^2`$, such as $`(\lambda ^2)^{n/2}`$ or $`\lambda ^{2n}\mathrm{ln}\lambda ^2`$, and arise from the low-momentum part of Feynman diagrams. Diagrams which are more infrared sensitive, i.e., have contributions $`(\lambda ^2)^{n/2}`$ or $`\lambda ^{2n}\mathrm{ln}\lambda ^2`$ for small values of $`n`$, are expected to have larger nonperturbative contributions. Linear infrared sensitivity, i.e., terms of order $`\sqrt{\lambda ^2}`$, are a signal of $`\mathrm{\Lambda }_{\mathrm{QCD}}`$ effects, quadratic sensitivity, i.e., terms of order $`\lambda ^2\mathrm{ln}\lambda ^2`$ are a signal of $`\mathrm{\Lambda }_{\mathrm{QCD}}^2`$ effects, etc. From Refs. and follows that the linear infrared sensitivity cancels in the upsilon expansion to order $`ϵ^2`$ (probably to all orders as well, but the demonstration of this appears highly non-trivial).
Substituting Eq. (13) into Eq. (12) and collecting terms of a given order in $`ϵ`$ gives
$$\mathrm{\Gamma }(\overline{B}X_ue\overline{\nu })=\frac{G_F^2|V_{ub}|^2}{192\pi ^3}\left(\frac{m_\mathrm{{\rm Y}}}{2}\right)^5\left[10.115ϵ0.035_{\mathrm{BLM}}ϵ^20.005_{\mathrm{BLM}}ϵ^3\frac{9\lambda _2\lambda _1}{2(m_\mathrm{{\rm Y}}/2)^2}+\mathrm{}\right],$$
(14)
where the BLM subscript indicates that only the corrections proportional to the highest power of $`\beta _0`$ have been kept. The complete order $`ϵ^2`$ term is $`0.041ϵ^2`$ . The perturbation series, $`10.115ϵ0.035_{\mathrm{BLM}}ϵ^20.005_{\mathrm{BLM}}ϵ^3`$, is far better behaved than the series in Eq. (12), $`10.17ϵ0.13_{\mathrm{BLM}}ϵ^20.12_{\mathrm{BLM}}ϵ^3`$, or the series expressed in terms of the $`\overline{\mathrm{MS}}`$ mass, $`1+0.30ϵ+0.19_{\mathrm{BLM}}ϵ^2+0.05_{\mathrm{BLM}}ϵ^3`$. The uncertainty in the decay rate using Eq. (14) is much smaller than that in Eq. (12), both because the perturbation series is better behaved, and because $`m_\mathrm{{\rm Y}}`$ is better known (and better defined) than $`m_b`$. The relation between $`|V_{ub}|`$ and the total semileptonic $`\overline{B}X_ue\overline{\nu }`$ decay rate is
$$|V_{ub}|=(3.06\pm 0.08\pm 0.08)\times 10^3\left(\frac{(\overline{B}X_ue\overline{\nu })}{0.001}\frac{1.6\mathrm{ps}}{\tau _B}\right)^{1/2},$$
(15)
The first error is obtained by assigning an uncertainty in Eq. (14) equal to the value of the $`ϵ^2`$ term and the second is from assuming a $`100`$MeV uncertainty in Eq. (13). The scale dependence of $`|V_{ub}|`$ due to varying $`\mu `$ in the range $`m_b/2<\mu <2m_b`$ is less than 1%. The uncertainty in $`\lambda _1`$ makes a negligible contribution to the total error. Of course, it is unlikely that $`(\overline{B}X_ue\overline{\nu })`$ will be measured without significant experimental cuts, for example, on the hadronic invariant mass (see Sec. III), but this method should reduce the uncertainties in such analyses as well.
The $`\overline{B}X_ce\overline{\nu }`$ decay depends on both $`m_b`$ and $`m_c`$. It is convenient to express the decay rate in terms of $`m_\mathrm{{\rm Y}}`$ and $`\lambda _1`$ instead of $`m_b`$ and $`m_c`$, using Eq. (13) and
$$m_bm_c=\overline{m}_B\overline{m}_D+\left(\frac{\lambda _1}{2\overline{m}_B}\frac{\lambda _1}{2\overline{m}_D}\right)+\mathrm{},$$
(16)
where $`\overline{m}_B=(3m_B^{}+m_B)/4=5.313`$GeV and $`\overline{m}_D=(3m_D^{}+m_D)/4=1.973`$GeV. We then find
$$\mathrm{\Gamma }(\overline{B}X_ce\overline{\nu })=\frac{G_F^2|V_{cb}|^2}{192\pi ^3}\left(\frac{m_\mathrm{{\rm Y}}}{2}\right)^5\mathrm{\hspace{0.17em}0.533}\times \left[10.096ϵ0.029_{\mathrm{BLM}}ϵ^2(0.28\lambda _2+0.12\lambda _1)/\mathrm{GeV}^2\right],$$
(17)
where the phase space factor has also been expanded in $`ϵ`$. For comparison, the perturbation series in this relation when written in terms of the pole mass is $`10.12ϵ0.06ϵ^2\mathrm{}`$ . Equation (17) implies
$$|V_{cb}|=(41.6\pm 0.8\pm 0.7\pm 0.5)\times 10^3\times \eta _{\mathrm{QED}}\left(\frac{(\overline{B}X_ce\overline{\nu })}{0.105}\frac{1.6\mathrm{ps}}{\tau _B}\right)^{1/2},$$
(18)
where $`\eta _{\mathrm{QED}}1.007`$ is the electromagnetic radiative correction. The uncertainties come from assuming an error in Eq. (17) equal to the $`ϵ^2`$ term, a $`0.25\mathrm{GeV}^2`$ error in $`\lambda _1`$, and a $`100`$MeV error in Eq. (13), respectively. The second uncertainty can be removed by determining $`\lambda _1`$, as discussed in Sec. II. Other applications, such as for nonleptonic decays, exclusive semileptonic decays and $`\overline{B}X_s\gamma `$ photon spectrum were studied in Refs. .
The most important uncertainty in this approach is the size of nonperturbative contributions to $`m_\mathrm{{\rm Y}}`$ other than those which can be absorbed into the $`b`$ quark mass. By dimensional analysis the size of this correction is of order $`a^3\mathrm{\Lambda }_{\mathrm{QCD}}^4`$, where $`a1/(m_b\alpha _s)`$ is the Bohr radius of the $`\mathrm{{\rm Y}}`$. Quantitative estimates, however, vary in a large range, and it is preferable to constrain such effects from data. The upsilon expansion yields parameter free predictions for $`\overline{(1x_B)}|_{x_B>1\delta }`$ defined in Eq. (8). The analog of Eq. (9) is
$$\overline{(1x_B)}|_{x_B>1\delta }=1\frac{m_\mathrm{{\rm Y}}}{2m_B}\left[1+0.011ϵ+0.019(ϵ^2)_{\mathrm{BLM}}1x_b|_{x_b>(2m_B/m_\mathrm{{\rm Y}})(1\delta )}\right],$$
(19)
For $`E_\gamma >2.1`$GeV this relation gives 0.111, whereas the central value from the CLEO data is around 0.093. Fig. 4 shows the prediction for $`\overline{(1x_B)}|_{x_B>1\delta }`$ as a function of $`\delta `$, both at order $`ϵ`$ and $`(ϵ^2)_{\mathrm{BLM}}`$. The perturbation expansion is very well behaved. In Eq. (19) nonperturbative contributions to $`m_\mathrm{{\rm Y}}`$ other than those which can be absorbed into the $`b`$ quark mass have been neglected. If the nonperturbative contribution to $`\mathrm{{\rm Y}}`$ mass, $`\mathrm{\Delta }_\mathrm{{\rm Y}}`$, were known, it could be included by replacing $`m_\mathrm{{\rm Y}}`$ by $`m_\mathrm{{\rm Y}}\mathrm{\Delta }_\mathrm{{\rm Y}}`$. For example, $`\mathrm{\Delta }_\mathrm{{\rm Y}}=+100`$MeV increases $`\overline{(1x_B)}`$ by 7%, so measuring $`\overline{(1x_B)}`$ with such accuracy will have important implications for the physics of quarkonia as well as for $`B`$ physics.
## V Conclusions
To conclude, let me emphasize the main points, and indicate what data would be important and useful in my opinion to address them:
* Experimental determination of $`\overline{\mathrm{\Lambda }}`$ and $`\lambda _1`$ from the semileptonic $`\overline{B}X_ce\overline{\nu }`$ lepton energy and hadron mass spectra will reduce the theoretical uncertainties in $`|V_{cb}|`$ and $`|V_{ub}|`$. (Need: Double tagged lepton spectrum with smaller errors.)
* Photon energy spectrum in $`\overline{B}X_s\gamma `$ gives complimentary information on $`\overline{\mathrm{\Lambda }}`$ and $`\lambda _1`$, even in the presence of an experimental cut on the photon energy. (Need: Spectrum with cut on $`E_\gamma `$ lowered; even a few hundred MeV can reduce the uncertainties significantly.)
* To distinguish $`\overline{B}X_ue\overline{\nu }`$ from $`\overline{B}X_ce\overline{\nu }`$, cutting on the hadronic invariant mass is theoretically cleaner than cutting on the lepton energy. (Need: Experimental cut on hadron mass as close to $`m_D`$ as possible, and a precise determination of $`\overline{\mathrm{\Lambda }}`$.)
* The upsilon expansion is equivalent to using a short distance $`b`$ quark mass, but it eliminates $`m_b`$ altogether from the theoretical predictions in favor of $`m_\mathrm{{\rm Y}}`$ in a simple and consistent manner. It raises several interesting theoretical questions, and has many important applications.
###### Acknowledgements.
I thank Adam Falk and Alan Weinstein for inviting me to give this talk, and Christian Bauer for providing Fig. 2. Fermilab is operated by Universities Research Association, Inc., under DOE contract DE-AC02-76CH03000.
|
no-problem/9904/cond-mat9904175.html
|
ar5iv
|
text
|
# Untitled Document
1. Introduction
In recent years, the dynamics of steps on crystal surfaces has attracted much interest, both experimentally \[1-6\] and theoretically \[7-11\]. Experimentally, equilibrium step fluctuations of isolated steps as well as steps on vicinal surfaces have been studied extensively, following pioneering STM measurements of Au(110) and Cu(11n) surfaces. The step fluctuations are quantified by the correlation function $`G(t)`$= $`(h(i,t)h(i,0))^2`$, where $`h(i,t)`$ denotes the position (or displacement) $`h`$ of the step at site $`i`$ and time $`t`$. Typically, the experimental data are described by a power law, $`G(t)t^\gamma `$, with the dynamic exponent $`\gamma `$ being in between about 1/4 and about 1/2. Theoretically, distinct atomic mechanisms driving the step dynamics have been identified, leading, indeed, at long times, to power laws, with $`\gamma `$ being 1/4, 1/3, and 1/2 in the limiting cases of step diffusion, terrace diffusion, and evaporation–condensation kinetics, respectively. These scenarios have been found in Langevin descriptions and confirmed, partly, in simulations on SOS models .
In this article, step fluctuations will be analysed in the framework of random walks of atoms detaching from the (reference) step, diffusing on the neighbouring terrace, and attaching again at the step at distance $`l`$ (in units of the lattice spacing on the surface) from the detachment site. The corresponding return probability distribution $`p(l)`$ is determined for various situations, to analyse the influence of the step roughness, of the presence of another reflecting or adsorbing step separated by $`d`$ lattice spacings from the reference step, of the interaction of the diffusing atom with the steps due to, e.g., elastic forces, as well as of the concentration of reflecting or adsorbing defects on the terrace. The case of a straight step with no obstacles on the terrace and without interactions between the atom and the step has been considered before, solving the related continuum diffusion equation, in the context of a Langevin theory, where $`p(l)`$ is called diffusion kernel . Here, results on $`p(l)`$ will be then incorporated in Monte Carlo simulations of the step fluctuations. We shall present simulational results for special choices of return probabilities connecting detachment and attachment step sites, including the three limiting cases mentioned above. One of the main points will be to elucidate the relation between those probability distributions and the step fluctuations characterised by $`G(t)`$. Indeed, deviations from the simple power law for $`G(t)`$ are found, which may play a relevant role in interpreting experimental observations.
The article is organized as follows. In the next section, the model used to describe step fluctuations is introduced, and the methods applied in the analysis, exact enumeration of random walks and Monte Carlo simulations, are introduced and outlined. Results on the return probability distribution $`p(l)`$ are given in section 3, followed by a discussion of simulational findings on $`G(t)`$. A brief summary will conclude the paper.
2. Model and methods
Let us consider a (reference) surface step of monoatomic height with $`2L+1`$ sites, $`i=L,L+1,\mathrm{},0,\mathrm{}L1,L`$. The positions of the step atoms, at time $`t`$, are denoted by $`h(i,t)`$. Setting the lattice spacing equal to one, $`h`$ is an integer. The step is perfectly straight, if $`h`$ is constant. The neighbouring terrace, consisting of sites $`(i,j)`$ on a square lattice, may be bordered by a straight step opposite to the reference step, separated by $`d`$ lattice spacings, see Fig. 1. The opposite step, with $`2L+1`$ sites, is supposed to either reflect (variant $`s_r`$) or absorb ($`s_a`$) an adatom diffusing on the terrace, mimicing either a descending step with a high Schwoebel–Ehrlich barrier or a rising step with a large sticking coefficient.
Specific boundary conditions may be used at the ends of the reference step, $`i=L`$ and $`i=L`$. For instance, the ends may be pinned or subject to periodic boundary conditions. Alternatively, the boundaries of the terrace perpendicular to the reference and opposite steps may be, e.g., chosen as straight reflecting (variant $`t_r`$) or absorbing ($`t_a`$) straight steps with $`d`$ sites.
Step fluctuations result from the detachment of an atom from the reference step at site $`i`$, $`hh1`$, and attachment at step site $`i+l`$, $`hh+1`$. The rate of this process depends on the energies involved and temperature. One may distinguish three limiting cases, (a) $`l=\pm 1`$, corresponding to step diffusion; (b) uncorrelated detachment and attachment sites, corresponding to evaporation–condensation kinetics; and (c) initial and final sites being connected by a random walk of the diffusing atom on the terrace neighbouring the step, corresponding to terrace diffusion. Several types of terrace diffusion are possible, reflecting various constraints on the random walk, see below. The time dependence of the step fluctuations may be quantified by the correlation function
$$G(t)=(h(i,t)h(i,0))^2$$
(1)
The brackets denote a thermal average.
We applied two methods in our study: (i) exact enumeration of random walks , to investigate different kinds of terrace diffusion, and (ii) Monte Carlo simulations , to compute $`G(t)`$.
(i) In enumerating random walks, we calculate the return probability distribution $`p(l)`$ of an atom detaching from site $`i_0`$ of the reference step onto the terrace (the step positions being then fixed once and for all, $`h_0(i)=h(i,t)`$) and attaching at step site $`i_0+l`$. In other words, the atom starts its random walk at terrace site $`(i_0,h_0(i_0)+2)`$, and one determines the probability that it reaches the site $`(i_0+l,h_0(i_0+l)+1)`$ first among all possible absorption sites at the reference step. The terrace is supposed to be bordered by either reflecting or absorbing steps perpendicular and opposite to the reference step. Obviously, one has
$$\underset{l}{}p(l)1$$
(2)
where the sum runs over all step sites. The identity holds when the boundary steps are reflecting. Usually, the starting point of the random walk is chosen at the center of the step, $`i_0=0`$.
The exact enumeration of $`p(l)`$ is based on calculating the probability $`p_{\mathrm{rw}}(i,j,n)`$ of the diffusing atom to be at terrace site $`(i,j)`$ in the $`n`$-th move of the random walk . The total probability, summing $`p_{\mathrm{rw}}(i,j,n)`$ over all sites, is conserved and equal to one. $`p_{\mathrm{rw}}`$ is related to the probability to jump from the neighbouring site $`(i^{},j^{})`$ to site $`(i,j)`$, $`w(i^{},j^{},i,j)`$, by
$$p_{\mathrm{rw}}(i,j,n)=\underset{i^{}j^{}}{}w(i^{},j^{},i,j)p_{\mathrm{rw}}(i^{},j^{},n1)+w_s(i,j)p_{\mathrm{rw}}(i,j,n1)$$
(3)
where $`w_s(i,j)`$ is the probability to stay at terrace site $`(i,j)`$. For an absorption site, at the reference or a boundary step, one has $`w_s(ij)=1`$ and $`w(i,j,i^{},j^{})`$=0, while for the other terrace sites $`w_s(i,j)=0`$.
On finite terraces, $`p_{\mathrm{rw}}(i,j,n)`$ can be calculated in a straightforward fashion for a variety of jump probabilities and boundary conditions, with the bookkeeping done by a computer programme (exact analytic results are known only for rather few special cases ). The return probability distribution $`p(l)`$ is given by
$$p(l)=p_{\mathrm{rw}}(i_0+l,h_0(i_0+l)+1,n)$$
(4)
considering indefinitely long random walks, $`n\mathrm{}`$. In practice, the length of the walk, $`n`$, depends on the convergence rate of $`p_{\mathrm{rw}}`$. We studied steps of length $`L600`$, with the width of the terrace $`d600`$. Typically, the random walks ended when the total probability of finding the atom on any non-absorption terrace site was smaller than $`10^8`$.
(ii) In our Monte Carlo simulations an atom at step site $`i`$ is moved to site $`i+l`$ with a (return) probability $`p_{\mathrm{da}}(l)`$. The form of $`p_{\mathrm{da}}(l)`$ is motivated by the findings on the random walks, $`p(l)`$. The move is then accepted, as usual, with a rate determined by the Boltzmann factor of the associated energy change $`\delta E`$, $`\mathrm{exp}(\delta E/k_BT)`$, where $`k_B`$ is the Boltzmann constant and $`T`$ the temperature . The energies at the step are assumed to be given by the number and depth of the kinks , as it is the case in the standard SOS model with the Hamiltonian
$$=ϵ\underset{[i,j]}{}|h(i,t)h(j,t)|$$
(5)
where the sum runs over neighbouring step sites $`[i,j]`$, $`j=i\pm 1`$.– To speed up the simulations, we used an algorithm with a dynamic time assignment .
The time $`t`$, elapsed during the simulation, is measured in terms of Monte Carlo attempts per step site pair (MCA), i.e. in one time unit one has tried to interchange, on average, one atom between each two step sites (the Monte Carlo time scale is linearly related to the real time). This interpretation corresponds to the situation where the time spent by the diffusing atom on the terrace is negligibly small compared to the mean time spent at the step. More realistic approaches using, e.g., kinetic Monte Carlo simulations for the surface dynamics, may suffer from other shortcomings, such as an ambiguity in the step position. Note that the average position of the entire step, $`\underset{i}{}h(i,t)/(2L+1)`$, does obviously not depend on time.
To compute the step fluctuations $`G(t)`$, see eq. (1), one has to define a starting time, $`t=0`$. Various choices are possible, corresponding to various initial step configurations, including straight and equilibrated steps. Perhaps closest to experiments, one may average over an ensemble of thermalized step configurations, possibly during a single Monte Carlo run (which will be called, in the following, dynamic averaging). In addition, one may impose different boundary conditions, for instance, by pinning the end positions of the step, at $`i=1`$ and $`2L+1`$, or applying periodic boundary conditions. In the later case, boundary effects are less severe.
Typically, we studied steps of length $`2L+1256`$, with periodic boundary conditions.
3. Random walks
Using the exact enumeration approach, the return probability distribution $`p(l)`$ has been calculated for four distinct cases, mimicing possible constraints on the terrace diffusion: (a) non–perturbed random walk, $`w(i^{},j^{},i,j)=1/4`$, with a straight reference step, see Fig. 2; (b) diffusing atom in an external potential, $`V=A/y^2`$, $`y`$ being the distance between the (straight) steps and the atom, due to, e.g., elastic interactions , see Fig.3. Note that due to the potential $`V`$ the jump probability $`w\mathrm{exp}(\delta V/k_BT)`$, with $`\delta V=V(i^{},j^{})V(i,j)`$, becomes anisotropic, favouring hops away from the step; (c) unbiased diffusion, $`w=1/4`$, on a terrace with quenched absorbing single–site defects of concentration $`c`$, with a straight reference step, see Fig.4; and (d) non–perturbed random walk with a rough reference step (the roughness mimics a thermalized reference step).
In each case, the wandering atom started at the center of the reference step, at terrace site $`(0,h_0(0)+2)`$. The three boundary steps were assumed to be straight and either reflecting or absorbing. For instance, ’$`s_r:t_a`$’ refers to a situation with a reflecting opposite step and absorbing terrace boundary steps perpendicular to it.
In general, one may distinguish four regimes, in which $`p(l)`$ exhibits different characteristic properties, depending on the return distance $`l`$. At very short distances of a few lattice spacings, $`l<l_0`$, $`p(l)`$ falls off rapidly, the concrete form being determined by details of the perturbations (step roughness, concentration of defects and strength $`A`$ of the elastic interaction).
At $`l_0ld`$, $`p(l)`$ acquires typically a power–law behaviour, $`p(l)l^\alpha `$. To analyse that regime, one may calculate the effective exponent
$$\alpha _{\mathrm{eff}}(l)=d\mathrm{ln}p(l)/d\mathrm{ln}l$$
(6)
being constant, $`\alpha _{\mathrm{eff}}=\alpha `$, if, indeed, the deacy of $`p(l)`$ follows a simple power–law. Increasing the return distance $`l`$ furthermore, the effect of the boundaries on the random walk shows up. Assuming $`Ld`$, first the presence of the opposite (vicinal) step affects the return probability, leading to a characteristic exponential decay of $`p(l)`$, $`p\mathrm{exp}(al)`$, both for reflecting, $`s_r`$, and absorbing, $`s_a`$, steps. Finally, the distribution $`p(l)`$ will be modified due to the perpendicular terrace boundaries, $`t_r`$ or $`t_a`$.
As illustrated in Fig. 2 for the non–perturbed case, $`\alpha _{\mathrm{eff}}`$ tends to approach $`\alpha =2`$, in agreement with the solution of the corresponding continuum diffusion equation . Consequently, one expects $`p(l)=p_0l^2`$ at sufficiently large return distances $`l`$ for indefinitely long isolated straight steps (note that $`\alpha `$ is expected to determine the value of the dynamic exponent $`\gamma `$ describing the time dependence of the step fluctuations $`G(t)`$ , see below). Actually, neither the interaction of the diffusing atom with the steps nor the step roughness seem to affect that value of $`\alpha `$. In fact, for a rough reference step, $`\alpha _{\mathrm{eff}}(l)`$ follows closely the form for a straight step, $`l>l_0`$ . Applying the potential $`V=A/y^2`$, we observe a systematic dependence of the proportionality factor $`p_0`$ on A, as shown in Fig. 3. $`p_0`$ is found to increase exponentially with the interaction strength $`A`$, $`p_0\mathrm{exp}(\eta A)`$, with $`\eta 1.23`$, at least for the strengths we considered, $`0A1`$. As depicted in Fig. 4, our findings do not rule out that the value of $`\alpha `$ may, however, depend on the concentration $`c`$ of defects on the terrace. Over an appreciable range of return distances $`l`$ one may notice a plateau–like behaviour in $`\alpha _{\mathrm{eff}}`$ at a value slightly, but definitely smaller than –2. E.g., at $`c=0.2`$, the plateau–like behaviour occurs at about $`2.2`$; at $`c=0.1`$, the effective exponent $`\alpha _{\mathrm{eff}}`$ seems to settle for some distances $`l`$ at a value below, but closer to –2. However, the data do not allow to rule out the possibility that $`\alpha =2`$ may be approached, now from below, at larger return distances $`l`$. Of course, it would be desirable to consider longer steps and larger terraces to study further this aspect. Such computations would be extremely computer time consuming, because one has to average over a large number of defect concentrations to get meaningful data (typically, we averaged over about 10 000 realizations).
With increasing $`l`$, the return probability distribution $`p(l)`$ of the diffusing atom approaches an exponential decay, $`p(l)\mathrm{exp}(gl)`$ in all four cases, $`(a)`$$`(d)`$. Accordingly, the exponential form describes $`p(l)`$ for large return distances $`l`$ in the presence of a neighbouring, descending or rising, step, separated by $`d`$ lattice spacings. For the unbiased random walk with straight steps the coefficient $`g`$, as follows from exact enumeration, agrees, to a high degree of accuracy, with the one obtained from the continuum diffusion equation , namely $`g=\pi /d`$ for an absorbing opposite step, and $`g=\pi /2d`$ for a reflecting, vicinal step. The various perturbations seem to have little effect: The decay form of $`p(l)`$ remains exponential, and even the coefficient $`g`$ seems to be rather robust.– In the unbiased case, $`(a)`$, we found that $`p(l,d)`$ seems to scale, at $`d1`$, as $`p(l,d)=p_s(l/d)/d^2`$, both in the power–law and the exponential regime .
Finally, with $`l`$ getting closer to $`L`$, the perpendicular terrace boundaries become relevant. As expected, $`p(l)`$ is lower when the boundaries are absorbing than in the reflecting situation. However, we did not explore this region in much detail, because those boundary conditions are somewhat artificial and extremely long random walks, see eq. (4), would be needed for quantitatively correct results.
4. Monte Carlo simulations
We simulated the fluctuations $`G(t)`$ of steps of length $`M=2L+1`$, with their ends being connected by periodic boundary conditions. An atom is detached at the randomly chosen site $`i`$ and attached at site $`i+l`$, see Fig. 1. The (return) probability of selecting site $`i+l`$ is denoted by $`p_{\mathrm{da}}(l)`$. The probability of accepting such an elementary move is given by the Boltzmann factor of the energy change, see eq. (5), associated with this process.
We studied mainly five distinct cases, with the choice of $`p_{\mathrm{da}}(l)`$ being motivated by results on the return probability distribution $`p(l)`$ of a random walk and by general physical considerations. In particular, we considered $`p_{\mathrm{da}}(l)`$ being (i) constant, corresponding to evaporation–condensation kinetics. Obviously, the mean step position is conserved in each move. We also simulated evaporation–condensation dynamics by choosing between detachment and attachment completely randomly, thus conserving the mean step position only in the time average; (ii) 1/2 for $`l=\pm 1`$ and 0 elsewhere, describing step diffusion; (iii) proportional to $`1/l^2`$ (the prefactor is determined by requiring that $`p_{\mathrm{da}}(l)`$ is normalized: $`p_{\mathrm{da}}(l)=1`$), as one may expect for fluctuations of an isolated step due to terrace diffusion, neglecting deviations from that form at short distances $`l`$ between attachment and detachment sites; (iv) proportional to $`\mathrm{exp}(l)`$, monitoring the possible effect of a neighbouring step on $`G(t)`$; and (v) proportional to $`1/l^3`$, to study the sensitivity of $`G(t)`$ on the value of $`\alpha `$ in $`p_{\mathrm{da}}(l)l^\alpha `$.
For indefinitely long steps, one may argue that $`G(t)`$ increases with time in the form of a power–law $`G(t)t^\gamma `$ \[7-10\], with the dynamic exponent $`\gamma `$ depending on the atomistic mechanism governing the fluctuations. To determine $`\gamma `$ in the various cases we simulated, one may calculate the effective dynamic exponent
$$\gamma _{\mathrm{eff}}(t)=d\mathrm{ln}G(t)/d\mathrm{ln}t$$
(7)
with $`\gamma _{\mathrm{eff}}(t)`$ approaching $`\gamma `$ at large times and long steps.
Results on $`\gamma _{\mathrm{eff}}(t)`$ for each of the five situations are depicted in Fig. 5, at fixed step length, $`M=256`$, and fixed temperature $`k_BT/ϵ=1`$. We used dynamic averaging over an ensemble of successive, equilibrated initial configurations. Evaporation–condensation kinetics was simulated by random attachment and detachment processes, i.e. without conserving the average step position in each Monte Carlo move.
At very early times, the step fluctuations are always diffusive, with $`G(t)=c_dt`$, corresponding to $`\gamma _{\mathrm{eff}}=1`$. The diffusion coefficient $`c_d`$ depends rather weakly on the transport mechanism, with the exception of the evaporation–condensation dynamics having a significantly larger diffusion coefficient.
Due to the rigidity of the step, the step meandering then slows down, with $`\gamma _{\mathrm{eff}}`$ approaching a plateau located at about 1/2, 1/3, or 1/4, depending on the form of the return probability, $`p_{\mathrm{da}}`$, see Fig. 5. Actually, the plateau at $`\gamma _{\mathrm{eff}}1/2`$, characterising evaporation–condensation, is reached most quickly. The plateau at 1/3 is realised for terrace diffusion with an isolated step, $`p_{\mathrm{da}}(l)1/l^2`$. The largest time is needed to approach the plateau at 1/4, characterising step diffusion as well as terrace diffusion with rapidly decaying return probabilities, $`p_{\mathrm{da}}(l)1/l^3`$ and $`\mathrm{exp}(l)`$. The last situation corresponds to terrace diffusion with pairs of steps. One may argue that the plateau signals the asymptotic power law increase of $`G(t)t^\gamma `$ for large times and indefinitely long steps. Indeed, the plateau values have been obtained before in Langevin descriptions \[7-9\] for infinitely extended steps (for $`3<\alpha <2`$, one may expect a continuously varying value, $`\gamma =1/(\alpha +1)`$ ).
Combining our results on the random walks and simulations, we conclude that the exponent $`\gamma =1/3`$ for an isolated step is robust against various perturbations of terrace diffusion, with the possible exception of defects on the terrace.
The estimates of the plateau values of $`\gamma _{\mathrm{eff}}`$ are confirmed by a Fourier analysis of the step fluctuations $`h(i,t)`$. They can be written in the form
$$h(i,t)=\underset{k=1}{\overset{M/2}{}}(a_k(t)\mathrm{sin}(2\pi ki/M)+b_k\mathrm{cos}(2\pi ki/M))$$
(8)
The fluctuation modes may be described by
$$G_k(t)=(a_k(t)a_k(0))^2+(b_k(t)b_k(0))^2$$
(9)
taking into account phase shifts. Based on Langevin descriptions, $`G_k(t)`$ is expected to converge rapidly towards equilibrium $`G_k(\mathrm{})`$ \[7-9\],
$$G_k(t)=G_k(\mathrm{})(1\mathrm{exp}(I_kt))$$
(10)
in agreement with the simulational data. More specificly, for small wavenumbers, $`I_k`$ follows closely the form $`I_kk^x`$, with $`x=1/\gamma `$, in accordance with the Langevin theory. However, $`I_k`$ shows pronounced deviations from the power–law behaviour at larger values of $`k`$, as shown in Fig. 6. This feature is not described by the Langevin theory .
At later times, after having passed the plateau, the effective dynamic exponent, $`\gamma _{\mathrm{eff}}`$, will eventually go to zero (as may be easily seen for shorter steps ) or, in the case of evaporation–condensation, to one, due to the constraint of conserving, or not, the average step position, see Fig. 5. For evaporation–condensation, the related diffusion coefficient is rather small and depends on the step length. In any event, the late–time behaviour reflects the finite length of the step.
The time dependence of $`\gamma _{\mathrm{eff}}`$, as depicted in Fig. 5, implies that an average effective exponent, $`\gamma _a`$, as obtained from a log–log plot of $`G(t)`$ in a given time range, may vary with experimental parameters when one is not in the truely asymptotic regime. In particular, the plateau in $`\gamma _{\mathrm{eff}}`$ may not yet have been reached or finite size effects may already matter. Note that, for instance, the crossover from the diffusive short–time behaviour to the subdiffusive step motion at later times is temperature dependent, thereby causing possibly a temperature dependence in $`\gamma _a`$. Similarly, the presence of a neighbouring step is expected to affect $`\gamma _a`$. In general, much care is needed in identifying and disentangling the numerous possible crossover effects.
5. Summary
We studied equilibrium step fluctuations in a somewhat idealized way by enumerating random walks on terraces with various constraints and using Monte Carlo techniques.
In the simulations, different atomistic mechanisms determining the equilibrium step fluctuations may be mimiced by special choices of the return probability for atoms attaching at distance $`l`$ from the detachment site of the step. The form of the return probability at large distances $`l`$ determines the time dependence of the step fluctuations $`G(t)`$ at late times $`t`$ for long steps. In particular, we simulated evaporation–condensation kinetics, step diffusion and (perfect) terrace diffusion. For $`\mathrm{𝑖𝑠𝑜𝑙𝑎𝑡𝑒𝑑}`$ steps, $`G(t)`$ is confirmed to increase with a power law $`G(t)t^\gamma `$, with $`\gamma `$ =1/2, 1/3, and 1/4, respectively. In the case of terrace diffusion for $`\mathrm{𝑝𝑎𝑖𝑟𝑠}`$ of steps, the dynamic exponent $`\gamma `$ approaches the value of step diffusion, $`\gamma =1/4`$. The simulational observations agree with and refine predictions of Langevin theory.– Similarly, a Fourier analysis of the step fluctuations, driven by the different atomic mechanisms, confirms and refines previous Langevin descriptions.
In general, crossover phenomena may mask the asymptotic behaviour of $`G(t)`$. Such phenomena can be caused by several reasons, including change from diffusive to subdiffusive step motions at early times, effect of the finite step length, defects on the terrace, and influence of neighbouring steps. As a result, the average effective dynamic exponent $`\gamma _a`$, as usually obtained from measurements, may vary with experimental parameters like temperature.
From the exact enumeration of random walks, the decay of the return probability distribution $`p(l)l^2`$ for unbiased terrace diffusion with straight steps is found to be robust against step roughness and (elastic) interactions between the diffusing atom and the neighbouring step. This fact implies the robustness of the value $`\gamma =1/3`$ for isolated steps. Moreover, for pairs of steps, the exponential decay of $`p(l)`$ persists in the presence of those perturbations, implying the robustness of $`\gamma =1/4`$ in that case.
Acknowledgement
We should like to thank H. P. Bonzel, T. L. Einstein, H. Ibach, and M. Giesen for useful discussions and remarks.
References
1. L. Kuipers, M. S. Hoogeman, and J. W. M. Frenken, Phys. Rev. Lett. 71 (1993) 3517.
2. M. Giesen-Seibert, R. Jentjes, M. Poensgen, and H. Ibach, Phys. Rev. Lett. 71 (1993) 3521.
3. M. Giesen, G. S. Schulze Icking–Konert, D. Stapel, and H. Ibach, Surf. Sci. 366 (1996) 229.
4. W. W. Pai, N. C. Bartelt, and J. E. Reutt–Robey, Phys. Rev. B 53 (1996) 15991.
5. P. Wang, H. Pfnür, S. V. Khare, T. L. Einstein, E. D. Williams, W. W. Pai, and J. E. Reutt–Robey, Phys. Rev. B 53 (1996) 15991.
6. S. Tanaka, N. C. Bartelt, C. C. Umbach, R. M. Tromp, and J. M. Blakely, Phys. Rev. Lett. 78 (1997) 3342.
7. N. C. Bartelt, T. L. Einstein, and E. D. Williams, Surf. Sci. 312 (1994) 411.
8. A. Pimpinelli, J. Villain, D. E. Wolf, J. J. Metois, J. C. Heyraud, I. Elkinani, and G. Uimin, Surf. Sci. 295 (1993) 143.
9. B. Blagojevic and P. M. Duxbury, in : Dynamics of Crystal Surfaces and Interfaces, Eds. P. M. Duxbury and T. Pence (Plenum, New York, 1997) p.1.
10. S. V. Khare and T. L. Einstein, Phys Rev. B57 (1998) 4782.
11. W. Selke and M. Bisani, Lecture Notes in Physics 519 (1999) 298.
12. I. Majid, D. Ben-Avraham, S. Havlin, and H. E. Stanley, Phys Rev. B30 (1984) 1626.
13. K. Binder and D. W. Heermann, Monte Carlo Simulations in Statistical Physics (Springer, Heidelberg, 1992).
14. M. Bisani, Diploma thesis, RWTH Aachen (1998).
15. W. Feller, An Introduction to Probability Theory and its Applications (John Wiley, New York, 1968).
16. B. S. Swartzentruber, Y.-W. Mo, R. Kariotis, M. G. Lagally, and M. B. Webb, Phys. Rev. Lett. 65 (1990) 1913.
17. J. D. Weeks, in : Ordering in Strongly Fluctuating Condensed Matter Systems, Ed. T. Riste (Plenum, New York, 1980) p.293.
18. A. B. Bortz, M. H. Kalos, and J. L. Lebowitz, J. Comp. Phys. 17 (1975) 10.
19. V. I. Marchenko and A. Y. Parshin, Sov. Phys. JETP 52 (1980) 129.
20. W. H. Press, S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery, Numerical Recipes in C (Cambridge University Press, Cambridge, 1992).
Figure Captions
* Geometry of the surface, showing the reference step and a descending ($`s_r`$) or rising ($`s_a`$) opposite step.
* Effective exponent $`\alpha _{\mathrm{eff}}`$ of the return probability distribution $`p(l)`$ for the unbiased random walk with a straight reference step ($`L=600`$, $`d=600`$). Absorbing terrace boundaries, $`t_a`$, were used.
* Return probability distribution $`p(l)`$ for random walks in a repulsive external potential $`V=Ay^2`$ at $`k_BT=1`$, for $`L=600`$, $`d=600`$, and with $`s_a:t_a`$ boundary conditions.
* Effective exponent $`\alpha _{\mathrm{eff}}`$ of $`p(l)`$ for random walks with defects on the terrace ($`d=50`$, $`L=100`$, concentration of defects $`c=0.2`$, averaged over 8778 realisations, with $`s_a:t_a`$ boundary conditions). The dashed line is based on smoothed data obtained by applying a Savitzky-Golay-filter .
* Effective dynamic exponent $`\gamma _{\mathrm{eff}}`$ as a function of time (measured in units of MCA) for various atomistic mechanisms and return probabilities. From top to bottom: evaporation–condensation (solid), $`p_{\mathrm{da}}(l)l^2`$ (solid), $`l^3`$ (solid), $`e^l`$ (dotted), $`=1/2`$ for $`l=\pm 1`$ (dashed). Data was smoothed using a Savitzky-Golay-filter. Steps with 256 sites, at $`k_BT/ϵ=1`$ were simulated.
* $`I_k`$ vs. $`k/M`$ (in units of $`2\pi `$) for different transport mechanisms: evaporation–condensation (plus), terrace diffusion with $`p_{\mathrm{da}}(l)l^2`$ (cross) and step diffusion with $`p_{\mathrm{da}}(l)=1/2`$ for $`l=\pm 1`$ (asterik). The lines correspond to the power–law behaviour expected from Langevin theory: $`I_kk^2`$ (solid), $`I_kk^3`$ (dashed) and $`I_kk^4`$ (dotted), respectively. Steps with 64 sites, at $`k_BT/ϵ=0.8`$, were simulated.
|
no-problem/9904/cond-mat9904150.html
|
ar5iv
|
text
|
# Single hole dynamics in the one dimensional 𝑡-𝐽 model
## Abstract
We present a new finite-temperature quantum Monte Carlo algorithm to compute imaginary-time Green functions for a single hole in the $`t`$-$`J`$ model on non-frustrated lattices. Spectral functions are then obtained with the Maximum Entropy method. Simulations of the one-dimensional case show that a simple charge-spin separation Ansatz is able to describe the overall features of the spectral function over the whole energy range for values of $`J/t`$ from $`1/3`$ to $`4`$. This includes the bandwidth $`W4t+J`$ and the compact support of the spectral function. The quasiparticle weight $`Z_k`$ is computed on lattices up to $`L=96`$ sites, and scales as $`Z_kL^{1/2}`$.
Understanding single hole dynamics in quantum antiferromagnets is a decisive step towards a comprehensive description of elementary excitations in strongly correlated systems. Experimental realizations are found in compounds such as $`SrCuO_2`$ , $`Na_{0.96}V_2O_5`$ for chains, $`Sr_{14}Cu_{24}O_{41}`$ for ladders and $`Sr_2CuO_2Cl_2`$ for planes. In particular chain compounds attract at present an increasing amount of interest in order to elucidate, whether signals of charge-spin separation as predicted from Luttinger-liquid theory can be observed experimentally. On the other hand, theoretical treatments based on Bethe-Ansatz (BA) results lead recently to a complete description of the spectral function of the Hubbard model at $`U=\mathrm{}`$ and the low energy sector in the nearest-neighbour (NN) $`t`$-$`J`$ model, where explicit results are obtained at the supersymmetric (SuSy) point . Further exact results - apart from exact diagonalizations which suffer from strong finite-size effects - are available only for the inverse-square exchange (ISE) $`t`$-$`J`$ model at the SuSy point. In order to be able to compare with experiments, it is crucial to extend such studies to realistic values of the parameters and possibly beyond the asymptotic low energy limit.
In this letter, we present a simple finite-temperature quantum Monte Carlo (QMC) algorithm capable of dealing with this issue for the NN $`t`$-$`J`$ model. For single-hole excitations and in the absence of frustration, the method is free of the notorious sign problem, and applicable to chains, n-leg ladders and planes. Here, we concentrate on chains. Our simulations lead to the conclusion that the overall features of the spectral functions are well described by a charge-spin separation Ansatz (CSSA) based on a mean-field slave-boson picture , where the hole spectral function is given by the convolution of the spectral functions of free holons and spinons. The agreement with the simulations is obtained over all energy scales and values of $`J/t`$ ranging from $`1/3`$ to $`4`$. At the SuSy point a more detailed understanding of the spectrum is achieved by supplementing the simple model with BA results. A finite-size scaling on chains up to $`L=96`$ sites shows that the quasi-particle weight $`Z_k`$ vanishes as $`1/\sqrt{L}`$, a result which was beyond numerical capabilities up to now.
Our starting point is the NN $`t`$-$`J`$ model,
$`H_{tJ}=t{\displaystyle \underset{<i,j>,\sigma }{}}\stackrel{~}{c}_{i,\sigma }^{}\stackrel{~}{c}_{j,\sigma }+J{\displaystyle \underset{<i,j>}{}}\left(\stackrel{}{S}_i\stackrel{}{S}_j{\displaystyle \frac{1}{4}}\stackrel{~}{n}_i\stackrel{~}{n}_j\right).`$ (1)
Here $`\stackrel{~}{c}_{i,\sigma }^{}`$ are projected fermion operators $`\stackrel{~}{c}_{i,\sigma }^{}=(1c_{i,\sigma }^{}c_{i,\sigma })c_{i,\sigma }^{}`$ , $`\stackrel{~}{n}_i=\underset{\alpha }{}\stackrel{~}{c}_{i,\alpha }^{}\stackrel{~}{c}_{i,\alpha }`$, $`\stackrel{}{S}_i=(1/2)\underset{\alpha ,\beta }{}c_{i,\alpha }^{}\stackrel{}{\sigma }_{\alpha ,\beta }c_{i,\beta }`$, and the sum runs over nearest neighbours. After a canonical transformation this model is cast into the form
$`\stackrel{~}{H}_{tJ}=+t{\displaystyle \underset{<i,j>}{}}P_{ij}f_i^{}f_j+{\displaystyle \frac{J}{2}}{\displaystyle \underset{<i,j>}{}}\mathrm{\Delta }_{ij}(P_{ij}1),`$ (2)
where $`P_{ij}=(1+\stackrel{}{\sigma }_i\stackrel{}{\sigma }_j)/2`$, $`\mathrm{\Delta }_{ij}=(1n_in_j)`$ and $`n_i=f_i^{}f_i`$. In this mapping, one uses the following identities for the standard creation ($`c_{i,\sigma }^{}`$) and annihilation ($`c_{i,\sigma }`$) operators $`c_i^{}=\gamma _{i,+}f_i\gamma _{i,}f_i^{},c_i^{}=\sigma _{i,}(f_i+f_i^{}),`$ where $`\gamma _{i,\pm }=(1\pm \sigma _{i,z})/2`$ and $`\sigma _{i,\pm }=(\sigma _{i,x}\pm i\sigma _{i,y})/2`$. The spinless fermion operators fulfill the canonical anticommutation relations $`\{f_i^{},f_j\}=\delta _{i,j}`$, and $`\sigma _{i,a},a=x,y,`$ or $`z`$ are the Pauli matrices. The constraint to avoid doubly occupied states transforms to the conserved and holonomic constraint $`_i\gamma _{i,}f_i^{}f_i=0`$.
The Green function in the spin up sector may be written as
$`G_{}(ij,\tau )=T\stackrel{~}{c}_{i,}(\tau )\stackrel{~}{c}_{j,}^{}=Tf_i^{}(\tau )f_j`$ (3)
where $`T`$ corresponds to the time ordering operator. Inserting complete sets of spin states the quantity above transforms as
$`G(ij,\tau )=`$ $`{\displaystyle \frac{\underset{\sigma _1}{}v|\sigma _1|e^{(\beta \tau )\stackrel{~}{H}_{tJ}}f_je^{\tau \stackrel{~}{H}_{tJ}}f_i^{}|\sigma _1|v}{\underset{\sigma _1}{}\sigma _1|e^{\beta \stackrel{~}{H}_{tJ}}|\sigma _1}}=`$ (6)
$`{\displaystyle \underset{\stackrel{}{\sigma }}{}}P(\stackrel{}{\sigma })\times {\displaystyle \frac{v|f_je^{\mathrm{\Delta }\tau \stackrel{~}{H}(\sigma _n,\sigma _{n1})}e^{\mathrm{\Delta }\tau \stackrel{~}{H}(\sigma _{n1},\sigma _{n2})}\mathrm{}e^{\mathrm{\Delta }\tau \stackrel{~}{H}(\sigma _2,\sigma _1)}f_i^{}|v}{\sigma _n|e^{\mathrm{\Delta }\tau \stackrel{~}{H}_{tJ}}|\sigma _{n1}\mathrm{}\sigma _2e^{\mathrm{\Delta }\tau \stackrel{~}{H}_{tJ}}|\sigma _1}}+𝒪(\mathrm{\Delta }\tau ^2)`$
$`={\displaystyle \underset{\stackrel{}{\sigma }}{}}P(\stackrel{}{\sigma })G(i,j,\tau ,\stackrel{}{\sigma })+𝒪(\mathrm{\Delta }\tau ^2)`$
Here $`m\mathrm{\Delta }\tau =\beta `$, $`n\mathrm{\Delta }\tau =\tau `$, $`\mathrm{\Delta }\tau t1`$ and $`\mathrm{exp}(\mathrm{\Delta }\tau \stackrel{~}{H}(\sigma _1,\sigma _2))`$ is the evolution operator for the holes, given the spin configuration $`(\sigma _1,\sigma _2)`$. In the case of single hole dynamics $`|v`$ is the vacuum state for holes, and $`P(\stackrel{}{\sigma })`$ is the probability distribution of a Heisenberg antiferromagnet for the configuration $`\stackrel{}{\sigma }`$, where $`\stackrel{}{\sigma }`$ is a vector containing all intermediate states $`(\sigma _1,\mathrm{}\sigma _n,\mathrm{}\sigma _m,\sigma _1)`$. The sum over spins is performed in a very efficient way by using a world-line cluster-algorithm . As the evolution operator for the holes is a bilinear form in the fermion operators, $`G(x,\tau ,\stackrel{}{\sigma })`$ can be calculated exactly. $`G(x,\tau ,\stackrel{}{\sigma })`$ contains a sum over all possible fermion paths between $`(i,0)`$ and $`(j,\tau )`$, where $`ij=x`$. This stands in contrast to the worm approach , where fermion paths are sampled stochastically. The numerical effort to calculate $`G(x,\tau ,\stackrel{}{\sigma })x,\tau `$ scales as $`L\tau `$. Spectral properties are obtained by inverting the spectral theorem
$`G(k,\tau )={\displaystyle \underset{\mathrm{}}{\overset{\mathrm{}}{}}}d\omega A(k,\omega ){\displaystyle \frac{\mathrm{exp}(\tau \omega )}{\pi (1+\mathrm{exp}(\beta \omega ))}}`$ (7)
with the Maximum Entropy method (MEM) . Since $`P(\stackrel{}{\sigma })`$ is the probability distribution for the quantum antiferromagnet, the algorithm does not suffer from sign problems on bipartite lattices and next neighbour interactions in any dimension. However, when the spin and charge dynamics evolve according to very different time scales ($`J<0.2t`$), $`G(x,\tau ,\stackrel{}{\sigma })`$ shows an increasing variance. Best results are obtained at the SuSy point and an appreciable range of $`J/t`$ may be considered as shown below.
We now concentrate on the one-dimensional $`t`$-$`J`$ model. The simulations were performed at temperatures $`T\mathrm{min}(J,t)/15`$, such that no appreciable changes with a further decrease in temperature can be seen: the results correspond to the zero temperature limit, a limit which is in general difficult to reach in other finite-temperature fermionic algorithms. We compare our results with the predictions of the CSSA, where free holons and spinons are described by
$`H={\displaystyle \frac{t_h}{2}}{\displaystyle \underset{<i,j>}{}}h_i^{}h_j{\displaystyle \frac{J_s}{2}}{\displaystyle \underset{<i,j>}{}}s_{i,\sigma }^{}s_{j,\sigma }.`$ (8)
Here the electron operator $`c_{i,\sigma }`$ is given by the product of a holon ($`h_i`$) and a spinon ($`s_{i,\sigma }`$) operator, $`c_{i\sigma }=s_{i,\sigma }h_i^{}`$, the holon being a boson and the spinon a spin-1/2 fermion. As a consequence of the above Ansatz , the dispersion relations of the free holons and spinons are given by $`ϵ_h=t_h\mathrm{cos}q_h`$ and $`ϵ_s=J_s\mathrm{cos}q_s`$ respectively, whereas the energy of the hole is $`E(k)=ϵ_hϵ_s`$ and by momentum conservation $`k=q_hq_s`$. We take $`t_h`$ and $`J_s`$ as two free parameters in contrast to a mean-field approximation, where they have to be calculated self-consistently. The spectral function is then given by a convolution of the spinon and holon Green functions. The lowest attainable energy ($`t_h`$) and highest one ($`t_h+J_s`$) define the bandwidth of the hole, $`2t_h+J_s`$. Since the full band-width obtained by considering the compact support of the spectral function at $`J=0`$ is known to be exactly $`4t`$ , we take $`t_h=2t`$. In order to determine $`J_s`$, we consider the overall bandwidth, as obtained from the simulation. As can be seen in Fig. 1, for all values of $`J`$,
the width of the density of states $`N(\omega )`$ scales approximately as $`4t+J`$ in the parameter range considered, leading to $`J_s=J`$.
Beyond predicting bandwidths, the CSSA describes accurately the support of the spectral function in the case $`J=0`$, when compared with exact results . If furthermore phase string effects are taken into account, the singularities of $`A(k,\omega )`$ related to holons and spinons can be reproduced. For finite $`J`$, the minimal (maximal) possible energy of a hole in CSSA is given by $`E(k)=F_k`$ ($`E(k)=F_k`$) for $`k<k_0`$ ($`k>k_0`$), where $`F_k\sqrt{J^2+4t^24tJ\mathrm{cos}\left(k\right)}`$ contains both holon and spinon contributions, and $`k_0`$ is determined by $`\mathrm{cos}(k_0)=J/(2t)`$. The remaining parts of the compact support are given by $`E(k)=2t\mathrm{sin}(k)`$ for $`k>k_0`$ (lower edge) and $`k<k_0`$ (upper edge) respectively. Such dispersions correspond to holons with momentum $`k+q_s`$, and a spinon with $`q_s=\pi /2`$ . As $`J2t`$, $`k_00`$ and the lower edge of the compact support is entirely determined by the dispersion of the holon.
We now compare the above predictions with our QMC data. Figure 2 shows $`A(k,\omega )`$ for $`J/t=0.4`$ (a), $`1.2`$ (b) and $`2`$ (c). In all cases the compact support is reproduced very well by the CSSA. The Ansatz also predicts singularities at the lower (upper) edge for $`k<k_0`$ ($`k>k_0`$), and when phase strings are considered along the edges and the holon lines ($`\pm 2t\mathrm{sin}(k)`$) for all momenta. The singularities along the lower holon line are also supported by a recent low energy theory . For all parameter values we observe dominant weight along the above mentioned lines. For $`J/t=0.4`$, we have checked that the results are consistent within the uncertainties of MEM with a peak along the edges and a further peak along the holon lines, signaled by a broad structure between the edges and the holon lines (Fig. 2.a). We observed such a behaviour for $`0.33J/t0.6`$. For $`J/t1.2`$ (Fig. 2.b and 2.c), the structure at the lower edge narrows considerably and the data are not any more consistent with an additional structure along the lower holon line for $`k<k_0`$, but only with a singularity for $`k>k_0`$. At $`J/t=2`$ the exact holon and spinon dispersions can be obtained by BA . Figure 2.c shows the comparison with the CSSA, where on the one side the original dispersions are used (full line) and on the other side, with the dispersions as given by BA (crosses). Whereas the BA holon dispersion reproduces very well the lower edge, showing that as anticipated by the CSSA, at the SuSy point that edge is completely determined by the holon dispersion, the full bandwidth is better described with the original dispersions. We assign the additional weight in the region $`k>\pi /2`$ to processes involving one holon and more than one BA spinon. In fact, that portion resembles the difference between the supports for one-holon/one-spinon and one-holon/three-spinon processes in the ISE model . In our case, no limitation on the possible number of spinons exists, such that in principle all odd number of them are allowed. It is interesting to notice that using a fermionic spinon one is able to describe both the case $`J=0`$ and $`J=2t`$. In the first case, the spinon in the exact solution is a fermion. At the SuSy point it is expected to be a semion and on the basis of our results, we conclude that the fermionic spinon contains all possible states with an odd number of semionic spinons.
Finally, we consider the quasiparticle residue $`Z_k=\left|\mathrm{\Psi }_0^{L1}|\stackrel{~}{c}_{k\sigma }|\mathrm{\Psi }_0^L\right|^2`$ at $`k=\pi /2`$ for $`J=2t`$.
As Fig. 2.c shows, the lower edge is very sharp and without prior knowledge, the question may arise whether we are dealing with a quasiparticle. $`Z_k`$ is related to the imaginary time Green function through:
$`\underset{\tau \mathrm{}}{lim}G(k,\tau )Z_k\mathrm{exp}\left[\tau \left(E_0^LE_0^{L1}\left(k\right)\right)\right].`$ (9)
Fig. 3.a shows $`G(\pi /2,\tau )\mathrm{exp}\left(\tau (E_0^LE_0^{L1}(\pi /2))\right)`$ versus $`\tau `$, where the energy difference is obtained by fitting the tail of $`G(\pi /2,\tau )`$ to a single exponential form, for several sizes. The thus estimated $`Z(\pi /2)`$ is plotted versus system size in Fig. 3.b. Our results are consistent with a vanishing quasiparticle weight $`Z(\pi /2)L^{1/2}`$ which is the scaling obtained by a combination of bosonization and conformal field theory . Since the CPU-times scales as $`V\beta `$ ($`V`$ is the volume) the determination of the $`Z`$-factor may be efficiently extended to higher dimensions, in contrast to determinantal algorithms for the Hubbard model that scale as $`V^3\beta `$.
In summary, we have developed a new QMC algorithm which allows the determination of single-hole dynamics in quantum antiferromagnets. This algorithm is extremely powerful in the sense, that the required CPU time scales as $`V\beta `$. For the one dimensional case, we showed that the spectral function is well described by a simple model with free spinons and holons with dispersions given by $`J`$ and $`2t`$ respectively. The comparison of our results at the supersymmetric point lead to a characterization of the excitation content of the spectra for this particular parameter, where additional information is available from the Bethe Ansatz solution. Finally we computed the quasiparticle weight and showed that it vanishes as $`L^{1/2}`$.
This work was supported by Sonderforschungsbereich 341 in Tübingen-Stuttgart. The numerical calculations were performed at HLRS Stuttgart and HLRZ Jülich. We thank the above institutions for their support.
|
no-problem/9904/hep-th9904035.html
|
ar5iv
|
text
|
# References
A Comment on the Entropy of Strongly
Coupled $`𝒩=4`$
N. Itzhaki
Department of Physics
University of California, Santa Barbara, CA 93106
sunny@physics.ucsb.edu
## Abstract
We propose a field theory argument, which rests on the non-renormalization of the two point function of the energy-momentum tensor, why the ratio between the entropies of strongly coupled and weakly coupled $`𝒩=4`$ is of order one.
The Maldacena conjecture and the entropy of near-extremal D3-branes imply that the ratio between the entropies, at fixed temperature, of strongly coupled and weakly coupled $`𝒩=4`$ is $`3/4`$. In $`𝒩=4`$, unlike 2D CFT, the entropy is not protected thus it is not surprising that the ratio is not $`1`$. It is surprising, however, that the ratio is not a function of the ’t Hooft coupling, $`\lambda =g_{YM}^2N`$, which vanishes when $`\lambda \mathrm{}`$.
The reason is the following perturbative argument.<sup>1</sup><sup>1</sup>1Though this argument is widely known we did not find it in the literature. A closely related discussion can be found in . At finite temperature, $`T`$, the expectation value of the fields is $`\varphi ^2=T^2.`$ As a result the potential term in SYM, which has the form $`Vg_{YM}^2[\varphi _i,\varphi _j]^2,`$ induces a mass, $`m^2\lambda T^2`$, for a generic field. At small ’t Hooft coupling the induced masses are much smaller then the temperature so to a good approximation the contribution to the entropy is of $`N^2`$ massless fields with a small correction which reduces the entropy.<sup>2</sup><sup>2</sup>2For a rigorous discussion on the weakly coupled region see . At large ’t Hooft coupling the induced masses are much larger then the temperature. Therefore, the contribution to the entropy from a generic field (not in the Cartan subalgebra of $`SU(N)`$) is suppressed at the strongly coupled region.
Since the argument above rests on perturbation theory it cannot be trusted all the way to the strongly coupled region and hence, strictly speaking, there is no contradiction with the Maldacena conjecture. Still, it is fair to say that it is somewhat disturbing that the only field theory argument available (as far as we know) leads to that conclusion. Especially, when a similar argument for SYM in $`1+1`$ dimensions leads to results that fit so nicely into the Maldacena conjecture for D1-branes .
The purpose of this short note is to put forward a field theory argument, which does not rest on the AdS/CFT correspondence, that implies that the entropy at large coupling is of the order of the entropy at weak coupling. The argument rests on the $`𝒩=4`$ non-renormalization theorem for the energy-momentum tensor two point function and therefore it cannot be generalised to two dimensional SYM which is a non-conformal theory and hence the R-symmetry cannot protect the two-point function.
We study SYM in a box whose volume is $`L_zA`$ with $`A=L_xL_y`$ and we take the limit $`L_x,L_yL_z`$. Consider the transformation $`x_3x_3(1\epsilon )`$ with $`\epsilon 1`$. The variation of the action under this transformation is
$$\delta S=\epsilon d^3x_0^{L_z}𝑑x_3T_{33}.$$
(1)
Therefore, the variation of the expectation value of $`T_{00}`$ is
$`\delta T_{00}(0)={\displaystyle 𝒟\varphi (e^{(S+\delta S)}e^S)T_{00}(0)}=\epsilon {\displaystyle d^3x_0^{L_z}𝑑x_3T_{00}(0)T_{33}(x)},`$ (2)
where $`𝒟\varphi `$ represents integration with respect to all fields. To calculate the integral we need to know the energy-momentum tensor two point function. On $`R^4`$ non-renormalization theorem protects the energy-momentum tensor two point function. Thus on $`R^4`$ we can use the free SYM result
$$T(0)T(x)=\frac{N^2}{x^8},$$
(3)
where we have suppressed numerical factors of order one and the Lorentz indices (for details see ).
However, what we need is not the two-point function in $`R^4`$ but rather in $`R^3\times S^1`$. In two dimensions the conformal transformation group contains the transformation from $`R^2`$ to $`R\times S^1`$. Therefore, the two-points function in $`R\times S^1`$ are determined by the two-points function in $`R^2`$ and the dimensions of the operators. This is an important ingredient in Cardy’s proof that the asymptotic growth of the number of state of a 2D CFT depends only on the central charge and not on the details of the CFT . In four dimensions, however, the conformal transformations do not contain the transformation from $`R^4`$ to $`R^3\times S^1`$. Thus, we do not know the exact form of the energy-momentum two point function for strongly coupled SYM on $`R^3\times S^1`$.<sup>3</sup><sup>3</sup>3For weakly coupled theories one can find directly on $`R^3\times S^1`$ the mode expansion of the relevant fields. So there is no need to start with the two-points function on $`R^4`$.
What we do know is that at distances smaller then $`L_z`$ the boundary condition is irrelevant and so eq.(3) is a good approximation at short distances. Therefore, for a given point on $`S^1`$ we can calculate the contribution to $`\delta T_{00}(x_3)`$ from the region $`|x_3x_3^{^{}}|<L_z`$. The integral divergent at short distances. The regularized result is $`\delta T_{00}(0)\frac{\epsilon N^2}{L_z^4}`$. Integrating over the volume we find the variation of the ground state energy, which yields after integration with respect to $`\delta L_z=\epsilon L_z`$, the ground state energy
$$E_0\frac{N^2A}{L_z^3}.$$
(4)
It is important to emphasis that we have assumed in the calculation of $`\delta T_{00}`$ that the integration over the whole region does not contain cancellations between the region where eq.(3) is a good approximation and the region where it is not. Such cancellations can, in principle, reduce the ground state energy in a significant way to yield $`E_0`$ which is suppressed at large coupling. Therefore, our argument is not a proof but rather a strong indication that the entropies ratio is of order one. In other words, we estimate the Casimir energy, which is a boundary condition effect, using an approximation which is not sensitive to the details of the boundary condition but only to the distance between the boundaries.
Eq.(4) implies that the partition function at low temperature (compared to $`L_z`$) is,
$$Z\mathrm{exp}\left(\frac{N^2A\beta }{L_z^3}\right).$$
(5)
Now we can use the standard argument of switching the roles of $`\beta `$ and $`L_z`$ <sup>4</sup><sup>4</sup>4See for a related discussion in the context of the AdS/CFT correspondence. to end up with the partition function of strongly coupled SYM at high temperature (compared to the size of the box)
$$Z\mathrm{exp}(N^2VT^3),$$
(6)
which agrees, up to a numerical factor in the exponent, with the partition function of weakly coupled SYM.
Acknowledgements
I would like to thank Aki Hashimoto for helpful discussions. Work supported in part by the NSF grant PHY97-22022.
|
no-problem/9904/cond-mat9904288.html
|
ar5iv
|
text
|
# Thermodynamics of the Superconducting Phase Transition in Ba0.6K0.4BiO3
## Abstract
We suggest that the transition to superconductivity in a single crystal of Ba<sub>0.6</sub>K<sub>0.4</sub>BiO<sub>3</sub> with a $`T_c`$ = 32K, and having critical fields with anomalous temperature dependencies and vanishing discontinuities in specific heat and magnetic susceptibility, may well be an example of a fourth order (in Ehrenfest’s sense) phase transition. We have derived a free energy functional for a fourth order transition and calculated (for the temperature range $`T_c/2<TT_c`$) the temperature dependence of the critical fields. We find $`H_{c1}(T)(1T/T_c)^3`$, $`H_0(T)(1T/T_c)^2`$ and $`H_{c2}(T)(1T/T_c)^1`$ in general agreement with experiments.
To appear in Physical Review Letters
In the Ehrenfest classification of phase transitions, an $`n^{th}`$ order transition is described by continuous derivatives, with respect to temperature and a mechanical variable (for example a magnetic field or pressure) up to order $`(n1)`$. The $`n^{th}`$ order derivatives are discontinuous. So far however, only first and second order transitions have been observed. There are no known examples of transitions higher in order than two.
We report below, what appears to be an example of a $`fourth`$ order phase transition. In the course of measuring the magnetization of superconducting Ba<sub>0.6</sub>K<sub>0.4</sub>BiO<sub>3</sub> (BKBO) , as a function of magnetic field (up to 27T) and temperature (1.3K to 350K) we were surprised to find no evidence of a discontinuity in the magnetic susceptibility. While this was an anomalous property, it was congruent with the other mystery about BKBO, that there is no discontinuity in specific heat either at $`T_c`$. Since in a second order phase transition, the boundary between the normal and superconducting phases satisfies,
$$\left(\frac{dH_{c2}}{dT}\right)^2=\frac{\mathrm{\Delta }C}{T_c\mathrm{\Delta }\chi },$$
(1)
with both $`\mathrm{\Delta }C`$ and $`\mathrm{\Delta }\chi `$ vanishing, a question arises concerning the order of this transition.
The answer is provided by the thermodynamic critical field $`H_0(T)`$. Since in the superconducting state, the thermodynamic critical field, $`H_0(T)`$, is given by ($`0<H<H_{c2}`$), $`𝐌𝑑𝐇=H_0^2/8\pi `$. This is the free energy of the superconducting state that is derived from the experimentally determined M(H,T); thus, in case of a second order phase transition, should have the temperature dependence $`F(T)=H_0^2/8\pi (1T/T_c)^2`$. That is, $`H_0(T)`$ would be linear in $`(1T/T_c)`$, apart from critical fluctuation effects which lead to a divergent specific heat. As shown in Fig. 1, with $`T_c=32`$ K $`H_0(T)(1T/T_c)^2`$. Since for an $`n^{th}`$ order phase transition, the critical field has an exponent of $`n/2`$, the transition here must be of $`fourth`$ order in the sense of Ehrenfest.
Further support for this assertion comes from the temperature dependence of other critical fields. In particular, we find experimentally that the lower critical field, the field which separates the Meissner state (no flux in the sample) from the Abrikosov state (partial flux penetration in the form of a vortex lattice), depends on temperature as $`H_{c1}(T)(1T/T_c)^3`$ as shown in Fig. 2. The upper critical field, which separates the Abrikosov state from the normal state is measured to be $`H_{c2}(T)(1T/T_c)^{1.2}`$ as shown in Fig. 3. This fact leads to an anomalous result, specific to this higher order phase transition. For a BCS superconductor, the ratio $`\kappa ^2=H_{c2}(T)/H_{c1}(T)`$ is a constant. Here it diverges approximately as $`(1T/T_c)^2`$. Both of the critical fields are inversely proportional to squares of the two length scales in the problem, the London penetration length $`\lambda `$, which controls the flux penetration and therefore the size of a vortex, and the superconducting coherence length $`\xi `$ which determines the stiffness of the local density of the superconducting electrons. In a BCS superconductor, these length scales are identical in their temperature dependence. To our knowledge there is no fundamental reason why $`\kappa `$ should be a constant.
In the following we derive a free energy functional which describes the properties of a $`fourth`$ order phase transition. Once we include the interaction of the superconductor with the magnetic field in the usual gauge invariant form, we also can derive the temperature dependencies of the critical fields. These results are in full accord with the experiments.
The free energy is derived following the requirement of a $`fourth`$ order phase transition, viz. $`F(T)=f_o(1T/T_c)^4`$ as a function of temperature. The free energy is in the spirit of a Ginzburg-Landau (GL) functional which is minimized with respect to the complex order parameter $`\psi =\mathrm{\Delta }e^{i\varphi }`$. The value at the minimum then is the thermodynamic free energy. The first two terms are self evident. Indeed it is important that the terms proportional to $`|\psi |^2`$ and $`|\psi |^4`$ be not present. The form of the spatial gradient term is also determined by the same considerations. The term below is the one with the lowest power of gradients. Higher power of gradients such as $`|\psi ^2\psi |^2`$ and $`|^3\psi |^2`$ are possible but they contribute higher order nonlinear contributions of the magnetic field and therefore are unnecessary for a stability analysis. They can be included for effects nonlinear in the magnetic field. The free energy functional appears as,
$$F_{\mathrm{IV}}(\psi ,T)=a|\psi |^6+b|\psi |^8+c|\psi ^2\psi |^2$$
(2)
Here $`a=a_o(T/T_c1)`$ and $`b`$ and $`c`$ are positive constants.
The minimum of this free energy corresponds to an order parameter amplitude $`\mathrm{\Delta }(T)(1T/T_c)^{1/2}`$. The specific heat is expected to be $`C_{\mathrm{IV}}(T)(1T/T_c)^2`$, and $`\chi =\frac{M}{H}(1H/H_{c2})^2`$. The thermodynamic discontinuities are in the fourth derivative of the free energy (or in the second derivative of specific heat as a function of temperature or the second derivative of the magnetic susceptibility $`\chi `$ with respect to the magnetic field). We see that in the common thermodynamic observables, there are no discontinuities, as seen in the experiments. It is conceivable that broad transitions that have been observed in the past, instead of being recognized as candidates for a higher order phase transition were forcibly squeezed into a second order framework. The Ehrenfest relation appropriate for a IV order phase transition is
$$\left(\frac{dH}{dT}\right)^4=\frac{\mathrm{\Delta }\frac{^2c}{T^2}}{T_c\mathrm{\Delta }\frac{^2\chi }{H^2}}$$
(3)
In the presence of a magnetic field the gradient term transforms as $`(+\frac{2\pi i}{\varphi _o}A)`$, where A is the vector potential. Here $`\varphi _o`$ is the flux quantum, $`\varphi _o=h/2e=2\times 10^{15}Tm^2`$. Thus Eq. (2), as always, is the basis for a study of both spatial thermodynamic fluctuations as well as magnetic field effects. We note that the penetration depth for a magnetic field, the coefficient of the $`A^2`$ in the generalized Eq. (2), diverges as
$$\lambda ^2(T)=\frac{4(2\pi )^3}{\varphi _o^2}c\mathrm{\Delta }^6(1T/T_c)^3$$
(4)
This too is in agreement with experiments, not as a direct measurement but that of $`H_{c1}(T)\varphi _o/\lambda ^2`$. This is shown in Fig. 2 with $`H_{c1}(T)`$ plotted as a function of temperature. Here the data on the lower critical field is limited by its size at temperatures close to $`T_c`$. At low temperatures (less than $`T_c/2`$), $`H_{c1}(T)`$ behaves linear in T and has the right intercept at $`T_c`$. The spatial fluctuations of the order parameter are still governed by $`\xi ^2=c/a(1T/T_c)^1`$. For $`H_{c2}(T)`$ we recognize that $`\varphi _o/\xi ^2=H_{c2}(T)`$. Experimentally, as shown in Fig. 3, the exponent is nearly one.
The proposal here rests on several critical assumptions. For example, Graebner et al have reported a very small specific heat discontinuity. The reported discontinuity is in fact anomalously small and roughly of the size of their experimental uncertainly. To estimate the expected discontinuity, consider the specific heat results in ref. The high temperature limit of the specific heat can be described by $`C(T)=\gamma T+\beta T^2`$ with $`\gamma 150mJ/moleK^2`$ as the electronic contribution to $`C(T)`$. This large $`\gamma `$ puts BKBO in the category of heavy fermion compounds and the expected $`\mathrm{\Delta }C`$ (of the order of $`\gamma T_c`$ should be nearly 5J/moleK, considerably more (by a factor of $`10^5`$) than the experimental uncertainty and the reported value in ref. Moreover, Hundley et al find that at low temperatures ($`T<T_c/2`$), the linear term in $`C(T)`$ disappears. The specific heat then is given by $`C=\beta ^{^{}}T^3`$, where $`\beta ^\mathrm{`}>\beta `$. But this larger $`\beta ^{^{}}`$ may well be due to the presence of nodes in the putative energy gap at the Fermi surface. For example point nodes in the energy gap give rise to a $`CT^3`$ augmenting the well known phonon contribution with the same power.
Another basis for the suggestion here is the temperature dependence of the lower critical field $`H_{c1}(T)`$. The cubic temperature dependence here is in contrast to the results of Grader et al where $`H_{c1}(T)`$ is linear, as expected for a second order BCS superconductor. However, closer inspection reveals that the $`H_{c1}(T)`$ values of Hall et al (the values used in this analysis) are at low temperatures $`T<T_c/2`$ in agreement with the results of Grader et al who employed in their study high quality microcrystals to eliminate spurious effects associated with sample inhomogeneities. The values given in ref for $`T>T_c/2`$, while in general agreement with ref can be seen to follow a straight line but extrapolate to a smaller $`T_c27K`$. When the zero field $`T_c32K`$ is included, it is impossible to avoid a curvature in the temperature dependence of $`H_{c1}(T)`$.
We note, in passing, that the relation $`H_{o}^{}{}_{}{}^{2}=H_{c1}H_{c2}`$ is still valid. The consequences, near $`T_c`$ of a divergent $`\lambda `$ are more curious. For example the central result that the flux expulsion happens more slowly in the mixed state is clear. That the vortex lattice appears more slowly and therefore the irreversibility field is smaller is less obvious. Other questions such as the symmetry of the vortex lattice are currently under study and will be reported later. Similarly, the surface energy of a normal-superconducting (N-S) domain wall, is negative and proportional to $`\lambda `$ and therefore larger than in a BCS case. It may well engender a more inhomogeneous ground state at $`H_{c2}`$. A numerical analysis of these questions is in progress and will be reported later.
It is important to note that the thermodynamic behavior changes for $`TT_c/2`$. This is clearly seen in several independent measurements, for example specific heat and critical fields. The discussion here is focussed on the order of the transition from the normal state and is therefore limited to the vicinity of $`T_c`$. However, a microscopic theory which might attempt to derive Eq. (2) will also have to include an explanation of this crossover behavior and possible existence of point nodes in the energy gap.
It might also contain an explanation of why the free energy does not contain terms such as $`\mathrm{\Psi }^2`$ and $`\mathrm{\Psi }^4`$. At present we can only speculate about a microscopic theory. In a sense, this question is equivalent to the seemingly deeper question: Why is the transition of order IV? In this paper, we have focussed on the properties of a IV order phase transition, but let’s speculate: for instance, the BCS/GL theory contains an overall factor of density of states at the Fermi surface. Suppose, as discussed in ref. , the density of states $`N(0)=0`$ for $`TT_c`$ and $`N(0)\mathrm{\Psi }^{2p}`$ for $`TT_c`$. This would be a transition from an insulator to a superconductor, the free energy for $`p=1`$ would not have a $`\mathrm{\Psi }^2`$ term and the order of the transition would be III. For p=2, the transition would be of order IV.
Now, addressing the relationship between fluctuations as developed for a II order phase transition and the framework: in a second order phase transition, including the critical effects, one might view the free energy as depending on temperature as $`F_o(T)=f_o(1T/T_c)^{(2\alpha )}`$ Thus the small quantity $`\alpha `$ is calculated by pseudo-perturbative schemes (such as Gaussian approximation or some version of renormalization group). It is clear, however, that a value of $`\alpha =1or2`$ is essentially beyond the realm of a perturbative approach. If the free energy exponent is significantly different from 2, then the unperturbed ground state could be a transition of order corresponding to the nearest integer, about which a calculation of fluctuations could be done in the future.
The conclusions presented here are the first part of a work in progress. We are currently working on determining (1) the magnitude of the fluctuations, and (2) whether there is an upper critical dimension and, if so, what it is. These, and other points of interest, will be presented in forthcoming publications.
In summary then, we have analyzed the thermodynamic properties of the superconducting phase transition in Ba<sub>0.6</sub>K<sub>0.4</sub>BiO<sub>3</sub> (BKBO). The absence of a discontinuity in specific heat and magnetic susceptibility, on transforming from the normal to superconducting state, shows that the phase transition cannot be of second order. The temperature dependence of the thermodynamic critical field shows that the transition is $`fourth`$ order. The conclusions about other critical fields, derived from a free energy developed for a fourth order phase transition, i.e. $`H_{c1}(T)(1T/T_c)^3`$ and $`H_{c2}(T)(1T/T_c)`$ are in accord with the experiments.
The work at LSU was supported by NSF grant No. DMR-9501419. We also acknowledge the support (in many ways) of the National High Magnetic Field Laboratory which is supported by the NSF Cooperative Agreement No. DMR 9016241. Significant part of the writing of this paper was done during a stay at the Theory Division at the Los Alamos National Laboratory. PK is grateful to Alan Bishop for his hospitality and the members of the T-11 group for discussions. We thank M. Graf, A. Hebard, Daryl Hess, J. R. Schrieffer, J. D. Thompson and J. W. Wilkins for their insights and discussion.
|
no-problem/9904/chao-dyn9904013.html
|
ar5iv
|
text
|
# Traversal-Time Distribution for a Classical Time-Modulated Barrier
## 1 Introduction
The problem of obtaining the time involved in the tunneling process in quantum mechanics is still a controversial issue, despite considerable efforts in recent years . In particular, in order to address this issue, some authors have analyzed the tunneling through time-modulated potential barriers . One of the pioneer works in this area is the model introduced by Büttiker and Landauer in 1982 in which they consider the transmission through a time-modulated rectangular barrier, and introduced a characteristic time for the process. However, in the above-mentioned papers, there is practically no mention of the corresponding classical problem; although the classical limit is straightforward when the potential barrier does not depend on time, it is far from trivial when the potential is time modulated.
In this paper I study the classical problem of a rectangular time-modulated potential barrier, in order to analyze in detail the traversal time distribution for an ensemble of classical particles. This classical model was inspired, in part, by the Büttiker-Landauer model mention above.
I will study first the case of a potential barrier located inside a rigid box . In this case, the classical orbits can be periodic, quasiperiodic or even chaotic, depending on the parameters and the initial conditions of the motion. In order to study the dynamics, I derive first an area-preserving map that allow us to find the orbits for all times. Then, I study the scattering problem of an ensemble of particles that interact with an oscillating rectangular potential barrier. In this case, I will show that the traversal time strongly depends on the arrival time of the incident particles.
There is a basic difference between these two problems: (1) In the first case, what we have is the bounded problem of an oscillating barrier inside a rigid box of finite size. This means that an incident particle interact with the barrier not once but an arbitrary number of times, since the particle can cross the barrier region and then, after bouncing elastically in the box, returns to the oscillating barrier. Then, the dynamics can become chaotic, since we have the main ingredients: on one hand, sensitive dependence on initial condition or arrival times due to the oscillating barrier, and on the other hand, bounded motion due to presence of the finite box. (2) In the second case, we have a scattering problem in which an incident particle interacts with the barrier only once. Of course, if this is the case, the problem is straightforward and there is only a single traversal time. But, if we consider an ensemble of $`N`$ noninteracting particles with slightly different initial conditions, say different initial velocities, then we can expect, in general, $`N`$ different traversal times that can exhibit a complex distribution of traversal times.
An approach to the problem of tunnelling times, that is closely related to the classical trayectories discussed here, is the Bohm trajectory point of view . This approach has been used by Leavens and Aers to give an unambiguous prescription for calculating traversal times that are conceptually meaningful within that interpretation. In particular, Leavens and Aers have treated in detail the case of a time-modulated rectangular barrier, using Bohm’s trajectory interpretation of quantum mechanics . They calculate, among other things, transmission time distributions, the transmission probability as a function of frequency and Bohm trajectories.
## 2 The model and the map
Let us study the classical dynamics of a particle in a one-dimensional box, inside of which there is an oscillating rectangular potential barrier . This problem consists of a particle moving in one dimension under the action of a time-dependent potential $`V(x,t)`$. Since the Hamiltonian of this system is time dependent, the total energy of the particle is not conserved. The Hamiltonian is given by $`H(x,p,t)=p^2/2m+V(x,t)`$, where
$$V(x,t)=V_0(x)+V_1(x)f(t).$$
(1)
The potential $`V_0(x)`$ goes to infinity when $`x<0`$ or $`x>l+b+L`$, is equal to the constant value $`V_0`$ when $`lxl+b`$, and otherwise is equal to zero. Thus, what we have is an infinite potential well with a rectangular potential barrier of width $`b`$ inside, as shown in Fig. 1a. This potential separates the box in three regions: region I, $`0x<l`$ of width $`l`$; region II, where the rectangular barrier is located, $`lxl+b`$ of width $`b`$; and region III, $`l+b<xl+b+L`$ of width $`L`$.
Clearly, the motion of a particle under the influence of the potential $`V_0(x)`$ is regular, that is, we have periodic orbits and the energy is conserved. However, if we add a time-dependent potential, we can obtain periodic, quasiperiodic and chaotic orbits, as we will show below. The potential $`V_1(x)`$ in eq. (1) is different from zero only inside the interval $`lxl+b`$, where it takes the constant value $`V_1`$. The function $`f(t)`$ in eq. (1) is assumed periodic with period $`\tau `$, that is, $`f(t+\tau )=f(t)`$. In this way, as shown in Fig. 1a, what we have is an oscillating potential barrier, with an amplitude which oscillates between $`V_0V_1`$ and $`V_0+V_1`$, with frequency $`\omega /2\pi `$ and period $`\tau =2\pi /\omega `$. We will take $`V_0>V_1`$.
Let us now derive a map that describes the dynamics of a particle under this potential. The motion is as follows: at the fixed walls at $`x=0`$ and $`x=l+b+L`$, the particle bounces elastically, changing the sign of the velocity but with the same absolute value. The other two points where there is a change in the velocity is at the borders of the potential barrier at $`x=l`$ and $`x=l+b`$. The rest of the time the velocity is constant. Thus, the particle can gain or loose kinetic energy at $`x=l`$ and $`x=l+b`$. The phase space for a typical orbit is depicted in Fig. 1b.
We can analyze the dynamics using a discrete map from the time $`t_n`$ when the particle hits the wall at $`x=0`$, until the next time $`t_{n+1}`$ when it hits this wall again. Let us denote by $`v_n`$ the velocity of the particle immediately after the $`nth`$ kick with the fixed wall at $`x=0`$, and by $`E_n`$ the corresponding total energy. Clearly, $`E_n=mv_{n}^{}{}_{}{}^{2}/2`$. After traveling the distance $`l`$, it arrives at the left side of the barrier after a time of flight $`l/v_n`$, where a change in the velocity occurs. To determine this change let us consider the following: In region I, the total energy of the particle is given by $`E_n=mv_{n}^{}{}_{}{}^{2}/2`$ which is just the kinetic energy, because in this region the potential energy is zero; when the particle enters region II, the kinetic energy $`E_n^{}`$ is changed to $`E_nV_0V_1f(t_n+l/v_n)`$, that is, the total energy minus the value of the potential energy at the time of arrival $`t_n+l/v_n`$. If we denote the new velocity by $`v_n^{}`$ (see Fig. 1b), then $`E_n^{}=mv_{n}^{}{}_{}{}^{2}/2`$ and we obtain in this way the change in energy as:
$$E_n^{}=E_nV_0V_1f\left(t_n+\frac{l}{v_n}\right).$$
(2)
Clearly, if the total energy is less than the potential energy at time $`t_n+l/v_n`$, then the particle cannot penetrate region II and simply reflects elastically and there is only a change in the sign of the velocity; thus the particle gets trapped in region I and returns to the wall at $`x=0`$. After a time lapse of $`2l/v_n`$ it will hit again the oscillating barrier and try again to cross it. If this time the total energy is greater than the potential energy, then the particle can cross the barrier region; otherwise, it bounces once more inside region I, and so on.
Now, once the particle overcomes the barrier, it crosses the region II without changing its velocity $`v_n^{}`$, even though the barrier is oscillating in time. When the particle arrives at the right side of the barrier at $`x=l+b`$, then another change in the velocity takes place, but this time the velocity increases in such a way that the kinetic energy $`E_n^{\prime \prime }`$ becomes
$$E_n^{\prime \prime }=E_n^{}+V_0+V_1f\left(t_n+\frac{l}{v_n}+\frac{b}{v_n^{}}\right),$$
(3)
where $`E_n^{\prime \prime }`$ is the energy in region III. Clearly, the time that it takes to arrive at the wall located at $`x=l+b+L`$ is $`l/v_n+b/v_n^{}+L/v_n^{\prime \prime }`$, where $`v_n^{\prime \prime }`$ is the velocity in region III (see Fig. 1b). After a time $`t_n+l/v_n+b/v_n^{}+2L/v_n^{\prime \prime }`$, the particle returns to the right side of the barrier after traveling twice the distance $`L`$ in region III, and enters once again region II. However, in general, the potential barrier has a different height, given by $`V_0+V_1f(t_n+l/v_n+b/v_n^{}+2L/v_n^{\prime \prime })`$. Therefore, the new kinetic energy $`E_n^{\prime \prime \prime }`$ inside region II is now given by
$$E_n^{\prime \prime \prime }=E_n^{\prime \prime }V_0V_1f\left(t_n+\frac{l}{v_n}+\frac{b}{v_n^{}}+\frac{2L}{v_n^{\prime \prime }}\right),$$
(4)
Here, once more, there is the possibility that the total energy in region III is less than the potential energy at time $`t_n+l/v_n+b/v_n^{}+2L/v_n^{\prime \prime }`$. In this case, the particle gets trapped in region III until it can escape by crossing the barrier region.
Finally, after a time $`b/|v_n^{\prime \prime \prime }|`$, where $`v_n^{\prime \prime \prime }`$ is the velocity in region II (see Fig. 1b), the particle arrives at the left side of the barrier at $`x=l`$, where the velocity varies once more depending on the height of the barrier at time $`t_n+l/v_n+b/v_n^{}+2L/v_n^{\prime \prime }+b/|v_n^{\prime \prime \prime }|`$. We will denote the velocity in region I, after this time, by $`v_{n+1}`$, because this is precisely the velocity after the next hit with the wall at $`x=0`$. The last part of this journey is covered in a time span of $`l/|v_{n+1}|`$; after this, the particle hits the wall at the origin at time $`t_{n+1}`$ and start again its trip to the oscillating barrier, and the whole process starts again.
Therefore we arrive at the following map in terms of energy and time:
$$E_{n+1}=E_n^{\prime \prime \prime }+V_0+V_1f(t_n+T_n)$$
(5)
and
$$t_{n+1}=t_n+T_n+\sqrt{\frac{m}{2}}\frac{l}{\sqrt{E_{n+1}}},$$
(6)
where $`T_n`$ is given by
$$T_n=\sqrt{\frac{m}{2}}\left(\frac{l}{\sqrt{E_n}}+\frac{b}{\sqrt{E_n^{}}}+\frac{2L}{\sqrt{E_n^{\prime \prime }}}+\frac{b}{\sqrt{E_n^{\prime \prime \prime }}}\right)$$
(7)
and $`E_n^{}`$, $`E_n^{\prime \prime }`$ and $`E_n^{\prime \prime \prime }`$ are given by eqs. (2-4), respectively.
Furthermore, it can be shown that, for this map, the Jacobian is exactly one, that is,
$$J=\frac{(E_{n+1},t_{n+1})}{(E_n,t_n)}=1.$$
(8)
This result indicates that this map is an area-preserving one .
Let us scale the time using the period $`\tau `$ of the function $`f(t)`$. We define the dimensionless quantities: $`\varphi _n=(2\pi /\tau )t_n`$ and $`\mathrm{\Phi }_n=(2\pi /\tau )T_n`$. In order to scale the energies we introduce the dimensionless variables: $`e_n=E_n/V_0`$, $`e_n^{}=E_n^{}/V_0`$, $`e_n^{\prime \prime }=E_n^{\prime \prime }/V_0`$ and $`e_n^{\prime \prime \prime }=E_n^{\prime \prime \prime }/V_0`$. With all this definitions we arrive at the following dimensionless map:
$$e_{n+1}=e_n^{\prime \prime \prime }+1+rf\left(\varphi _n+\mathrm{\Phi }_n\right),$$
(9)
and
$$\varphi _{n+1}=\varphi _n+\mathrm{\Phi }_n+\frac{2\pi M}{\sqrt{e_{n+1}}},(mod2\pi )$$
(10)
where $`M=l/(w\tau )`$, $`r=V_1/V_0`$ and $`w=(2V_0/m)^{1/2}`$. Here, $`\mathrm{\Phi }_n`$ is given by
$$\mathrm{\Phi }_n=2\pi M\left(\frac{1}{\sqrt{e_n}}+\frac{b}{l}\frac{1}{\sqrt{e_n^{}}}+\frac{2L}{l}\frac{1}{\sqrt{e_n^{\prime \prime }}}+\frac{b}{l}\frac{1}{\sqrt{e_n^{\prime \prime \prime }}}\right).$$
(11)
This map, although more complicated, resembles the structure of the Fermi Map .
## 3 Numerical results
Let us now analyze numerically the map obtained above. First of all, we notice that we have four dimensionless parameters: the width of the barrier $`b/l`$ scaled with the length of region I; the length $`L/l`$ of region III scaled with $`l`$; the ratio of the amplitude of oscillation of the barrier scaled with its height $`r=V_1/V_0`$; and $`M=l/(w\tau )`$. The parameter $`M`$ is the ratio of the time of flight $`l/w`$ in region I of Fig. 1a, with velocity $`w`$, and the period $`\tau `$ of oscillation of the barrier. That is, $`M`$ measures the number of oscillations of the barrier since the particle leaves the wall at $`x=0`$ until it arrives at the left side of the barrier. On the other hand, we will take the periodic function as: $`f(\varphi _n)=\mathrm{sin}(\varphi _n)`$.
If we fix the barrier position within the one-dimensional box, and choose a width, then we are fixing the parameters $`b/l`$ and $`L/l`$; the remaining two parameters $`M`$ and $`r`$ will control the type of motion. In what follows, we take the symmetric case, $`b/l=1`$ and $`L/l=1`$, which corresponds to the oscillating barrier centered inside the box, and an oscillating amplitude of $`r=0.5`$.
In Fig. 2 we show the energy-phase space $`(e_n,\varphi _n)`$ for $`M=4.7`$, using the map given by eqs. (9-11). We plot several orbits that correspond to different initial conditions. We can clearly see that, for this system, we have a phase space with a mixed structure, in which we have periodic, quasiperiodic and chaotic orbits. Some of the fixed points of the map can be seen surrounded by elliptic orbits. We notice a fine structure of smaller islands in the chaotic region, as is usually the case for other maps .
The quantity that we want to analyze in detail is the traversal time in the barrier region, that is, the time it takes the particle to cross the region where the barrier is oscillating. We can obtain this quantity simply as $`b/v_n^{}`$ or $`b/|v_n^{\prime \prime \prime }|`$ (see Fig. 1b). The structure of this traversal or dwell time depends strongly on the type of orbit. Clearly, if we have a periodic orbit, then this time will take only two possible values, since $`v_n^{}`$ and $`v_n^{\prime \prime \prime }`$ does not change with $`n`$. On the other hand, if the orbit is quasiperiodic, the velocity can vary in a full range of values. In this case, the traversal time can vary only in a limited range. However, when we have a chaotic orbit, the variation can display a very rich structure .
For the bounded problem, where the oscillating barrier is confined within a box, we can obtain a chaotic dynamics as shown in Fig. 2. However, if we remove the walls and leave only the oscillating barrier, we end up with an open system of the scattering type. In this case, we cannot have chaotic dynamics, since the particle interacts with the barrier only once. However, we can study not a single particle, but an ensemble of noninteracting particles, each of them with different initial conditions.
In Fig. 3 we show a space-time diagram of trayectories for an ensemble of incident particles. In this case, and for the rest of the figures, we take $`r=0.5`$ and $`M=77.7`$. I use dimensionless distance $`x/l`$ and dimensionless time $`t`$, which is the time scaled with $`l/w`$. Since $`l=b`$, then $`l/w`$ is the time it takes to cross the barrier region with a velocity $`w=(2V_0/m)^{1/2}`$. The barrier is located between $`x/l=1`$ and $`x/l=2`$, and is indicated by horizontal dashed lines in Fig. 3. We take an ensemble of initial conditions in which the initial velocity is constant and the initial phase is uniformily distributed. We see from Fig. 3 that only a subset of particles in the ensemble can cross the barrier region and that the traversal time is different for each particle. This is due to the fact that each particle is influenced differently by the time-modulated barrier, depending on the arrival time. That is, different arrival times mean different barrier amplitudes.
The traversal time is defined as the time it takes to cross the region where the barrier is oscillating, and is given by $`b/v_n^{}`$. Since we scale this traversal time with the time $`b/w`$, the dimensionless form is given by $`1/\sqrt{e_n^{}}`$. For the particles in the ensemble, this time is shown in Fig. 4. We notice that in many cases the dimensionless time $`t1`$; however, there are some others cases for which $`t1`$. These large peaks occur when the arrival time is such that the total energy is just above the barrier heigth, and thus the velocity inside the barrier region is very small and consequently the traversal time is very large. We can see a strong variation in the traversal time, that leads to a broad distribution of times. On the other hand, since the minimum velocity in the barrier region is zero, then there is no upper bound for the dwell time, and it can acquire very large values, as seen in Fig. 4.
The traversal time distribution is depicted in Fig. 5. This normalized distribution has a long-time tail which is a power law. In Fig. 6 we show the same distribution in a log-log plot that clearly shows that this is indeed a distribution with a power-law tail of the form $`p(t)t^\alpha `$, with $`\alpha 3`$. The straight (dashed) line in this figure has a slope of $`3`$.
Another quantity of interest is the transmission coefficient, defined as the number of particles that cross the barrier region, divided by the total number of particles in the ensemble. In Fig. 7 we show this transmission coefficient as a function of $`M`$. Remember that $`M=l/(w\tau )`$ and is, therefore, proportional to the frequency of oscillation of the barrier. We can see in this figure that the transmission coefficient vary strongly with $`M`$, in particular for low frequencies ($`M1`$). On the other hand, for higher frequencies ($`M1`$), the transmission coefficient tend towards a constant value. This last result indicates that for $`M1`$, the oscillating potential barrier acts as an effective potential barrier of average height $`V_0`$.
Finally, in Fig. 8 we show the average traversal time as a function of $`M`$. Again we can see strong fluctuations of this quantity. Since the distribution of traversal times is a power law with an exponent $`\alpha 3`$, we can expect these large fluctuations; although the first moment of the distribution is finite in this case, the second or higher moments can diverge, leading to these large fluctuations, as is usually the case for Lévy distributions .
## 4 Concluding remarks
In this paper, the dynamics of the classical problem of an oscillating rectangular potential barrier is analyzed. When the oscillating barrier is located within a one-dimensional box, we have a bounded problem and the corresponding classical dynamics can have a mixed phase space structure comprising periodic, quasiperiodic and chaotic orbits. For the scattering problem of a single oscillating barrier, a distribution of traversal times with a power-law tail is obtained. This Lévy-type distribution of times leads to large fluctuations of the average traversal time as a function of the frequency of oscillation of the barrier; therefore, it is difficult to obtain a characteristic time to the process of crossing the classical oscillating barrier. These large fluctuations arise due to the sensitive dependence on initial conditions, typical of the dynamics of chaotic systems. In particular, for our problem, the quantity that controls the traversal time is the time of arrival at the barrier. Thus, we obtain a sensitive dependence on the time of arrival for the classical case. The possible role for the tunneling time problem, if any, of the sensitive dependence on the time of arrival and the difficulty to obtain a characteristic traversal time in the classical domain, remains to be seen.
## 5 Figure Captions
Fig. 1 a) Potential well with a rectangular time-modulated potential barrier of width b. The height of the barrier oscillates harmonically between $`V_0+V_1`$ and $`V_0V_1`$. b) Typical orbit in phase space, showing a general change in the velocity for one iteration of the map (solid line) and a second iteration (dashed line).
Fig. 2 Phase space $`(e_n,\varphi _n)`$, for $`M=4.7`$ and $`r=0.5`$, showing periodic, quasiperiodic and chaotic orbits for different initial conditions.
Fig. 3 Space-time diagram of trayectories for an ensemble of incident particles with the same velocity and different phases. In this case $`M=77.7`$ and $`r=0.5`$. The horizontal dashed lines indicate the barrier region.
Fig. 4 Traversal time for an ensemble of incident particles. In this case $`M=77.7`$ and $`r=0.5`$.
Fig. 5 Traversal time distribution for the case $`M=77.7`$ and $`r=0.5`$.
Fig. 6 Log-log plot of the traversal time distribution of Fig. 5, clearly showing a power law. The slope of the dashed line is $`3`$.
Fig. 7 Transmission coefficient as a function of $`M`$, for $`r=0.5`$.
Fig. 8 Average traversal time as a function of $`M`$, for $`r=0.5`$.
|
no-problem/9904/quant-ph9904023.html
|
ar5iv
|
text
|
# Entanglement-Assisted Classical Capacity of Noisy Quantum Channels
\[
## Abstract
Prior entanglement between sender and receiver, which exactly doubles the classical capacity of a noiseless quantum channel, can increase the classical capacity of some noisy quantum channels by an arbitrarily large constant factor depending on the channel, relative to the best known classical capacity achievable without entanglement. The enhancement factor is greatest for very noisy channels, with positive classical capacity but zero quantum capacity. We obtain exact expressions for the entanglement-assisted capacity of depolarizing and erasure channels in $`d`$ dimensions.
\]
Prominent among the goals of quantum information theory are understanding entanglement and calculating the several capacities of quantum channels. Physically, a quantum channel can be pictured as the transfer of some quantum system from sender to receiver. If the transfer is intact and undisturbed, the channel is noiseless; if the quantum system interacts enroute with some other system, a noisy quantum channel results. Quantum channels can be used to carry classical information, and, if they are not too noisy, to transmit intact quantum states and to share entanglement between remote parties. Unlike classical channels, which are adequately characterized by a single capacity, quantum channels have several distinct capacities. These include a classical capacity $`C`$, for transmitting classical information, a quantum capacity $`Q`$, for transmitting intact quantum states, a classically-assisted quantum capacity $`Q_2`$, for transmitting intact quantum states with the help of a two-way classical side-channel, and finally $`C_E`$, the entanglement-assisted classical capacity, which we define as a quantum channel’s capacity for transmitting classical information with the help of unlimited prior pure entanglement between sender and receiver . In most cases, only upper and lower bounds on these capacities are known, not the capacities themselves .
Entanglement, eg in the form of Einstein-Podolsky-Rosen (EPR) pairs of particles shared between two parties, interacts in subtle ways with other communications resources. By itself, prior entanglement between sender and receiver confers no ability to transmit classical information, nor can it increase the capacity of a classical channel above what it would have been without the entanglement. This follows from the fact that local manipulation of one of two entangled subsystems cannot influence the expectation of any local observable of the other subsystem . This is sometimes loosely called the constraint of causality, because its violation would make it possible to send messages into one’s past.
On the other hand, it is well known that prior entanglement can enhance the classical capacity of quantum channels. In the effect known as superdense coding, discovered by Wiesner , the classical capacity of a noiseless quantum channel is doubled by prior entanglement. In other words, $`C_E=2C`$ for any noiseless quantum channel. We show that for some channels this enhancement persists, and even increases, as the channel is made more noisy, even after the channel has become so noisy that its quantum capacities $`Q`$ and $`Q_2`$ both vanish, and the channel itself can be simulated by local actions and classical communication between sender and receiver .
This is perhaps surprising, since it might seem that any quantum channel that can be classically simulated ought to behave like a classical channel in all respects—in particular not having its capacity increased by prior entanglement. In fact there is no contradiction, because, as we shall see, even when a quantum channel can be classically simulated, the simulation necessarily involves some amount of forward classical communication from the sender (henceforth “Alice”) to the receiver (“Bob”); and this information is never less than the channel’s entanglement-assisted capacity. Thus for any quantum channel, $`CC_EFCCC`$, where $`FCCC`$ denotes the forward classical communication cost, ie the forward classical capacity needed, in conjunction with other resources, to simulate the quantum channel.
To illustrate these inequalities consider a specific example, the 2/3-depolarizing qubit channel, which transmits the input qubit intact with probability 1/3 and replaces it by a random qubit with probability 2/3. As is well known, this noisy quantum channel, sometimes referred to as the classical limit of teleportation, can be simulated classically by the following “measure/re-prepare” procedure: A third party chooses a random axis R and tells both Alice and Bob. Then Alice measures the input qubit along this axis and tells Bob the one-bit result, after which Bob prepares an output qubit in the same state found by Alice’s measurement. Evidently the FCCC of this procedure is 1 bit, but the best known classical capacity of a 2/3 depolarizing channel (realized by encoding 0 and 1 as $`|0`$ and $`|1`$ on the input side and measuring in the same basis on the output side) is about 0.0817 bits, the capacity of a classical binary symmetric channel of crossover probability 1/3. As we shall show, the $`C_E`$ of the 2/3 depolarizing channel is about 0.2075 bits, more than twice the unassisted value, but still safely less than the 1 bit forward classical cost of simulating the channel by measure/re-prepare, which we denote $`FCCC_{MR}`$.
Suppose we wished to simulate not a 2/3-depolarizing channel, but a 5/6-depolarizing channel. Clearly this could be done by simulating the 2/3-depolarizing channel then further depolarizing its output. But a more economical simulation would be for Alice to send her one-bit measurement result to Bob not through a noiseless classical channel but through a noisy classical channel of correspondingly lesser capacity. If she sent it through a binary symmetric channel of randomization probability 1/2 (equivalent to a crossover probability 1/4), the 5/6-depolarizing channel would be have been simulated at an $`FCCC_{MR}`$ of only $`1H_2(1/4)0.1887`$ bits per channel use, where $`H_2`$ is the binary Shannon entropy $`H_2(p)=p\mathrm{log}_2p(1p)\mathrm{log}_2(1p)`$. This is of course greater than the 5/6-depolarizing channel’s best known classical capacity of $`1H_2(5/12)0.02013`$. The 5/6-depolarizing channel’s entanglement-assisted capacity must lie between these two bounds.
We now develop these ideas further to obtain an exact expression for $`C_E`$ for an important class of channels, the $`d`$-dimensional depolarizing channel $`𝒟_x^{(d)}`$ of depolarization probability $`x`$. This is the channel that transmits a $`d`$-state quantum system intact with probability $`1x`$ and randomizes its state with probability $`x`$. We show that in the high-depolarization limit $`x1`$ this channel’s entanglement-assisted capacity is $`d+1`$ fold higher than the best known lower bound on the classical capacity of the same channel without prior entanglement. This lower bound, the “one-shot” classical capacity $`C_1,`$ is defined as the maximum classical information that can be sent through a single use of the channel, without prior entanglement, by an optimal choice of source states at the channel input and an optimal measurement at the channel output. For this highly symmetric channel, this optimum can be achieved by assigning equal probability $`1/d`$ to each state of an arbitrary orthonormal basis $`\{|0,|1\mathrm{}|d1\}`$ at the channel input, and performing a complete von Neumann measurement in the same basis at the channel output. This causes the quantum channel to behave as a $`d`$-ary symmetric classical channel of randomization probability $`x`$, giving a capacity
$$C_1(𝒟_x^{(d)})=\mathrm{log}_2dH_d(1x\frac{(d1)}{d}),$$
(1)
where $`H_d(p)=p\mathrm{log}_2(p)(1p)\mathrm{log}_2((1p)/(d1))`$ is the Shannon entropy of a $`d`$-ary distribution consisting of one element of probability $`p`$ and $`d1`$ elements each of probability $`(1p)/(d1)`$. This input ensemble is known to be optimal, for a one-shot use of the channel, because it saturates the Holevo bound $`C_1\mathrm{log}_2d\overline{S(\rho _i)}`$, on the one-shot capacity , where $`\overline{S(\rho _i)}`$ is the average von Neumann entropy of the output states $`\rho _i`$.
Similarly, it is easy to generalize the measure/re-prepare construction to show that a $`d`$-dimensional depolarizing channel can be simulated classically whenever $`xd/(d+1)`$, at a cost
$$FCCC_{MR}(𝒟_x^{(d)})=\mathrm{log}_2(d)H_d\left(dx(d(1/d))\right).$$
(2)
The simulation is performed by having Alice measure in a pre-agreed random basis, send Bob the result through a $`d`$-ary symmetric noisy classical channel, after which he re-prepares an output state in the same basis. Figure 1 compares the definitions of asymptotic capacity $`C`$ and one-shot capacity $`C_1`$, and illustrates the measure/re-prepare technique for simulating some noisy quantum channels classically.
So far, we have only given lower and upper bounds on $`C_E`$, without calculating $`C_E`$ itself. To do so we use modified versions of the well-known superdense coding and teleportation protocols to obtain tighter lower and upper bounds, respectively, which in the case of depolarizing and erasure channels coincide, thereby establishing $`C_E`$ exactly for these channels. We treat the case where $`𝒩`$ is a generalized depolarizing channel $`𝒟_x^{(d)}`$ first.
Clearly $`C_E`$ for any noisy channel $`𝒩`$ can be lower-bounded by the entanglement-assisted capacity via a particular protocol, namely superdense coding with the noisy quantum channel $`𝒩`$ substituted for the usual noiseless return path for Alice’s half of a shared maximally entangled EPR state $`\mathrm{\Psi }`$. This version of superdense coding is illustrated in Figure 2a, and we shall use $`C_{Sd}(𝒩)`$ to denote the entanglement-assisted capacity of $`𝒩`$ via this protocol. Conversely (Figure 2b), $`C_E(𝒩)`$ can be upper-bounded by the forward classical communication cost of simulating $`𝒩`$, not by measure/re-prepare, but by a version of teleportation in which the requisite amount of noise is introduced by substituting a noisy classical channel N for the usual noiseless classical arm of the teleportation procedure (the classical channel N operates on a $`d^2`$-letter classical alphabet, in contrast to the $`d`$-letter alphabet used by the channel n in the measure/re-prepare simulation of Fig 1c). This upper bound follows from the fact that even in the presence of prior shared entanglement, the FCCC of simulating a quantum channel cannot be less than its classical capacity; otherwise a violation of causality would occur. Whenever a quantum channel $`𝒩`$ can be simulated by teleportation with a noisy classical arm we use $`FCCC_{Tp}(𝒩)`$ to denote the forward classical communication cost of doing so.
In the case of depolarizing channels the two bounds coincide, because of the readily verified fact that superdense coding and teleportation map each $`x`$-depolarizing $`d`$-dimensional quantum channel into an $`x`$-randomizing $`d^2`$-ary symmetric classical channel and vice versa. Thus for all depolarizing channels $`𝒟_x^{(d)}`$,
$$C_E=C_{Sd}=FCCC_{Tp}=2\mathrm{log}_2dH_{d^2}(1x\frac{d^21}{d^2}).$$
(3)
From equations 1 and 3 it can be seen that in the high-noise limit $`x1`$, the enhancement factor $`C_E/C_1`$ approaches $`d+1`$. Thus prior entanglement can increase classical capacity by an arbitrarily large factor. For large $`d`$, $`C_E/C_12`$ for most $`x`$, rising sharply near $`x=1`$.
We now turn to the quantum erasure channel, which is unusual among noisy quantum channels in that its capacities $`C`$, $`Q`$ and $`Q_2`$ are known exactly . A quantum erasure channel transmits its $`d`$-dimensional input state intact with probability $`1x`$ and with probability $`x`$ replaces the input by a unique $`(d+1)`$’st state, called an erasure symbol, orthogonal to all the input states. If the channels $`𝒩`$ and N in Figure 2 are taken to be, respectively, a $`d`$-dimensional quantum erasure channel and a $`d^2`$-dimensional classical erasure channel, the superdense coding and teleportation bounds can again easily be shown to coincide, providing an entanglement-assisted capacity $`C_E=2(1x)\mathrm{log}d,`$ exactly twice the erasure channel’s ordinary classical capacity.
Figure 3 left shows all the capacities of the quantum erasure channel. These capacities are of interest not only in their own right, but also because they upper-bound the corresponding capacities of the depolarizing channel, since a quantum erasure channel can simulate a depolarizing channel by having the receiver substitute a fully depolarized state for every erasure symbol he receives.
Returning to the depolarizing channel, we are in the peculiar position of knowing its entanglement-assisted FIG. 3.: Left: Capacities of the quantum erasure channel.
Right: Bounds on the asymptotic classical capacity $`C`$ of the qubit depolarizing channel. classical capacity $`C_E`$ without knowing its ordinary unassisted classical capacity $`C`$. The latter is generally believed to be equal to the one-shot unassisted capacity $`C_1`$, but the possibility cannot be excluded that a higher capacity might be achieved asymptotically by supplying entangled inputs to multiple instances of the channel (this cannot occur for $`C_E`$, where any larger capacity would exceed $`FCCC_{Tp}`$, violating causality). The range of possible values for the depolarizing channel’s unassisted classical capacity $`C`$ is bounded below by its known $`C_1`$, and above by its known $`C_E`$ and by the known unassisted classical capacity $`(1x)\mathrm{log}d`$ of the quantum erasure channel. Figure 3 right shows these bounds for the qubit case $`d=2`$.
Although the depolarizing channel’s unassisted capacity $`C`$ remains unknown in absolute terms for all $`d`$, the bounds $`C_1C(1x)\mathrm{log}d`$ become increasingly tight relative to $`C`$ as $`d\mathrm{}`$, because, as can readily be verified, the difference between the bounds approaches $`H_2(x)`$ in this limit. Similarly, the depolarizing channel’s unassisted quantum capacity $`Q`$ is upper bounded by the erasure channel’s quantum capacity, $`\mathrm{max}\{0,12x\}\mathrm{log}d`$, and lower bounded by the depolarizing channel’s quantum capacity via random hashing , $`\mathrm{log}dS([𝒩I](\mathrm{\Psi }))`$. Here $`[𝒩I](\mathrm{\Psi })`$ is the mixed state formed by sending half a maximally entangled $`dd`$ pair $`\mathrm{\Psi }`$ through the noisy channel. Again the difference between the bounds approaches $`H_2(x)`$ as $`d\mathrm{}`$.
The equality between $`FCCC_{Tp}`$ and $`C_{Sd}`$, which makes $`C_E`$ exactly calculable for depolarizing channels, holds for all “Bell-diagonal” channels , those that commute with superdense coding and teleportation, so that $`Tp(Sd(𝒩))=𝒩`$ . For example the qubit dephasing channel, which subjects its input to a $`\sigma _z`$ Pauli rotation with probability $`x/2`$, has $`C=1`$ independent of $`x`$, while $`C_E=2H_2(x/2)`$. For other channels, it can be shown that
$$C_E=\underset{\mathrm{\Psi }}{\mathrm{max}}\{S(\rho )+S(𝒩(\rho ))S([𝒩I](\mathrm{\Psi }))\},$$
(4)
where $`\mathrm{\Psi }`$ is a bipartite pure state in $`dd`$ and $`\rho `$ is its partial trace over the second party. This capacity can be achieved asymptotically by applying superdense coding to a Schumacher-compressed version of $`\rho ^n`$ for large $`n`$, and evaluating the resulting classical capacity by Holevo’s formula ; that $`C_E`$ can be no higher can be shown using Holevo’s formula and the strong subadditivity property of quantum entropy.
A channel’s entanglement-assisted quantum capacity $`Q_E`$ may be defined as its maximum rate for transmitting intact qubits with the help of prior entanglement but no classical communication. By teleportation and superdense coding, $`Q_E=C_E/2`$ for all channels. Naturally, $`Q_E`$ upper bounds the unassisted quantum capacity $`Q`$, but in most instances, eg the depolarizing channel, tighter upper bounds are known.
We thank Howard Barnum, Herb Bernstein, David DiVincenzo, Richard Jozsa, Barbara Terhal, Joy Thomas, and Bill Wootters for helpful discussions. CHB, AVT, and JAS acknowledge support by the U.S. Army Research Office under contract DAAG55-98-C-0041, and AVT under DAAG55-98-1-0366.
|
no-problem/9904/hep-ph9904492.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
The investigation of heavy quarks production in high energy hadron collisions provides a method for studying the internal structure of hadrons. Some problems are the same in hadroproduction and photo-/electroproduction processes. So, the review of the situation of the heavy quark hadroproduction can be useful for the interpretation of HERA data.
In this talk we present a short review of heavy quark hadroproduction. The theoretical predictions are usually obtained in the NLO parton model . The assumptions which are used for simplifications of the computations are considered in Sect. 2. In the case of one-particle distributions even LO ($`\alpha _s^2`$) parton model with collinear approximation is enough for the data description, NLO contributions ($`\alpha _s^3`$) only change the normalizations. On the other hand, in the case of two-particle distributions, see Sect. 3, the collinear approximation has failed, and it is necessary to account for the transverse momenta of the incident partons. The possibility to include the transverse momenta of the incident partons in the framework of semihard theory , where the virtualities and polarizations of the gluons are taken into account, is considered in Sect. 4. In Ref. we presented the results for main and simplest subprocess, $`gg\overline{Q}Q(\alpha _s^2)`$ for hadroproduction, and $`\gamma g\overline{Q}Q(\alpha _s)`$ for photo- and electroproduction.
## 2 Conventional NLO parton model
The conventional NLO parton model expression for the heavy quark hadroproduction cross sections has the factorization form :
$$\sigma (abQ\overline{Q})=\underset{ij}{}𝑑x_i𝑑x_jG_{a/i}(x_i,\mu _F)G_{b/j}(x_j,\mu _F)\widehat{\sigma }(ijQ\overline{Q}),$$
(1)
where $`G_{a/i}(x_i,\mu _F)`$ and $`G_{b/j}(x_j,\mu _F)`$ are the structure functions of partons $`i`$ and $`j`$ in the colliding hadrons $`a`$ and $`b`$, $`\mu _F`$ is the factorization scale (i.e. virtualities of incident partons) and $`\widehat{\sigma }(ijQ\overline{Q})`$ is the cross section of the subprocess which is calculated in perturbative QCD. The last cross section can be written as a sum of LO and NLO contributions, $`\widehat{\sigma }(ijQ\overline{Q})=\alpha _s^2(\mu _R)\sigma _{ij}^{(o)}+\alpha _s^3(\mu _R)\sigma _{ij}^{(1)}`$, where $`\mu _R`$ is the renormalization scale, and $`\sigma _{ij}^{(o)}`$ as well as $`\sigma _{ij}^{(1)}`$ depend practically only on one variable $`\rho =\frac{4m_Q^2}{\widehat{s}}`$ , $`\widehat{s}=x_ix_js_{ab}`$.
The expression (1) corresponds to the process shown schematically in Fig. 1 with
$$q_{1T}=q_{2T}=0.$$
(2)
The main contribution to the cross section at small $`x`$ is known to come from gluon-gluon fusion, $`i=j=g`$.
The principal uncertainties of any numerical QCD calculation of heavy flavour production are connected with the unknown values of the parameters: both scales, $`\mu _F`$ and $`\mu _R`$<sup>1</sup><sup>1</sup>1These uncertainties should disappear when one sums up all the high order contributions. Sometimes people say that strong scale dependence of the calculated results in LO or NLO means the large contribution of high order diagrams and weak dependence means their small contribution. Of course, it is not true. Strong scale dependence of NLO results means only strong scale dependence of high order contributions but at some fixed scale value the last ones can be numerically small. Weak scale dependence of NLO results means weak scale dependence of high order terms but they can be numerically large., and the exact value of heavy quark mass, $`m_Q`$. The values of both scales should be of the order of hardness of the considered process, however nobody can say what is better to use for scales, $`m_Q`$, $`m_T=\sqrt{m_Q^2+p_T^2}`$ or $`\widehat{s}`$. The phenomenological parton densities are sometimes (at very small $`x`$) in contradiction with the general properties of perturbative QCD. However it is just the region that dominates in the heavy quark production at high energies<sup>2</sup><sup>2</sup>2In the case of charm production, $`m_c`$ = 1.4GeV, at LHC, $`\sqrt{s}`$ = 14 TeV, the product $`x_1x_2`$ of two gluons (both $`x_1`$ and $`x_2`$ are the integral variable) is equal to $`410^8`$.. Another problem of parton model is the collinear approximation. The transverse momenta of the incident partons, $`q_{iT}`$ and $`q_{jT}`$ are assumed to be zero, and their virtualities are accounted for only via structure functions; the cross sections $`\sigma _{ij}^{(o)}`$ and $`\sigma _{ij}^{(1)}`$ are assumed to be independent on these virtualities.
The NLO parton model calculations of the total cross sections of $`c\overline{c}`$ and $`b\overline{b}`$ production, as functions of the beam energy, for $`\pi ^{}N`$ and $`pN`$ collisions can be found in . These results depend strongly (on the level of several times) on the numerical values of quark masses as well as on the both scales, $`\mu _F`$ and $`\mu _R`$. Some experimental data are in contradiction with each other, however generally they are in agreement with NLO parton model predictions.
The NLO contributions to one-particle distributions lead only to renormalization of LO results, practically without correction of the shapes of a distributions . It means that instead of more complicate calculation of $`p_T`$, or rapidity distributions, in NLO, it is enough to calculate them in LO, and multiply after by K-factor
$$K=\frac{LO+NLO}{LO},$$
(3)
which can be taken, say, from the results for total production cross sections. The comparison of LO + NLO calculations with LO multiplied by K-factor is presented in Fig. 2 taken from Ref. .
The values of K-factors and their energy and scale dependences for several sets of structure functions were calculated in Refs. .
The experimental data for $`x_F`$-distributions of D-mesons produced in $`\pi N`$ interactions are in agreement with the parton model distributions for bare quarks, as one can see in Fig. 3 taken from . It means that the fragmentation processes are not important here, or they are compensated by, say, recombination processes. The shape of $`x_F`$-distributions does not depend practically on the mass of $`c`$-quark.
The data on one-particle $`p_T`$-distributions,including the hadronic colliders data for the case of beauty production, also can be described by the NLO parton model, see .
## 3 Azimuthal correlations and failure of the collinear approximation
The azimuthal angle $`\varphi `$ is defined as an opening angle between two produced heavy quarks, projected onto a plane perpendicular to the beam. In the LO parton model this angle between them is exactly $`180^o`$. In the case of NLO parton model a distribution over $`\varphi `$ angle appears .
The investigation of such distributions is very important. In one-partricle distributions, the sum of LO and NLO contributions of the parton model practically coinsides with the LO contribution multiplied by $`K`$-factor. So we can not control the magnitudes of LO and NLO contributions separately. In the case of azimuthal correlations all difference from the trivial $`\delta (\varphi \pi )`$ distribution comes from NLO contribution.
The experimental data on azimuthal correlations are claimed (see and Refs. therein) to be in disagreement with the NLO predictions, for the cases of charm pair hadro- and photoproduction at fixed target energies. The level of disagreements can be seen in Fig. 4 (solid histograms) taken from Ref. . These data can be described , assuming the comparatively large intrinsic transverse momenta of incoming partons ($`k_T`$ kick). For each event, in the longitudinal centre-of-mass frame of the heavy quark pair, the $`Q\overline{Q}`$ system is boosted to rest. Then a second transverse boost is performed, which gives the pair a transverse momentum equal to $`\stackrel{}{p}_T(Q\overline{Q})+\stackrel{}{k}_T(1)+\stackrel{}{k}_T(2)`$; $`\stackrel{}{k}_T(1)`$ and $`\stackrel{}{k}_T(2)`$ are the transverse momenta of the incoming partons, which are chosen randomly, with their moduli distributed according to
$$\frac{1}{N}\frac{dN}{dk_T^2}=\frac{1}{k_T^2}\mathrm{exp}(k_T^2/k_T^2).$$
(4)
The dashed and dotted histograms in Fig. 4 correspond to the NLO parton model prediction, supplemented with the $`k_T`$ kick, with $`k_T^2=0.5`$ GeV<sup>2</sup> and $`k_T^2=1`$ GeV<sup>2</sup>, respectively. We see that with $`k_T^2=1`$ GeV<sup>2</sup> it is possible to describe the data.
However the large intrinsic transverse momentum significantly changes one-particle $`p_T`$-distributions of heavy flavour hadrons, which were in good agreement with the data. The solid curves in Fig. 5 taken from represent the NLO parton model predictions for charm quarks $`p_T`$-distributions which are in agreement with the data. The effect of the $`k_T`$ kick results in a hardening of the $`p_T^2`$ spectrum. On the other hand, by combining the $`k_T`$ kick with $`k_T^2=1`$ GeV<sup>2</sup> and the Peterson fragmentation , the theoretical predictions slightly undershoot the data (dot-dashed curves).
The $`k_T`$ kick can only very weakly change the $`x_F`$-distributions of produced $`c`$-quarks, Fig. 3, and after accounting the fragmentation these distributions can become too soft.
Let us consider why the conventional NLO parton model with collinear approximation works reasonably for one-particle diustributions, and, at the same time, it is in disagreement with the data on azimuthal correlations.
The contribution of the processes of Fig. 1, which governs the heavy quark production can be written<sup>3</sup><sup>3</sup>3We omit for simplicity all factors which are non-essential here. as a convolution of initial transverse momenta distributions, $`I(q_{1T})`$ and $`I(q_{2T})`$, with squared modulo of perturbative QCD matrix element, $`|M(q_{1T},q_{2T},p_{1T},p_{2T})|^2`$ :
$$\sigma _{QCD}(Q\overline{Q})d^2q_{1T}d^2q_{2T}I(q_{1T})I(q_{2T})|M(q_{1T},q_{2T},p_{1T},p_{2T})|^2.$$
(5)
Now there are two possibilities:
i) the essential values of initial transverse momenta are much smaller than the transverse momenta of produced heavy quarks, $`q_{iT}p_{iT}`$, and
ii) all transverse momenta are of the same order, $`q_{iT}p_{iT}`$.
In LO for the first case (i) we can wait that one-dimentional $`p_T`$ distributions should be more broad than the distributions on the transverse momenta of the quark pair, because in LO $`q_{1T}+q_{2T}=p_{1T}+p_{2T}`$. NLO gives here a correction, numerically not very large, see Fig. 2. In this case one can replace both the initial distributions $`I(q_{iT})`$ by $`\delta `$-functions, $`\delta (q_{iT})`$. This reduces the expression (5) to the very simplified one :
$$\sigma _{coll.}(Q\overline{Q})|M(0,0,p_{1T},p_{2T})|^2,$$
(6)
in total agreement with Weizsaecker-Williams approximation in QED.
In the second case (ii) the distributions on the transverse momenta of the quark pair should be the same, or even more broad than the one-particle $`p_T`$-distributions, and namely this situation is realised . In this case we can not wait a priory that the Weizsaecker-Williams approximation will give good results, however it works quite reasonably in the case of one-particle distributions. In the case of distribution on the transverse momentum of the heavy-quark pair, $`p_T^2(Q\overline{Q})`$ we measure (only approximately in NLO) the distribution over the sum of transverse momenta of incident gluons. In this case it seems to be senseless to replace the distributions $`I(q_{1T})`$ and $`I(q_{2T})`$ by $`\delta `$-functions, and to expect a reasonable agreement with the data. The same can be said about the azimuthal correlations.
The $`k_t`$ kick effectively accounts for the transverse momenta of incident partons. It uses the expression which can be written symbollically as
$$\sigma _{kick}(Q\overline{Q})I(q_{1T})I(q_{2T})|M(0,0,p_{1T},p_{2T})|^2,$$
(7)
and the main difference from the general QCD expression Eq. (5) is that due to absence of $`q_{iT}`$ in the matrix element the values of $`k_T^2`$ in Eq. (4) should be different for different processes and kinematical regions. The reason is that in Eq. (5) the values $`I(q_{iT})`$ decrease at large $`q_{iT}^2`$ as a weak power (see next Sect.), i.e. comparatively slowly, and more important is the $`q_{iT}^2`$ dependence of the matrix element. In the last one the corrections of the order of $`q_{iT}^2/\mu ^2`$, where $`\mu ^2`$ is the QCD scale, are small enough when $`q_{iT}^2/\mu ^2<<1`$ and they start to suppress a matrix element value when $`q_{iT}^2/\mu ^21`$.
## 4 Heavy quark production in semihard approximation
Let us consider another approach, when the transverse momenta of incident gluons in the small-$`x`$ region appear from the diffusion of transverse momenta in the gluon evolution<sup>4</sup><sup>4</sup>4The similar approach based on $`k_T`$-factorization formulae can be found, for example, in Refs. .. This diffusion is described by the function $`\phi (x,q^2)`$ determined as
$$\phi (x,q^2)=4\sqrt{2}\pi ^3\frac{d[xG(x,q^2)]}{dq^2},$$
(8)
where $`G(x,q^2)`$ is usual gluon structure function.
In principle the function $`\phi (x,q^2)`$ which determine the probability to find gluon with fixed value of transverse momentum, $`q_T`$, depends on three variables, $`x`$, $`q_T`$ and gluon virtuality $`q^2`$. However at small $`x`$ in LLA $`q_T^2q^2`$, and it leads to comparatively weak dependence of $`\phi (x,q^2)`$ on $`q_T^2`$ (strongly different from exponential dependence in Eq. (4)) due to weak $`q^2`$-dependences of phenomenological structure functions.
The exact expression for gluon $`q_T`$-distributions can be obtained, as a solution of the nonlinear evolution equation. The calculations result in difference from our $`\phi (x,q^2)`$ function only about 10-15%.
The matrix element $`M_{QQ}`$ accounting for the gluon virtualities and polarizations is much more complicate than the parton model one. That is why we consider only LO contribution of the subprocess $`ggQ\overline{Q}`$. The differential cross section of heavy quarks hadroproduction has the form
$$\frac{d\sigma _{pp}}{dy_1^{}dy_2^{}d^2p_{1T}d^2p_{2T}}=\frac{1}{(2\pi )^8}\frac{1}{(s)^2}d^2q_{1T}d^2q_{2T}\delta (q_{1T}+q_{2T}p_{1T}p_{2T})$$
$$\times \frac{\alpha _s(q_1^2)}{q_1^2}\frac{\alpha _s(q_2^2)}{q_2^2}\phi (q_1^2,y)\phi (q_2^2,x)|M_{QQ}|^2.$$
(9)
Here $`s=2p_ap_b`$ and $`y_{1,2}^{}`$ are the quarks’ rapidities in the hadron-hadron c.m.s. frame.
Eq. (9) enables to calculate straightforwardly all distributions concerning heavy flavour one-particle, or pair production. However there exists a problem coming from infrared region. Gluon structure function in Eq. (8) is not determined at small virtualities, so the function $`\phi (x,q_2^2)`$ is unknown at the small values of $`q_2^2`$ and $`q_1^2`$. To solve this problem we will use the direct consequence of Eq. (8)
$$xG(x,q^2)=xG(x,Q_0^2)+\frac{1}{4\sqrt{2}\pi ^3}_{Q_0^2}^{q^2}𝑑q_1^2\phi (x,q_1^2),$$
(10)
and rewrite the integrals in the Eq. (9) as the sum of four contributions. The first one is determined by the product of two gluon distributions, $`G(x,Q_0^2)`$ and $`G(y,Q_0^2)`$, and it is the same as the conventional LO parton model expression. Next three terms contain the corrections to the parton model. If the initial energy is not high enough, the first term dominates. In the case of very high energy the first term can be considered as a small corrections, and our results are differ from the conventional ones. In the cases when the collinear approximation is available, our results only slightly differ from the parton model.
It is illustrated in Fig. 6 taken from , where we compare our calculations of $`b`$-quark $`p_T`$-distributions with the experimental results of CDF collaboration and with two variants of parton model calculations. The distributions over the azimuthal angle $`\varphi `$ can be found in .
The essential values of $`q_{1T}`$ and $`q_{2T}`$ in our calculations increase with increase the value of $`p_T^{min}`$ of detected $`b`$-quark. In the language of $`k_T`$ kick it means that the values of $`k_T^2`$ will be also increased.
## 5 Conclusion
The experimental results on total cross sections for charm and beauty production are in agreement with the conventional parton model predictions, using reasonable values of QCD scales and quark masses. The data on $`x_F`$ and $`p_T`$ distributions are also in reasonable agreement with parton model without any fragmentation functions<sup>5</sup><sup>5</sup>5The $`p_T`$ distributions of charm photoproduction measured by E691 Coll. are more hard than the NLO parton model predictions , that can be considered as an argument for including a non-perturbative fragmentation function. However, the $`p_T`$ slope of the calculated spectrum depends strongly on the charm quark mass, and the measured charm production cross section has very strange energy dependence ..
Moreover, the shapes of one-particle LO and NLO distributions practically coinside. It means that instead of calculation the NLO contributions, it is enough to calculate only LO contributions, and rescale them using K-factor taken, say, from the calculated ratio of total cross sections.
In the case of distribution over the total transverse momentum of the produced quark pair, or azimuthal correlations, the conventional NLO parton model with collinear approximation can not describe the data. The $`k_T`$ kick allows one to describe these data, however the problems with one-particle $`p_T`$-distributions appear, which can be solved by introducing the fragmentation function. The last way should produce the problems in description of $`x_F`$-spectra. Moreover, it seems that the $`k_T^2`$ values should depend on the process and the kinematical regions.
Another possibility to solve the problems of initial transverse momenta is to use semihard theory, accounting for the virtual nature of the interacting gluons, as well as their transverse motion and different polarizations. It results in a qualitative differences with the LO parton model predictions . In Ref. the values of $`F_2(x,Q^2)`$ were calculated using phenomenological gluon structure functions, and the infrared contributions to $`F_2(x,Q^2)`$ were investigated in details. The possible estimations of the shadow corrections in the processes of heavy flavour production can be found in Refs. .
I am grateful to M.G.Ryskin and A.G.Shuvaev for multiple discussions, to M.L.Mangano for very useful critical comments and to E.M.Levin who participated at the early stage of this activity. I thank the Organizing Commitee of HERA Monte Carlo Workshop for financial support. This work is supported by grant NATO OUTR.LG 971390.
|
no-problem/9904/cond-mat9904382.html
|
ar5iv
|
text
|
# Photon assisted tunneling in a superconducting SET transistor induced by the Josephson oscillations of a SQUID
## Abstract
Josephson oscillations generated by a SQUID were used to measure photon assisted tunneling in a superconducting single electron tunneling (SET) transistor. The SQUID was fabricated only a few microns from the SET transistor. The close proximity of the generator to the SET transistor allowed for a continuous sweep of the excitation frequency from 10 GHz to 190 GHz. The amplitude of the Josephson oscillations were tuned by adjusting the flux through the SQUID loop while the frequency was independently adjusted by controlling the bias voltage across the SQUID. Current peaks which could be attributed to photon assisted tunneling in the SET transistor were observed for bias voltages less than $`4\mathrm{\Delta }/e`$.
Many artificially structured microsystems such as quantum dots or single-electron tunneling transistors have characteristic energies on the order of 1 meV. The frequencies needed to probe these energies spectroscopically are tens to hundreds of GHz. This is an inconvenient frequency range since there are few tunable sources of radiation available in this range which can be coupled to a sample at low temperature while isolating the sample from room temperature blackbody radiation.
One suitable radiation source is a Josephson junction. When a Josephson junction is biased above the critical current, voltage oscillations appear with a frequency $`f=2eV_{dc}/h`$, where $`V_{dc}`$ is the dc voltage across the junction. By adjusting the average voltage across the junction, the frequency of these oscillations can be tuned. We have used such Josephson oscillators to measure photon assisted tunneling in a superconducting single electron tunneling (SET) transistor.
In the experiments, two Al/AlO<sub>x</sub>/Al Josephson junctions were used in parallel as in a superconducting quantum interference device (SQUID). Normally SQUIDs are used to measure magnetic flux but in this case we used the flux modulation of the critical current of the SQUID to tune the power of the radiation that was generated. The SQUID loop formed by the two junctions was 10 $`\mu `$m<sup>2</sup> resulting in an inductance of about 3 pH. The Josephson junctions had critical currents of 1 $`\mu `$A and were shunted by a 30 $`\mathrm{\Omega }`$ Pt shunt. Figure 1a shows a schematic diagram of the circuit and Fig. 1b is an electron microscope image of the sample showing the coupling capacitors $`C_1`$ and $`C_2`$ and the positions of the SET transistor and the SQUID. The generator was located just 10 $`\mu `$m from the SET transistor.
It was possible to tune the frequency of the Josephson oscillations continuously from 10 GHz to 190 GHz. An essential feature of this experiment is that the entire circuit is smaller than a wavelength at all relevant frequencies. This means that the coupling of the radiation from the generator to the sample circuit was not strongly frequency dependent as is typically the case with larger circuits where microwave resonances affect the coupling. Another advantage of the small size of the circuit is that the entire circuit could be placed in a microwave tight enclosure with filtered dc feedthroughs. This prevents blackbody radiation in the GHz range from interfering with the experiment. The frequency and amplitude of the SQUID generator were adjusted with dc signals and the detection current flowing through the SET was also a dc signal.
Because the impedance of the SQUID is much smaller than the resistance of the SET transistor, the SQUID generator can best be modeled as an rf voltage source with a source impedance equal to the resistive shunt. The SQUID generates an rf voltage,
$$V_{rf}(t)=\frac{R\left(I^2I_c^2\right)}{II_c\mathrm{sin}(\omega t)}.$$
(1)
Here $`I_c`$ is the critical current of the SQUID, $`I`$ is the SQUID bias current, $`R`$ is the shunt resistance, and $`\omega =\frac{2eI_cR}{\mathrm{}}\sqrt{\left(I/I_c\right)^21}`$.
The SET transistor consisted of two low capacitance Al/AlO<sub>x</sub>/Al tunnel junctions in series. Since these junctions were much smaller than the junctions in the SQUID, it was necessary to fabricate the SET junctions in a separate fabrication step from the SQUID junctions. The junctions had capacitances of $`C_1`$ = 52 aF and $`C_2`$ = 95 aF, resistances of $`R_1`$ = 400 k$`\mathrm{\Omega }`$ and $`R_2`$ = 90 k$`\mathrm{\Omega }`$, and a gate capacitance of $`C_g`$ = 7.3 aF. A contour map of the current through the SET transistor as a function of gate voltage and bias voltage in the absence of applied radiation is shown in Fig. 2. Very little current was observed for bias voltages less than $`4\mathrm{\Delta }/e`$. No supercurrent was observed because of the high resistance of the sample and the Josephson-quasiparticle cycle was not observed because $`e/C_\mathrm{\Sigma }>4\mathrm{\Delta }/e`$. Here $`C_\mathrm{\Sigma }=C_1+C_2+C_g`$
is the total capacitance of the island. There is a fairly sharp onset of current that corresponds to the threshold for sequential quasiparticle tunneling in the SET transistor. This threshold forms the zigzag pattern that appears between $`4\mathrm{\Delta }/e`$ and $`4\mathrm{\Delta }/e+e/C_\mathrm{\Sigma }`$ in Fig. 2.
When a bias voltage is applied across the SQUID,
Josephson oscillations occur which enable additional tunnel processes through the SET transistor via photon assisted tunneling. Figure 3 shows the current through the SET transistor when 55 GHz radiation is applied. The threshold for sequential quasiparticle tunneling corresponds to the sharp rise of current at bias voltages at 0.8 mV. The two higher plateaus arise from photon assisted tunneling through one of the junctions and then direct tunneling through the other junction. The lower plateau is due to photon assisted tunneling through both junctions.
Figure 4 shows the positions of the thresholds for photon assisted tunneling for four frequencies (0 GHz, 30 GHz, 79 GHz, and 160 GHz) in the $`V_bV_g`$ plane. Currents above 100 pA are plotted as contours and the positions of smaller current peaks and thresholds are plotted as solid points. The gray diamond shaped regions indicate where each quasiparticle is assisted by one photon in tunneling through each junction. This corresponds to the lower diamond-shaped plateau in Fig. 3. A constant voltage across junction 1 corresponds to a line with a slope of $`C_g/C_2`$ in the $`V_bV_g`$ plane and a constant voltage across junction 2 correspond to a line with a slope of $`C_g/(C_1+C_g)`$ in the $`V_bV_g`$ plane. For this reason the thresholds all follow lines with these slopes. The positions of the photon assisted tunneling thresholds correspond to a constant voltage across one of the junctions that is $`nhf/e`$ less than the voltage at which the original threshold occurs. Here $`n`$ is an integer that indicates the order of the photon assisted tunneling peak.
Figure 4a shows the current when no microwave radiation is applied. In addition to the threshold for the
sequential tunneling of quasiparticles, small current peaks occur when the BCS density of states of the island and one of the leads aligns. These current peaks arise from the tunneling of thermally activated quasiparticles in a process known as singularity matching and are indicated by the triangles. A line of current peaks was also observed when the voltage across one of the junctions equals $`2\mathrm{\Delta }/e`$. If the bias across a junction is greater than $`2\mathrm{\Delta }/e`$, then tunneling can proceed through that junction at the normal rate. These current peaks are indicated by the circles.
Figures 4b-4d show the current through the SET transistor when high frequency radiation from the SQUID is applied. Again the solid triangles represent the singularity matching peaks and the circles represent the transition to normal rate tunneling in one of the two junctions. The solid squares represent the thresholds for photon assisted tunneling. These squares line along the dotted lines which indicate the positions of thresholds for photon assisted tunneling that are expected for the frequency of the radiation applied. The thresholds for sequential quasiparticle tunneling and for singularity matching do not move as a function of frequency but the thresholds for photon assisted tunneling scale linearly with frequency.
Photon assisted tunneling should also show a characteristic scaling as a function of rf power. The amplitude of the $`n`$th order peak should modulate as $`J_n^2\left(\frac{eV_{rf}}{hf}\right)`$, where $`J_n(x)`$ is the $`n`$th order Bessel function. In the experiment, the zeroth order peak decreases and the higher order peaks increase as the power increases as would be expected from the Bessel function relations at low rf
powers. However, the power generated by the SQUID was not sufficient to see the amplitude of the peaks oscillate as one would expect for high rf powers. From Eqn. 1, one can estimate that the amplitude of the rf signal should be approximately $`V_{rf}I_cR`$ = 30 $`\mu `$V. The maximum amplitude of the rf voltage that a SQUID can deliver is limited by the $`I_cR`$ product of the Josephson junctions. Since the $`I_cR`$ product for overdamped aluminum junctions is limited to about 0.1 mV, these Josephson generators will not generate enough power to observe oscillations in the amplitude of the photon assisted tunneling peaks for frequencies higher than about 3 GHz. Higher microwave powers could be achieved by using arrays of junctions or other materials. Typical $`I_cR`$ products for overdamped niobium or high $`T_c`$ junctions are about 1 mV. This leads to oscillations in the amplitude of the photon assisted tunneling peaks for frequencies up to about 30 GHz for single junction oscillators.
In conclusion, high frequency oscillations generated by a SQUID were used to irradiate a superconducting SET transistor. This radiation induced photon assisted tunneling in the SET transistor over a span of frequencies from 10 GHz to 190 GHz. A linear scaling with the frequency was observed. An advantage of using SQUIDs as the rf source is that microwave resonances in the measurement circuit were not observed so that it was possible to tune the frequency continuously. Placing the rf oscillator on the same chip as the sample to be measured also makes it possible to enclose the entire measurement circuit in a microwave tight box which minimizes external interference such as the blackbody radiation of room temperature equipment.
###### Acknowledgements.
Support from Esprit project 22953, CHARGE, is gratefully acknowledged.
|
no-problem/9904/quant-ph9904063.html
|
ar5iv
|
text
|
# MAXIMUM-LIKELIHOOD ALGORITHM FOR QUANTUM TOMOGRAPHY
## 1 Introduction
Several years after the first demonstration , optical homodyne tomography has become a well established tool for measuring quantum statistical properties of optical radiation. What is particularly fascinating, this technique provides practical means to visualise the measured quantum state in the form of the Wigner function. This success is a result of combining a complete quantum mechanical measurement of field quadratures with a filtered back-projection algorithm used in medical imaging.
The purpose of this contribution is to trace some other analogies between quantum state reconstruction and classical image processing, with the motivation to develop novel numerical methods for quantum tomography. Our interest will be focused on imperfect detection , which has deleterious effects on quantum interference phenomena exhibited by non-classical states . As we will discuss in Sec. 2, such effects can be related in the phase space representation to image blurring. Restoration of blurred images is a well known problem in classical imaging, and a number of methods has been developed for this purpose. Specifically, we shall briefly describe in Sec. 3 the Richardson algorithm (known also in statistics as the expectation-maximization algorithm ), which provides an effective iterative procedure to perform image deblurring.
An interesting question is, whether classical deblurring methods can be transferred to quantum tomography. We discuss this, in the case of the Richardson algorithm, in Sec. 4. The answer is not straightforward: the Richardson algorithm assumes positive definiteness of the original undegraded image, and this condition is not satisfied by the quantum mechanical Wigner function. We will show that this difficulty can be overcome by expressing the Wigner function in terms of the phase space displaced photon distribution. This yields an iterative algorithm for reconstructing the Wigner function, which incorporates compensation for detection losses in a numerically stable way .
## 2 Imperfect detection and image blurring
Homodyne detection is a realization of the quantum mechanical measurement of field quadratures only in the idealized loss-free limit. In practice, a fraction of the signal field is always lost due to the mode-mismatch and the non-unit efficiency of photodiodes. The homodyne statistics collected using a realistic setup is described by the distribution :
$$h(x;\theta )=_{\mathrm{}}^{\mathrm{}}\text{d}x^{}\frac{1}{\sqrt{\pi (1\eta )}}\mathrm{exp}\left(\frac{(x\sqrt{\eta }x^{})^2}{1\eta }\right)x_\theta ^{}|\widehat{\rho }|x_\theta ^{},$$
(1)
where $`\theta `$ is the local oscillator phase, $`\eta `$ characterizes the overall detector efficiency, and $`x_\theta ^{}|\widehat{\rho }|x_\theta ^{}`$ denote diagonal elements of the density matrix $`\widehat{\rho }`$ in the quadrature basis.
Application of the back-projection transformation to $`h(x;\theta )`$ yields, instead of the Wigner function, a generalized phase space quasidistribution function $`P(q,p;s)`$ with the ordering parameter $`s=(1\eta )/\eta `$. This function can be expressed as a convolution of the Wigner function $`W(q,p)`$ with a gaussian function:
$$P(q,p;|s|)=\text{d}q^{}\text{d}p^{}\frac{1}{\pi |s|}\mathrm{exp}\left(\frac{1}{|s|}[(qq^{})^2+(pp^{})^2]\right)W(q^{},p^{}).$$
(2)
Thus what we reconstruct from imperfect homodyne data, is a blurred version of the Wigner function. The question is, whether we can get rid of this blurring in numerical processing of experimental data.
A similar problem appears in the following classical context: suppose we observe an image using an imperfect apparatus (for example ill-matched glasses), which generates some blurring. Such blurring can be described by a so-called point spread function specifying the shape generated by a single point of the original image. The observed degraded image is consequently given by a convolution of the original image with the point spread function. Using this language, we can assign the following names to the terms of Eq. (2):
| $`W(q^{},p^{})`$ | | original image |
| --- | --- | --- |
| $`{\displaystyle \frac{1}{\pi |s|}}\mathrm{exp}\left({\displaystyle \frac{1}{|s|}}[(qq^{})^2+(pp^{})^2]\right)`$ | | point spread function |
| $`P(q,p;|s|)`$ | | degraded image |
The common problem now is the restoration of the original image from the degraded one, assuming that the point spread function is known.
## 3 Image restoration
An analytical way to deconvolve Eq. (2) is to apply the Fourier transform, which maps a convolution onto a direct product. Dividing both the sides by the Fourier-transformed point spread function and evaluating the inverse Fourier transform thus yields an explicit expression for the original image. However, this procedure is very sensitive to statistical fluctuations and numerical truncation errors, which makes its practical application a very delicate matter. These problems have been noted also in the context of quantum tomography .
The numerical instability of the Fourier deconvolution has led to the development of techniques dedicated for image restoration. The basic observation is that statistical noise does not allow us to specify precisely the original image that was ’behind’ the blurred data. In principle, the measured degraded image could could be generated by a variety of original images. However, comparing various original images we intuitively expect that some of them were more likely to generate the measured data than other ones. The maximum-likelihood methodology quantifies this intuition, and selects as an estimate the original image which maximizes the likelihood.
In order to discuss this idea in detail, we shall consider a discretized version of Eq. (2):
$$p_\nu =\underset{n}{}A_{\nu n}w_n,$$
(3)
where $`w_n`$ is the original image, $`A_{\nu n}`$ is the point spread function, and $`p_\nu `$ is the degraded image. Note that this formulation is more general compared to Eq. (2), because it allows the point spread function to be of different form for each ’element’ of the original image indexed with $`n`$. The likelihood can be quantified using the function
$$=\underset{\nu }{}p_\nu \mathrm{ln}\left(\underset{n}{}A_{\nu n}w_n\right)$$
(4)
which has a rigorous derivation when the degraded image is observed as a histogram of events governed by Poissonian statistics. The likelihood function for quantum measurement has been discussed in Ref. , where its maximization has been proposed as a method for quantum state estimation.
In classical imaging, it is natural to assume that $`w_n`$, as well as $`A_{\nu n}`$ as a function of $`\nu `$ for each $`n`$, are positive definite distributions with sum equal to one. Under these assumptions, it is possible to find the maximum of the likelihood function $``$ via simple iteration:
$$w_n^{(i+1)}=\underset{\nu }{}p_\nu \frac{A_{\nu n}w_n^{(i)}}{_mA_{\nu m}w_m^{(i)}},$$
(5)
which is the essence of the Richardson algorithm for image restoration . A simple heuristic derivation of this algorithm can be found in Ref. .
Of course, the maximum-likelihood approach is not a magic wand solving unconditionally the problem of image restoration. With increasing blurring, the quality of the reconstructed image worsens, and the convergence of the iterative algorithm may be very slow. In many cases, however, it offers superior performance compared to the Fourier deconvolution technique.
## 4 Quantum tomography
An obvious difficulty with applying the iterative restoration algorithm to quantum tomography is that the object to be reconstructed in the quantum case, i.e. the Wigner function, is not positive definite. Nevertheless, there are some other quantum mechanical reconstruction problems, where positivity constraints appear in a natural way. An interesting and nontrivial example is determination of the photon number distribution from random phase homodyne statistics . The relation between the phase-averaged homodyne statistics and the photon number distribution $`\varrho _n`$ is given by
$$\frac{1}{2\pi }_\pi ^\pi \text{d}\theta h(x;\theta )=\underset{n}{}A_n(x)\varrho _n,$$
(6)
where $`A_n(x)`$ describe contributions to the homodyne statistics generated by different Fock states $`|n`$. This formula, after discretization of $`x`$, is exactly of the form assumed in Eq. (3). Thus we arrive at the following formal analogy:
| $`\varrho _n`$ | | original image |
| --- | --- | --- |
| $`A_n(x)`$ | | point spread function |
| $`{\displaystyle \frac{1}{2\pi }}{\displaystyle _\pi ^\pi }\text{d}\theta h(x;\theta )`$ | | degraded image |
which allows us to apply directly the iterative reconstruction algorithm . In this procedure, there is one a priori parameter: it is the cut-off of the distribution $`\varrho _n`$ specifying the maximum number of photons.
Reconstruction of the photon distribution may seem to be quite distant from the starting point of our considerations, which was deblurring of the Wigner function. However, let us recall that the Wigner function can be represented as an alternating sum of the photon distribution $`\varrho _n(q,p)`$ corresponding to the phase space displaced state $`\widehat{D}^{}(q,p)\widehat{\varrho }\widehat{D}(q,p)`$:
$$W(q,p)=\frac{1}{\pi }\underset{n=0}{\overset{\mathrm{}}{}}(1)^n\varrho _n(q,p)$$
(7)
Obviously, we can apply this formula to evaluate the Wigner function at $`q=p=0`$. What would be of interest, is the generalization of the maximum-likelihood algorithm to determination of an arbitrarily displaced photon distribution $`\varrho _n(q,p)`$. This would yield a numerically stable procedure for reconstructing the Wigner function from homodyne statistics, even in the case of the non-unit detection efficiency.
Surprisingly, this generalization is quite straightforward. The basic observation is that the displacement transformation has a simple effect on the homodyne statistics, shifting it by $`\sqrt{\eta }(q\mathrm{cos}\theta +p\mathrm{sin}\theta )`$ for a given local oscillator phase $`\theta `$. Consequently, we have the relation
$$\frac{1}{2\pi }_{\mathrm{}}^{\mathrm{}}\text{d}\theta h(x+\sqrt{\eta }q\mathrm{cos}\theta +\sqrt{\eta }p\mathrm{sin}\theta ;\theta )=\underset{n}{}A_n(x)\varrho _n(q,p),$$
(8)
which can be readily implemented in the iterative algorithm. Thus, we have arrived at the following two-step algorithm for quantum tomography: for a given phase space point $`(q,p)`$, construct the phase-averaged histogram according to the right-hand side of Eq. (8), and iteratively reconstruct $`\varrho _n(q,p)`$. Then, calculate the value of the Wigner function according to Eq. (7). In Fig. 1 we present Monte Carlo simulated reconstruction of the Wigner function for a Schrödinger cat state detected using a homodyne setup with the efficiency $`\eta =90\%`$.
The standard filtered back-projection algorithm used in quantum tomography is based on the inverse Radon transform, whose integral kernel is singular. Therefore, a regularization scheme is necessary in processing experimental data. This aspect has a counterpart in the maximum-likelihood algorithm. In this approach, we have the cut-off for the photon distribution which can be regarded as a regularization parameter. Its proper choice is an important matter. Setting it too small perturbes the reconstructed photon distribution. On the other hand, the larger number of $`\varrho _n`$s, the slower iterations converge. The expected shape of the photon distribution can be quite easily predicted, if we roughly know the region of the phase space where the Wigner function is localized. For this purpose it is useful to recall the semiclassical picture of projections on Fock states as rings in the phase space characterized by the radius $`\sqrt{2n}`$. The photon distribution is nonzero over the range of $`n`$ for which the corresponding rings overlap with the localization region for the Wigner function.
Truncation of the photon distribution can be introduced as a regularization scheme also in the standard linear reconstruction approach. In such a scheme, the Wigner function would be evaluated from a finite part of the photon distribution reconstructed using the pattern function technique. However, properties of the reconstructed photon distribution make this method very sensitive to statistical noise. This can be straightforwardly seen in the most regular case of $`\eta =1`$. For large $`n`$, the error of $`\varrho _n`$ tends to a fixed nonzero value , and moreover deviations of consecutive $`\varrho _n`$s are strongly anticorrelated. Consequently, the alternating sum defined in Eq. (7) accumulates the statistical uncertainty of the photon distribution .
Let us also note that in principle we could apply the restoration algorithm to homodyne histograms described by Eq. (1), in order to obtain deblurred quadrature distributions $`x_\theta |\widehat{\varrho }|x_\theta `$. In this case statistical fluctuations would play a much more significant role. The advantage of using Eq. (8) is that we use the whole available sample of experimental data to determine a relatively small number of parameters $`\varrho _n(q,p)`$.
Acknowledgements I would like to thank Prof. G. Mauro D’Ariano for his hospitality during my stay in Pavia supported by INFM, and Prof. Krzysztof Wódkiewicz for comments on the manuscript. I have benefited a lot from discussions with Zdenek Hradil on maximum-likelihood methods in quantum state reconstruction. This research is supported by Komitet Badań Naukowych, grant 2P03B 013 15.
|
no-problem/9904/cond-mat9904200.html
|
ar5iv
|
text
|
# Density of states in coupled chains with off-diagonal disorder
Since the works of Dyson, it has been known that one dimensional systems with off-diagonal disorder may show anomalous behavior in their density of states and their localization properties . This anomalous behavior has reappeared in many reincarnations, ranging from Dyson’s original application of masses coupled with springs, to the random hopping model , quantum XY spin chains , supersymmetric quantum mechanics , one-dimensional classical diffusion in a random medium , narrow-gap semiconductors , and charge density wave materials . In all these systems, it has been shown that both the density of states and the localization length diverge at zero energy. These two divergencies are closely linked by a theorem by Thouless .
Recently, it was found that the localization length of a wire consisting of $`N`$ of such chains shows a remarkable dependence on the parity of $`N`$ : At zero energy, i.e. at the center of the band, the localization length diverges for odd $`N`$ (odd $`N`$ includes the pure one-dimensional case $`N=1`$), while wavefunctions remain localized for even $`N`$. Moreover, it was found that staggering in the hopping parameter (as the consequence of a Peierls instability), may lead to an additional set of delocalized states for all $`N>1`$, regardless of parity . For comparison, in the pure one-dimensional case $`N=1`$ staggering always enhances localization. In this paper we consider the density of states (d.o.s.) for the $`N`$-chain wire with off-diagonal disorder. Unlike in the one-dimensional case, this is a problem that requires separate attention, since the Thouless theorem which links d.o.s. and localization length does not hold for $`N>1`$ coupled chains. Moreover this problem is also of relevance for the random flux model, which is a special example of off-diagonal randomness, and whose localization properties and density of states near $`\epsilon =0`$ are the subject of an ongoing debate .
To be specific, we consider the d.o.s. for the Schrödinger equation
$$\epsilon \psi _n=\psi _n=t_n^{}\psi _{n+1}t_{n1}^{}\psi _{n1},$$
(1)
where $`\psi _n`$ is an $`N`$-component wavefunction and $`t_n`$ an $`N\times N`$ hopping matrix. The index $`n=1,\mathrm{},L`$ labels the site index along the wire. The length of the wire, measured in units of the lattice spacing, is $`L`$. The reason why this system can display anomalous behavior near zero energy is the existence of a particle-hole or chiral symmetry, which is absent in the presence of on-site disorder: Under a mapping $`\psi _n(1)^n\psi _n`$, the Hamiltonian $``$ changes sign, $``$. As a result, the eigenvalues of $``$ occur in pairs $`\pm \epsilon `$. There are two mechanisms by which the chiral symmetry is known to affect the d.o.s. near $`\epsilon =0`$. First, level repulsion of the eigenvalue $`\epsilon `$ with its mirror image $`\epsilon `$ causes a universal suppression of the d.o.s. near $`\epsilon =0`$ . This suppression appears on the scale of a level spacing, is independent of the geometry, but becomes unimportant in the thermodynamic limit. The second mechanism, for which several different descriptions exist , is special for a (quasi) one-dimensional geometry, and survives in the thermodynamic limit. It is responsible for the divergence of the d.o.s. in the pure one-dimensional case. Below we present a calculation of the d.o.s. in the multi-chain case, combining ideas from the Fokker-Planck approach to localization in multichannel quantum wires and the calculation of the d.o.s. in the one-dimensional random hopping model .
We first state our main results. We find that the parity dependence that was previously obtained for the localization properties is also present in the density of states. For odd $`N`$, the d.o.s. diverges at zero energy according to
$$\rho (\epsilon )\frac{w^2}{|\epsilon \mathrm{ln}^3(w^2/\epsilon )|},N\text{ odd},$$
(2)
where $`w`$ is a dimensionless parameter governing the randomness in the $`t_n`$. The bandwidth is chosen as the unit of energy. In Eq. (2) and in the remainder of the paper we assume $`\epsilon >0`$. The d.o.s. for $`\epsilon <0`$ then follows from $`\rho (\epsilon )=\rho (\epsilon )`$. The form of the divergence (2) is independent of whether time-reversal symmetry is broken or not, i.e. of whether the hopping matrices $`t_n`$ are generically real or complex. For even $`N`$, in contrast, the density of states strongly depends on the presence or absence of time-reversal symmetry (labeled by the parameter $`\beta =1`$ or $`2`$, respectively),
$$\rho (\epsilon )|\epsilon /w^2|^{\beta 1}|\mathrm{ln}(w^2/\epsilon )|,N\text{ even}.$$
(3)
In the presence of time-reversal symmetry, $`\rho (\epsilon )`$ shows a logarithmic divergence as $`\epsilon 0`$, while in the absence of time-reversal symmetry a pseudo-gap is opened, $`\rho (\epsilon )`$ vanishes at $`\epsilon =0`$. Such a strong dependence on time-reversal symmetry is a remarkable result. An effect of comparable magnitude appears in the suppression of the gap by a weak magnetic field in a normal metal in the proximity of a superconductor. In the remainder of this paper we derive the results (2) and (3) assuming a specific statistical model for the hopping matrices $`t_n`$ in Eq. (1). The details of this model are not relevant, the singular behavior of the d.o.s. near zero energy being governed by the fundamental symmetries of the Hamiltonian $``$ only. As an illustration of the general validity of our result, we close with a comparison to numerical simulations.
As we are interested in the d.o.s. of the random hopping model (1) in the thermodynamic limit, the boundary conditions at the two ends of the chain are not important. For convenience we choose hard wall boundary conditions, $`t_0=t_L=0`$, $`L`$ being the length of the chain. We can solve Eq. (1) recursively in terms of a sequence of hermitian $`N\times N`$ matrices $`a_n`$,
$`a_n\psi _{n+1}`$ $`=`$ $`t_n^{}\psi _n,a_n=t_n^{}\left(\epsilon +a_{n1}\right)^1t_n^{}.`$ (4)
The boundary condition at $`n=0`$ implies $`a_0=0`$. Evaluating the Schrödinger equation at $`n=L`$ then yields that $`\epsilon `$ is an eigenvalue of $``$ if and only if
$$\epsilon \psi _L=t_{L1}^{}\psi _{L1}=a_{L1}\psi _L,$$
(5)
i.e. if $`a_{L1}`$ has an eigenvalue $`\epsilon `$, or alternatively, in view of Eq. (4), if $`a_L`$ has a diverging eigenvalue.
As a statistical model for the $`t_n`$ that contains all the relevant symmetries, we parameterize the $`t_n`$ in terms of the generators
$$t_n=e^{W_n},$$
(6)
where $`W_n`$ is a real (complex) matrix for $`\beta =1`$ ($`2`$), and choose the matrices $`W_n`$ from independent Gaussian distributions with mean and variance given by
$`(W_n)_{\mu \nu }^{}(W_n)_{\rho \sigma }^{}`$ $`=`$ $`{\displaystyle \frac{1}{2}}w^2\beta [\delta _{\mu \rho }\delta _{\nu \sigma }{\displaystyle \frac{1\eta }{N}}\delta _{\mu \nu }\delta _{\rho \sigma }],`$ (7)
$`(W_n)_{\mu \nu }`$ $`=`$ $`{\displaystyle \frac{1}{2}}(1)^n\mathrm{\Delta }\delta _{\mu \nu }.`$ (8)
Here $`\eta `$ governs the fluctuations of $`\text{tr}W_n`$ and $`\mathrm{\Delta }`$ measures the staggering of the hopping parameter. ($`\mathrm{\Delta }`$ is the gap size that the staggering would induce in the absence of disorder.) We assume that the disorder and staggering are weak, and that the energy is small compared to the bandwidth ($`w^2,\mathrm{\Delta },\epsilon 1`$).
The matrix $`a_L`$ has eigenvalues $`\alpha _\mu `$, that can be parameterized as $`\alpha _\mu \mathrm{tan}(\varphi _\mu /2)`$ ($`\mu =1,\mathrm{},N`$). As we have discussed below Eq. (5), the energy $`\epsilon `$ is an eigenvalue of the Hamiltonian, if and only if there is an angle $`\varphi _\mu `$ with $`\varphi _\mu =\pi `$. For general $`\epsilon `$, however, none of the $`\varphi _\mu `$ will be equal to $`\pi `$. Nevertheless, we can use the angles $`\varphi _\mu `$ to compute the (disorder averaged) density of states. Hereto we first note that $`a_{L+2}=a_L`$ in the absence of disorder, staggering, and for $`\epsilon =0`$. Then, taking disorder, staggering, and a finite energy into account, and considering the length $`L`$ as a fictitious “time”, the angles $`\varphi _\mu `$ perform a Brownian motion on the unit circle, which is such that upon increasing $`L`$, they move around the circle in positive direction. The rate at which the $`\varphi _\mu `$ pass through $`\pi `$ as we increase $`L`$, i.e. their current, equals the number of states per unit length $`N(\epsilon )`$ with energy between $`0`$ and $`\epsilon `$. (This is a generalization of the node-counting theorem used to compute the d.o.s. for $`N=1`$ .) For comparison, we remark that, in the absence of disorder, the angles $`\varphi _\mu `$ move around at a constant speed $`\epsilon `$, resulting in a constant d.o.s. With disorder, their motion acquires a random (Brownian) component, which dramatically affects their average speed, and hence the density of states, as we shall see below.
For a quantitative description a different parameterization of the eigenvalues $`\alpha _\mu `$ proves to be more convenient,
$$\alpha _\mu =\mathrm{tan}(\varphi _\mu /2)=e^{u_\mu }.$$
(9)
The variables $`u_\mu `$ are restricted to the two branches $`\text{Im}u_\mu =0`$ and $`\text{Im}u_\mu =\pi `$ in the complex plane, see Fig. 1a. We refer to these as lower and upper branches, respectively. Noting that the $`u_\mu `$ are related to the angles $`\varphi _\mu `$ on the unit circle, we see that a (fictitious) particle with coordinate $`u_\mu `$ that vanishes on one of the branches at $`\pm \mathrm{}`$ reappears at the opposite branch, as indicated by the arrows in Fig. 1a. Upon increasing $`L`$ by two, the $`u_\mu `$ change according to $`u_\mu u_\mu +\delta u_\mu `$, where, to lowest order in $`w`$, $`\epsilon `$, and $`\mathrm{\Delta }`$, the average and variance of the increments $`\delta u_\mu `$ are
$`\delta u_\mu `$ $`=`$ $`2\epsilon \mathrm{cosh}u_\mu +2\mathrm{\Delta }+w^2\beta {\displaystyle \underset{\nu \mu }{}}\mathrm{coth}{\displaystyle \frac{u_\mu u_\nu }{2}},`$ (10)
$`\delta u_\mu \delta u_\nu `$ $`=`$ $`4w^2\left[\delta _{\mu \nu }(1\eta )/N\right].`$ (11)
Taking $`L`$ as a continuous variable, their distribution function $`P(u_1,\mathrm{},u_N;L)`$ obeys the Fokker-Planck equation
$`{\displaystyle \frac{P}{L}}`$ $`=`$ $`w^2{\displaystyle \underset{\mu ,\nu }{}}{\displaystyle \frac{}{u_\mu }}\left(\delta _{\mu \nu }{\displaystyle \frac{1\eta }{N}}\right)J{\displaystyle \frac{}{u_\nu }}J^1P`$ (13)
$`{\displaystyle \underset{\mu }{}}{\displaystyle \frac{}{u_\mu }}\left(\epsilon \mathrm{cosh}u_\mu +\mathrm{\Delta }\right)P,`$
$`J`$ $`=`$ $`{\displaystyle \underset{\mu <\nu }{}}\mathrm{sinh}^\beta [(u_\mu u_\nu )/2].`$ (14)
For large $`L`$ the solution of Eq. (14) acquires a steady state carrying a current $`j(\epsilon )`$ equal to the integrated d.o.s. $`N(\epsilon )`$ in the thermodynamic limit. Unfortunately, except in the case $`N=1`$, where a solution in closed form is possible , it is notoriously problematic to find the steady-state solution of a Fokker-Planck equation of the type (14), due to the lack of detailed balance . Instead, below we present a qualitative analysis of the Brownian motion process described by Eq. (14), illustrating the mechanisms that lead to the anomalous behavior of the d.o.s. near $`\epsilon =0`$. We give a detailed account of the case $`\eta =1`$, followed by a brief discussion of the general case $`\eta 1`$ and a comparison with numerical simulations.
Let us first identify the relevant parameters in the Brownian motion process (14). There are $`N`$ fictitious Brownian “particles” with coordinates $`u_\mu `$ ($`\mu =1,\mathrm{},N`$) and diffusion coefficient $`D=w^2`$. Three forces $`F_w`$, $`F_\mathrm{\Delta }`$, and $`F_\epsilon `$ act on the particles, arising from the presence of disorder, staggering, and energy, respectively. The force $`F_w=(w^2\beta /2)\mathrm{coth}[(u_\mu u_\nu )/2]`$ is a repulsive interaction, with a hard-core on the same branch and a soft core on different branches. (We adopt a convention where the particles have unit mobility.) Staggering causes a constant force field $`F_\mathrm{\Delta }=\mathrm{\Delta }`$ that favors motion to the right on both branches. Finally, $`F_\epsilon =\pm \epsilon \mathrm{cosh}u`$ pushes the particles to the left (right) on the lower (upper) branch and thus causes the nonzero steady-state current. For small energies $`\epsilon w^2`$, motion is governed by the force $`F_\epsilon `$ for large $`|u|`$, $`|\text{Re}u|u_{\mathrm{max}}`$, where
$$u_{\mathrm{max}}=\mathrm{ln}(w^2/\epsilon ),$$
(15)
while diffusion and the forces $`F_w`$ and $`F_\mathrm{\Delta }`$ are dominant for small $`u`$, $`|\text{Re}u|u_{\mathrm{max}}`$. Following Ref. , we now approximate our model by truncating the branches at $`|\text{Re}u|=u_{\mathrm{max}}`$, see Fig. 1b, and adding a one-way move towards the upper (lower) branches at the end-points, see Fig. 1b. These simplifications don’t alter the functional dependencies of $`\rho (\epsilon )`$ for $`\epsilon 0`$, though they may affect prefactors.
For $`N=1`$, the “particle” with coordinate $`u`$ needs a time $`(2u_{\mathrm{max}})^2/2D`$ to diffuse to the end point of a branch, as one can verify from a solution of the one-dimensional diffusion equation on the line $`u_{\mathrm{max}}<u<u_{\mathrm{max}}`$, with hard wall boundary conditions at $`u=u_{\mathrm{max}}`$ and absorbing wall boundary conditions at $`u=u_{\mathrm{max}}`$ . Hence the current is $`j(\epsilon )=N(\epsilon )D/4u_{\mathrm{max}}^2`$, and after differentiation with respect to $`\epsilon `$, one finds Eq. (2) for the density of states. Diffusion effectively speeds up the particle, explaining the enhancement of the d.o.s. relative to the clean case . For $`N=2`$, the picture is completely different. As a result of their mutual repulsion, the two particles with coordinates $`u_1`$ and $`u_2`$ get trapped near the end points, say at $`u_1=u_{\mathrm{max}}`$ and $`u_2=u_{\mathrm{max}}+i\pi `$. Now the particles have to diffuse out of their traps against their repulsive interaction, until they eventually meet at $`\text{Re}u_1=\text{Re}u_2`$ and the repulsive force $`F_w`$ starts to favor travel (see Figs. 2a and b). Such a process costs a large time, which can be calculated from the diffusion equation for two particles on a line with hard wall boundary conditions at $`\text{Re}u_1=u_{\mathrm{max}}`$ and $`\text{Re}u_2=u_{\mathrm{max}}`$ and absorbing boundary conditions at $`\text{Re}u_1=\text{Re}u_2`$. We find a current $`j(\epsilon )=(4u_{\mathrm{max}}F_w^3/D^2)\mathrm{exp}(2F_wu_{\mathrm{max}}/D)`$, resulting in the d.o.s. (3). (The prefactor $`u_{\mathrm{max}}`$ arises from the degeneracy of the meeting point on the line.) We conclude that the d.o.s. for two coupled chains has only a logarithmic singularity for real hopping disorder, and a pseudogap for complex disorder. The strong $`\beta `$-dependence of the d.o.s. stems from the $`\beta `$-dependence of the interaction force $`F_w`$.
The qualitative behavior of the d.o.s. for general $`N>2`$ depends crucially on the parity of $`N`$ and closely resembles the scenarios we have outlined above for $`N=1`$ and $`N=2`$. If $`N`$ is even, all particles get “trapped” near the ends of the branches, half of them on the lower branch near $`u_{\mathrm{max}}`$, and half of them on the upper branch near $`u_{\mathrm{max}}`$, like in the case $`N=2`$, cf. Fig. 2d. The repulsive interaction force $`F_w`$ is smallest for the two particles that are closest to the origin. These two particles dominate the current, resulting in a d.o.s. of the form (3). If $`N`$ is odd, on the other hand, the picture is like that of the case $`N=1`$, see Fig. 2c. All particles get trapped at the two ends, except for one “free” particle, for which diffusion is not slowed down by the interaction forces (the repulsive forces from the other particles cancel exactly). Hence, for odd $`N`$, the d.o.s. is of the form (2). This even-odd effect is reminiscent of that found for the conductance at $`\epsilon =0`$, where the existence of a delocalized state could be attributed to the existence of a similar “free” transmission eigenvalue .
The effect of staggering $`\mathrm{\Delta }`$ is to add a constant force pointed to the right on both branches. As a function of the staggering strength, the system alternates between behavior corresponding to even and odd $`N`$. Repeating the above analysis, we find that with staggering, the d.o.s. shows the maximum (2) if $`\mathrm{\Delta }=(N+12j)w^2\beta /2`$, $`j=1,\mathrm{},N`$, whereas it shows the minimum (3) if $`\mathrm{\Delta }=(N2j)w^2\beta /2`$, $`j=1,\mathrm{},N1`$. For all other values of $`\mathrm{\Delta }`$, for $`\epsilon 0`$ we have
$$\rho (\epsilon )\epsilon ^{\beta n1},n=\underset{j=1,\mathrm{},N}{\mathrm{min}}\left|\frac{2\mathrm{\Delta }}{w^2\beta }N1+2j\right|.$$
(16)
The case of arbitrary $`\eta 1`$ is not much different from the case $`\eta =1`$ we considered above; up to prefactors the $`\epsilon `$-dependence of the d.o.s. is not changed. The only exception is the case $`\eta =0`$, $`N=2`$, when the center of mass $`u_1+u_2`$ is pinned. As a result, the degeneracy giving rise to the logarithm in Eq. (3) is lifted, and the logarithmic prefactor vanishes.
We conclude with a comparison to numerical simulations for the d.o.s. in a quantum wire on a square lattice with a width between $`N=1`$ and $`N=4`$, and a length $`L`$ between $`10^3`$ and $`10^5`$. For $`\beta =1`$ and also for $`N=1`$ the hopping amplitudes are taken from a uniform distribution in the interval \[0.5,1.5\], while for $`\beta =2`$ the random flux model is used, where the randomness is introduced only via the random phases of the hopping amplitudes. Results of an average over $`4\times 10^4`$$`10^6`$ disorder realizations are shown in Fig. 3. The agreement with our theoretical results, Eqs. (2) and (3), is excellent.
We would like to thank C. W. J. Beenakker, D. S. Fisher, and B. I. Halperin for discussions. PWB acknowledges support by the NSF under grant nos. DMR 94-16910, DMR 96-30064, and DMR 97-14725. CM acknowledges a fellowship from the Swiss Nationalfonds. The numerical computations were performed at the Yukawa Institute Computer Facility.
|
no-problem/9904/astro-ph9904156.html
|
ar5iv
|
text
|
# Weak Gravitational Lensing and Cluster Mass Estimates
## 1. Introduction
Clusters of galaxies are excellent cosmological probes. Their size suggests that their constituents provide a fair sample of the universe. Their structure and hydrodynamic state provide information on their formation and evolution, and thus upon models of structure formation. Measurements of the abundance of clusters of a given mass allows constraint of the amplitude of mass fluctuations in the universe; measurements of abundance evolution can be used to constrain the mass density $`\mathrm{\Omega }_\mathrm{m}`$.
Many of these approaches depend upon some knowledge of the mass, or mass distribution, of the cluster. Most techniques for measuring cluster masses are based upon some equilibrium assumption which relates the cluster mass to an observable such as the temperature of the intracluster plasma or the velocity dispersion of cluster galaxies. Recently, however, it has become feasible to measure the surface density distribution of a cluster through observations of weak gravitational lensing of the background galaxy field by the cluster. An attractive quality of this technique is that no assumptions about the dynamical or thermodynamical state of the cluster components need be made. In other words, weak lensing analyses probe the mass distribution directly.
However, analyses of the mass distribution of a cluster drawn from weak lensing observations are not without problems (for a recent review, see Mellier Mellier (1999)). Many of these relate to details of the procedure adopted to go from the observed ellipticity distribution to the mass, or from instrumental effects. We will not discuss these in this Letter. We are interested here in the degree to which weak lensing mass estimates of clusters are affected by lensing from material outside but nearby the cluster. This is a source of systematic error which is not well understood (though it has been alluded to in earlier work, e.g. Miralda-Escude ME (1991); Wambsganss, Cen & Ostriker WamCenOst (1998)).
Since clusters form in overdense regions, the volume surrounding a cluster is likely to contain infalling overdense matter (Gunn & Gott GunGot (1972)). This infalling matter could add to an observed lensing signal and result in an overestimate of the cluster mass. It is possible that this bias could be quite severe. In modern hierarchical models of structure formation, such as the Cold Dark Matter (CDM) model, numerical simulations imply that clusters form primarily at the intersections of filaments in a web of cosmic structure, accreting additional diffuse mass and smaller collapsed objects that drain along these filaments. It is thus reasonable to expect a beaded filamentary structure surrounding most clusters of galaxies. Tentative observational evidence of filamentary structure near clusters has been reported recently (Kull & Boehringer KulBoe (1999), Kaiser et al. MS0302 (1999)). A filament lying near the line-of-sight will also lens the background galaxies, and therefore contribute spuriously to the lensing signal. If the observed lensing signal were attributed solely to the cluster, the inferred cluster mass could be much larger than its actual mass.
In this Letter, we use mock clusters from numerical simulations to explore the significance of the systematic mass overestimation induced by the additional lensing signal of both diffuse and filamentary material near the cluster. We find that this effect can be quite significant and must be considered when evaluating the results of lensing mass reconstruction techniques. In the next section, we describe the numerical data and our procedure for evaluating the errors introduced into cluster mass estimates by nearby material. Section 3 describes the results of our analyses. We discuss these results and outline plans for future study at the end.
## 2. Method
Weak lensing mass reconstruction techniques produce a map of shear or convergence. These are integrals of the mass along the line-of-sight times a projection kernel. This kernel is quite wide in the redshift direction, scaling as $`D_LD_{LS}/D_S`$ where $`D_L`$ is the distance from the observer to the lens, $`D_S`$ from the observer to the source and $`D_{LS}`$ from the lens to the source<sup>1</sup><sup>1</sup>1For a distribution of source distances, one takes an appropriate integral of this expression. (Mellier Mellier (1999)). Under the assumption that the cluster is the most massive object along the line-of-sight and is well localized in space (the thin-lens approximation), the convergence map is proportional to the projected surface density map of the lensing cluster itself.
Any additional mass located near the cluster and along the line-of-sight will also contribute to the lensing signal. Since the kernel is such a slowly varying function of distance, material even large distances from the cluster will contribute within the thin lens approximation. For a source at $`z=1`$ and a cluster at $`z0.5`$ the kernel changes by only 1% within $`\pm 40h^1`$Mpc of the cluster in a universe with $`\mathrm{\Omega }_\mathrm{m}=0.3=1\mathrm{\Omega }_\mathrm{\Lambda }`$, with similar results in other cosmologies. As a result, weak lensing observations will probe the projected density of a cluster plus all of the material in its vicinity. Note that this “nearby” material is essentially “at” the redshift of the cluster for the purposes of lensing, and so cannot be distinguished by using extra information such as source redshifts.
To study the effect of the surrounding mass upon the projected mass inferred from lensing observations of the simulated clusters, we have examined the mass distribution around several clusters of galaxies extracted from a large cosmological simulation. The simulated clusters were taken from the X-Ray Cluster Data Archive of the Laboratory for Computational Astrophysics of the National Center for Supercomputing Applications (NCSA), and the Missouri Astrophysics Research Group of the University of Missouri. To produce these clusters, a particle-mesh N-body simulation incorporating adaptive mesh refinement was performed in a volume $`256h^1`$Mpc on a side. Regions where clusters formed were identified; for each cluster, the simulation was then re-run (including a baryonic fluid) with finer resolution grids centered upon the cluster of interest. In the adaptive mesh refinement technique, the mesh resolution dynamically improves as needed in high-density regions. The “final” mesh scale at the highest resolution was $`15.6h^1`$ kpc, allowing good resolution of the filamentary structure around the cluster. Inside the cluster, the characteristic separation between the smallest–mass particles, given by $`d=\left(4\pi r_{200}^3/3N\right)^{1/3}`$ with $`N`$ the number of particles inside the region, was approximately 86 kpc for all three clusters examined here. The code is described in detail in Norman & Bryan (TheCode (1999)).
The clusters used here were extracted at $`z=0`$ from simulations of a $`\mathrm{\Lambda }`$CDM model, with parameters $`\mathrm{\Omega }_\mathrm{m}=0.3`$, $`\mathrm{\Omega }_\mathrm{B}=0.026`$, $`\mathrm{\Omega }_\mathrm{\Lambda }=0.7`$, $`h=0.7`$, and $`\sigma _8=0.928`$. In this Letter we observe these clusters as if they were at $`z=0.5`$. In future work we plan to investigate the dependence of these results on cosmology and on cluster redshift.
Since the number density of rich clusters is approximately $`\varphi _{}10^5`$ Mpc<sup>-3</sup>, the typical separation between them is $`\varphi _{}^{1/3}40h^1`$Mpc. This is a characteristic scale for filaments: volumes containing a cluster and with one-dimensional extent $`40h^1`$Mpc should also contain much of the nearby filamentary structure. Three such volumes, containing a rich cluster (Clusters 0, 2 and 4) as well as satellites and filaments, were extracted from the archive. We selected clusters that did not appear to be mergers or have a large secondary mass concentration nearby. Such systems might be excluded observationally from studying, for instance, the galaxy line-of-sight velocity distribution. For each volume, the “center” of our cluster was determined using a maximum-density algorithm. As the extracted volumes were not spherical, it was possible that some lines of sight could contain more mass than others simply by geometry. To avoid such biases we restrict our analysis to particles that lay within the largest sphere, centered on the cluster, which was contained entirely within the extracted volume. The radii of these spheres, $`R_{\mathrm{sphere}}`$, are listed in Table 1, along with other properties of the clusters. Note that these radii are large compared to the projected values of $`r_{200}`$ obtained for each cluster; thus no significant radial surface density gradient is introduced by a decreasing chord length through the sphere with radius. It is also important to note that since our volumes are by necessity limited, our results should be interpreted as a lower limit to the size of the effect; the magnitude of the lensing kernel is still significant at the edge of our spherical volume.
Each of the clusters we examined was surrounded by a large amount of mass. Most of this material appeared by eye to be collapsed into “beads” along a filamentary structure, although a small number of clumps could be found outside the filaments. We show a projection of a fraction of the points from the simulation of cluster 4 in Fig. 1. The filamentary structure and satellites are easily evident. Note that this filamentary structure extends well beyond our radius $`R_{\mathrm{sphere}}`$. No single projection can show the full 3D nature of the structure, in which the filamentarity is even more apparent. Since much of this mass is at low density it is unlikely it would be a site for galaxy formation or otherwise emit light. Thus this structure would not be easy to constrain by observations of redshifts near the cluster.
We observed each of the three selected clusters from $`10,000`$ randomly chosen viewing angles. For each cluster and viewing angle, the projected surface density map was constructed and used to estimate $`r_{200}`$, the radius within which the mean interior density contrast is 200. In three dimensions, this radius is defined in terms of the enclosed mass by
$$M\left(<r_{200}\right)=200\times \left(\frac{4\pi }{3}\right)\mathrm{\Omega }_\mathrm{m}\rho _{\mathrm{crit}}r_{200}^3.$$
(1)
The projected estimate of $`r_{200}`$ was extracted from the surface density map by considering the radius of the circle, centered on the cluster, which contained the amount of mass given by Eq. (1) above, i.e.
$$_0^{2\pi }d\theta _0^{r_{200}}RdR\mathrm{\Sigma }(R,\theta )=M\left(<r_{200}\right)$$
(2)
with $`\mathrm{\Sigma }(R,\theta )`$ the surface density on the map in terms of a two-dimensional radius $`R`$. This radius was compared to the cluster’s true $`r_{200}`$, extracted from the three–dimensional mass distribution. The ratio of the projected mass to true mass is given simply by the cube of the ratio of the estimated value of $`r_{200}`$ to the true value. For each cluster, a value of this ratio was obtained for each viewing angle.
## 3. Results
Our main result is displayed in Fig. 2, where we show the distribution of projected vs. “true” cluster mass in each of the three simulated clusters. We have checked that the features in the histogram do not come from shot noise due to discrete particles in the simulation. However, the “spikiness” is due to a discrete number of objects in the neighborhood of the cluster. A small lump of matter near the cluster will project entirely within $`r_{200}`$ for a fraction of the lines of sight. For any such line of sight, the effect on the projected value of $`r_{200}`$ is identical.
We expect the ratio $`M_{200}/M_{\mathrm{true}}`$ to be greater than unity since only additional mass can be included in the projection. The size of the smallest offset from unity for clusters 0 and 4 is approximately twice what would be expected for material uniformly distributed at the mean density. This suggests that matter near the cluster is itself clustered and at higher than mean density. The width of the histogram in Fig. 2, as a fraction of the true mass, depends on the true mass of the cluster. Though we have only a few clusters in this study, it appears that the mass in nearby material is not proportional to the cluster mass, thus the relative effect of this material is smaller the larger the cluster. The total mass in the sphere, $`M_{\mathrm{tot}}`$, is also listed in Table 1 for reference.
The signal shown in Fig. 2 comes from (primarily filamentary) material outside the cluster and is not the well known projection effect arising from cluster asphericity. To verify this, we repeated the procedure described above for a subset of particles aimed at selecting the cluster alone. This was done by first selecting out particles with a local density contrast of greater than 70 (chosen because density profiles near $`r^2`$ reach a local density contrast near 70 at a mean interior density contrast of 200); a small sphere containing the cluster but little nearby material was then cut out of this subset. The histogram produced by viewing the clearly prolate cluster at a large number of randomly chosen viewing angles produced a much narrower distribution, with a maximum offset of less than $`10\%`$ in the mass ratio and a mean offset of approximately half that value.
While it is beyond the scope of this Letter to perform a detailed modelling of any observational weak lensing strategy, we show in Fig. 3 two sample profiles obtained from aperture densitometry on our noiseless projected mass maps. Specifically, for Cluster 4, we show the profile along the lines of sight giving the largest and smallest $`r_{200}`$, for comparison. The $`\zeta `$ statistic is the mean value of the convergence, $`\kappa `$, within a disk of radius $`r_1`$ minus the mean value within an annulus $`r_1rr_2`$ (Fahlman et al. Fahlman (1994), Kaiser K95 (1995)). Here we calculate $`\zeta `$ directly from the projected mass, though observationally it would be computed from the tangential shear. We have taken $`r_2=800^{\prime \prime }`$. Such a large radius is not (currently) achievable observationally, but it minimizes the impact of objects nearby the cluster and provides a lower limit on the size of the projection effect. We have explicitly checked that reducing the radius to half this value does not change our result.
In calculating the convergence $`\kappa `$, the cluster was again assumed to be at a redshift of $`0.5`$, with the lensed sources at a redshift of $`1.0`$. In any real observation, of course, the lensed sources will span a range of redshifts. For material very close to the cluster, such as here, this will not affect our conclusions, and the error introduced by incorrectly modelling the redshift distribution of the background sources is not the subject of this work. In addition to being easy to estimate, the $`\zeta `$ statistic is robust and minimizes contamination by foreground mass (Mellier Mellier (1999)) because it is insensitive to the sheet mass degeneracy. This does not, however, remove the sensitivity to clustered material, as can be seen in Fig. 3. The mass which would be inferred from Fig. 3 along two lines of sight differ by a factor of $`1.7`$.
While the distribution of the projection effect varies from cluster to cluster, it seems clear that positive biases in projected mass of $`20\%`$ are typical, with biases above $`30\%`$ not uncommon. Furthermore, we emphasize again that these estimates are in fact lower limits; some lines of sight through the untruncated volume produced overestimates as large as $`80\%`$ or more. While appropriate modeling of a mean density profile outside $`r_{200}`$ (drawn perhaps from simulations such as these) can be used to produce an unbiased estimator, the width of these histograms implies a large amount of scatter around such an estimator of the true (unprojected) mass. It is clear that this effect can be quite significant and must be taken into account when attempting to understand the results of mass reconstruction analyses.
## 4. Conclusions
Clusters of galaxies are part of the large-scale structure of the universe and observations of them should be considered within this context. The filamentary structure near a cluster can contain a reasonable fraction of the mass of a cluster in tenuous material. Should a filament lie close to the line-of-sight to a cluster it will contribute to the weak lensing signal and positively bias the projected mass. We have shown that such a bias could easily be 30% (see Fig. 2).
Weak lensing remains one of the best methods for determining the mass of clusters of galaxies. However methods which obtain the mass from estimates of the projected surface density must consider the effect outlined in this Letter. This is clearly of particular significance for attempts to determine the mass function of clusters directly through surveys of weak lensing-determined masses.
We have not attempted to address the detailed question of how much this filamentary signal would affect a particular reconstruction algorithm; the answer is no doubt algorithm dependent. We hope to return to this issue in future work, as well as to consider the effect of cosmological model and evolution with cluster redshift.
The authors would like to thank Greg Bryan and Greg Daues for assistance in understanding the archive data, and Gordon Squires and Albert Stebbins for useful conversations on lensing. This research was supported by the NSF.
|
no-problem/9904/hep-ph9904384.html
|
ar5iv
|
text
|
# SEARCH FOR THE NEUTRINO MAGNETIC MOMENT IN THE NON-EQUILIBRIUM REACTOR ANTINEUTRINO ENERGY SPECTRUM
## SEARCH FOR THE NEUTRINO MAGNETIC MOMENT IN THE NON-EQUILIBRIUM REACTOR ANTINEUTRINO ENERGY SPECTRUM
V.I. Kopeikin, L.A. Mikaelyan, V.V. Sinev
RRC ”Kurchatov Institute”, Kurchatov Sq., 1, Moscow-123182, Russia.
Resume of seminar talks given at Kurchatov Institute, March 1999 y.
We study the time evolution of the typical nuclear reactor antineutrino energy spectrum during reactor ON period and the decay of the residual antineutrino spectrum after reactor is stopped. We find that relevant varia-tions of the soft recoil electron spectra produced via weak and magnetic $`\stackrel{~}{\nu }_e,e`$ scattering process can play a significant role in the current and planned searches for the neutrino magnetic moment at reactors.
## 1 Introduction
Efforts are currently being done to observe the neutrino magnetic moment below the limit $`\mu _\nu <210^{10}\mu _B`$ that was found in previous $`\stackrel{~}{\nu }_e,e`$ scattering experiments at SAVANNAH RIVER, KRASNOYARSK and ROVNO reactors . The KURCHATOV-PNPI collaboration is plan-ning for KRASNOYARSK new studies of low kinetic energy recoil electrons in $`\stackrel{~}{\nu }_e,e`$ experiment with a Si semiconductor multi detector. The MUNU collaboration experiment at BUGEY with a gas TPC chamber is in the final state of preparation .
The dominant contribution to the soft recoil electron produced in the $`\stackrel{~}{\nu }_e,e`$ scattering comes from the low energy part of the reactor $`\stackrel{~}{\nu }_e`$ energy spectrum. This part of the spectrum is strongly time dependent: it never comes to saturation during the reactor operating run and does not vanish after the reactor is shut down, the time when the background is usually measured.
Here we consider the time evolution of the typical reactor $`\stackrel{~}{\nu }_e`$ energy spectrum and discuss relevant variations of the $`\stackrel{~}{\nu }_e,e`$ scattering recoil electron spectra.
## 2 TIME VARIATION OF THE REACTOR <br>ANTINEUTRINO SPECTRUM
1. Three components contribute the reactor $`\stackrel{~}{\nu }_e`$ energy spectrum
$`\rho \left(E\right)/fissMeV`$:
$$\rho \left(E\right)=^F\rho \left(E\right)+^U\rho \left(E\right)+\mathrm{\Delta }\rho \left(E\right).$$
(1)
Here, the term $`{}_{}{}^{F}\rho \left(E\right)`$ represents the radiation of the $`{}_{}{}^{235}U`$, $`{}_{}{}^{239}Pu`$, $`{}_{}{}^{238}U`$ and $`{}_{}{}^{241}Pu`$ fission fragments. The second term stems from the chain of the $`\beta `$-decays which follow neutron radiative capture in $`{}_{}{}^{238}U`$:
$${}_{}{}^{238}U(n,\gamma )^{239}U\frac{\beta }{23.5min}^{239}Np\frac{\beta }{2.36days}^{239}Pu$$
(2)
The last term in Eq.(1) accounts for the antineutrinos (and neutrinos) induced by the neutron interactions with other materials in the reactor core. As discussed in Ref. this term adds no more than 1% to the total reactor $`\stackrel{~}{\nu }_e`$ flux and is disregarded here.
Till recently the term $`{}_{}{}^{F}\rho `$ has traditionally been identified with the reactor $`\stackrel{~}{\nu }_e`$ spectrum. The contribution of the chain (2) antineutrinos is however quite sizable for all reactors where neutrino experiments are running or planned. In the ROVNO, BUGEY and CHOOZ PWR-type reactors about 1.2 $`\stackrel{~}{\nu }_e`$ per fission come from this source.
2. For each of the four isotopes $`{}_{}{}^{235}U`$, $`{}_{}{}^{239}Pu`$, $`{}_{}{}^{238}U`$ and $`{}_{}{}^{241}Pu`$ the evolution of the neutrino spectra $`\rho (E,t)`$ have been calculated vs time t since the beginning of the fission process. The subsequent decay of the spectra during reactor OFF period have been followed.
The base used for these calculations involves data on 571 fission fragments, data on nuclear isomers and delayed neutron emission.
3. Calculations show that in PWR reactors about 2/3 of all antineu-trinos belong typically to the energy range below E = 1,5 MeV. This part of the $`\stackrel{~}{\nu }_e,e`$ spectrum $`\rho (E,330)`$ and it’s component due to fission fragments $`\rho (E,330)`$ at the end of the reactor 330 day ON period are presented in Fig.1.
The evolution of the $`\stackrel{~}{\nu }_e,e`$ spectrum during PWR reactor ON period and its decay after the reactor is shut down is illustrated in Fig.2a,b.
## 3 RECOIL ELECTRON ENERGY SPECTRA
1. The recoil-electron spectra $`S^W`$(T) and $`S^M`$(T) in $`cm^2/MeVfiss.`$ units, (T is the recoil-electron kinetic energy) for weak (W) and magnetic (M) scattering of reactor antineutrinos are found by convolution the $`\stackrel{~}{\nu }_e,e`$ spectra $`\rho \left(E\right)`$ with the differential cross sections for monoenergy antineu-trino:
$$\frac{d\sigma ^W}{dT}=g_F^2\frac{m}{2\pi }\left[4x^4+\left(1+2x^2\right)\left(1\frac{T}{E}\right)^22x^2\left(1+2x^2\right)\frac{mT}{E^2}\right]$$
(3)
$$\frac{d\sigma ^M}{dT}=\pi r_0^2\frac{\mu _\nu ^2}{\mu _B^2}\left(\frac{1}{T}\frac{1}{E}\right),$$
(4)
where m is the electron mass, $`g_F^2\frac{m}{2\pi }=4.3110^{45}cm^2/MeV`$, $`x=sin^2\theta _W=0.232`$ is the Weinberg parameter, $`\pi r_0^2=2.49510^{25}cm^2`$.
2. Calculated recoil-electron spectra $`S^W`$(T,330) and $`S^M`$(T,330) at the end of the reactor ON period are shown in Fig.3. In searches for the neutrino magnetic moment, weak $`\stackrel{~}{\nu }_e,e`$ scattering plays the role of the reactor-correlated background. We note, that in order to keep this background at sufficiently low level one should try to study recoil electrons at not too high energies. As an example, the recoil electron energies $`T>100`$ keV seem to be ”high” to search for $`\mu _\nu =210^{11}\mu _B`$ while the range $`T<700`$ keV is tolerably low for $`\mu _\nu =510^{11}\mu _B`$.
Calculated time variations of the recoil-electron spectra during PWR reactor ON and OFF periods for weak and magnetic scattering are presented in Fig.4a,b.
## 4 DISCUSSION AND CONCLUSIONS
In practice the situation is not as simple as presented above. There occur deviations from the standard operating schedule: reactor can be stopped for a few days, or it can operate at a reduced level of power etc. For each particular experiment a comprehensive analysis of the reactor operation data should be carried out including all details.
The main result of this study is that effects due to neutrino relaxa-tion play not negligible role in sensitive searches for the neutrino magne-tic moment at reactors.
## Acknowledgments
This work is supported by RFBR, projects 97-02-16031, 96-15-96640.
|
no-problem/9904/physics9904050.html
|
ar5iv
|
text
|
# Short distance relativistic atom-atom forcesContribution for third issue of the Bulletin of the Asia Pacific Center for Theoretical Physics, Seoul, Korea.
About fifty years ago two important papers appeared describing novel interactions. One, by Casimir, discussed the case of two interacting walls and the other, by Casimir and Polder, considered the interactions between an atom and a wall and between two atoms . The history and some of the many interesting aspects of these interactions, their derivations, and their importance in field theory and atomic and molecular physics are summarized elsewhere . Indeed, recent experiments give strong evidence of the reality of both the atom-wall and the wall-wall interactions predicted in those two papers.
Here I will focus on the connection between the QED result of Casimir and Polder and other results for relativistic atom-atom interactions at short distances of the order of, say, $`20a_0`$. The interaction between an electron and an ion will also be considered. For typical atomic systems these relativistic effects are very small corrections to the non-relativistic potentials arising from the van der Waals and Coulomb interactions for, respectively, the atom-atom interaction and the electron-ion interaction.
Casimir and Polder used QED and old-fashioned perturbation theory and their result was subsequently duplicated by other authors with different methods, cf. . One way to write their result for the interaction potential $`V(R)`$ between two spherically symmetric atoms separated by a distance $`R`$ is as a one-dimensional integral
$$V(R)=\frac{1}{\pi R^6}_0^{\mathrm{}}𝑑\omega \mathrm{exp}(2\alpha \omega R)[\alpha _d(i\omega )]^2P(\omega \alpha R),$$
(1)
with $`P(x)=x^4+2x^3+5x^2+6x+3`$ and $`\alpha `$ the fine structure constant. Atomic units $`\mathrm{}=m=e=1`$ are used throughout and in these units $`c=1/\alpha `$. The function $`\alpha _d(i\omega )`$ is the dynamic electric dipole polarizability at imaginary frequency,
$$\alpha _d(i\omega )=\underset{u}{}f_u/[(E_uE_0)^2+\omega ^2],$$
(2)
where $`f_u`$ is the oscillator strength of state $`u`$ and $`E_uE_0`$ is the transition frequency between the states $`u`$ and $`0`$, with $`0`$ denoting the ground state of the atom, and the summation in (2) includes an integration over continuum states. The function $`\alpha _d(i\omega )`$ is a smooth function of $`\omega `$ with no singularities. The limit of $`V(R)`$ for asymptotically large separations of the atoms can be obtained from the Casimir-Polder integral (1) yielding
$$V(R)\frac{23}{4\pi }\frac{[\alpha _d(0)]^2}{\alpha R^7},R\mathrm{}.$$
(3)
What about the limit for small $`R`$? The result is
$$V(R)\frac{3}{\pi R^6}_0^{\mathrm{}}𝑑\omega [\alpha _d(i\omega )]^2,R137,$$
(4)
and upon integration (4) yields
$$V(R)C_6/R^6,$$
(5)
where $`C_6`$ is the van der Waals constant expressed as a double sum over oscillator strengths and the correct form of the atom-atom interaction at short distances (say $`20a_0`$) is reproduced. The van der Waals constant is of vast importance for all sorts of molecular spectroscopic and atomic collision problems, of course. The result for two H atoms is $`C_6=\mathrm{6.499\hspace{0.17em}026\hspace{0.17em}705}\mathrm{}`$ and for studies of atomic collisions at ultracold temperatures $`C_6`$ plays a crucial role in characterizing the interactions . So it is nice to see QED connect nicely with non-relativistic molecular quantum mechanics. What is the next correction?
If one more term is retained in the small $`R`$ expansion then
$$V(R)=\frac{C_6}{R^6}+\alpha ^2\frac{W_4}{R^4}+𝒪(\alpha ^3/R^3)$$
(6)
where
$$W_4=\frac{1}{\pi }_0^{\mathrm{}}𝑑\omega \omega ^2[\alpha _d(i\omega )]^2.$$
(7)
By integrating (7), the coefficient $`W_4`$ can be expressed as a double sum over oscillator strengths and it was evaluated for a small number of diatomic systems using various approximations, both semi-empirical and computational . The result for two H atoms is $`W_4=\mathrm{0.462\hspace{0.17em}807}`$. The derivations above assume that the two atoms are well-separated and accordingly do not include considerations involving electron exchange.
How do the results above connect with results from the Breit-Pauli approximation to the Dirac equation? The van der Waals potential was shown above to be the short range limit of the QED result; yet it is also the long-range limit of the molecular interaction potential. The full power of quantum-chemical methods (recognized in the 1998 Nobel Prize in Chemistry) enables, at least in principle, calculation of the molecular potential by solution of the nonrelativistic Schrödinger equation. Relativistic effects are treated using perturbation theory on the terms in the Breit-Pauli Hamiltonian (or for molecules containing high-$`Z`$ atoms by solution of the Dirac equation.) The connection to (6) was given by Power and Zienau who showed using perturbation theory that the matrix element of the orbit-orbit interaction $`H_{\mathrm{oo}}`$ reproduces the second term in (6) as $`R`$ increases,
$$\stackrel{~}{0}|H_{\mathrm{oo}}|\stackrel{~}{0}\alpha ^2\frac{W_4}{R^4},RR_0,$$
(8)
where $`R_0`$ is of order, say, $`10`$ to $`20a_0`$ and $`|\stackrel{~}{0}`$ is the molecular ground electronic state wave function. Therefore there is a smooth connection between the relativistic and Casimir-Polder results.
This relativistic $`R^4`$ term might be studied by incorporating it into theoretical calculations of collisions of ultra-cold atoms, particularly for H-H, H-Li, and Li-Li where high precision determinations of the molecular potentials are possible. There are of course additional subtle effects to be accounted for such as deviations from the Born-Oppenheimer approximation through isotope effects and nonadiabatic terms (nonlocal terms arising from the action of the nuclear kinetic energy operator on the electronic wave function) and additional relativistic terms like the $`p^4`$ and Darwin terms, for example, but these are unrelated to the Casimir-Polder result.
The Casimir-Polder-type interaction between an electron and an ion is closely related to that of the atom-atom interaction (3). Kelsey and Spruch exhibited the result for asymptotic separations,
$$U(R)\frac{11}{4\pi }\frac{\alpha \alpha _d(0)}{R^5},R\mathrm{},$$
(9)
where $`R`$ now denotes the electron-ion distance. They obtained (9) using QED and old-fashioned perturbation theory and they considered the possibility of measurement of this potential through spectroscopy of the Rydberg states of atoms. Later, the integral form of $`U(R)`$, analogous to (1), was obtained yielding an expression not particularly more complicated than (1) and which can be obtained essentially by replacing the polarizability $`\alpha _d(i\omega )`$ of one of the atoms by the quantity $`1/\omega ^2`$, which is an excellent approximation to the polarizability of the weakly bound electron . (Some care is required, however, due to the additional Coulomb interaction present for the ion-electron case, see for details.) The limit of the electron-ion “Casimir-Polder” potential for small $`R`$ for an electron interacting with an ion (of net charge $`Z1`$) is
$$U(R)=\frac{\alpha ^2}{Z^2}\frac{1}{R^4}+𝒪(\alpha ^3/R^3),R137/Z^2.$$
(10)
Similarly to the atom-atom case, the relativistic $`R^4`$ term in the ion-electron interaction was derived alternatively using $`H_{\mathrm{oo}}`$ with perturbation theory on the non-relativistic wave function of the Rydberg atom providing a connection to the QED result (10). This term is a small correction to the much larger Coulomb interaction between the two charged particles, but nevertheless, through much theoretical work by Drachman, Drake, and others , there is definitive evidence for the first term of (10) from a long series of careful measurements of energies of Rydberg states of the helium atom by Lundeen and collaborators . At present the asymptotic part of $`U(R)`$, (9), has not been measured and Hessels and collaborators conclude from their measurements that there is, in fact, no experimental evidence for deviations from (10). Additional experiments are in progress and it will be interesting to see if the ion-electron Casimir effect will be verified. From a theoretical point of view there are interesting connections at short and long distance between the order $`𝒪(\alpha ^3/R^3)`$ QED corrections in (6) and (10).
This work was supported in part by the National Science Foundation through a grant for the Institute for Theoretical Atomic and Molecular Physics at the Smithsonian Astrophysical Observatory and Harvard University.
|
no-problem/9904/cond-mat9904241.html
|
ar5iv
|
text
|
# Is the ‘Finite Bias Anomaly’ in planar GaAs-Superconductor junctons caused by point-contact like structures?
## I Introduction
The current-voltage (I-V) relation of normal metal -superconductor (NS) interfaces - is strongly modified both by wave interference phenomena (producing quasi-bound Andreev levels) and ballistic transport at the interface. Several groups have observed such novel superconducting phenomena at the NS interface between a superconductor and semiconductor -. Based on the success of Marsh et al. - in fabricating In and Sn ‘alloyed’ contacts to a two-dimensional electron gas (2DEG) formed at the AlGaAs/GaAs interface, we have studied ‘alloyed’ In contacts to the 2DEG. We find the mechanism for producing highly transmissive NS contacts is In growth into the AlGaAs ‘guided’ along a preferred crystallographic direction. For an AlGaAs/GaAs heterojunction with a oriented surface, we find that In growth into the AlGaAs occurs preferentially along the {111} crystallographic planes. This unusual type of In growth into the AlGaAs produces an ‘inverted pyramid’ or ‘field emission’ point contact tip as shown in Fig. 1(a). Similar microstructure for metallic contacts to GaAs has been observed for both AuGeNi and Au metallizations. This guiding of In into the AlGaAs also allows the In to maintain its superconducting properties. An AlInGaAs alloy, formed by diffusing In into AlGaAs, would simply be a normal metal.
For such a crystallographically defined point contact metallization, we find the closer one can grow the tip of the point contact to the 2DEG without contacting it, the higher the transmission coefficient of an electron incident from the NS contact into the 2DEG. For such nearly ballistic transport through the NS junction, a corresponding excess current results . Growing the In down into direct contact with the 2DEG, on the other hand, results in a low-transmission normal metal - insulator - superconductor (NIS) contact and its corresponding Giaever tunneling I-V characteristic . We postulate that In in direct contact with the 2DEG depletes the electrons around it, forcing the superconducting electrons to tunnel through a large depletion layer near the contact, shown schematically in Fig. 1(b). Electron depletion around a metallic contact to GaAs is commonly known as a Schottky barrier. A similar mechanism for forming highly transmissive AuGeNi contacts to GaAs was originally postulated by Braslau -. This mechanism for forming highly transmissive tunneling type AuGeNi contacts to GaAs is the reason some AlGaAs/GaAs heterojunction transistors can operate at low temperatures with low contact resistance.
In this paper we correlate transmission electron microscope (TEM) photographs of the superconducting In contacts to the resulting I-V characteristics of the NS junctions. All the different $`dI/dV`$ characteristics shown in this paper are from nominally identical samples, grown and prepared from the same GaAs wafer at the same time. The only differences between the samples is in post process contact annealing, and hence in the contact geometry.
Changes in contact geometry produce widely different $`dI/dV`$ characteristics. In point contacts grown near the 2DEG result in ballistic transport of electrons through the NS contact and an excess current. In in direct contact with the 2DEG produces lower transmission contacts and Giaever tunneling. Since the GaAs semiconductor forming the normal metal is also weakly localized, we are able to observe weak localization corrections to Giaever tunneling . In In/GaAs junctions where some region of the contact is transmissive, that portion of the contact will produce a conductance drop around zero bias . The incoherent addition of the conductance from different regions of the same ‘contact’ could therefore generate the ‘finite bias anomaly’ seen in Ref. .
## II Formation of the NS Contact
A cross section of the unnannealed In/GaAs heterostructure is shown in Fig. 2(a). An undoped Al<sub>0.3</sub>Ga<sub>0.7</sub>As spacer layer, followed by a Si doped Al<sub>0.3</sub>Ga<sub>0.7</sub>As layer and a 50 $`A^o`$ protective Si doped GaAs layer, was grown on an undoped (semi-insulating) (100) GaAs substrate. The resulting mobility of the 2DEG at liquid nitrogen temperature was about 125,000 $`cm^2`$/$`Vs`$. The In contacts were deposited by thermal evaporation and liftoff. A top view of the device is shown in Fig. 2(b). Two In pads, each of dimension 3.2mm$`X`$2.5mm and seperated by a nominal gap of 4$`\mu `$, were deposited on top of the heterostructure using thermal evaporation lift off. After annealing, we diffused In through the AlGaAs barrier layer to contact the 2DEG. The annealing temperature was varied between 500<sup>o</sup>C and 600<sup>o</sup>C. For temperatures of 450<sup>o</sup>C or less the In failed to contact the 2DEG, while annealing temperatures greater than 700<sup>o</sup>C caused thermal deterioration of the interface. Large In grains, 1-2 microns in diameter, grew on the contacts after annealing.
A TEM micrograph of Sample 1, annealed for a relatively short duration e.g. at 550 $`{}_{}{}^{o}C`$ for 2 minutes, is shown in Fig. 3(a). Fig. 3(a) shows that Indium starts growing into (100) AlGaAs/GaAs preferentially along the $`<`$111$`>`$ directions. The growth seems to be guided by the {111} crystallographic planes of GaAs inclined at angles of about 55<sup>o</sup> to the surface. Therefore for short annealing times we get ‘spikes’ of Indium descending towards the interface, forming an array of point contacts pictured in Fig. 3(a). The spikes are rather non-uniform in size and irregular in their penetration depths, probably nucleating at defects in the suface oxide of AlGaAs. In grain growth on the wafer surface therefore has little effect on the final microstructure and geometry of the NS contact. The net total conductance of the junction is determined by the sum of the conductance of all the point contacts in parallel. This growth mechanism is similar to that for alloyed Au-Ge-Ni contacts with GaAs in which most of the conduction is through isolated Ge rich islands formed on the GaAs.
A TEM micrograph of Sample 2, annealed at 550<sup>o</sup>C for 6 minutes and then at 650<sup>o</sup>C for 3 minutes, is shown in Fig. 3(b). For the longer annealing times and higher annealing temperatures used in sample 2, more such In spikes grow from the deposited In contact. Fig. 3(b) shows these In spikes coalesce and penetrate completely through the AlGaAs/GaAs interface to physically touch the 2DEG, as depicted schematically in Fig. 1(b). The conductance characteristics of Samples 1 and 2 differ dramatically as described in the next section.
A high magnification TEM photograph of one of the inverted pyramid type In spikes in Sample 1 is shown in Fig. 4. The (100) GaAs surface is towards the top of Fig. 4. The penetration of In into the AlGaAs is clearly guided by {111} crystallographic planes. One can see in the high resolution TEM picture that the In indeed follows the {111} AlGaAs planes for several atoms. The In-AlGaAs boundary then moves abruptly along the direction for 1-2 atoms before continuing along the {111} planes. The detailed TEM picture in Fig. 4 shows that while the {111} planes strongly guide the growth of In into (100), the In does not exactly follow those planes. However, the overall structure of the In contacts still resembles a point contact.
## III Current-Voltage Characteristics
Hall measurements were made on the 2DEG at the interface at a temperature of 400 mK. From the slope of transverse resistance $`R_{xy}`$ we estimate the carrier density to be $`1.8X10^{11}`$ $`cm^2`$. The mobility was 215,000 $`cm^2/Vs`$ which yielded a mean free path $`l`$ of about 0.4 $`\mu `$. From the low-field magnetoresistance, we estimate the phase breaking length $`l_\varphi `$ to be about 1.6 $`\mu `$, which is less than the 4 $`\mu `$m separation between the two In pads in Fig. 2(b). We conclude that the conductance of the devices is essentially equivalent to two NS junctions in series separated by a series resistor (the 2DEG).
### A NS Junction with an excess current
In Fig. 5(a) we show the conductance characteristics of Sample 1 at a base temperature of 100 mK. Fig. 5(a) shows an increase in conductance of about 10-12% around a range of $`\pm `$ 6mV. The 10-12% excess conductance around zero voltage shows that the junction behaves like a moderately transmissive junction , with the majority of the diffused Indium spikes forming transmissive interfaces with the 2DEG of the GaAs. The large oscillations of the conductance in Fig. 5(a) are reproducible with thermal recycling of the device. The oscillations are therefore likely a consequence of electron wave interference due to scattering from the fixed In point contacts to the 2DEG. The parasitic resistance of the 2DEG in series with the two NS junctions stretches the dI/dV versus V characteristics along the voltage axis, and also suppresses any changes observed along the dI/dV axis. Series resistance of the 2DEG explains why we observe only a 10-12% excess conductance instead of an excess conductance approaching 100% in Fig. 5(a).
We can crudely estimate the interface transmission in Fig. 5(a) by comparing it to an ideal, ballistic NS junction . A perfectly transmissive interface shows a 100% rise of conductance when the voltage bias satisfies $`|eV|\mathrm{\Delta }`$, where $`2\mathrm{\Delta }`$ superconducting energy gap . Setting $`2\mathrm{\Delta }=`$ 6mV (for two NS junctions in series), and using the BCS formula $`2\mathrm{\Delta }=3.5k_BT_c`$, gives a critical temperature $`T_c=20`$K. Since the actual critical temperature of the contacts is about $`3.4`$K, we infer a 1mV drop across the two NS interfaces and 5mV across the semiconductor. The total conductance at the gap voltage can be read directly from Fig. 5(a) as about 9mS, from which we infer a series resistance of about 92.6$`\mathrm{\Omega }`$ and a resistance of the two NS interfaces in series of about 18.5$`\mathrm{\Omega }`$ (when the voltage across the NS interface is less than the energy gap). When the voltage across the two NS interfaces is large, we can read directly off Fig. 5(a) a conductance of about 7.75mS, or a total resistance of about 129$`\mathrm{\Omega }`$. Subtracting the series resistance, we see the device resistance changes from 18.5$`\mathrm{\Omega }`$ when $`V_{\mathrm{interface}}\mathrm{\Delta }`$ to about 36.4$`\mathrm{\Omega }`$ when $`V_{\mathrm{interface}}\mathrm{\Delta }`$, roughly a 97% increase in background conductance. Since strong wave interference is present in Fig. 5(a), this comparison should be regarded as giving an order of magnitude estimate. However, after correcting for series resistance, this estimate indicates the junction in Fig. 5(a) is a nearly ballistic NS interface.
### B Giaever-type Tunnel Junctions
In Fig. 5(b) we show the differential conductance for Sample 2 at a temperature of 400mK. The differential conductance is suppressed around zero voltages, indicating Giaever tunneling and a lower transmission contact. We can conclude from Figs. 1 and 5 that In in intimate contact with the 2DEG forms a relatively low transmission contact, whereas In nearby but not directly in contact with the 2DEG forms a high transmission contact. The temperature dependence of the differential conductance for Samples 1 and 2 is also consistent with a ballistic contact and a low transmission contact, respectively. The differential conductance of Sample 1 is relatively constant with temperature, while the differential conductance of Sample 2 greatly decreases as the temperature is lowered (indicating thermionic emission).
We have taken several NS contacts displaying Giaever tunneling at low temperature and annealed them for longer times, attempting to obtain a ballistic NS interface. In all cases further annealing simply makes the Giaever tunneling characteristic more pronounced, indicating that further annealing lowers the interface transmission. The room temperature conductance of the sample improves with further annealing, however, simply due to an increase of the effective contact area. Since the room temperature conduction mechanism through the contact of Sample 2 is thermionic emission, it simply scales with the contact area. Further annealing improves the room temperature conductance, but worsens the low temperature conductance. There is an optimal annealing time where In grows down to almost reach the 2DEG, but is not in physical contact with the 2DEG. Any further annealing after this point degrades interface transmission.
### C Effect of Weak Localization on Giaever Tunneling
For an NS junction of length L, obeying the condition $`lLl_\varphi `$, electrons which initially failed to Andreev refelct from the NS interface can backscatter again to the NS interface. Therefore, weak localization inside the normal conductor gives the electrons additional opportunities for Andreev reflection. The net effect of Giaever tunneling at the NS junction combined with weak localization inside the normal conductor is an enhancement of the total Andreev reflection probability at the Fermi level, leading to an additional conductance peak around zero bias voltage .
Fig. 6 shows the differential conductance for the NS junction annealed at 550<sup>o</sup>C for 3 minutes, which we call Sample 3. Sample 3 displays the conductance peak around zero bias voltage first observed in Ref. and explained in Ref. . Disorder assisted backscattering can cause a zero bias conductance peak of magnitude up to 10% of the background conductance value. In our case the zero bias peak is about a 1.5% increase over the background conductance. The Giaever tunneling feature is also spread over a large voltage range larger than 1mV. Parasitic resistance from the 2DEG again explains the smaller zero bias conductance peak and the spreading out of the dI/dV versus V along the voltage axis. Since the height of the zero bias conductance peak saturates by 300 mK Fig. 6(a), the peak is not the precursor of a supercurrent between the two contacts. The supercurrent should be negligibly small in any case, since the sample satisfies $`lLl_\varphi `$. We adequately filtered RF noise away from the devices to observe supercurrents in other superconductor - semiconductor samples with closer pad separation. We would also have observed such a supercurrent if it were present in this sample.
Fig. 6(b) shows the magnetic field dependence of the conductance for Sample 3. For a junction of length L and width W, where $`L,W>L_\varphi `$, the magnetic field required to destroy this zero bias conductance peak is of the order $`B_c=\varphi _0/L_{\varphi }^{}{}_{}{}^{2}`$. In Sample 3 we observe $`B_c`$ 80 Gauss. The calculated field is $`B_c`$ 20 Gauss, nearly four times smaller than the observed value. It is possible that the $`L_\varphi `$ is overestimated, i.e. it may be around $`L_\varphi `$ 0.8 $`\mu `$m. Since these numbers are not precise data fits based on any quantitative theory, it is comforting that we obtain roughly the coherence length obtained by previous weak localization measurements on the 2DEG. The shift in the background conductance with the magnetic field in Fig. 6(b) is also due to the parasitic magnetoresistance of the 2DEG.
### D Anamalous Weak Localization Corrections to Giaever Tunneling
On most samples where we observed weak localization corrections to Giaever tunneling, we obtained conductance characteristics similar to those in Fig. 6. However, Fig. 7 shows the conductance characteristics for an NS junction annealed at 500<sup>o</sup> C for 2 minutes which we call Sample 4. Overall Sample 4 displays a background of Giaever tunneling. A zero bias conductance peak (similar to the one in Fig. 6) continues to develop for temperatures down to about 800 mK in Fig. 7(a). At a temperature of 650 mK in Fig. 7(a), however, a conductance dip begins developing around zero bias. This dip in conductance around zero bias, which is superposed on the broader conductance peak, is nearly fully developed by 300 mK as shown in Fig. 7(a). There is little change in the differential conductance between 300 mK and 180 mK in Fig. 7(a). This anomalous dip feature superimposed on the weak localization correction to Giaever tunneling is reproducible on thermally cycling the NS junction back to room temperature and again down to mK temperatures.
Marmorkos, Beenakker, and Jalabert have numerically simulated the conductance of an NS junction in contact with a dirty normal metal. For low transmission interfaces they numerically observe, in Fig. 2 of Ref. , the zero bias conductance peak associated with the weak localization corrections to Giaever tunneling. However, for high transmission between the NS interface and normal conductor, the numerical simulations of Ref. reveal that the conductance peak changes into a conductance dip around zero bias. Ref. therefore shows that the same weak localization phenomena which causes a zero bias conductance peak in low transmission contacts causes a zero bias conductance dip for highly transmissive NS interfaces.
The numerical simulation in Ref. offers one possible way to explain the conductance dip around zero bias we observe in Sample 4. The overall conductance of Sample 4 displays Giaever tunneling. Therefore, the majority of the NS interface area in Sample 4 has an additional tunneling barrier between the superconductor and 2DEG, namely the depletion region shown in Fig. 1(b). However, the type of NS junctions we form by diffusion In into AlGaAs/GaAs are inhomogeneous enough that a significant fraction of the sample can form a transmissive NS interface of the type shown in Fig. 1(a). The conductance we observe in Fig. 7 will be a parallel combination of these two different types of NS junctions, as shown schematically in Fig. 8. For voltages away from $`V=0`$, the slow variation of the background Giaever tunneling conductance dominates the dI/dV curve. For voltages very close to zero bias, the weak localization phenomena at the transmissive regions dominate and leads to the observed conductance peak and dip.
The peak at finite bias in Figs. 7 was first observed by Poirier et al. , who called it the ‘finite bias anomaly’. One problem with using the simulations of Mormorkos et al. to explain a conductance dip around zero bias is that it requires a high interface transmission, whereas the data of Poirier et al. (and our own data) show a Giaever tunneling background (low average interface transmission). The inherent inhomogeneity of supposedly planar superconducting In contacts to the 2DEG in AlGaAs/GaAs we have demonstrated in this paper overcomes this difficulty. A few high transmission point emitters can produce the conductance dip around zero bias, whereas the majority of the contact can maintain low overall interface transmissivity. The weak localization dip around zero bias can therefore peacefully coexist with a Giaever tunneling background conductance.
The weak localization correction to the conductance of a ballistic NS junction could have been more clearly observed in Sample 1, were it actually present in that sample. Similarly Sample 2 (and several other samples we measured) did not exhibit the weak localization correction to the Giaever tunneling conductance. The exact impurity configuration near a particular NS interface will determine whether or not the weak localization correction to the conductance appears in any given sample. Perhaps it is therefore not surprising that the weak localization correction to the conductance can be observed only in a fraction of the samples.
A different mechanism which splits the zero bias conductance peak in NI<sub>1</sub>NI<sub>2</sub>S junctions was developed in Ref. . Weak localization inside the middle N region produces the zero bias conductance peak. If the two insulators I<sub>1</sub> and I<sub>2</sub> have two different transmission coefficients, the zero bias conductance peak is split as shown in Fig. 3 of Ref. . These two barriers, having different transmissivity, is the same mechanism proposed by Poirier et al. to account for the ‘finite bias anomaly’. Ref. proposed a model which used the Schottky barrier at the NS interface to produce I<sub>2</sub>, and an impurity inside the semiconductor as I<sub>1</sub>. The spacing between I<sub>1</sub> and I<sub>2</sub> is L, a random number set by the impurity configuration. The McMillan-Rowell resonance nearest the Fermi level survives in the conductance of an NI<sub>1</sub>NI<sub>2</sub>S junction upon averaging over different L, producing a finite bias anomaly whose voltage is set by the average L.
The composite ‘point emitter’ model for the contact developed in this paper may also provide some support for this NI<sub>1</sub>NI<sub>2</sub>S model for the ‘finite bias anomaly’. An electron moving through the 2DEG past a point emitter would see that emitter as a scattering center, equivalent to an insulating barrier. The distance between the emitters in Sample 1 is of the order 100nm, less than the electron phase coherence length. In Sample 1, therefore, could be regarded as a type of NI<sub>1</sub>NI<sub>2</sub>NI<sub>3</sub> … S junction. Each normal metal region N would also be weakly localized. This model may also produce a finite bias anomaly, but would require further numerical support. A two-dimensional numerical simulation, where the electrons could actually move around the point emitter scattering centers, would be required to confirm this picture.
## IV Conclusions
We have measured the differential conductance of superconductor-normal metal junctions formed by diffusing Indium into AlGaAs/GaAs heterostructures. In grows into a AlGaAs/GaAs heterostructure having a oriented surface preferentially along the {111} crystallographic planes. Instead of a planar diffusion profile, we therefore find that In forms ‘inverted pyramids’ or point contacts to the 2DEG. Supposedly ‘planar’ superconducting In contacts to the electron gas in an AlGaAs/GaAs heterojunction are therefore actually composed of many point emitters. Correlating the contact microstructure observed on different samples with the differential conductance spectroscopy of the NS contact allowed us both to explain many observed features in the conductance and to determine the mechanism of superconducting (ohmic) contacts to the 2DEG in this materials system.
For NS junctions annealed at a moderate temperature for a short times, so that the In point contacts do not physically touch the 2DEG, we obtain highly transmissive NS junctions. Due to the contact inhomogeneity, the point emitters nucleate and grow at different rates into the semiconductor. We observed wave interference between these different superconducting emitters in transmissive NS junctions. For identically prepared NS junctions annealed at higher temperatures and for longer times, so that the In point contacts grow together and have direct physical contact with the 2DEG, we obtain a lower transmission NS interface and Giaever tunneling. This is due to a depletion layer which forms around the In which directly touches the 2DEG. Further annealing simply increases the effective strength of the interface barrier between N and S, as regions of the In which previously were not in direct physical contact with the 2DEG come in contact with the 2DEG.
Since the semiconductor forming N is disordered and has a reasonable phase coherence length, weak localization corrections to the differential conductance around zero bias voltage are also observed in this materials system. This zero bias conductance peak is a correction to Giaever tunneling which has been previously observed by several other groups . We observed an additional dip inside this zero bias conductance peak which develops in some samples at low temperature . One possible explanation for the additional dip is due to contact inhomogeneity, where a small percentage of the contact is a nearly ballistic NS interface while most of the NS contact area remains in the tunneling limit. This conductance dip around zero bias voltage is therefore possible evidence for the predicted weak localization correction to the conductance of ballistic normal metal - superconductor junctions in Ref. . In any case, explanations for this ‘finite bias anomaly’ should account for the actual non-planar physical structure of the superconducting contact.
## V Acknowledgments
We gratefully acknowledge support from the David and Lucile Packard Foundation and from the MRSEC of the National Science Foundation under grant No. No. DMR-9400415. We thank Tamer Rizk, Richard Riedel, Manoj Samanta and Supriyo Datta for many useful discussions.
<sup>1</sup> Present address: Intel Corporation, RN2-40, 2200 Mission College Blvd. Santa Clara, CA 95052.
<sup>2</sup> Present address: Xilinx, 2100 Logic Dr., San Jose, CA 95124.
<sup>3</sup> Present Address: Yale University, Department of Electrical Engineering, New Haven, CT 06520.
<sup>4</sup> Present address: Dept. of Physics, University of North Florida, Jacksonville, FL 32224.
|
no-problem/9904/astro-ph9904242.html
|
ar5iv
|
text
|
# Optical counterparts of cosmological GRBs due to heating of ISM in the parent galaxy
## 1 Introduction
In the cosmological model for Gamma-Ray Bursts energy production in $`\gamma `$-region must be enormous large $`10^{51}10^{53}`$ ergs, during several seconds. Such huge energy flux interacts with the matter in the parent galaxy leading to its heating and formation of counterparts in another spectral region. We investigate results of this interaction without specifying the mechanism of GRB formation.
The problem is solved of the respond of interstellar medium with normal chemical composition on propagation through it of a short powerful gamma-ray pulse. Spherical symmetry was accepted; different densities of interstellar gas and burst energies were considered.
## 2 Basic equations
The basic equations, describing this problem are :
$$\frac{\varrho }{t}+\frac{1}{r^2}\frac{}{r}(r^2\varrho u)=0,$$
(1)
$`{\displaystyle \frac{u}{t}}+u{\displaystyle \frac{u}{r}}={\displaystyle \frac{1}{\varrho }}{\displaystyle \frac{p}{r}}+F_\gamma ,`$ (2)
$`{\displaystyle \frac{ϵ}{t}}+u{\displaystyle \frac{ϵ}{r}}={\displaystyle \frac{p}{\varrho }}{\displaystyle \frac{1}{r^2}}{\displaystyle \frac{}{r}}(r^2u)+H_\gamma C_\gamma .`$ (3)
Where
$`\begin{array}{c}p=(n_i+n_n+n_e)kT,ϵ=\frac{3}{2}\frac{kT}{\mu m_u},\end{array}`$ (5)
$`n_n`$, $`n_i`$, $`n_e`$ are concentrations of neutral atoms, ions and electrons correspondently.
For the $`\gamma `$-ray signal interacting with matter due to Thomson scattering of photons on electrons with cross-section $`\sigma _T`$ we have:
$$F_\gamma =\frac{1}{c}\frac{L}{4\pi r^2}\frac{\mu _e\sigma _T}{m_u},$$
(6)
For $`\gamma `$-rays with $`h\nu \text{B}_e^{(\text{a,i})}`$ (B<sub>e</sub> is the binding energy of electrons in atoms or ions) the cross-section is almost the same for free and bound electrons. Heating of the gas by the light signal with the spectrum $`\frac{dL}{dE}=\frac{L}{E_{max}}e^{E/E_{max}}`$ is given by (Bisnovatyi-Kogan & Blinnikov B-KB (1980)):
$`H_\gamma ={\displaystyle \frac{L}{4\pi r^2}}{\displaystyle \frac{\mu _e\sigma _T}{m_u}}{\displaystyle \frac{E_{max}4kT}{m_ec^2}},`$ (7)
where $`L`$ is the energy flux of the signal; $`E`$ is the energy of photons. This formula is valid for $`Em_ec^2`$. For photons with energies $`<10`$ keV the main process of interaction with the gas is photoabsorption by ions of heavy elements. To take in to account gas heating by soft X-rays photons we enhance term (7) using $`E_{max}2`$ MeV. Because of lack of information about soft X-ray spectra of GRBs and variations in interstellar media density such assumption seems to be quite good. On the other hand, completely ignoring gas heating by X-rays we get a counterpart by less then $`2^m`$ fainter. Cooling of optically thin plasma by free-free and free-bound transitions is given by approximating function (Cowie et al. CowieMcKee (1981)):
$`C_\gamma ={\displaystyle \frac{\mathrm{\Lambda }(T)n^2}{\varrho }},n=n_n+n_i`$ (8)
$`\mathrm{\Lambda }(T)=\{\begin{array}{cc}0,\hfill & \text{when }T<10^4\text{K;}\hfill \\ 1.010^{24}T^{0.55},\hfill & \text{when }10^4\text{K}<T<10^5\text{K;}\hfill \\ 6.210^{19}T^{0.6},\hfill & \text{when }10^5\text{K}<T<410^7\text{K;}\hfill \\ 2.510^{27}T^{0.5},\hfill & \text{when }T>410^7\text{K.}\hfill \end{array}`$
Electron density ($`n_e`$) is given by the Elvert formula.
This system was solved numericaly using a full conservative difference scheme with flux corrected transport, because there is a strong density gradient in the solution (shock wave).
## 3 Main results
For GRB with the energy $`10^{52}`$ erg we considered following densities of the surrounding interstellar medium: $`10^5\text{cm}^3,300\text{cm}^3,0.25\text{cm}^3,1.610^3\text{cm}^3`$. Main results of calculations in optically thin approximation are presented in our publications Timokhin & Bisnovatyi-Kogan (Timokhin1 (1995, 1996)). According these calculations a counterpart of cosmological GRB due to the influence of GRB on the surrounding medium could be observed if density of the medium is greater then $`300\text{cm}^3`$. For such dense interstellar clouds optically thin approximation is not valid because of large optical depth in main emission lines.
Then we considered a simplified picture of conversion of UV photons into optical ones in opticaly thick regions allowed us to get optical and UV light curves of counterparts (see details in Bisnovatyi-Kogan & Timokhin Timokhin3 (1997)). We considered GRBs with energies $`10^{52}`$,$`10^{51}`$ ergs and following initial conditions in a gaseous cloud of radius $`R`$:
* $`n_0=10^5\text{cm}^3,T_0=20\text{K},R=1.5\text{pc}`$
* $`n_0=10^4\text{cm}^3,T_0=20\text{K},R=5\text{pc}`$
The results of calculations are presented in Fig. 1. Dashed lines give the lower boundary for luminosity, and solid lines - the upper one. For GRB with total energy output $`10^{51}`$ and $`10^{52}`$ ergs maximal luminosities of counterparts differ by one order of magnitude too. It means that relative maximum brightness of the counterparts depends weakly on the total energy output. The total duration of the counterpart radiation decreases 2 times when the total output decreases 10 times. A more detailed description of the results is given in Bisnovatyi-Kogan & Timokhin (Timokhin3 (1997)).
## 4 Conclusions
It is shown, that counterparts of cosmological GRB due to interaction of gamma-radiation with dense interstellar media are ”longliving” objects, existing for years after GRB. To distinguish GRB counterpart from a superova event, having similar energy output, it is necessary to take into account its unusual light curve and spectrum. In optical region of the spectrum the strongest emission lines are H<sub>α</sub> and H<sub>β</sub>.
The volume occupied by dense interstellar medium in the parent galaxy is relatively small, so the probability to get bright counterpart of such kind is also small. Nevertheless, discovery of even one optical counterpart of GRB with properties described above could be a decisive argument in favor of the cosmological nature of GRB.
###### Acknowledgements.
Part of this work was supported by Russian Foundation of Basic Research (RFFI) under grant 96-02-16553.
|
no-problem/9904/cond-mat9904316.html
|
ar5iv
|
text
|
# Orbital Dynamics: The Origin of Anomalous Magnon Softening in Ferromagnetic Manganites
\[
## Abstract
We study the renormalization of magnons by charge and coupled orbital-lattice fluctuations in colossal magnetoresistance compounds. The model considered is an orbitally degenerate double-exchange system coupled to Jahn-Teller active phonons. The modulation of ferromagnetic bonds by low-energy orbital fluctuations is identified as the main origin of the unusual softening of the zone-boundary magnons observed experimentally in manganites.
\]
Recently, distinct new features in the spin dynamics of the ferromagnetic manganese oxide compound Pr<sub>0.63</sub>Sr<sub>0.37</sub>MnO<sub>3</sub> have been reported . Striking deviations from the predictions of the canonical double-exchange (DE) theory were observed. In particular, an anomalous softening of magnons at the zone boundary was found even well below the Curie temperature $`T_C`$. It is noted that these anomalies are closely related to the reduced values of $`T_C`$: For higher-$`T_C`$ compounds no considerable deviations from a simple cosine dispersion is observed .
These experimental findings seem to be of high importance. They in fact indicate that some very specific features of magnetism in colossal magnetoresistance manganites have still to be identified. In this paper we propose a mechanism which might explain the above experimental observations. Our basic idea is the following: The strength of ferromagnetic interaction in a given bond strongly depends on which orbital is occupied by an $`e_g`$ electron (Fig. 1). Suppose that orbitals and Jahn-Teller (JT) distortions are disordered in the ferromagnetic phase. It is then evident that temporal fluctuations of orbitals affect the short-wavelength magnons through a strong modulation of exchange bonds. Quantitatively, the effect is expected to be controlled by the characteristic time scale of orbital fluctuations: A slowing-down of the dynamics of the coupled system of orbitals and JT phonons should lead to a stronger magnon renormalization. In other words, the observed zone-boundary magnon softening is interpreted in this picture as a precursor effect in the proximity of static orbital-lattice ordering.
More specifically, we calculate the dispersion of one-magnon excitation at zero temperature. First, we map the ferromagnetic Kondo-lattice model onto a Hamiltonian of interacting magnons and spinless fermions. Yet the fermions carry orbital quantum number, the fluctuations of which the magnon can scatter on. Second, we show that as the orbitals are strongly coupled to the lattice there is an indirect coupling of magnons to JT phonons via the orbital sector. Put another way, the orbital-lattice coupling produces a low-energy component in the orbital fluctuation spectrum at phonon frequencies. We then calculate the self-energy corrections to the magnon dispersion perturbatively (employing a $`1/S`$ expansion). Charge fluctuations are found to produce a moderate softening of magnons throughout the Brillouin zone. The effect of orbital and lattice fluctuations on the magnon dispersion is more pronounced and becomes dramatic as static order in the orbital-lattice sector is approached.
We start with a model describing the ferromagnetic coupling of double-degenerate $`e_g`$-band electrons to an otherwise noninteracting system of localized spins $`𝑺_t`$ on a cubic lattice:
$`H`$ $`=`$ $`{\displaystyle \underset{ij_\gamma }{}}t_\gamma ^{\alpha \beta }\left(e_{i\sigma \alpha }^{}e_{j\sigma \beta }+\text{H.c.}\right)J_H𝑺_{it}𝒔_{ie}`$ (2)
$`+{\displaystyle \underset{\alpha }{}}Un_{i\alpha }n_{i\alpha }+{\displaystyle \underset{\alpha \beta }{}^{}}\left(U^{}J_H\widehat{P}\right)n_{i\alpha }n_{i\beta }`$
with $`\gamma \{x,y,z\}`$ and $`\widehat{P}=(𝒔_{i\alpha }𝒔_{i\beta }+\frac{3}{4})`$. Hereafter, the indices $`\alpha /\beta `$ and $`\sigma `$ stand for orbital and spin quantum numbers of $`e_g`$ electrons, respectively; summation over repeated indices is implied and double counting is excluded from the primed sum. The spin operator $`𝒔_{i\alpha }`$ acts on orbital $`\alpha `$ and $`𝒔_{ie}=_\alpha 𝒔_{i\alpha }`$. The last two terms in Eq. (2) describe intra- (inter-) orbital Coulomb interaction $`U`$ ($`U^{}`$) and Hund’s coupling between $`e_g`$ electrons in doubly occupied states. Correlations are assumed to be strong: $`U,U^{},(U^{}J_H)t`$. The important point is the peculiar orbital and bond dependencies of the electron hopping matrix elements. In orbital basis $`\alpha \{|3z^2r^2,|x^2y^2\}`$ (see Fig. 1) one has :
$`t_{x/y}^{\alpha \beta }=t\left(\begin{array}{cc}1/4& \sqrt{3}/4\\ \sqrt{3}/4& 3/4\end{array}\right),t_z^{\alpha \beta }=t\left(\begin{array}{cc}1& 0\\ 0& 0\end{array}\right).`$
We work in the limit $`J_H\mathrm{}`$. Then the combined action of the on-site Hund’s “ferromagnetism” and the spin-diagonal nature of electron hopping results in a global ferromagnetic ground state. Two different contributions to the spin stiffness should be noticed: Conventional double exchange due to metallic charge motion and superexchange which accounts for the kinetic energy gain by virtual hoppings of $`e_g`$ electrons. We now derive the effective Hamiltonian describing the spin excitations at $`TT_C`$.
Double exchange.— At $`J_H\mathrm{}`$, $`𝑺_t`$ and the $`e_g`$ spin $`𝒔_e`$ are not independent anymore: $`𝒔_{ie}=n_{ie}𝑺_t/2S_t`$. They form a total on-site spin $`𝑺_i`$ with spin value $`S_t+\frac{1}{2}`$ if an $`e_g`$ electron is present, $`n_{ie}=1`$, or simply $`S_t`$ otherwise. The unification of band and local spin subspaces implies spin-charge separation. That is, the spin component $`b_{i\sigma }`$ of the $`e_g`$ electron $`e_{i\sigma \alpha }=b_{i\sigma }c_{i\alpha }`$ is decoupled and “absorbed” by the total spin, and we are left with a charged fermion $`c_{i\alpha }`$ carrying the orbital index. At $`TT_C`$, the Schwinger boson $`b_i`$ simply becomes a part of the magnon operator $`b_i`$, namely $`b_i=b_i/(2S)^{1/2}`$, while $`b_i`$ is almost condensed, $`b_i=(n_{ie}n_i)^{1/2}1b_i^{}b_i/4S`$. Hereafter, $`S=S_t+\frac{1}{2}`$. Further, we keep the relevant leading $`1/S`$ terms only and assume small doping $`x`$. The kinetic energy then reads
$`H_{\text{kin}}`$ $`=`$ $`{\displaystyle \underset{ij_\gamma }{}}t_\gamma ^{\alpha \beta }c_{i\alpha }^{}c_{j\beta }+{\displaystyle \frac{1}{2S}}{\displaystyle \underset{ij_\gamma }{}}t_\gamma ^{\alpha \beta }`$ (4)
$`\times c_{i\alpha }^{}c_{j\beta }\left({\displaystyle \frac{1}{2}}b_i^{}b_i+{\displaystyle \frac{1}{2}}b_j^{}b_jb_i^{}b_j\right)+\text{H.c.}`$
The first term here describes the fermionic motion in a ferromagnetic background (yet this motion is strongly correlated in the presence of orbital disorder ), while the second term controls the spin dynamics and spin-fermion interaction. The latter term would come out from a spin-wave expansion of an effective Heisenberg model $`J_{\text{DE}}(𝑺_i𝑺_j)`$ with $`J_{\text{DE}}=(2S^2)^1t_\gamma ^{\alpha \beta }c_{i\alpha }^{}c_{j\beta }`$ if one considers the fermionic sector on average as in mean-field treatments of the DE model. However, the bond variable $`c_{i\alpha }^{}c_{j\beta }`$ is a fluctuating complex quantity and the spin structure of Eq. (4) is therefore not of Heisenberg form. Explicitly ($`TT_C`$):
$`H_{\text{kin}}`$ $`=`$ $`{\displaystyle \underset{ij_\gamma }{}}t_\gamma ^{\alpha \beta }c_{i\alpha }^{}c_{j\beta }`$ (6)
$`\times \left[{\displaystyle \frac{3}{4}}+{\displaystyle \frac{1}{4S^2}}\left(S_i^zS_j^z+S_i^{}S_j^+\right)\right]+\text{H.c.}`$
In the classical limit for spins, one recovers from Eq. (6) the effective fermionic model with phase-dependent hopping . In addition to the above peculiarities, the band/local duality of spin in the DE system further results in the following important point: Going to momentum representation in Eq. (4) one immediately realizes that the magnon-fermion interaction vertex does not vanish at zero-momentum transfer. What is wrong? The boson $`b_i`$ should not be regarded as a physical magnon. The true Goldstone particle of the DE model is the following object with both local spin and itinerant features:
$`B_i`$ $`=`$ $`b_i\left[n_{ic}+\sqrt{{\displaystyle \frac{2S1}{2S}}}\left(1n_{ic}\right)\right]`$ (7)
$``$ $`b_i{\displaystyle \frac{1}{4S}}\left(1n_{ic}\right)b_i.`$ (8)
The composite character of the physical magnon $`B_i`$ is the price one has to pay for spin-charge separation and for the rearrangement of the original Hilbert space. The itinerant component of $`B_i`$ is of order $`1/S`$ only. However, the spin stiffness is itself of the same order, and the $`1/S`$ correction in Eq. (8) is of crucial importance to ensure spin dynamics consistent with the Goldstone theorem. Now, commuting Eq. (8) with Eq. (4) one finds the mean-field magnon dispersion $`\omega _𝒑`$ and the correct momentum structure of the magnon-fermion scattering vertex:
$`[B_𝒑,H]`$ $`=`$ $`\omega _𝒑B_𝒑+{\displaystyle \frac{t}{2S}}{\displaystyle \underset{𝒒}{}}A_𝒑^{\alpha \beta }(𝒌)c_{𝒌\alpha }^{}c_{𝒌𝒒,\beta }B_{𝒑+𝒒},`$ (10)
$`A_𝒑^{\alpha \beta }(𝒌)=\gamma _𝒌^{\alpha \beta }\gamma _{𝒌+𝒑}^{\alpha \beta }.`$
Here, $`\omega _𝒑=zD_s\left(1\gamma _𝒑\right)`$, $`D_s=J_{\text{DE}}S`$, $`J_{\text{DE}}`$ is defined above, $`z=6`$, and the form factors $`\gamma _𝒌=z^1_𝜹\mathrm{exp}(i𝒌𝜹)`$, $`\gamma _𝒌^{\alpha \beta }=(zt)^1_𝜹t_\delta ^{\alpha \beta }\mathrm{exp}(i𝒌𝜹)`$. The physics behind the second term in Eq. (10) is the temporal fluctuations of the “exchange constant” due to the charge and orbital dynamics. From now on, the conventional diagrammatic method with bare magnon-fermion vertex given in Eq. (10) can be used.
Now we turn to the correlations in the fermionic band. We separate the charge and orbital degrees of freedom by the parameterization $`c_{i\alpha }=h_i^{}f_{i\alpha }`$, where the slave boson $`h_i`$ and the fermion $`f_{i\alpha }`$ represent the density and orbital fluctuations, respectively . Further, approximating $`\delta (c_{i\alpha }^{}c_{j\beta })=h_j^{}h_if_{i\alpha }^{}f_{j\beta }+f_{i\alpha }^{}f_{j\beta }h_j^{}h_i`$, one finds from Eq. (10) two different contributions to the magnon self energy.
Superexchange.— At low dopings, the virtual charge transfer across the Hubbard gap becomes of importance. In limit of large $`J_H`$, transitions to the high-spin intermediate state with energy $`U_1=(U^{}J_H)U^{},U`$ dominate. The corresponding superexchange Hamiltonian is then obtained to be:
$$H_{\text{SE}}=J_{\text{SE}}\underset{ij_\gamma }{}\left[𝑺_i𝑺_j+S(S+1)\right]\left(1\tau _i^\gamma \tau _j^\gamma \right).$$
(11)
Here, $`J_{\text{SE}}=(t^2/U_1)[2S(2S+1)]^1`$ and $`\tau ^{x/y}=\left(\sigma ^z\pm \sqrt{3}\sigma ^x\right)/2`$, $`\tau ^z=\sigma ^z`$ with Pauli matrices $`\sigma ^{x/z}`$ acting on orbital subspace $`f_\alpha `$. Eq. (11) is the generalization of orbitally degenerate superexchange models for arbitrary values of spin. Superexchange is of ferromagnetic nature because the $`J_H\mathrm{}`$ limit is assumed, and the exchange strength depends on orbital orientations . In an orbitally disordered state one can represent the orbital part of Eq. (11) in terms of fluctuating bond operators $`f_{i\alpha }^{}f_{j\beta }`$ . Then the effect of superexchange is simply to add a new term
$$A_𝒑^{\alpha \beta }(𝒌,𝒒)=x_0\left(\gamma _𝒌^{\alpha \beta }+\gamma _{𝒌𝒒}^{\alpha \beta }\gamma _{𝒌+𝒑}^{\alpha \beta }\gamma _{𝒌𝒒𝒑}^{\alpha \beta }\right)$$
(12)
to the scattering amplitude in Eq. (10) and to renormalize $`D_s(J_{\text{DE}}+J_{\text{SE}})S=t\chi _0(x+x_0)/2S`$. Here the constant $`x_0=2\chi _0t/U_1`$ with mean-field parameter $`\chi _0=f_{i+c}^{}f_i\frac{1}{2}(1x)`$ defines the effective doping level below which the superexchange contribution becomes of importance.
To summarize up to now: We have developed a quantum theory of the double-exchange model at $`T<T_C`$ which also includes the correlation effects present in transition metal oxides. The obvious advantage of this approach over earlier treatments is that the theory has a transparent structure adjusted to describe the low-energy spin dynamics, and the large atomic scale $`J_H`$ does not appear in the calculations, either. Yet the corrections $`t/J_H`$ can be accounted for perturbatively, if necessary.
Indirect magnon-phonon coupling via orbital sector.— An important piece of physics relevant to the experiment is still missed in model (2), that is the Jahn-Teller orbital-lattice coupling . This can be written as:
$$H_{\text{JT}}=g_0\left(Q_{2i}\sigma _i^x+Q_{3i}\sigma _i^z\right),$$
(13)
where $`Q_2`$ and $`Q_3`$ are the distortions of appropriate symmetry . The deformation energy including intersite correlations is given by
$$H_{\text{ph}}=\frac{1}{2}K\underset{i}{}𝑸_i^2+K_1\underset{ij_\gamma }{}\stackrel{~}{Q}_{3i}^\gamma \stackrel{~}{Q}_{3j}^\gamma ,$$
(14)
with $`\stackrel{~}{Q}_3^{x/y}=(Q_3\pm \sqrt{3}Q_2)/2`$, $`\stackrel{~}{Q}_3^z=Q_3`$, and $`𝑸=(Q_2,Q_3)`$. In general, the JT interaction strongly mixes the orbital and lattice dynamics, leading to a fluctuation of exchange bonds at low (phonon) frequencies. Here, we treat this problem only perturbatively as shown in Fig. 2.
We arrive at the following effective spin-phonon Hamiltonian:
$$H_{\text{s-ph}}=\underset{𝒑𝒒}{}\left(𝒈_{𝒑𝒒}𝑸_𝒒\right)B_𝒑^{}B_{𝒑+𝒒}.$$
(15)
The coupling constants $`g_{𝒑𝒒}^{(\alpha )}=g_0a_0(\eta _𝒒^{(\alpha )}\eta _𝒑^{(\alpha )}\eta _{𝒑+𝒒}^{(\alpha )})/S`$, where $`\eta _𝒒^{(2)}=\sqrt{3}(c_xc_y)/2`$, $`\eta _𝒒^{(3)}=c_z\frac{1}{2}c_x\frac{1}{2}c_y`$, $`c_\alpha \mathrm{cos}q_\alpha `$, and the parameter $`a_0=t(x+x_0)(f_{i+c}^{}f_i)(\sigma _i^z)_{\omega =0}`$.
Magnon self energies— We are now in the position to calculate the renormalization of the magnon dispersion. The leading $`1/S^2`$ corrections are shown in Fig. 3.
Charge and orbital susceptibilities in Figs. 3(a) and 3(b), respectively, are calculated using mean-field Green’s functions in slave boson $`h`$ and fermion $`f`$ subspaces. For the spectral density of JT phonons in Fig. 3(c) we use
$$\rho _\pm ^{\text{ph}}(\omega ,𝒒)=\frac{1}{\pi }\frac{\omega }{\omega _𝒒^\pm }\frac{\mathrm{\Gamma }}{(\omega \omega _𝒒^\pm )^2+\mathrm{\Gamma }^2}$$
(16)
accounting phenomenologically for the damping $`\mathrm{\Gamma }`$ due to coupling to orbital fluctuations. The phonon dispersions $`\omega _𝒒^\pm =\omega _0^{\text{ph}}[\kappa _{1𝒒}\pm (\kappa _{2𝒒}^2+\kappa _{3𝒒}^2)^{1/2}]^{1/2}`$ with $`\omega _0^{\text{ph}}=(K/M)^{1/2}`$ follow from Eq. (14). Here, $`\kappa _{1𝒒}=1+k_1(c_x+c_y+c_z)`$, $`\kappa _{2𝒒}=k_1\eta _𝒒^{(2)}`$, $`\kappa _{3𝒒}=k_1\eta _𝒒^{(3)}`$, and $`k_1=K_1/K`$. The expressions obtained from the diagrams in Fig. 3 contain summations over momentum space which we perform numerically. We find the effect of charge fluctuations on the magnon spectrum to be quite featureless and moderate (see Fig. 4),
which is due to the fact that the spectral density of charge fluctuations lies well above the magnon band. On contrary, the relatively low-energy fluctuations of the orbital and lattice degrees of freedom are found to affect the spin-wave dispersion in a peculiar way, particularly in $`(0,0,q)`$ and $`(0,q,q)`$ directions. Fig. 4 shows the magnon dispersion. The hopping amplitude $`t=0.4`$ eV is chosen to fit the spin stiffness in Pr<sub>0.63</sub>Sr<sub>0.37</sub>MnO<sub>3</sub>; further we use $`U_1=4`$ eV . The phonon contribution depends on the quantity $`(g_0a_0)^2/2KE_{\text{JT}}a_0^2`$. We set $`E_{\text{JT}}a_0^2=0.004`$ eV , $`\omega _0^{\text{ph}}=0.08`$ eV , $`\mathrm{\Gamma }=0.04`$ eV. Our key observation is the crucial effect of intersite correlations of JT distortions, controlled by $`k_1`$, on the magnon dispersion (see Fig. 5). To explain the experimental data we are forced to assume ferro-type correlations ($`k_1<0`$). We interpret this surprising result in the following way: Conventionally $`k_1>0`$ corresponding to AF order of JT distortions and orbitals is expected in undoped compounds . However, in the doped case charge mobility prefers a ferro-type local orientation of orbitals which minimizes the kinetic energy. This competition between JT and kinetic energies can be simulated by tuning $`k_1`$. At large enough doping, ferro-type orbital correlations are expected to prevail, hence effectively $`k_1<0`$. In fact, a ferro-type ordering of orbitals leading to a layered AFM spin structure is experimentally observed at dopings about $`x=0.5`$. As this instability is approached, low-energy, quasi static fluctuations of exchange bonds develop, yielding a magnon evolution as shown in Fig. 5. Remarkably, the small $`q`$ spin stiffness is not affected by this physics, while a strong reduction of $`T_C`$ by soft magnons at the zone boundary is predicted. This explains the origin of the anomalous enhancement of the $`D_s/T_C`$ ratio in low-$`T_C`$ manganites .
In summary, we have presented a theory of spin dynamics in a model relevant to manganites, emphasizing particularly the interplay between double-exchange physics and orbital-lattice dynamics. The unusual magnon dispersion in lower-$`T_C`$ compounds is explained as due to the proximity to orbital-lattice ordering. Apparently, strongly correlated orbital fluctuations play a crucial role in the physics of manganites.
We would like to thank H. Y. Hwang for stimulating discussions.
|
no-problem/9904/hep-ph9904386.html
|
ar5iv
|
text
|
# References
FLAVOURED MULTISKYRMIONS.
V.B.Kopeliovich and W.J.Zakrzewski<sup>∗∗</sup>
Institute for Nuclear Research of the Russian Academy of Sciences, Moscow 117312, Russia
<sup>∗∗</sup>Department of Mathematical Sciences, University of Durham, Durham DH1 3LE, UK
Static properties of multiskyrmions with baryon numbers up to 8 are calculated, based on the recently given rational map ansaetze. The spectra of baryonic systems with strangeness, charm and bottom are considered within a ”rigid oscillator” version of the bound state soliton model. It is suggested that the recently observed negatively charged nuclear fragment can be considered as a quantized strange multiskyrmion with $`B=6`$ or $`7`$. In agreement with previous observation, baryonic systems with charm or bottom have more chance to be bound by the strong interactions than strange baryonic systems. PACS 14.20.Mr, 14.20. Lq 1. The topological soliton models, and the Skyrme model among them , are attractive because of their simplicity and the possibility that they may describe well various properties of low energy baryons. The models of this kind provide also a very good framework within which to investigate the possibility of the existence of nuclear matter fragments with unusual properties, such as flavour being different from $`u`$ and $`d`$ quarks. In addition to being important by itself, this issue can have consequences in astrophysics and cosmology. It is well known that the relativistic many-body problems cannot be solved directly using the existing methods, and the chiral soliton approach may allow to overcome some of these difficulties.
The description of skyrmions with large baryon numbers is complicated because the explicit form of the fields was not known. A recent remarkable observation that the fields of the $`SU(2)`$ skyrmions can be approximated accurately by rational map ansaetze giving the values of masses close to their precise values, has simplified considerably their studies. Similar ansaetze have also been recently presented for $`SU(N)`$ skyrmions (which are not embeddings of $`SU(2)`$ fields).
Here we use the $`SU(2)`$ rational map ansaetze as the starting points for the calculation of static properties of bound states of skyrmions necessary for their quantization in the $`SU(3)`$ collective coordinates space. The energy and baryon number densities of the $`B=3`$ configuration have tetrahedral symmetry, for $`B=4`$ \- the octahedral (cubic) one , for $`B=5`$ \- $`D_{2d}`$-symmetry, for $`B=6`$ \- $`D_{4d}`$, for $`B=7`$ \- dodecahedral symmetry, and for $`B=8`$ \- $`D_{6d}`$ \- symmetry , etc. The minimization, with the help of a 3-dimensional variational $`SU(3)`$ program , lowers the energies of these configurations by few hundreds of $`Mev`$ and shows that they are local minima in the $`SU(3)`$ configuration space. The knowledge of the “flavour” moment of inertia and the $`\mathrm{\Sigma }`$-term allows us then to estimate the flavour excitation energies. The mass splittings of the lowest states with different values of strangeness, charm or bottom are calculated within the rigid oscillator version of the bound state approach. The binding energies of baryonic systems ($`BS`$) with different flavours are also estimated.
2. Let us consider simple $`SU(3)`$ extentions of the Skyrme model : we start with $`SU(2)`$ skyrmions (with flavour corresponding to $`(u,d)`$ quarks) and extend them to various $`SU(3)`$ groups, $`(u,d,s)`$, $`(u,d,c)`$, or $`(u,d,b)`$. We take the Lagrangian density of the Skyrme model, which in its well known form depends on parameters $`F_\pi ,F_D`$ and $`e`$ and can be written in the following way :
$$=\frac{F_\pi ^2}{16}Trl_\mu l^\mu +\frac{1}{32e^2}Tr[l_\mu ,l_\nu ]^2+\frac{F_\pi ^2m_\pi ^2}{16}Tr\left(U+U^{}2\right)+$$
$$+\frac{F_D^2m_D^2F_\pi ^2m_\pi ^2}{24}Tr\left(1\sqrt{3}\lambda _8\right)\left(U+U^{}2\right)+\frac{F_D^2F_\pi ^2}{48}Tr\left(1\sqrt{3}\lambda _8\right)\left(Ul_\mu l^\mu +l_\mu l^\mu U^{}\right).$$
$`\left(1\right)`$
$`USU\left(3\right)`$ is a unitary matrix incorporating chiral (meson) fields, and $`l_\mu =U^{}_\mu U`$. In this model $`F_\pi `$ is fixed at the physical value: $`F_\pi `$ = $`186`$ Mev. $`M_D`$ is the mass of $`K,D`$ or $`B`$ meson.
The flavour symmetry breaking $`(FSB)`$ in the Lagrangian $`(1)`$ is of the usual form, and was sufficient to describe the mass splittings of the octet and decuplets of baryons . The Wess-Zumino term, not shown here, plays an important role in the quantization procedure, but it does not contribute to the static masses of classical configurations .
We begin our calculations with $`USU(2)`$, as was mentioned above. The classical mass of $`SU(2)`$ solitons, in most general case, depends on $`3`$ profile functions: $`f,\alpha `$ and $`\beta `$. The general parametrization of $`U_0`$ for an $`SU(2)`$ soliton we use here is given by $`U_0=c_f+s_f\stackrel{}{\tau }\stackrel{}{n}`$ with $`n_z=c_\alpha `$, $`n_x=s_\alpha c_\beta `$, $`n_y=s_\alpha s_\beta `$, $`s_f=sinf`$, $`c_f=cosf`$, etc.
The flavour moment of inertia enters directly in the procedure of quantization -, and for arbitrary $`SU(2)`$ skyrmions is given by :
$$\mathrm{\Theta }_F=\frac{1}{8}\left(1c_f\right)\left[F_D^2+\frac{1}{e^2}\left(\left(\stackrel{}{}f\right)^2+s_f^2\left(\stackrel{}{}\alpha \right)^2+s_f^2s_\alpha ^2\left(\stackrel{}{}\beta \right)^2\right)\right]d^3\stackrel{}{r}.$$
$`\left(2\right)`$
It is simply connected with $`\mathrm{\Theta }_F^{(0)}`$ for the flavour symmetric case: $`\mathrm{\Theta }_F=\mathrm{\Theta }_F^{(0)}+(F_D^2/F_\pi ^21)\mathrm{\Gamma }/4,`$ $`\mathrm{\Gamma }`$ is defined in $`(3)`$ below. The isotopic moments of inertia are the components of the corresponding tensor of inertia , in our case this tensor of inertia is close to unit matrix multiplied by $`\mathrm{\Theta }_T`$. The quantities $`\mathrm{\Gamma }`$ (or $`\mathrm{\Sigma }`$-term), which defines the contribution of the mass term to the classical mass of solitons, and $`\stackrel{~}{\mathrm{\Gamma }}`$ also are used in the quantization procedure:
$$\mathrm{\Gamma }=\frac{F_\pi ^2}{2}\left(1c_f\right)d^3\stackrel{}{r},\stackrel{~}{\mathrm{\Gamma }}=\frac{1}{4}c_f\left[\left(\stackrel{}{}f\right)^2+s_f^2\left(\stackrel{}{}\alpha \right)^2+s_f^2s_\alpha ^2\left(\stackrel{}{}\beta \right)^2\right]d^3\stackrel{}{r}.$$
$`\left(3\right)`$
The masses of solitons, moments of inertia, $`\mathrm{\Gamma }`$ and $`\stackrel{~}{\mathrm{\Gamma }}`$ are presented in the Table below.
| $`B`$ | $`M_{cl}`$ | $`\mathrm{\Theta }_F^{\left(0\right)}`$ | $`\mathrm{\Theta }_T`$ | $`\mathrm{\Gamma }`$ | $`\stackrel{~}{\mathrm{\Gamma }}`$ | $`\omega _s`$ | $`\omega _c`$ | $`\omega _b`$ | $`\mathrm{\Delta }ϵ_s`$ | $`\mathrm{\Delta }ϵ_c`$ | $`\mathrm{\Delta }ϵ_b`$ |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| $`1`$ | $`1.702`$ | $`2.04`$ | $`5.55`$ | $`4.83`$ | $`15.6`$ | $`0.309`$ | $`1.542`$ | $`4.82`$ | | | |
| $`3`$ | $`4.80`$ | $`6.34`$ | $`14.4`$ | $`14.0`$ | $`27`$ | $`0.289`$ | $`1.504`$ | $`4.75`$ | $`0.041`$ | $`0.01`$ | $`0.03`$ |
| $`4`$ | $`6.20`$ | $`8.27`$ | $`16.8`$ | $`18.0`$ | $`31`$ | $`0.283`$ | $`1.493`$ | $`4.74`$ | $`0.020`$ | $`0.019`$ | $`0.06`$ |
| $`5`$ | $`7.78`$ | $`10.8`$ | $`23.5`$ | $`23.8`$ | $`35`$ | $`0.287`$ | $`1.505`$ | $`4.75`$ | $`0.027`$ | $`0.006`$ | $`0.05`$ |
| $`6`$ | $`9.24`$ | $`13.1`$ | $`25.4`$ | $`29.0`$ | $`38`$ | $`0.287`$ | $`1.504`$ | $`4.75`$ | $`0.019`$ | $`0.017`$ | $`0.05`$ |
| $`7`$ | $`10.6`$ | $`14.7`$ | $`28.7`$ | $`32.3`$ | $`44`$ | $`0.282`$ | $`1.497`$ | $`4.75`$ | $`0.017`$ | $`0.021`$ | $`0.06`$ |
| $`8`$ | $`12.2`$ | $`17.4`$ | $`33.4`$ | $`38.9`$ | $`47`$ | $`0.288`$ | $`1.510`$ | $`4.77`$ | $`0.018`$ | $`0.014`$ | $`0.02`$ |
Characteristics of the bound states of skyrmions with baryon numbers up to $`B=8`$. The classical mass of solitons $`M_{cl}`$ is in $`Gev`$, moments of inertia, $`\mathrm{\Gamma }`$ and $`\stackrel{~}{\mathrm{\Gamma }}`$ \- in $`Gev^1`$, the excitation frequences for flavour $`F`$, $`\omega _F`$ in $`Gev`$. The parameters of the model $`F_\pi =186Mev,e=4.12`$. The accuracy of calculations is better than $`1\%`$ for the masses and few $`\%`$ for other quantities. The $`B=1`$ quantities are shown for comparison. $`\mathrm{\Delta }ϵ_{s,c,b}`$, in $`Gev`$, are the changes of binding energies of lowest $`BS`$ with flavour $`s,c`$ or $`b`$, $`|F|=1`$, in comparison with usual $`(u,d)`$ nuclei (see Eq.(14)).
3. To quantize the solitons in $`SU(3)`$ configuration space, in the spirit of the bound state approach to the description of strangeness proposed in and used in , we consider the collective coordinates motion of the meson fields incorporated into the matrix $`U`$:
$$U(r,t)=R\left(t\right)U_0\left(O\left(t\right)\stackrel{}{r}\right)R^{}\left(t\right),R\left(t\right)=A\left(t\right)S\left(t\right),$$
$`\left(4\right)`$
where $`U_0`$ is the $`SU(2)`$ soliton embedded into $`SU(3)`$ in the usual way (into the left upper corner), $`A(t)SU(2)`$ describes $`SU(2)`$ rotations, $`S(t)SU(3)`$ describes rotations in the “strange”, “charm” or “bottom” directions, and $`O(t)`$ describes rigid rotations in real space.
$$S\left(t\right)=exp\left(i𝒟\left(t\right)\right),𝒟\left(t\right)=\underset{a=4,\mathrm{}7}{}D_a\left(t\right)\lambda _a,$$
$`\left(5\right)`$
$`\lambda _a`$ are Gell-Mann matrices of the $`(u,d,s)`$, $`(u,d,c)`$ or $`(u,d,b)`$ $`SU(3)`$ groups. The $`(u,d,c)`$ and $`(u,d,b)`$ $`SU(3)`$ groups are quite analogous to the $`(u,d,s)`$ one. For the $`(u,d,c)`$ group a simple redefiniton of hypercharge should be made. For the $`(u,d,s)`$ group, $`D_4=(K^++K^{})/\sqrt{2}`$, $`D_5=i(K^+K^{})/\sqrt{2}`$, etc. And for the $`(u,d,c)`$ group $`D_4=(D^0+\overline{D}^0)/\sqrt{2}`$, etc.
The angular velocities of the isospin rotations are defined in the standard way: $`A^{}\dot{A}=i\stackrel{}{\omega }\stackrel{}{\tau }/2.`$ We shall not consider here the usual space rotations explicitly because the corresponding moments of inertia for $`BS`$ are much greater than isospin moments of inertia, and for lowest possible values of angular momentum $`J`$ the corresponding quantum correction is either exactly zero (for even $`B`$), or small.
The field $`D`$ is small in magnitude, at least, of order $`1/\sqrt{N_c}`$, where $`N_c`$ is the number of colours in $`QCD`$. Therefore, an expansion of the matrix $`S`$ in $`D`$ can be made safely. To the lowest order in field $`D`$ the Lagrangian of the model $`(1)`$ can be written as
$$L=M_{cl,B}+4\mathrm{\Theta }_{F,B}\dot{D}^{}\dot{D}\left[\mathrm{\Gamma }_B\left(\frac{F_D^2}{F_\pi ^2}m_D^2m_\pi ^2\right)+\stackrel{~}{\mathrm{\Gamma }}_B\left(F_D^2F_\pi ^2\right)\right]D^{}Di\frac{N_cB}{2}\left(D^{}\dot{D}\dot{D}^{}D\right).$$
$`\left(7\right)`$
Here and below $`D`$ is the doublet $`K^+,K^0`$ ($`D^0,D^{}`$, or $`B^+,B^0`$). We have kept the standard notation for the moment of inertia of the rotation into the “flavour” direction $`\mathrm{\Theta }_F`$ for $`\mathrm{\Theta }_s,\mathrm{\Theta }_c`$ or $`\mathrm{\Theta }_b`$ (the index $`c`$ denotes the charm quantum number, except in $`N_c`$). The contribution proportional to $`\stackrel{~}{\mathrm{\Gamma }}_B`$ is suppressed in comparison with the term $`\mathrm{\Gamma }`$ by the small factor $`(F_D^2F_\pi ^2)/m_D^2`$, and is more important for strangeness. The term proportional to $`N_cB`$ in $`(7)`$ arises from the Wess-Zumino term in the action and is responsible for the difference of the excitation energies of strangeness and antistrangeness (flavour and antiflavour in general case) .
Following the canonical quantization procedure the Hamiltonian of the system, including the terms of the order of $`N_c^0`$, takes the form :
$$H_B=M_{cl,B}+\frac{1}{4\mathrm{\Theta }_{F,B}}\mathrm{\Pi }^{}\mathrm{\Pi }+\left(\mathrm{\Gamma }_B\overline{m}_D^2+\stackrel{~}{\mathrm{\Gamma }}_B\left(F_D^2F_\pi ^2\right)+\frac{N_c^2B^2}{16\mathrm{\Theta }_{F,B}}\right)D^{}D+i\frac{N_cB}{8\mathrm{\Theta }_{F,B}}\left(D^{}\mathrm{\Pi }\mathrm{\Pi }^{}D\right).$$
$`\left(8\right)`$
$`\overline{m}_D^2=\left(F_D^2/F_\pi ^2\right)m_D^2m_\pi ^2`$. The momentum $`\mathrm{\Pi }`$ is canonically conjugate to variable $`D`$. Eq. $`(8)`$ describes an oscillator-type motion of the field $`D`$ in the background formed by the $`(u,d)`$ $`SU(2)`$ soliton. After the diagonalization which can be done explicitly following , the normal-ordered Hamiltonian can be written as
$$H_B=M_{cl,B}+\omega _{F,B}a^{}a+\overline{\omega }_{F,B}b^{}b+O\left(1/N_c\right),$$
$`\left(9\right)`$
with $`a^{}`$, $`b^{}`$ being the operators of creation of strangeness, i.e., antikaons, and antistrangeness (flavour and antiflavour) quantum number, $`\omega _{F,B}`$ and $`\overline{\omega }_{F,B}`$ being the frequences of flavour (antiflavour) excitations. $`D`$ and $`\mathrm{\Pi }`$ are connected with $`a`$ and $`b`$ in the following way :
$$D^i=\left(b^i+a^i\right)/\sqrt{N_cB\mu _{F,B}},\mathrm{\Pi }^i=\sqrt{N_cB\mu _{F,B}}\left(b^ia^i\right)/\left(2i\right)$$
$`\left(10\right)`$
with
$$\mu _{F,B}=\left[1+16\left(\overline{m}_D^2\mathrm{\Gamma }_B+\left(F_D^2F_\pi ^2\right)\stackrel{~}{\mathrm{\Gamma }}_B\right)\mathrm{\Theta }_{F,B}/\left(N_cB\right)^2\right]^{1/2}.$$
For the lowest states the values of $`D`$ are small: $`D[16\mathrm{\Gamma }_B\mathrm{\Theta }_{F,B}\overline{m}_D^2+N_c^2B^2]^{1/4},`$ and increase, with increasing flavour number $`|F|`$ like $`(2|F|+1)^{1/2}`$. As was noted in , deviations of the field $`D`$ from the vacuum decrease with increasing mass $`m_D`$, as well as with increasing number of colours $`N_c`$, and the method works for any $`m_D`$ (and also for charm and bottom quantum numbers).
The excitation frequences $`\omega `$ and $`\overline{\omega }`$ are:
$$\omega _{F,B}=N_cB\left(\mu _{F,B}1\right)/\left(8\mathrm{\Theta }_{F,B}\right),\overline{\omega }_{F,B}=N_cB\left(\mu _{F,B}+1\right)/\left(8\mathrm{\Theta }_{F,B}\right).$$
$`\left(11\right)`$
As was observed in , the difference $`\overline{\omega }_{F,B}\omega _{F,B}=N_cB/(4\mathrm{\Theta }_{F,B})`$ coincides, to the leading order in $`N_c`$ with the expression obtained in the collective coordinates approach .
The $`FSB`$ in the flavour decay constants, i.e. the fact that $`F_K/F_\pi 1.22`$ and $`F_D/F_\pi =1.7\pm 0.2`$ (we take $`F_D/F_\pi =1.5`$ and $`F_B/F_\pi =2`$) leads to the increase of the flavour excitation frequences, in better agreement with data for charm and bottom . It also leads to some increase of the binding energies of $`BS`$ .
The behaviour of static characteristics of multiskyrmions and flavour excitation frequences shown in the Table is similar to that obtained in for toroidal configurations with $`B=2,3,4`$. The flavour inertia increases with $`B`$ almost proportionally to $`B`$. The frequences $`\omega _F`$ are smaller for $`B3`$ than for $`B=1`$.
4. The terms of the order of $`N_c^1`$ in the Hamiltonian, which depend on the angular velocities of rotations in the isospin and the usual space and which describe the zero-mode contributions are not crucial but important for the numerical estimates of spectra of baryonic systems.
In the rigid oscillator model the states predicted do not correspond to the definite $`SU(3)`$ or $`SU(4)`$ representations. How this can be remedied was shown in . For example, the state with $`B=1`$, $`|F|=1`$, $`I=0`$ should belong to the octet of $`(u,d,s)`$, or $`(u,d,c)`$, $`SU(3)`$ group, if $`N_c=3`$.
Here we consider quantized states of $`BS`$ which belong to the lowest possible $`SU(3)`$ irreps $`(p,q)`$, $`p+2q=3B`$: $`p=0,q=3B/2`$ for even $`B`$, and $`p=1,q=(3B1)/2`$ for odd $`B`$. For $`B=3,\mathrm{\hspace{0.17em}5}`$ and $`7`$ they are $`35,\mathrm{\hspace{0.17em}80}`$ and $`143`$-plets, for $`B=4,\mathrm{\hspace{0.17em}6}`$ and $`8`$ \- $`28`$, $`55`$ and $`91`$-plets. Since we are interested in the lowest energy states, we discuss here the baryonic systems with the lowest allowed angular momentum,ie $`J=0`$, for $`B=4,\mathrm{\hspace{0.33em}6}`$ and $`8`$. For odd $`B`$ the quantization of $`BS`$ meets some difficulties, but the correction to the energy of quantized states due to nonzero angular momentum is small and decreases with increasing $`B`$ since the corresponding moment of inertia increases proportionally to $`B^2`$. Moreover, the $`J`$-dependent correction to the energy cancels in the differences of energies of flavoured and flavourless states which we discuss.
For the energy difference between the state with flavour $`F`$ belonging to the $`(p,q)`$ irrep, and the ground state with $`F=0`$ and the same angular momentum and $`(p,q)`$ we obtain:
$$\mathrm{\Delta }E_{B,F}=\left|F\right|\omega _{F,B}+\frac{\mu _{F,B}1}{4\mu _{F,B}\mathrm{\Theta }_{F,B}}\left[I\left(I+1\right)T_r\left(T_r+1\right)\right]+\frac{\left(\mu _{F,B}1\right)\left(\mu _{F,B}2\right)}{8\mu _{F,B}^2\mathrm{\Theta }_{F,B}}I_F\left(I_F+1\right),$$
$`\left(12\right)`$
$`T_r=p/2`$ is the quantity analogous to the “right” isospin $`T_r`$, in the collective coordinates approach , and $`\stackrel{}{T_r}=\stackrel{}{I}_{bf}\stackrel{}{I_F}`$. Clearly, the binding energy of multiskyrmions is cancelled in Eq. $`(12)`$. For the states with maximal isospin $`I=T_r+|F|/2`$ the energy difference can be simplified to:
$$\mathrm{\Delta }E_{B,F}=\left|F\right|\left[\omega _{F,B}+T_r\frac{\mu _{F,B}1}{4\mu _{F,B}\mathrm{\Theta }_{F,B}}+\frac{\left(\left|F\right|+2\right)}{8\mathrm{\Theta }_{F,B}}\frac{\left(\mu _{F,B}1\right)^2}{\mu _{F,B}^2}\right].$$
$`\left(13\right)`$
This difference depends on the flavour moment of inertia but not on $`\mathrm{\Theta }_T`$. In the case of antiflavour excitations we have the same formulas, with the substitution $`\mu \mu `$. For even $`B,T_r=0`$, for odd $`B`$, $`T_r=1/2`$ for the lowest $`SU(3)`$ irreps. It follows from $`(12)`$ and $`(13)`$ that when some nucleons are replaced by flavoured hyperons in $`BS`$ the binding energy of the system changes by
$$\mathrm{\Delta }ϵ_{B,F}=\left|F\right|\left[\omega _{F,1}\omega _{F,B}\frac{3\left(\mu _{F,1}1\right)}{8\mu _{F,1}^2\mathrm{\Theta }_{F,1}}T_r\frac{\mu _{F,B}1}{4\mu _{F,B}\mathrm{\Theta }_{F,B}}\frac{\left(\left|F\right|+2\right)}{8\mathrm{\Theta }_{F,B}}\frac{\left(\mu _{F,B}1\right)^2}{\mu _{F,B}^2}\right]$$
$`\left(14\right)`$
For strangeness Eq. $`(14)`$ is negative indicating that stranglets should have binding energies smaller than those of nuclei, or can be unbound. Since $`\mathrm{\Theta }_{F,B}`$ increases with increasing $`B`$ and $`m_D`$ this leads to the increase of binding with increasing $`B`$ and mass of the ”flavour”, in agreement with . For charm and bottom Eq. $`(14)`$ is positive for $`B3`$, see the Table for the case $`|F|=1`$.
The nuclear fragments with sufficiently large values of strangeness (or bottom) can be found in experiments as fragments with negative charge $`Q`$, according to the well known relation $`Q=T_3+(B+S)/2`$ (similarly for the bottom number). One event of a long lived nuclear fragment with mass about $`7.4Gev`$ was reported in . Using the above formulas it is not difficult to establish that this fragment can be the state with $`B=S=6`$, or $`B=7`$ and strangeness $`S=3`$. In view of some uncertainty of present calculation - the rigid oscillator version of the model leads to overestimation of flavour excitation energies - greater values of strangeness, by $`1`$ or $`2`$ units can be necessary to obtain the observed value of mass.
As in the $`B=1`$ case the absolute values of masses of multiskyrmions are controlled by the poorly known loop corrections to the classic masses, or the Casimir energy. And as was done for the $`B=2`$ states, , the renormalization procedure is necessary to obtain physically reasonable values of the masses. This generates an uncertainty of about several tens of $`Mev`$, as the binding energy of the deuteron is $`30Mev`$ instead of the measured value $`2.23Mev`$, so $`30Mev`$ characterises the uncertainty of our approach . But this uncertainty is cancelled in the differences of binding energies $`\mathrm{\Delta }ϵ`$ shown in the Table.
5. Using rational map ansaetze as starting configurations we have calculated the static characteristics of bound skyrmions with baryon numbers up to $`8`$. The excitation frequences for different flavours - strangeness, charm and bottom - have been calculated using a rigid oscillator version of the bound state approach of the chiral soliton models. This variant of the model overestimates the mass splitting of strange hyperons when $`FSB`$ in decay constant $`F_K`$ is included, but works better for $`c`$ and $`b`$ flavours . Our previous conclusion that $`BS`$ with charm and bottom have more chances to be bound respectively to strong decay than strange $`BS`$ is reinforced by the present investigation. This conclusion takes place also in $`FS`$ case, $`F_D=F_\pi `$.
Consideration of the $`BS`$ with “mixed” flavours is possible in principle, but would be technically more involved. Our results agree qualitatively with the results of where the strangeness excitation frequences had been calculated within the bound state approach. The difference is, however, in the behaviour of excitation frequences: we have found that they decrease when the baryon number increases from $`B=1`$ thus increasing the binding energy of corresponding $`BS`$.
The charmed baryonic systems with $`B=3,\mathrm{\hspace{0.17em}4}`$ were considered in within a potential approach. The $`B=3`$ systems were found to be very near the threshold and the $`B=4`$ system was found to be stable with respect to the strong decay, with a binding energy of $`10Mev`$. Further experimental searches for the baryonic systems with flavour different from $`u`$ and $`d`$ could shed more light on the dynamics of heavy flavours in baryonic systems.
This work has been supported by the UK PPARC grant: PPA/V/S/1999/00004.
VBK is indebted to the Institute for Nuclear Theory at the University of Washington where the work was initiated for its hospitality and DOE for support.
References
|
no-problem/9904/hep-th9904028.html
|
ar5iv
|
text
|
# The Generating Formula for The Solutions to The Associativity Equations
(March 9, 1999)
## Abstract
An exact formula for the solutions to the WDVV equation in terms of horizontal sections of the corresponding flat connection is found.
1. Introduction. The associativity equations (or WDDV equations) appeared in the classification problem for the topological field theories at the early 90’s, see , . During the last years these equations has attracted a great interest due to connections with the enumerative geometry (Gromov – Witten invariants ), quantum cohomology and the Whitham theory .
In it was noticed that the classification problem for the topological field theories is equivalent to the classification of the Egoroff metrics of special type. Egoroff metrics are flat diagonal metrics
$$ds^2=\underset{i=1}{\overset{n}{}}h_i^2(u)(du^i)^2,u=(u^1,\mathrm{},u^n)$$
(1)
such that $`_ih_j^2(u)=_jh_i^2(u)`$, where $`_i=/u^i`$. It turns out that for every Egoroff metric with the additional constraints $`_{j=1}^n_jh_i=0`$ the functions
$$c_{kl}^m(x)=\underset{i=1}{\overset{n}{}}\frac{x^m}{u^i}\frac{u^i}{x^k}\frac{u^i}{x^l},$$
(2)
where $`x^k(u)`$ are the flat coordinates of the metric (1), satisfy
$$c_{ij}^k(x)c_{km}^l(x)=c_{jm}^k(x)c_{ik}^l(x),$$
(3)
which are the associativity conditions for the algebra $`\varphi _k\varphi _l=c_{kl}^m\varphi _m`$. Moreover, there exists a function $`F(x)`$ such that its third derivatives are equal to
$$\frac{^3F(x)}{x^kx^lx^m}=c_{klm}(x)=\eta _{mi}c_{kl}^i(x),\text{where}\eta _{pq}=\underset{i=1}{\overset{n}{}}h_i^2(u)\frac{u^i}{x^p}\frac{u^i}{x^q}.$$
(4)
In the topological field theory with $`n`$ primary fields $`\varphi _k`$ function $`F`$ plays role of the partition function. There also exist constants $`r^m`$ such that $`\eta _{kl}=r^mc_{klm}(x)`$.
Equations (3) and the existence of the function $`F`$ such that (4) holds are equivalent to the compatibility conditions of the linear system (see )
$$_k\mathrm{\Phi }_l\lambda c_{kl}^m\mathrm{\Phi }_m=0,$$
(5)
where $`\lambda `$ is a spectral parameter.
In few special cases the function $`F`$ was found explicitly (see , , ). However, in general case an expression for $`F`$ in terms of the flat sections of the connection $`_k=/x^k\lambda c_{kl}^m`$ has been unknown. The main goal of these notes is to write down such an expression. It was motivated by the results of , where explicit formulas for the algebraic-geometrical solutions to the associativity equations were found. Though the general formula of (Theorem 5.1) is correct, in the final formula (Theorem 5.2) one of the terms was missed. Here we take an opportunity to fix it.
2. Let us consider a solution $`\beta _{ij}(u)=\beta _{ji}(u)`$ to the Darboux – Egoroff system
$$_k\beta _{ij}=\beta _{ik}\beta _{kj},ijk;\underset{m=1}{\overset{n}{}}_m\beta _{ij}=0,ij.$$
(6)
Following we fix the unique Egoroff metric (1) corresponding to this solution by defining the Lame coefficients $`h_i(u)`$ from the system
$$_jh_i(u)=\beta _{ij}(u)h_j(u),ij,_ih_i(u)=\underset{ji}{}\beta _{ij}(u)h_j(u),$$
(7)
and initial conditions $`h_i(0)=1`$, $`i=1,\mathrm{},n`$. This system is compatible due to (6). The condition $`\beta _{ij}=\beta _{ji}`$ implies that the above defined metric is in fact Egoroff metric.
The flat coordinates $`x^1,\mathrm{},x^n`$ of this metric can be found from the linear system
$$_i_jx^k=\mathrm{\Gamma }_{ij}^i_ix^k+\mathrm{\Gamma }_{ji}^j_jx^k,ij;_i_ix^k=\underset{j=1}{\overset{n}{}}\mathrm{\Gamma }_{ii}^j_jx^k,$$
(8)
where $`\mathrm{\Gamma }_{ij}^k`$ are Christoffel symbols: $`\mathrm{\Gamma }_{ij}^i=_jh_i/h_i`$, $`\mathrm{\Gamma }_{ii}^j=(2\delta _{ij}1)(h_i_jh_i)/(h_j^2)`$. We choose the following initial conditions: $`x^k(0)=0`$, $`_{k,l}\eta _{kl}_ix^k(0)_jx^l(0)=\delta _{ij}`$. Here $`\eta _{kl}`$ is the fixed symmetric nondegenerate matrix.
The Darboux – Egoroff system can be regarded as the compatibility conditions of the following linear system:
$`_j\mathrm{\Psi }_i(u,\lambda )`$ $`=\beta _{ij}(u)\mathrm{\Psi }_j(u,\lambda ),ij`$ (9)
$`_i\mathrm{\Psi }_i(u,\lambda )`$ $`=\lambda \mathrm{\Psi }_i(u,\lambda ){\displaystyle \underset{ki}{}}\beta _{ik}(u)\mathrm{\Psi }_k(u,\lambda ),`$ (10)
Here $`\mathrm{\Psi }_i=(\mathrm{\Psi }_i^1,\mathrm{},\mathrm{\Psi }_i^n)`$, $`i=1,\mathrm{},n`$, are vector-functions which are formal power series in $`\lambda `$. There exists a unique solution of this linear system with the initial conditions $`\mathrm{\Psi }_i^k(0,\lambda )=\lambda _ix^k(0)`$.
It follows from (9), (10) that the expansion of $`\mathrm{\Psi }_i`$ has the form
$$\mathrm{\Psi }_i(u,\lambda )=\underset{s=0}{\overset{\mathrm{}}{}}\frac{_i\xi _s^k(u)}{h_i(u)}\lambda ^s,$$
where $`\xi _0^k=r^k`$ are constants (we will define them later), $`\xi _1^k(u)=x^k(u)`$ are flat coordinates and $`\xi _s^k`$ for $`s2`$ obey the following equations
$`_i_j\xi _s^k`$ $`=\mathrm{\Gamma }_{ij}^i_i\xi _s^k+\mathrm{\Gamma }_{ji}^j_j\xi _s^k,ij,`$ (11)
$`_i_i\xi _s^k`$ $`={\displaystyle \underset{j=1}{\overset{n}{}}}\mathrm{\Gamma }_{ii}^j_j\xi _s^k+_i\xi _{s1}^k,`$ (12)
with the initial conditions $`\xi _s^k(0)=0`$, $`_i\xi _s^k(0)=0`$. The equations
$$\frac{^2\xi _s^m}{x^kx^l}=\underset{p=1}{\overset{n}{}}c_{kl}^p\frac{\xi _{s1}^m}{x^p},$$
(13)
where $`c_{kl}^p`$ are defined by (2), follow immediately from (11), (12). We will denote $`\xi _2^k(u)`$ and $`\xi _3^k(u)`$ by $`y^k(u)`$ and $`z^k(u)`$, respectively.
System (9,10) implies that $`\lambda \mathrm{\Psi }_i=_{j=1}^n_j\mathrm{\Psi }_i`$ and, therefore, $`_{i=1}^n_i\xi _s^k(u)=\xi _{s1}^k`$ for $`s1`$. The last equality for $`s=1`$ is, in fact, the definition of the constants $`r^k`$.
3. Let $`\psi `$ be given by the formula
$$\lambda \psi (u,\lambda )=\underset{i=1}{\overset{n}{}}h_i(u)\mathrm{\Psi }_i(u,\lambda ).$$
(14)
It’s straightforward to check that $`_i\psi (u,\lambda )=h_i(u)\mathrm{\Psi }_i(u,\lambda )`$. As a formal power series on $`\lambda `$ the $`k`$-th component of the vector-function $`\psi (u,\lambda )`$ has the form
$$\psi ^k(u,\lambda )=r^k+x^k(u)\lambda +y^k(u)\lambda ^2+z^k(u)\lambda ^3+\underset{s=4}{\overset{\mathrm{}}{}}\xi _s^k(u)\lambda ^s.$$
(15)
Note that (13) implies that the functions $`\mathrm{\Phi }_k(x)=\psi (x)/x^k`$ satisfy (5). Moreover,
$$\lambda \psi (x)=\underset{k=1}{\overset{n}{}}r^k\mathrm{\Phi }_k(x)$$
and thus the function $`\psi `$ can be regarded as a generating function for the flat sections of the above-defined connection $`_k`$.
###### Lemma 1.
The vector-functions $`\mathrm{\Psi }_i(u,\lambda )`$ satisfy the equations
$$\mathrm{\Psi }_i(u,\lambda ),\mathrm{\Psi }_j(u,\lambda )=\delta _{ij}\lambda ^2,$$
(16)
the scalar product given by the matrix $`\eta _{kl}`$: $`A,B=\eta _{kl}A^kB^l`$.
###### Lemma 2.
The scalar product
$$\frac{1}{\lambda }_i\psi (u,\lambda ),\psi (u,\lambda )$$
(17)
does not depend on $`\lambda `$.
Both lemmas can be proved in the similar way. First we establish the required properties of the scalar products (16), (17) at the point $`u=0`$. Then we show that they satisfy certain differential equations. The Uniqueness Theorem implies both statements.
###### Corollary.
The functions $`x^k(u)`$, $`y^k(u)`$ and $`z^k(u)`$ satisfy the following relation:
$$\underset{q=1}{\overset{n}{}}\eta _{kq}y^q=\underset{p,q=1}{\overset{n}{}}\eta _{pq}\left(x^q\frac{y^p}{x^k}r^q\frac{z^p}{x^k}\right).$$
(18)
Proof. Consider the scalar product $`{\displaystyle \frac{1}{\lambda }}{\displaystyle \frac{\psi }{x^k}}(u,\lambda ),\psi (u,\lambda )`$. Since $`\psi /x^k`$ is the linear combination of $`\psi /u^i`$ Lemma 2 implies that this scalar product is independent of $`\lambda `$. On the other hand, it can be presented as a power series in $`\lambda `$
$$\frac{1}{\lambda }\frac{\psi }{x^k},\psi ^\sigma =\frac{1}{\lambda }\underset{p,q=1}{\overset{n}{}}\eta _{pq}\left(\delta _k^p\lambda +\frac{y^p}{x^k}\lambda ^2+\frac{z^p}{x^k}\lambda ^3+\mathrm{}\right)\left(r^qx^q\lambda +y^q\lambda ^2z^q\lambda ^3+\mathrm{}\right)$$
(19)
(here $`\psi ^\sigma (u,\lambda )=\psi (u,\lambda )`$). Therefore, all but the first coefficients of the series (19) should equal zero. Applying this argument to the coefficient of $`\lambda ^2`$ we obtain (18).
###### Theorem.
The function $`F(x)=F(u(x))`$ defined by the formula
$$F(u)=\frac{1}{2}\underset{p,q=1}{\overset{n}{}}\eta _{pq}\left(x^q(u)y^p(u)r^qz^p(u)\right)$$
(20)
satisfies the equation (4).
Proof. Let us notice that Corollary implies $`F/x^k=_{q=1}^n\eta _{kq}y^q`$. Now the statement of the Theorem is direct implication of (2) and (13) for $`s=2`$.
From the previous formulas it follows that $`F`$ satisfies the renormalization group type equation:
$$F(x)\underset{k=1}{\overset{n}{}}x^k\frac{F}{x^k}=\underset{p,q=1}{\overset{n}{}}\eta _{pq}r^qz^p.$$
In our next paper we hope to obtain more general formula, where the function $`F`$ depends on infinitely many variables corresponding to gravitational descendants of the primary fields $`\varphi _k`$.
|
no-problem/9904/astro-ph9904371.html
|
ar5iv
|
text
|
# Detection of X-ray emission from the host clusters of 3CR quasars
## 1 Introduction
The two most powerful FR II radio sources in the nearby Universe – Cyg A and 3C295 – are each located at the centre of a dense, moderately rich cluster of galaxies. While such an environment is exceptional for a low-redshift FR II galaxy, it appears to be common around powerful radio objects at earlier epochs. Above a redshift of 0.5, radio-loud objects (both the quasars and radio galaxies) are inferred to lie in clusters of galaxies of moderate optical richness. The evidence for such an environment includes optical and near-IR galaxy counts (Yee & Green 1987; Yates et al 1989; Hill & Lilly 1991; Ellingson et al 1991; Dickinson 1997), high gas pressures within a radius of 30 kpc (Crawford & Fabian 1989; Forbes et al 1990; Bremer et al 1992; Durret et al 1994), cD-type host galaxy profiles (Best et al 1998), a gravitational arc (Deltorn et al 1997), and lensing shear of surrounding field galaxies (Bower & Smail 1997). The properties of the radio source itself also imply the presence of a confining medium: a large-scale working surface on which the jets form the radio lobes; a steep radio spectrum; and a high minimum pressure in regions of relaxed radio structure (Bremer et al 1992). A Faraday depolarization asymmetry (Garrington & Conway 1991), the distortion and compression of high-redshift radio source morphologies (Hintzen et al 1983; Barthel & Miley 1988) and sources with very high Faraday rotation measures (Carilli et al 1994; Carilli et al 1997) all corroborate the inference of a dense, clumpy medium surrounding the radio source. Thus it appears that the deepest potential wells we can readily pinpoint at $`z1`$ are those around powerful radio sources.
The cluster distribution at high redshift can provide a stringent cosmological test (see e.g. Donahue et al 1998), and can also be compared to the X-ray luminosity function of clusters at low redshift (eg Ebeling et al 1997). Whilst it may result in a sample of clusters biased to only those that can host an active nucleus, using radio sources to identify the location of deep potential wells is a promising way of finding clusters out to and beyond a redshift $`z1`$ (Crawford 1997). Current X-ray surveys of clusters detected from the ROSAT All-Sky Survey (eg Ebeling et al 1998) do not reach sufficiently faint flux levels, and studies of deep serendipitous X-ray pointings (e.g. Rosati et al 1998) cover only a small fraction of the sky. The first step, however, is simply to confirm that powerful radio quasars beyond a redshift of a half really do lie at the centre of clusters of galaxies.
The clearest way to determine directly the presence of a cluster of galaxies is to detect thermal X-ray emission from its hot intracluster medium. A certain degree of success has been achieved in detecting and spatially resolving the X-ray emission around distant ($`0.5<z<2`$) radio galaxies using ROSAT (Crawford & Fabian 1993; Worrall et al. 1994; Crawford & Fabian 1995a, 1996a,b; Crawford 1997; Dickinson 1997; Hardcastle, Lawrence & Worrall 1998; Carilli et al 1998). Any X-rays emitted by the central bright nucleus of radio galaxies are assumed to be absorbed along the line of sight, as observed for the powerful low-redshift radio galaxy Cygnus-A (Ueno et al 1994). The inferred bolometric luminosity of the X-ray sources associated with the radio galaxies is $`0.718\times 10^{44}`$$`\mathrm{erg}\mathrm{s}^1`$, easily compatible with that expected from moderately rich clusters of galaxies around the radio sources. There could also be a contribution to the extended X-ray emission from inverse Compton scattering of the hidden quasar radiation (eg Brunetti, Setti & Comastri 1997).
In the case of radio quasars, however, the X-ray detection of the spatially extended environment is complicated by the presence of bright spatially-unresolved X-ray emission from the active nucleus. The ROSAT PSPC did not combine the necessary sensitivity with a sufficiently good point-response function, needed to both detect and resolve any cluster emission around quasars. Upper limits of $`1.63.5\times 10^{44}`$$`\mathrm{erg}\mathrm{s}^1`$(in the 0.1-2.4 keV rest-frame band) to X-ray emission from the environment of three radio-loud quasars have been derived from ROSAT HRI data (Hall et al 1995, 1997) assuming the cluster emission profile is modelled by a King law.
We have also obtained ROSAT HRI data to spatially resolve and detect the extended emission from the intracluster medium around each of a small sample of intermediate-redshift radio-loud quasars. The detection of such a component is, however, complicated by the wobble of the spacecraft during the observation. This occurs on a $`402`$ s period, and when the attitude of the spacecraft is not well reconstructed, leads to smearing of the point-spread function (PSF). The bright emission from the quasar nucleus can then contaminate the outer regions where we hope to detect emission from any surrounding cluster, and this has so far hindered our progress in interpreting the data. In this paper, however, we present an analysis of our ROSAT HRI data taken of seven intermediate-redshift (0.1$`<z<`$0.8) radio-loud quasars, which employs a new correction for the spacecraft wobble derived by Harris et al. (1998).
A contemporaneous and independent analysis of an overlapping dataset using this technique has been carried out by Hardcastle & Worrall (1999), who obtain similar results.
## 2 Observations and analysis
We use the ROSAT data of intermediate-redshift, radio-loud quasars for which there is prior evidence from other wavebands for a cluster environment (see notes on individual quasars for details). We preferentially selected quasars of only moderate X-ray luminosity in order to minimise the contrast between the nuclear emission and any cluster emission. These targets were supplemented by data available from the ROSAT public archive on 3C273 and 3C215. We also include the observations of H1821+643 to form a comparison to the results of Hall etal (1997). The observations used, and details of the quasars are listed in Table 1.
### 2.1 Wobble-correction technique
The spatial analysis of ROSAT HRI observations is often complicated by smearing of the image on the order of 10 arcsec (Morse 1994). This degradation of the instrinsic resolution of the HRI instrument (5 arcsec) can be induced by errors in the aspect solution associated with the wobble of the ROSAT spacecraft, or with the reacquisition of the guide stars. To counteract this effect, we use the wobble-correction technique of Harris et al (1998) which minimizes the spatial smearing of the sources.
The technique is based on the simple assumption that in the case of a stable roll angle (i.e. the same guide star configuration) the aspect error is repeated through each cycle of the wobble. We thus select data only from the longest constant roll angle interval of an observation and folded these data over the ROSAT 402 s wobble phase. This phase was grouped into a number of intervals (5, 10 or 20) in order to calculate phase-resolved sub-images. The centroid of each of the sub-images was calculated by fitting a 2d-gaussian to the brightest source near the field center. The wobble-corrected HRI events list is reconstructed by adding the sub-images which have been shifted to the centroid position of the uncorrected HRI image. If a sub-image contains too few photons with which to determine the centroid, then the position of the sub-image is not changed. This makes sure that source extension cannot be produced by a failure of the centroid determination in one or more sub-images.
For X-ray sources with countrates $``$0.1$`\mathrm{ct}\mathrm{s}^1`$the method can reduce the full width half maximum by about 30 per cent (cf. Harris et al. 1998). The efficacy of this technique has been tested on HRI observations of low-luminosity X-ray AGN, some of which show no sign of extended X-ray emission once corrected for wobble (Lehmann et al. 1999). We cannot apply the wobble-correction technique to all possible data on suitable targets in the literature, as it can only be applied to observations that have a sufficient number of photons within a constant spacecraft roll angle (cf. Table 1).
### 2.2 Radial profile analysis
Our aim is to search for spatially extended X-ray emission originating in any intracluster medium around the quasars. We have derived a background-subtracted radial profile from the un-corrected and wobble-corrected HRI data of the X-ray source associated with each quasar. First we determined the centroid position of the X-ray source by fitting a 2D Gaussian. Then we calculated the counts per arcsec<sup>2</sup> within the rings around the centroid position from 0 arcsec to 2.5 arcsec in steps of 0.5 arcsec, from 2.5 arcsec to 15 arcsec in steps of 2.5 arcsec, from 15 arcsec to 100 arcsec in steps of 12.5 arcsec and from 100 arcsec to 1000 arcsec in steps of 100 arcsec. The background value was calculated as the median from the 6 rings between 300 arcsec and 900 arcsec. Finally we subtracted the background value from each ring.
First, however, we need to accurately model the HRI PSF which will allow us to remove the contaminating spill-over of light from the bright quasar nucleus. The standard HRI PSF of David et al (1995) does not provide a good fit to the radial profile of observed point sources (see Fig 1). This deviation is particularly acute where the PSF shows a sharp drop at radii between 10 and 30 arcsec, where we expect the contrast between the nuclear source and any extended cluster emission to begin to show. Instead we have determined a good analytical characterization to an empirical PSF derived from observations of 21 ROSAT Bright survey stars, each of which has undergone the same wobble correction procedure as the quasar data. Our best fit is the sum of two Gaussians and a power-law component (Fig 1). Assuming that this PSF forms a good model for the spillover of quasar nuclear light, we fix the relative normalizations and widths of the three components and leave only the overall normalization $`n`$ of this profile as the only free variable in the function:
$$I(r)=n\{e^{0.5(r/4.5001)^2}+4.376e^{0.5(r/2.8644)^2}+0.9346r^{1.6569}\}$$
We follow the same procedure for the profile fitting of each quasar. First we fit the profile by the empirically-derived HRI PSF alone, allowing its normalization to vary freely. We then fit this PSF (with normalization still free to vary) in combination with each of two models chosen to represent any extended emission. The first model is a broken power-law, with slope $`r^1`$ for radii $`r<R`$ and $`r^{2.1}`$ for $`rR`$, where the break $`R`$ and the absolute normalization of the extended component are free to vary. This is an approximation to the X-ray surface brightness profile of the gas in a typical cluster of galaxies containing a cooling flow, and $`R`$ then corresponds to the cooling radius (eg Crawford & Fabian 1995b). The second model employed is a projected King law, with index fixed at -1.5, and the core radius $`R`$ and normalization left as free parameters. Given the errors inherent in whether such simple models truly characterize the extended emission, we do not convolve the extended emission models with the PSF. The relative normalization between the PSF and extended components are not always very well determined, so we also derive what should be regarded as a lower limit to the presence of any extended component by assuming the nuclear emission accounts for all the light in the X-ray core. We fit the PSF to the quasar radial profile within the inner 1–5 arcsec and then subtract this model and fit the residuals by each of the cluster models. We execute these 5 model fits to the profiles out to a radius of 50 arcsec (11 data points), yielding 10, 8, and 9 degrees of freedom for the psf only, psf+extended component models, and the fit of the extended component model to the residual after subtraction of the normalised HRI PSF. We then repeat the fits to the profile out to a radius of 100 arcsec (15 data points), yielding 14, 12 and 13 degrees of freedom to the fits as above. The fitting analysis is carried out first for each quasar image in the absence of any wobble correction, and then for the images corrected using different phase intervals.
The detailed results are summarized in Table 2, where the reduced-$`\chi ^2`$ is given for each fit. We tabulate the best-fit parameters of the profile fits out to a radius of 50 arcsec and then 100 arcsec in turn: $`R`$ (in arcsec) representing either the break in the broken power-law model, or the core radius in the King law model; the integrated luminosity from the extended component as a percentage of the total luminosity of the X-ray source; the X-ray luminosities (in the observed 0.1-2 keV energy band) of the quasar component ($`L_X^{QSO}`$) and that of the cluster component ($`L_X^{cl}`$) assuming a power-law of photon index 2 and thermal bremsstrahlung emission at a temperature of $`kT=4\mathrm{keV}`$ respectively. (At the redshift of our quasars this observed band carries about half of the bolometric luminosity for the thermal spectrum.) The errors are derived from propagating the $`\mathrm{\Delta }\chi ^2=1`$ confidence limits of the fit parameters. Errors are not shown when the fit was insufficiently robust to extract errors on all parameters of interest. Table 2, however, demonstrates the full range of values obtained from the ten model fits employed for each of the phase intervals and allows one to assess the variation of each parameter from the systematic uncertainties of PSF normalization and extended component model employed. A comparison of some of the better fits to the radial profile of each quasar (those shown in bold font in Table 2) are displayed in Figure 2. These plots clearly show that there are significant differences between the PSF-only fit to the profile, and the fits that include a model for extended emission. In all this analysis we necessarily assume that any extended component is both centred on the quasar (in no case do we see any evidence for a secondary off-centre peak), and derive its properties such as scale and luminosity assuming that it is at the redshift of the quasar.
The present data cannot rule out a contribution to the extended component of X-ray emission from the active nuclei of close companion galaxies to each of the quasars. Such emission would, of course, provide further support for a clustered environment. The probability of getting an unassociated X-ray source within an aperture of 1 square arcminute centred on a quasar is less than $`10^3`$, at the flux level of the extended emission. Thus there is little chance of the extended emission component being due to contamination by fore- or back-ground sources.
## 3 Results for individual quasars
### 3.1 3C48
Given its proximity to a very luminous source of photoionizaton, the low ionization state observed in the spatially extended oxygen line emission around this 3C48 led Fabian et al (1987) to deduce a high density environment around this quasar. The inferred gas pressure of 3-8$`\times 10^5`$$`\mathrm{cm}^3\mathrm{K}`$ within 30 kpc of the quasar core is consistent with confinement of the extended emission-line region by an intracluster medium There is, however, no strong evidence for a rich cluster of galaxies from optical images (Yee, Green & Stockman 1986; Yates etal 1989).
The fit to the radial profile is substantially improved by the addition of an extended component, the best fits being obtained in all cases when this is represented by a King law. The extended component requires a very consistent value for the core radius $`R`$ of around 5-6 arcsec in all fits (1 arcsec corresponds to 6.2 kpc at the redshift of the quasar<sup>1</sup><sup>1</sup>1We assume a value for the Hubble constant of $`H_0=50`$$`\mathrm{km}\mathrm{s}^1\mathrm{Mpc}^1`$ and a cosmological deceleration parameter of $`q_0=0.5`$ throughout this paper.), and accounts for 10-16 per cent of the total X-ray source. The full variation of its luminosity is $`510\times 10^{44}`$$`\mathrm{erg}\mathrm{s}^1`$, with most of the values derived being to the lower end of this range.
### 3.2 3C215
This quasar lies in a densely clustered environment (Ellingson et al. 1991; Hintzen 1984), and the radio source has a very complex structure suggestive of deflection and distortion of the radio jet to the south-east by some external medium. The two sides of the radio source show asymmetric Faraday depolarization, which can be interpreted as due to differing lines of sight through a depolarizing cluster medium (Garrington, Conway & Leahy 1991). Crawford & Fabian (1989) inferred a high gas pressure of over $`3\times 10^5`$$`\mathrm{cm}^3\mathrm{K}`$ from the ionization state of the extended line emission within 30 kpc of the quasar nucleus.
We have extracted radial profiles from three observations of 3C215. Smearing of the image seems to have badly affected observation number 800753n00, as $`R`$ decreases substantially after correction of the image. In all cases, the reduced $`\chi ^2`$ of the fit improves from the addition of an extended component to the HRI PSF model, although there is little preference shown between the King and broken power-law models. In the fits allowing free normalization of the PSF component the $`R`$ derived is in the range 4-9 arcsec (where 1 arcsec corresponds to 6.5 kpc at the redshift of the quasar), with the broken power-law model always yielding the larger values of $`R`$. The luminosity of the extended component ranges over 1.6-6.3$`\times 10^{44}`$$`\mathrm{erg}\mathrm{s}^1`$, and account for 11-40 per cent of the total X-ray emission from this source; the higher values are derived from the King law models.
### 3.3 3C254
3C254 is a quasar with a very asymmetric radio source, and was discovered by us (Forbes et al 1990) to lie in a spectacular emission-line region extending out to radii of 80 kpc. The optical nebula is again at low ionization, with an inferred pressure of $`>10^6`$$`\mathrm{cm}^3\mathrm{K}`$ within 30kpc (Forbes et al 1990; Crawford & Vanderriest 1997). The kinematics and distribution of the line-emitting gas in combination with the radio morphology strongly suggest that the radio plasma to the east of the quasar is interacting with a dense clumpy environment (Bremer 1997; Crawford & Vanderriest 1997). The radio source itself also shows asymmetric depolarization (Liu & Pooley 1991). Optical continuum images show an overdensity of faint objects around the quasar consistent with a location in a compact cluster or group (Bremer 1997). The HRI image of 3C254 shows a detached secondary source of X-ray emission approximately 35 arcsec east of the quasar, but at a very low fraction of the total quasar luminosity. There is no optical counterpart to this source on the Space Telescope Science Institute Digitized Sky Survey.
The fitting of the profile is improved by the addition of an extended component to the HRI PSF model, although it cannot discriminate between a King or broken power-law model. The extended emission contains 12-19 per cent of the total X-ray luminosity of the source. The characteristic scale length $`R`$ of the extended emission component varies over 9-15 arcsec (1 arcsec corresponds to 8.1 kpc at the redshift of the quasar), with a luminosity of 5-9$`\times 10^{44}`$$`\mathrm{erg}\mathrm{s}^1`$.
### 3.4 3C273
3C273 is famous for its jet, which can be seen in the outermost contours of our corrected images. The jet is known to emit X-rays at around 16 arcsec (nearer the core than the radio and optical features of the jet), but the emission from the jet contributes less than 0.5 per cent of the total X-ray luminosity of the source associated with the quasar (Harris & Stern 1987). From optical images, 3C273 may be a member of a poor cluster of galaxies (Stockton 1980).
All the wobble-corrected profiles of 3C273 require the addition of an extended component (preferably a King law). Each of the models for the extended emission yields slightly (and consistently) different results: the broken power-law fits tend to have $`R`$ of 8.7 arcsec (where 1 arcsec corresponds to 3.6 kpc at the redshift of the quasar) and contain 5 per cent of the total X-ray luminosity, at 7$`\times 10^{44}`$$`\mathrm{erg}\mathrm{s}^1`$. The King law fits show a greater variation on the parameters, but have $`R`$ of only $``$4 arcsec, and $`11`$ per cent of the total luminosity at 1.3-2.0$`\times 10^{45}`$$`\mathrm{erg}\mathrm{s}^1`$. The extended emission we find in the environs of this quasar is sufficiently luminous that it cannot easily be ascribed to the jet.
### 3.5 3C275.1
3C275.1 was the first quasar discovered from optical galaxy counts to be located at the centre of a rich cluster of galaxies (Hintzen et al 1981; Hintzen 1984; Ellingson et al 1991). The radio source is only slightly bent, but the two sides display an asymmetry in the Faraday depolarization (Garrington et al 1991). The quasar is embedded in a host galaxy with a continuum spatial profile and absolute magnitude typical of a bright cluster cD; this is in turn surrounded by a large (100 kpc) optical emission-line nebula (Hintzen & Romanishin 1986; Hintzen & Stocke 1986). Crawford & Fabian (1989) deduce a pressure within this gas of $`>3\times 10^5`$$`\mathrm{cm}^3\mathrm{K}`$ at radii of $`<`$20 kpc and thus also the presence of an intracluster medium (Crawford & Fabian 1989).
The fit to the X-ray radial profile improves slightly with the addition of an extended component to the HRI PSF. The models all imply a surprisingly broad core, so much so that fitting a second component to data only within a 50 arcsec radius is not robust. The King and broken power-law models yield very similar results, with $`R`$36 arcsec (where 1 arcsec corresponds to 7.4 kpc at the redshift of the quasar), and 24 per cent of the total X-ray luminosity of the source at $`L_X3\times 10^{44}`$$`\mathrm{erg}\mathrm{s}^1`$.
### 3.6 3C281
3C281 is known to lie in a rich cluster (Yee & Green 1987) and Bremer et al. (1992) infer a pressure exceeding 2$`\times 10^6`$$`\mathrm{cm}^3\mathrm{K}`$ in the extended emission-line gas within a radius of 20kpc. After the wobble-correction has been applied, the X-ray source associated with the quasar shows a distinct elongation to either side of the core along a position angle of 45west of north. The direction of this elongation is, however, at odds with that of the radio source which has an axis 10east of north.
The profiles extracted from both observations of 3C281 both show an improved fit from the addition of an extended component (preferably a King law). The fractional luminosity of this component is high, ranging over 42-67 per cent at 1.1-1.7$`\times 10^{45}`$$`\mathrm{erg}\mathrm{s}^1`$(the higher values obtained with the King model fits). The characteristic radius of this component is 5-6 arcsec (where 1 arcsec corresponds to 7.6 kpc at the redshift of the quasar).
### 3.7 3C334
This quasar lies in a clustered environment (Hintzen 1984) and has a pressure within the extended emission-line region of over $`6\times 10^5`$$`\mathrm{cm}^3\mathrm{K}`$ at 30 kpc from the quasar core (Crawford & Fabian (1989). The quasar again shows a depolarization asymmetry (Garrington et al 1991), and narrow-band imaging by Hes (1995) suggests that the \[OII\] line emission is extended along the same position angle of $`150`$as the strong radio jet to the south-east of the quasar core.
The fits to the wobble-corrected profiles show an improvement to the fit with the addition of an extended component, preferably a King model. The lengthscale $`R`$ of this component lies in the range 5-10 arcsec (where 1 arcsec corresponds to 7.4 kpc at the redshift of the quasar), and contains 8-22 per cent of the total X-ray luminosity at 3-8$`\times 10^{44}`$$`\mathrm{erg}\mathrm{s}^1`$.
### 3.8 H1821+643
H1821+643 is radio-quiet quasar which is luminous in the infrared. The X-ray emission was found to have a significant extended component by Hall et al (1997). It is included here as a comparison object.
The fits improve dramatically with the inclusion of an extended component. With the 50 arcsec apertures, the King model is often preferred, whereas the 100 arcsec apertures show a marked improvement for the broken power-law model. The characteristic radius of this extended component, $`R`$, shows a wide range of 29-81 arcsec (where 1 arcsec corresponds to 5.5 kpc at the redshift of the quasar), but clear separation according to both extended model and outer profile radius employed. The extended component contains from 10–19 per cent of the X-ray luminosity, at 8–20$`\times 10^{44}`$$`\mathrm{erg}\mathrm{s}^1`$.
## 4 Summary of results
We tabulate average properties of the extended emission component from the profile fits to each quasar in Table 3, where these averages are derived only from the fits where the normalizations of the PSF and model for the extended emission are both allowed to vary (labelled as PSF+broken power-law and PSF+King law in Table 2). The values are averaged from the fits only to the wobble-corrected images (and from each phase interval employed). As is clear from both Tables 2 and 3, where the scatter in properties is sufficiently low for comparison to be made (3C48, 3C215, 3C273 and 3C281), systematic differences can be seen between parameters derived using the King and broken power-law fits. The King law always gives a higher X-ray luminosity (and thus higher overall percentage of the total luminosity), but a smaller characteristic radius $`R`$ than the broken power-law model.
The average bolometric luminosities (from the profile fit out to 100 arcsec) span the range of 6-43$`\times 10^{44}`$$`\mathrm{erg}\mathrm{s}^1`$, with values of 13 and 18$`\times 10^{44}`$$`\mathrm{erg}\mathrm{s}^1`$typical for the broken power-law and King models respectively. The King models give values of $`R`$ in the range 33-38 kpc, with the exception of 14 kpc for the fits to 3C273; for the broken power-law model the values are 31-49 kpc. The three quasars, 3C334, 3C254 and 3C275.1, for which we do not differentiate between parameters derived from each model have an average $`R`$ of 49, 94 and 270 kpc, respectively.
Where a preference between the two models for the extended emission can be seen, it is nearly always for the King law over the broken power-law model (in 3C48, 3C273, 3C281 and 3C334). An exception is the 100 arcsec profile fits to 1821+643, which prefers the broken power-law fits. The continuing increase in luminosity of the extended X-ray emission between the 50 and 100 arcsec profile fits also shows that this component is truly extended over cluster-wide scales of $`400800`$ kpc radius.
We note that the discrepancy between the observed, corrected profile and the PSF is least for 3C273 and 3C48, which are the nearest 3CR quasars in our sample. If there are systematic errors associated with the wobble correction that we are unaware of, then these objects will be the most seriously affected. It is important that the extended emission from these objects be confirmed with Chandra.
## 5 Discussion
We plot the average bolometric luminosity of the extended cluster component (from Table 3) against quasar redshift in Figure 3. For comparison we plot the bolometric luminosity of the X-ray source associated with the distant radio galaxies 3C277.2, 3C294, 3C324, 3C356, 3C368 (from Crawford & Fabian 1996b) and 1138-262 (Carilli et al 1998), and the clusters surrounding the two nearby FR II radio galaxies Cygnus A (Ueno et al 1994) and 3C295 (Henry & Henriksen 1986). The observed countrates of the distant radio galaxies have been converted to luminosities assuming the same 4 keV thermal bremsstrahlung model used to obtain luminosities for the quasar extended emission.
The luminosities we have derived for the environment of our quasars are brighter than the upper limits of 1.6 – 3.5$`\times 10^{44}`$$`\mathrm{erg}\mathrm{s}^1`$(rest-frame 0.1-2.4 keV) to any cluster emission surrounding three radio-loud quasars in Hall et al (1995, 1997). We note, however, that those upper limits have been obtained from images not corrected for satellite wobble. They also assume therefore that the quasar light follows the standard HRI PSF derived by David et al (1995) and accounts for all the light in the innermost bin. Our quasar host clusters are consistent with the luminosity of 3.7$`\times 10^{45}`$$`\mathrm{erg}\mathrm{s}^1`$detected by Hall et al (1997) for the environment of the radio-intermediate quasar H1821+643 at a redshift $`z`$=0.297.
The inferred bolometric luminosities of the extended components we have found here are completely reasonable for moderately rich clusters of galaxies at low redshift. They are comparable to the luminosities of the clusters associated with the powerful radio galaxies Cygnus A and 3C295. They are however (Fig. 1) more luminous than the extended X-ray emission detected around more distant 3CR radio galaxies above redshift one. Whether this indicates evolution, a problem for radio galaxy/quasar unification, or is a result of small number statistics must await the compilation of a complete sample, which are study is not.
All extended models have a central cooling time considerably shorter than a Hubble time. We therefore explore the properties of the implied cooling flows occurring around these quasars by deriving some approximate parameters from the broken power-law fits to the profiles. We attribute all the X-ray luminosity of the extended component within radius $`R`$ to thermal bremsstrahlung from gas with electron density $`n`$ (where $`nr^1`$) at a temperature of 4 keV. The cooling time of the gas at $`R`$ (except in the case of 3C275.1) is then between about 1–3 billion yr. We then estimate the cooling flow radius $`r_{\mathrm{CF}}`$ at which the cooling time is $`10^{10}`$ yr and obtain a rough indication of the mass deposition rate within that radius from the ratio of the mass of gas within $`r_{\mathrm{CF}}`$ to $`10^{10}`$ yr. The derived values are shown in Table 4. Note they are of course subject to not only the appropriateness of the fixed slopes chosen for our original broken power-law model, but also to the true gravitational potential of any cluster, and the amount of gravitational work done on the cooling gas. The values should be regarded as uncertain by a least a factor of 2. They may be underestimated by a factor of at least 2 if the gas temperatures are significantly higher than the 4 keV assumed and there is internal absorption such as is common in low redshift cooling flows.
Note that the radius of the surface brightness break which we infer is in the range 40–90 kpc, and is similar to the break radius in the profile of the cluster around IRAS 09104 ($`60`$ kpc, Crawford & Fabian 1995b). This is likely to be the radius of the core of the gravitational potential of the cluster; the $`r^1`$ profile then occurs within there since that gas is cooling at approximately constant pressure. Such small gravitational core radii are characteristic of relaxed lensing cluster cores such as are associated with massive cooling flows (Allen 1998). The large break radii found for 3C275.1 and H1821+643 do not agree with this picture and require more detailed images.
We can also use our cooling flow parameters to derive a gas pressure $`P`$ at a radius of 30 kpc (see Table 4) for comparison to the pressures derived from the completely independent method using the ionization state of the extended optical emission lines (Crawford & Fabian 1989; Bremer et al 1992; Crawford & Vanderreist 1997). The pressures derived from the optical nebulosities mostly underestimate those derived from the X-ray profile fits (Figure 4) by a factor of up to 10. Given that the gas pressures are mostly derived from the optical nebulosity using a conservative underestimate to the crucial but unknown UV and soft X-ray band of the ionizing nuclear spectrum, this discrepancy is not surprising. Support for this interpretation of the disparity in derived pressures is found in Crawford et al (1991) where a better knowledge of the ionizing continuum of the nucleus of 3C263 was found to increase the optically-derived pressure by up to an order of magnitude. In addition, we note that 3C254, the quasar with the best agreement between the two pressure values is the only one where UV HST data has been used to constrain the shape of the ionizing continuum (Crawford & Vanderriest 1997).
We note that the derived cluster luminosity for 3C273 is high and implies the presence of a rich cluster which is not seen at other wavelengths. At its relatively low redshift of $`z=0.16`$, such a cluster should be obvious in the optical band. As mentioned already, it has the profile most susceptible to systematic errors, and the absence of an optical cluster may argue for the existence of such errors. If the PSF is then uncertain by a relative amount equal to the observed 3C273 profile and our empirical PSF, then it will not change greatly our results on the other quasars except perhaps for 3C48.
Inverse-Compton scattering of quasar radiation could still contribute to our extended component of X-ray emission. If a significant process, the X-ray source would appear asymmetric and lop-sided, as (back) scattering by the electrons in the more distant radio lobe should be stronger than in the nearer lobe. The quality of our current data is insufficient to show any significant asymmetry, and so we are unable to assess the contribution from this process. Observations of these sources with Chandra and XMM should clarify its relative importance to the total emission.
## 6 Conclusions
The seven powerful radio-loud quasars studied here appear to be surrounded by luminous extended X-ray emission. The spatial properties of the emission are consistent with an origin in thermal emission from an intracluster medium. The radiative cooling time of the gas within $`50`$ kpc of the quasars is only a few billion years or less, indicating the presence of strong cooling flows of hundreds of $`\mathrm{M}_{}\mathrm{yr}^1`$. The high pressure of that gas is sufficient to support the extended optical nebulosities seen around many of these quasars and may play a rôle in shaping the properties of the radio sources, such as structure, depolarization and possibly even fuelling and evolution (Fabian & Crawford 1990). Such a rôle is not clear from a comparison of radio properties with the neighbouring optical galaxy density (Rector et al 1995).
A small number of luminous clusters without central AGN have so far been found beyond redshift 0.5, and the discovery of a cluster at $`z0.8`$ with luminosity $`10^{45}`$$`\mathrm{erg}\mathrm{s}^1`$(Donahue et al 1998) provides strong evidence for a low density universe. We have shown that the study of the environment of powerful radio-loud quasars is a promising way of extending the discovery of similarly luminous clusters both in numbers and to higher redshifts.
## 7 Acknowledgements
We thank Steve Allen for advice on the profiles of massive cooling flows. CSC and ACF thank the Royal Society for financial support. This work has been supported in part by the DLR (format DARA GmbH) under grant 50 OR 9403 5 (GH and IL). This research has made use of the NASA/IPAC Extragalactic Database (NED) and the Leicester Database and Archive Service (LEDAS).
|
no-problem/9904/astro-ph9904188.html
|
ar5iv
|
text
|
# Future Cosmic Microwave Background Experiments
## 1. Prologue
The satellite missions MAP and Planck dominate any view of future measurements of the anisotropy of the Cosmic Microwave Background. We will attempt to look beyond and around those two experiments and see what sorts of physical questions other future projects might address.
The reader has several advantages over the authors which we will not try to counter. First, many of the experiments which are in the near term future for us will be in the present or past for the reader, so we do not focus on evaluating detailed anticipated technical capabilities for a short list of such experiments. Readers who wish to pursue that approach might start at http://www.astro.ubc.ca/people/scott/cmb.html, or other similar web-pages for up to date information and links.
Second, readers of the rest this volume will be in a better position than we are to make informed judgements about the ideal strategies for measuring, avoiding or understanding foreground sources. Therefore, even though we think that this aspect of anisotropy will be an increasingly important and sophisticated part of the field, we have not put much emphasis on it here. As a crude aid to understanding how well future experiments are equipped to cope with foreground sources we have included a column giving the number of independent frequency channels for each experiment listed in Table 1.
A view of the present situation, indicated in Figure 1 (see Smoot & Scott 1998 for more details), sets the context for our view of the future. Even at a casual and sceptical glance these experiments seem to be converging on a power spectrum which has a peak in it. This is confirmed by careful quantitative analysis of combined data sets (Bond, Jaffe & Knox 1998). Collectively these CMB measurements already tell us a number of fundamental things about the sort of Universe that we live in (see Lawrence, Scott & White 1999). The prospects for future measurements look very bright indeed. Announcements of the value of $`\mathrm{\Omega }_0`$, for example, are likely to (continue to) come from experiments carried out from the best terrestrial sites or suspended from stratospheric balloons, during the next few years. However, the full belief of the community in any detailed cosmological conclusions will and should await the satellite results.
Despite the steadily improving quality of experiments, we believe that none of the more recent experiments in Figure 1 would have stood as a convincing discovery of primordial anisotropy had it not been for COBE (this remark certainly applies to our own experiment, BAM, as much as to any other experiment). What was critical in the discovery was the understanding of the roles of galactic contributions and systematic errors, provided by COBE’s all-sky coverage and comparatively stable operating environment. It was also crucial for the discovery process that the DMR on COBE and the FIRS balloon program showed a consistent fluctuation amplitude and, later analysis showed, correlated structures observed at very different wavelengths. Many experimenters had reassured themselves by making plots showing the similarity of the FIRS and DMR power spectra, before the end of the day on which the DMR results were announced.
There is a lesson arising from the history of the measurement of the intensity spectrum of the CMB which may be useful here. There were plenty of experiments prior to 1990 which appeared to have sufficient sensitivity to perform useful measurements, many of these with no obvious source of systematic error.<sup>1</sup><sup>1</sup>1We will decline to provide examples here, reminded as we are of Winston Churchill’s failed attempt to maintain parliamentary courtesy when he said that half of the members present were not asses. The successful 1990 experiments (Gush et al. 1990, Mather et al. 1990) were performed out of the atmosphere, they were differential and they were carried out with a fanatical attention to avoiding systematic errors as the primary design guideline. The results were clear and reliable enough to render moot any lingering debates about inconsistencies between previous experiments. One should not be surprised to see a very similar scenario play itself out in the near term anisotropy measurements.
## 2. Near Term Future Experiments
Table 1 lists the properties of some future experiments. The list is meant to be illustrative of current planning; experiments are included which are past the proposal stage and for which no results are yet available. Some of the listed experiments already have data. Of course many experiments which have already produced some results and are therefore not on this list will also produce future results. All of the listed experiments involve dedicated, custom-built instrumentation. The control of systematic errors which this allows puts these experiments well ahead of attempts to use existing general purpose facilities.
Sufficient sensitivity is achievable, sometimes through great technical effort. The various detection technologies available each impose constraints on experimental design and come with their own set of sources of potential systematic errors. Any serious discussion of specific systematic errors is beyond the scope of this article but we include some naive examples to illustrate the problem. Either a $`100`$mK change in the temperature of an ideal aluminum mirror or a $`200`$mK change in the atmosphere above a stratospheric balloon causes a radiative signal 25 times larger per pixel than the MAP systematic error budget!
### 2.1. Systematic Errors
The careful CMB experimenter is not paranoid, but knows that the Universe is in fact trying to ruin the experiment. The standard answer to the question of what level of systematic error is tolerable is that there is no systematic way to handle systematic errors and, therefore, that any level of systematic error is a concern. We will ignore this good advice for a moment and try to estimate an answer.
If the goal of an experiment is to get a rough estimate of the power spectrum of the sky, a systematic error amounting to $`10\%`$ of the signal amplitude contributes about $`1\%`$ to the power spectrum. Even if the signals are correlated in some surprising way and this estimate is wrong by a factor of a few, the effect is not likely to mask the presence of the main acoustic peak, for example. This fact is what has allowed us to get so far without a better understanding of diffuse foreground sources.
On the other hand, there are important questions whose answer requires correlating many pixels in a map together in order to pick out a fairly weak efect. Measuring amplitudes of non-Gaussian statistics of a map or searching for intensity-polarization correlations are examples. In these cases the requirement for what level of signal systematic errors can contribute to a map becomes very stringent. The amplitude of systematic errors should be below the experiment’s single pixel variance divided by the square root of the number of pixels to be averaged. As a numerical example, in an experiment with $`0.13^{}`$ pixels and $`30\mu `$K variance averaging 1/10 of the sky, one needs to know that systematic errors are less than $`0.06\mu `$K rms for the average value not to be tainted. This is 50 times better than any experiment we have heard of. The lesson is: to produce maps of the CMB which merit careful scrutiny, avoid systematic errors like the plague.<sup>2</sup><sup>2</sup>2Winston Churchill also said ‘One ought never to turn one’s back on a threatened danger and try to run away from it. If you do that, you will double the danger. But if you meet it promptly and without flinching, you will reduce the danger by half’.
### 2.2. Detection Techniques
Detectors fall into two broad categories: coherent detectors, in which the radiative electric field, including its phase, is amplified before detection; and incoherent detectors, which measure total radiative power within some frequency band.
There are two types of very low noise coherent amplifier: InP high electron-mobility transistors (HEMTs); and superconductor-insulator-superconductor (SIS) mixers. HEMTs can be operated at temperatures as warm as room temperature. The noise performance gets better as they are cooled, down to $`4`$K, although amplifiers exhibit gain fluctuations at low temperature. Recently HEMT amplifiers have been made to work at frequencies well above $`100`$GHz – noise performance is better at lower frequencies. SIS mixers are typically quieter than HEMTs and can operate at frequencies as high as $`1`$THz. However they must be cooled to $`4`$K to operate. Either of these coherent amplifiers can be used in a single telescope where the signal is amplified and detected, or as part of an interferometer in which case amplified signals from several telescopes are each multiplied with a local oscillator signal yielding lower frequency outputs which are then correlated to produce interference fringes.
The advantages of coherent detectors are that they are fast, stable, not sensitive to microphonic pick-up and involve simple cryogenics. HEMTs also have the important practical advantage that many aspects of detector performance can be verified at room temperature, which greatly speeds up new instrument development. The disadvantage is that they are not as sensitive to broad band signals as incoherent detectors are.
Incoherent detectors, in this case bolometers, can be an order of magnitude more sensitive than HEMT and SIS systems. They can be made to operate with background limited performance (BLIP), where fundamental thermodynamic fluctuations in the incident radiation field dominate over detector noise. In addition they can be made to be sensitive to a broad range of wavelengths. However, physical device size scales with wavelength and so it is easier to make small bolometers sensitive. Typically bolometers are designed for frequencies above $`50`$GHz. Bolometers are often susceptible to microphonic and radio-frequency pick-up. They are non-linear and therefore they must be characterized in their experimental operating condition, which can be very difficult for balloon and satellite experiments. They need cumbersome cryogenics to reach their operating temperatures of $`0.3`$K or colder. However, their extraordinary sensitivity and broad frequency coverage often outweigh these disadvantages. Table 2 lists some detector properties for the experiments in Table 1.
#### Interferometers
The idea of building a dedicated interferometer to study anisotropy of the CMB is not new, but improvements in detectors, and especially in broad bandwidth correlators has made this a very promising option, which is being actively pursued by several groups.
Interferometers do a good job of rejecting the effects of atmospheric variations compared to beam-chopped single telescope instruments. Measurements take place essentially instantaneously, on time scales associated with the interference bandwidth, and on these time scales the atmosphere does not vary. Also, interferometers measure at slightly higher angular resolution than a single telescope of the same overall size, and in any case can easily be built for higher resolution than the currently planned space missions. This advantage will be important in exploring the expected Sunyaev-Zel’dovich forest, especially if they can also be made to work above $`200`$GHz.
Unlike the case for conventional radio interferometers, the individual telescopes here are crowded quite close together to keep angular resolution modest. Often, all the telescopes are mounted on a single pointed platform, which eliminates the need for signal delays before the correlators. See White et al. (1997) for an analysis of the performance of these interferometers for measuring anisotropies.
#### Satellites
Assuming that neither suffers any serious mishap, MAP and Planck will produce definitive measurements of the primary anisotropy of the CMB, at a reliability level which the other experiments can not attain. The reliability arises form the long observation period, complete sky coverage and, primarily, the extraordinarily good observing environment at L2. Even during the 90-odd day period during which it makes its way past the moon and out to L2 to start the nominal observation program, MAP will be in a much more thermally and radiatively stable observing environment than any previous CMB experiment.
## 3. After MAP and Planck
What will the important experimental questions be after MAP and Planck succeed? Clearly, measurements of the polarization of the CMB, which are explored elsewhere in this volume, will be very exciting. We also expect that studying diffuse foreground emission will become very exciting and active, as our ability to measure and identify these sources of emission develops. However, that topic is covered in the whole rest of this book so we need not consider it further here! For the remainder of this article we will discuss various ideas for what might be conceivable in the more long term future.<sup>3</sup><sup>3</sup>3Ignoring the sound advice of Winston Churchill, who said ‘It is a mistake to try to look too far ahead. The chain of destiny can only be grasped one link at a time.’
### 3.1. Anisotropy
#### Statistics
Can phases contain information which is not more easily seen in the power spectrum? In principle, of course the answer is yes. But in practice, it seems clear that the smart money has to be on the negative answer. So although it would always be foolish to neglect to search for other signals, we expect that the vast majority of the primary anisotropy information will be contained in the power spectrum. Partly this is because the signals seem likely to be close to Gaussian, but also because the power spectrum (or the variance as a function of scale) is such a robust quantity – specific patterns on the sky may require lots of phase-correlation to produce them, but much of that simply specifies the specific realization, rather than containing information about the underlying model. The supremacy of the power spectrum will certainly cease to be true for foreground signals, or indeed for a range of astrophysical processing effects that come in at smaller angular scales.
One could imagine mounting a specific search for, e.g. point or line sources on the sky, as specific examples of non-Gaussian signals. One question to ask, then, is what sort of strategy one would design to carry this out most efficiently (and convincingly). We find it hard to see how to avoid the conclusion that you would end up making a map, perhaps deeper and with higher resolution than otherwise, but a map nevertheless. Hence we suspect that the search for non-Gaussian signals is unlikely to be a strong driver for the design of future experiments, even if it plays a stronger role in the data analysis.
A great deal of effort has been going into the study of non-Gaussian signals, e.g. using Minkowski functionals, wavelets, etc. Given how many such tests have already been applied to COBE, we imagine that every reasonable statistic will be measured for all future large data-sets. In particular we foresee an increased interest in the investigation of non-Gaussian statistics for various sorts of foreground signal.
#### Angular scales
Ignoring foregrounds, how far out in $`\mathrm{}`$ is far enough? Planck seems sufficient for the primary signal. But that may change, depending on what we learn about foregrounds and the secondary signals, caused by various astrophysical effects, which conceptually lie in the ‘grey-area’ between background and foregrounds. There seems to be a growing interest in these astrophysical signals at small scales, and we see no reason for that to change. It may be that the smallest angular scales are ultimately best probed with interferometers. We expect there to be secondary signal information down to the angular scales of distant galaxies, i.e. $`\mathrm{}\mathrm{few}\times 10^4`$.
#### A CMB Deep Field
What might we learn from a CMB deep field? Of course, irrespective of the answer to that question, it will be done anyway! Non-Gaussian signals from higher-order effects at small-scales would certainly show up in such a map. On scales where there has been significant growth of structure, and certainly on non-linear scales, we would expect there to be significant non-Gaussianity. There seems little doubt that at some point, when the instrumentation has matured, it will be worthwhile to carry out such a CMB Deep Field. Exactly how non-Gaussian (or in other ways surprising) the small-scale signals turn out to be will determine how far beyond ‘cosmic variance’ it is worth integrating.
#### Sunyaev-Zel’dovich effects
We are sure that, motivated by the impressive results of today’s experimenters, investigation of S-Z effects will continue to grow as a sub-field. Particularly exciting is the idea of ‘blank sky’ searches for the ‘S-Z Forest’, or ionized gas tracing out the Cosmic Web.
The thermal Sunyaev-Zel’dovich effect, or inverse Compton scattering of the CMB photons through hot gas, gives a temperature fluctuation of roughly $`(kT_\mathrm{e}/m_\mathrm{e}c^2)\tau `$, where $`T`$ and $`m_\mathrm{e}`$ are the temperature and mass of the electron, respectively, and $`\tau `$ is the optical depth through the ionized gas. There is also a spectral shape, distinct from the CMB blackbody, of a well-known form (see e.g. Sunyaev & Zel’dovich 1980) Detailed studies of the thermal S-Z effect for particular clusters will provide constraints on the morphology, clumping, thermal state of the gas and projected mass, amongst other things. The power spectrum of these fluctuations peaks at $`\mathrm{}2000`$ typically, with $`Q_{\mathrm{flat}}`$ amplitude of a few $`\mu `$K, although with great variation between models. Detailed investigation of this power spectrum might further distinguish between cosmological models, and between ideas for cluster formation. The power spectrum for the kinematic effect, and for related effects (due to variations in potential, for example) are generally much smaller.
Several of these ‘higher-order’ Sunyaev-Zel’dovich type effects are potentially measurable for individual clusters, and will doubtless be attempted in the future (see the review by Birkinshaw 1998). Certainly the kinematic effect (which depends on the line-of-sight velocity and is $`(v_{}/c)\tau `$) can be measured for some clusters. However, this effect has the same spectrum as a CMB fluctuation, and so the small-angle CMB anisotropies act as a source of ‘noise’, making is difficult to measure the velocities to better than a few hundred $`\mathrm{km}\mathrm{s}^1`$. One further effect uses the polarization in the CMB scattered by the kinematic S-Z effect, which depends on the cluster’s transverse velocity (actually $`(v^2/c^2)\tau `$). In principle, together with the kinematic S-Z effect itself, this gives a means of estimating the full 3 dimensional velocity of clusters. Although difficult to measure, this polarization signal has a frequency dependence which may help to disentangle it from other effects (Audit & Simmons 1999). There are other spectral signals expected from non-thermal electron populations, for example in the lobes of radio sources. However, the utilisation of such measurements to study the lobe properties will require extremely high angular resolution.
#### Other secondary effects
There are several other known secondary effects (see, e.g. Hu et al. 1995, and other contributions to this volume), and surely many other unknown ones!
One effect which has been studied in some detail is a second-order coupling between density and velocity, usually referred to as the Vishniac effect. In a sense this can be thought of as specific case of the kinematic Sunyaev-Zel’dovich effect. For Cold Dark Matter type models the signal is typically $`1\mu `$K and peaking at $`\mathrm{}\mathrm{few}\times 10^3`$ (e.g. Hu & White 1996, Jaffe & Kamionkowski 1998). Although certainly difficult to measure, it is nevertheless feasible, and worth pursuing, since measurement can give direct information about reionization. Additional structure ar small scales (e.g. from an isocurvature mode) could increase the signal. In addition there will be polarization effects, although these are likely to be really small.
Patchy reionization (discussed elsewhere in this volume) is just another S-Z effect, and tends to be dominated by the kinematic source from moving bubbles of gas as the Universe undergoes reionization. The amplitude of this signal seems likely to be smaller than for the Vishniac effect, although it is as yet unclear what the predictions will be for realistic models which include inhomogeneous reionization (with radiative transfer through voids etc.), distributions of sources, and other complications. Again there may be polarization effects, and correlations with other signals, which, in principle, could be used to pull the signal out. In addition there may also be a measurable S-Z signal from the Ly$`\alpha `$ forest, on scales well below an arcminute, and with amplitude perhaps as high as a few $`\mu `$K (Loeb 1996).
Rees-Sciama, or varying potential fluctuations tend to be rather small in amplitude ($`<10^6`$ in fractional temperature change), but not negligibly so. Here again there are a number of effects, in particular those caused by time-varying potentials in the light-crossing time, and those caused by potentials moving across the line of sight (e.g. Tuluie, Laguna & Anninos 1996). These will have CMB-like spectra, and the signal will be dominated by non-linear structures (meaning that the statistics will be highly non-Gaussian). The effects may peak at relatively small angles $`\mathrm{}\mathrm{few}\times 100`$, but there they will be well below the primary signal, and hard to disentangle. So the best prospects for detection may be at smaller scales, where the primary power spectrum is falling off. Detection may also be easier using correlations with other signals. And certainly such signals are unlikely to be Gaussian, and so may be teased out of the data by looking at their statistics (e.g. Spergel & Goldberg 1999).
Gravitational lensing affects the CMB power spectrum by smearing the anisotropies, thereby smoothing out features in the power spectrum. The temperature field is affected by this smearing, so that combinations of derivatives can be used to extract the lensing signal directly, at least in principle (Seljak & Zaldarriaga 1999). The projected matter field can be reconstructed through a combination of this technique and correlations with other signals (Zaldarriaga & Seljak 1999). For example, the large angle signal caused by the variation in gravitational potential (the ‘ISW effect’) may be correlated with the lensing signal in open or $`\mathrm{\Lambda }`$-dominated models. However, the level of such a correlated signal is not likely to be large. One can easily imagine searching for all sorts of other correlations, for example the lensing signal with S-Z signals, with surveys at other wavelengths, e.g. large-scale structure, X-ray maps, etc.
#### Spatial-spectral signals
At the moment the only significant signal which mixes both spatial and spectral deviations is the S-Z effect. Although we have no specific ideas, we imagine that other such effects, involving perhaps different scattering processes, are likely to be developed in the near future. Although we expect the effect to be quite small, we mention as an example that Rayleigh scattering, which would spectrally filter anisotropy signals, has been omitted from calculations. In addition there could in principle be resonant line scattering from molecules in clouds at high redshift. Searches for such mixed spatial-spectral signals seem likely to become more important as multi-frequency data-sets improve in quality and quantity.
### 3.2. Non-anisotropy
Non-anisotropy measurements are heroically hard to do; certainly such things are worth pursuing, but the immediate returns are not as obvious as for the current anisotropy prospects. On the other hand, we expect that effort will fairly soon return to this direction when the ‘easy’ results have been mined from the primary power spectrum. Here we simply list a number of possibilities. Figure 3 shows the form of some of the standard distortions to the CMB spectrum.
A FIRAS/COBRA style total emission measurement of the spectrum of the sky will almost entirely be dominated by foregrounds outside of the 20 to $`400`$GHz frequency window in which the spectrum is already well measured. The present FIRAS limits allow parametrized spectral distortions as large as 10s of mK, easily larger than the measurement uncertainty in a careful experiment, but 100 to 1000 times dimmer than diffuse galactic emission at those wavelengths! Perhaps a multi-frequency measurement with appropriate angular resolution and sky coverage will allow a reliable extrapolation to zero galactic emission, but it will not be easy. Details for commonly considered distortions are listed below.
#### Compton distortions
$`y`$-distortions have essentially already been measured, in the sense that the sum of all the S-Z detected structure will give the uniform Compton-distortion over the sky. Certainly this gives a lower limit, which seems likely to be the bulk of the detectable signal (barring unforeseen exotic processes). The size of this signal is estimated to be $`y10^6`$ (e.g. Colafrancesco et al. 1997), depending on the cosmology. After the Planck mission (and S-Z investigations from ground-based interferometers) we will have a very precise estimate for the uniform $`y`$-distortion (and indeed an estimate of its power spectrum as well). Between this underlying signal, and the FIRAS upper limit on a full-sky distortion, there will remain only a rather narrow window to search for possible isotropic $`y`$-distortions from other sources (such as late energy injection, unrelated to cluster formation). Since there are no immediate candidates for such processes, and the window is rather narrow, we don’t see this as a particularly strong motivation for mounting a next generation FIRAS mission.
#### Free-free emission
For late energy releases, free-free emission leads to a distortion in the CMB spectrum, which increases towards lower frequency. This seems to be the type of distortion which is most feasible to measure in the near future for realistic models of the Universe. The best upper limits at the moment imply free-free optical depths of order $`Y_{\mathrm{ff}}<10^5`$ (e.g. Nordberg & Smoot 1998). Since this distortion increases at lower frequencies, then it is best investigated at the lowest frequency at which foreground signals can be dealt with, which means somewhere around $`5`$GeV. The expected signal at these frequencies may be as high as $`100\mu `$K, corresponding to $`Y_{\mathrm{ff}}`$ only about an order of magnitude below the current limits. The planned experiments ARCADE and DIMES (Kogut 1996) may be able to reach into the parameter space for realistic models, and help us understand more about the early ionized stages of the intergalactic medium. One nice thing about free-free is that lowering the temperature of the ionized medium increases the distortion (approximately $`n_\mathrm{e}^2/\sqrt{T_\mathrm{e}}`$), even although it decreases the Compton distortion. Hence good limits on $`Y_{\mathrm{ff}}`$ imply either low reionization redshifts or high electron temperatures, and limits on $`y`$ would restrict $`T_\mathrm{e}`$, so that direct limits on $`z_{\mathrm{reion}}`$ could be obtained.
#### Chemical potential
Current limits on $`\mu `$-type distortions are at the $`10^5`$ level. Note that this allows about $`15`$mK at $`1`$GHz within the error budget of the measurements, which is about 0.1% of the galactic signal. So pushing that limit further down is going to be tricky! The way to do this would presumably be to make a spectral map of the sky and extrapolate to zero galaxy (essentially what FIRAS did). So how big could a signal be?
Some amount of $`\mu `$-distortion is unavoidable, since it is generated by the damping of small-scale perturbations. For realistic models the value is likely to be around $`\mathrm{few}\times 10^8`$ (Hu, Scott & Silk 1994), which seems unlikely ever to be measurable. Of course various exotic processes, including energy injection at redshifts $`z10^5`$ could give much higher values of $`\mu `$. Limits could be set by experiments which also constrain free-free signals. However, we see no compelling reason currently to invest heavily in future experiments seeking to measure $`\mu `$ itself. Of course, if any hint of signal were to turn up then that would be extremely exciting (since unexpected) – in that case further investigation of the turn-off in the distortion at low frequencies would probe an otherwise unexplored early epoch.
#### Recombination lines
When the Universe recombined, every atom emitted at least one Lyman photon, or else got from the first excited state to the ground state via the two-photon process (see Seager, Scott & Sasselov 1999 for more details). This is a lot of photons, waiting there to be discovered! Mere measurement of the background of these photons would be an unprecedented confirmation of the Big Bang paradigm, that the Universe became neutral at $`z1000`$. Further investigation of these recombination lines would be a direct probe of the recombination process, and might provide further cosmological constraints. For example, the strengths of the residual Ly$`\alpha `$ feature and the two-photon feature will depend on the baryon density and on the expansion rate, hence allowing measurements of $`\mathrm{\Omega }_\mathrm{B}`$ and $`H`$ at $`z1000`$.
The problem is that the main recombination lines lie at wavelengths $`\lambda 150\mu `$m, where the signal from the galaxy is orders of magnitude stronger. The way to try to find the signal is then presumably to have enough spatial information to be able to extrapolate to zero Galaxy, and at the same time to have adequate spectral information to distinguish the relatively wide spectral feature. If all else failed it might be possible to rely on the dipole to extract the cosmological signal, but that would be even more difficult. So we might envisage an experiment with reasonable sky coverage, low angular resolution, but at least 3 spectral channels (say in the range 100–200$`\mu `$m) to extract the wide line. The spectral resolution would have to be good enough to distinguish this from a roughly isotropic component of warm interstellar dust – but that should be possible given that the spectral shape of the recombination lines is calculable (Dell’Antonio & Rybicki 1993, Boschan & Biltzinger 1998).
#### 21 cm line studies
If the Universe became reionized at redshifts between 5 and 20 there should be a spectral feature due to red-shifted $`21`$cm emission from neutral hydrogen which appears today at $`70`$ to $`240`$MHz (see Shaver et al. 1999). This emission can be seen against the CMB provided that there are spatial or spectral signatures (e.g. Tozzi, et al. 1999) and a mechanism which decoupled the electron spin temperature from the CMB. In principle, such studies, using the proposed Square Kilometer Array for example, could provide information about the processes that marked the end of the so-called Dark Ages, i.e. the reionization process and the formation of the first structures. This endeavor is sometimes called ‘cosmic tomography’.
#### Other diagnostics of the ‘Dark Ages’
There are of course other ways of probing the end of the Dark Ages, and even into that epoch, many of which might come from entirely different wavelengths, for example the near infra-red with NGST. However, we imagine that the microwave band will continue to be important in furnishing new ideas for exploring the domain between $`z=5`$ and $`z=1000`$. One recent speculative idea involves searching for masers which may come from structures at either the recombination or reionization epochs (Spaans & Norman 1997). There will surely be other such ideas in the coming years.
#### Measurements of $`T_{\mathrm{CMB}}(z)`$
The currently best value for the CMB temperature is $`T_0=2.725\pm 0.001`$K (Mather et al. 1999). It seems unclear why anyone should care about a more precise measurement than that! Before the existence of the CMB was even suspected, there was evidence for excess excitation in line ratios of certain molecules, notably cyanogen , in the interstellar medium (McKellar, 1941). This method has more recently been used to constrain the CMB temperature at high redshifts ($`z\mathrm{few}`$) using line with excitation temperatures of the relevant energy (e.g. Songaila et al. 1994, Roth & Bauer 1999). Measurements of other line ratios etc. can be used to set limits on the variation of fundamental physical constants (e.g. Webb et al. 1999). In a similar way, detailed measurement of the blackbody shape indicates that certain combinations of fundamental constants have not varied much since $`z1000`$.
## 4. Epilogue
Assuming that MAP and Planck are fully successful, and that the current suite of ground- and balloon-based experiments also return exquisite data, what then? Will this be then end of the study of the CMB?<sup>4</sup><sup>4</sup>4Churchill warned that ‘success is never final’. He also pointed out that ‘it is a good thing for an uneducated man to read books of quotations’. Eventually we can imagine a time when the primordial anisotropies have been measured so accurately that there are diminishing returns from further generations of satellite missions, and when small scale measurements, involving non-Gaussian signals, mixed spatial-spectral signals, and other complications, have moved firmly into the regime of ‘messy astrophysics’. However, there will be further primordial information to unlock from ever more ambitious polarization experiments. Certainly the CMB should not be looked at in isolation – although it is the dominant diffuse extragalactic background, there are several others to study (see Figure 4). And if that doesn’t leave the future still filled with exciting and challenging possibilities, there’s always the cosmic neutrino background!
##### Acknowledgments.
We thank the editors for their patience.
## References
Audit, E., & Simmons, J. F. L., 1999, MNRAS, submitted, \[astro-ph/9812310\]
Birkinshaw, M, 1999, Physics Reports, in press \[astro-ph/9808030\]
Bond, J. R., Jaffe, A. H., & Knox, L. E., 1998, ApJ, in press, \[astro-ph/9808264\]
Boschán, P., & Biltzinger, P., 1998, A&A, 336, 1
Colafrancesco, S., Mazzotta, P., Rephaeli, Y., & Vittorio, N., 1997, ApJ, 479, 1 \[astro-ph/9703121\]
Dell’Antonio, I. P., & Rybicki, G. B., 1993, in ‘Observational Cosmology’, ASP Conf. Ser. Vol. 51
Dwek, E., & Arendt, R. G., 1998, ApJ, 508, L9
Gush, H. P., Halpern, M., & Wishnow, E. H., 1990, Phys Rev. Lett., 65, 537
Hauser, M. G., et al., 1998, ApJ, 508, 25
Hu, W., Scott, D., & Silk, J., 1995, ApJ, 430, L5 \[astro-ph/9402045\]
Hu, W., Scott, D., Sugiyama, N., & White, M., 1995, Phys. Rev., D52, 5498 \[astro-ph/9505043\]
Hu, W., & White, M., 1996, A&A, 315, 33 \[astro-ph/9507060\]
Jaffe, A. H., & Kamionkowski, M., 1998, Phys. Rev., D58, 43001 \[astro-ph/9801022\]
Kappadath, S. C., et al., 1999, HEAD, 31, 3503
Kogut, A., 1996, in Proc. XVI Moriond Astrophysics meeting \[astro-ph/9607100\]
Lagache, G., Abergel, A., Boulanger, F., Désert, F. X., & Puget, J.-L., A&A, 344, 322 \[astro-ph/9901059\]
Lawrence, C. L., 1998, in ‘Evolution of Large Scale Structure’, ed. A.J. Banday et al., in press,
Lawrence, C. L., Scott, D., & White, M., 1999, PASP, in press, \[astro-ph/9810446\]
Leinert, C. H., et al., 1998, A&AS, 127, 1
Loeb, A., 1996, ApJ, 471, L1 \[astro-ph/9511075\]
Mather, J. C., et al., 1990, ApJ, 354, L37
Mather, J. C., Fixsen, D., Shafer, R. A., Mosier, C., & Wilkinson, D. T., 1999, ApJ, 512, 511 \[astro-ph/9810373\]
McKellar, A., 1941, Publ. Dom. Astrophys. Obs, Victoria, B.C., 7, 251
Miyaji, T., et al., 1998, A&A, 334, L13
Nordberg, H. P., & Smoot, G. F., 1998, ApJ, submitted \[astro-ph/9805123\]
Pozzetti, L., et al., 1998, MNRAS, 298, 1133
Roth, K. C., & Bauer, J. M., 1999, ApJ, 515, L57
Seager, S., Sasselov, D., & Scott, D., 1999, ApJ, submitted
Seljak, U., & Zaldarriaga, M., 1999, Phys. Rev. Lett., 82, 2636 \[astro-ph/9810092\]
Shaver, P. A., Windhorst, R. A., Madau, P., & de Bruyn, A. G., 1990, A&A, submitted \[astro-ph/9901320\]
Smoot, G. F., 1997, Proc. of the Strasbourg NATO School, in press \[astro-ph/9705101\]
Smoot, G. F., & Scott, D., 1998, in Caso C., et al., Eur. Phys. J., C3, 1, the Review of Particle Physics. p.$`\mathrm{\hspace{0.17em}127}`$ \[astro-ph/9711069\]
Songaila, A., et al., 1994, Nature, 371, 43
Spaans, M., & Norman, C. A., 1997, ApJ, 483, 87
Spergel, D. N., & Goldberg, D. M., 1999, ApJ, in press \[astro-ph/9811252\]
Sreekumar, P., et al., 1998, ApJ, 494, 523
Sunyaev, R. A., & Zel’dovich, Ya. B., 1980, ARAA, 18, 537
Tozzi, P., Madau, P., Meiksin, A., & Rees, 1999, ApJ, submitted \[astro-ph/9903139\]
Tuluie, R., Laguna, P., & Anninos, P., 1996, ApJ, 463, 15
Webb, J. K., et al., 1999, Phys. Rev. Lett., 82, 884
White, M., Carlstrom, J. E., Dragovan, M., & Holzapfel, S. W. L., 1997, ApJ, in press \[astro-ph/9712195\]
Zaldarriaga, M., & Seljak, U., 1999, Phys. Rev. D, in press \[astro-ph/9810257\]
Anisotropy Results
COBE: Hinshaw, G. et al., 1996, ApJ, 464, L17
FIRS: Ganga, K., Page, L., Cheng, E., & Meyers, S., 1993, ApJ, 432, L15
Ten.: Gutiérrez, C. M., et al., 1997, ApJ, 480, L83
BAM: Tucker, G. S., et al., 1997, ApJ, 475, L73
SP91: Schuster, J., et al., 1993, ApJ, 412, L47 (Revised, see SP94 reference)
SP94: Gundersen, J. O., et al., 1994, ApJ, 443, L57
Sask.: Netterfield, et al., 1997, ApJ, 474, 47
Pyth.: Platt, S. R., et al., 1997, ApJ, 475, L1
PyV: Coble, K., et al., 1999, ApJ, in press \[astro-ph/9902195\]
ARGO: de Bernardis, P., et al., 1994, ApJ, 422, L33; Masi, S., et al., 1996, ApJ463, L47
QMAP: de Oliveira-Costa, A., et al., 1998, ApJ, 509, L77; Herbig, T., et al., 1998, ApJ, 509, L73; Devlin, M., et al., 1998, ApJ, 509, L69
IAB: Piccirillo, L., & Calisse, P., 1993, ApJ, 413, 529
MAX: Tanaka, S. T., et al., 1996, ApJ, 468, L81; Lim, M., et al., 1996, ApJ, 469, L69
MSAM: Cheng, E. S., et al., 1996, ApJ, 456, L71; Cheng, E. S., et al., 1997, ApJ, 488, L59; Wilson, G. W., et al., ApJ, in press \[astro-ph/9902047\]
CAT: Scott, P. F. S., et al., 1996, ApJ, 461, L1; Baker, J. C., 1997, Proc. Particle Physics and Early Universe Conf., (http://www.mrao.cam.ac.uk/ peuc/astronomy/papers/baker/baker.html); Baker, J. C., et al., in preparation
WD: Tucker, G. S., Griffin, G. S., Nguyen, H. T., & Peterson, J. B., 1993, ApJ, 419, L45; Ratra, B., et al., 1998, ApJ, 505, 8
OVRO: Readhead, A. C. S., et al., 1989, ApJ, 346, 566
Ring: Leitch, E. M., et al., 1998, ApJ, in press \[astro-ph/9807312\]
SuZIE: Church, S. E., et al., 1997, ApJ, 484, 523
ATCA: Subrahmayan, R., Ekers, R. D., Sinclair, M., & Silk, J., 1993, MNRAS, 263, 416
RT: Jones, M. E., 1997, Proc. Particle Physics and Early Universe Conference (http://www.mrao.cam.ac.uk/ ppeuc/astronomy/papers/jones/jones.html); Jones, M. E., et al., in preparation
VLA: Partridge, B., et al., 1997, ApJ, 483, 38
|
no-problem/9904/astro-ph9904342.html
|
ar5iv
|
text
|
# Neutron Disks Around Black Holes
## 1 Introduction
Angular momentum in accretion disks around black holes must deviate from a Keplerian distribution, since the presence of ion, radiation or inertial pressure gradient forces become as significant as the gravitational and centrifugal forces (see Chakrabarti 1996a; Chakrabarti 1996b and references therein). The inertial pressure close to a black hole is high, because, on the horizon, the inflow velocity must be equal to the velocity of light. For causality, the velocity of sound must be less than the velocity of light. In fact, in the extreme equation of state of $`P=\frac{c^2}{3}\rho `$ (where $`c`$ is the velocity of light and $`P`$ and $`\rho `$ are the isotropic pressure and mass density respectively), the sound speed is only $`c/\sqrt{3}`$. Thus, the flow must pass through a sonic point and become supersonic before entering into the horizon. A flow which must pass through a sonic point must also be sub-Keplerian (Chakrabarti 1996b and references therein), and this causes the deviation. If the accretion rate is low, the flow cools down only by inefficient bremsstrahlung and Comptonization processes, unless the magnetic field is very high (Shvartsman 1971; Rees 1984; Bisnovatyi-Kogan 1998). This hot flow can undergo significant nucleosynthesis depending on the inflow parameters. Earlier, in the context of thick accretion disks calculations of changes in composition inside an accretion disk were carried out (Chakrabarti et al. 1987; Hogan & Applegate 1987; Arai & Hashimoto 1992; Hashimoto et al. 1993), but the disk models used were not completely self-consistent, in that neither the radial motion, nor the cooling and heating processes were included fully self-consistently. Secondly, only high accretion rates were used. As a result, the viscosity parameter required for a significant nuclear burning was extremely low ($`\alpha _{vis}<10^4`$). In the present paper, we do the computation after including the radial velocity in the disk and the heating and cooling processes. We largely follow the solutions of Chakrabarti (1996b) to obtain the thermodynamic conditions along a flow.
Close to a black hole horizon, the viscous time scale is so large compared to the infall time scale that the specific angular momentum $`\lambda `$ of matter remains almost constant and sub-Keplerian independent of viscosity (Chakrabarti 1996a,b; Chakrabarti 1989). Because of this, as matter accretes, the centrifugal force $`\lambda ^2/x^3`$ increases much faster compared to the gravitational force $`GM/x^2`$ (where $`G`$ and $`M`$ are the gravitational constant and the mass of the black hole respectively, $`\lambda `$ and $`x`$ are the dimensionless angular momentum and the radial distance from the black hole). As a result, close to the black hole (at $`x\lambda ^2/GM`$) matter may even virtually stop to form standing shocks (Chakrabarti 1989). Shock or no-shock, as the flow slows down, the kinetic energy of matter is converted into thermal energy in the region where the centrifugal force dominates. Hard X-rays and $`\gamma `$-rays are expected from here (Chakrabarti & Titarchuk, 1995). In this centrifugal pressure supported hot ‘boundary layer’ (CENBOL) of the black hole (Chakrabarti et al. 1996) we find that for low accretion rates, $`{}_{}{}^{4}He`$ of the infalling matter is completely photo-dissociated and no $`{}_{}{}^{7}Li`$ could be produced. In this region, about ten to twelve percent of matter is found to be made up of pure neutrons. These neutrons should not accrete very fast because of very low magnetic viscosity associated with neutral particles (Rees et al. 1982) while protons are dragged towards the central black hole along with the field lines. Of course, both the neutrons and protons would have ‘normal’ ionic viscosity, and some slow accretion of protons (including those produced after neutron decay) would still be possible. In contrast to neutron stars, the neutron disks which we find are not dense. Nevertheless, they can participate in the formation of neutron rich isotopes and some amount of deuterium. They can be eventually dispersed into the galaxy through jets and outflows, which come out of CENBOL (Chakrabarti 1998; Das & Chakrabarti 1998) thereby possibly influencing the metallicity of the galaxy.
On the equatorial plane, where the viscosity is the highest, a Keplerian disk deviates to become sub-Keplerian very close to the black hole (Chakrabarti & Titarchuk 1995; Wiita 1982). Away from the equatorial plane, viscosity is lower and the flow deviates from a Keplerian disk farther out. This is because the angular momentum transport is achieved by viscous stresses. Weaker the viscosity, longer is the distance through which angular momentum goes to match with a Keplerian disk. When the viscosity of the disk is decreased on the whole, the Keplerian disk recedes from the black hole forming quiescence states when the objects become very faint in X-rays (Ebisawa et al. 1996). Soft photons from the Keplerian disk are intercepted by this sub-Keplerian boundary layer (CENBOL) and photons are energized through Compton scattering process. For higher Keplerian rates, electrons and protons cool down completely and the black hole is in a soft state (Tanaka & Lewin 1995). Here, bulk motion Comptonization produces the power-law tail of slope $`\alpha 1.5`$ (Chakrabarti & Titarchuk 1995; Titarchuk et al. 1997). For lower Keplerian rates, the Compton cooling is incomplete and the temperature of the boundary layer remains close to the virial value,
$$T_p\frac{1}{2k}m_pc^2\frac{x_g}{x}=5.2\times 10^{11}\left(\frac{10}{x/x_g}\right){}_{}{}^{\mathrm{o}}\mathrm{K}.$$
$`(1)`$
In this case, bremsstrahlung is also important and the black hole is said to be in a hard state with energy spectral index $`\alpha `$ ($`F_\nu \nu ^\alpha `$, where $`\nu `$ is the frequency of the photon) close to $`0.5`$. In Eq. (1), $`m_p`$ is the mass of the proton, $`x_g=2GM/c^2`$ is the Schwarzschild radius of the black hole, and $`c`$ is the velocity of light. (In future, we measure the distances and velocities in units of $`x_g`$ and $`c`$.) In this low Keplerian rate, electrons are cooler typically by a factor of $`(m_p/m_e)^{1/2}`$ unless the magnetic field is very high. Present high energy observations seem to support the apparently intriguing aspects of black hole accretion mentioned above. For instance, the constancy of (separate) spectral slopes in soft and hard states has been observed by many (Ebisawa et al. 1994; Miyamoto et al. 1991; Ramos et al. 1997; Grove et al. 1998; Vargas et al. 1997). ASCA observations of Cygnus X-1 seem to indicate that the inner edge of the Keplerian component is located at around $`15R_g`$ (instead of $`3R_g`$) (Gilfanov et al. 1997). HST FOS observations of the black hole candidate A0620-00 in quiescent state seems to have very faint Keplerian features (McClintock et al. 1995) indicating the Keplerian component to be farther out at low accretion rates. Bulk motion Comptonization close to the horizon has been considered to be a possible cause of the power-law tail in very soft states (Crary et al., 1996; Ling et al. 1997; Cui et al. 1997). However, some alternative modes may not be ruled out to explain some of these features.
This observed and predicted dichotomy of states of black hole spectra motivated us to investigate the nuclear reactions thoroughly for both the states, but we report here the results obtained in the more important case, namely, when the flow is hotter, i.e., for hard states. We use $`255`$ nuclear elements in the thermo-nuclear network starting from protons, neutrons, deuterium etc. till $`{}_{}{}^{72}Ge`$ and the nuclear reaction rates valid for high temperatures. We assume that accretion on the galactic black hole is taking place from a disk where matter is supplied from a normal main sequence star. That is, we choose the abundance of the injected matter to be that of the sun. Because of very high temperature, the result is nearly independent of the initial composition, as long as reasonable choices are made. When accretion rates are higher, the advective region becomes cooler and very little nucleosynthesis takes place, the results are presented elsewhere (Mukhopadhyay 1998; Mukhopadhyay & Chakrabarti 1998).
As hot matter approaches a black hole, photons originated by the bremsstrahlung process, as well as those intercepted from the Keplerian disk, start to photo-dissociate deuterium and helium in the advective region. There are two challenging issues at this stage which we address first: (a) Thermodynamic quantities such as density and temperature inside a disk are computed using a thin disk approximation, i.e., the vertical height $`h(x)`$ at a radial distance $`x`$ very small compared to $`x`$ ($`h(x)<<x`$), and assuming the flow to be instantaneously in vertical equilibrium. However, at a low rate, it is easy to show that the disk is optically thin in the vertical direction $`_0^{h(x)}\rho \sigma 𝑑h<1`$ ($`\sigma `$ is the Thomson scattering cross-section). However, soft photons from the Keplerian disk enter radially and $`_1^{x_s}\rho \sigma 𝑑x>1`$, generally. In fact, this latter possibility changes the soft photons of a few keV from a Keplerian disk to energies up to $`1`$MeV by repeated Compton scattering (Sunyaev & Titarchuk 1980; Chakrabarti & Titarchuk 1995) while keeping the photon number strictly constant. The spectrum of the resultant photons emitted to distant observers becomes a power law $`F_\nu \nu ^\alpha `$ instead of a blackbody, where $`\alpha 0.5`$ for hard state and $`\alpha 1.5`$ for soft states of a black hole. (b) Now that the spectrum is not a blackbody, strictly speaking, the computation of photo-disintegration rate that is standard in the literature (which utilizes a Planckian spectrum) cannot be followed. Fortunately, this may not pose a major problem. As we shall show, the standard photo-disintegration rate yields a lower limit of the actual rate that takes place in the presence of power-law photon spectra. Thus, usage of the correct rate obtainable from a power-law spectrum would, if anything, strengthen our assertion about the photo-disintegration around a black hole. After photo-disintegration by these hard photons, all that are left are protons and neutrons. The exact location where the dissociation actually starts may depend on the detailed photon spectrum, i.e., optical depth of this boundary layer and the electron temperature.
The plan of the present paper is the following: in the next section, we present briefly the hydrodynamical model using which the thermodynamic quantities such as the density and temperature inside the inner accretion disk are computed. We also present the model parameters we employ. In Sect. 3, we present results of nucleosynthesis inside a disk. Finally, in Sect. 4, we present out concluding remarks.
## 2 Model Determining the Thermodynamic Conditions
We chose the units of distance, time and mass to be $`2GM/c^2`$, $`2GM/c^3`$ and $`M`$ where, $`G`$ is the gravitational constant, $`M`$ is the mass of the black hole, and $`c`$ is the velocity of light. To keep the problem tractable without sacrificing the salient features, we use a well understood model of the accretion flow close to the black hole. We solve the following equations (Chakrabarti 1996a,b) to obtain the thermodynamic quantities:
(a) The radial momentum equation:
$$\vartheta \frac{d\vartheta }{dx}+\frac{1}{\rho }\frac{dP}{dx}+\frac{\lambda _{Kep}^2\lambda ^2}{x^3}=0,$$
$`(2a)`$
(b) The continuity equation:
$$\frac{d}{dx}(\mathrm{\Sigma }x\vartheta )=0,$$
$`(2b)`$
(c) The azimuthal momentum equation:
$$\vartheta \frac{d\lambda (x)}{dx}\frac{1}{\mathrm{\Sigma }x}\frac{d}{dx}(x^2W_{x\varphi })=0,$$
$`(2c)`$
(d) The entropy equation:
$$\mathrm{\Sigma }vT\frac{ds}{dx}=\frac{h(x)\vartheta }{\mathrm{\Gamma }_31}(\frac{dp}{dx}\mathrm{\Gamma }_1\frac{p}{\rho })=Q_{mag}^++Q_{nuc}^++Q_{vis}^+Q^{}=Q^+g(x,\dot{m})Q^+=f(\alpha ,x,\dot{m})Q^+.$$
$`(2d)`$
Here, $`Q^+`$ and $`Q^{}`$ are the heat gained and lost by the flow, and $`\dot{m}`$ is the mass accretion rate in units of the Eddington rate. Here, we have included the possibility of magnetic heating $`Q_{mag}^+`$ (due to stochastic fields; Shvartsman 1971; Shapiro, 1973; Bisnovatyi-Kogan, 1998) and nuclear energy release $`Q_{nuc}^+`$ as well (cf. Taam & Fryxall 1985) while the cooling is provided by bremsstrahlung, Comptonization, and endothermic reactions and neutrino emissions. A strong magnetic heating might equalize ion and electron temperatures (e.g. Bisnovatyi-Kogan 1998) but this would not affect our conclusions. On the right hand side, we wrote $`Q^+`$ collectively proportional to the cooling term for simplicity (purely on dimensional grounds). We use the standard definitions of $`\mathrm{\Gamma }`$ (Cox & Giuli 1968),
$$\mathrm{\Gamma }_3=1+\frac{\mathrm{\Gamma }_1\beta }{43\beta },$$
$$\mathrm{\Gamma }_1=\beta +\frac{(43\beta )^2(\gamma 1)}{\beta +12(\gamma 1)(1\beta )}$$
and $`\beta (x)`$ is the ratio of gas pressure to total pressure,
$$\beta (x)=\frac{\rho kT/\mu m_p}{\rho kT/\mu m_p+\overline{a}T^4/3+B(x)^2/4\pi }$$
Here, $`\overline{a}`$ is the Stefan constant, $`k`$ is the Boltzman constant, $`m_p`$ is the mass of the proton, $`\mu `$ is the mean molecular weight. Using the above definitions, Eq. (2d) becomes,
$$\frac{43\beta }{\mathrm{\Gamma }_1\beta }[\frac{1}{T}\frac{dT}{dx}\frac{1}{\beta }\frac{d\beta }{dx}\frac{\mathrm{\Gamma }_11}{\rho }\frac{d\rho }{dx}]=f(\alpha ,x,\dot{m})Q^+.$$
$`(2e)`$
In this paper, we shall concentrate on solutions with constant $`\beta `$. Actually, we study in detail only the special cases, $`\beta =0`$ and $`\beta =1`$, so we shall liberally use $`\mathrm{\Gamma }_1=\gamma =\mathrm{\Gamma }_3`$. We note here that unlike self-gravitating stars where $`\beta =0`$ causes instability, here this is not a problem. Similarly, we shall consider the case for $`f(\alpha ,x,\dot{m})`$ = constant, though as is clear, $`f0`$ in the Keplerian disk region and probably much greater than $`0`$ near the black hole depending on the efficiency of cooling (governed by $`\dot{m}`$, for instance). We use the Paczyński-Wiita (1980) potential to describe the black hole geometry. Thus, $`\lambda _{Kep}`$, the Keplerian angular momentum is given by, $`\lambda _{Kep}^2=x^3/2(x1)^2`$, exactly same as in general relativity. $`W_{x\varphi }`$ is the vertically integrated viscous stress, $`h(x)`$ is the half-thickness of the disk at radial distance $`x`$ (both measured in units of $`2GM/c^2`$) obtained from vertical equilibrium assumption (Chakrabarti 1989) $`\lambda (x)`$ is the specific angular momentum, $`\vartheta `$ is the radial velocity, $`s`$ is the entropy density of the flow. The constant $`\alpha `$ above is the Shakura-Sunyaev (1973) viscosity parameter modified to include the pressure due to radial motion ($`\mathrm{\Pi }=W+\mathrm{\Sigma }\vartheta ^2`$, where $`W`$ and $`\mathrm{\Sigma }`$ are the integrated pressure and density respectively; see Chakrabarti & Molteni (1995) in the viscous stress. With this choice, $`W_{x\varphi }`$ keeps the specific angular momentum continuous across of the shock.
For a complete run, we supply the basic parameters, namely, the location of the sonic point through which flow must pass just outside the horizon $`X_{out}`$, the specific angular momentum at the inner edge of the flow $`\lambda _{in}`$, the polytropic index $`\gamma `$, the ratio $`f`$ of advected heat flux $`Q_+Q_{}`$ to heat generation rate $`Q^+`$, the viscosity parameter $`\alpha _{vis}`$ and the accretion rate $`\dot{m}`$. The derived quantities are: $`x_{tr}`$ where the Keplerian flow deviates to become sub-Keplerian, the ion temperature $`T_p`$, the flow density $`\rho `$, the radial velocity $`v_r`$ and the azimuthal velocity $`\lambda /x`$ of the entire flow from $`x_{tr}`$ to the horizon. Temperature of the ions obtained from above equations is further corrected using a cooling factor $`F_{Comp}`$ obtained from the results of radiative transfer of Chakrabarti & Titarchuk (1995). Electrons cool due to Comptonization, but they cause the ion cooling also since ions and electrons are coupled by Coulomb interaction. $`F_{Comp}`$, chosen here to be constant in the advective region, is the ratio of the ion temperature computed from hydrodynamic (Chakrabarti 1996b) and radiation-hydrodynamic (Chakrabarti & Titarchuk 1995) considerations.
## 3 Results of Nucleosynthesis Calculations
In the first example, we start with a relativistic flow (polytropic index $`\gamma =4/3`$) with the accretion rate $`\dot{M}=0.01\dot{M}_{Edd}`$, where, $`\dot{M}_{Edd}`$ is the Eddington accretion rate. We use the mass of the central black hole to be $`M=10M_{}`$ throughout. We choose a very high viscosity and the corresponding $`\alpha `$ parameter (Shakura & Sunyaev 1973) is $`0.2`$ in the sub-Keplerian regime. The cooling is not as efficient as in a Keplerian disk: $`Q^{}0.9Q^+`$, where, $`Q^+`$ and $`Q^{}`$ are the heat generation and heat loss rates respectively. The specific angular momentum at the inner edge is $`\lambda _{in}=1.65`$ (in units of $`2GM/c`$). The flow deviates from a Keplerian disk at $`4.15`$ Schwarzschild radii. It is to be noted that $`Q^{}`$ includes all possible types of cooling, such as bremsstrahlung, Comptonization as well as cooling due to neutrino emissions. We assume that the flow is magnetized so that only ions have larger viscosity. Due to poor supply of the soft photons from Keplerian disks, the Comptonization in the boundary layer is not complete: we assume a standard value (Chakrabarti, & Titarchuk 1995) in this regime: $`F_{Comp}0.1`$, i.e., ions (in te radiation-hydrodynamic solution) are one-tenth as hot as obtained from the hydrodynamic solutions. \[For high accretion rate, $`\dot{m}>0.3`$, $`F_{Comp}0.001`$ and ions and electrons both cool to a few KeV ($`10^7`$ <sup>o</sup>K)\]. The typical density and temperature near the marginally stable orbit are $`\rho _{x=3}8.5\times 10^8`$ gm cm<sup>-3</sup> and $`7.5\times 10^9`$ <sup>o</sup>K respectively where the thermonuclear depletion rates $`N_A<\sigma v>`$ for the $`Dp+n`$, $`{}_{}{}^{4}HeD+D`$ and $`{}_{}{}^{4}He+^4He=^7Li+p`$ reactions are given by $`1.6\times 10^{14}`$ gm<sup>-1</sup> s<sup>-1</sup>, $`4\times 10^3`$ gm<sup>-1</sup> s<sup>-1</sup> and $`1.9\times 10^{12}`$ gm<sup>-1</sup> s<sup>-1</sup> respectively. Here, $`N_A`$ is the element abundance on the left, $`\sigma `$ is the reaction cross-section, $`v`$ is the Maxwellian average velocity of the reactants. At these rates, the time scales of these reactions are given by, $`4\times 10^5`$s, $`5\times 10^{11}`$s and $`4\times 10^{20}`$s respectively indicating that the deuterium burning is the fastest of the reactions. In fact, it would take about a second to burn initial deuterium with $`Y_D=10^5`$. The $`{}_{}{}^{7}Li`$ does not form at all because the $`{}_{}{}^{4}He`$ dissociates to $`D`$ much faster.
The above depletion rates have been computed assuming Planckian photon distribution corresponding to ion temperature $`T_p`$. The wavelength $`\lambda _{Planck}`$ at which the brightness is highest at $`T=T_p`$ is shown in Fig. 1 in the dashed curve (in units of $`10^{11}`$ cm). Also shown is the average wavelength of the photon $`\lambda _{Compton}`$ (solid curve) obtained from the spectrum $`F_\nu \nu ^\alpha `$. The average has been performed over the region $`2`$ to $`50`$keV of the photon energy in which the hard component is usually observed
$$<F_\nu >=\frac{_{\nu _{min}}^{\nu _{max}}F_\nu 𝑑\nu }{_{\nu _{min}}^{\nu ^{max}}𝑑\nu }=\nu _{Compton}^\alpha $$
$`(3)`$
where, $`\nu _{min}`$ and $`\nu _{max}`$ are computed from $`2`$ and $`50`$keV respectively. The average becomes a function of the energy spectral index $`\alpha `$ ($`F_\nu \nu ^\alpha `$), which in turn depends on the ion and electron temperatures of the medium. We follow Chakrabarti & Titarchuk (1995) to compute these relations. We note that $`\lambda _{Compton}`$ is lower compared to $`\lambda _{Planck}`$ for all ion temperatures we are interested in. Thus, the disintegration rate with Planckian distribution that we employed in this paper is clearly a lower limit. Our assertion of the formation of a neutron disk should be strengthened when Comptonization is included.
Figure 2 shows the result of the numerical simulation for the disk model mentioned above. Logarithmic abundance of neutron $`Y_n`$ is plotted against the logarithmic distance from the black hole. First simulation produced the dash-dotted curve for the neutron distribution, forming a miniature neutron torus. As fresh matter is added to the existing neutron disk, neutron abundance is increased as neutrons do not fall in rapidly. Thus the simulation is repeated several times in order to achieve a converging steady pattern of the neutron disk. Although fresh neutrons are deposited, the stability of the distribution is achieved through neutron decay and neutron capture reactions. Results after every ten iterations are plotted. The equilibrium neutron torus remains around the black hole indefinitely. The neutron abundance is clearly very significant (more than ten per cent!).
We study yet another case where the accretion rate is smaller ($`\dot{m}=0.001`$) and the viscosity is so small ($`\alpha =0.01`$) and the disk so hot that the sub-Keplerian flow deviates from a Keplerian disk farther away at $`x=85.1`$. The polytropic index is that of a mono-atomic (ionized) hot gas $`\gamma =5/3`$. The Compton cooling factor is as above since it is independent of the accretion rates as long as the rate is low (Sunyaev & Titarchuk 1980; Chakrabarti & Titarchuk 1995). The cooling is assumed to be very inefficient because of lower density: $`Q^{}0.4Q^+`$. The specific angular momentum at the inner edge of the disk is $`\lambda _{in}=1.55`$. In Fig. 3, we show the logarithmic abundances of proton (p), helium ($`{}_{}{}^{4}He`$) and neutron (n) as functions of the logarithmic distance from the black hole. Note that $`{}_{}{}^{4}He`$ dissociates completely at a distance of around $`x=30`$ where the density and temperatures are $`\rho =2.29\times 10^{11}`$ gm cm<sup>-3</sup> and $`T=6.3\times 10^9`$ <sup>o</sup>K. Maximum temperature attained in this case is $`T_{max}=3.7\times 10^{10}`$ <sup>o</sup>K. Both the neutrons and protons are enhanced for $`x<30`$, the boundary layer of the black hole. This neutron disk also remains stable despite neutron decay, since new matter moves in to maintain equilibrium. The $`{}_{}{}^{7}Li`$ abundance is insignificant.
## 4 Concluding Remarks
In this paper, we have shown that hot flows may produce neutron disks around black holes, where neutron abundance is significant. However, unlike neutron stars, the formation of which is accompanied by the production of neutron rich isotopes, neutron disks do not produce significant neutron rich elements. Some fragile elements, such as deuterium, could be produced in the cooler outflows as follows:
Neutrons and protons may be released in space through winds which are produced in the centrifugal barrier. These winds are common in black hole sources and earlier they have been attributed to the dispersal of magnetic fields to the galactic medium (Daly & Loeb 1990; Chakrabarti et al. 1994). Recently, Chakrabarti (1998) and Das & Chakrabarti (1998), through a first ever self-consistent calculation of outflows out of accretion, found that significant winds can be produced and for low enough accretion rates, disks may even be almost evacuated causing the formation of quiescence and inactive states such as what is observed in V404 Cyg and our Galactic centre. If the temperature of the wind falls off as $`1/z`$ and density as $`z^{3/2}`$ (as is expected from an outflow of insignificant rotation), the deuterium synthesis rate $`n+pD`$, increases much faster very rapidly than the reverse ($`Dn+p`$) process. For instance, with density and temperature mentioned as in the earlier section, at $`z=30x_g`$, the forward rate ($`N_A<\sigma v>`$) is $`0.12\times 10^5`$ while the reverse rate is much higher: $`6.7\times 10^{13}`$. This results in the dissociation of deuterium. However, at $`z=300x_g`$, the above rates are $`1.8\times 10^8`$ and $`9.6\times 10^6`$ respectively and at $`z=3000x_g`$, the above rates are $`1.3\times 10^8`$ and $`10^{165}`$ respectively. Thus a significant deuterium could be produced farther out, say, starting from a distance of $`10^3x_g`$. Ramadurai & Rees (1985) suggested deuterium formation on the surface of ion tori. As we establish here, this process may be feasible, only if these tori are vertically very thick: $`z(x)10^3x_g`$. In any case, deuterium would be expected to form in winds and disperse.
In a typical case of a disk with an accretion rate of $`\dot{M}\dot{M}_{Edd}`$, the temperature is lower, but the density is higher. In that case, the photo-dissociation of $`{}_{}{}^{4}He`$ is insignificant and typically the change in abundances of some of the elements, such as $`{}_{}{}^{16}O`$, $`{}_{}{}^{20}Ne`$ etc. could be around $`\mathrm{\Delta }Y10^3`$ not as high as that of the neutron as in above cases where $`\mathrm{\Delta }Y_n0.1`$. One could estimate the contamination of the galactic metalicity due to nuclear reactions as we do for realistic models. Assume that, on an average, all the $`N`$ stellar black holes of equal mass $`M`$ have a non-dimensional accretion rate of around $`\dot{m}1`$ ($`\dot{m}=\dot{M}/\dot{M}_{Edd}`$). Let $`\mathrm{\Delta }Y_i`$ be the typical change in composition of this matter during the run and let $`f_w`$ be the fraction of the incoming flow that goes out as winds and outflows, then in the lifetime of a galaxy (say, $`10^{10}`$yrs), the total ‘change’ in abundance of a particular species deposited to the surroundings by all the stellar black holes is given by:
$$<\mathrm{\Delta }Y_i>10^9(\frac{\dot{m}}{1})(\frac{N}{10^6})(\frac{\mathrm{\Delta }Y_i}{10^3})(\frac{f_w}{0.1})(\frac{M}{10M_{}})(\frac{T_{gal}}{10^{10}})(\frac{M_{gal}}{10^{13}M_{}})^1.$$
$`(4)`$
We here assume a conservative estimate that there are $`10^6`$ such stellar black holes (there number varies from $`10^8`$ (van den Heuvel 1992, 1998) to several thousands (Romani, 1998) depending on assumptions made) and the mass of the host galaxy is around $`10^{13}M_{}`$ and the lifetime of the galaxy during which such reactions are going on is about $`10^{10}`$Yrs. We believe that $`<\mathrm{\Delta }Y_i>10^9`$ is quite reasonable for a typical case when $`\mathrm{\Delta }Y_i10^3`$ and a fraction of ten percent of matter is blown off as winds. When $`\mathrm{\Delta }Y_i0.1`$ or the outflow rate is higher (particularly in presence of strong centrifugal barrier) the contamination would be even higher.
It is to be noted that our assertion of formation of neutron disks around a black hole for very low accretion rate $`\dot{M}0.0010.01\dot{M}_{Edd}`$ is different from that of the earlier results (Hogan, & Applegate, 1987) where $`\dot{M}10\dot{M}_{Edd}`$ was believed to be the more favourable accretion rate. This is because in last decades the emphasis was on super-Eddington thick accretion tori. More recent computations suggest that advective regions are not as hot when the rates are very high. Another assertion of our work is that $`{}_{}{}^{7}Li`$ should not be produced in accretion disks at all. This is not in line with earlier suggestions (Jin 1990) also. That is because unlike earlier case where the spallation reaction $`{}_{}{}^{4}He`$ +$`{}_{}{}^{4}He`$ was dealt with in isolation, we study this in relation to other reactions prevalent in the disk. We find that $`{}_{}{}^{4}He`$ could be dissociated much before it can contribute to spallation. However, our work supports Ramadurai & Rees’ (1985) conjecture that deuterium may be produced in the outer regions of the disk provided the disk is at least as thick as $`10^3x_g`$.
In the process of performing the simulation we were faced with a challenge which was never addressed earlier in the literature. The problem arises because the inflow under consideration is optically thin vertically, but optically thick horizontally. As a result, photons emitted form a power-law spectrum. Question naturally arises, whether these power-law photons are capable of photo-disintegration. We find that the answer is yes and that the calculation of usual photo-disintegration gives a lower limit of the changes in the composition. In the extreme conditions close to the black hole, such processes are sufficiently effective to produce neutron disks around black holes.
We thank Mr. A. Ray for carefully reading the manuscript.
|
no-problem/9904/astro-ph9904045.html
|
ar5iv
|
text
|
# On the Cosmological Aspects of Observed High Energy Cosmic Phenomena
### Introduction
As is known, there is presently no physical theory of the universe : the Standard Cosmological Model (SCM) seems to fail to explain many Observed Universe features because of the singularity problem. The old question why the universe is on average uniform and isotropic is still unanswered. The situation is aggravated, as new observational data on high-energy processes become available. In the suggested alternative cosmology the relativistic properties of a universe matter outside the Observed Universe are revealed. It gives us a clue for explaining major SCM difficulties.
## 1 CBR as a “neo-ether”
In the SCM the whole Universe, which is the Observed Universe, represents a massive absolute reference frame matched with a Cosmic Background Radiation (SBR). The latter is even more perfect absolute reference system because it does not have peculiar velocity dispersion. Actually, this issue is similar to the one formulated in the question : why the Universe is so well “tuned” and “aligned”. However, the “neo-ether” issue emphasizes more distinctly a confrontation of the SCM with conventional Physics in interpretation of the cosmological observations. If the Universe has such an inherent property as an absolute reference system then special and general relativistic theories must be redone. It is not easy to accept a CBR explanation of where the matter goes to in the expanding universe. But it is unacceptable to blame the general relativity theory for a decrease of a CBR energy density by one expansion factor faster than a total energy density of a massive corpuscular matter .
The situation radically changes if one suggests that the Observed Universe is not the whole Universe but an ensemble of material objects in the Grand Universe. The latter is a multitude of different ensembles. Then a CBR as ether (an absolute reference frame) and other “strange” universe attributes become local features of the individual material system. This idea is put in a basis of the suggested alternative cosmological concept.
## 2 Antimatter issue
Antimatter is apparently present in the Observed Universe in small quantities while Physics shows no preference of matter over antimatter. If the whole universe is really baryon asymmetric then Physics must be deeply revised. The only way to save a baryon symmetry and a baryon charge conservation is again to suggest that the Observed (Home) Universe actually is for some reason a matter-made material system. One may call our Home Universe a representative of multitude of typical universes evenly made of either matter or antimatter and chaotically dispersed in a Grand Universe space. This is a continuation of the above idea of our Home Universe being a material system limited in volume. Thus, resolving the confrontation in baryon symmetry issue we can return to the characteristics of the whole Universe (Grand Universe, GU, for short) and to the concept of its evolution.
## 3 A Universe without absolute reference frame
An absence of the absolute reference system in the GU means a Lorentz invariance of a coordinate-momentum distribution function of a GU matter. It is known from statistics of a relativistic non-interacting gas that this function takes a form :
$$F(x_1,x_2,x_3;p_1,p_2,p_3)dx_1dx_2dx_3dp_1dp_2dp_3=Constdx_1dx_2dx_3dp_1dp_2dp_3$$
$`\left(1\right)`$
It characterizes a fully chaotic motion. In this idealized model an integral over a momentum distribution (1) diverges. It shows unlimited sky brightness (Olbers paradox). In practice a bolometer placed in GU space should measure a certain average energy density. It integrates radiation (matter flux) over a source distribution, which is a “last scattering sphere”. Hence, a real momentum (energy) distribution must have a smooth cut–off of upper energies. This distribution is also a Lorentz–invariant because a last scattering sphere does not depend on a choice of a reference system.
In terms of a velocity distribution of relativistic particles a special relativity theory gives a formula equivalent to (1), (see, for example, ) :
$$F\left(\beta \right)d\beta =Const\gamma ^4\beta ^2d\beta $$
$`\left(2\right)`$
for $`0<\beta <\beta _{max}`$ , where $`\beta _{max}`$ characterizes a cut–off parameter.
The above relativistic distribution of GU matter is seen unchanged in any free reference system. All forms of matter are supposed to be subjected to this law. Hence, a thought observer can not distinguish between his states of relative motion in principle. Evidently, he can find any reference system among multitude of local ones, none of them being absolute. The important conclusion is that a GU matter can exist in a state of chaotic relativistic motion, which is seen identical for observers in any moving free reference system. In space with a matter in this state any kind of an “absolute” reference system allowing to detect any effect due to the relative motion does not exist. Now we can develop the alternative cosmological concept suggesting that the GU matter is evolving in a selfcreating manner being in a baryon symmetric steady state, which is found as a Lorentz–invariant (chaotic) relativistic motion. The GU space is thought to be of Minkovski’s type. This is a scene for a cyclic evolution of a multitude of typical universes. Our Home Universe is one of them.
## 4 On a GU matter theory
The Grand Universe is an open system: there is no physical boundary. This is a primary reason for a GU matter being in a stationary state of a relativistic motion. Another important property of the GU system is a presence of two complementary entities, that is matter and antimatter. We may describe such a system in terms of a global conservation of sum of two kinds of energies : one is “locked” in a relativistic mass of matter and antimatter, the other is “released” in a matter-antimatter annihilation. Imagine for a moment that an initial state of the system is “locked” energy. Than an immediate process of energy release will start in a form of explosion (a GU version of “inflation” model). The other extreme case would be a state of full matter-antimatter annihilation. A “pendulum” starts going back through the process of pair creation restoring a selfsustained state of a relativistic motion. There is no energy dissipation, and the system acquires an equilibrium state in a form of continuing matter-antimatter annihilation and pair creation in parallel. Probably, this state may be described in terms of a maximal entropy of the GU system.
It is clear now that we describe the idea of a generalized matter transport theory, which is different from Boltzmann’s one first of all in above discussed two aspects : openness of the system, and matter annihilation/creation. Besides, we should introduce into this theory the basic laws of matter interactions, from gravity mechanics and physics, and nuclear physics, in particular. Boltzmann’s statistics of gravitating objects is known as a mathematically very complex problem. Hence, we may think about model approximations on different levels.
The most surprising what was found by author in a simple GU model is a statistical separation of matter and antimatter with a formation of an hierarchy structure, including evolving “typical universes” up to the size of our Home Universe. Evidently, only the smallest chances are given to a “typical universe” to reach a mature age.
We expect that a development of a GU matter transport theory will give us a basis for a physical theory of Cosmology with a following development of specific models of a universe evolution, galaxy formation, cosmological nucleonic synthesis and others. It seems that the theory requires only one adjustable parameter : a matter density. We can not exclude even that a more general model may be developed with “matter density long range waves”. In any case, the purpose of this theory would be solutions of the GU transport equation describing both a stationary Lorentz-invariant momentum distribution (discussed above) and an “evolutionary function” describing dynamics characteristics of matter structure, mass distribution of gravitationally linked systems (like evolving typical universes), in particular.
For further discussing the main topic of this report we need a qualitative picture of the GU scenario of our Home Universe evolution.
## 5 Our Home Universe evolution
Any typical universe should start with some gravitating coagulant. Its growth to an embryonic or a mature state may be thought as the result of a probabilistic survival. Over a lifetime it continually interacts with the GU environment, some of the interactions being random “catastrophic” events. The mass distribution function mentioned must characterize this process that is actually fluctuations of the GU matter. This function evidently is a monotonous one rapidly decreasing. A “tail” of the function is due to the extremely rare and huge fluctuations what are typical universe formations. A random “soft” collision between them resulted in either further growth or a major annihilation. Typical universes as they reach a “mature” size become unaffected while capturing a “small stuff” (due to a good statistical averaging). For survival they compete with each other on a similar size level. In this scenario any typical universe is a gravitationally linked relativistic ensemble characterized by a momentum distribution similar to one of the GU matter but with a lower cut–off parameter. Eventually a growing universe becomes vulnerable to the collision with a quite smaller anti–counterpart. The criterion for this stage may be written roughly in a form:
$$Gm/Rc^2\left(\gamma 1\right)<1$$
$`\left(3\right)`$
where : $`m`$ — mass, $`R`$ — mean radius, $`\gamma `$ — effective Lorentz factor, $`G`$ — gravitational constant, $`c`$ — speed of light.
Tracing back to redshifts $`z10`$ we can imagine our Home Universe being about one order smaller and about three orders denser than at present, what gives the criterion (3) a value close to the critical. If so, a mechanism triggering our early Home Universe to decay might be an abrupt mass drop due to an accidental collision with a smaller antimatter universe. As a result, we observe our present “expanding” Universe originated from some pre-expansion stage. Evidently, galaxies (many of them were formed at pre-expansion stage) are aligned in a Hubble’s flow reflecting an initial relativistic distribution of a universe matter. So, chaos turned into an order.
To complete the rough picture of Our Universe in the GU concept frame we should note that stars, galaxies and clusters are inner formations in a general GU structure hierarchy. A luminous matter is expected to be only a small fraction of a total mass that provides for a self-sustained galaxy evolution. Hence, a “dark matter”, which is an ordinary matter, should be the dominant part in a typical universe. If so, it has to be in equilibrium with a thermal radiation. In our expanding Universe the observed 2.7 K CBR bears information on a mean surface temperature of dark matter bodies (dust clouds included). In accordance with the GU concept a mass distribution of dark matter is expected to be broad enough what makes a universe space quite transparent. Dark matter naturally participates in an “expansion” process. One can easily find that this concept perfectly explains all observed CBR features, a temperature decrease inversely proportional to the expansion factor, in particular. In a closed system like Our Universe, a thermal radiation performs an adiabatic expansion work because of a pressure gradient across the universe volume. Hence, there is no paradox with a “missing energy”.
Now we are ready to return to the main topic of this report : to give a qualitative explanation of high-energy cosmic corpuscular and gamma rays, gamma bursts, quasars and jetting objects in the frame of the suggested alternative Cosmology.
## 6 Cosmic rays
According to the GU concept a space outside typical universes is filled with a baryon symmetric matter in relativistic motion characterized by a Lorentz–invariant momentum distribution. The matter includes highly energetic gas and all sorts of macroscopic objects. Any typical universe evolves due to the active interaction with this outer GU background. Our Home Universe being also exposed to this background radiation must reveal products resulted from a transport of this radiation through the Universe medium. So called primary cosmic rays (its high–energy tail) should be actually a secondary radiation from interuniverse sources. In previous author’s work this cosmic rays transport model is substantiated by numerical assessments. Cosmic rays researchers long time looked for natural mechanisms of acceleration. This is one of the old cosmological problems. The GU concept suggests an explanation of this phenomenon, paradoxically, as a contrary process of a moderation of primary “inter–universe” particles (both corpuscular and gamma quants) with much higher initial energies than observed.
## 7 Star explosions and gamma bursts
Gamma bursts discovered not long ago have been so far as enigmatic as cosmic rays. They happen to flash randomly and uniformly in the Universe space. Our simple explanation is as follows. The Universe space in fact contains an appreciable amount of antimatter in a form of material objects left from the moment of “Big Bump”; some captured later from the outer space. As mentioned before, a “dark matter” partially consists of it. Some objects might be single stars. Hence, high–energy cosmic gamma rays should have a component due to the annihilation processes. A proper physical model is needed to distinguish this component from many other sources. But part of a radiation of an annihilation type is thought to be clearly observable due to its pulse character. There must be comparatively rare events of collision of matter and antimatter solid objects of big masses. In particular, the whole star may be involved in an annihilation event observed as a gamma burst of huge energy. The data already available on the identification of luminous “disappearing” objects as gamma burst sources. This explanation is given here as a hypothesis in the GU concept frame for a numerical test.
## 8 Quasars
We have to explain several quasar features : small size, big power density, age comparable with one of the Universe, power variability. According to the GU concept, quasars are “afterglow” resulting from the “Big Bump” what was a collision of our early Universe with some smaller typical universe made of antimatter (scenario discussed in previous sections). It must be the most catastrophic event in Our Universe evolution. A huge amount of matter has been annihilated with following release of energy in different forms of radiation, gamma radiation, in particular. A quasar “engine” should be a comparably small power–generating ensemble of material objects with a high surface–to–volume ratio. It keeps burning out in a non–exploding manner due to a continually incoming antimatter gas stream. One may explain its varying power by gas pressure instability. We think that so–called jetting objects have the same physical nature. A rough estimate of the quasar phenomenon in a simplified model shows a feasibility of matter–antimatter annihilation power generating mechanism. Detailed numerical analysis will be made after a development of the whole GU concept.
## 9 Conclusion
Now again the “hot issues” in a cosmological field are the old problems of the Standard Cosmological Model : is the Universe flat or open, and what is a true value of a Hubble’s constant. As often happened in a science history, a resolution of a problem could be in finding that a problem is incorrectly formulated due to the wrong basic postulate. In the suggested alternative cosmology the answers to the above questions are : our Universe is neither open, nor flat, and an exact value of a Hubble’s constant does not exist at all, for a peculiar velocity field has been initially created and further developed in both transversal and radial directions. The key element in a theory of the alternative cosmology must be the idea of a self–creation. A baryon symmetric world in Minkovski’s space seems to be a perspective concept for a development of a cosmological theory having an explanatory and predictive power and being falsifiable.
### Acknowledgement
I want to express my gratitude to Professor Nikoly Rabotnov for interesting discussions of cosmological problems. I am also thankful to Professor Ryan Gieed for his interest and support of my work.
|
no-problem/9904/nucl-th9904049.html
|
ar5iv
|
text
|
# 𝜋𝑁𝑁 coupling determined beyond the chiral limit
## Abstract
Within the conventional QCD sum rules, we calculate the $`\pi NN`$ coupling constant, $`g_{\pi N}`$, beyond the chiral limit using two-point correlation function with a pion. We consider the Dirac structure, $`i\gamma _5`$, at $`m_\pi ^2`$ order, which has clear dependence on the PS and PV coupling schemes for the pion-nucleon interactions. For a consistent treatment of the sum rule, we include the linear terms in quark mass as they constitute the same chiral order as $`m_\pi ^2`$. Using the PS coupling scheme for the pion-nucleon interaction, we obtain $`g_{\pi N}=13.3\pm 1.2`$, which is very close to the empirical $`\pi NN`$ coupling. This demonstrates that going beyond the chiral limit is crucial in determining the coupling and the pseudoscalar coupling scheme is preferable from the QCD point of view.
preprint: TIT/HEP-419/NP
QCD sum rule is a framework which connects hadronic parameters with QCD parameters. In this framework, a correlation function is introduced in terms of interpolating fields constructed from quark and gluon fields. The interpolating field is constructed so that its coupling to the hadron of concern is expected to be strong while its couplings to other higher resonances are hoped to be small. Then the correlator is calculated by Wilson’s operator product expansion (OPE) in the deep Euclidean region ($`q^2=\mathrm{}`$) and matched with the phenomenological “ansatz” to extract the hadron’s information in terms of QCD parameters.
One interesting quantity to be determined is the pion-nucleon coupling constant, $`g_{\pi N}`$. As the coupling is empirically well-known, successful reproduction of this quantity may provide a solid framework to determine other meson-baryon couplings as well as a better understanding of nonperturbative nature of hadrons. For this direction, the two-point correlation function for the nucleon interpolating field $`J_N`$,
$`\mathrm{\Pi }(q,p)=i{\displaystyle d^4xe^{iqx}0|T[J_N(x)\overline{J}_N(0)]|\pi (p)},`$ (1)
may be useful and it is often used in calculating $`g_{\pi N}`$ . Alternative approach is to consider the correlation function without pion but in an external axial field . This provides the nucleon axial charge, $`g_A`$, which can be converted to $`g_{\pi N}`$ with the help of the Goldberger-Treiman relation. Our interest in this work is to provide a reasonable value of $`g_{\pi N}`$ using Eq. (1) because its extension to other meson-baryon couplings seems to be more straightforward. Further advantage in using Eq. (1) is to provide a criterion for the PS-PV coupling schemes for the pion-nucleon interaction as will be discussed below.
The correlation function, Eq. (1), contains various independent Dirac structures, each of which can be in principle used to calculate $`g_{\pi N}`$. For example, Ref. uses the $`\gamma _5\overline{)}p`$ structure while Ref. uses the $`i\gamma _5`$ structure in the soft-pion limit. In the recent calculations , we proposed to use the $`\gamma _5\sigma _{\mu \nu }`$ structure in studying $`g_{\pi N}`$ as this structure is independent of the pseudoscalar (PS) and pseudovector (PV) coupling schemes employed in the phenomenological side. This sum rule contains very small contribution from the transition $`NN^{}`$, and the result is insensitive to the continuum threshold. Therefore, this structure provides a value of $`g_{\pi N}`$ independent of the coupling schemes. However, the result from this Dirac structure, $`g_{\pi N}10`$, is not quite satisfactory. Certainly a further improvement of this sum rule may be needed for the future extension to other SU(3) mesons.
Various improvements can be sought for. These may include a question related to the use of Ioffe’s nucleon current for the correlator, higher order corrections in the OPE, or corrections associated with the chiral limit. The last possibility for the improvement is interesting because $`g_{\pi N}`$ from the $`\gamma _5\sigma _{\mu \nu }`$ sum rule is rather close to the one in the chiral limit than its empirical value. In Ref. , the calculation is performed beyond the soft-pion limit by taking the leading order of the pion momentum $`p_\mu `$, but for the rest of the correlator the chiral limit, $`p^2=m_\pi ^2=0`$, is taken. Thus, it is not clear whether the calculation is performed beyond the chiral limit and this may cause the discrepancy with the empirical $`g_{\pi N}`$.
In this paper, we pursue an improvement by presenting a QCD sum rule calculation beyond the chiral limit. Specifically, we consider the Dirac structure, $`i\gamma _5`$, at the order, $`p^2=m_\pi ^2`$. The sum rule for the structure, $`i\gamma _5`$, is, first of all, technically less involved when the calculation is done beyond the chiral limit. Secondly, even beyond the chiral limit, this structure is clearly PS-PV coupling-scheme dependent. Therefore, the successful reproduction of the empirical value for $`g_{\pi N}`$ may provide an important QCD constraint for the pion-nucleon interaction type.
To see the coupling scheme dependence more clearly, we use the PS and PV Lagrangians
$`_{ps}=g_{\pi N}\overline{\psi }i\gamma _5𝝉𝝅\psi ;_{pv}={\displaystyle \frac{g_{\pi N}}{2m}}\overline{\psi }\gamma _5\gamma _\mu 𝝉^\mu 𝝅\psi ,`$ (2)
in constructing the phenomenological side of the correlator, Eq. (1). Then the correlator is expanded in terms of the pion momentum $`p_\mu `$. Using the PS Lagrangian, we obtain for the $`i\gamma _5`$ structure ,
$`g_{\pi N}\lambda ^2\left[{\displaystyle \frac{1}{q^2m^2}}{\displaystyle \frac{pq}{(q^2m^2)^2}}+{\displaystyle \frac{p^2}{(q^2m^2)^2}}\right]+\mathrm{}.`$ (3)
Here $`\lambda `$ is coupling of $`J_N`$ to the physical nucleon, $`m`$ is nucleon mass. Note that the first term is the phenomenological part of the sum rule in the soft-pion limit . The second term containing $`pq`$ is not the same chiral order as $`m_\pi ^2`$. Thus at $`p^2=m_\pi ^2`$, the phenomenological correlator takes the form,
$`m_\pi ^2{\displaystyle \frac{g_{\pi N}\lambda ^2}{(q^2m^2)^2}}+\mathrm{}.`$ (4)
The ellipses indicate the contribution when $`J_N`$ couples to higher resonances. This includes the continuum contribution whose spectral density is usually parameterized by a step function with a certain threshold $`S_\pi `$, and single pole terms associated with the transitions, $`NN^{}`$ .
On the other hand, with the PV Lagrangian, the similar recipe yields the correlator at the order $`m_\pi ^2`$,
$`{\displaystyle \frac{m_\pi ^2}{2}}{\displaystyle \frac{g_{\pi N}\lambda ^2}{(q^2m^2)^2}}+\mathrm{},`$ (5)
Note that in the PV case, there is no soft-pion limit as it should be. This PV correlator contains an additional residue of $`1/2`$ compared to the PS correlator. Thus, $`g_{\pi N}`$ determined from the PV coupling scheme is twice of the one from the PS coupling scheme.
In the construction of this sum rule, the pion mass, $`m_\pi ^2`$, will be taken out as an overall factor. The rest correlator will be used to construct the sum rule. Then, a consistent treatment should be made also in the OPE side. Namely, from the Gell-Mann$``$Oakes$``$Renner relation,
$$2m_q\overline{q}q=m_\pi ^2f_\pi ^2,$$
(6)
the vanishing limit of the pion mass, $`m_\pi ^20`$, is consistent with the chiral limit, $`m_q0`$. Therefore, for the sum rule with $`m_\pi ^2`$ taken out as an overall factor, the quark-mass term should be kept in the OPE side. Clearly, this aspect has been overlooked in our previous calculations and needs to be implemented.
To construct the OPE side, we consider the correlation function with a charged pion,
$`\mathrm{\Pi }(q,p)=i{\displaystyle d^4xe^{iqx}0|T[J_p(x)\overline{J}_n(0)]|\pi ^+(p)}.`$ (7)
Here $`J_p`$ is the proton interpolating field suggested by Ioffe ,
$`J_p=ϵ_{abc}[u_a^TC\gamma _\mu u_b]\gamma _5\gamma ^\mu d_c,`$ (8)
and the neutron interpolating field $`J_n`$ is obtained by replacing $`(u,d)(d,u)`$. In the OPE, we keep the quark-antiquark component of the pion wave function and use the vacuum saturation hypothesis to factor out higher dimensional operators in terms of the pion wave function and the vacuum expectation value.
For the sum rule with the $`i\gamma _5`$ structure, we replace the quark-antiquark component of the pion wave function as follows,
$`0|u_a^\alpha (x)\overline{d}_a^{}^\beta (0)|\pi ^+(p){\displaystyle \frac{\delta _{aa^{}}}{12}}(i\gamma _5)^{\alpha \beta }0|\overline{d}(0)i\gamma _5u(x)|\pi ^+(p).`$ (9)
At $`p^2=m_\pi ^2`$ order, the matrix element in the left-hand side is replaced as
$`0|\overline{d}(0)i\gamma _5u(x)|\pi ^+(p)m_\pi ^2{\displaystyle \frac{\sqrt{2}\overline{q}q}{3f_\pi }},`$ (10)
where the overall normalization of the pion wave function at the second moment has been used. Another contribution at $`m_\pi ^2`$ order is obtained by moving a gluon tensor from a quark propagator into the quark-antiquark component. This constitutes the three particle wave function whose overall normalization is relatively well-known. From Ref. ,
$`0|G_{\mu \nu }^A(0)u_a^\alpha (x)\overline{d}_b^\beta (0)|\pi ^+(p)={\displaystyle \frac{if_{3\pi }}{32}}m_\pi ^2t_{ab}^A(\gamma _5\sigma _{\mu \nu })^{\alpha \beta },`$ (11)
where Its value is uncertain by an error $`\pm 0.0005`$ GeV<sup>2</sup> depending on the renomalization scale . However, the contribution from Eq. (11) is small in our sum rule as we will discuss below. Thus, this error in $`f_{3\pi }`$ is negligible in our sum rule. $`f_{3\pi }=0.003`$ GeV<sup>2</sup> and the color matrices $`t^A`$ are related to the Gell-Mann matrices via $`t^A=\lambda ^A/2`$.
As we have discussed, the linear terms in quark mass should be kept in the OPE for the sum rule at $`m_\pi ^2`$ order. The quark-mass dependent terms can be obtained by first taking the limit, $`p_\mu 0`$, in the quark-antiquark component, Eq. (9), while picking up linear terms in quark-mass from the other part of the correlator For complete quark propagator including the linear order in quark mass, see Ref. . Note that gluonic tensor used there has opposite sign of that in Ref. . This is just a matter of how one defines the covariant derivative but, in practice, this sign difference should be carefully noted.. It turns out that the condensates, $`m_q\overline{q}q`$ and $`m_q\overline{q}g_s\sigma 𝒢qm_qm_0^2\overline{q}q`$, contribute to the OPE of the $`i\gamma _5`$ structure. The Gell-Mann$``$Oakes$``$Renner relation is used to convert $`m_q\overline{q}q`$ to $`m_\pi ^2f_\pi ^2/2`$. Therefore, the quark-mass terms give additional contributions to the sum rule at $`m_\pi ^2`$ order.
Collecting all the OPE terms contributing to the $`i\gamma _5`$ structure at $`m_\pi ^2`$ order, the OPE side (after taking out the isospin factor $`\sqrt{2}`$ as well as $`m_\pi ^2`$ as overall factors) takes the form
$`ln(q^2)\left[{\displaystyle \frac{\overline{q}q}{12\pi ^2f_\pi }}+{\displaystyle \frac{3f_{3\pi }}{4\sqrt{2}\pi ^2}}\right]+f_\pi \overline{q}q{\displaystyle \frac{1}{q^2}}+{\displaystyle \frac{1}{72f_\pi }}\overline{q}q{\displaystyle \frac{\alpha _s}{\pi }}𝒢^2{\displaystyle \frac{1}{q^4}}{\displaystyle \frac{1}{3}}m_0^2f_\pi \overline{q}q{\displaystyle \frac{1}{q^4}}`$ (12)
Note, we use the pion decay constant $`f_\pi =0.093`$ GeV here. The second and fourth terms come from the quark-mass dependent terms. It turns out that these are important in stabilizing the sum rule, justifying the inclusion of quark-mass terms in the OPE. The second term in the bracket comes from gluonic contribution combined with the quark-antiquark component, Eq. (11). Its contribution is about 4 times smaller than the first term in the bracket. Except for this term, all others contain the quark condensate. This feature provides very stable results when this sum rule combined with the nucleon chiral-odd sum rule.
We now match the OPE with its pseudoscalar phenomenological part, Eq.(4). To saturate the correlator with the low-lying resonance, we perform Borel transformation and obtain,
$`g_{\pi N}\lambda ^2e^{m^2/M^2}[1+AM^2]=`$ (13)
$`M^4E_0(x_\pi )\left[{\displaystyle \frac{\overline{q}q}{12\pi ^2f_\pi }}+{\displaystyle \frac{3f_{3\pi }}{4\sqrt{2}\pi ^2}}\right]f_\pi \overline{q}qM^2+{\displaystyle \frac{1}{72f_\pi }}\overline{q}q{\displaystyle \frac{\alpha _s}{\pi }}𝒢^2{\displaystyle \frac{1}{3}}m_0^2f_\pi \overline{q}q.`$ (14)
The contribution from $`NN^{}`$ is denoted by the unknown constant, $`A`$. The continuum contribution is included in the factor, $`E_n(x_\pi S_\pi /M^2)=1(1+x_\pi +\mathrm{}+x_\pi ^n/n!)e^{x_\pi }`$ where $`S_\pi `$ is the continuum threshold, which we take 2.07 GeV<sup>2</sup> corresponding to the Roper resonance. In our analysis, we take standard values for the QCD parameters,
$`\overline{q}q`$ $`=`$ $`(0.23\mathrm{GeV})^3;{\displaystyle \frac{\alpha _s}{\pi }}𝒢^2=(0.33\mathrm{GeV})^4;m_0^2=0.8\mathrm{GeV}^2.`$ (15)
In figure 1, we plot $`g_{\pi N}\lambda ^2[1+AM^2]`$ as a function of the Borel mass $`M^2`$. To see the sensitivity to the continuum threshold, we also plot the curve with $`S_\pi =2.57`$ GeV<sup>2</sup>, which is very close to the one with $`S_\pi =2.07`$ GeV<sup>2</sup>. The two curves differs only by 2% at $`M^2=1`$ GeV<sup>2</sup>, indicating that our sum rule is insensitive to the continuum threshold. The highest dimensional term in the OPE contributes appreciably for $`M^20.6`$ GeV<sup>2</sup>, more than 20 % of the total OPE. Thus, the truncated OPE therefore will be reliable for $`M^20.6`$ GeV<sup>2</sup>. The slope of the curve for $`M^20.6`$ GeV<sup>2</sup> is small, indicating that the unknown single pole term denoted by $`A`$ is small.
To eliminate the dependence on the unknown strength $`\lambda `$ in our sum rule, we divide Eq. (14) by the nucleon chiral-odd sum rule and obtain,
$`{\displaystyle \frac{g_{\pi N}}{m}}[1+AM^2]=`$ (16)
$`\left\{M^4E_0(x_\pi )\left[{\displaystyle \frac{1}{3f_\pi }}+{\displaystyle \frac{3f_{3\pi }}{\overline{q}q\sqrt{2}}}\right]+{\displaystyle \frac{8\pi ^2f_\pi }{3}}M^2{\displaystyle \frac{\pi ^2}{18f_\pi }}{\displaystyle \frac{\alpha _s}{\pi }}𝒢^2+{\displaystyle \frac{4\pi ^2}{3}}m_0^2f_\pi \right\}`$ (17)
$`\times \left\{M^4E_1(x_N){\displaystyle \frac{\pi ^2}{6}}{\displaystyle \frac{\alpha _s}{\pi }}𝒢^2\right\}^1,`$ (18)
where $`x_N=S_N/M^2`$ with $`S_N`$ being the continuum threshold for the nucleon sum rule. Note that the dependence on the quark condensate has been mostly canceled in the ratio, leaving a slight dependence in the term $`f_{3\pi }`$. Additional source of the uncertainty associated with the gluon condensate is also very small as it is canceled in the ratio. One important uncertainty comes from the parameter $`m_0^2`$, which however appears only in the highest dimensional OPE. Therefore, its contribution will be suppressed in the Borel window chosen. The error from QCD parameters is estimated numerically and it is about $`\pm 1.2`$ in determining $`g_{\pi N}`$. For the continuum threshold in the nucleon sum rule, we take $`S_N=S_\pi `$. This choice is made because at the chiral limit the $`i\gamma _5`$ sum rule is equivalent to the nucleon chiral-odd sum rule; these two are related by chiral rotation . This equivalence provides the Goldberger-Treiman relation with $`g_A=1`$ . This choice for the continuum is also supported from modeling higher resonance contributions to the correlator based on effective models . We determine $`g_{\pi N}`$ and $`A`$ by fitting the RHS with a straight line within the appropriately chosen Borel window. The dependence on the Borel mass is mainly driven by the nucleon sum rule. The maximum Borel mass is determined by restricting the the continuum contribution from the nucleon sum rule while the minimum Borel mass is obtained by restricting the highest OPE term from the $`\pi NN`$ sum rule. These gives the common window of the two sum rules, $`0.65M^21.24`$. By fitting the RHS with a straight line within this window, we obtain $`g_{\pi N}=13.3\pm 1.2`$, where the quoted error comes from the QCD parameters. This is remarkably close to its empirical value of 13.4.
In getting this result, it is essential to go beyond the chiral limit. Since the empirical $`g_{\pi N}`$ should include the chiral corrections, it is indeed natural to go beyond the chiral limit in the determination of $`g_{\pi N}`$. One important observation made in this work is that the quark-mass dependent terms shouldn’t be treated separately from the sum rule proportional to $`m_\pi ^2`$ as they constitute the same chiral order via the Gell-Mann$``$Oakes$``$Renner relation. The quark-mass terms are found to be very important in stabilizing the sum rule. Our findings, remarkable agreement with the empirical value and the insensitiveness on the QCD parameters, may provide a solid ground in constructing sum rules for other meson-baryon couplings. The predictive power of QCD sum rules can be substantially enhanced. One application of our sum rule to the $`\eta NN`$ coupling is in progress . Furthermore, our result provides a QCD constraint for the type of the pion-nucleon coupling. The value of $`g_{\pi N}`$ quoted above is based on the PS coupling scheme. With the PV coupling scheme, we would have obtained the value twice of the quoted above. \[See Eq.(5).\] Any error in our approach can not produce the value of $`g_{\pi N}`$ consistent with the PV coupling scheme. Therefore our work suggests that the PS scheme is preferable for the pion-nucleon coupling from the QCD point of view.
In summary, we have developed a QCD sum rule for $`\pi NN`$ coupling beyond the chiral limit for the first time. The quark-mass dependent terms are combined to the sum rule proportional to $`m_\pi ^2`$ and they are very important in this sum rule. A remarkable agreement with the empirical value of $`g_{\pi N}`$ was obtained with very small errors. This sum rule provides the first QCD constraint for the type of the pion-nucleon coupling, in favor of the pseudoscalar coupling.
###### Acknowledgements.
The author is indebted to T. Hatsuda who has drawn the attention to this problem. The author also thanks M. Oka, and S. H. Lee for useful discussions. This work is supported by Research Fellowships of the Japan Society for the Promotion of Science.
|
no-problem/9904/quant-ph9904033.html
|
ar5iv
|
text
|
# Squashed States of Light: Theory and Applications to Quantum Spectroscopy Running Title: Quantum Spectroscopy with Squashed Light
## 1 Introduction
Squeezed states of light are nonclassical . The foremost consequence of this is that they can produce a homodyne photocurrent having a noise level below the shot-noise limit. The shot-noise limit is what is predicted by a theory in which the light is classical, with no noise, but the process of photo-electron emission is treated quantum-mechanically.
There is, however, a simple way to produce a sub-shot-noise photocurrent without squeezed light: modulating the light incident on the photodetector by a current originating from that very detector. This was first observed around the same time as the first incontestable observation of squeezing .
The sub-shot noise spectrum of an in-loop photocurrent is not regarded as evidence for squeezing for a number of reasons. First, the two-time commutation relations for an in-loop field are not those of a free field . This means that it is possible to reduce the fluctuations in the measured (amplitude) quadrature without increasing those in the other (phase) quadrature. Second, attempts to remove some of the supposedly low-noise light by a beam splitter yields only above shot-noise light, as verified experimentally .
Because of these differences, the presence of a sub-shot noise photocurrent spectrum for in-loop light has by and large been omitted from discussions of squeezing . Nevertheless, it has been argued that the in-loop field can justifiably be called “sub-Poissonian” (even if not squeezed), because a perfect quantum-non-demolition (QND) intensity meter for the in-loop light would register the same sub-shot-noise statistics as the (perfect) in-loop detector. Furthermore, it was proposed in Ref. that the apparent in-loop noise reduction could be used to improve the signal to noise ratio for a measurement of a modulation in the coupling coefficient of such a QND intensity meter.
Following on from Ref. , it has been shown that in-loop optical noise suppression may have other, more practical, applications. Buchler et al. showed that such in-loop light can suppress radiation pressure noise in a gravitational wave detector by a factor of two. Even more strikingly, I recently showed that a two-level atom coupled to the in-loop field can exhibit linewidth narrowing exactly analogous to that produced by squeezed light .
In this paper I will discuss the properties of “squashed” states of light, as the in-loop analogues of squeezed states of light are called in Ref. . Section 3 covers the general theory of squashed states of light, including states which are squashed in both quadratures and states which are simultaneously squashed and squeezed. In Section 4, I generalize the analysis of Ref. by considering the effects of these more general state of light on a two-level atom. But to begin, I review the properties of squeezed states of light in the following section.
## 2 Squeezed States
### 2.1 Single-mode squeezing
The noise in the quadratures of a single-mode light field of annihilation operator $`a`$,
$$x=a+a^{},y=ia+ia^{},$$
(1)
is limited by the Heisenberg uncertainty relation
$$V_xV_y\left|\frac{1}{2}[x,y]\right|^2=1,$$
(2)
where $`V`$ denotes variance. Squeezed states of light are states such that one of the quadrature variances, say $`V_x`$, is less than one . Clearly the other quadrature variance, say $`V_y`$, must be greater than one. If the equality is attained in the uncertainty relation then these are called minimum uncertainty squeezed states. The only other sort of minimum uncertainty state is the coherent state with $`V_x=V_y=1`$.
### 2.2 Continuum Squeezing
Single-mode squeezing was generalized early to multimode squeezing . In this work I wish to consider the limit of an infinite number of modes: the electromagnetic continuum. Considering polarized light propagating in one direction, only a single real-valued index is needed for the modes, and we take that to be the mode frequency $`\omega =k`$ (using units such that the speed of light is one). Then the continuum field operators $`b(k)`$ obey
$$[b(k),b^{}(k^{})]=\delta (kk^{}).$$
(3)
For light restricted in frequency to a relatively narrow bandwidth $`B`$ around a carrier frequency $`\mathrm{\Omega }`$ it is possible to convert from the frequency domain to the time or distance domain by defining
$$b(t)=_{\mathrm{\Omega }B}^{\mathrm{\Omega }+B}𝑑kb(k)e^{i(k\mathrm{\Omega })t}.$$
(4)
These obey
$$[b(t),b^{}(t^{})]=\delta (tt^{}),$$
(5)
where $`b^{}(t)[b(t)]^{}`$, and the $`\delta `$ function is actually a narrow function with width of order $`B^1`$. The operator $`b^{}(t)b(t)`$ can be interpreted as the photon flux operator.
Defining continuum quadrature operators
$$X(t)=b(t)+b^{}(t),Y(t)=ib(t)+ib^{}(t),$$
(6)
one obtains
$$[X(t),Y(t^{})]=2i\delta (tt^{}).$$
(7)
The singularity in the associated uncertainty relations can be avoided by quantifying the uncertainty by the spectrum
$$S^X(\omega )=\stackrel{~}{X}(\omega )X(0)_{\mathrm{ss}}\stackrel{~}{X}(\omega )_{\mathrm{ss}}X(0)_{\mathrm{ss}}.$$
(8)
Then one can derive the finite uncertainty relations
$$S^X(\omega )S^Y(\omega )1.$$
(9)
For very high frequencies the spectra always go to unity. This represents the shot-noise or vacuum noise level. However for finite frequencies it is possible to have for example $`S_X(\omega )<1`$. This indicates squeezing of the $`X`$ quadrature. The uncertainty in the conjugate quadrature would of course be increased.
The quadrature operators $`X(t),Y(t)`$ can be directly measured (one at a time) using homodyne detection . For detection of efficiency $`ϵ1`$ the photocurrent $`I_{\mathrm{hom}}^X`$ has the same statistics as (and therefore can be represented by) the operator
$$I_{\mathrm{hom}}^X(t)=\sqrt{ϵ}X(t)+\sqrt{1ϵ}\xi _X(t),$$
(10)
where $`\xi _X(t)`$ is a Gaussian white noise term. Here the normalization has been chosen so that the photocurrent spectrum
$$S_{\mathrm{hom}}^X(\omega )=ϵS^X(\omega )+(1ϵ),$$
(11)
remains equal to unity (the shot noise limit) at high frequencies, where $`S^X(\omega )=1`$.
### 2.3 An Atom in a Squeezed Bath
Now consider the situation where a two-level atom is immersed in a beam of squeezed light with annihilation operator $`b_0(t)`$. If the degree of mode matching of the squeezed light to the atom’s dipole radiation mode is $`\eta `$, then the dipole coupling Hamiltonian in the rotating-wave approximation is
$$H_0(t)=i[\sqrt{\eta }b_0(t)+\sqrt{1\eta }\nu (t)]\sigma ^{}(t)+\mathrm{H}.\mathrm{c}.$$
(12)
Here $`\sigma `$ is the atomic lowering operator and the atomic linewidth has been set to unity. The operator $`b_0(t)`$ represents the squeezed field with spectra $`S_0^X(\omega )`$ and $`S_0^Y(\omega )`$, and $`\nu (t)`$ represents the vacuum field interacting with the atom, satisfying $`\nu (t)\nu ^{}(t^{})=\delta (tt^{})`$.
Now if the quadrature spectra of the squeezed light are much broader than the atomic linewidth then one can make the white noise approximation that they are constant. For minimum-uncertainty squeezing, one has
$$S_0^X(\omega )=L=1/S_0^Y(\omega ),$$
(13)
and the atom will obey the master equation
$$\dot{\rho }=(1\eta )𝒟[\sigma ]\rho +\frac{\eta }{4L}𝒟[(L+1)\sigma (L1)\sigma ^{}]\rho ,$$
(14)
where $`𝒟[A]BA^{}BA\frac{1}{2}\{A^{}A,B\}`$ as usual. This leads to the following atomic dynamics:
$`\mathrm{Tr}[\dot{\rho }\sigma _x]`$ $`=`$ $`\gamma _x\mathrm{Tr}[\rho \sigma _x],`$ (15)
$`\mathrm{Tr}[\dot{\rho }\sigma _y]`$ $`=`$ $`\gamma _y\mathrm{Tr}[\rho \sigma _y],`$ (16)
$`\mathrm{Tr}[\dot{\rho }\sigma _z]`$ $`=`$ $`\gamma _z\mathrm{Tr}[\rho \sigma _z]C,`$ (17)
where
$`\gamma _x`$ $`=`$ $`\frac{1}{2}\left[(1\eta )+\eta L\right],`$ (18)
$`\gamma _y`$ $`=`$ $`\frac{1}{2}\left[(1\eta )+\eta L^1\right],`$ (19)
$`\gamma _z`$ $`=`$ $`\gamma _x+\gamma _y,C=1.`$ (20)
Note that for $`L<1`$ the decay rate of the $`x`$ component of the atomic dipole is reduced below the vacuum level of $`\frac{1}{2}`$, while the decay rate of the other component is increased. The reduction or increase in the decay rates are directly attributable to the reduction or increase in the fluctuations of the respective quadrature of the input continuum field. The prediction of this effect by Gardiner began the study of quantum spectroscopy (that is, the interaction of nonclassical light with matter) .
For sufficiently large $`\eta `$ this effect would be easily detectable experimentally in the fluorescence power spectrum of the atom (into the vacuum modes):
$`P(\omega )`$ $`=`$ $`{\displaystyle \frac{1\eta }{2\pi }}\stackrel{~}{\sigma }^{}(\omega )\sigma (0)_{\mathrm{ss}}`$ (21)
$`=`$ $`{\displaystyle \frac{(1\eta )(\gamma _zC)}{8\pi \gamma _z}}\left[{\displaystyle \frac{\gamma _x}{\gamma _x^2+\omega ^2}}+{\displaystyle \frac{\gamma _y}{\gamma _y^2+\omega ^2}}\right].`$ (22)
For $`L<1`$ the spectrum consists of two Lorentzians, one with a sub-natural linewidth and one with a super-natural linewidth. The overall linewidth (defined as the full-width at half-maximum height) is reduced.
This line-narrowing is only noticeable if the degree of mode matching $`\eta `$ of the squeezed light to the atom is significant, which is hard to do with a squeezed beam. One way around this is to make the atom strongly coupled to a microcavity, which can be driven by a squeezed beam. The microcavity enhances the atomic decay rate, but a squeezed input should suppress this enhancement in one quadrature. Unfortunately, experiments to date have failed to see this suppression, due to imperfections of various kinds .
## 3 Theory of Squashed States
### 3.1 Generation of Squashed States
Consider the feedback loop shown in Fig. 1. The field entering the modulator is $`b_0(t)=\frac{1}{2}[X_0(t)+iY_0(t)]`$. The quadrature operators are assumed to have independent statistics defined by the spectra $`S_0^X(\omega )`$, $`S_0^Y(\omega )`$. The modulator simply adds a coherent amplitude to this field. There are various ways of achieving this, one of which is discussed in Ref. . The field exiting is in any case given by
$$b_1(t)=\frac{1}{2}\left[X_0(t)+iY_0(t)+\chi (t)+i\upsilon (t)\right],$$
(23)
where $`\chi (t),\upsilon (t)`$ are real functions of time. This field now enters a homodyne detection device, set up so as to measure the $`X`$ quadrature. If the efficiency of the measurement is $`ϵ_X`$ then the photocurrent is given by
$$I_{\mathrm{hom}}^X(t)=\sqrt{ϵ_X}\left[X_0(t)+\chi (t)\right]+\sqrt{1ϵ_X}\xi _X(t).$$
(24)
Now, through the feedback loop, this current may determine the classical field amplitudes $`\chi ,\upsilon `$. Obviously in the case of measuring the $`X`$ quadrature, the only interesting results will come from controlling $`\chi (t)`$, and we set
$$\chi (t)=_0^{\mathrm{}}g_Xh(s)I_{\mathrm{hom}}^X(t\tau s)/\sqrt{ϵ_X}𝑑s.$$
(25)
Here $`\tau `$ is the minimum delay in the feedback loop, $`h(s)`$ is the feedback loop response function normalized to $`_0^{\mathrm{}}h(s)𝑑s=1`$ and $`g_X`$ is the round-loop gain . Taking the fourier transform of this expression and substituting into Eqs. (23) and (24) yields
$$\stackrel{~}{b}_1(\omega )=\frac{1}{2}\left[\frac{\stackrel{~}{X}_0(\omega )+\sqrt{\theta _X}g_Xe^{i\omega \tau }\stackrel{~}{h}(\omega )\stackrel{~}{\xi }_X(\omega )}{1g_Xe^{i\omega \tau }\stackrel{~}{h}(\omega )}+i\stackrel{~}{Y}_0(\omega )+i\upsilon (t)\right].$$
(26)
where
$$\theta _Xϵ_X^11.$$
(27)
The $`X`$ quadrature spectrum of this light is
$$S_1^X(\omega )=\frac{S_0^X(\omega )+\theta _Xg_X^2|\stackrel{~}{h}(\omega )|^2}{|1g_Xe^{i\omega \tau }\stackrel{~}{h}(\omega )|^2}.$$
(28)
Say we are only interested in frequencies well inside the bandwidth of $`\stackrel{~}{h}(\omega )`$, much less than $`\tau ^1`$, and much less than the bandwidth of $`S_0^X(\omega )`$. Then we can replace $`e^{i\omega \tau }\stackrel{~}{h}(\omega )`$ by unity and $`S_0^X(\omega )`$ by a constant $`L`$. This gives
$$S_1^X=\frac{L+g_X^2\theta _X}{(1g_X)^2}\frac{L}{1+L/\theta _X},$$
(29)
where the minimum is achieved for negative feedback $`g_X=L/\theta _X`$.
Evidently this minimum is less than $`L`$. This means that even starting with shot-noise limited light ($`L=1`$) it is possible to produce sub-shot-noise light. However, it is important to note that this is not squeezed light in the ordinary sense. For example, it is impossible to remove any of the squeezed light by putting a beam splitter in the path of the in-loop beam. Under the above conditions the resulting out-of-loop beam actually has a noise level above the shot noise . Nevertheless, the fluctuations of the in-loop light do produce genuine physical effects in other circumstances, as investigated in Sec. 4.
### 3.2 Violation of the Uncertainty Relations
A curious point in the apparent squeezing of the $`X`$ quadrature is that the feedback has no effect on the $`Y`$ quadrature of $`b_1`$. From Eq. (26), $`S_1^Y=L^1`$, assuming a minimum uncertainty input $`b_0`$ and $`\upsilon =0`$. Thus the uncertainty relation (9) is violated for this in-loop light. For this reason, the in-loop light exhibiting sub-shot-noise fluctuations has been called “squashed light” . By feedback, the noise in one quadrature can be squashed (reduced), but there is no “squeezing” of phase-space area into an increased noise in the other quadrature.
The reason that the uncertainty relation (9) is violated is that the commutation relations (5) are no longer valid for time differences $`|tt^{}|>\tau `$, the minimum feedback loop delay. This is a direct consequence of the feedback loop. It is important to realize that the parts of the in-loop field separated in time by greater than $`\tau `$ never actually exist together. That is because the propagation time from the modulator to the detector is necessarily less than $`\tau `$. Thus the fundamental commutation relations between parts of the field at different points in space at the same time are never violated.
For freely propagating fields there is no real distinction between space and time separations, but for an in-loop field it is a crucial distinction. The temporal anticorrelations in the in-loop squeezed light only exist for time separations greater than $`\tau `$, and hence greater than the time for which any part of the in-loop light exists. There is never any anticorrelation between parts of the in-loop field in existence at any given time. By contrast, conventional squeezed light can propagate for an arbitrarily long time before detection, so the anticorrelations are between parts of the field which can exist simultaneously (even if they may not actually do so in a given experiment).
### 3.3 Simultaneous Squashing in Both Quadratures
It is interesting now to consider feedback in both the $`X`$ and $`Y`$ quadratures. Obviously one cannot simultaneously measure both of these quadratures with unit efficiency, but one can measure $`Y`$ with efficiency $`ϵ_Y1ϵ_X`$. Carrying through the same sort of analysis as above shows that the $`Y`$ quadrature spectrum can be simultaneously reduced to
$$S_1^Y=\frac{L^1}{1+L^1/\theta _Y}.$$
(30)
For the special case of $`L=1`$ (a vacuum input), one finds
$$S_1^Y+S_1^X=2ϵ_Xϵ_Y1.$$
(31)
For feedback based on heterodyne detection (which is equivalent to homodyne detection on both quadratures with equal efficiency), one can have $`S_1^Y=S_1^X=1\frac{1}{2}ϵ`$, which goes to one half in the limit of perfect detectors.
### 3.4 Simultaneous Squeezing and Squashing
In the general case of $`L1`$, the sum of the quadrature spectra need not even be greater than one. Rather, for fixed $`L<1`$ and fixed $`ϵ=ϵ_X+ϵ_Y`$ one finds, for $`ϵ_X=0`$ and $`ϵ_Y=ϵ`$
$$S_1^Y+S_1^X=L+\frac{ϵ^11}{L(ϵ^11)+1}0,$$
(32)
where the limit of zero noise in both quadratures is approached for $`L0`$ and $`ϵ1`$. For example, with experimentally realisable parameters of 6dB squeezing and detection efficiency $`ϵ=0.95`$ , one could obtain $`S_1^X=0.25`$ and $`S_1^Y=0.05`$, giving $`S_1^Y+S_1^X=0.30`$, compared to the limit of $`2`$ implied by the uncertainty relation (9).
## 4 Application to Quantum Spectroscopy
Since the quadrature spectra calculated above apply to an in-loop field, which cannot be extracted using a beam splitter, it might seem that they have no physical significance. However, this is not the case. As I showed recently , placing a two-level atom in an squashed bath leads to linewidth narrowing of one atomic dipole quadrature, entirely analogous to that produced by a squeezed bath. The master equation is not identical, however, because the non-squashed quadrature is still shot-noise limited, so there is no line broadening of the other atomic dipole quadrature.
In this work I generalize the results of Ref. to include light which is simultaneously squashed in both quadratures, or simultaneously squeezed and squashed. If the atom is coupled to the in-loop beam $`b_1(t)`$ with mode-matching $`\eta `$ then the atomic Hamiltonian is
$$H(t)=i[\sqrt{\eta }b_1(t)+\sqrt{1\eta }\nu (t)]\sigma ^{}(t)+\mathrm{H}.\mathrm{c}.$$
(33)
Following the methods of Ref. , the expression for $`b_1`$ is modified from (26) by the addition of the atom’s radiated field in the direction of the beam, $`\sqrt{\eta }\stackrel{~}{\sigma _x}(\omega )`$, to the input operator $`\stackrel{~}{X}_0(\omega )`$. Thus the total Hamiltonian can be written
$$H(t)=H_{\mathrm{fb}}(t)+H_0(t),$$
(34)
where $`H_0(t)`$ is as given in Eq. (12) and the Hamiltonian due to the feedback is
$`H_{\mathrm{fb}}(t)`$ $`=`$ $`\lambda _X\frac{1}{2}\sigma _y(t)\left\{\sigma _x(t^{})+[X_0(t^{})+\sqrt{\theta _X}\xi _X(t^{})]\sqrt{\eta }\right\}`$ (35)
$`+\lambda _Y\frac{1}{2}\sigma _x(t)\left\{\sigma _y(t^{})+[Y_0(t^{})+\sqrt{\theta _Y}\xi _Y(t^{})]\sqrt{\eta }\right\}.`$
The feedback parameters are defined as
$$\lambda _X=\frac{g_X\eta }{1g_X};\lambda _Y=\frac{g_Y\eta }{1g_Y}.$$
(36)
for the feedback of the homodyne current $`I_{\mathrm{hom}}^Y`$ are defined analogously to those from the feedback of $`I_{\mathrm{hom}}^X`$.
In Eq. (35) the limit of broad-band feedback has been taken, with $`\stackrel{~}{h}(\omega )e^{i\omega \tau }`$ in Eq. (26) set to unity. This Markov approximation is justified provided the bandwidth of the feedback is very large compared to the characteristic rates of response of the system . In the present context the rate of atomic decay is unity so we require, for instance, $`\tau 1`$. For a typical electro-optic feedback loop with a bandwidth in the MHz range, the atom would have to have be metastable to satisfy this inequality. The precise requirements for the validity of the Markov approximation will be investigated in a future publication. Of course even in the broad-band limit the feedback from the measurement of a particular part of the field must act after that part of the field has interacted with the atom. This is the reason for the use of the time argument $`t^{}`$ rather than $`t`$ in Eq. (35).
Now to describe the evolution generated by the total Hamiltonian (33), the theory of homodyne detection and feedback in the presence of white noise is required. This was first detailed in Ref. , generalizing the earlier work in Refs. and . The basic equation is
$$d\rho (t)=\mathrm{exp}[iH_{\mathrm{fb}}(t)dt]\mathrm{exp}[iH_0(t)dt]\rho (t)\mathrm{exp}[iH_0(t)dt]\mathrm{exp}[iH_{\mathrm{fb}}(t)dt]\rho (t).$$
(37)
Here the ordering of the unitary operators has been chosen such that the time delay of the feedback has been taken into account and one can replace $`t^{}`$ in Eq. (35) by $`t`$.
In Eq. (37) the expectation value indicates an average over the bath operators $`b_0(t)`$ and $`\nu (t)`$, and the detector noise terms $`\xi _X(t),\xi _Y(t)`$. This is effected by making replacements such as
$$[X_0(t)dt]^2Ldt;\nu (t)dt\nu ^{}(t)dtdt.$$
(38)
The result is
$`\dot{\rho }`$ $`=`$ $`(1\eta )𝒟[\sigma ]\rho +{\displaystyle \frac{\eta }{4L}}𝒟[(L+1)\sigma (L1)\sigma ^{}]\rho `$ (39)
$`i\lambda _X[\frac{1}{2}\sigma _y,\frac{1}{2}\{(L+1)\sigma (L1)\sigma ^{}\}\rho +\rho \frac{1}{2}\{(L+1)\sigma ^{}(L1)\sigma \}]`$
$`+i\lambda _Y[\frac{1}{2}\sigma _x,i\frac{1}{2}\{(L^1+1)\sigma +(L^11)\sigma ^{}\}\rho +i\rho \frac{1}{2}\{(L^1+1)\sigma ^{}(L^11)\sigma \}]`$
$`+{\displaystyle \frac{\lambda _X^2(L+\theta _X)}{\eta }}𝒟\left[{\displaystyle \frac{\sigma _y}{2}}\right]\rho +{\displaystyle \frac{\lambda _Y^2(L^1+\theta _Y)}{\eta }}𝒟\left[{\displaystyle \frac{\sigma _x}{2}}\right]\rho .`$
This again produces the atomic dynamics of Eqs. (15)–(17), but with
$`\gamma _x`$ $`=`$ $`\frac{1}{2}\left[(1\eta )+\eta L(1+\lambda _X/\eta )^2+\lambda _X^2\theta _X/\eta \right]=\frac{1}{2}\left[(1\eta )+\eta S_X\right],`$ (40)
$`\gamma _y`$ $`=`$ $`\frac{1}{2}\left[(1\eta )+\eta L^1(1+\lambda _Y/\eta )^2+\lambda _Y^2\theta _Y/\eta \right]=\frac{1}{2}\left[(1\eta )+\eta S_Y\right],`$ (41)
$`\gamma _z`$ $`=`$ $`\gamma _x+\gamma _y,C=1+\lambda _X+\lambda _Y.`$ (42)
In the above equations, the introduction of the spectrum $`S^X`$ is based on Eq. (29), with the identification $`\lambda _X=\eta g_X/(1g_X)`$, and similarly for $`Y`$. Note that the expressions for $`\gamma _x`$ and $`\gamma _y`$ depend upon the quadrature spectra (in the absence of the atom) in precisely the same way as those for pure squeezed (not squashed) light in Eqs. (18) and (19). This suggests that the natural explanation for the change in the decay rates is again the reduced fluctuations of the input light. It seems that squashed fluctuations are much the same as squeezed fluctuations as far as the atom is concerned. The one difference is that the constant $`C`$ in the equation of motion for $`\sigma _z`$ is also altered by the feedback.
Consider now the case as in Sec. 3.4 where $`L<1`$, $`ϵ_X=0`$ and $`ϵ_Y=ϵ`$. Then choosing $`\lambda _Y=\eta /(1+L\theta )`$, so as to minimize the in-loop $`Y`$ quadrature spectrum, gives
$`\gamma _x`$ $`=`$ $`\frac{1}{2}\left[(1\eta )+\eta L\right],`$ (43)
$`\gamma _y`$ $`=`$ $`\frac{1}{2}\left[(1\eta )+\eta \theta /(1+L\theta )\right],`$ (44)
$`\gamma _z`$ $`=`$ $`\gamma _x+\gamma _y,C=1\eta /(1+\theta L),`$ (45)
where $`\theta =ϵ^11`$ as usual. Choosing $`ϵ=0.95`$ and $`L0.25`$, as before, gives $`\gamma _z10.85\eta `$. That is, in the limit of $`\eta 1`$, the rate of decay of the atomic population would be slowed by $`85\%`$. In the ideal limit of $`L0`$, and $`ϵ,\eta 1`$, the atom would be frozen in its initial state and would not decay at all.
## 5 Conclusion
In this work I have presented for the first time the theory for a new class of in-loop light, namely light which may be both squashed (in either or both quadratures) and squeezed. Squeezing here refers to conventional quantum noise reduction, whereas squashing refers to the noise reduction produced by the feedback loop. Even without a squeezed input it is possible to reduce the noise in both quadratures of the in-loop field below the shot-noise limit. With a squeezed input it is possible, in principle, to reduce the noise in both quadratures to zero (by squeezing one and squashing the other).
I next derived the effect of this arbitrarily squeezed and squashed light on an in-loop atom. The calculated in-loop spectra precisely reflect the noise to which the atom responds, provided the bandwidth of the squeezing and the bandwidth of the feedback are much greater than the atomic linewidth. As the quantum fluctuations seen by the atom are reduced, the decay rates for the quadratures of the atom’s dipole are reduced. In the limit that the atom is coupled only to in-loop light which is perfectly squeezed in one quadrature and perfectly squashed in the other, the atomic decay rates vanish and the atom’s dynamics are frozen.
As discussed above, the main experimental difficulty with seeing squeezing-induced line-narrowing is related to efficiently coupling the squeezed light onto the atom. Using squashed light rather than squeezed light would not overcome this difficulty. However, highly squashed light should be easier to generate than highly squeezed light, because it is limited only by the homodyne detector efficiency. Also, it can be produced at any frequency for which a coherent source and the appropriate electro-optic equipment is available. These factors, plus the intriguing possibility of observing simultaneous linewidth narrowing on both atomic quadratures, suggest an important role for squashed light in experimental quantum spectroscopy.
|
no-problem/9904/cond-mat9904179.html
|
ar5iv
|
text
|
# One-dimensional transport in bundles of single-walled carbon nanotubes
## Abstract
We report measurements of the temperature and gate voltage dependence for individual bundles (ropes) of single-walled nanotubes. When the conductance is less than about $`e^2/h`$ at room temperature, it is found to decrease as an approximate power law of temperature down to the region where Coulomb blockade sets in. The power-law exponents are consistent with those expected for electron tunneling into a Luttinger liquid. When the conductance is greater than $`e^2/h`$ at room temperature, it changes much more slowly at high temperatures, but eventually develops very large fluctuations as a function of gate voltage when sufficiently cold. We discuss the interpretation of these results in terms of transport through a Luttinger liquid.
The strength and extended length of single-walled carbon nanotubes makes it quite straightforward to attach metallic electrodes to them. This has enabled several recent studies of the transport properties of individual tubes and ropes (ordered bundles of tubes) ref1 ; ref2 ; ref3 ; ref4 ; ref5 . In most of these studies the conductance at low temperature $`T`$ is found to be dominated by Coulomb blockade (CB). Here we also include measurements on ropes with high conductance and weak $`T`$ dependences which do not show CB ref11 . We analyse the characteristics of all our devices in the light of predictions that electrons in nanotubes should form Luttinger liquids.
Each device consists of an individual nanotube rope ref6 , containing between 1 and $`20`$ tubes lying on a thermally grown SiO<sub>2</sub> surface and contacted with gold electrodes patterned by electron beam lithography. The electrode separation is 0.2 or 0.5 $`\mu `$m, and the metallically doped silicon substrate beneath the 0.3- or 1.0-$`\mu `$m thick SiO<sub>2</sub> is used as a gate electrode. We have two varieties of devices: ‘end-contacted’, where the electrode metal is deposited on top of the rope; and ‘bulk-contacted’, where the rope is deposited on top of prefabricated electrodes. An atomic force microscope (AFM) image of a typical device is inset to Fig. 1. In all the measurements reported here, the two-terminal dc current-voltage ($`I`$-$`V`$) characteristics were measured in a cryostat with the device bathed in helium.
Although transistor-like behavior is sometimes seen ref4 , we consider here only devices whose conductance $`G`$ is approximately independent of gate voltage $`V_\mathrm{g}`$ at room temperature. Fig. 1 illustrates the behavior of a typical (end-contacted) device as it is cooled down. As $`T`$ decreases, $`G`$ drops steadily, and below $`T50\mathrm{K}`$, Coulomb blockade (CB) peaks develop in $`G`$ vs $`V_\mathrm{g}`$ ref1 ; ref2 . The right inset is a grey scale plot of $`\mathrm{d}I/\mathrm{d}V`$ vs $`V_\mathrm{g}`$ and $`V`$ at 4.2 K. A standard analysis of this bias spectroscopy plot ref2 ; ref8 yields $`C_\mathrm{g}/C0.15`$ and a charging energy $`U=e^2/C10\mathrm{meV}`$, where $`C_\mathrm{g}`$ is the capacitance to the gate and $`C`$ is the total capacitance. The excitation spectrum for each charge state is resolved here at 4.2 K, allowing us to estimate a mean level spacing $`\mathrm{\Delta }2`$ meV. From the regularity of this plot we infer that the conductance is largely determined by a single quantum dot. We consider in this paper only devices which show such regularity in their spectroscopy plots. The values of $`U`$ and $`\mathrm{\Delta }`$ are consistent with charging a rope segment of length $`L`$ roughly equal to the contact separation (in this case $`L`$ = 0.5 $`\mu `$m). This is usual for end-contacted ropes. For bulk-contacted devices on the other hand, $`U`$ is smaller and corresponds to charging the full length of the rope. This implies that in both cases the electrode-tube interfaces act as tunnel contacts to the rope which form the dot. Our terminology was chosen to reflect this: end-contacted means that the current flows from the metal into the end of the active rope segment, and bulk-contacted means it flows into the bulk.
In Fig. 2 the average conductance $`G_{\mathrm{av}}`$ is plotted against $`T`$ on a log-log scale for several end-contacted (Fig. 2a) and bulk-contacted (Fig. 2b) ropes. In all cases, $`G_{\mathrm{av}}`$ decreases monotonically as $`T`$ decreases, irrespective of the room temperature conductance $`G_{\mathrm{RT}}`$, which varies greatly. That said, we can distinguish the following two categories of behavior. In the first, which contains most devices, $`G_{\mathrm{RT}}<e^2/h`$, and the conductance decreases by up to an order of magnitude before it is Coulomb blockaded at low $`T`$. For these devices we do not plot $`G`$ at the lowest temperatures, where CB dominates, and at higher $`T`$ we multiply $`G`$ by a correction factor to compensate for the effect of classical CB ref3 . Note that this factor is between one and two, and has no major qualitative effect on any of the data. In the second category, which contains only end-contacted devices, $`G_{\mathrm{RT}}>e^2/h`$, the dependence on $`T`$ is much weaker, and there is no Coulomb blockade at 4.2 K.
We begin by discussing the results (open symbols in Fig. 2) for devices in the first category, having $`G_{\mathrm{RT}}<e^2/h`$. Note that the $`T`$ dependence for all bulk-contacted ropes in Fig. 2b is very similar, and closely resembles a power law, $`GT^{\alpha _{\mathrm{bulk}}}`$, with $`\alpha _{\mathrm{bulk}}0.37`$. For the end-contacted ropes in Fig. 2a the $`T`$ dependences are steeper, and (at least below $`100`$ K) are roughly described by another power law, $`GT^{\alpha _{\mathrm{end}}}`$ with $`\alpha _{\mathrm{end}}0.7`$. We have previously argued that these results are consistent with the existence of Luttinger liquids (LL) in nanotubes ref3 ; ref9 ; ref10 . Tunneling from a metal contact into an LL is expected to be suppressed at low energies, resulting in power laws for the conductance, $`GT^\alpha `$ (for $`eV<k_\mathrm{B}T`$) and $`\mathrm{d}I/\mathrm{d}VV^\alpha `$ (for $`eV>k_\mathrm{B}T`$). Given a Luttinger parameter of $`g0.28`$ for a metallic nanotube ref3 ; ref9 ; ref10 , the exponent is predicted to be $`\alpha =(g^11)/40.65`$ for tunneling into the end of the tube, and $`\alpha =(g^1+g2)/80.24`$ for tunneling into the bulk ref3 . These numbers are in fair agreement with the measured quantities $`\alpha _{\mathrm{end}}`$ and $`\alpha _{\mathrm{bulk}}`$ above. Moreover, the prediction that the same exponent $`\alpha `$ should be seen in the bias as in the $`T`$ dependence is also borne out. This is illustrated in the inset to Fig. 2b, where at high bias the $`\mathrm{d}I/\mathrm{d}V`$ vs $`V`$ traces for a bulk-contacted device converge parallel to a line (on this log-log plot) of slope $`\alpha _{\mathrm{bulk}}=0.37`$.
We come now to the second category of devices, having $`G_{\mathrm{RT}}>e^2/h`$. Fig. 3 shows the characteristics of a device with $`G_{\mathrm{RT}}3e^2/h`$, (diamonds in Fig. 2a), at a series of $`T`$ ref11 . In strong contrast to Fig. 1, there is almost no $`T`$ dependence above 100 K. At lower $`T`$, large, reproducible fluctuations start to develop in $`G`$ vs $`V_\mathrm{g}`$. In Figs. 3b and 3c we plot the mean conductance $`G_{\mathrm{av}}`$ and the standard deviation $`\sigma _G`$ of these fluctuations. The bias spectroscopy plot inset to Fig. 3a exhibits a well behaved, symmetric cross structure. This implies that the conductance is dominated by transmission through a single object, symmetrically coupled to the two contacts. The fluctuations are almost unaffected by a magnetic field of 7 T. This is consistent with their arising from a nanotube rope, whose thickness is much smaller than the magnetic length (24 nm) at this field.
Nevertheless, these fluctuations are quantitatively and qualitatively different from the CB oscillations in Fig. 1. We notice that $`G_{\mathrm{av}}`$ remains above $`2e^2/h`$, while $`G`$ fluctuates by as much as $`2e^2/h`$, and even occasionally peaks above $`G_{\mathrm{RT}}`$. These facts are at odds with standard Coulomb blockade. Rather, we suggest that high-transparency contacts exist between the gold and one or more metallic nanotubes within the rope. Apart from the high conductance, further evidence for this comes from the very small value of $`C_\mathrm{g}/C0.01`$ deduced from the spectroscopy plot, which means that the rope is much more weakly sensitive to $`V_\mathrm{g}`$ than are similar devices that behave as quantum dots. This is consistent with very strong coupling to the electrodes. It is therefore possible that the fluctuations are related to transport through the LL in a tube. Any amount of disorder is expected to completely suppress the conductance of a LL at sufficiently low $`T`$. At high $`T`$ a single impurity produces a power law suppression of $`G`$ from its ideal value $`G_0=4e^2/h`$ ref9 . In Fig. 3b the dashed line is a fit of $`G_{\mathrm{av}}`$ to the form $`G_0aT^b`$, yielding $`b=0.22`$. Although the fit appears quite good, we make no claim to its validity, as the behavior of a disordered LL remains to be addressed theoretically.
In conclusion, when the conductance of a nanotube rope device is dominated by tunneling from the contacts, power law dependences on temperature and bias are observed which are consistent with the predicted existence of Luttinger liquids in nanotubes. Further, we find evidence that one can obtain highly transparent contacts between gold and nanotubes, which should soon enable experimental investigations of the intrinsic conducting properties of such a Luttinger liquid.
We would like to thank Poul-Erik Lindelof, Jia Lu, Reinhold Egger, Anders Kristensen and Leon Balents, for help and discussions.
|
no-problem/9904/cond-mat9904214.html
|
ar5iv
|
text
|
# Frictional Drag Between Coupled 2D Hole Gases in GaAs/AlGaAs Heterostructures
## Abstract
We report on the first measurements of the drag effect between coupled 2D-hole gases. We investigate the coupling by changing the carrier densities in the quantum wells, the widths of the barriers between the gases and the perpendicular magnetic field. From the data we are able to attribute the frictional drag to phonon coupling, because the non-parabolicity allows to tune the Fermi wavevector and the Fermi velocity separately and, thereby, to distinguish between phonon- and plasmon-dominated coupling.
The coupling mechanisms between two closely spaced two dimensional (2d) charge systems in semiconductors have found much interest recently. Even if tunneling between the layers can be neglected, the layers are coupled by the electron-electron interactions which leads to a transfer of momenta and a frictional drag between them. The drag force can be measured by passing a drive current $`I_{drive}`$ through one layer and measuring the resulting voltage drop $`V_{drag}`$ in the other one . The coupling strength is usually stated as the transresistivity $`\rho _T=(W/L)(V_{drag}/I_{drive})`$ where $`(W/L)`$ is the width to length ratio of the sample. The transresistivity has been derived theoretically based on RPA and is found to depend on the imaginary parts of the susceptibilities of the two layers $`\chi _1,\chi _2`$, on the interlayer interaction $`V_{12}`$ and on the dielectric constant $`ϵ_{12}`$ of the combined system :
$$\rho _T_0^{\mathrm{}}𝑑qq^3\frac{d\omega }{2\pi }|\frac{V_{12}(q,\omega )}{ϵ_{12}}|^2\frac{Im\chi _1(q,\omega )Im\chi _2(q,\omega )}{sinh^2(\mathrm{}\omega /2k_BT)},$$
(1)
Two-dimensional electron gases (2DEGs) in GaAs/AlGaAs heterostructures in zero magnetic field have been found to couple via Coulomb interaction , via the excitation of coupled plasmons , and via the exchange of phonons with $`q2k_F`$ . At zero magnetic fields the Coulomb coupling is weak compared to the other two mechanisms. At temperatures exceeding about $`0.2T_F`$ coupled plasmon modes are excited. This is most efficient if the Fermi velocities $`v_F`$ of the two layers coincide. The phonon coupling, on the other hand, is maximum if the $`k_F`$ values in the two layers are identical. Very recently, it has been suggested that coupled electron-phonon modes may exist under certain conditions leading to a vanishing of $`ϵ_{12}`$ around $`q=2k_F`$ . No evidence for the existence of these coupled modes has been reported yet.
No experimental work has been published on the frictional drag between two 2d hole gases (2DHGs) and only preliminary data exist on mixed (2DEG/2DHG) systems . In these systems the phonon coupling should be larger because of the larger effective hole masses. On the other hand $`T_F`$ is also smaller for the same reason and it is not clear if phonon or plasmon coupling should be expected to dominate at low temperatures. However, hole systems offer a unique possibility to discriminate between the different mechanisms because their energy dispersion curves are nonparabolic and, moreover, this nonparabolicity can be tuned by varying the shape of the quantum wells by external fields or by doping. Therefore, these systems offer the possibility to tune both $`k_F`$ and $`v_F`$ and to achieve coincidences of either of these quantitites at different densities in the two layers. The study of the drag between electron and hole gases will furthermore give information about the (in)congruences of the Fermi surfaces which limits the possibility to observe superfluidity in coupled 2DEG/2DHG systems by the reduction of the phase space for Cooper-pair-like scattering.
In this Letter we report the first, to our knowledge, measurements of the frictional drag between two 2DHGs as well as more detailed results on coupled 2DEG/2DHG systems. In both cases, the frictional drag is measured as function of the charge densities in the two layers. We also present data on samples having widely varying barrier thicknesses and we will discuss the effect of temperature and of magnetic field. We find a very asymmetric behavior with respect to the densities which is in contrast to the previously studied purely electronic systems, and we present evidence that the coupling is well described by phonon exchange and not by plasmon interaction.
The coupled 2DHG samples are prepared in two $`20nm`$ thick quantum wells in GaAs/AlGaAs heterostructures. Remote doping was achieved using carbon with a spacer layer thickness of typically $`20nm`$. Six samples are produced with GaAs quantum wells separated by $`\mathrm{Al}_{.3}\mathrm{Ga}_{.7}\mathrm{As}`$ barriers with thicknesses varying from $`30`$ to $`190nm`$. In some samples the doping of one well is placed inside the barrier leading to a strong asymmetry between the layers. The samples are shaped as a Hall bar geometry with $`80\mu m`$ width and $`800\mu m`$ length. Ohmic contacts to both layers are made by diffusion of Au and Zn. Separate contacts to the two layers are achieved by using the standard selective depletion technique . In this case, metallic front gates and p-doped buried backgates are used. Two more gates cover the main part of the Hall bar and allow the independent variation of the carrier densities. Typical hole mobilities at $`4K`$ are between $`40.000cm^2/Vs`$ and $`80.000cm^2/Vs`$ which are reasonably good values for hole gases on (001) surfaces. The hole concentration can be varied typically from zero to about $`510^{11}\mathrm{cm}^2`$. Details of the coupled 2DHG/2DEG system are already described in . In this case the distance between the two layers is 340 nm.
The drag measurements are done by passing drive currents of $`100nA`$ at a frequency of about $`1Hz`$ through one of the layers and using lock-in technique to measure the resulting drag voltage in the other layer. The integrity of the signal is first controlled by checking that all leakage currents are unmeasurably small and cannot influence the signal. Second, the linearity between drag voltage and drive current value is confirmed, and third, the drag and drive layers are interchanged and the signals are found to be identical. Measurements are done in a standard cryostat at temperatures between 1.5 and 10 K and in magnetic fields up to 11 T.
First, we present data on the dependence of the transresistivity $`\rho _T`$ on the hole concentrations. The respective densities are determined by Shubnikov-deHaas measurements at 1.5 K . In Fig. 1 we show $`\rho _T`$ as a combined grey tone - contour plot at 2.8 K as a function of the upper and lower hole densities for a coupled 2DHG sample with $`d=140nm`$. In comparison to coupled 2DEG structures we find a rather complicated dependence of $`\rho _T`$ on the densities in the two layers. Particularly remarkable is the fact that there is no symmetry in the data if the two densities $`p_{upper}`$ and $`p_{lower}`$ are interchanged. At small densities in both layers ($`310^{11}cm^2`$) the maximal coupling strength is found along a ”ridge” (marked by heavy dots) running just below the condition of equal densities in the two layers. At higher densities, however, this ridge splits into two ones which run nearly vertically and horizontally in the figure (also marked by dots). This asymmetry with respect to the densities in the two layers must be consequence of the asymmetric doping of the two wells, because it is also seen in the other asymmetrically doped samples but not in the symetrically doped ones where we find a very broad maximum of the coupling centered around the line of equal densities.
The asymmetric doping leads to an effective electric field in the respective quantum wells which causes the splitting of the highest hole subband into two. This effect has recently been studied in detail in just one hole layer . The splitting of the subbands automatically leads to different values of $`k_F`$ and $`v_F`$ in the two bands, and $`v_F`$ is no longer proportional to $`k_F`$ because of the strong nonparabolicity of the dispersion curves. Both quantities are available from the dispersion curves which we calculate in a self-consistent Hartree approximation based on a $`4\times 4kp`$-method. Examples are shown in Fig. 2 (a) and (b). The $`k_F`$ of one branch is identical for the two layers even if their densities are out of balance. The loci of identical $`k_F`$ and of identical $`v_F`$ are plotted in Fig.2 (c) and (d), respectively. The cross marks the case of Fig.2 (a) and (b). Here we use the values obtained for the direction in the plane. In the plots of (c) and (d) the solid lines correspond to the case where hole subbands with the same quantum number are coupled while the dashed lines indicates coupling between bands with different quantum numbers. A second line of this type lies far outside of the plot range.
Comparison of these plots with the data of Fig.1 shows that only the phonon coupling between hole bands with identical quantum numbers agrees with the experimental data. In Fig. 1 the calculated locus of equal $`k_F`$ values in the equivalent bands is indicated as a solid line. If a similar analysis is made for the -direction, one finds that the loci for coinciding $`k_F`$ respectively $`v_F`$ values do not differ significantly from those shown in Fig.1 and Fig.2. We conclude that the phonon exchange is the main source of coupling between the hole layers. The fact that only identical branches of the hole dispersion curves couple with each other is most likely due to the acoustical anisotropy of the GaAs which cause that only certain phonon modes (e.g. longitudinal or transversely polarised ones) couple to either one branch of the dispersion curves .
Similar experimental data have been obtained in a coupled 2DEG/2DHG system with a barrier of 340 nm. The transresistance as function of hole and electron concentration is shown in Fig.3. These data are obtained at 5 K. Similarly to the case of coupled hole gases one finds an asymmetric behavior with respect to the two densities, particularly at large densities. We calculate again the $`k_F`$ and $`v_F`$ values of this 2DHG layers and compare them with $`k_{F,e}=\sqrt{2\pi n}`$ and $`v_{F,e}=\mathrm{}k_{F,e}/m^{}`$ of the 2DEG. The only satisfactory match with the experimental data is obtained using phonon coupling (i.e. matching the $`k_F`$ values) between the electrons and only one (the one with the heavier hole mass) of the hole branches. The resulting locus of equal $`k_F`$ values is plotted in Fig.3 as the heavy line. In this case an angular average of the $`k_F`$ values of the hole gas is used. An interesting result of Fig.3 is that an approximate congruence as it is realised along the ridge of maximum coupling between the hole and the electron Fermi surfaces requires in general quite unequal densities between the two gases. In contrast to the case of coupled 2d hole gases, the ridge positions of the coupled 2DEG/2DHG system shifts significantly with increasing temperature because the phase space for scattering processes changes very differently for the electrons and the holes, respectively .
The coupling between two 2d charge gases depends on the distance between the two layers. Theoretical studies of the interaction via plasmons predict a strong decrease of the transresistivity with distance $`\rho _Td_{eff}^3`$ . Here $`d_{eff}`$ is the distance between the center of gravity of the wave functions of the respective charge layers. On the other hand, the theory based on the exchange of coupled phonons predicts a logarithmic decrease with distance . In Fig. 4 we show data of $`\rho _T/T^2`$ for six coupled 2DHG systems for two different temperatures as function of $`d_{eff}`$ at matched densities of $`310^{11}cm^2`$. There is some scatter in the data which is probably due to different preparation conditions of the samples which were fabricated over a time span of more than one year. Nevertheless, the comparison of the data with the two predicted thickness dependences shows clearly that the logarithmic dependence is a better description of the data. A fit using $`\rho _T/T^2ln(l_{ph}/d_{eff})`$ with $`l_{ph}`$ being the mean free path of the phonons, as suggested in for the case of damped phonons, gives $`l_{ph}300nm`$. This number is small compared to the mean free paths deduced from thermal conductivity or heat pulse data for high quality GaAs but agrees quite well with the typical length scale of inhomogeneities along MBE grown layers as observed by AFM studies .
Finally we investigate the dependence of the transresistivity on perpendicular magnetic fields. In earlier studies in 2DEG/2DEG systems a dramatic increase of $`\rho _T`$ was observed except right under the quantum Hall effect conditions where $`\rho _T`$ vanishes . A less dramatic increase was seen in the 2DHG/2DEG systems . The 2DHG/2DHG systems are different because quantisation effects are small in most of our magnetic field and temperature regimes. Experimental data are shown in Fig. 5 where $`\rho _T`$ is plotted as function of a perpendicular magnetic field. The barrrier thickness is $`40nm`$, the carrier-densitiy is $`2.510^{11}cm^2`$ in both wells. At the lowest temperature T=1.5 K, $`\rho _T`$ reflects the Shubnikov-de-Haas-oscillations of $`\rho _{xx}`$ but the increase of the maxima with field is less than with the previously studied systems involving electrons . At higher temperatures these quantisation effects disappear, but now $`\rho _T`$ shows a $`decrease`$ with $`increasing`$ magnetic field. At even higher temperatures the $`\rho _T`$ seems to become nearly independent of the magnetic field. Interestingly, this behavior is just the opposite to the one of $`\rho _{xx}`$ which $`increases`$ with magnetic field in the same experimental range. The general decrease of $`\rho _T`$ at intermediate temperatures and its flattening at higher temperatures is also observed with densities which were not matched between the two layers and with samples having wider barriers. This behavior can be qualitatively understood from the behavior of the susceptibility $`Im\chi (q,\omega )`$ at $`q2k_F`$ in magnetic fields. At zero magnetic fields this function has a strong maximum near $`q2k_F`$ on which most of the analysis of this paper is based. This maximum disappears with magnetic field as was shown theoretically by Glasser . Thus, the coupling via phonons with $`q2k_F`$ is weakened in magnetic field. At quantising fields which correspond to the usual situation in systems involving electrons, this argument is no longer applicable because other types of interactions are dominant .
In conclusion we report the first data of the frictional drag between coupled 2d hole gases. By variation of doping profiles and the application of gate voltages we vary $`k_F`$ and $`v_F`$ independently from each other and establish that the coupling mechanism is dominated by phonon coupling at wavevectors $`2k_F`$. The coupling between 2d electron and 2d hole gases can be described within the same model. We find a logarithmic dependence of the coupling on the distance between the layers which agrees with a theoretical prediction of coupled phonon-plasmon modes. For the coupled hole gases we find a decrease with magnetic field as long as we are in the classical regime, which is consistent with the expected behavior of the susceptibilities at large wave vectors.
This work was supported by the BMBF under the grant BM621/4. S.J.C gratefully acknowledges financial support by the DAAD.
|
no-problem/9904/astro-ph9904397.html
|
ar5iv
|
text
|
# The Late Time Light Curve of SN 1998bw Associated with GRB980425
## 1 Introduction
In their search for an optical counterpart for the gamma ray burst GRB980425, Galama et al. (1998a) detected SN 1998bw in the galactic arm of ESO 184-G82(EOP 184-82), which Tinney et al. (1998) determined to have a red shift of $`0.0085\pm 0.0002`$. The supernova’s light curves rose sharply after the burst, and its spatial coordinates were well within the burst’s error box, strongly suggesting a connection between the two events. The probability of their independence was estimated by Galama et al. (1998a) at $`1.1\times 10^4`$. However, BeppoSAX also detected a fading x-ray source (generally thought to be the hallmark of the burst counterpart) at a position inconsistent with SN 1998bw (Pian et al. 1998, Piro et al. 1998, Pian et al. 1999), so the relationship between SN 1998bw and GRB980425 is unclear. Further observations showed that SN 1998bw is positionally coincident with a second BeppoSAX x-ray source which has faded by a factor of two in brightness from 26 April to 10 November 1999, which is consistent with x-ray emission from a supernova plus the galaxy (Pian et al. 1999).
SN 1998bw has peculiar and unique properties other than a possible association with a Gamma-Ray Burst. Its spectrum is unique (although two Type Ic supernovae have somewhat similar spectra; see Iwamoto et al. 1998) and displays ejection velocities measured from the blue wings of the Ca II line as high as $`60,000kms^1`$ (Kulkarni et al. 1998). Its emissions at radio wavelengths increased much more quickly than other supernovae, and it is also the most luminous supernova to date at radio wavelengths (Kulkarni et al. 1998). These coincidences of unusual properties greatly strengthen the connections between GRB 980425 and SN 1998bw. In general, a concensus has emerged that the burst is related to the supernova, and this has inspired much research detailing connections between the two phenomena.
Due to the unique and pivotal nature of SN 1998bw, it is imperative that the light curve be tracked in a wide range of optical bands as long as possible. Galama et al. (1998a) tracked the U, B, V, R, and I light curves for 58 days after the burst, and these showed a typical peak as generally seen for supernovae of many types. In the interests of recording as much data as possible for such an unprecedented event, we followed up on their results with further observations in the B, V, and I filters.
## 2 Observations
The data were obtained using the Yale 1-m telescope, at the Cerro Tololo Inter-American Observatory in Chile, from 27 June through 28 October 1998. Our series of observations commenced as soon as the refurbishing of the Yale 1-m telescope had been compeleted and the CCD camera installed. The images’ pixel size was 0.30”, with a field of view of $`10.2^{}\times 10.2^{}`$. Our exposure times were always 300 seconds per image, using B, V, and I filters. Our typical seeing had a FWHM of $`1.2^{\prime \prime }`$. We obtained 139 measures of the brightness of SN 1998bw.
The images were first processed with the normal procedure for overscan correction, bias subtraction and flat fielding. Our photometric analyses were made with IRAF’s program “APPHOT”. We used apertures of only three pixel radius to minimize interference from the parent galaxy, which was a sufficient size because of the high signal-to-noise ratio. Our background annuli were constructed with inner and outer radii of thirty and forty pixels centered on the star, with the sky background taken as the mode within this annulus. The deduced sky background for the supernova is close to that deduced for isolated stars.
For each image, observed magnitudes were recorded for SN 1998bw and for five comparison stars. We used the updated magnitudes provided by Galama et al. (1998b), choosing numbers 4, 6, 7, 9, and 10 because they were relatively bright. Comparison star #2 was excluded because its data was just eratic enough to arouse suspicion that it might be a small amplitude variable star, although is was not definitively identified as such.
We were able to estimate our magnitude uncertainties by comparing the standard stars with each other in a variety of images. These errors were determined to be substantially dependent on the apparent magnitudes of the stars due to the normal Poisson variations, which is why we used the brightest ones for our data analysis. Monitoring the differences between the standard stars in each image also enables us to catch photometric problems with the standard stars (due to cosmic rays, bad columns, etc.). A few nights had large uncertainty due to clouds or bad seeing. Our results are that in general we have systematic uncertainty of 0.02 mag added in quadrature with the statistical errors reported by IRAF. For the supernova, the statistical errors are generally substantially smaller than our systematic errors in the early portions of our light curve. The comparison stars are fainter than the supernova, yet our use of the average of five stars as our ‘standard’ improves the accuracy of this ‘standard’ to $`0.01`$ mag. In all, the uncertainties in our supernova magnitudes typically range from 0.02 to 0.04 mag.
SN 1998bw appears in the spiral arm of its host galaxy, so we must consider the effects of the galaxy light in our photometry. Fortunately, the supernova was quite bright during the entire duration of our study and the contribution of light from the spiral arm is minimal. To be quantitative, we have measured the surface brightness of the center of the spiral arm on both sides of the supernova and compared this with the total brightness within our photometric aperture centered on the supernova. Images of the galaxy from before the supernova show the brightness along the spiral arm to be uniform along the position of the supernova, so we know that there are no significant knots or stars at the supernova position. At the beginning of our light curve, the contamination from galaxy light in our photometry aperture varied from $`0.5\%`$ to $`0.8\%`$ for the three filters. So we have a systematic error which is smaller than our quoted uncertainties that will make the supernova slightly fainter than tabulated. Ideally, we should wait several years for the transient to fade to invisibility, then get further images with our same equipment and subtract off the galaxy light; but in the meantime the systematic error is known to be small.
For each image, we compared the instrumental magnitude of the supernova to the average instrumental magnitude of the five standard stars. This difference was then applied to the average of the standard stars’ actual magnitudes taken from Galama et al. (1998b) to determine the actual magnitude of SN 1998bw. Our results are plotted in Fig. 1 (along with Galama’s earlier results) and tabulated in Table 1.
We are impressed with the remarkable linearity of our portion of the light curves. The best fit lines to our data are displayed in Figure 1, and we see no significant systematic deviation from perfect lines at any time or in any color. Our limits on systematic deviations are $`<0.05`$ for our entire 123 day observation time. In the B filter, the light is declining at $`0.0141\pm 0.0002`$ magnitudes per day, which corresponds to a radioactive half-life of $`53.4\pm 0.8`$ days. For V, these figures are $`0.0194\pm 0.0003`$ magnitudes per day, which a corresponding radioactive half-life of $`40.9\pm 0.7`$ days. For I, the figures are $`0.0181\pm 0.0003`$ magnitudes per day and a half-life of $`41.6\pm 0.7`$ days.
The B light curve has a somewhat slower decay than in the V and I bands. For an extinction of $`A_v=0.2`$ (Galama et al. 1998), the supernova’s B-V was 0.82 mag at the beginning of our observation period and around 0.3 mag toward the end. The extinction corrected V-I rose by a small amount, from 0.53 to 0.60 mag, during the same time period.
With our B, V, and I light curves, we can approximate the bolometric light curve for radiation from the ultraviolet to the infrared. We have done this by first correcting for galactic extinction ($`A_v=0.20`$), converting our magnitudes into $`f_\nu `$, adopting a power law spectrum from B to V and from V to I, adopting a Rayleigh-Jeans spectrum for lower frequencies than I, adopting a Wien spectrum for higher frequencies than B, and integrating the spectrum. For JD2450996 (68 days after the burst) we get a bolometric flux of $`1.1\times 10^{11}ergcm^2s^1`$ or a bolometric luminosity of $`2.0\times 10^{42}ergs^1`$. For JD2451098 (170 days after the burst) we get a bolometric flux of $`2.1\times 10^{12}ergcm^2s^1`$ or a bolometric luminosity of $`3.9\times 10^{41}ergs^1`$. For the conversion to luminosity, we adopted a velocity of $`2550kms^1`$ and a Hubble Constant of $`65kms^1Mpc^1`$, for a distance of 39 Mpc. These calculated luminosities have significant uncertainties arising from the extinction ($`10\%`$), bolometric correction ($`30\%`$), and distance ($`20\%`$), so that overall errors perhaps as large as $`50\%`$ might be present. The effective half-life for this decline is 44 days.
## 3 Comparison with Other Supernova
SN1998bw has many unique and extreme properties, however at first look, its light curve appears to be that of a normal supernova. Is the light curve unique? We will compare our light curve with those of Type Ia, Ib, Ic, and II supernovae in turn.
For the majority of Type Ia events, the B light curve fades by 1.1 magnitudes in the first 15 days afters peak (Hamuy 1996a), while SN 1998bw has the same drop. The usual slope of the late time B light curve is $`0.01516\pm 0.0024mag/day`$ for Type Ia events (Barbon et al. 1984) and is easily compatible with that of SN 1998bw. However, the decline rates in V and I differ substantially between most of the Type Ia events and SN 1998bw (0.0184 versus 0.024 and 0.0181 versus 0.041 magnitudes per day respectively) from 70-80 days after peak. A second important difference is that almost all Type Ia events display a prominent bump in the I band light curves from 20-50 days after peak (Hamuy et al. 1996a, Riess et al. 1999), while SN 1998bw does not show any sign of such a bump. A third difference is that SN 1998bw has a peak absolute magnitude of $`18.88\pm 0.05`$ (Galama et al. 1998a) whereas the majority of Type Ia events have peak absolute magnitudes of $`19.26`$ (Hamuy et al. 1996b). However, uncertainties in distance and extinction can perhaps be as large as a third of a magnitude, so this third difference may not be significant. A fourth difference is that the extinction corrected color at peak of SN 1998bw is $`BV=0.47\pm 0.07`$ mag (Galama et al. 1998a) while the usual value for Type Ia events is $`0.00mag`$ (Hamuy et al. 1996b). So in all, the light curve of SN 1998bw looks similar to that of Type Ia events, yet detailed parameters are quantitatively different.
Perhaps a closer match can be found with the anomalous Type Ia SN 1991bg (Leibundgut et al. 1993; Fillipenko et al. 1992). This event had a substantially redder peak color ($`BV0.8`$), a much lower peak absolute magnitude ($`M_B=16.62`$), and no bump in the I band light curves. While the detailed light curve (and spectrum) of SN 1991bg is still different from that SN 1998bw, we note that many of the properties are more like those of SN 1991bg than of normal Type Ia events.
Type Ib and Ic light curves have not been characterized as closely as those of Type Ia supernovae. Nevertheless, enough is known (e.g., Uomoto & Kirshner, 1986; Ensman and Woosley 1988; Clocchiatti et al. 1997) to find similarities and differences with SN 1998bw. The overall light curve of Type Ib and Ic events is the same for SN 1998bw with similar decline rates over the first 15 days. The late time decline rate of Type Ic events vary, apparently with two classes, as slow and fast decliners. The 60-180 day decline rates of roughly 0.016 magnitudes per day are seen for the Type Ic events SN1983N and SN1983V (Clocchiatti et al. 1997) which is comparable to that for SN 1998bw. The color evolution for Type Ic events is similar to that of SN 1998bw, both at peak and at late times. The peak absolute magnitude of Type Ic events vary about $`M_B17.5`$, yet are all significantly fainter than SN 1998bw. However, with the few well measured Type Ic events having a wide scatter, the luminosity of SN 1998bw may not be unusual. In all, SN 1998bw appears to have a light curve within the class of Type Ic events.
Type II supernovae vary greatly in their light curve shape and color (Patat et al. 1993). The colors, peak absolute magnitudes, decline rate of SN 1998bw from peak, and late time decline rate are all within the normal range for the IIL subclass. Nevertheless, there are some subtle distinctions; such as a total lack of any indication of a plateau in the I band and the switch to the late time decline rate only $`30`$ days after peak.
In all, the light curve of SN1998bw is fully consistent with those of Type Ic supernovae, in keeping with the spectral classification and physical models.
## 4 Comparison with Models
The decay rate of the tail is so close to an exponential that we suggest that this is no coincidence. In addition, the measured decline rate corresponds to that expected from the decay of radio active cobalt (with a half-life of 78.5 days) as modified by the effects due to the expansion of the shell (Colgate and McKee 1969). Hence, it is reasonable to take our light curve as strong evidence that the late time light curve of SN 1998bw is being powered by the decay of cobalt, with the difference in slope caused by the leakage of gamma radiation from the shell.
Three detailed models have been presented seeking to explain the light curve of SN1998bw. Iwamoto et al. (1998) and Iwamoto (1999) model the event as an extremely energetic explosion of a massive star stripped down to its carbon/oxygen core. Woosley, Eastman, and Schmidt (1998) independently present a similar model with similar results. Höflich, Wheeler, and Wang (1998) and Wheeler, Höflich, and Wang (1999) present a model with an aspherical explosion in the nondegenerate C/O core of a massive star. An asymmetric event can account for the observed polarization. All three models account for the early light curve, the early colors, and the early spectrum with generally acceptable accuracy.
Iwamoto et al. (1998) and Iwamoto (1999) present predictions for the late-time V light curve of SN1998bw (see Figure 1 of Iwamoto 1999) as a perfectly straight line in a log-log plot of flux versus time since the burst, for a predicted power law with slope -2.75. This power law prediction does not agree with the observed exponential decline. However, K. Nomoto (1999, private communication) has presented a more detailed light curve prediction which shows a more complicated shape (neither a power law nor an exponential) than presented in Iwamoto (1999). The model V magnitude declines by 2.80 magnitudes from 27 June to 28 October with deviations from a simple exponential curve defined by the end points of up to 0.22 mag. For comparison, our data shows a decline of 2.26 mag over this same time and maximum departures from a simple exponential decline of $`<0.05mag`$.
One way to distinguish the three models is by the explosion energy and the ejected $`{}_{}{}^{56}Ni`$ mass. The spherically symmetric models have energies and nickel masses of around $`3\times 10^{52}erg`$ and $`0.7M_{}`$, while the aspherical models have $`2\times 10^{51}erg`$ and $`0.2M_{}`$. Wheeler, Höflich, and Wang point out that if the late time light curve tracks the radioactive decay line then the ejected nickel mass can be determined, and that this might prove the simplest discriminant between models.
One possible method to measure the nickel mass from our light curve is to scale the luminosity in the tail from another supernova of known late-time luminosity and nickel mass. The best case for comparison might be SN1987A which has a well measured light curve (Hamuy et al. 1988), a well known distance (50 kpc; McCray 1993), and a well known nickel mass ($`0.069\pm 0.003M_{}`$; McCray 1993). On day 170 after the core collapse, SN1998bw had an extinction corrected V magnitude of 17.60, while SN1987A had an extinction corrected magnitude of 4.39. If SN1987A were placed at 39 Mpc, then its V magnitude should appear 1.25 mag fainter than we observed for SN1998bw. This implies a nickel mass 3.2 times larger, or that SN1998bw has $`0.22M_{}`$. The dominant uncertainty arises from the distance to SN1998bw, for which peculiar velocities of up to $`400kms^1`$ and uncertainties in the Hubble Constant of up to $`10kms^1Mpc^1`$ yield nickel mass uncertainties of $`0.09M_{}`$.
The procedure in the previous paragraph can only be approximate, in particular since there might be substantial leakage of the gamma rays from the expanding nebula. Such leakage could explain why our observed decline is steeper than that associated with $`{}_{}{}^{56}Co`$ decay (Clocchiatti & Wheeler 1997). Such leakage would lower the late-time luminosity and lower the deduced nickel mass. The V light curve of SN1987A declined with the $`{}_{}{}^{56}Co`$ rate whereas the V light curve of SN1998bw declined at a roughly twice the rate. The effect of leakage on our previous derived nickel mass can only be estimated within specific models, yet it is likely that our $`0.22\pm 0.09M_{}`$ value must be regarded as a lower limit.
We expect that our late-time light curve will provide a set of observations useful for refining and constraining individual models of the unique SN1998bw. In particular, our data might constrain the quantity of ejected $`{}_{}{}^{56}Ni`$ so as to decide between symmetric and asymmetric models. Another challenge to models is to explain the near perfect exponential shape of the light curve even though the slope is not that of the $`{}_{}{}^{56}Co`$ decay.
We will continue to monitor the brightness of SN1998bw in 1999. However, the background light from its host galaxy is increasingly a problem for exact photometry.
|
no-problem/9904/hep-ph9904465.html
|
ar5iv
|
text
|
# A PHENOMENOLOGICAL OUTLOOK ON THREE-FLAVOR ATMOSPHERIC NEUTRINO OSCILLATIONS
## 1 Introduction
The recent atmospheric neutrino data from the Super-Kamiokande (SK) experiment are in excellent agreement with the hypothesis of flavor oscillations generated by nonzero neutrino mass and mixing in the $`\nu _\mu \nu _\tau `$ channel . Such hypothesis is consistent with all the SK data, including sub-GeV $`e`$-like and $`\mu `$-like events (SG$`e,\mu `$) , multi-GeV $`e`$-like and $`\mu `$-like events (MG$`e,\mu `$) , and upward-going muon events (UP$`\mu `$) , and is also corroborated by independent atmospheric neutrino results from the MACRO and Soudan-2 experiments, as well as by the finalized upward-going muon data sample from the pioneering Kamiokande experiment . Oscillations in the $`\nu _\mu \nu _\tau `$ channel are also compatible with the negative results of the reactor experiment CHOOZ in the $`\nu _\mu \nu _e`$ channel .
However, it has been realized that dominant $`\nu _\mu \nu _\tau `$ oscillations plus subdominant $`\nu _\mu \nu _e`$ oscillations are also consistent with SK+CHOOZ data, and lead to a much richer three-flavor oscillation phenomenology . A detailed $`3\nu `$ analysis of the SK observations, including the full 33 kTy data sample, can be found in Ref. . Here we report and comment briefly the results of our updated analysis, based on the recent 45 kTy SK data . The theoretical framework is based on the so-called one mass scale dominance , which has been used also for three-flavor oscillation studies of pre-SK atmospheric and reactor neutrino data in Refs. .
## 2 $`3\nu `$ analysis of SK phenomenology (45 kTy)
In the hypothesis that the two lightest neutrinos $`(\nu _1,\nu _2)`$ are effectively degenerate $`(m_1^2m_2^2)`$ (one mass scale dominance), it can be shown that atmospheric neutrinos probe only $`m^2m_3^2m_{1,2}^2`$, together with mixing matrix elements $`U_{\alpha i}`$ related to $`\nu _3`$, namely:
$$\mathrm{Parameter}\mathrm{space}(m^2,U_{e1}^2,U_{e2}^2,U_{e3}^2),$$
(1)
where $`U_{e1}^2+U_{e2}^2+U_{e3}^2=1`$ for unitarity. The unitarity constraint can be conveniently embedded in a triangle plot , whose corners represent the flavor eigenstates, while the heights projected from any inner point represent the $`U_{\alpha 3}^2`$’s.
Within this framework, we analyze 30 data points, related to the zenith distributions of sub-GeV events (SG $`e`$-like and $`\mu `$-like, 5+5 bins), multi-GeV events (MG$`e`$,$`\mu `$ 5+5 bins) and upward-going muons (UP$`\mu `$, 10 bins), using the latest 45 kTy SK sample . We also consider the rate of events in the CHOOZ reactor experiment (one bin), which constrains the $`\nu _e`$ disappearance channel. Constraints are obtained through a $`\chi ^2`$ statistic, as described in Ref. .
Figure 1 shows the regions favored at 90% and 99% C.L. in the triangle plot, for representative values of $`m^2`$. The CHOOZ data, which exclude a large horizontal strip in the triangle, appear to be crucial in constraining three-flavor mixing. Pure $`\nu _\mu \nu _e`$ oscillations (right side of the triangles) are excluded by SK and CHOOZ independently. The center of the lower side, corresponding to pure $`\nu _\mu \nu _\tau `$ oscillations with maximal mixing, is allowed in each triangle both by SK and SK+CHOOZ data. However, deviations from maximal $`(\nu _\mu \nu _\tau )`$ mixing, as well as subdominant mixing with $`\nu _e`$, are also allowed to some extent. Such deviations from maximal $`2\nu `$ mixing are slightly more constrained than in the previous analysis of the 33 kTy SK data .
Figure 2 shows the constraints on the mass parameter $`m^2`$ for unconstrained three-flavor mixing. The best fit value is reached at $`m^23\times 10^3`$ eV<sup>2</sup>, and is only slightly influenced by the inclusion of CHOOZ data. However, the upper bound on $`m^2`$ is significantly improved by including CHOOZ. As compared with the corresponding plot in Ref. (33 kTy), this figure shows that the 45 kTy data are in better agreement with the oscillation hypothesis (lower $`\chi ^2`$). Moreover, the favored range of $`m^2`$ is restricted by $`10\%`$ with respect to Ref. .
Figures 1 and 2 clearly show the tremendous impact of the SK experiment in constraining the neutrino oscillation parameter space. Prior to SK, the data could not significantly favor $`\nu _\mu \nu _\tau `$ over $`\nu _\mu \nu _e`$ oscillations, and could only put weak bounds on $`m^2`$ (see Refs. ).
Finally, Fig. 3 shows the best-fit zenith distributions of SG$`e,\mu `$, MG$`e,\mu `$, and UP$`\mu `$ events, normalized to the no-oscillation rates in each bin. There is excellent agreement between data and theory, especially for the $`\mu `$ distributions. The nonzero value of $`U_{e3}^2`$ at best fit leads to a slight expected electron excess in the MG$`e`$ sample for $`\mathrm{cos}\theta 1`$. The observed electron excess is, however, somewhat larger than expected, both for SG$`e`$’s and for MG$`e`$’s. A significant reduction of the statistical error is needed to probe possible MG$`e`$ distortions, which would be unmistakable signals of subdominant $`\nu _\mu \nu _e`$ oscillations.
## 3 Outlook
The Super-Kamiokande data are compatible with three-flavor oscillations dominated by $`\nu _\mu \nu _\tau `$ transitions. The amplitude of the $`\nu _\mu \nu _e`$ channel is not necessarily zero, although being strongly constrained by both SK and CHOOZ. A contribution from the $`\nu _\mu \nu _e`$ channel might explain part of the electron excess observed in SK, especially for multi-GeV $`e`$-like events. Higher statistics is needed to test such interpretation. A definite progress in confirming the oscillation hypothesis, and in constraining the mass-mixing parameters, emerges from a comparison of the 33 kTy and 45 kTy SK data analyses.
Acknowledgments. G.L.F. thanks the organizers of the workshop WIN’99 for kind hospitality.
## References
|
no-problem/9904/hep-lat9904013.html
|
ar5iv
|
text
|
# LATEST RESULTS FROM LATTICE QCD FOR EXOTIC HYBRID MESONS.
## 1 Introduction
The quark model predicts that the charge conjugation (C) and parity (P) of a meson with spin $`S`$ and orbital angular momentum $`L`$ is
$$P=(1)^{L+1}C=(1)^{L+S}$$
(1)
States with quantum numbers not produced by eq. 1, such as
$$J_{exotic}^{PC}=1^+,0^+,2^+,0^{}$$
(2)
are known as exotics . Exotic states are allowed by QCD.
There are a number of different possibilities for the structure of an exotic state. An exotic signal could be: a hybrid meson, which is quark and anti-quark and excited glue, or bound state of two quarks and two anti-quarks ($`\overline{Q}\overline{Q}QQ`$). The two most popular guesses for the structure of the ($`\overline{Q}\overline{Q}QQ`$) state are either a molecule of two mesons or diquark anti-diquark bound state. In this paper I review the latest lattice results for the masses of exotic hybrid mesons, concentrating on the $`1^+`$ state, obtained from lattice QCD.
## 2 Lattice simulations of exotic mesons
Many numerical predictions of QCD can be determined from the path integral
$$c(t)𝑑U𝑑\psi 𝑑\overline{\psi }\underset{\underset{¯}{x}}{}O(\underset{¯}{0},0)O(\underset{¯}{x},t)^{}e^{S_FS_G}$$
(3)
where $`S_F`$ is the fermion action (some lattice discretization of the Dirac action) and $`S_G`$ is the pure gauge action. The path integral in eq. 3 is put on the computer using a clever finite difference formalism , due to Wilson, that maintains gauge invariance. The physical picture for eq. 3 is that a hadron is created at time 0, from where it propagates to the time t, where it is destroyed. The fermion integration can be done exactly in eq. 3 to produce the fermion determinant. Simulations that include the effect of the determinant are very expensive computationally, so typically it is not included in the simulation (the quenched approximation). However there has been some recent work that includes the effect of the determinant on the light exotic spectrum.
The standard interpolating operator for the pion, which can be used in eq. 3, is
$$O_\pi (\underset{¯}{x},t)=\overline{\psi }(\underset{¯}{x},t)\gamma _5\psi (\underset{¯}{x},t)$$
(4)
which has the correct $`J^{PC}=0^+`$ quantum numbers. One possible interpolating operator for an exotic $`1^+`$ particle is
$$O_{1^+}(\underset{¯}{x},t)=\overline{\psi }(\underset{¯}{x},t)\gamma _jF_{ij}(\underset{¯}{x},t)\psi (\underset{¯}{x},t)$$
(5)
where $`F`$ is the QCD field strength tensor. It is essential to use operators that have some spatial separation between the quarks in the meson to get a good signal. Recently the MILC collaboration has attempted to measure the “wave function” of the $`1^+`$ hybrid meson in coulomb gauge . Unfortunately the operator used did not have the correct charge conjugation quantum number, so the published wave function is incorrect.
In this formalism a gauge invariant interpolating operator, for any possible exotic hybrid particle or four particle state can be constructed. The dynamics then determines whether the resulting state has a narrow decay width, which can be detected experimentally. In the large $`N_c`$ (number of colours) limit both exotic hybrid mesons and non-exotic mesons have widths vanishingly small compared to their masses.
The data from the simulation is extracted using a fit model :
$$c(t)=a_0exp(m_0t)+a_1exp(m_1t)+\mathrm{}$$
(6)
where $`m_0`$ ($`m_1`$) is the ground (first excited) state mass and the dots represent higher excitations. Although in principle excited state masses can be extracted from a multiple exponential fit, in practice this is numerically non-trivial because of the noise in the data from the simulation. Simulations that involve the calculation of the properties of exotic hybrid mesons are harder than those that concentrate on the non-exotic hadrons, because the signal to noise ratio is worse for exotic mesons than for $`\overline{Q}Q`$ mesons.
In an individual lattice simulation there are errors from the finite size of the lattice spacing and the finite lattice volume. State of the art lattice simulations in the quenched theory, run at a number of different lattice spacings and physical volumes and extrapolate the results to the continuum and infinite volume. For the exotic mesons, this is done for heavy quarks (see section 4), but as yet, the continuum extrapolation has not been done for light exotic mesons.
How successful is lattice QCD in practice? One way to tell is to compare the predictions for the masses from lattice QCD for well known particles (proton, $`\rho `$, etc.) with experiment. The most accurate quenched calculation to date has recently been completed by the CP-PACS collaboration . From the masses of 11 light hadrons, they conclude that the quenched approximation disagrees with experiment by at most 11%. The comparison of results from simulations that include dynamical fermions with experiment is less clear, because of their high computational cost (see Kenway for a review of the latest results).
The results from lattice QCD also provide insight into the underlying dynamics of light hadrons. Lattice QCD simulations can test the various assumptions made in models of the QCD dynamics. For example there are a number of models of exotic states based on the idea of a bound diquark anti-diquark pair . A critical assumption in diquark models is that two quarks actually do cluster to form a diquark. This assumption has recently been tested in a lattice gauge theory simulation by the Bielefeld group , where they found no deeply bound diquark state in Landau gauge.
## 3 Results for light exotic mesons
In the last year the MILC collaboration have repeated their initial simulations using the clover fermion action. The clover action is “closer” to the continuum than the Wilson fermion action, because it has the leading order lattice spacing terms removed. There are also new results for the hybrid masses from the SESAM collaboration (reported by Lacock and Schilling) , that include some effects from dynamical sea fermions.
The results for the mass of the $`1^+`$ exotic state are summarised in table 1. All the results from the various simulations are essentially consistent. with the mass of the $`1^+`$ state around $`2`$ GeV. In table 2 we show the physical parameters for each of the simulations. The interpolating operators used to create the exotic meson states in the MILC calculations are different to those used in the UKQCD and SESAM simulations .
The observation that the results for the mass of the $`1^+`$ hybrid meson are consistent for very different simulations gives us confidence in the final result. Although I would prefer to see simulation results at lighter quark masses. Simulations at the point where the ratio of the pseudoscalar mass to vector mass ($`M_{PS}/M_V0.5`$) are possible with current algorithms and computers .
## 4 Results for heavy exotic mesons
There has been a lot of work on calculating the spectroscopy of $`\overline{c}c`$ and $`\overline{b}b`$ mesons from lattice gauge theory (see Davies for a review). The main technical complication in heavy quark simulations is that the lattice spacing of current simulations is not smaller than the heavy quark mass. So various effective field theory Lagrangian approximations to QCD are simulated.
The NRQCD (nonrelativistic QCD) Lagrangian is one such effective field theory approximation to QCD, with the expansion parameter equal to the velocity squared. NRQCD has been particularly successful in simulating the Upsilon spectrum , but is less well converged for charmomium, (particularly for spin splittings), because the charm quarks move with higher velocity . The NRQCD Langrangian correct up to $`O(Mv^2)`$ is
$$^{NRQCD}=\overline{\psi }(\frac{^2}{2M}\frac{c_0^4}{8M^2}\frac{c_1\sigma B}{2M})\psi $$
(7)
where $`c_0`$ and $`c_1`$ are coefficients obtained by a perturbative matching procedure to QCD. In table 3 the results of all the recent NRQCD simulations of the $`\overline{b}bg`$ hybrids in the quenched approximation are compiled. In the hybrid meson simulations no spin terms are included in the Lagrangian ($`c_1=0`$ in eq. 7), so the $`1^+`$, $`0^+`$, and $`2^+`$ states are degenerate. Both the results from the CP-PACS collaboration and from Juge, Kuti and Morningstar were shown to be independent of the lattice spacing and lattice volume. For example, the CP-PACS collaboration found that the masses of the hybrids were independent of the box size above $`1.2fm`$.
The “asymmetric” comment in table 3 refers to the technique of treating space and time asymmetrically. A smaller lattice spacing was used in the time direction than in the space direction, which allowed the signal to be seen for further, for a given spatial volume. This technique has helped to reduce the errors. The practicalities of this idea were demonstrated by Morningstar and Peardon for the glueball spectrum.
In table 4 the results of the mass splitting of the $`1^+`$ states and the $`1S`$ state are shown in the charmonium system. The MILC collaboration used the standard Wilson and clover actions to simulate the charm quark in their simulations of heavy exotic mesons, as previous work has shown this to be reliable .
The first results for heavy exotic hybrids were done in the adiabatic surfaces approach, where the effect of the excited glue is subsumed in a potential measured on the lattice (see for a review). Juge, Kuti and Morningstar have completed a systematic study of these potentials. The NRQCD approach is a more accurate approximation to QCD than the adiabatic potential technique; however Juge, Kuti and Morningstar found that the adiabatic potential approach reproduced the level splittings from NRQCD up to 10 %. A preliminary result for the calculation of the adiabatic surfaces with the effects of dynamical fermions included has been reported by Bali and collaborators. No dramatic differences between the quenched theory result were observed.
## 5 Conclusions
All the lattice simulations agree that the light $`1^+`$ state has a mass of $`2.0(2)GeV`$. The first simulation that included the effects of dynamical fermions has not changed the result.
The experimental results for light exotics are reviewed by S.U. Chung , so I just briefly compare the lattice results to experiment. There is an experimental signal for a $`1^+`$ state at $`1.4`$ GeV with a decay into $`\eta \pi `$ from E852, Crystal barrel, VES, and KEK. It is surprising that this state is only seen in the $`\eta \pi `$ channel, as this decay is theoretically suppressed relative to other decays . The E852 collaboration have also reported a signal for $`1^+`$ state decaying into $`\rho \pi `$ with a mass of around $`1.6`$ GeV. The decay width is in reasonable agreement with theoretical calculations .
Clearly the agreement between the possible experimental signals for the $`1^+`$ states and the lattice results is very poor. The errors on the lattice results for $`1^+`$ states are large relative to the errors on $`\overline{Q}Q`$ states. To quantify the disagreement between experiment and lattice results the systematic errors on the lattice simulation results should be reduced. In particular the masses of the quarks used in the lattice simulation should be reduced. It is possible that the states seen experimentally are really $`\overline{Q}\overline{Q}QQ`$ states, in which case the operators used in the lattice simulations would not couple strongly to them.
Although the adiabatic lattice potential approach is not expected to be a good description of the physics of light hybrids, we note that the results of Juge, Kuti and Morningstar , show that the splitting between the ground and first excited state is about $`200`$ MeV, in broad agreement with the experimental results of the E852 collaboration. Although no insight is gained about the different decay widths.
To definitely identify an exotic hybrid meson requires both the calculation of the mass as well as the decay widths. There has been very little work on hadronic decays on the lattice. The most obvious hadronic process to study using lattice gauge theory is the $`\rho \pi \pi `$ decay, however there have only been a few attempts to calculate the $`g_{\rho \pi \pi }`$ coupling The GF11 lattice group has recently computed the decay widths for the decay of the $`0^{++}`$ glueball to two pseudoscalars .
The MILC collaboration have started to investigate the mixing between the operator in eq. 5 and the operator ($`\pi a_1`$) eq. 8.
$$\overline{\psi }^a\gamma _5\psi ^a\overline{\psi }^b\gamma _5\gamma _i\psi ^b$$
(8)
which has the quantum numbers $`1^+`$. This type of correlator would naively be expected to yield the decay width of the $`1^+`$ state to $`\rho `$, and $`a_1`$. Unfortunately the analysis of Maiani and Testa shows that the matrix element required in the computation of the decay width is hidden beneath an unphysical term that increases exponentially with time. The origin of the unphysical term comes from the requirement that both final mesons should be onshell and is deeply related to theory being defined in Euclidean space (required for us to have a well defined theory to simulate). Some information may be extracted for onshell processes , using the methods proposed by Michael .
## 6 Acknowledgements
I thank Chris Michael and Doug Toussaint for discussions about exotic mesons.
|
no-problem/9904/math9904032.html
|
ar5iv
|
text
|
# Warwick EuroConference (July 1996)
### Full title:
Recent Trends in Algebraic Geometry – EuroConference on Algebraic Geometry (Warwick, July 1996), Editors: Klaus Hulek (chief editor), Fabrizio Catanese, Chris Peters and Miles Reid, CUP, May 1999
### Contents:
Victor V. Batyrev: Birational Calabi–Yau $`n`$-folds have equal Betti numbers, 1–11 (alg-geom/9710020)
Arnaud Beauville: A Calabi–Yau threefold with non-Abelian fundamental group, 13–17 (alg-geom/9502003)
K. Behrend: Algebraic Gromov–Witten invariants, 19–70 (compare alg-geom/9601011)
Philippe Eyssidieux: Kähler hyperbolicity and variations of Hodge structures, 71–92
Carel Faber: Algorithms for computing intersection numbers on moduli spaces of curves, with an application to the class of the locus of Jacobians, 93–109 (alg-geom/9706006)
Marat Gizatullin: On some tensor representations of the Cremona group of the projective plane, 111–150
Y. Ito and I. Nakamura: Hilbert schemes and simple singularities, 151–233
Oliver Küchle and Andreas Steffens: Bounds for Seshadri constants, 235–254 (alg-geom/9601018)
Marco Manetti: Degenerate double covers of the projective plane, 255–181 (see further math.AG/9802088)
David R. Morrison: The geometry underlying mirror symmetry, 283–310 (alg-geom/9608006)
Shigeru Mukai: Duality of polarized K3 surfaces, 311–326
Roberto Paoletti: On symplectic invariants of algebraic varieties coming from crepant contractions, 327–346
Kapil H. Paranjape: The Bogomolov–Pantev resolution, an expository account, 347–358 (math.AG/9806084)
Tetsuji Shioda: Mordell–Weil lattices for higher genus fibration over a curve, 359–373
Bernd Siebert: Symplectic Gromov–Witten invariants, 375–424
Claire Voisin: A generic Torelli theorem for the quintic threefold, 425–463
P.M.H. Wilson: Flops, Type III contractions and Gromov–Witten invariants on Calabi–Yau threefolds, 465–483
## Foreword
The volume contains a selection of seventeen survey and research articles from the July 1996 Warwick European algebraic geometry conference. These papers give a lively picture of current research trends in algebraic geometry, and between them cover many of the outstanding hot topics in the modern subject. Several of the papers are expository accounts of substantial new areas of advance in mathematics, carefully written to be accessible to the general reader. The book will be of interest to a wide range of students and nonexperts in different areas of mathematics, geometry and physics, and is required reading for all specialists in algebraic geometry.
The European algebraic geometry conference was one of the climactic events of the 1995–96 EPSRC Warwick algebraic geometry symposium, and turned out to be one of the major algebraic geometry events of the 1990s. The scientific committee consisted of A. Beauville (Paris), F. Catanese (Pisa), K. Hulek (Hannover) and C. Peters (Grenoble) representing AGE (Algebraic Geometry in Europe, an EU HCM–TMR network) and N.J. Hitchin (Oxford), J.D.S. Jones and M. Reid (Warwick) representing Warwick and British mathematics. The conference attracted 178 participants from 22 countries and featured 33 lectures from a star-studded cast of speakers, including most of the authors represented in this volume.
#### The expository papers
Five of the articles are expository in intention: among these a beautiful short exposition by Paranjape of the new and very simple approach to the resolution of singularities; a detailed essay by Ito and Nakamura on the ubiquitous ADE classification, centred around simple surface singularities; a discussion by Morrison of the new special Lagrangian approach giving geometric foundations to mirror symmetry; and two deep and informative survey articles by Behrend and Siebert on Gromov–Witten invariants, treated from the contrasting viewpoints of algebraic and symplectic geometry.
#### Some main overall topics
Many of the papers in this volume group around a small number of main research topics. Gromov–Witten invariants was one of the main new breakthroughs in geometry in the 1990s; they can be developed from several different starting points in symplectic or algebraic geometry. The survey of Siebert covers the analytic background to the symplectic point of view, and outlines the proof that the two approaches define the same invariants. Behrend’s paper explains the approach in algebraic geometry to the invariants via moduli stacks and the virtual fundamental class, which essentially amounts to a very sophisticated way of doing intersection theory calculations. The papers by Paoletti and Wilson give parallel applications of Gromov–Witten invariants to higher dimensional varieties: Wilson’s paper determines the Gromov–Witten invariants that arise from extremal rays of the Mori cone of Calabi–Yau 3-folds, whereas Paoletti proves that Mori extremal rays have nonzero associated Gromov–Witten invariants in many higher dimensional cases. The upshot is that extremal rays arising in algebraic geometry are in fact in many cases invariant in the wider symplectic and topological setting.
Another area of recent spectacular progress in geometry and theoretical physics is Calabi–Yau 3-folds and mirror symmetry. This was another major theme of the EuroConference that is well represented in this volume. The paper by Voisin, which is an extraordinary computational tour-de-force, proves the generic Torelli theorem for the most classical of all Calabi–Yau 3-folds, the quintic hypersurface in $`^4`$. The survey by Morrison explains, among other things, the Strominger–Yau–Zaslow special Lagrangian interpretation of mirror symmetry. Beauville’s paper gives the first known construction of a Calabi–Yau 3-fold having the quaternion group of order 8 as its fundamental group. The paper by Batyrev proves that the Betti numbers of a Calabi–Yau 3-fold are birationally invariant, using the methods of $`p`$-adic integration and the Weil conjectures; the idea of the paper is quite startling at first sight (and not much less so at second sight), but it is an early precursor of Kontsevich’s idea of motivic integration, as worked out in papers of Denef and Loeser. Several other papers in this volume (those of Ito and Nakamura, Mukai, Shioda and Wilson) are implicitly or explicitly related to Calabi–Yau 3-folds in one way or another.
#### Other topics
The remaining papers, while not necessarily strictly related in subject matter, include some remarkable achievements that illustrate the breadth and depth of current research in algebraic geometry. Shioda extends his well-known results on the Mordell–Weil lattices of elliptic surfaces to higher genus fibrations, in a paper that will undoubtedly have substantial repercussions in areas as diverse as number theory, classification of surfaces, lattice theory and singularity theory. Faber continues his study of tautological classes on the moduli space of curves and Abelian varieties, and gives an algorithmic treatment of their intersection numbers, that parallels in many respects the Schubert calculus; he obtains the best currently known partial results determining the class of the Schottky locus. Gizatullin initiates a fascinating study of representations of the Cremona group of the plane by birational transformations of spaces of plane curves. Eyssidieux gives a study, in terms of Gromov’s Kähler hyperbolicity, of universal inequalities holding between the Chern classes of vector bundles over Hermitian symmetric spaces of noncompact type admitting a variation of Hodge structures. Küchle and Steffens’ paper contains new twists on the idea of Seshadri constants, a notion of local ampleness arising in recent attempts on the Fujita conjecture; they use in particular an ingenious scaling trick to provide improved criteria for the very ampleness of adjoint line bundles.
Manetti’s paper continues his long-term study of surfaces of general type constructed as iterated double covers of $`^2`$. He obtains many constructions of families of surfaces, and proves that these give complete connected components of their moduli spaces, provided that certain naturally occuring degenerations of the double covers are included. This idea is used here to establish a bigger-than-polynomial lower bound on the growth of the number of connected components of moduli spaces. In more recent work, he has extended these ideas in a spectacular way to exhibit the first examples of algebraic surfaces that are proved to be diffeomorphic but not deformation equivalent.
The Fourier–Mukai transform is now firmly established as one of the most important new devices in algebraic geometry. The idea, roughly speaking, is that a sufficiently good moduli family of vector bundles (say) on a variety $`X`$ induces a correspondence between $`X`$ and the moduli space $`\widehat{X}`$. In favourable cases, this correspondence gives an equivalence of categories between coherent sheaves on $`X`$ and on $`\widehat{X}`$ (more precisely, between their derived categories). The model for this theory is provided by the case originally treated by Mukai, when $`X`$ is an Abelian variety and $`\widehat{X}`$ its dual; Mukai named the transform by analogy with the classical Fourier transform, which takes functions on a real vector space to functions on its dual. It is believed that, in addition to its many fruitful applications in algebraic geometry proper, this correspondence and its generalisations to other categories of geometry will eventually provide the language for mathematical interpretations of the various “dualities” invented by the physicists, for example, between special Lagrangian geometry on a Calabi–Yau 3-fold and coherent algebraic geometry on its mirror partner (which, as described in Morrison’s article, is conjecturally a fine moduli space for special Lagrangian tori). Mukai’s magic paper in this volume presents a Fourier–Mukai transform for K3 surfaces, in terms of moduli of semi-rigid sheaves; under some minor numerical assumptions, he establishes the existence of a dual K3 surface, the fact that the Fourier–Mukai transform is an equivalence of derived categories, and the biduality result in appropriate cases.
The paper of Ito and Nakamura is the longest in the volume; it combines a detailed and wide-ranging expository essay on the ADE classification with an algebraic treatment of the McKay correspondence for the Kleinian quotient singularities $`^2/G`$ in terms of the $`G`$-orbit Hilbert scheme. The contents of their expository section will probably come as a surprise to algebraic geometers, since alongside traditional aspects of simple singularities and their ADE homologues in algebraic groups and representation of quivers, they lay particular emphasis on partition functions in conformal field theory with modular invariance under $`\mathrm{SL}(2,)`$ and on II<sub>1</sub> factors in von Neumann algebras. Their study of the $`G`$-Hilbert scheme makes explicit for the first time many aspects of the McKay correspondence relating the exceptional locus of the Kleinian quotient singularities $`^2/G`$ with the irreducible representations of $`G`$; for example, the way in which the points of the minimal resolution can be viewed as defined by polynomial equations in the character spaces of the corresponding irreducible representations, or the significance in algebraic terms of tensoring with the given representation of $`G`$. Ito and Nakamura and their coworkers are currently involved in generalising many aspects of the $`G`$-orbit Hilbert scheme approach to the resolution of Gorenstein quotient singularities and the McKay correspondence to finite subgroups of $`\mathrm{SL}(3,)`$, and this paper serves as a model for what one hopes to achieve.
#### Thanks to all our sponsors
The principal financial support for the EuroConference was a grant of ECU40,000 from EU TMR (Transfer and Mobility of Researchers), contract number ERBFMMACT 950029; we are very grateful for this support, without which the conference could not have taken place. The main funding for the 1995–96 Warwick algebraic geometry symposium was provided by British EPSRC (Engineering and physical sciences research council). Naturally enough, the symposium was one of the principal activities of the Warwick group of AGE (European Union HCM project Algebraic Geometry in Europe, Contract number ERBCHRXCT 940557), and financial support from Warwick AGE and the other groups of AGE was a crucial element in the success of the symposium and the EuroConference. We also benefitted from two visiting fellowships for Nakamura and Klyachko from the Royal Society (the UK Academy of Science). Many other participants were covered by their own research grants.
The University of Warwick, and the Warwick Mathematics Institute also provided substantial financial backing. All aspects of the conference were enhanced by the expert logistic and organisational help provided by the Warwick Math Research Centre’s incomparable staff, Elaine Greaves Coelho, Peta McAllister and Hazel Graley.
Klaus Hulek and Miles Reid, November 1998
|
no-problem/9904/astro-ph9904091.html
|
ar5iv
|
text
|
# The Linear Polarization of Sagittarius A* I. VLA Spectro-polarimetry at 4.8 and 8.4 GHz
## 1 Introduction
The compact non-thermal radio source Sgr A\* has long been recognized as a massive black hole candidate. Recent results from stellar proper motion studies indicate that there is a dark mass of $`2.6\times 10^6M_{\mathrm{}}`$ enclosed within 0.01 pc (Genzel et al. 1997, Ghez et al. 1998). Very long baseline interferometry studies at millimeter wavelengths have shown that the intrinsic radio source coincident with the dark mass has a size that is less than 1 AU and a brightness temperature greater than $`10^9`$ K (Rogers et al. 1994, Bower & Backer 1998, Lo et al. 1998). Together these points are compelling evidence that Sgr A\* is a cyclo-synchrotron emitting region surrounding a massive black hole. Nevertheless, specific details of the excitation of high energy electrons, their distribution and the accretion of infalling matter onto Sgr A\* are unknown (e.g., Falcke, Mannheim & Biermann 1993, Melia 1994, Narayan et al. 1998, Mahadevan 1998).
Linear polarization stands as one of the few observables of Sgr A\* not extensively investigated observationally or theoretically. However, we expect linear polarization to arise from the cyclo-synchrotron radiation that is responsible for the radio to millimeter wavelength spectrum. A homogeneous, optically-thin, synchrotron source with a uniform magnetic field has a fractional polarization of 70%. Measured fractional polarizations in AGN are typically a few percent at wavelengths shorter than 6 cm where the compact cores dominate the highly-polarized radio lobes in the total flux (e.g., Aller, Aller & Hughes 1992). However, polarization VLBI images sometimes show regions of significantly enhanced polarization (Brown, Roberts & Wardle 1994).
The polarization of Sgr A\* may prove to be as important a diagnostic of models for the radio to millimeter spectrum as it has been for AGN. Detection of linear polarization in AGN has firmly established synchrotron emission as the radiation mechanism. Comparison of the evolution of linear polarization to the evolution of total intensity has provided a strong argument for the existence of shocks in the relativistic jets of AGN (e.g., Hughes, Aller and Aller 1985). Detection of similar correlations in polarized and total intensity variations in Sgr A\* would be convincing evidence for a jet. Other models may have unique signatures for polarized intensity variations.
Sgr A\* is located in a region with strong magnetic fields and high electron density. The image of Sgr A\* is significantly scatter-broadened by intervening thermal plasma (e.g., Lo et al. 1998), as are the images of many masers in the Galactic Center region (Frail et al. 1994). Furthermore, nonthermal filaments in the Galactic Center region show RMs which vary on the arcsecond scale and are as large as 4000 $`\mathrm{rad}\mathrm{m}^2`$ (Yusef-Zadeh, Wardle & Parastaran 1997). Such large RMs can effectively depolarize a signal detected with a large bandwidth.
In §2, we discuss the effect of large RMs on a polarized signal and our Fourier transform technique for detecting large RMs. In §3.1, we present VLA continuum observations at 4.8 GHz. In §3.2 and §3.3, we present VLA spectro-polarimetric observations at 4.8 GHz and 8.4 GHz. These observations are sensitive to a wide-range of RMs. In §4, we consider other effects of interstellar matter on a polarized signal from Sgr A\*. And in §5, we discuss the consequences of our upper limits for the polarization on models for Sgr A\*. In a future paper, we will address millimeter polarization observations of Sgr A\*.
## 2 Searching for Large RMs
In an ionized and magnetized region right and left circularly polarized waves will have different indices of refraction. This leads to a wavelength-dependent delay between circular polarizations which is equivalent to a rotation of the position angle $`\chi `$ of a linearly polarized signal
$$\chi _F=\mathrm{RM}\lambda ^2,$$
(1)
where RM is the rotation measure. This rotation of $`\chi `$ is equivalent to a rotation in the two-dimensional Stokes $`Q`$ and $`U`$ space.
A linearly polarized signal will be significantly depolarized in an observing bandwidth $`\mathrm{\Delta }\nu `$ if $`\chi `$ rotates by more than one radian, or if the RM exceeds
$$\mathrm{RM}_{\mathrm{max}}=\frac{1}{2}\frac{1}{\lambda ^2}\frac{\nu }{\mathrm{\Delta }\nu }.$$
(2)
If the bandwidth $`\mathrm{\Delta }\nu `$ is split into $`n\times \delta \nu `$ channels, a search can be made for RMs larger than $`\mathrm{RM}_{\mathrm{max}}`$. When $`\chi `$ wraps through more than one turn, $`n\pi `$ ambiguities make it impossible through a linear least squares fit to detect RMs. Fourier transforming the complex visibility $`P=Q+iU`$ with respect to $`\lambda ^2`$ searches for large RMs without loss of sensitivity. The maximum RM detectable in this scheme can be found by replacing $`\mathrm{\Delta }\nu `$ with $`\delta \nu `$ in Equation 2. In addition to detecting RMs that exceed $`\mathrm{RM}_{\mathrm{max}}`$, the technique is sensitive to multiple RMs from the same object. A more detailed analysis of this technique can be found in Killeen et al. (1999).
A continuum observation with the VLA at 4.8 GHz with 50 MHz has $`\mathrm{RM}_{\mathrm{max}}10^4\mathrm{rad}\mathrm{m}^2`$. Splitting the band into 256 channels increases $`\mathrm{RM}_{\mathrm{max}}`$ by two orders of magnitude to $`3.5\times 10^6\mathrm{rad}\mathrm{m}^2`$. The minimum fully-sampled RM detectable in a spectro-polarimetric data set, $`\mathrm{RM}_{\mathrm{min}}`$, is approximately equal to $`\mathrm{RM}_{\mathrm{max}}`$ for a continuum data set with the same total bandwidth.
The RM can be found to better accuracy than $`\mathrm{RM}_{\mathrm{min}}`$. We estimate the error to be
$$\sigma _{\mathrm{RM}}=\frac{\mathrm{RM}_{\mathrm{min}}}{\mathrm{SNR}}.$$
(3)
SNR is the ratio of the peak amplitude in Fourier space to the off-peak root-mean-square noise.
## 3 Observations and Data Reduction
### 3.1 VLA Continuum Polarimetry at 4.8 GHz
The VLA of the National Radio Astronomy Observatory<sup>1</sup><sup>1</sup>1The National Radio Astronomy Observatory is a facility of the National Science Foundation operated under cooperative agreement by Associated Universities, Inc. observed Sgr A\* on 10 and 18 April 1998 in the A array at 4.8 GHz with a bandwidth of 50 MHz. Instrumental calibration was performed with the compact sources 1741-038 and 1748-253. The right-left phase difference was set with observations of 3C 286. Only baselines longer than 100 $`k\lambda `$ were used for Sgr A\*. Several nearby calibrator sources, GC 441, W56 and W109, were also observed (Backer & Sramek 1999). All sources were self-calibrated and imaged in Stokes $`I,Q`$ and $`U`$.
We summarize the measured polarized and total intensities of Sgr A\* and the calibrators in Table 1. The rms noise in the Sgr A\* map is 74 $`\mu \mathrm{Jy}`$. Consistency between the results on the two days indicates the accuracy of the results. Polarization was reliably detected from all sources but Sgr A\* and GC 441. The measured polarization at the position of Sgr A\* is 0.1%. This value is equal to the average off source fractional polarization in the map and is, therefore, an upper limit. The maximum RM detectable with this bandwidth is $`10^4\mathrm{rad}\mathrm{m}^2`$.
### 3.2 VLA Spectro-polarimetry at 4.8 GHz
The VLA observed Sgr A\* in the A array in a spectro-polarimetric mode at 4.8 GHz on 27 November 1992. Observations were carried out in 8 consecutively-spaced frequency bands of 6.25 MHz each. Each band was divided into 32 separate frequency channels. The bands covered the frequency range from 4832 MHz to 4882 MHz. Five scans of 2.5 minutes apiece on Sgr A\* were interleaved with six scans of 2.5 minutes apiece on NRAO 530 in each frequency band. Amplitude, phase and polarization calibration were performed separately for each band. Polarization calibration was performed with NRAO 530 alone and with NRAO 530 and Sgr A\*, producing similar final results. The right-left phase difference was set for each band with an observation of 3C 286.
For each source, the spectral data were time-averaged and exported from AIPS for further processing. A bandpass correction was applied. The complex polarization was then Fourier-transformed with respect to $`\lambda ^2`$. Sampling effects were removed through a one-dimensional CLEAN method. The CLEAN method permits a better estimate of the RM peak and of the noise level. The sampling sidelobes are readily visible for 3C 286 and NRAO 530 in Figure 1. Our tests with noise data and with synthetic signals indicate that the CLEAN method does not generate false signals and improves the accuracy of peak determination. Applying CLEAN to the 4.8 GHz NRAO 530 data reduced the noise in the spectrum from 1.6 mJy to 0.26 mJy.
The range of fully-sampled RM is $`10^4\mathrm{rad}\mathrm{m}^2`$ to $`3.5\times 10^6\mathrm{rad}\mathrm{m}^2`$. The Fourier amplitude for each source is shown in Figure 1 and the results are summarized in Table 2. These images are without bandpass correction and dirty-beam removal. We also calculate and plot the Fourier transform for a distribution of Gaussian noise. Strong peaks at low RM are apparent for both 3C 286 and NRAO 530, as expected. The measured values are consistent with the known RMs of these sources: $`1\pm 2\mathrm{rad}\mathrm{m}^2`$ for 3C 286 and $`63\pm 5\mathrm{rad}\mathrm{m}^2`$ for NRAO 530 (Rusk 1988).
No strong peak is apparent for Sgr A\* at any RM. The maximum Fourier amplitude for Sgr A\* is $`0.15\%`$ at $`\mathrm{RM}=2.1\times 10^6\mathrm{rad}\mathrm{m}^2`$. Imaging Sgr A\* with and without a RM correction produced a peak polarization of $`0.2\%`$. This is equal to the fractional polarization of thermal ionized gas in the vicinity of Sgr A\*, indicating that we are limited by residual instrumental polarization.
### 3.3 VLA Spectro-polarimetry at 8.4 GHz
The VLA observed Sgr A\* in the A array in a spectro-polarimetric mode at 8.4 GHz, also on 27 November 1992. Observations were carried out in 7 frequency bands of 6.25 MHz each. Each band was divided into 32 separate frequency channels. Five bands covered the frequency range from 8405 MHz to 8437 MHz. Two other bands were centered at 8150 MHz and 8700 MHz. Five scans of 2.5 minutes apiece on Sgr A\* were interleaved with six scans of 2.5 minutes apiece on NRAO 530 in each frequency band. Amplitude, phase and polarization calibration were performed separately for each band. Polarization calibration was performed with NRAO 530 alone and with NRAO 530 and Sgr A\* together, producing similar final results. The right-left phase difference was set for each band with an observation of 3C 286. The sources W56, 1741-312, GC 441, W109 and 1748-253 were observed for two minutes in the three 6.25 MHz bands centered at 8150 MHz, 8420 MHz and 8700 MHz. The results for all sources were the same using all frequency bands or only the inner 5 bands. The same reduction steps were taken for the 8.4 GHz data as for the 4.8 GHz data.
The Fourier amplitudes for all sources are shown in Figures 2 and 3. These images are without bandpass correction and dirty-beam removal. The results are summarized in Table 3. These data are sensitive to $`3.5\times 10^5<|\mathrm{RM}|<1.5\times 10^7\mathrm{rad}\mathrm{m}^2`$. There are strong detections of linear polarization in 3C 286 and NRAO 530 at RMs consistent with zero. Significant detections were also made for W56, W109, 1741-312 and 1748-253, also at RMs consistent with zero. No polarization was detected in GC 441. The errors in RM for these secondary calibrators are larger due to the sparser frequency coverage and shorter observing time.
For Sgr A\*, we detect a peak in the Fourier spectrum of 0.17% at $`\mathrm{RM}=24000\pm 37000\mathrm{rad}\mathrm{m}^2`$. Imaging Sgr A\* with and without RM corrections, we find a fractional polarization of 0.1%. Off-source fractional polarizations are typically 0.1%, again implying that we are limited by residual instrumental polarization.
We tested noise models to see if we could reproduce a weak signal at non-zero RM. We used an input signal with $`\mathrm{RM}=0`$ at 0.15% of the peak intensity of Sgr A\* and noise that matched that of Sgr A\*. This is the model plotted in Figure 2. The measured RM peak wandered within the error range.
## 4 Interstellar Propagation Effects
The interstellar medium may depolarize a linearly polarized radio wave in two ways: significant rotation of the polarization position angle through the observing bandwidth; and, differential Faraday rotation along the many paths that contribute to the scatter-broadened image of Sgr A\*. We have already addressed the first effect in §2 and found in §3 that Sgr A\* is not depolarized by RMs less than $`1.5\times 10^7\mathrm{rad}\mathrm{m}^2`$. We now consider the second effect.
The scattering region will depolarize the signal if $`\delta \chi _F\pi `$. For our observing wavelengths, $`\delta \mathrm{RM}=900\mathrm{rad}\mathrm{m}^2`$ and $`\delta \mathrm{RM}=2400\mathrm{rad}\mathrm{m}^2`$. Over the scattering size of 50 mas at 4.8 GHz, this corresponds to $`\delta \mathrm{RM}/\delta \theta =18000\mathrm{rad}\mathrm{m}^2\mathrm{arcsec}^1`$.
Observed variations in RM in the GC region are many orders of magnitude below those necessary to depolarize Sgr A\*. Observations on the arcsecond to arcminute scale of a nonthermal filament within 1 degree of Sgr A\* find a maximum $`\delta \mathrm{RM}/\delta \theta =250\mathrm{rad}\mathrm{m}^2\mathrm{arcsec}^1`$ (Yusef-Zadeh, Wardle & Parastaran 1997). Extrapolation of the RM structure function to the scattering size implies $`\delta \mathrm{RM}50\mathrm{rad}\mathrm{m}^2`$. However, these observations are made on a much larger scale than the scattering disk of Sgr A\* and the scattering medium is believed to be inhomogeneous.
Could the more extreme conditions necessary to depolarize Sgr A\* exist in the Galactic Center scattering region? The RM is expressed as
$$\mathrm{RM}=0.8n_eBL\mathrm{rad}\mathrm{m}^2,$$
(4)
where $`n_e`$ is the electron number density in cm<sup>-3</sup>, $`B`$ is the magnetic field parallel to the line of sight in $`\mu \mathrm{G}`$ and $`L`$ is the size scale in pc. Since $`L`$ must be a fraction of the scattering diameter, we find $`L0.1\theta _{SgrA}D_{SgrA}10^4\mathrm{pc}`$. Yusef-Zadeh et al. (1994) argued that the photo-ionized skins of molecular clouds in the GC region have a similar length scale, milliGauss fields and $`n_e10^4\mathrm{cm}^3`$. This matches the depolarization condition if the regions are fully turbulent. However, if the constraints on the outer scale of turbulence derived by Lazio & Cordes (1998) are correct, then $`L10^7\mathrm{pc}`$. In this case, the RM condition and pressure balance between the magnetic and thermal components can only be satisfied if $`B10\mathrm{mG}`$ and $`n_e10^6\mathrm{cm}^3`$. These conditions are extreme even for the GC region. The largest magnetic fields as measured for OH masers are on the order of a few milliGauss (e.g., Yusef-Zadeh et al. 1996). Ionized densities measured for H II regions on the arcsecond scale ($`0.1`$ pc) are significantly less than $`10^5\mathrm{cm}^3`$ (Mehringer et al. 1993). No depolarization is predicted for the higher temperature and lower density model of Lazio & Cordes (1998) for $`B<1\mathrm{G}`$. The conditions necessary to depolarize at 8.4 GHz are even more extreme. We conclude, therefore, that the conditions necessary to depolarize Sgr A\* are unlikely to occur in the scattering region.
We consider now whether depolarization may occur in the accretion region of Sgr A\*, where the electron density and magnetic field strength are large but the length scale is smaller. If we consider the simplest model of spherical infall with $`\dot{M}=10^4M_{\mathrm{}}y^1`$ and equipartition between particle, magnetic and gravitational energy (Melia 1994), then the change $`\delta \mathrm{RM}`$ over an interval $`\delta r`$ at a radius $`r`$ from Sgr A\* is
$$\delta \mathrm{RM}=1.2\times 10^{14}r^{11/4}\delta r\mathrm{rad}\mathrm{m}^2,$$
(5)
where we have expressed $`r`$ and $`\delta r`$ in units of the gravitational radius $`r_g=2GM/c^2=7.8\times 10^{11}\mathrm{cm}`$ for a $`2.6\times 10^6M_{\mathrm{}}`$ black hole. This relation only holds for $`r10^3`$, where the temperature falls below $`10^9\mathrm{K}`$. The RM inside of this radius is negligible unless there is a separate population of cold electrons. We consider the effect of cold electrons in more detail in the following Section. If the scattering screen is at a distance of 100 pc from Sgr A\*, then the image will be an average of ray paths over a tangential length scale $`l2\times 10^5r`$. Fluctuations in the RM will depolarize Sgr A\* at a given radius if $`\delta \mathrm{RM}>900\mathrm{rad}\mathrm{m}^2`$ and $`l>l_0`$, where $`l_0`$ is the outer scale of turbulence. Assuming $`\delta rr`$, we find that depolarization will occur only if $`l_010^5\mathrm{pc}`$. Although this scale is much smaller than the outer scale in the local ISM (Armstrong, Rickett & Spangler 1995), it is a scale that may be pertinent to the dense, energetic environment of the accretion region.
## 5 An Intrinsically Weakly Polarized Sgr A\*
A polarization fraction less than 1% is uncommon in compact radio sources at wavelengths shorter than 6 cm (Aller, Aller & Hughes 1992). However, optically thick quasar cores observed with VLBI are frequently weakly polarized (Cawthorne et al. 1993). Such cores may be analogous to the radio source in Sgr A\*, which, due to its low power, may not produce the strong shocks in the jet that are the source for higher polarization regions in quasars. Weak polarization is more common in radio galaxies than quasars or blazars and it is also more common in compact-double sources or sources with irregular morphologies (Aller, Aller & Hughes 1992). A notable source with a very low polarization fraction ($`<0.1\%`$) is the radio galaxy 3C 84, which has a very irregular morphology.
If the radio to millimeter spectrum of Sgr A\* does arise in a jet, the low power of this jet or environmental effects in the Galactic Center region may limit the magnetic field order. For the case of a spherically symmetric emitting region, an ordered magnetic field may depolarize the source, as well. Alternatively, low energy electrons in the synchrotron environment of Sgr A\* may Faraday depolarize the source. The ADAF model and the Bondi-Hoyle accretion model predict the presence of non-relativistic electrons in the accretion region.
Observations at millimeter wavelengths may resolve many of the questions raised in this paper. Interstellar effects are reduced such that depolarization in the scattering region is extremely unlikely and depolarization in the accretion region must occur at radii less than 0.01 pc. Furthermore, the synchrotron emission arises from a more compact and presumably more homogeneous region. The source may also have less synchrotron self-absorption at millimeter wavelengths, although this is not required by all models. We will report in a future paper on millimeter polarimetric observations of Sgr A\*.
We thank Alok Patnaik for enlightening discussions. HF is supported by DFG grant Fa 358/1-1&2.
|
no-problem/9904/astro-ph9904202.html
|
ar5iv
|
text
|
# Investigating the Near-Infrared Properties of Planetary Nebulae II. Medium Resolution Spectra
## 1. Introduction
This is the second of two papers describing the results of surveys examining the properties of planetary nebulae (PNe) as observed in the near-infrared ($`\lambda =12.5`$ µm). The first paper (Latter et al. 1995; hereafter Paper I) presented an infrared imaging survey; here we present the results of a near-infrared spectral survey.
There are several reasons why knowledge of the near-infrared (near-IR) characteristics of PNe is important, as described in Paper I. In order to interpret the imaging results and to learn more about the physical conditions in the nebulae, the spectra of these objects must be examined to understand the processes responsible for the emission. There are many emission lines present in the 1 – 2.5 µm spectral region, most notably those due to recombination lines of H I, and lines of vibrationally excited H<sub>2</sub>. Also present are atomic lines of He I and \[Fe II\], and emission from other molecular species such as CO and C<sub>2</sub>. These lines can act as diagnostic tools to probe the physical conditions inside the nebula, sampling different regions and ranges of temperature, density, and excitation than is seen by observing the optical line emission. Some PNe also exhibit in the near-IR strong continuum emission from hot dust. This emission becomes significant longward of $`\lambda =2`$ µm in many of the PNe, requiring near-IR spectroscopy to detect it and to differentiate between line and continuum emission sources. Finally, the lower optical depth of the PNe in the IR as compared to optical wavelengths allows us to potentially see into regions of the nebula that are obscured by dust.
There have been several previous surveys that have explored the properties of PNe in the infrared. Early spectroscopic and photometric surveys (e.g., Gillett, Merrill, & Stein 1972; Cohen & Barlow 1974) determined that there was an excess of IR emission over what was expected from reflected continuum emission from the central star. Other photometric surveys in the following years (Whitelock 1985; Persi et al. 1987) determined the primary sources of IR emission to be stellar continuum, thermal dust emission, and thermal & line emission from the nebula itself. The near-IR color characteristics of most PNe are unique and can be used to identify new PNe and post-AGB objects (Garcia-Lario et al. 1990).
Recently there have been more detailed spectral observations of PNe in the near-IR. Hrivnak, Kwok, & Geballe (1994) surveyed a set of proto-PNe in the H and K bands. In these objects the H I Brackett lines were observed in absorption, and most objects had CO absorption or emission, indicating recent mass loss events. Rudy et al. (1992, and references therein) and Kelly & Latter (1995) have surveyed several PNe and proto-PNe in the $`\lambda =0.51.3`$ µm range. Dinerstein & Crawford (1998) have completed a survey of a set of PNe in the K-band, focusing on excitation of molecular hydrogen.
There are several unique aspects of the survey results presented here that were made possible by the KSPEC spectrograph (see the instrument description below). First, because of KSPEC’s high sensitivity and simultaneous sampling of the full near-infrared spectral range, we were able to obtain data on a comparatively large number of objects (41) in a short period of time. Using the relatively narrow and short slit of the spectrograph, we sampled different regions of the PNe to examine the emission throughout the nebula. In most of the other surveys described above, larger beams were used that included much or all of the object. Another aspect of the data presented here is because of the cross-dispersed design of the instrument, the entire $`\lambda =1.12.5`$ µm range is obtained at once, eliminating the possibility of telescope pointing errors or other fluctuations affecting the relative line strengths in the spectra. Finally, the slit-viewing detector allowed precise positioning and guiding, so the region of the PN being observed was well known for each spectrum. The PNe observed in this survey were chosen to overlap with the near-IR imaging survey (Paper I), along with several other optically-bright PNe and unresolved objects that were not included in the imaging survey.
## 2. Observations and Data Reduction
The observations were performed on several runs during the period 1992 October through 1994 September at the University of Hawaii 2.2m telescope on Mauna Kea, using the near-IR KSPEC spectrograph (Hodapp et al. 1994). KSPEC is a $`\lambda =12.5`$ µm cross-dispersed spectrograph that has a separate slit-viewing IR array for acquiring the source and guiding. The full spectral range is obtained in a single exposure, resulting in accurate relative line measurements and highly efficient data acquisition. The diffraction orders are well-matched to the atmospheric transmission windows, with the K band in 3rd order (1.9 – 2.5 µm), H in 4th order (1.45 – 1.8 µm) and J in 5th order (1.15 – 1.32 µm), with a resolution R $``$ 700. The 1$`\times `$6 arcsecond slit provides a small aperture that was used to sample different spatial regions of the nebula to search for spectral variations.
Table 1 lists the nebulae observed and details of the observations. Integration times for each exposure ranged from a few seconds for the extremely bright sources to 5 minutes for faint sources. The off-axis guider was used to keep a consistent on-source slit position. Multiple “Fowler” sampling was used to reduce the read noise. The number of samples for a particular integration ranged from 4 to 16, with more samples used for the longer integration times. The spectra were reduced using IRAF; the extraction and processing of the spectral data were done using the functions in the noao.twodspec and noao.onedspec packages. Alternating source and sky integrations of the same length were taken and differenced to remove sky and telescope background flux. Dome flats were used to correct for pixel-to-pixel gain variations in the array. Stars of known spectral type (either G0 or A0) were observed at the same airmass as the nebulae immediately before and after the PNe observations and were used for correction of the instrumental response and sky transmission. Individual lines were removed from the stellar spectra, and then normalized using a blackbody function of T = 5920 K for the G0 stars and T = 10800 K for the A0 standards. The spectra were wavelength-calibrated using observations of an Argon reference lamp. The wavelength values used for lines greater than 1.1 µm are from Rao et al. (1966); for lines less than 1.1 µm, the wavelengths were taken from Wiese, Smith, & Miles (1969) and corrected to vacuum wavelengths. The average 1$`\sigma `$ uncertainty in the measured wavelengths of the lines is about 5 Å.
Infrared photometric standard stars were observed in the same way as the PNe and used to flux calibrate the spectra. Absolute calibration is difficult with these spectra because not all of the flux from the star enters the narrow slit during a single integration, and the amount differs for each integration depending on how well the star is centered on the slit (the seeing at 2 µm during typical observations was 0$`\stackrel{}{\mathrm{.}}`$5 – 1$`\stackrel{}{\mathrm{.}}`$0). The amount of light lost was estimated by the following method: the full width at half maximum (FWHM) brightness of the standard star was measured along the spatial direction of the slit, in the spectrum with maximum flux for that star. It was then assumed that the point spread function (PSF) is well represented by a two-dimensional Gaussian distribution with the measured FWHM, and the amount of flux falling outside of the slit was then calculated. This was typically 20 – 30% of the total light for a single integration. The calibration for each star was corrected by this factor, along with corrections for airmass. Comparing results from different standard stars taken throughout the night indicated that this method is accurate to approximately 20%. For the observations of the PNe, no correction was applied for the slit width or length. The length along the slit of the extracted regions for the three bands were 2$`\stackrel{}{\mathrm{.}}`$0 (J), 3$`\stackrel{}{\mathrm{.}}`$5 (H), and 4$`\stackrel{}{\mathrm{.}}`$0 (K).
## 3. Results
The spectra are presented in Figures 1 – 32. One to three PNe spectra are plotted in each figure. Tables 2 through 7 list the line identifications and extracted fluxes with uncertainties for the spectra shown in the Figures. The PN in this section are listed in Tables 2, 3, and 4. There are a number of features not identified (indicated by question marks in the tables). These features tend to appear above the $`3\sigma `$ level and must be considered real, though confirming spectra would be valuable. Our search for possible identifications for these lines has been careful, but perhaps not exhaustive. In addition to the relatively low S/N of these lines, the moderate wavelength resolution is not sufficient to differentiate between the several possible identifications for each line.
The PNe spectra are separated into four groups that share common characteristics. These are H I recombination line-dominated, H I recombination line + H<sub>2</sub> emission, H<sub>2</sub> – dominated, and continuum-dominated. A fifth group of objects is included that contains two objects that were at one time classified as PNe, but are now generally regarded as being H II regions (M 1–78 and K 4–45; Acker et al. 1992). We do not discuss these objects further, but include them for comparison. Within each group, the NGC objects are listed first, followed by the remaining PNe in alphanumeric order. The morphological classifications are given according to Balick (1987), unless otherwise noted.
### 3.1. H I - line dominated
The line emission in these PNe is dominated by lines of H I and He I. In the J-band, the Paschen $`\beta `$ (Pa$`\beta `$) line is the most intense, with contributions from lines of He I, \[Fe II\], and O I. In the H band, the Brackett series of H I dominate, with He I emission at 1.7002 µm and \[Fe II\] emission at 1.6440 µm present in some PNe. In the K band, the brightest line is usually Brackett $`\gamma `$ (Br$`\gamma `$), with strong lines of He I at 2.058 and 2.112 µm. When the H I lines are strong enough, one begins to see the Pfund series lines starting near 2.35 µm where they are just beginning to be separated at this resolution. There are also two unidentified lines at 2.199 and 2.287 µm (Geballe et al. 1991) that appear in several PNe in this category. A few of the spectra shown here have contributions from central star continuum flux that is larger towards shorter wavelengths, or warm dust continuum which is stronger at longer wavelengths.
#### 3.1.1 NGC 1535
NGC 1535 is classified as early round, and its near-IR spectrum is dominated by emission lines of H I. It has a bright ionized shell of emission which is surrounded by a fainter halo (e.g., see Schwarz, Corradi, & Melnick 1992). Near-IR images were presented in Paper I. This PN has previously been observed to have H<sub>2</sub> lines in absorption in the far-UV (Bowers et al. 1995). Recent observations by Luhman et al. (1997) failed to detect H<sub>2</sub> in emission in the $`v=10`$ S(1) line. They attributed the earlier detection of absorption at shorter wavelengths to the interstellar medium, or a region in the PN itself of a much smaller size than the ionized zone.
The spectrum of NGC 1535 shown in Figure 1 was obtained at a position centered on the brightest part of the ring directly west (W) of the central star. We also fail to detect the H<sub>2</sub> emission in the $`v=10`$ S(1) line, at a 1 $`\sigma `$ level of about $`5\times 10^{17}`$ ergs cm<sup>-2</sup> s<sup>-1</sup> Å<sup>-1</sup>. There is some indication of emission from H<sub>2</sub> in the $`v=10`$ Q(1) and $`v=10`$ Q(3) lines at the long wavelength end of the spectrum. However, the spectrum is noisier in this region and there is confusion with the Pfund-series H I lines, therefore the H<sub>2</sub> lines are not detected above the 3 $`\sigma `$ level.
#### 3.1.2 NGC 2022
NGC 2022 is an early elliptical PN, but is morphologically very similar to NGC 1535 in optical images (e.g., Schwarz et al. 1992). The main difference between the two is a different relative outer halo size as compared to the inner ring (the halo is relatively smaller in NGC 2022). Zhang & Kwok (1998) also find similar parameters with their morphological fits of these two PNe. Near-IR images of this PN were presented in Paper I.
NGC 2022 is also spectrally similar to NGC 1535, as seen in Figure 1. The spectrum of NGC 2022 was taken centered on the ring directly east (E) of the central star. The dominant emission lines are those of atomic hydrogen.
#### 3.1.3 NGC 2392
The PN NGC 2392 (the “Eskimo nebula”) is another double-shell nebula; however, this PN has a significant amount of structure in the inner ring and outer shell. Spectrophotometric (Barker 1991) and kinematic studies (Reay, Atherton, & Taylor 1983; O’Dell, Weiner, & Chu 1990) that have been carried out with optical imaging and spectroscopy have revealed the abundances and ionization states and velocities of the various components. Near-IR images of this PN were presented in Paper I.
The spectrum of NGC 2392 in Figure 1 was taken centered on the brightest part of the ring directly E of the central star. This third early-type round PN differs from the other two in Figure 1 primarily from the bright He I line at 2.058 µm, and the \[Fe II\] lines in the J and H bands of the spectrum.
#### 3.1.4 NGC 3242
NGC 3242 is an early elliptical with several interesting morphological features. In addition to the bright elliptical ring, there are several filaments and knots of emission in the central region, and two ansae that are placed roughly along the major axis of the elliptical emission. Also, there is a larger faint halo that envelopes the inner structure.
Spectra acquired at three different positions on the nebula are shown in Figure 2, on the SE knot (NGC 3242SE), on the E section of the bright ring (NGC 3242E), and on the SW halo (NGC 3242H). The spectra are similar in all locations; the bright H I lines are present in all positions, along with stellar continuum at the shorter wavelengths. One difference is that the He II line at 2.189 µm are much brighter in the E ring than in the SE knot or halo position.
#### 3.1.5 NGC 6210
NGC 6210 is a fairly compact PN with a core – halo morphology similar to other ellipticals. Phillips & Cuesta (1996) performed a visible wavelength spectroscopic study that revealed a complex velocity structure which suggests multiple shells and possibly “jets” at various position angles. The near-IR image presented in Paper I does not reveal much of the structure. This PN is one of several in which Geballe et al. (1991) detected unidentified emission at 2.286 µm (but not at 2.199 µm).
Three spectra are presented for this PN, and shown in Figure 3. They were acquired with the slit centered on the core (Core), 1″ east (E1), and 3″ east (E3). The core position shows a contribution from stellar continuum in the $`\lambda =11.8`$ µm region that decreases successively in the E1 and E3 positions. The unidentified feature at 2.287 µm is detected at the E1 position, but not the 2.199 line, similar to the findings of Geballe et al. (1991).
#### 3.1.6 NGC 6543
NGC 6543 (the “Cat’s Eye”) has a complex morphology, with a high degree of symmetry. Recent HST imaging (Harrington & Borkowski 1994) has shown more clearly the structure of rings, shock fronts, jets, and fast, low-ionization emission-line regions (FLIERS) present in this PN.
The spectrum shown in Figure 4 was taken at the position S of the central star and slightly E, where the two emission arcs cross and create a local emission maximum (see Paper I, Figure 6a). The spectra is similar to the other PN in the H I/He I - dominated class, and also show the unidentified lines at 2.199 and 2.286 µm.
#### 3.1.7 NGC 6572
This young PN has a bipolar morphology in the near-IR, with its major axis in the N-S direction, and a bright ring structure closer to the central star (Paper I). The near-IR spectrum from 0.77 to 1.33 µm was measured by Rudy et al. (1991), and the UV and optical spectrum was recently obtained by Hyung, Aller, & Feibelman (1994), who found evidence of variability.
Figure 5 shows two slit positions on the PN, one on the nebula center, and one on the brightest location in the E lobe of the PN. Both the core and E lobe spectra show a strong contribution from stellar continuum flux. There is strong H I and He I line emission, and relatively strong unidentified line emission at 2.199 and 2.286 µm.
#### 3.1.8 NGC 6790
This PN has been shown by radio continuum observations to be an elliptically-shaped nebula with a diameter of roughly an arcsecond (Aaquist & Kwok 1990). Aller, Hyung, & Feibelman (1996) obtained UV and optical spectra and suggest that NGC 6790 is a relatively young object, slightly more evolved than Hb 12. Kelly & Latter (1995) obtained a 0.9 – 1.3 µm spectrum and find similarities to Hb 12 and AFGL 618.
The spectrum of NGC 6790 in Figure 4 shows a stellar contribution because the slit was centered on the object, and includes the star as well as the nebular emission that is typical of this class of PN.
#### 3.1.9 NGC 6803
This compact elliptical PN has a uniformly bright disk with no apparent structure in the optical, and is surrounded by a fainter halo about twice the size of the bright shell (Schwarz et al. 1992). In the KSPEC imaging channel, however, the PN was seen to be double-lobed, with the lobes on the minor axis of the bright elliptical region. Spectra were obtained at two positions, one centered on the E lobe, and the other position 5 arcseconds NW in the halo region. The lobe spectrum shows bright H I and He I lines, with a stellar continuum contribution out to about 1.5 µm. There is also unidentified line emission at 2.286 µm. The halo spectrum is quite different, with weak Pa$`\beta `$ and continuum emission, and is probably reflected star and nebular emission.
#### 3.1.10 NGC 6826
This elliptical PN is morphologically similar to NGC 3242, with a bright elliptical inner ionized ring, ansae along the major axis, and a fainter halo that envelopes the system (Balick 1987). The I-band image in Paper I shows evidence for a shell between the inner bright ring and the outer halo, and at the same radial distance as the ansae.
The spectra are shown in Figure 7. The bright core is dominated by stellar continuum emission. The other positions show strong H I line emission. The spectrum labeled SW Lobe was taken on the inner bright ring directly SW of the central star. The SW Halo position was taken in the halo midway between the bright ring and the outer edge of the PN. The lobe and halo emission is similar, except for relatively brighter lines of He I at 1.7002 µm in the lobe. There is also some continuum emission at the short wavelength end of the nebular positions, which is probably scattered light from the central star.
#### 3.1.11 NGC 7009
NGC 7009 (the “Saturn” nebula) is an elliptical with an interesting twisted symmetry in its shell and in the various filaments and knots of emission. Balick et al. (1998) recently published HST images that show the “microstructures” in this PN. The images show that the inner knots are actually groups of FLIERs, and jets in \[N II\] are seen that terminate at the tips of the nebula.
Two positions were sampled, one in the halo region on the W edge of the PN, and one in the N part of the nebula. Both positions show a stellar continuum emission contribution from the central star. The halo emission is similar to the spectrum taken on the north edge of the PN.
#### 3.1.12 NGC 7662
NGC 7662 is a triple-shell elliptical, with a bright inner ring, a fainter outer shell, and a very faint nearly circular halo (Hyung & Aller 1997). This PN also has a large number of complex microstructures, recently examined using HST imaging by Balick et al. (1998). They suggest that there is a prolate elliptical bubble around the central star aligned perpendicular to the bright ring. The bright ring is interpreted as a torus seen at roughly 30 inclination.
The spectrum shown in Figure 9 was taken on the bright ring directly SE of the central star. The N end of the slit was near the central star, so some stellar continuum is seen in the spectrum, which is otherwise dominated by H I and He I lines.
#### 3.1.13 IC 351
This compact PN has a double-lobed structure with a round halo (Hua & Grundseth 1986; Aaquist & Kwok 1990; Manchado et al. 1996). Feibelman, Hyung, & Aller (1996) obtained UV and visible light spectra of IC 351 that show it to be a high excitation nebula, but without the presence of the usual silicon lines, and suggest that the silicon atoms could be locked up in grains.
Our spectrum (Figure 9) taken centered on the PN suffers somewhat from an incomplete subtraction of OH airglow lines, due to the sky frames not being taken properly for the on-source images. The OH lines show up in emission mainly in the H and K-band portions of the spectrum. However, the major features of H I and He I emission lines can be seen.
#### 3.1.14 IC 418
The spectrum of this well-studied young, low-excitation PN was previously shown to be dominated by lines of H I and He I in the near-IR, with a hot dust continuum (Willner et al. 1979; Zhang & Kwok 1992; Hodapp et al. 1994). Hora et al. (1993) and Paper I showed broad- and narrow-band near-IR images of the PN, showing the elliptical, double-lobed structure in the IR.
Three positions in the nebula were observed to determine the spectral variations across the object. The positions observed were on the central star, the peak of the E lobe, and in the E halo region outside of the bright ring (Figure 10). In the central position, stellar continuum is visible, rising towards shorter wavelengths. The nebular lines are similar in the central and lobe positions. The halo emission is almost devoid of lines; there is some faint Pa$`\beta `$ present as well as Br$`\gamma `$. This is possibly reflected from the bright lobes. The main component of the halo emission is a weak continuum that rises toward longer wavelengths.
#### 3.1.15 IC 2149
This peculiar PN has a bright core and a roughly bipolar nebula extending approximately E-W, but does not show the usual H<sub>2</sub> signature of bipolar PN. The two spectra shown in Figure 11 were taken centered on the bright central star, and on the E lobe. The core spectrum shows strong stellar continuum, with the nebular lines superimposed. The E lobe emission is primarily from lines of H I and He I, in addition to weak continuum emission which is probably reflected from the central star.
#### 3.1.16 IC 3568
This round PN consists of spherical shells, an inner bright one and a outer halo. Balick et al. (1987) showed that the structure was consistent with simple hydrodynamic models of PN that are shaped by interior stellar winds. The spectrum taken on the N edge of the PN is shown in Figure 12. The predominant features are H I and He I emission lines, and low level continuum emission which is probably reflected from the central star.
#### 3.1.17 IC 4593
Bohigas & Olguin (1996) obtained spectroscopy and imaging of this PN which has two inner shells surrounded by an outer highly excited halo. IC 4593 is unusual in that the condensations outside of the inner region are located asymmetrically in the SW region.
Two spectra were taken on this PN, one positioned on the central star, and the other at a position 3″ E (Figure 12). The core shows bright stellar continuum, with nebular lines superimposed, and Pa$`\beta `$ absorption. The E spectrum has the typical H I and He I emission lines. The sky subtraction was not of high quality for this spectrum, which resulted in OH airglow lines showing up in emission in the H and K spectral regions.
#### 3.1.18 J 320
Images of J 320 (Balick 1987) show it to have a central region that is elongated roughly E-W, but the low level flux has a N-S elongation indicating a shell or streamer extending in this direction. Three spectra for J 320 are shown in Figure 13, centered on the core (C), offset 1″ N, and offset 1$`\stackrel{}{\mathrm{.}}`$5 N. The core has a stronger contribution from stellar continuum, but otherwise the spectra are similar. We therefore detect no spectral differences between the N extension and the nebula near the core.
#### 3.1.19 M 4–18
This young, low-excitation PN was recently imaged with HST by Dayal et al. (1997) and shows a toroidal shell surrounding the central star. There is strong mid-IR emission from warm ($``$ 200 K) dust that has a similar morphology but the position angle of the dust emission maxima is orthogonal to those shown in H$`\alpha `$ emission. The spectrum of the core of this PN shown in Figure 14 has strong stellar continuum, as well as H I and He I line emission. The slit was positioned N-S across the compact ring of the PN, so the ionized regions of the nebula are included in this spectrum.
### 3.2. H I - line and H<sub>2</sub> emission
The emission lines present in the spectra of this PNe group contain those mentioned in the previous section plus lines of molecular hydrogen. The H<sub>2</sub> lines are strongest in the K band, although in some objects there are lines visible in the H and J bands as well. In the objects where more than one slit position was measured, there is often a large change in the relative line strength of H I versus H<sub>2</sub> emission, indicating that the emission is being produced in different regions of the nebula. In general, the H<sub>2</sub> emission is more likely to be in the outer regions of the PNe, whereas the H I emission lines more closely trace the ionized regions and has similar morphology to the visible appearance. The PNe in this group all have some bipolar symmetry in their shape, most being of the “butterfly” morphology characterized by narrow equatorial regions and large bipolar lobes (e.g., M 2–9, Hb 12). However, some are classified as elliptical based on the shape of the brightest components. Tables 5 and 6 list the line identifications and extracted fluxes for the PNe in this section.
#### 3.2.1 Molecular hydrogen excitation in PNe
A near-infrared H<sub>2</sub> emission line spectrum can occur through slow electric quadrupole vibration-rotation transitions in the ground electronic state. Because the allowed transitions are such that $`T_{\mathrm{ex}}1000`$ K is required to produce a detectable near-IR H<sub>2</sub> spectrum, special excitation conditions must exist when the near-IR spectrum is present. In Paper I we discussed mechanisms of H<sub>2</sub> excitation in PNe; see also Kastner et al. (1996). If detected in sufficient number, the observed H<sub>2</sub> line ratios are an excellent diagnostic for determining the relative importance of shocks and UV photons in a photodissociation region (PDR) for the excitation of the H<sub>2</sub> emission. Even if the excitation mechanism cannot be determined, the presence of H<sub>2</sub> emission is important to understanding the conditions in PNe and how they evolve through wind interactions and photodissociation.
An analysis of near-IR H<sub>2</sub> emission can determine an ortho-to-para (O/P) ratio, the rotational excitation temperature $`T_{ex}(J)`$, and the vibrational excitation temperature $`T_{ex}(v)`$ of the molecules. If the rotational and vibrational excitation temperatures differ, then UV excitation is indicated. This is most readily determined by comparing the column densities in the upper state vibration-rotation levels with the upper state energy (in temperature units; see Hora & Latter 1994, 1996 for a full discussion; see also Black & van Dishoeck 1987; Sternberg & Dalgarno 1989). When many H<sub>2</sub> lines are detected, especially those from highly excited levels that fall in the J band, a much stronger case can be made for the importance of UV excitation than does the traditional $`v=21`$ S(1) to $`v=10`$ S(1) line ratio (e.g. Hora & Latter 1994, 1996). The comparison of the column densities to the upper state energy levels has been done for the PNe with detected H<sub>2</sub> emission and the results are summarized in Table 8. Only the rotational excitation temperatures are listed. For collisionally- (shock-) excited spectra, the rotational and vibrational excitation temperatures are coupled and the same. For UV excited spectra, the vibrational temperature is the result of a cascade through levels and not a thermal process. The rotational levels are easily thermalized by collisions. The observed O/P ratio is in general rather uncertain, especially if only $`v=1`$ lines are detected. We did not attempt to determine the O/P ratio for objects without sufficient line detections. An observed O/P ratio lower than 3 indicates that the H<sub>2</sub> emission is not thermally excited. A subthermal O/P ratio as determined from near-IR spectra is not caused by the UV excitation process itself, but is a function of chemistry and density in the PDR (e.g., Hora & Latter 1996; Black & van Dishoeck 1987), and might not be indicative of the “true” O/P abundance ratio of the H<sub>2</sub> (see Sternberg & Neufeld 1999). Selected excitation diagrams for several objects that are discussed below are shown in Figure 33.
#### 3.2.2 NGC 40
The morphological classification of NGC 40 is middle elliptical (Balick 1987), which seems to contradict the previously observed strong correlation between bipolar morphology and H<sub>2</sub> detection. However, if one examines the low-level emission in the N and S regions of this PN (see Paper I), one can see material that has broken through and expanded beyond the elliptical shell defined by the E and W bright lobes. Mellema (1995) has found the morphology consistent with models of “barrel”-shaped PNe, which have roughly cylindrical emission regions slightly bowed outwards at the equatorial plane, and less dense polar regions. Higher resolution and more sensitive optical imaging has recently been carried out by Meaburn et al. (1996) show gas escaping from the polar regions of the PN, with other filamentary structure in the outer halo.
This is the first reported detection of H<sub>2</sub> in NGC 40. The spectrum was taken centered on the W lobe, and the H<sub>2</sub> lines are relatively weak compared to the H I and He I lines from the ionized gas in this region. The H<sub>2</sub> emission was not detected in the narrowband imaging surveys of Paper I or Kastner et al. (1996), so the molecular emission must be confined to a region near the bright ionized gas that dominates the spectrum. The data suggest that the H<sub>2</sub> is shock-excited. However, insufficient line detections make this result less than firm.
#### 3.2.3 NGC 2440
NGC 2440 is a bipolar PN with complex morphological and spectral structure. In the optical, the nebula is bipolar with the major axis in roughly the E-W direction for the large outer lobes (Balick 1987; Schwartz 1992). However, there are two bright lobes near the core that are positioned along an axis roughly perpendicular to the major axis of the outer lobes. There are two fainter knots that are also along a roughly E-W axis, but not aligned with the outer lobes. There are filaments and knots throughout the lobes. Lopez et al. (1998) finds up to three outflowing bipolar structures in the lobes, and find from their kinematic study that the inner bright lobes (their lobes “A” and “B”) are the emission maxima from a radially-expanding toroid viewed nearly in the plane of the sky.
In the near-IR, the inner pairs of lobes are also prominent, but the large E-W lobes are not visible (see Paper I). Instead, there is a circular outer halo visible in H<sub>2</sub> that is not quite centered on the inner lobe structure. Also visible are faint H<sub>2</sub> “spikes” that extend from the center to the circular outer halo, roughly in the equatorial plane of the large optical E-W lobes (Latter & Hora 1998).
The spectra shown in Figures 15 and 16 were taken at three different positions in the PN: on the N lobe (of the innermost bright pair of lobes), on the fainter E knot, and on a clump of H<sub>2</sub> emission located on the NE edge of the outer circular halo (see Paper I, Figure 4a; it is the clump visible at the upper left corner of the “H<sub>2</sub> sub” image). The N lobe exhibits H I and He I lines from the ionized gas in this region, but also has significant H<sub>2</sub> emission. There is also strong \[Fe II\] emission at 1.64 and 1.257 µm. The E knot also displays similar H I, He I, \[Fe II\], and H<sub>2</sub> emission, although fainter. In contrast to the inner regions, the NE clump spectrum in Figure 16 is dominated by H<sub>2</sub> emission, with the only H I lines detected being Pa$`\beta `$ and Br$`\gamma `$. There is strong \[Fe II\] emission at 1.64 and 1.257 µm in this region as well. The excitation analysis for the three positions observed showed that they are UV-excited, except for the E knot position for which there is insufficient data. The low value of the observed O/P ratio is suggestive of the H<sub>2</sub> emission arising from a PDR at this location as well.
Since the inner region of NGC 2440 is morphologically complex and any line of sight through the PN is likely to intersect several distinct regions, it is probably the case that the ionized and molecular zones are not mixed as the spectra might seem to indicate, but that the slit simply includes several nebular components, or is looking through a PDR and is sampling both the molecular and the recently ionized gas.
#### 3.2.4 NGC 6720
NGC 6720 (the “Ring Nebula”) is probably the best-known PN, and is the archetype for the ring or elliptical morphology that characterizes the brightest part of the nebula. The emission is not consistent with a uniform prolate shell, however, since the ratio of flux between the edge and center of the ring is higher than expected from a limb-brightened shell (Lame & Pogge 1994). Balick et al. (1992) have suggested that NGC 6720 is actually a bipolar PN viewed along the polar axis, based on narrow-band imaging and high-resolution spectroscopic observations. This view is supported by the presence of H<sub>2</sub> in the nebula and halo, which correlates strongly with bipolar morphology. Guerrero, Manchado, & Chu (1997) draw different conclusions, however, based on their chemical abundance and kinematic study of the nebula. They argue that the Ring has a prolate ellipsoid structure, with a halo of remnant red giant wind.
Our spectra of the Ring (Figure 17) were obtained at two positions, one on the bright ring directly N of the central star, and the second position several arcseconds further north, off the bright ring but on a moderately bright (in H<sub>2</sub>) position in the halo. Both positions show bright H<sub>2</sub> emission, with the lobe position also showing contributions from emission lines of H I and He I from the ionized gas, as one would expect based on the visible wavelength and IR images showing the distribution of the line emission. The ring spectrum is strongly UV excited, indicating it is the PDR interface to the outer molecular shell.
#### 3.2.5 NGC 7026
The late elliptical PN NGC 7026 has two bright lobes on either side (E-W) of the central star, with fainter bipolar emission extending roughly N-S from the core. Cuesta, Phillips, & Mampaso (1996) obtained optical spectra and imaging of this object and found kinematically complex structure, with several separate outflows at the outer edge of an inner spherical shell, and suggested that the primary shell may be undergoing breakup in transition to a more typical bipolar outflow structure.
Two positions were sampled in NGC 7026, shown in Figure 18, centered on the E and W bright lobes near the central star. The lobe spectra are nearly identical, as one might expect from the symmetry in this PN. The H<sub>2</sub> emission in this PN is fairly weak at these positions. This might be due to the H<sub>2</sub> being concentrated in other regions of the PN, and not in the bright ionized lobes that were sampled by the spectra presented here. We are unable to determine the excitation mechanism.
#### 3.2.6 NGC 7027
NGC 7027 is one of the most highly studied PN at all wavelengths, particularly in the infrared because of its brightness and wealth of spectral features. Treffers et al. (1976) obtained a spectrum for $`\lambda =0.92.7`$ µm with a beam that included the entire nebula. They identified the major near-IR spectral components, including the first detection of H<sub>2</sub> lines in a PN, and the first detection of the unidentified line at 2.29 µm. Since then, several near-IR spectra have been published, including Scrimger et al. (1978), Smith, Larson, & Fink (1981), Rudy et al. (1992), and Kelly & Latter (1995).
The spectra shown in Figure 19, one taken centered on the W bright lobe, and the other at the brightest position in H<sub>2</sub> of the NW lobe (see Paper I). Both show H I and H<sub>2</sub> emission; the H<sub>2</sub> is relatively stronger in the NW position than in the bright lobe. Narrowband imaging has shown that the H I and He I emission is primarily in the bright inner ring of the nebula, and the H<sub>2</sub> emission is in what appears to be bipolar lobes outside of this shell (Graham et al. 1993a,b; Paper I; Latter et al. 1998). It has been argued before based on morphology that the H<sub>2</sub> is in a PDR (Graham et al. 1993a). Our data clearly demonstrate this to be the case, with the H<sub>2</sub> showing a strongly UV excited spectrum in a relatively high density medium (see Figure 33b).
#### 3.2.7 BD+303639
The young PN BD+303639 is well-studied in the infrared, and is remarkable primarily because of its large IR emission excess. It has many similarities to NGC 7027, including its IR morphology and the presence of H<sub>2</sub> in the near-IR and unidentified IR (UIR) emission features in the mid-IR spectrum, which are usually attributed to polycyclic aromatic hydrocarbons (PAHs). Rudy et al. (1991) obtained a $`\lambda =0.461.3`$ µm spectrum of BD+303639; high-resolution visible and near-IR images were recently obtained by Harrington et al. (1997) and Latter et al. (1998), and ground-based near- and mid-IR images have been presented by Hora et al. (1993), Paper I, and Shupe et al. (1998).
Three positions in BD+303639 were sampled in the spectra presented in Figure 20; the emission peak on the N lobe of the ring, the E side of the ring, and on the H<sub>2</sub> emission region located approximately 3″ E of the ring. These spectra show a steady progression of decreasing emission from the ionized gas and increasing molecular emission as one moves east. As for NGC 7027, the H<sub>2</sub> emission in BD+303639 is UV excited (Figure 33a) and defines the PDR (see also Shupe et al. 1998).
#### 3.2.8 Hubble 12
Hubble 12 (Hb 12) has been notable primarily because it represents one of the clearest cases known of UV excited near-IR fluorescent H<sub>2</sub> emission (Dinerstein et al. 1988; Ramsay et al. 1993). Our Hb 12 results from this survey and our imaging survey were presented in a previous paper (Hora & Latter 1996); the spectra are reproduced here for comparison with the rest of the survey.
Dinerstein et al. had mapped the inner structure and found it to be elliptical surrounding the central star; our deep H<sub>2</sub> images showed the emission to be tracing the edges of a cylindrical shell around the star, with faint bipolar lobes extending N-S. We also detected \[Fe II\] line emission at 1.64 µm in a position along the edge of the shell. The H<sub>2</sub> line ratios observed were in excellent agreement with predictions by theoretical H<sub>2</sub> fluorescence calculations, and no significant differences were found between the excitation in the two positions of the nebula that were sampled (see also Luhman & Rieke 1996).
#### 3.2.9 IC 2003
IC 2003 is a round, high-excitation PN that has a ring of emission, with a bright knot on the S edge (Manchado et al. 1996; Zhang & Kwok 1998). Feibelman (1997) obtained IUE spectra of this PN that shows a wealth of nebular and stellar lines. The IR spectra presented in Figure 22 taken in the center of the PN shows that there is little continuum from the nebula; the emission is primarily from lines of H I in the J, H, and K bands. There is strong unidentified emission at 2.286 µm but none detected at 2.199 µm. H<sub>2</sub> emission is tentatively detected in the K-band, in the $`v=10`$ S(1), $`v=32`$ S(1), and $`v=10`$ Q(1) lines. Each of the lines are detected at roughly a 2$`\sigma `$ level. The line fluxes are not reliable or numerous enough to allow fitting of the line ratios.
#### 3.2.10 IRAS 21282+5050
The young, carbon-rich PN IRAS 21282+5050 has been identified as having an 07(f)-\[WC11\] nucleus (Cohen & Jones 1987) with possibly a binary at its center. Strong <sup>12</sup>CO has been detected in a clumpy expanding shell (Likkel et al. 1988) with elongated emission N-S. Shibata et al. (1989) believe the elongated emission suggests the presence of a dust torus in the E-W direction; however, Meixner et al. (1993) was evidence for a clumpy, expanding elliptical envelope. The elongated structure is also seen in the visible (Kwok et al. 1993). Weak continuum flux at 2 and 6 cm suggests a young PN just beginning to be ionized (Likkel et al. 1994; Meixner et al. 1993). Kwok et al. (1993) believe there has been a recent sharp drop in luminosity based on the measured CO/FIR ratio. Weak HCO<sup>+</sup> and <sup>13</sup>CO are present (Likkel et al. 1988).
Two positions were sampled on IRAS 21282+5050, centered on the bright core, and offset approximately 3″ N and 3″ W. The spectrum of the offset position is shown in Figure 22. The core is dominated by continuum emission from the central star. Also present are both emission lines from the ionized gas, and H<sub>2</sub> features in the K-band. The nebula is compact, about 4″ in diameter at K (Paper I). The slit therefore samples a slice through the entire nebula, and as a result this spectrum does not necessarily imply that the molecular and ionized gas is mixed. The Lobe spectrum shows primarily lines of H<sub>2</sub> (with the OH night sky lines showing up in absorption because of imperfect sky subtraction in this spectrum). The lack of emission lines due to H I and He I in the Lobe spectrum indicates that the H<sub>2</sub> emission is predominantly in the outer regions of the PN. The data are suggestive of shock excitation, but this should be considered tentative.
#### 3.2.11 M 1–16
M 1–16 is a PN with a near-IR bright central region and bipolar lobes with fast winds extending at least 35″ from the core (Schwartz et al. 1992; Aspin et al. 1993; Sahai et al. 1994). Several spectra were obtained in this PN scanning across the central region; the two positions shown in Figure 23 are on the core position and 1″ S of the core. Both positions show H<sub>2</sub> emission; the S position is slightly brighter in both H<sub>2</sub> and the ionized nebular lines. Our data reveal that the H<sub>2</sub> is UV excited in both regions observed. This had be suggested earlier by Aspin et al. (1993). The core shows a slight rise towards long wavelengths indicating emission from warm dust continuum.
#### 3.2.12 M 1–92
M 1–92 (“Minkowski’s Footprint”) is a bipolar proto-planetary nebula similar in near-IR appearance to AFGL 618, and has evidence of highly collimated outflows along the bipolar axis (Paper I; Trammell & Goodrich 1996 and references therein). Two positions were sampled in M 1–92, one in the core and one on the NW bipolar lobe. There are problems with the sky background subtraction in both spectra, which are most prominent in the $`\lambda =1.92.1`$ µm region of the spectrum, but also contribute to a lower signal to noise ratio (S/N) over the whole dataset. Nevertheless, the primary characteristics are apparent. The core region is dominated by strong warm dust continuum emission. There is also weak Br$`\gamma `$ and Pa$`\beta `$ emission, but the other most other H I and He I lines are too weak to be detected. The lobe position shows weak H<sub>2</sub> emission. The emission appears to be shock-excited, but the low excitation suggested by our data is suggestive of UV excitation. Data of higher S/N are required to discern the dominant excitation mechanism. There might also be H<sub>2</sub> emission near the core that is being masked by the strong continuum emission. In both positions, there also seems to be CO bandhead emission at $`\lambda =2.32.5`$ µm although the S/N is not high in these regions.
#### 3.2.13 M 2–9
M 2–9 (the “Butterfly”) is a highly symmetric bipolar nebula, with lobes extending from opposite sides of a bright central core, nearly in the plane of the sky. Bright knots of emission are visible in the lobes at the N and S ends. Our results for M 2–9 from this survey were previously presented in Hora & Latter (1994), and some of the spectra are reproduced in Figures 25 and 26 for comparison. High-resolution imaging in several near-IR lines indicated that the lobes had a double-shell structure, with the inner shell dominated by H I and He I line emission from ionized gas and continuum emission scattered from the central source, and the outer shell (the “Lobe O” spectrum in Figure 25) of the lobes showing strong H<sub>2</sub> emission which exhibit a spectrum consistent with UV excitation in a PDR. The core region shows a strong dust continuum component, as well as emission lines of H I, He I, Fe II, \[Fe II\] and O I. The N knot has strong \[Fe II\] emission, with relatively weaker H I, He I, and H<sub>2</sub> emission.
#### 3.2.14 Vy 2–2
Vy 2–2 is a compact PN, so very little is known about its morphology. The spectrum obtained in this survey was taken centered on the bright core and the slit sampled most or all of the emission from this object. The spectrum contains stellar continuum, lines of H I and He I emission from the nebula, and weak H<sub>2</sub> emission. This detection confirms the indication of H<sub>2</sub> emission as reported by Dinerstein et al. (1986). The spectrum shown in Figure 26 is similar to others in this category, such as BD+303639 and NGC 2440, where several nebular components are superimposed because of the position and size of the slit. As for those objects, the H<sub>2</sub> spectrum in Vy 2–2 is also UV excited.
### 3.3. H<sub>2</sub> dominated
The PNe in this group have spectra that primarily contain emission lines of H<sub>2</sub>. These objects all have bipolar morphology, and most are young or proto-PNe (PPNe). The PPNe also have warm continuum dust emission or stellar continuum that is strongest in the core. Table 7 lists the line identifications and fluxes for the PNe in this section.
#### 3.3.1 NGC 2346
NGC 2346 is a PN with faint bipolar lobes seen clearly in H<sub>2</sub> emission (e.g., Paper I). The brightest part of the nebula is in the “equatorial” region near the central star where the bipolar lobes meet. Walsh, Meaburn, & Whitehead (1991) performed deep imaging and spectroscopy that showed the full extent of the lobes, and they model the PN as two ellipsoidal shells that are joined near the central star. The distribution of the H<sub>2</sub> emission is similar to the optical (Zuckerman & Gatley 1988; Kastner et al. 1994; Paper I).
The near-IR spectrum of NGC 2346 in Figure 27 is dominated by UV-excited H<sub>2</sub> emission, as shown in Figure 33d. The spectrum was obtained with the slit positioned on the bright condensation to the W of the central star. There is also weak Pa$`\beta `$ and Br$`\gamma `$ emission seen, which is possibly reflected from near the central star.
#### 3.3.2 J 900
The PN J 900 is a bipolar nebula with an unusual “jet”-like structure and an outer shell structure that is seen primarily in H<sub>2</sub> emission (Shupe et al. 1995; Paper I). The spectrum of J 900 shown in Figure 27 was obtained at a position N of the brighter lobe just NW of the central star, centered on the “jet” of emission. Problems with sky-subtraction caused the J and H-band portions of the spectrum continuum to be slightly negative. There is no detected continuum in any part of the spectrum. The H<sub>2</sub> spectrum is shock-excited in a moderate velocity wind (Figure 33e).
#### 3.3.3 AFGL 618
AFGL 618 is a carbon-rich, bipolar reflection nebula with a relatively hot central star ($`30,000`$ K), similar spectra in the two lobes, and the eastern lobe is significantly brighter than the other. In this as in other ways, the object bears a great resemblance to AFGL 2688 (see below), despite the fact that their central star temperatures differ by about a factor of 5. The visible spectrum shows numerous emission lines characteristic of ionized gas (Westbrook et al. 1975; Schmidt & Cohen 1981) which are scattered by dust into the line of sight, with a small H II region surrounding the central object (Carsenty & Solf 1982; Kelly, Latter, & Rieke 1992). The near-IR spectrum of AFGL 618 is also dominated by rotation-vibration lines of H<sub>2</sub> (Thronson 1981, 1983; Latter et al. 1992; Paper I).
AFGL 618 exhibits a rich spectrum of molecular line emission (Lo and Bechis 1976; Knapp et al. 1982; Cernicharo et al. 1989; Kahane et al. 1992; Martin-Pintado & Bachiller 1992; Bachiller et al. 1997; Young 1997). The lines detected include <sup>12</sup>CO, <sup>13</sup>CO, C<sup>17</sup>O, C<sup>18</sup>O, CS, NH<sub>3</sub>, HCN, HCO<sup>+</sup>, CN, and C I.
Two of the positions sampled are presented here in Figure 28 – the core spectrum and one taken 2$`\stackrel{}{\mathrm{.}}`$4 E of the core. Both spectra show strong H<sub>2</sub> emission, along with \[Fe II\] and weak Pa$`\beta `$ and Br$`\gamma `$. In addition, the core has a warm dust continuum that is apparent throughout the spectrum, and clear CO bandhead features in the 2.3 – 2.4 µm region. The CO features are also present but at lower levels in the 2$`\stackrel{}{\mathrm{.}}`$4 E spectrum position. Our analysis of the H<sub>2</sub> spectrum confirms the earlier results by Latter et al. (1992) – the spectrum is dominated by a shock-heated component, but a UV excited component is clearly present as well (Figure 33c).
#### 3.3.4 AFGL 2688
AFGL 2688 (the “Egg Nebula”) is a bipolar reflection nebula (Ney et al. 1975) at visible and near-infrared wavelengths. It has a central star that exhibits the spectrum of a carbon-rich supergiant (Crampton et al. 1975; Lo & Bechis 1976). Similar in visible appearance to AFGL 915, each lobe shows two “jets” or “horns” extending away from the central region (Crampton et al. 1975; Latter et al. 1993; Sahai et al. 1998a). The lobes have identical spectra at visible wavelengths, but their brightness differs significantly (Cohen & Kuhi 1977). The near-IR spectrum is dominated by H<sub>2</sub> rotation-vibration lines (Thronson 1982; Beckwith 1984; Latter et al. 1993). A central source is seen in the mid-IR and longer wavelengths, with fainter extended emission along the axis of the nebula (Hora et al. 1996). There is an enigmatic equatorial region seen in H<sub>2</sub> emission and might be traced by other molecular species, such as HCN (Latter et al. 1993; Bieging & Ngyuen-Quang-Rieu 1996; Sahai et al. 1998b).
Similar to AFGL 618, this object also has a rich molecular content. SiC<sub>2</sub> is seen in absorption (Cohen & Kuhi 1977); this feature is usually found in stars of the highest carbon abundance. Strong absorption features of C<sub>3</sub> and emission in C<sub>2</sub> (Crampton et al. 1975) are present, while C<sub>2</sub> is also seen in absorption in reflected light from the lobes (Bakker et al. 1997). The CO $`J=10`$ line shows three distinct velocity structures (Kawabe et al. 1987; Young et al. 1992).
Our results for this object from this survey were previously presented in Hora & Latter (1994, 1995) and our narrowband imaging in Latter et al. (1993). The spectra are reproduced here for comparison with the rest of the survey. Spectra were obtained at several positions in the nebula, including positions along the N lobe, and in the equatorial region (see Hora & Latter 1994 for details). The emission is segregated; the core is dominated by continuum emission, there are emission lines of C<sub>2</sub> and CN further from the core along the lobes, and the H<sub>2</sub> emission is confined to the ends of the lobes and in the equatorial region in what appears to be a ring or toroidal structure (Latter et al. 1993; Sahai et al. 1998b). Our analysis of the H<sub>2</sub> line ratios showed that the emission is collisionally excited in shocks, with no discernible difference between the emission in the lobes and the equatorial region.
### 3.4. Continuum - dominated
These are young PNe or PPNe that have strong warm dust continuum and little line emission. The strongest component is in general the core, with most of the emission from an unresolved point source. In some of the nebulae, emission structure extends a few arcseconds from the core region. Also, in objects such as AFGL 915, they are associated with larger optical nebulae that extend arcminutes from the core. In this survey, only the regions near the core were sampled.
#### 3.4.1 AFGL 915
AFGL 915 (the “Red Rectangle”) is a carbon-rich biconical reflection nebula with a metal-depleted spectroscopic binary at its center (Cohen et al. 1975). The nebula appears axially symmetric and shows spikes running tangent to the edge of the bicone. Surrounding the post-AGB star at its center is a circumbinary disk viewed edge-on (Jura, Balm, & Kahane 1995) which could be oxygen-rich (Waters et al. 1998).
C<sub>2</sub> and CN are not detected near the binary, though C<sub>2</sub> is present in emission in the reflection lobes. CH<sup>+</sup> (0,0) and (1,0) are detected in emission (Bakker et al. 1997; Balm & Jura 1992). CO is underabundant, with relatively weak emission and broad wings detected (Dayal & Bieging 1996; Greaves & Holland 1997; Loup et al. 1993; Bujarrabal et al. 1992). Glinski et al. (1997) found CO and C I in the UV in both absorption and emission. They expect strong CO overtone emission in the IR based on their observations of hot CO emission and absorption in the UV. The object shows ERE (extended red emission) from about $`\lambda =5400`$ to 7200 Å and a set of emission bands around 5800 Å (Schmidt, Cohen, & Margon 1980) whose carriers might be the same material as the carriers of the DIBs (diffuse interstellar bands). This object also shows strong emission in the PAH bands at 3.3, 7.7, and 11.2 µm (Cohen et al. 1975), which are located predominantly in the lobes and spikes of emission (Bregman et al. 1993; Hora et al. 1996).
Spectra taken at two different positions are shown, one centered on the core, and the other at 4″ S of the core. Both show strong warm dust continuum, and the core also has strong CO bandhead emission features in the $`\lambda =2.32.4`$ µm range.
#### 3.4.2 IRC+10420
IRC+10420 is a highly evolved, OH/IR star that is thought to be in a post-red supergiant phase (Jones et al. 1993). The central star seems to have changed spectral type, transitioning recently to an early A type (Oudmaijer 1998). Oudmaijer et al. (1996) detected several of the hydrogen lines in absorption and emission in the near-IR, and Oudmaijer (1998) presented a high-resolution 0.38 – 1 µm spectrum showing a large number of emission and absorption lines. Recent HST imaging by Humphreys et al. (1997) shows that the circumstellar environment around this star is extremely complex, with spherical outer shells that extend to a diameter of 6 arcseconds, and several inner condensations. In the near- and mid-infrared, bipolar lobes are visible that extend $`2`$ arcseconds from the core.
The spectrum of IRC+10420 shown in Figure 31 was taken centered on the object. The slit length includes the inner few arcseconds of the object, although it is dominated by the bright core. The observed spectrum shows a bright and relatively featureless continuum. Some H I lines, e.g., Pa$`\beta `$, are seen in absorption.
#### 3.4.3 M 2–56
The PPN M 2–56 is a bipolar nebula with a bright central core. It is similar in morphology to AFGL 618, although it seems to be at an earlier evolutionary stage since it does not appear to have an H II region (Trammell, Dinerstein, & Goodrich 1993; Goodrich 1991). The spectrum of M 2–56 shown in Figure 31 was taken centered on the core of this PPN. The dominant feature is a hot dust continuum that is most prominent in the K band region of the spectrum. There are some residual features in the spectrum from imperfect sky subtraction, mostly in the K band.
## 4. Discussion and Summary
### 4.1. Spectral Categories
The PNe spectra presented in this paper were grouped according to spectral characteristics as described above. The groups are an efficient way to present the data, but also can be seen to correlate strongly with other characteristics of the PN.
#### 4.1.1 Morphology
The group of H I - line dominated PNe is composed of primarily elliptical or round PNe, along with the peculiar or irregular nebulae of the sample. In general these PNe are well-known from optical studies, identified either by their morphology or their optical spectra. Many of the PNe in this group of the sample, however, do have IR “excess” continuum emission from warm dust, which in some cases prompted their inclusion in this sample.
The spectral groups with molecular and/or dust continuum emission are primarily bipolar. This classification includes objects such as NGC 6720 which have a ring morphology but are thought to be bipolar viewed pole-on; Hb 12 which is brightest in H<sub>2</sub> in the equatorial region and along the outer edges of the lobes; M 2–9 which is brightest in H<sub>2</sub> at the edges of the lobes with no equatorial emission other than at the core; and AFGL 2688 which is brightest along the axis of the bipolar lobes, with H<sub>2</sub> emission in the equatorial plane. Clearly this is a heterogeneous group with a wide range of emission and morphological differences that imply a range of evolutionary tracks and states.
#### 4.1.2 The Carbon-to-Oxygen Ratio
Carbon stars, although a small fraction of all AGB stars, return about half of the total mass injected into the ISM by all AGB stars, since they have on average much higher mass loss rates ($`>10^4\mathrm{M}_{\mathrm{}}\mathrm{yr}^1`$) than do O-rich objects. The carbon-to-oxygen (C/O) abundance ratio in PNe has previously been shown to correlate with morphology (Zuckerman & Aller 1986), with bipolar PNe tending to be carbon-rich. It is therefore expected that the C/O ratio also correlates with the spectral classifications presented here. This is in general the case, with the H I - line dominated PNe having C/O ratios less than or about 1, whereas the remaining categories which are dominated by the bipolar PNe have C/O ratios $`>`$ 1, as reported by Zuckerman & Aller (1986) and Rola & Stasińska (1994). Rola & Stasińska discuss problems with previous determinations of the C/O ratio, and use different criteria that result in a slightly lower percentage of carbon-rich PNe (35%) than others. Their ratios are used in the discussion below.
The morphology of PN also has been shown to depend on the progenitor mass (see Corradi & Schwarz 1995), with the bipolar PN being more massive than other morphological types. This relationship, along with the link between carbon abundance and morphology, suggests that carbon stars are the progenitors of bipolar PN and those with a large amount of molecular material. The mechanisms that cause massive carbon-rich stars to preferentially form bipolar PN are still not understood.
There are some exceptions to the correlation of morphological type to C/O ratio; in particular, NGC 6543 has a much higher value (9.55) than the others in the class. In the other extreme, NGC 2346 stands out as having a low C/O ratio (0.35) compared to other bipolar PNe in the H<sub>2</sub> \- dominated group. This object is a much more evolved object than the others in its group (e.g., AFGL 2688), and exhibits weak Pa$`\beta `$ emission, showing that the ionized gas is present although weak relative to the molecular emission in the nebula.
### 4.2. Spectral Sampling of Morphological Features
This survey has differed from many previous investigations in that a short, narrow slit was used to obtain the data, rather than a large beam that could include most or all of the nebula. Because of this, one cannot easily use the spectra presented here to model the PNe in a global sense, if that requires a measurement of the total flux from the object. Also, if a complete census of emission lines were required, some might be missed if there were variations of emission characteristics across the nebula and certain regions were not sampled.
The spatial selectivity that prevents viewing the entire PN at once, however, has proven to be an advantage when trying to examine various aspects of the PN, including variations across the nebula, as a function of distance from the central star, or in examining certain morphological features. For example, in M 2–9 and NGC 2440, the emission of the lobe walls and emission knots were separately sampled, which showed the large spectral differences in these regions. This information is important for modeling the structure and formation of the PN. Another reason why the small aperture is useful is that if the emission from a group of lines such as H<sub>2</sub> is to be modeled, it is important to compare the emission from a clump of material where the conditions do not vary greatly over its size. For example, the emission from H<sub>2</sub> present very close to the central star in a strong UV field could be quite different from H<sub>2</sub> emission from the outer parts of the halo. Also, the small slit has aided in detecting weak H<sub>2</sub> emission from several PNe such as NGC 40, where detection would have been difficult if the central star and the rest of the nebula could not be excluded from the measurement.
### 4.3. Summary of Molecular Hydrogen Emission in PNe
A long-standing problem in the interpretation of H<sub>2</sub> emission from interstellar and circumstellar environments is understanding the excitation mechanism. Three fundamental mechanisms are possible. One is excitation of a near-infrared fluorescence spectrum resulting from a rotational-vibrational cascade in the ground electronic state following electronic excitation by the absorption a UV photon in the Lyman and Werner bands (Black & van Dishoeck 1987). A second excitation mechanism is collisional excitation in a warm gas ($`T_\mathrm{K}1600`$ K). While UV excitation in a low density gas produces an easily identifiable spectrum, the level populations can be driven to produce thermal line ratios when the UV flux is large and densities begin to exceed $``$ $`10^4`$ cm<sup>-3</sup> (Sternberg & Dalgarno 1989). Detailed spectral and morphological analysis are often required to determine an origin of the near-IR spectrum. A third excitation mechanism is formation of H<sub>2</sub> on the surfaces of dust grains and in the gas phase. While potentially important in isolated regions of certain objects, we do not consider this to be generally important in PNe and PPNe relative to the other two processes. This is because molecular formation in PNe is relatively slow compared to dissociation rates.
In PNe and PPNe, the situation can be complicated by both dominant excitation mechanisms being present simultaneously, and in different forms. Several ways of exciting near-IR H<sub>2</sub> emission have been identified as possible: direct thermal excitation in warm gas created behind moderate velocity shocks, direct excitation by UV photons from the hot central star, somewhat indirectly by collisional excitation in warm gas created by rapid grain streaming (e.g. Jura & Kroto 1990), and excitation through absorption of Ly$`\alpha `$ photons (by an accidental resonances with the B$`{}_{}{}^{1}\mathrm{\Sigma }_{u}^{+}\mathrm{X}^1\mathrm{\Sigma }_g^+`$ $`v=12`$ P(5) and R(6) transitions of H<sub>2</sub>) which can be generated in a nearby strong shock (e.g., Black & van Dishoeck 1987).
The first two mechanisms have been identified in several PNe, such as thermal excitation in AFGL 2688 (e.g., Hora & Latter 1994; Sahai et al. 1998a), pure UV excitation in a low density gas around Hb 12 (e.g., Dinerstein et al. 1988; Hora & Latter 1996; Luhman & Rieke 1996), and UV excitation in a high density gas in M 2–9 (Hora & Latter 1994) and NGC 7027 (Graham et al. 1993b; this paper). A combined spectrum was found from a detailed analysis of AFGL 618 (Latter et al. 1992; this paper). While the form of the excitation might be apparent for these and other objects, it is not always evident what is the source of the warm gas or UV photons. Winds are present in AFGL 2688 which could directly heat the gas through shocks, but considerable grain streaming is likely taking place as well (see Jura & Kroto 1990).
Very fast winds and dissociating shocks are present in AFGL 618, M 2–9 (e.g., Kelly, Latter, & Hora 1998), and M 1–16 (Sahai et al. 1994; Schwarz 1992), and all show clear evidence of UV excitation. In addition, the photon path to the H<sub>2</sub> emitting regions is not in a direct line-of-sight to the central star, which for photons coming from the central star suggests scattering in what is a fairly low density medium. Alternatively, we are seeing in each of these objects excitation of H<sub>2</sub> at the bipolar lobe walls by UV photons generated within strong shocks produced by the fast winds. This hypothesis was explored through detailed modeling by Latter et al. (1992) of AFGL 618, but the high relative intensity of the thermally excited emission and poor spatial resolution limited this analysis. The presence of very fast winds in the lobes of each of these objects, and the presence of UV excited H<sub>2</sub> emission at the lobe walls strongly suggests that indirect excitation of the H<sub>2</sub> is occurring by interactions with photons generated by wind-produced shocks. Detailed modeling of sensitive, high spatial resolution spectra is required. It is also evident, in general, that without detailed spectra, H<sub>2</sub> is a rather poor diagnostic of overall conditions in PNe and PPNe.
If we conclude that all of the ways to excite H<sub>2</sub> in PNe and PPNe listed above are present and important, what does this imply for our understanding of these objects and the utility of H<sub>2</sub> as a diagnostic? It is now well understood that the presence of molecular emission from PNe and PPNe is tied to the morphology of the objects such that if molecular emission is present, the object has a bipolar morphology (e.g., Zuckerman & Gatley 1988; Latter et al. 1996; Kastner et al. 1996, and references therein). We have argued that H<sub>2</sub> emission is excited in multiple ways in PNe and PPNe. While special conditions are required for H<sub>2</sub> emission to be seen in near-IR spectra, the conditions that drive the excitation are common in all PNe and PPNe and are not clearly dependent on morphological type. A conclusion that can be drawn from this argument alone is that molecular material is present in nebulae with a bipolar morphology and a significant amount of molecular material is $`not`$ present in other morphological types. Therefore, objects that have a bipolar morphology must have a dense, high mass envelope in which the molecular material can be shielded and survive dissociation for relatively long times – suggesting a high mass loss rate and a high mass progenitor star. A correlation between bipolar morphology and high mass progenitor stars has been found by others (e.g., Corradi & Schwarz 1995). It is apparent that the presence of H<sub>2</sub> emission in a PN is not tied to directly to the morphology, but that the bipolar morphology is intimately related to the density and mass of the circumstellar envelope, and therefore the mass of the progenitor star. Why high mass, high mass loss rate asymptotic giant branch stars shed material in an axisymmetric, not spherical, way remains a mystery.
We thank Xander Tielens and David Hollenbach for useful discussions and encouragement. We acknowledge support from NASA grant 399-20-61 from the Long Term Space Astrophysics Program. WBL was supported during part of this study by a National Research Council Research Associateship.
|
no-problem/9904/nucl-th9904042.html
|
ar5iv
|
text
|
# (Hybrid) Baryons in the Flux–Tube Model
## Abstract
We construct baryons and hybrid baryons in the non–relativistic flux–tube model of Isgur and Paton. The motion of the flux–tube with the three quark positions fixed, except for centre of mass corrections, is discussed. It is shown that the problem can to an excellent approximation be reduced to the independent motion of a junction and strings.
Hybrids are bound states where there is an explicit excitation in the gluon field of QCD. Particularly, “hybrids baryons” may be viewed as quark–quark–quark–glue composites.
The experimental interest in hybrid baryons centers around the excited baryon resonance ($`<2.2`$ GeV) program at TJNAF, mostly in Hall B. Hybrid baryon production is expected. If hybrid baryons obey similar decay selection rules to hybrid mesons, they may be distinguishable based on their strong decays.
Hybrid baryons have only been constructed in the MIT bag model bag . We are motivated to build a model consistent<sup>1</sup><sup>1</sup>1The hybrid meson interquark potential is consistent with that evaluated from lattice gauge theory paton85 . with predictions from QCD lattice gauge theory, i.e. the Isgur–Paton non–relativistic flux–tube model paton85 . This model is motivated from the strong coupling limit of the hamiltonian lattice gauge theory formulation of QCD (HLGT).
This talk will deal with fixed quark positions relative to each other. However, we shall allow the quarks with fixed relative positions to move in order to work in the centre of mass frame. This is called the “quasi–adiabatic” approximation.
The model is motivated from the strong coupling limit of HLGT, where there are “flux–lines” which play the role of glue. In the spirit of the adiabatic approximation, where quarks do not respond to the influence of glue, we neglect operators which make quarks move. The remaining operator taking you away from the strong coupling limit is the plaquette operator. This induces motion of the “flux–line” between the quarks perpendicular to its rest position. We model the flux–line by “beads”, all with the same mass, which is fixed from the energy in the linear flux–line. The beads are seperated along their rest positions by a finite lattice spacing, and are allowed to move perpendicular to their rest position. The beads are attracted to each other by a linear potential, vibrating in various string modes.
There is a second essential ingredient from HLGT. Three flux–lines can come together at a point called the “junction”. A plaquette operator cannot move the junction in the lowest order of perturbation theory so that the junction is taken to have a different (higher) mass associated with it than the other beads.
The final picture of a (hybrid) baryon is that of three quarks, each connected via a line of beads to the junction in a Mercedes Benz configuration. The three quarks define a plane. The “equilibrium configuration” is the lowest energy configuration: the junction is located such that there are angles of $`120^o`$ between each of the “triads” that connect each of the quarks to the junction, and the beads all lie on the triads. The junction and beads then vibrate with respect to the equilibrium configuration. There are two important motions which are expected to have physical significance: (1) the motion of the junction perpendicular and within the plane relative to the junction rest position, called the “junction motion”; (2) the motion of the beads in the two directions perpendicular to the line connecting the quark to the junction, called the “string motion”. If the angles between two of the quarks suspended at the third quark is larger than $`120^o`$ the equilibrium configuration is not the Mercedes Benz configuration. This issue is not considered further here.
We now make the small oscillation approximation, where the beads and junction move near to the equilibrium configuration. We make sure that we work in the centre of mass system, and therefore make the “quasi–adiabatic” approximation. The hamiltonian is written in terms of the junction and string motion coordinates.
We have demonstated that the hamiltonian can be seperated into a part $`H_J`$ which corresponds to the motion of the junction in the potential one would use if there were no beads in the problem, with an effective junction mass related to its own mass and the mass of the beads. Another part $`H_S`$ is the independent motion of three strings with respect to a fixed junction, with an effective bead mass that is related to the bead and junction masses. There is also an “interaction term” $`H_{int}`$ where the strings corresponding to different quarks interact with each other and various string modes associated with the same quark interact with each other, and where the junction interacts with the various string modes.
We shall now demonstate that the interaction term gives a minor contribution. The free parameters in the model (and the values used for the numerical simulation) are the string tension ($`0.18`$ GeV<sup>2</sup>), the ratio of the junction and bead masses (1) and the quark masses ($`0.33`$ GeV). We shall perform a simulation where there is one bead between each quark and the junction and the quarks form an equilateral triangle with the lengths of the triads equal to a typical value of $`2.5`$ GeV<sup>-1</sup>.
First we solve the exact problem numerically. The frequencies parallel and perpendicular to the plane are (in GeV)
| Parallel | 0.607 | 0.607 | 0.924 | 1.08 | 1.08 |
| --- | --- | --- | --- | --- | --- |
| Perpendicular | 0.828 | | 0.924 | 0.924 | 1.37 |
where the bold faced frequencies are clearly lower than the others. In fact, if we set $`H_{int}=0`$ then we again obtain the same number of frequencies, and here quote the results for the lowest lying frequencies corresponding to the junction motion (in GeV):
| Parallel | 0.614 | 0.614 |
| --- | --- | --- |
| Perpendicular | 0.869 |
We hence conclude that the lowest frequencies are the ones corresponding to the junction motion, and that we can neglect $`H_{int}`$ safely for the lowest frequency. In retrospect, the reason why $`H_{int}`$ can be neglected is because we chose the physically appropriate coordinates for the problem: the junction and string motions.
To compare the frequencies of the full hamiltonian and the junction frequencies of the hamiltonian with $`H_{int}=0`$, we calculate the deviation $`ϵ`$ of them from one another. $`ϵ_{}=1\%`$ and $`ϵ_+=1\%`$ for the low and high parallel frequencies respectively (in this case the frequencies are equal). $`ϵ=5\%`$ perpendicular to the plane. For the lowest frequency, it is hence sufficient to work with a hamiltonian of the form $`H_J+H_S`$ from now on, and there are three types of hybrid baryons: the one corresponding to junction motion perpendicular to the plane which is always the heaviest; and two corresponding to motion parallel to the plane. For a generic quark configuration, one of the parallel frequencies is always below or equal to the other, so that they are not usually degenerate. The flux–tube model thus contains three low lying hybrid baryons, corresponding to vibrations along three perpendicular axes, but each with a different excitation frequency above the baryon.
To access the error in the hamiltonian with $`H_{int}=0`$ more fully, we vary parameters around the central values above, one at a time; with the quark mass up to the charm quark mass of $`1.5`$ GeV, the ratio of the junction to the bead mass up to 10, and with triads with lengths from $`0.55`$ GeV<sup>-1</sup> (the triad lengths not being equal in general). We found that that $`ϵ_{}\stackrel{<}{}5\%`$ and $`ϵ_+\stackrel{<}{}6\%`$ for parallel frequencies and $`ϵ\stackrel{<}{}40\%`$ for the perpendicular frequency. The error introduced by neglecting $`H_{int}`$ is therefore rather minimal for the lowest frequency. Hence, to a good approximation, the dynamics of the lowest frequency can be simplified to junction and string motion which are independent of one another.
Significant progress has been made towards building a realistic flux–tube model of (hybrid) baryons. We have constructed the full multibead hamiltonian in the quasi–adiabatic approximation and in the small oscillations approximation. We demonstated that the junction bead decouples from the other beads to a high degree of accuracy.
|
no-problem/9904/astro-ph9904018.html
|
ar5iv
|
text
|
# ABSTRACT
## ABSTRACT
We describe BeppoSAX observations of the black hole candidates LMC X–1 and LMC X–3 performed in Oct. 1997. Both sources can be modelled by a multicolor accretion disk spectrum, with temperature $`1`$ keV. However, there is some evidence that a thin emitting component coexists with the thick disk at these temperatures. In the direction of LMC X–1, we detected a significant emission above 10 keV, which we suspect originates from the nearby source PSR 0540-69. For LMC X–1, we estimate an absorbing column density of $`6\times 10^{21}`$ cm<sup>-2</sup>, which is almost ten times larger than that found for LMC X–3. In both sources, we find no indication of emission or absorption features whatsoever.
## 1 INTRODUCTION
LMC X–3 and LMC X–1 are two luminous persistent X–ray binaries. LMC X–3 has an orbital period of 1.7 days and is one of the most secure black hole candidates ($`M_\mathrm{X}10M_{}`$). For LMC X–1 there is still some uncertainty regarding the optical counterpart. The most probable optical candidate has a period of 4.23 d, which implies a companion (hole) mass of $`5M_{}`$ (Cowley et al. 1995).
The two objects have soft X–ray spectra reminiscent of Cyg X–1 in the high/soft state. A high energy tail extending above 10 keV is detected in both systems (Ebisawa, Mitsuda & Inoue 1989; Treves et al. 1990).
We observed the two sources with BeppoSAX in Oct. 1997, which allowed us to construct simultaneous spectra in a broad energy band (0.1–100 keV). All the four narrow field instruments (LECS 0.1–4 keV, MECS 1.8–10 keV, HPGSPC 7–70 keV, PDS 12–150 keV) performed nominally; the MECS functioning with two units. Here we are presenting some preliminary results on the X–ray spectral distribution. HPGSPC data analysis is not discussed in this paper.
Our observations correspond to a low intensity state of LMC X–3, while LMC X–1 appears stable on month/year time scale, as apparent from the the XTE–ASM light curves.
## 2 BeppoSAX SPECTRA
Data reduction followed the standard procedure. Total exposure times and count rates are reported in Table 1. Results of the spectral analysis described below are reported in Table 2.
* In the field of view of the MECS and PDS there is the source PSR 0540-69. In the $`[210]`$ keV band, the pulsar is fainter than LMC X–1 by a factor $``$ 10. The pulsar spectrum is rather hard, and a power–law fit of the MECS data yields an energy index $`\mathrm{\Gamma }=2.01\pm 0.07`$. In principle, the pulsar emission could account of (part of) the signal detected in the PDS. To test this hypothesis, we first fitted separately the LECS+MECS data of LMC X–1 with an absorbed disk blackbody, obtaining a totally unacceptable fit ($`\chi ^2/dof>3`$). The inclusion of an optically thin free–free component greatly improves the fit ($`\chi ^2/dof=0.89`$; see Table 2 and Fig. 1). We did not find any narrow emission or absorption feature. The column density is quite large, $`N_\mathrm{H}6.5\times 10^{21}`$ cm<sup>-2</sup>. Then, we fitted the PDS data alone. They are well represented by a hard power–law, which is completely consistent, in terms of flux and spectral index, with the extrapolation in the PDS energy range of the MECS spectrum of the pulsar. We therefore suspect that the pulsar is responsible of the high energy emission detected in the PDS.
* In this source, we did not find any positive detection in the PDS. A variable hard X–ray tail was observed in the PDS in only one of the two one–month spaced BeppoSAX science verification phase observations performed in 1996, while in the other it had an upper limit comparable to ours (Siddiqui et al. 1998). As in the case of LMC X–1, a fit with an absorbed multicolor disk spectrum showed evidence of features in the 0.8–2 keV range ($`\chi ^2`$ null probability $`\stackrel{<}{}0.1`$%). The inclusion of a free–free component reduced the $`\chi ^2`$ to an acceptable value (see Table 2 and Fig. 2). Again, we did not find any narrow emission or absorption feature above 2 keV. The absorbing column lies in the range reported by previous X–ray observations ($`N_\mathrm{H}5.7\times 10^{20}`$ cm<sup>-2</sup>, Treves et al. 1988).
## 3 CONCLUSIONS
Our analysis shows that in LMC X–1 and LMC X–3 an optically thick accretion disk coexists with an optically thin, X–ray emitting gas, with comparable temperatures. This may indicate that the thermal emission from the innermost region of the accretion disk is modified by electron scattering. In LMC X–1 direction there is also indication of a component at much higher energy, responsible for the power–law emission in the PDS, which however we suspect arises from a nearby pulsar. Previous claims of the presence of a hard tail in LMC X–1 were based on non imaging instruments, where confusion with the nearby pulsar can not be excluded. LMC X–1 is found to be severely absorbed at low energies. Theoretical interpretations of the results are in progress and will be presented in a forthcoming paper.
## 4 REFERENCES
* Cowley, A.P., Schmidtke, P.C., Anderson, A.L., and McGrath, T.K., 1995, PASP, 107, 145.
* Ebisawa, K., Mitsuda, K., and Inoue, H., 1989, PASJ, 41, 519.
* Siddiqui, H., et al., 1998, in prep.
* Treves, A., et al., 1988, ApJ, 325, 119.
* Treves, A., et al., 1990, ApJ, 364, 266.
|
no-problem/9904/cond-mat9904136.html
|
ar5iv
|
text
|
# Glassy timescale divergence and anomalous coarsening in a kinetically constrained spin chain
## Abstract
We analyse the out of equilibrium behavior of an Ising spin chain with an asymmetric kinetic constraint after a quench to a low temperature $`T`$. In the limit $`T0`$, we provide an exact solution of the resulting coarsening process. The equilibration time exhibits a ‘glassy’ divergence $`t_{\mathrm{eq}}=\mathrm{exp}(\text{const}/T^2)`$ (popular as an alternative to the Vogel-Fulcher law), while the average domain length grows with a temperature dependent exponent, $`\overline{d}t^{T\mathrm{ln}2}`$. We show that the equilibration time $`t_{\mathrm{eq}}`$ also sets the timescale for the linear response of the system at low temperatures.
Even after decades of research, understanding the dynamics of glasses remains a challenging problem (see e.g. ). One of the main features of glassy systems is that their relaxation time $`\tau `$ increases quickly as the temperature $`T`$ is lowered. A popular representation of this increase (for so-called ‘fragile’ glasses ) is the Vogel-Fulcher (VF) law, $`\tau \mathrm{exp}[\text{const}/(TT_0)]`$. This predicts that $`\tau `$ diverges at temperature $`T_0`$, and the latter has therefore been associated with the temperature at which a true thermodynamic glass transition (achievable only in the limit of infinitely slow cooling) would take place. However, other functional forms for $`\tau (T)`$ that have been proposed do not exhibit singularities at any finite $`T`$, indicating the absence of a thermodynamic glass transition. Among these, the exponential inverse temperature squared (EITS) form $`\tau \mathrm{exp}(\text{const}/T^2)`$ is popular. Experimentally, it is difficult to distinguish between VF and EITS behavior due to obvious limitations on the longest accessible timescales; both can represent the experimentally observed $`\tau (T)`$ in many materials . Thus analytical results are desirable to shed light on this controversy. In this work we solve a simple dynamical model exhibiting glassy dynamics and find EITS behavior.
To model relaxation in glassy systems theoretically, one can postulate some kind of quenched disorder, either in terms of some underlying microscopic Hamiltonian (as is done in spin glasses) or more phenomenologically by making assumptions about the phase space of the system (e.g. in terms of hierarchical or ultrametric structures or energy barrier distributions ). So far the main theoretical justification for either VF or EITS behavior comes from the latter approach; the EITS law, for example, is motivated by considering activated dynamics in a landscape of Gaussian distributed energy barriers .
The alternative approach is to consider simple models whose dynamics directly induce glassiness. Examples include systems with kinetic constraints or entropic barriers , and driven diffusive models . Such an approach is more obviously relevant to the dynamics of structural glasses (where quenched disorder is absent) since one does not need additional arguments that relate quenched and dynamically ‘self-induced’ disorder . The present work provides a first example where EITS behavior emerges directly from a microscopic model without imposed quenched disorder; instead energy barriers arise naturally from dynamical constraints.
We consider a chain of spins in a uniform field, whose dynamics is nontrivial due to an asymmetric kinetic constraint. This model was introduced by Jäckle and Eisinger and has recently been rediscovered . We study in particular the behavior after a quench to a low temperature $`T0`$. We solve the resulting coarsening dynamics exactly in this limit and find two main results: Firstly, the equilibration time of the system diverges as $`t_{\mathrm{eq}}\mathrm{exp}(1/T^2\mathrm{ln}2)`$ (EITS behaviour). Secondly, before equilibrium is reached, the average domain length grows as $`\overline{d}t^{T\mathrm{ln}2}`$, with an exponent that varies continuously with temperature. This novel anomalous coarsening is a consequence of the dynamical constraint, which produces scale-dependent energy barriers which grow as the logarithm of the domain size. Finally, we show that $`t_{\mathrm{eq}}`$ is not just the timescale for equilibration after a quench, but in fact is also the timescale for relaxation of spin-spin correlations in equilibrium (at low $`T`$); this relaxation time therefore also has an EITS divergence at low $`T`$.
The model comprises a chain of $`L`$ spins $`s_i\{0,1\}`$ where $`1iL`$; periodic boundary conditions imply that the left neighbor of $`s_1`$ is $`s_L`$. The dynamics for a given temperature $`T`$ are defined as follows: At any time, only spins whose left neighbor is up (i.e., has the value 1) can flip. For such ‘mobile’ spins, the rate for down-flips $`10`$ is 1, while the rate for up-flips $`01`$ is $`ϵ=\mathrm{exp}(1/T)`$. Detailed balance is obeyed, and the stationary distribution is the Boltzmann distribution for the trivial Hamiltonian $`H=_{i=1}^Ls_i`$. For low temperatures the equilibrium concentration $`c=ϵ/(1+ϵ)`$ of up-spins is small. Since these spins facilitate the dynamics, the system evolves slowly for small $`T`$. Moreover to eliminate an up-spin one first has to generate an adjacent up-spin.
Thus there are energy barriers in the system’s evolution.
We will be interested mainly in the behavior after a quench from equilibrium at some high initial temperature $`T_\mathrm{i}\stackrel{>}{}1`$ to $`T1`$. The basic objects that we use for the description of the system are domains. As shown by the vertical lines in $`\mathrm{}1|0001|1|1|01|001|1|1|01|0\mathrm{},`$ a domain consists of an up-spin and all the down-spins that separate it from the nearest up-spin to the left. The length $`d`$ of a domain also gives the distance between the up-spin at its right edge and the nearest up-spin to the left. Note that adjacent up-spins are counted as separate domains of length $`d=1`$. In equilibrium, the distribution of domain lengths and its average are
$$P_{\mathrm{eq}}(d)=ϵ/(1+ϵ)^d,\overline{d}_{\mathrm{eq}}=1+1/ϵ.$$
(1)
Now consider what happens after a deep quench to $`T1`$, $`ϵ1`$. The equilibrium concentration of up-spins at the final temperature $`T`$ is $`c=1/\overline{d}=ϵ+𝒪(ϵ^2)`$; hence the equilibrium probability of finding an up-spin within a chain segment of finite length $`d`$ is $`𝒪(dϵ)`$ and tends to zero for $`ϵ0`$. In this limit ($`ϵ0`$ at fixed $`d`$), the flipping down of up-spins therefore becomes irreversible to leading order. In terms of domains, this means that the coarsening dynamics of the system is one of coalescence of domains: an up-spin that flips down merges two neighbouring domains into one large domain. During such an irreversible coarsening process, no correlations between the lengths of neighboring domains can build up if there are none in the initial state . For the present model the equilibrated initial state consists of domains independently distributed according to (1). Therefore a ‘bag model’ or ‘independent interval approximation’ for the dynamics, which is defined by neglecting correlations between domains, becomes exact in the low-temperature limit (always taken at fixed $`d`$).
We now estimate the typical rate $`\mathrm{\Gamma }(d)`$ at which domains of length $`d`$ disappear by coalescing with their right neighbors. Because domain coalescence corresponds to the flipping down of up-spins, $`\mathrm{\Gamma }(d)`$ can also be defined as follows. Consider an open spin chain of length $`d`$, with a ‘clamped’ up-spin ($`s_0=1`$) added on the left. Starting from the state $`(s_0,s_1,\mathrm{},s_d)`$ = $`10\mathrm{}01`$, $`\mathrm{\Gamma }^1(d)`$ is the typical time needed to reach the empty state $`10\mathrm{}00`$ where spin $`s_d`$ has ‘relaxed’. Any instance of this relaxation process can be thought of as a path connecting the two states. Call the maximum number of ‘excited’ spins (up-spins except $`s_0`$) encountered along a path its height $`h`$. One might think that the relaxation of spin $`s_d`$ needs to proceed via the state 11…1, giving a path of height $`d`$. In fact, the minimal path height $`h(d)`$ is much lower and given by
$$h(d)=n+1\text{for}2^{n1}<d2^n$$
(2)
where $`n=0,1,\mathrm{}`$ This result is easily understood for $`d=2^n`$ . To relax the $`2^n`$-th spin $`s_{2^n}`$, one can first flip up $`s_{2^{n1}}`$ and use it as an ‘anchor’ for relaxing $`s_{2^n}`$. The corresponding path is (with $`s_{2^{n1}}`$ and $`s_{2^n}`$ underlined) $`1\mathrm{}\underset{¯}{0}\mathrm{}\underset{¯}{1}`$ $``$ $`1\mathrm{}\underset{¯}{1}\mathrm{}\underset{¯}{1}`$ $``$ $`1\mathrm{}\underset{¯}{1}\mathrm{}\underset{¯}{0}`$ $``$ $`1\mathrm{}\underset{¯}{0}\mathrm{}\underset{¯}{0}`$ and reaches height $`h(2^n)=h(2^{n1})+1`$; the $`+1`$ arises because the anchor stays up while the spin $`2^{n1}`$ to its right is relaxed. Continuing recursively, one arrives at $`h(2^n)=h(1)+n`$; but $`h(1)=1`$ because the only path for the relaxation of $`s_1`$ is $`1110`$. To prove (2) more generally, define $`d(h)`$ as the length of the largest single domain that can be relaxed by a path of height $`h`$. Because of detailed balance, any relaxation path can be reversed, yielding a path of the same height from the empty state to the state $`10\mathrm{}01`$. In the same way, let us define $`l(h)`$ to be the maximal length of any spin configuration (ending in an up spin) that can be reached from the empty state by a path of height $`h`$. One then has $`d(h+1)=l(h)+1`$ because to relax $`s_{d(h+1)}`$ one needs to flip up its left neighbor while exciting no more than $`h`$ additional spins. A second relation is obtained from the relaxation of a configuration realizing the bound $`l(h)`$. Such a configuration contains $`h`$ excited spins (due to its maximal length). To relax the first of these, no extra excitations are allowed (because of the ceiling $`h`$ on path height); for the relaxation of the 2<sup>nd</sup>, 3<sup>rd</sup>… $`h`$-th spin, a maximum of 1, 2… $`h1`$ excitations are available. Summing the maximal length change at each step then gives $`l(h)=_{h^{}=0}^{h1}d(h^{}+1)`$. The above two recursions for $`l(h)`$ and $`d(h)`$, combined with $`d(1)`$ $`=`$ $`l(1)`$ $`=`$ 1, yield $`l(h)=2^h1`$ and $`d(h)=2^{h1}`$, proving (2).
At this stage we already see the key feature of the dynamics: the energy barrier for the relaxation of spin $`s_d`$ is $`h(d)1`$ (the $`1`$ comes from the one excited spin ($`s_d`$) in the initial state). The rate for this relaxation is therefore $`\mathrm{\Gamma }(d)=𝒪(\mathrm{exp}[(h(d)1)/T])=𝒪(ϵ^{h(d)1})`$ . Then eq. (2) tells us that the relaxation rate for domains of size $`d`$ is $`\mathrm{\Gamma }(d)\mathrm{exp}(\mathrm{ln}d/T\mathrm{ln}2)`$. Thus the energy barrier for the growth of domains increases logarithmically with domain size, giving a typical domain size growing as $`\overline{d}t^{T\mathrm{ln}2}`$. Also, since $`\overline{d}_{\mathrm{eq}}\mathrm{exp}(1/T)`$ the equilibration time will grow according to an EITS law $`t_{\mathrm{eq}}\mathrm{exp}(1/T^2\mathrm{ln}2)`$.
From the scaling of $`\mathrm{\Gamma }(d)`$, the coarsening dynamics in the limit $`ϵ0`$ naturally divides into stages distinguished by $`n=h(d)1=0,1,\mathrm{}`$ During stage $`n`$, the domains with lengths $`2^{n1}<d2^n`$ disappear; we call these the ‘active’ domains. This process takes place on a timescale of $`𝒪(\mathrm{\Gamma }^1(d))=𝒪(ϵ^n)`$; because the timescales for different stages differ by factors of $`1/ϵ`$, we can treat them separately in the limit $`ϵ0`$. During stage $`n`$, the distribution of inactive domains ($`d>2^n`$) changes only because such domains can be created when smaller domains coalesce. Combining this with the (exact) bag model discussed above, we have for $`d>2^n`$
$$_\tau P(d,\tau )=\underset{2^{n1}<d^{}2^n}{}P(dd^{},\tau )[_\tau P(d^{},\tau )].$$
(3)
The term in square brackets is the rate at which active domains disappear; $`d^{}2^n`$ because inactive domains do not disappear. We use the rescaled time $`\tau =tϵ^n`$; during stage $`n`$ of the dynamics and in the limit $`ϵ0`$, it can take on any positive value $`\tau >0`$. The initial condition for (3) is the domain length distribution at the end of stage $`n1`$ of the dynamics, which we call $`P_n(d)`$ = $`P(d,\tau 0)`$. To calculate $`P_{n+1}(d)`$ = $`P(d,\tau \mathrm{})`$, introduce the generating function $`G(z,\tau )=_{2^{n1}<d}P(d,\tau )z^d`$, and its analog for the active domains, $`H(z,\tau )=_{2^{n1}<d2^n}P(d,\tau )z^d`$. From (3), one then finds
$`_\tau [G(z,\tau )H(z,\tau )]=G(z,\tau )_\tau H(z,\tau ).`$
This can be integrated to give $`[1G(z,\mathrm{})]/[1G(z,0)]=\mathrm{exp}[H(z,0)H(z,\mathrm{})]`$. But at the end of stage $`n`$, all domains that were active during that stage have disappeared, and so $`H(z,\mathrm{})=0`$. Defining the initial condition for $`G`$ as $`G_n(z)G(z,0)=_{2^{n1}<d}P_n(d)z^d`$ and similarly for the active generating function $`H_n(z)H(z,0)`$, we then have finally
$$G_{n+1}(z)1=[G_n(z)1]\mathrm{exp}[H_n(z)].$$
(4)
This exact result relates the domain length distributions $`P_n(d)`$ and $`P_{n+1}(d)`$ at the end of stages $`n1`$ and $`n`$ of the dynamics, as expressed through their generating functions. Iterating it from a given initial distribution $`P_0(d)`$ gives $`P_n(d)`$ for all $`n=1,2,\mathrm{}`$ We do this numerically by expressing (4) directly in terms of the probability distributions; the exponential is thus expanded into a series of convolutions of increasing order. Fig. 1 shows the results for the case where $`P_0(d)`$ is the equilibrium distribution (1) corresponding to an initial temperature of $`T_\mathrm{i}=\mathrm{}`$. Not unexpectedly, a scaling limit is approached for large $`n`$: The rescaled distributions $`\stackrel{~}{P}_n(x)=2^{n1}P_n(d)`$, where the scaled domain size is $`x=d/2^{n1}`$, converge to a limiting distribution $`\stackrel{~}{P}(x)`$ which is independent of the initial condition. Invariance under (4) gives an equation for the corresponding Laplace transforms $`g(s)`$ and $`h(s)`$ of $`\stackrel{~}{P}(x)`$
$`g(2s)1=[g(s)1]\mathrm{exp}[h(s)].`$
We find a self-consistent solution
$`\stackrel{~}{P}(x)`$ $`=`$ $`{\displaystyle \underset{m=1}{\overset{\mathrm{}}{}}}{\displaystyle \frac{(1)^{m1}}{m!}}{\displaystyle _1^{\mathrm{}}}{\displaystyle \underset{r=1}{\overset{m}{}}}{\displaystyle \frac{dx_r}{x_r}}\delta \left({\displaystyle \underset{s=1}{\overset{m}{}}}x_sx\right)`$ (5)
$`=`$ $`\mathrm{\Theta }(x1){\displaystyle \frac{1}{x}}\mathrm{\Theta }(x2){\displaystyle \frac{\mathrm{ln}(x1)}{x}}+\mathrm{}`$ (6)
where $`\mathrm{\Theta }(x)`$ is the Heaviside step function. This series has singularities in the $`k`$-th derivative at the integer values $`x=k+1`$, $`k+2`$, …The calculated $`\stackrel{~}{P}(x)`$ agrees well with the previous results obtained by direct iteration of (4) (Fig. 1). The average domain length in the scaling limit is given by $`\overline{d}_n=2^{n1}\overline{x}`$; from the results for $`\stackrel{~}{P}(x)`$ we find $`\overline{x}=\mathrm{exp}(\gamma )=1.78\mathrm{}`$, where $`\gamma `$ is Euler’s constant.
In order to compare the results to simulations, consider starting from an equilibrated state at some initial temperature, say $`T_\mathrm{i}=\mathrm{}`$, quench the system to temperatures $`T1`$ at time $`t=0`$ and observe its time evolution. If the results are plotted against the scaled time variable $`\nu =\mathrm{ln}(t)/\mathrm{ln}(1/ϵ)=T\mathrm{ln}t`$, then for $`T0`$ the $`n`$-th stage of the dynamics shrinks to the point $`\nu =n`$. In this limit we predict that, for $`n1<\nu <n`$, the domain length distribution is $`P_n(d)`$ as defined by the recursion (4). The average domain length $`\overline{d}`$ will follow a ‘staircase’ function, jumping at $`\nu =n`$ from $`\overline{d}_n=_dP_n(d)d`$ to $`\overline{d}_{n+1}`$. In the large $`\nu `$ scaling regime, this tells us that $`2^{\nu 1}\overline{x}\overline{d}2^\nu \overline{x}`$ (where $`\overline{x}=1.78\mathrm{}`$ from above), or $`\frac{1}{2}\overline{d}/(\overline{x}t^{T\mathrm{ln}2})1`$ when expressed in terms of ordinary time $`t`$.
We can therefore say that the system coarsens with an exponent that depends on temperature and is given by $`T\mathrm{ln}2`$ to lowest order in $`T`$. By extrapolating this coarsening law to the equilibrium domain length $`\overline{d}_{\mathrm{eq}}=\mathrm{exp}(1/T)+𝒪(1)`$, we then also have that the dominant divergence of the equilibration time of the system for $`T0`$ is $`t_{\mathrm{eq}}=\mathrm{exp}(1/T^2\mathrm{ln}2)`$.
In Fig. 2, we show the results of simulations for a range of values of $`ϵ=\mathrm{exp}(1/T)`$. We used a waiting time Monte Carlo algorithm combined with an efficient binary tree representation for the positions of the mobile spins. This let us access far larger systems ($`L=2^{15}`$) and longer times (up to $`t=10^{10}`$) than in previous simulations . The plateaus in $`\overline{d}(\nu )`$ that develop with decreasing $`ϵ`$ can clearly be seen, and their values are in good agreement with the predicted theoretical values. We also obtained the domain length distributions on the plateaus, by taking data at the minima of $`(d/d\nu )\overline{d}(\nu )`$ w.r.t. $`\nu `$. These are shown in Fig. 1 for the cases $`n=1,2,3`$ and are again in good agreement with our theory.
Our result for the equilibration time $`t_{\mathrm{eq}}=\mathrm{exp}(1/T^2\mathrm{ln}2)`$ is based on the extrapolation of the finite-$`\overline{d}`$ coarsening behavior $`\overline{d}t^{T\mathrm{ln}2}`$ into the equilibrium region $`\overline{d}=𝒪(1/ϵ)`$, where it is no longer strictly valid. We now show, however, that the same timescale is obtained from the initial decay of the spin-spin correlation function at equilibrium at low temperature $`T`$. It turns out that due to the asymmetric constraint the correlation function is site diagonal, $`(s_i(0)c)(s_j(t)c)=\delta _{ij}c[R(t)c]`$ . Here $`R(t)`$ is the probability that an up-spin at $`t=0`$ is also up at a later time $`t`$. With increasing $`t`$, it decays from $`R(0)=1`$ to the equilibrium concentration of up-spins, $`c=ϵ/(1+ϵ)`$. To find the initial decay of $`R(t)`$, consider again timescales $`t=𝒪(ϵ^\nu )`$ for finite $`\nu `$ and $`ϵ0`$. For $`\nu =n+0`$, all domains of length $`d2^n`$ will have disappeared. Therefore only up-spins that bounded longer domains at $`t=0`$ will have an $`𝒪(1)`$ probability of still being up. From the equilibrium distribution (1), one sees that they constitute a fraction $`(1+ϵ)^{2^n}`$ of the up-spins at $`t=0`$, and hence $`R(\nu =n+0)12^nϵ+𝒪(ϵ^2)`$ . Neglecting corrections of $`𝒪(ϵ^2)`$, the quantity $`\mathrm{ln}R(\nu )`$ thus lies between $`2^{\nu 1}ϵ`$ and $`2^\nu ϵ`$ (for $`\nu >0`$). Reverting to ordinary time, we have $`1/2[\mathrm{ln}R(t)]/(t/t_{\mathrm{eq}})^{T\mathrm{ln}2}1`$ for short times $`(t/t_{\mathrm{eq}})^{T\mathrm{ln}2}1`$. The relevant timescale that enters here is exactly the equilibration time $`t_{\mathrm{eq}}=\mathrm{exp}(1/T^2\mathrm{ln}2)`$ found above. We can thus identify the equilibration time for coarsening after a quench, with the equilibrium relaxation time; both have an EITS-divergence at low $`T`$.
Finally, we discuss briefly the spin-spin autocorrelation function for longer times $`(t/t_{\mathrm{eq}})^{T\mathrm{ln}2}=𝒪(1)`$, where the analysis becomes more involved . We have tackled this problem by extending the concept of domains to that of ‘superdomains’ which are bounded by up-spins that remain up on a given timescale. Combining this with a plausible hypothesis for the behavior of the relaxation timescales $`\mathrm{\Gamma }^1(d)`$ for $`d=𝒪(1/ϵ)`$, the following scenario seems likely : In the limit $`T0`$, $`R(t)`$ first decays linearly with the rescaled time variable $`\delta =(t/t_{\mathrm{eq}})^{T\mathrm{ln}2}`$. This is compatible to lowest order with a stretched exponential relaxation. But then the decay becomes much faster, and $`R`$ actually decays to zero at a finite value of $`\delta `$. (For nonzero $`T`$, there is a crossover into a slower decay, presumably exponential in $`t`$, at late times.) It would also be of interest to study the relaxation times of similar models in dimension $`D>1`$ .
Acknowledgements: Both authors are grateful for financial support from the Royal Society.
|
no-problem/9904/astro-ph9904073.html
|
ar5iv
|
text
|
# The 4 Year COBE DMR data is non-Gaussian
## I Introduction
In a recent letter apjl1 we have claimed that the 4 Year COBE DMR data exhibits evidence of non Gaussianity; our quantitative claim was that the hypotheses that this data set is due to a Gaussian random process can be ruled out at the $`98\%`$ confidence level. This came somewhat as a surprise given that this data set was lauded as strong evidence for the inflationary paradigm: its statistics were thought to be consistent with Gaussianity kog96a .
We have performed an extensive analysis of the 4 year COBE DMR data set and have been unable to find a non cosmological origin for the non-Gaussian signal. The resuls of this analysis are presented in bigpaper . In this report I will only address a few often raised questions:
* Why does this statistic get a different result from all the ones that were previously used?
* Isn’t the effect we are seeing just some systematic effect of the analysis?
* If it is signal what could it be?
## II The Statistic and Result
In our analysis we propose, and work with, an estimator for the normalized bispectrum. To construct such an estimator we work in the spherical harmonic representation $`a_\mathrm{}m`$ and consider the tensor product of $`3`$ $`\mathrm{\Delta }T_{\mathrm{}}=_ma_\mathrm{}m`$s One is interested in rotationally invariant quantities. These can be trivially obtained if one rewrites the tensor product in terms of the total angular momentum basis. The coefficient of the singlet will be the higher order invariant we are looking for. This procedure leads to an estimator of the bispectrum
$`\widehat{B}_{\mathrm{}}`$ $`=`$ $`\alpha _{\mathrm{}}{\displaystyle \underset{m_1m_2m_3}{}}𝒲_{m_1m_2m_3}^{\mathrm{}\mathrm{}\mathrm{}}a_{\mathrm{}m_1}a_{\mathrm{}m_2}a_{\mathrm{}m_3}`$
$`\alpha _{\mathrm{}}`$ $`=`$ $`{\displaystyle \frac{1}{(2\mathrm{}+1)^{\frac{3}{2}}}}\left(𝒲_{000}^{\mathrm{}\mathrm{}\mathrm{}}\right)^1`$ (1)
where $`𝒲_{m_1m_2m_3}^{\mathrm{}_1\mathrm{}_2\mathrm{}_3}`$ are the Wigner 3J coefficients. $`\widehat{B}_{\mathrm{}}`$ may then be divided by the appropriate power of an estimator for $`C_{\mathrm{}}`$ in order to make it dimensionless, and suitably normalised.
$`I_{\mathrm{}}^3`$ $`=`$ $`\left|{\displaystyle \frac{\widehat{B}_{\mathrm{}}}{(\widehat{C}_{\mathrm{}})^{3/2}}}\right|`$ (2)
Let us comment on a number of features of the estimator. Firstly this statistic is global on pixel space, i.e. all the estimators are a function of all the pixel values. This means that it will be very good at identifying the scale dependence of non-Gaussianity (in the same way as the $`C_{\mathrm{}}`$s are very discrimative of the scale dependence of the variance of the fluctuations); however it performs very poorly at identifying the location of any signal on the map. In this it contrasts with statistics which have been applied to the COBE data until now: the three point correlation function, the statistics of peaks and topological measures are all defined on pixel space kog96a . Secondly we have assumed statistical isotropy in constructing this estimator. We know, however, that the COBE data has a number of anisotropic features which violate this assumption, in particular the presence of the galaxy (which must be removed) and the anisotropic sky coverage of the observation pattern. The only way we can truly assess the significance of the $`I_{\mathrm{}}^3`$ is by comparing them to an ensemble of $`I_{\mathrm{}}^3`$ measured on Gaussian sky maps generated with exactly the same characteristics as that of the 4 year COBE DMR data. The assumptions of the Monte Carlo must be carefully checked.
We have tested the inverse noise variance weighted, average maps of the 53A, 53B, 90A and 90B COBE-DMR channels, with monopole and dipole removed, at resolution 6, in ecliptic pixelization. We use the extended galactic cut of banday97 , and benn96 to remove most of the emission from the plane of the Galaxy. We apply our statistics to the DMR maps before and after correction for the plausible diffuse foreground emission outside the galactic plane as described in kog96b , and COBE . To estimate the $`I_{\mathrm{}}^3`$s we set the value of the pixels within the galactic cut to 0 and the average temperature of the cut map to zero. We then integrate the map multiplied with spherical harmonics to obtain the estimates of the $`a_\mathrm{}m`$s and apply equations 1 and 2. We then compare the estimates to the distributions generated from Monte Carlos simulations of Gaussian maps.
Since the $`I_{\mathrm{}}^3`$ distributions are non Gaussian we generalize the $`\chi ^2`$ for a set of probability functions $`P_{\mathrm{}}(I_{\mathrm{}}^3)`$ associated with observations $`\{I_{\mathrm{}}^3\}`$ by defining the following functional
$$X^2=\frac{1}{N}\underset{\mathrm{}}{}X_{\mathrm{}}^2=\frac{1}{N}\underset{\mathrm{}}{}(2\mathrm{log}P_{\mathrm{}}(I_{\mathrm{}}^3)+\beta _{\mathrm{}}),$$
(3)
where the constants $`\beta _{\mathrm{}}`$ are defined so that for each term of the sum $`X_{\mathrm{}}^2=1`$. The definition reduces to the usual $`X^2`$ for Gaussian $`P_{\mathrm{}}`$. Again, we build a $`X^2`$ for the COBE-DMR data by means of Monte Carlo simulations. We proceed as follows. First we compute the distributions $`P(I_{\mathrm{}}^3)`$, for $`\mathrm{}=2,\mathrm{},18`$, for a Gaussian process as measured subject to our galactic cut, and pixel noises. These $`P(I_{\mathrm{}}^3)`$ were inferred from 25000 realizations (see Fig. 1). From these distributions we then build the $`X^2`$ defined in (3), taking special care with the numerical evaluation of the constants $`\beta _{\mathrm{}}`$. We call this function $`X_{COBE}^2`$. We then find its distribution $`F(X_{COBE}^2)`$ from 10000 random realizations. This is very well approximated by a $`\chi ^2`$ distribution with 12 degrees of freedom (Fig. 2). We then compute $`X_{COBE}^2`$ with the actual observations and find $`X_{COBE}^2=1.81`$. One can compute $`P(X_{COBE}^2<1.81)=0.98`$. Hence, it would appear that we can reject Gaussianity at the $`98\%`$ confidence level.
## III Is it a systematic effect?
The non-Gaussianity we are finding in the 4 year COBE DMR maps is surprising and fascinating enough that we have gone through an exhaustive test of all possible systematic effects. We summarize them in the following list:
> 1. Foregrounds contamination:
>
> * Dust (using the DIRBE sky maps and also the Schlegel et al dust model)
> * Synchrotron (with the Haslam template)
> * Foreground corrected maps
> 2. Noise model:
>
> * Anisotropic sky coverage
> * Noise correlations between different pixels
> * Analysis of noise templates
> 3. Galactic cut:
>
> * Dependence on shape (“custom” versus constant elevation)
> * Dependence on elevation
> * Dependence on monopole and dipole subtraction, before or after the cut, with or with out galaxy.
> 4. Systematic templates
>
> * Spurious offsets induced by the cut.
> * Instrument susceptibility to the Earth magnetic field.
> * Callibration errors .
> * Errors due to incorrect removal of the COBE Doppler and Earth Doppler signals.
> * Errors in correcting for emissions from the Earth, and eclipse effects.
> * Artifacts due to uncertainty in the correction for the correlation created by the low-pass filter on the lock-in amplifiers (LIA) on each radiometer
> * Errors due to emissions from the moon, and the planets.
> 5. Assumptions in Monte Carlos:
>
> * Dependence on tilt
> * Dependence on smooth versus discontinuous power spectrum
> * Dependence on beam shape
> * Dependence on pixelization.
Let us just highlight the foreground tests. In figure 3 (left panel) we plot the $`I_{\mathrm{}}^3`$s estimated directly from the dust maps produced by the DIRBE $`100\mu m`$ and $`240\mu m`$ channels and from the dust template constructed in sfd98 using the DIRBE and IRAS maps. The feature we identify in the DMR data is not there. Furthemore, if we use the dust templates to subtract any foreground contribution (see figure 3 right panel) the value of $`I_{16}^3`$ actually increases!
We have found that none of the effects listed above change our result by much, in fact for most corrections the confidence level rises to over $`99\%`$. We therefore claim that the non-Gaussianity we find in the COBE DMR data is not due to any of the known systematics.
## IV The nature of the signal in $`I_{16}^3`$
The structure of the non-Gaussian signal is truly intriguing. It manifests itself as a spectral “spike” at $`\mathrm{}=16`$; it is difficult to associate such a pattern to some known or speculated source of fluctuations. If the signal is cosmological, the minimal inflationary models cannot be right. On the other hand it is not obvious that the main competitor to inflation, topological defects, could explain this type of non-Gaussianity. Topological defects are non-Gaussian, but in ways which are often more subtle than commonly thought. An interesting possibility was recently proposed by Peebles peeb . This is an isocurvature model in which the underlying fluctuations are not a Gaussian random field, but the square of a Gaussian random field. The model is based on non minimal inflation, but produces fluctuations radically different from minimal inflationary fluctuations.
A more useful exercise is to try and understand, given the characteristics of the experiment, what the power spectrum of such a non-Gaussian signal might be. The simplest thing to consider is combination of three signals. On very large scales (upto $`\mathrm{}=1214`$) we have a Gaussian sky signal. It may be fundamentally Gaussian or merely a manifestation of the law of large numbers. For $`\mathrm{}14`$ the sky signal is non-Gaussian, and this should manifest itself all the way upto much higher $`\mathrm{}`$s. However the instrument noise of the 4 Year COBE DMR data sets starts to dominate the $`I_{\mathrm{}}^3`$s at $`\mathrm{}=1820`$, where the signal to noise drops below unity. Given that the noise has been shown to be extremely well characterized by a Gaussian, the sky map will manifest itself as Gaussian for these higher $`\mathrm{}`$s.
Following they publications of our result, a number of groups have reported similar results: Pando et al pando have applied a wavelet based technique and find evidence that the non-Gaussian signal is localized in the northern hemisphere (a result we tentatively confirm bigpaper ) while Novikov et al nov have applied topological tests to detect non-Gaussianity in the COBE DMR data. According to skeptics, this may merely reflect a change in the psychological prior, triggered by our work. More seriously one should remember that the work performed by us and by these groups makes use of the same data set. Therefore this work provides an independent confirmation of our analysis of the DMR maps. In the very least this may mean a revision of the data analysis techniques which are currently in vogue for power spectrum estimation from CMB data sets: the underlying assumption in these Bayesian techniques is that the data set is Gaussian (greatly simplifying the estimation algorithms) gorski97 . However the fact that COBE data set seems to be non-Gaussian raises the question weather the current estimates of the $`C_{\mathrm{}}`$s of the COBE data set are as accurate as they are claimed to be.
If the sky is truly non-Gaussian in the way we describe above, then we are at the threshold of uncovering its statistical nature with the higher resolution experiments that are coming online in particular with the BOOMERANG LDB experiment (described elsewhere in these proceedings) and ultimately with the MAP and Planck Surveyor satellite experiments.
## Acknowledgments
We thank the organizers for an excellent meeting. We thank JNICT, NASA-ADP, NASA-COMBAT, NSF, RS, Starlink, TAC for support.
|
no-problem/9904/astro-ph9904108.html
|
ar5iv
|
text
|
# The Cosmic Microwave Background and Particle Physics11footnote 1To appear in Annu. Rev. Nucl. Part. Sci. (1999)
## 1 Overview of the Cosmic Microwave Background
In 1948, Alpher & Hermann realized that if light elements were produced in a hot big bang, as Gamow and others had suggested , then the Universe today should have a temperature of about 5 K. When Penzias & Wilson discovered an anomalous background in 1964, consistent with a blackbody spectrum at a temperature of $`3`$ K , Dicke and his collaborators immediately recognized it as the radiation associated with this nonzero cosmological temperature . Subsequent observations that confirm a remarkable degree of isotropy (apart from a dipole , which can be interpreted as our motion of $`627\pm 22`$ km s<sup>-1</sup> with respect to the blackbody rest frame ) suggest an extragalactic origin for this cosmic microwave background (CMB). Strong upper limits to any angular cross-correlation between the CMB temperature and the extragalactic X-ray background intensity suggest that the CMB comes from redshifts greater than those ($`z24`$) probed by the active galactic nuclei and galaxy clusters that produce the X-ray background. This evidence, as well as the exquisite blackbody spectrum of the CMB , further supports the notion that this radiation is the cosmological blackbody postulated by Alpher & Hermann.
Although they have a Planck spectrum, CMB photons are not in thermal equilibrium. The mean free path for scattering of photons in the Universe must be huge, or else we would not see galaxies and quasars out to distances of thousands of Mpc.<sup>2</sup><sup>2</sup>2Mpc$`=3.3\times 10^6`$ light years$`=3.09\times 10^{24}`$ cm. So where did these photons come from? At early times ($`t\begin{array}{c}<\hfill \\ \hfill \end{array}10^5`$ y; redshifts $`z\begin{array}{c}>\hfill \\ \hfill \end{array}1000`$), the temperature of the Universe exceeded an eV, so the Universe consisted of a plasma of free electrons and light nuclei. CMB photons were tightly coupled to this plasma via Thomson scattering from the free electrons. At a redshift of $`z1000`$, the temperature dropped below a few eV, and electrons and nuclei combined to form atoms. At this point, photons ceased interacting. A detailed analysis of “recombination” and the almost simultaneous (although slightly later) decoupling of photons shows that CMB photons last scattered near a redshift of $`z1100`$ .
When we look at these CMB photons coming to us from all directions in the sky, we are therefore looking directly at a spherical surface in the Universe that surrounds us at a distance of $`10^4`$ Mpc, as it was when the Universe was only about 300,000 years old. The temperature of the CMB is found to be the same, to roughly one part in $`10^5`$, in every direction on the sky. This remarkable isotropy poses a fundamental conundrum for the standard big-bang theory. When these photons last scattered, the size of a causally connected region of the Universe was roughly 300,000 light years, and such a region subtends an angle of only one degree on the sky. Thus, when we look at the CMB, we are looking at roughly 40,000 causally disconnected regions of the Universe. How is it, then, that each of these has the same temperature to one part in $`10^5`$? This is the well-known isotropy, homogeneity, or horizon problem.
Another fundamental question in cosmology today is the origin of the large-scale structure of the galaxy distribution. The simplest and most plausible explanation is that the observed inhomogeneities grew from tiny density perturbations in the early Universe via gravitational instability. Mass from underdense regions is drawn towards overdense regions, and in this way, small primordial perturbations are amplified into the structure we see in the Universe today. New support for this hypothesis was provided by the Cosmic Background Explorer (COBE) detection of temperature differences in the CMB of roughly one part in $`10^5`$ . Heuristically, density perturbations induce gravitational-potential perturbations at the surface of last scatter; photons that arrive from denser regions climb out of deeper potential wells and thus appear redder than those from underdense regions (the Sachs-Wolfe effect ). Thus, the temperature fluctuations seen with COBE provide a snapshot of the tiny primordial perturbations that gave rise to the large-scale structure we see in the Universe today. But this raises a second question: If large-scale structure grew via gravitational infall from tiny inhomogeneities in the early Universe, where did these primordial perturbations come from?
Before COBE, there was no shortage of ideas for the origin of large-scale structure, and, quite remarkably, all causal mechanisms for producing primordial perturbations have come from new ideas in particle theory: primordial adiabatic perturbations from inflation , late-time phase transitions , a loitering Universe , scalar-field ordering , topological defects (such as cosmic strings , domain walls , textures , or global monopoles ), superconducting cosmic strings , isocurvature axion perturbations , etc.
However, after COBE, primordial adiabatic perturbations (perturbations to the total density with equal fractional number-density perturbations in each species in the Universe) seem to provide the only workable models. Such perturbations are produced naturally during inflation, a period of exponential expansion in the early Universe driven by the vacuum energy associated with some new scalar field . With adiabatic perturbations, hotter regions at the surface of last scatter are embedded in deeper potential wells, so the reddening due to the gravitational redshift of the photons from these regions partially cancels the higher intrinsic temperatures. When normalized to the amplitude of density perturbations indicated by galaxy surveys, alternative models generically produce a larger temperature fluctuation than that measured by COBE . Recently, more detailed calculations of the expected CMB-anisotropy amplitude have led proponents of topological defects, the primary alternative to inflation, to concede that these models have difficulty accounting for the origin of large-scale structure .
Although inflation now seems to provide the best candidate for the origin of large-scale structure, the primary attraction of inflation was originally that it provided (and still does) the best (if not only) solution of the horizon problem. For these reasons, inflation has taken center stage in cosmology. Although inflation was for a long time speculative physics beyond the realm of experimental tests, we are now entering a new era in which the predictions of inflation will be tested with unprecedented precision by CMB measurements.
The primary focus of this article is therefore to review the predictions of inflation and how they will be tested with the CMB. Although inflation currently seems to provide the most promising paradigm for the origin of large-scale structure, it is not yet well established. Moreover, although the simplest topological-defect models seem to be ruled out, it is still certainly plausible that some more involved models may be able to account for large-scale structure. We therefore review the CMB predictions of topological-defect models. We also discuss a number of other promising links between the CMB and particle physics that do not necessarily have to do with the origin of structure, e.g. dark matter, neutrino properties, decaying particles, cosmological magnetic fields from early-Universe phase transitions, parity violation, gravity theories, time variation of fundamental parameters, and baryogenesis scenarios.
We are unfortunately unable to cover the larger bodies of excellent work on the CMB in general, nor on the intersections between particle physics and cosmology more generally. Fortunately, a number of excellent reviews cover those subjects, to which we cannot do justice here. Lyth and Riotto review particle-physics models of inflation; Liddle and Lyth discuss structure formation in inflation-inspired cold-dark-matter models. Lidsey et al review the production of density perturbations and reconstruction of the inflaton potential from the power spectra of density perturbations and gravitational waves. White, Scott, and Silk review the CMB and structure formation, and Hu and White provide a brief review of the theory of CMB polarization. Finally, see References for reviews of the Sunyaev-Zeldovich effect (the scattering of CMB photons from hot gas in clusters of galaxies), an intriguing and potentially very important probe of the physics of clusters.
## 2 Cosmic Microwave Background Observables
### 2.1 The Frequency Spectrum
Standard cosmology predicts the CMB frequency spectrum to be that of a perfect blackbody,
$$S(\nu ;T)=\frac{2hc^2\nu ^3}{e^x1},$$
(1)
where $`x=hc\nu /kT`$, $`h`$ is Planck’s constant, $`c`$ is the velocity of light, $`\nu `$ is the frequency, $`k`$ is Boltzmann’s constant, and $`T`$ is the temperature. Of the infinitude of possible distortions to this spectrum, two common forms often considered in the literature—Bose-Einstein and Compton distortions—could arise from basic physical processes before recombination.
If photons are released into the Universe from some nonthermal process (e.g. decay of a massive particle) when the temperature of the Universe exceeds roughly 1 keV (redshifts $`z\begin{array}{c}>\hfill \\ \hfill \end{array}10^6`$ when the age of the Universe is $`t\begin{array}{c}<\hfill \\ \hfill \end{array}10^7`$ sec), they will come into complete thermal equilibrium with the photons in the primordial plasma. More precisely, they attain kinetic equilibrium through Compton scattering, double Compton scattering, and bremsstrahlung, and they attain chemical equilibrium (chemical potential $`\mu =0`$) because the rate for photon-number-changing processes (e.g. $`\gamma \gamma \gamma \gamma \gamma `$) that maintain a chemical potential $`\mu =0`$ exceeds the expansion rate. Therefore, if any electromagnetic energy is released into the Universe at such early times, it will have no observable effect on the CMB. However, if photons are released at later times (but still before recombination), they can distort the CMB frequency spectrum .
#### 2.1.1 Bose-Einstein distortion
Nonthermal photons produced in the redshift range $`10^5\begin{array}{c}<\hfill \\ \hfill \end{array}z\begin{array}{c}<\hfill \\ \hfill \end{array}3\times 10^6`$ (temperatures $`T0.11`$ keV and ages $`t10^{79}`$ sec) can still attain kinetic equilibrium, but they will not attain chemical equilibrium, as interactions that change the photon number occur less rapidly than the expansion rate. If electromagnetic energy is released at these times, the CMB frequency dependence will be that of a Bose-Einstein gas with a nonzero chemical potential,
$$S_\mu (\nu ;T,\mu )=\frac{2hc^2\nu ^3}{e^{x+\mu }1},$$
(2)
where $`\mu `$ is the (dimensionless) chemical potential. The Far Infrared Absolute Spectrophotometer (FIRAS) result for $`\mu `$ is $`\mu =1\pm 10\times 10^5`$ or a 95% confidence-level upper limit of $`|\mu |<9\times 10^5`$ . It is possible that values of $`\mu `$ as small as $`10^6`$ could be probed by a future satellite mission .
#### 2.1.2 Compton distortion
If photons are released at later times ($`z\begin{array}{c}<\hfill \\ \hfill \end{array}10^5`$) but still before recombination ($`z1100`$; temperatures $`T1100`$ eV and times $`t10^{913}`$ sec), they do not have enough time to come to either thermal or kinetic equilibrium and wind up producing a “Compton distortion” of the form
$$S_y(\nu ;T,y)=\frac{2hc^2\nu ^3}{e^x1}\left(1+yx\frac{1}{1e^x}\left[x\mathrm{coth}(x/2)4\right]\right),$$
(3)
to linear order in $`y`$ (the Kompaneets or “Compton-$`y`$” parameter). If some CMB photons were rescattered after recombination by a hot intergalactic gas, this would also produce a Compton-$`y`$ distortion. The FIRAS result for this type of distortion is $`y=1\pm 6\times 10^6`$ or an upper limit of $`|y|<15\times 10^6`$ at the 95% confidence level . The consensus among the experimentalists we have surveyed seems to be that it would be difficult to improve on this limit.
### 2.2 Temperature and Polarization Power Spectra
The primary aim of forthcoming CMB satellite experiments, such as NASA’s Microwave Anisotropy Probe (MAP) and the European Space Agency’s Planck Surveyor , will be to map the temperature $`T(\widehat{𝐧})`$ of the CMB and its linear polarization, described by Stokes parameters $`Q(\widehat{𝐧})`$ and $`U(\widehat{𝐧})`$, as functions of position $`\widehat{𝐧}`$ on the sky. Several temperature-polarization angular correlation functions, or equivalently, power spectra, can be extracted from such maps. These quantities can be compared with detailed predictions from cosmological models.
#### 2.2.1 Harmonic analysis for temperature anisotropies and polarization
Temperature Anisotropies
The temperature map can be expanded in spherical harmonics,
$$\frac{T(\widehat{𝐧})}{T_0}=1+\underset{lm}{}a_{(lm)}^\mathrm{T}Y_{(lm)}(\widehat{𝐧}),$$
(4)
where the mode amplitudes are given by
$$a_{(lm)}^\mathrm{T}=\frac{1}{T_0}𝑑\widehat{𝐧}T(\widehat{𝐧})Y_{(lm)}^{}(\widehat{𝐧});$$
(5)
this follows from orthonormality of the spherical harmonics. Here, $`T_0=2.728\pm 0.002`$ K is the cosmological mean CMB temperature .
Linear polarization
The Stokes parameters (where $`Q`$ and $`U`$ are measured with respect to the polar $`\widehat{\theta }`$ and azimuthal $`\widehat{\varphi }`$ axes) are components of a $`2\times 2`$ symmetric traceless tensor with two independent components,
$$𝒫_{ab}(\widehat{𝐧})=\frac{1}{2}\left(\begin{array}{cc}Q\left(\widehat{𝐧}\right)& U(\widehat{𝐧})\mathrm{sin}\theta \\ U(\widehat{𝐧})\mathrm{sin}\theta & Q(\widehat{𝐧})\mathrm{sin}^2\theta \end{array}\right),$$
(6)
where the subscripts $`ab`$ are tensor indices, and $`Q`$ and $`U`$ are given in temperature units. Just as the temperature is expanded in terms of spherical harmonics, the polarization tensor can be expanded ,
$$\frac{𝒫_{ab}(\widehat{𝐧})}{T_0}=\underset{lm}{}\left[a_{(lm)}^\mathrm{G}Y_{(lm)ab}^\mathrm{G}(\widehat{𝐧})+a_{(lm)}^\mathrm{C}Y_{(lm)ab}^\mathrm{C}(\widehat{𝐧})\right],$$
(7)
in terms of tensor spherical harmonics, $`Y_{(lm)ab}^\mathrm{G}`$ and $`Y_{(lm)ab}^\mathrm{C}`$. It is well known that a vector field can be decomposed into a curl (C) and a curl-free (gradient) (G) part. Similarly, a $`2\times 2`$ symmetric traceless tensor field can be decomposed into a tensor analogue of a curl and a gradient part; the $`Y_{(lm)ab}^\mathrm{G}`$ and $`Y_{(lm)ab}^\mathrm{C}`$ form a complete orthonormal basis for the “gradient” (i.e. curl-free) and “curl” components of the tensor field, respectively.<sup>3</sup><sup>3</sup>3Our G and C are sometimes referred to as the “scalar” and “pseudo-scalar” components , respectively, or with slightly different normalization as E and B modes (although these should not be confused with the radiation’s electric- and magnetic-field vectors). Lengthy but digestible expressions for the $`Y_{(lm)ab}^\mathrm{G}`$ and $`Y_{(lm)ab}^\mathrm{C}`$ are given in terms of derivatives of spherical harmonics and also in terms of Legendre functions in Reference . The mode amplitudes in Equation 7 are given by
$`a_{(lm)}^\mathrm{G}`$ $`=`$ $`{\displaystyle \frac{1}{T_0}}{\displaystyle 𝑑\widehat{𝐧}𝒫_{ab}(\widehat{𝐧})Y_{(lm)}^{\mathrm{G}ab}(\widehat{𝐧})},`$ (8)
$`a_{(lm)}^\mathrm{C}`$ $`=`$ $`{\displaystyle \frac{1}{T_0}}{\displaystyle 𝑑\widehat{𝐧}𝒫_{ab}(\widehat{𝐧})Y_{(lm)}^{\mathrm{C}ab}(\widehat{𝐧})},`$ (9)
which can be derived from the orthonormality properties of these tensor harmonics . Thus, given a polarization map $`𝒫_{ab}(\widehat{𝐧})`$, the G and C components can be isolated by first carrying out the transformations in Equation 9 to the $`a_{(lm)}^\mathrm{G}`$ and $`a_{(lm)}^\mathrm{C}`$ and then summing over the first term on the right-hand side of Equation 7 to get the G component and over the second term to get the C component.
#### 2.2.2 The power spectra
Theories for the origin of large-scale structure predict that the mass distribution in the Universe is a single realization of a statistically isotropic random field. In other words, the Fourier components $`\stackrel{~}{\delta }(\stackrel{}{k})`$ of the fractional density perturbation $`\delta (\stackrel{}{x})=[\rho (\stackrel{}{x})\overline{\rho }]/\overline{\rho }`$ \[where $`\rho (\stackrel{}{x})`$ is the density at comoving position $`\stackrel{}{x}`$ and $`\overline{\rho }`$ is the universal mean density\] are random variables that have expectation values $`\stackrel{~}{\delta }(\stackrel{}{k})=0`$ and covariance given by
$$\stackrel{~}{\delta }(\stackrel{}{k})\stackrel{~}{\delta }(\stackrel{}{k}^{})=(2\pi )^3\delta _D(\stackrel{}{k}+\stackrel{}{k}^{})P_s(k).$$
(10)
Here $`P_s(k)`$ is the scalar power spectrum (so called because density perturbations produce scalar perturbations to the spacetime metric), or alternatively, the power spectrum for the spatial mass distribution. Statistical isotropy demands that the power spectrum depends only on the amplitude (rather than orientation) of $`\stackrel{}{k}`$.
Because the temperature perturbation and polarization of the CMB are due to density perturbations, the $`a_{(lm)}^\mathrm{X}`$ must be random variables with zero mean, $`a_{(lm)}^\mathrm{X}=0`$, and covariance,
$$\left(a_{(l^{}m^{})}^\mathrm{X}^{}\right)^{}a_{(lm)}^\mathrm{X}=C_l^{\mathrm{XX}^{}}\delta _{ll^{}}\delta _{mm^{}},$$
(11)
for $`\mathrm{X},\mathrm{X}^{}=\{\mathrm{T},\mathrm{G},\mathrm{C}\}`$. The statistical independence of each $`lm`$ mode (i.e. the presence of the Kronecker deltas) is a consequence of statistical isotropy. The scalar spherical harmonics $`Y_{(lm)}`$ and the gradient tensor spherical harmonics $`Y_{(lm)}^\mathrm{G}`$ have parity $`(1)^l`$, whereas the curl tensor spherical harmonics $`Y_{(lm)}^\mathrm{C}`$ have the opposite parity, $`(1)^{l+1}`$. Thus, $`C_l^{\mathrm{TC}}=C_l^{\mathrm{GC}}=0`$ if the physics that gives rise to temperature anisotropies and polarization is parity-invariant. In this case, the two-point statistics of the CMB temperature-polarization map are completely specified by the four sets of moments, $`C_l^{\mathrm{TT}}`$, $`C_l^{\mathrm{TG}}`$, $`C_l^{\mathrm{GG}}`$, and $`C_l^{\mathrm{CC}}`$. Nonzero $`C_l^{\mathrm{TC}}`$ or $`C_l^{\mathrm{GC}}`$ would provide a signature of cosmological parity breaking.
#### 2.2.3 Angular correlation functions
The temperature two-point correlation function is
$$C^{\mathrm{TT}}(\alpha )=\frac{\mathrm{\Delta }T(\widehat{𝐦})}{T}\frac{\mathrm{\Delta }T(\widehat{𝐧})}{T}_{\widehat{𝐦}\widehat{𝐧}=\mathrm{cos}\alpha },$$
(12)
where the average is over all pairs of points on the sky separated by an angle $`\alpha `$. It can be written in terms of the temperature power spectrum (i.e. the $`C_l^{\mathrm{TT}}`$) as
$$C^{\mathrm{TT}}(\alpha )=\underset{l}{}\frac{2l+1}{4\pi }C_l^{\mathrm{TT}}P_l(\mathrm{cos}\alpha ),$$
(13)
where $`P_l(\mathrm{cos}\alpha )`$ are Legendre polynomials. Likewise, orthonormality of Legendre polynomials guarantees that the multipole coefficients, $`C_l^{\mathrm{TT}}`$, can be written as integrals over the product of the correlation function and a Legendre polynomial. Thus, specification of the correlation function is equivalent to specification of the power spectrum, and vice versa. CMB theorists and experimentalists have now adopted the convention of showing model predictions and presenting experimental results as power spectra ($`C_l`$) rather than as correlation functions, and we subsequently stick to this convention. Auto- and cross-correlation functions for the Stokes parameters and temperature-polarization cross-correlation functions can also be defined and written in terms of the polarization and temperature-polarization power spectra , but we do not list them here.
In practice, the temperature intensity (or polarization) can never be determined at a given point on the sky; it can only be measured by a receiver of some finite angular resolution (referred to as a “finite beamwidth”). Thus, the correlation function in Equation 13 cannot be measured; one can only measure a smoothed version. Likewise, a finite beamwidth $`\theta _{\mathrm{fwhm}}`$ (at full-width half maximum) limits determination of the power spectrum to multipole moments $`l\begin{array}{c}<\hfill \\ \hfill \end{array}200(\theta _{\mathrm{fwhm}}/1^{})^1`$.
The Differential Microwave Radiometer (DMR) experiment aboard COBE produced the first map of the temperature of the CMB. The receivers also provided some information on the polarization, but the sensitivity was not sufficient to detect the signal expected in most cosmological models. Measurements of the CMB intensity were made at three different frequencies (31.5, 53, and 90 GHz) near the blackbody peak to disentangle the possible contribution of foreground contaminants (e.g. dust or synchrotron emission) from the Galaxy, as these would have a frequency spectrum that differs from a blackbody. The DMR beamwidth was $`7^{}`$, so the temperature power spectrum was recoverable only for $`l\begin{array}{c}<\hfill \\ \hfill \end{array}15`$. MAP, scheduled for launch in the year 2000, will map the sky with an angular resolution better than $`0.3^{}`$ ($`l\begin{array}{c}<\hfill \\ \hfill \end{array}700`$). MAP should have sufficient sensitivity to see the polarization, although probably not enough to map the polarization power spectra precisely (the polarization is expected to be roughly an order of magnitude smaller than the temperature anisotropy). The Planck Surveyor, scheduled for launch around 2007, will map the temperature and polarization with even finer angular resolution (out to $`l\begin{array}{c}<\hfill \\ \hfill \end{array}20003000`$). Its sensitivity should be sufficient to map the polarization power spectra expected from density perturbations (discussed below) with good precision.
### 2.3 Gaussianity
Angular three-point and higher $`n`$-point temperature correlation functions can be constructed analogously to the two-point correlation functions in Equation 12. Fourier analogs of higher-order correlation functions can be defined. In particular, the temperature bispectrum $`B(l_1,l_2,l_3)`$ is the $`l`$-space version of the temperature three-point correlation function. It is defined by
$$a_{(l_1m_1)}^\mathrm{T}a_{(l_2m_2)}^\mathrm{T}a_{(l_3m_3)}^\mathrm{T}=\left(\begin{array}{ccc}l_1& l_2& l_3\\ m_1& m_2& m_3\end{array}\right)B(l_1,l_2,l_3),$$
(14)
where the array is the Wigner $`3j`$ symbol. This particular $`m`$ and $`l`$ dependence follows from the assumption of statistical isotropy. Closely related statistics include the skewness and kurtosis (respectively, the three- and four-point correlation functions at zero lag) and higher cumulants . As discussed further below, inflationary models predict the primordial distribution of perturbations to be perfectly (or very nearly) Gaussian. Gaussianity dictates that all the odd-$`n`$ $`n`$-point correlation functions vanish and that for even $`n`$, the higher $`n`$-point correlation functions can be given in terms of the two-point correlation function.
Numerous other statistical tests of CMB Gaussianity have also been proposed, including (but not limited to) topology of temperature contours and the related Minkowski functionals , temperature peak statistics , Fourier space patterns , and wavelet analysis .
## 3 Predictions of Inflation
If the energy density of the Universe is dominated by matter or radiation, then the expansion of the Universe is decelerating. If so, the horizon grows more rapidly than the scale factor. In such a Universe, objects that are now beyond our horizon and therefore inaccessible to us will eventually enter the horizon and become visible. Thus, the observable Universe contains more information and is more complicated at later times. Inflation postulates the existence of a period of accelerated expansion in the early Universe. In such a Universe, the scale factor grows more rapidly than the horizon. Thus, objects currently visible to any given observer will eventually exit that observer’s horizon (in much the same way that objects that fall into a black hole disappear when they pass through the black hole’s event horizon). A period of accelerated expansion therefore makes the Universe simpler and smoother. As we now discuss, this accelerated expansion also generically drives the observable Universe to be flat and provides a mechanism for producing primordial density perturbations and gravitational waves.
### 3.1 Scalar-Field Dynamics
Inflation supposes the existence of some new scalar field $`\varphi `$ (the “inflaton”), with a potential $`V(\varphi )`$ that roughly resembles either of those shown in Figure 1. The shape is not particularly important. All we require is that, at some time in the history of the Universe, the field is displaced from the minimum of the potential, and then it rolls slowly How slowly is slowly enough? This is determined by the Friedmann equation,
$$H^2\left(\frac{\dot{a}}{a}\right)^2=\frac{8\pi G\rho }{3}\frac{k}{a^2}=\frac{8\pi G}{3}\left(\frac{1}{2}\dot{\varphi }^2+V(\varphi )\right)\frac{k}{a^2},$$
(15)
which governs the time $`t`$ dependence of the scale factor $`a(t)`$ of the Universe (the dot denotes derivative with respect to time), as well as the scalar-field equation of motion,
$$\ddot{\varphi }+3H\dot{\varphi }+V^{}(\varphi )=0,$$
(16)
where $`V^{}=dV/d\varphi `$. In Equation 15, $`\rho `$ is the energy density of the Universe, which is assumed to be dominated by the inflaton potential-energy density $`V(\varphi )`$ and kinetic-energy density $`\dot{\varphi }^2/2`$. The term $`k/a^2`$ is the curvature term, and $`k>0`$, $`k<0`$, or $`k=0`$ for a closed, open, or flat Universe, respectively. Note that the expansion acts as a friction term for the scalar-field equation of motion in Equation 16. If
$$ϵ\frac{m_{\mathrm{Pl}}^2}{16\pi }\left(\frac{V^{}}{V}\right)^21,$$
(17)
and
$$\eta \frac{m_{\mathrm{Pl}}^2}{8\pi }\left[\frac{V^{\prime \prime }}{V}\frac{1}{2}\left(\frac{V^{}}{V}\right)^2\right]1,$$
(18)
where $`m_{\mathrm{Pl}}=1.22\times 10^{19}`$ GeV is the Planck mass, then the field rolls slowly enough so that the requirement for acceleration \[$`p<\rho /3`$, where $`p=(1/2)\dot{\varphi }^2V`$ is the pressure and $`\rho =(1/2)\dot{\varphi }^2+V`$ is the energy density\] is satisfied. (Note that the definitions of $`ϵ`$ and especially of $`\eta `$ may differ in some papers.)
The identity of the inflaton remains a mystery. It was originally hypothesized to be associated with a Higgs field in grand unified theories, but it may also have something to do with Peccei-Quinn symmetry breaking, a dilaton field, electroweak-symmetry breaking , some new pseudo-Nambu-Goldstone symmetry , supersymmetry , or some other new physics. As discussed below, the primary predictions of slow-roll inflation do not depend on the details of the physics responsible for inflation but rather on some gross features that are easily quantified.
### 3.2 The Geometry
Given any inflationary potential $`V(\varphi )`$, the equations of motion in Equations 16 and 17 can be solved numerically, if not analytically. Heuristically, during inflation, the potential $`V(\varphi )`$ is roughly constant, and $`\dot{\varphi }^2V(\varphi )`$. If the curvature term is appreciable initially, it rapidly decays relative to the potential term as the Universe expands, and the solution for the scale factor approaches an exponential, $`a(t)e^{Ht}`$. If $`k`$ is nonzero initially, the curvature term is then driven exponentially to zero during the inflationary epoch. In other words, if the duration of inflation is sufficiently long to place the observable Universe in a causally connected pre-inflationary patch, then the curvature radius is generically inflated to an exponentially (and unobservably) large value. In the language above, any initial nonzero curvature disappears beyond the horizon during accelerated expansion. Thus, the first prediction of slow-roll inflation is that the Universe should be flat today; i.e. the total density of all components of matter should sum to the critical density.
#### 3.2.1 “Open inflation”
It is, of course, mathematically possible that inflation did occur but that the inflationary epoch was prematurely terminated at just the right time so that the Universe today would be open with density $`\mathrm{\Omega }_00.3`$. Such a model requires some additional mechanism (e.g. another prior period of inflation and/or some arbitrary new “feature” in the inflaton potential) to solve the isotropy problem as well as to produce density perturbations. Several such open-inflation models have recently been constructed , motivated by observations that suggest $`\mathrm{\Omega }_00.3`$. The predictions of a scale-invariant spectrum and Gaussian perturbations (discussed below) are the same as in ordinary inflation, but the Universe would be open. We do not find these models even nearly as compelling as the ordinary slow-roll models that produce a flat Universe, although some theorists may disagree. Fortunately, the correct model will not be determined by debate; forthcoming CMB measurements, described below, should distinguish conclusively between these two classes—simple and elegant versus complicated and unappealing—of inflationary models.
### 3.3 Density Perturbations, Gravitational Waves, and the Inflationary Observables
#### 3.3.1 Production of density perturbations
Density perturbations are produced as a result of novel quantum-mechanical effects (analogous to the production of Hawking radiation from black holes) that occur in a Universe with accelerating expansion . This process has been reviewed in detail recently , so here we review the physics only heuristically. Consider perturbations $`\delta \varphi (\stackrel{}{x},t)`$ (as a function of comoving position $`\stackrel{}{x}`$) to the homogeneous slowly rolling field $`\varphi (t)`$. These perturbations satisfy a massless Klein-Gordon equation, and the equation of motion for each Fourier mode $`\stackrel{~}{\delta \varphi }(\stackrel{}{k})`$ is that of a simple harmonic oscillator in an expanding Universe. At sufficiently early times, when the wavelength of any given Fourier mode is less than the Hubble radius $`H^1`$, it undergoes quantum-mechanical zero-point oscillations. However, if the expansion is accelerating, then the physical wavelength of this comoving scale grows faster than the Hubble radius and eventually becomes larger than $`H^1`$. At this point, crests and troughs of a given Fourier mode can no longer communicate, and the zero-point fluctuation becomes frozen in as a classical perturbation $`\delta \varphi (\stackrel{}{x})`$ to the scalar field. Because the inflaton potential is not perfectly flat, this induces perturbations to the density $`\delta \rho (\stackrel{}{x})=(V/\varphi )\delta \varphi (\stackrel{}{x})`$.
#### 3.3.2 Production of gravitational waves
Tensor perturbations to the spacetime metric (i.e. gravitational waves) satisfy a massless Klein-Gordon equation. A stochastic background of gravitational waves are therefore produced in the same way as classical perturbations to the inflaton are produced . Moreover, the power spectra for the inflaton-field perturbations and for the tensor metric perturbations should be identical. The power spectrum of density perturbations is a little different from that for gravitational waves because a density perturbation is related to a scalar-field perturbation by $`\delta \rho =(V/\varphi )\delta \varphi `$. The production of scalar and tensor perturbations depends only on the expansion rate during inflation. If the expansion rate were perfectly constant during inflation, it would produce flat scalar and tensor power spectra, $`P_sk`$ (the “Peebles-Harrison-Zeldovich” spectrum) and $`P_t(k)\mathrm{constant}`$.
#### 3.3.3 Inflationary observables
A constant expansion rate is an oversimplification because the field must in fact be rolling slowly down the potential during inflation. Given any specific functional form for the potential, it is straightforward, using the tools of quantum field theory in curved spacetimes (see e.g. ), to predict precisely the functional forms of $`P_s(k)`$ and $`P_t(k)`$. Measurement of these power spectra could then be used to reconstruct the inflaton potential . Since the field must be rolling fairly slowly during inflation, a good approximation (in most models) can be obtained by expanding about a constant expansion rate. In this slow-roll approximation, the primordial scalar power spectrum is <sup>4</sup><sup>4</sup>4Note that this is the spectrum for the primordial perturbations. After the Universe becomes matter dominated at a redshift $`z10^4`$, density perturbations grow via gravitational infall, and the growth factor depends on the wave number. Therefore, the power spectrum for matter today is different from the primordial spectrum (it becomes $`k^4`$ times the primordial spectrum at large $`k`$), but it is straightforward to relate the primordial and current power spectra.
$$P_s=A_sk^{n_s},$$
(19)
and the primordial power spectrum for tensor perturbations is
$$P_t=A_tk^{n_t}.$$
(20)
The amplitudes $`A_t`$ and $`A_s`$ and power-law indices $`n_s`$ and $`n_t`$ have come to be known as the “inflationary observables.” These parameters can provide information on the inflaton potential. In the slow-roll approximation, the power-spectrum indices are roughly constant and given by
$$n_s=14ϵ+2\eta ,n_t=2ϵ,$$
(21)
where $`ϵ`$ and $`\eta `$ are the slow-roll parameters given in Equations 17 and 18. Strictly speaking, $`ϵ`$ and $`\eta `$ may change (logarithmically with $`k`$) during inflation , but, as the name implies, the field rolls slowly during slow-roll inflation, so the running of the spectral indices is, for all practical purposes, very small.
The amplitudes $`A_s`$ and $`A_t`$ are similarly fixed by the inflaton potential, but their precise values depend on Fourier conventions and on how the scale factor today is chosen. However, $`A_s`$ and $`A_t`$ are proportional, respectively, to the amplitude of the scalar and tensor contributions to $`C_2^{\mathrm{TT}}`$, the quadrupole moment of the CMB temperature. In terms of the slow-roll parameter $`ϵ`$ and height $`V`$ of the inflaton potential during inflation, these CMB observables are
$`𝒮`$ $``$ $`6C_2^{\mathrm{TT},\mathrm{scalar}}=0.66{\displaystyle \frac{V}{ϵm_{\mathrm{Pl}}^4}}`$
$`𝒯`$ $``$ $`6C_2^{\mathrm{TT},\mathrm{tensor}}=9.2{\displaystyle \frac{V}{m_{\mathrm{Pl}}^4}}.`$ (22)
For nearly scale-invariant spectra, COBE fixes $`C_2^{\mathrm{TT}}=C_2^{\mathrm{TT},\mathrm{scalar}}+C_2^{\mathrm{TT},\mathrm{tensor}}=(1.0\pm 0.1)\times 10^{10}`$. In terms of the slow-roll parameters, the tensor-to-scalar ratio is usually defined to be
$$r\frac{𝒯}{𝒮}=13.7ϵ.$$
(23)
Comparing Equation 23 with Equation 21, we observer that the observables $`n_t`$ and $`r`$ must satisfy a consistency relation, $`n_t=0.145r`$, in slow-roll models.
To summarize, slow-roll inflation models (which account for the overwhelming majority of inflation models that appear in the literature) are parameterized by (a) the height $`V`$ of the inflaton potential (i.e. the energy scale of inflation), (b) $`ϵ`$, which depends on the first derivative $`V^{}`$ of the inflaton potential, and (c) $`\eta `$, which depends additionally on the second derivative $`V^{\prime \prime }`$.
The discussion above suggests that because the inflaton is always rolling down the potential, the scalar spectral index must be $`n_s<1`$. Although this may be true for simple single-field inflation models, more complicated models (e.g. with multiple fields or with different potentials) may produce “blue” spectra with $`n_s>1`$ .
COBE alone already constrains $`V^{1/4}2.3\times 10^{16}`$ GeV. With some additional (but reasonable) modeling, the COBE constraint can be combined with current degree-scale CMB-anisotropy measurements and large-scale-structure observations to reduce this to $`V^{1/4}1.7\times 10^{16}`$ GeV (e.g. ). The COBE anisotropy implies $`n_s=1.1\pm 0.3`$ if it is attributed entirely to density perturbations , or $`n_t=0.2\pm 0.3`$ if it is attributed entirely to gravitational waves. Therefore, barring strange coincidences, the COBE spectral index and relations above seem to suggest that if slow-roll inflation is right, then the scalar and tensor spectra must both be nearly scale-invariant ($`n_s1`$ and $`n_t0`$).
### 3.4 Character of Primordial Perturbations
#### 3.4.1 Adiabatic versus isocurvature
The density perturbations produced by quantum fluctuations in the inflaton field are referred to as adiabatic, curvature, or isentropic perturbations. These are perturbations to the total density of the Universe, or equivalently, scalar perturbations to the spacetime metric. Adiabaticity further implies that the spatial distribution of each species in the Universe (e.g. baryons, photons, neutrinos, dark matter) is the same—that is, the ratio of number densities of any two of these species is everywhere the same.
Adiabatic perturbations can be contrasted with primordial isocurvature, or equivalently, pressure or entropy perturbations, perturbations to the ratios between the various species in the Universe (usually in a Universe with a homogeneous total density). Such varying ratios would set up perturbations to the pressure or equivalently to the entropy. When two initially causally-disconnected regions with different pressures come into causal contact, the pressure perturbations push matter around, thus seeding large-scale structure.
Axion Inflation
Although adiabatic perturbations are generically produced during inflation, it is also possible to obtain isocurvature perturbations. One example is isocurvature perturbations to an axion density from quantum fluctuations in the Peccei-Quinn field during inflation . As discussed below, comparison of the measured amplitude of CMB anisotropies with the amplitude of galaxy clustering essentially rules out these models. Inflation models that produce both adiabatic and isocurvature perturbations have also been considered ; future experiments should tightly constrain the relative contributions of the two types of perturbations.
#### 3.4.2 Causal versus acausal
Perturbations produced by inflation are said to be “super-horizon” or “acausal.” This simply refers to the fact that inflation produces a primordial (meaning before matter-radiation equality, when gravitational amplification of perturbations can begin) spectrum of perturbations of all wavelengths, including those much larger than the Hubble length at any given time. This is to be contrasted, for example, with “causal” models of structure formation, in which perturbations are generated by causal physics on scales smaller than the horizon. Since inflation implies distance scales much larger than the Hubble length can be within a causally connected pre-inflationary patch, the term acausal is really a misnomer.
#### 3.4.3 (Nearly) Gaussian distribution of perturbations
If the inflaton potential is flat enough for the slow-roll approximation to be valid, then each Fourier mode of the inflaton perturbation evolves independently; that is, the inflaton behaves essentially like an uncoupled massless scalar field. As a result, inflation predicts that the primordial density field is a realization of a Gaussian random field: each Fourier mode is decoupled from every other, and the probability distribution for each is Gaussian.
Of course, Gaussianity is an approximation that becomes increasingly valid in the slow-roll limit, in which the inflaton perturbation can be treated as a noninteracting scalar field. Deviations from Gaussianity are generally accepted to be small, and most theorists have adopted a pure Gaussian distribution as a prediction of inflation. However, the deviations in some models might be observable, and if so, would shed light on the physics responsible for inflation .
This can be quantified more precisely with the three-point statistic ,
$$I_l^3\frac{1}{(2l+1)^{3/2}(C_l^{\mathrm{TT}})^{3/2}}\left(\begin{array}{ccc}l& l& l\\ 0& 0& 0\end{array}\right)B(l,l,l).$$
(24)
In slow-roll models with smooth inflaton potentials, the prediction for this quantity is (L Wang, M Kamionkowski, manuscript in preparation)
$$\sqrt{l(l+1)}I_l^3=\frac{2}{m_{\mathrm{Pl}}^2}\sqrt{\frac{3V}{ϵ}}(3ϵ2\eta ).$$
(25)
Thus, in slow-roll models, one expects $`I_0\begin{array}{c}<\hfill \\ \hfill \end{array}10^6`$ (unless for some unforeseen reason $`ϵ`$ is extremely small and $`\eta `$ is not), too small to be observed. A larger non-Gaussian signal may conceivably arise if there is a glitch in the inflaton potential, but even this non-Gaussianity would be extremely small (L Wang, M Kamionkowski, manuscript in preparation). Detection of nonzero $`I_0`$ would thus rule out the simplest slow-roll models.
Note that the theory predicts that the primordial distribution of perturbations is Gaussian. When the Universe becomes matter dominated, and density perturbations undergo gravitational amplification, an initially Gaussian distribution will become non-Gaussian . Such departures from initial Gaussianity have a specific form and may be probed as consistency checks of inflation with galaxy surveys that probe the matter distribution today.
### 3.5 Brief Overview of Models
A huge literature is devoted to construction of inflationary models (for a comprehensive review, see ). Here we follow the classification of Dodelson et al . Models can be regarded as either large-field, small-field, or hybrid models. Linear models live at the border of large- and small-field models. In large-field (small-field) models, the inflaton moves a distance $`\mathrm{\Delta }\varphi `$ that is large (small) compared with the Planck mass during inflation. Hybrid models introduce a second scalar field and allow a broader range of phenomenology.
The models can be distinguished experimentally by the values of $`V`$, $`ϵ`$, and $`\eta `$ that they predict, or equivalently by the set of $`V`$, $`r`$, and $`n_s`$, as shown in Figure 2. Examples of large-field models are single-field models with polynomial potentials, $`V(\varphi )(\varphi /\mu )^p`$ (with $`p>1`$), or in the $`p\mathrm{}`$ limit, exponential potentials, $`V(\varphi )\mathrm{exp}(\varphi /\mu )`$.<sup>5</sup><sup>5</sup>5Exponential potentials are sometimes referred to as “power-law inflation,” since the scale factor grows as a power law during inflation in these models. The potentials in these models resemble qualitatively the potential shown in Figure 1a. These models have $`V^{\prime \prime }>0`$ and predict $`ϵ=[p/(p2)]\eta >0`$ and $`r7[p/(p+2)](1n_s)`$. Thus, a large tensor amplitude is expected for large $`p`$ (and therefore for exponential potentials as well) and for a sufficiently large deviation of $`n_s`$ from unity.
Figure 1b shows a potential typical of a small-field model. These are the types of potentials that often occur in spontaneous symmetry breaking and can be approximated by $`V(\varphi )[1(\varphi /\mu )^p]`$. These models have $`V^{\prime \prime }<0`$. Demanding that the field move a distance that is small compared with $`m_{\mathrm{Pl}}`$ requires that $`(\varphi /\mu )1`$, and in this limit, $`ϵ=[p/2(p1)]|\eta |(\varphi /\mu )^p\eta `$, $`\eta <0`$, and $`r7(1n_s)ϵ/|\eta |`$. Note that the slow-roll condition $`\varphi \mu `$ implies that $`ϵ1`$, so $`ϵ\eta `$. It thus follows that $`n_s1+2\eta `$, and that the tensor amplitude in these models is expected to be very small. Note that both small- and large-field models predict $`n_s<1`$.
Linear models live at the border of small- and large-field models. They have potentials $`V(\varphi )\varphi `$ (i.e. they have $`V^{\prime \prime }=0`$) and predict $`ϵ\eta >0`$ and $`r(7/3)(1n_s)`$.
Although hybrid models generally involve multiple scalar fields, they can be parameterized by a single-field model with a potential $`V[1+(\varphi /\mu )^p]`$. These models have $`ϵ>0`$ and
$$\frac{\eta }{ϵ}=\frac{2(p1)}{p}\left(\frac{\varphi }{\mu }\right)^p\left[1+\frac{p2}{2(p1)}\left(\frac{\varphi }{\mu }\right)^p\right]>0.$$
(26)
Unlike small- or large-field models, hybrid models can (although are not required to) produce blue spectra, $`n_s>1`$. Although both $`r`$ and $`n_t`$ depend only on $`ϵ`$ and are thus related, there is no general relation between $`r`$ and $`n_s`$ in hybrid models. The tensor amplitude is only constrained to be smaller than it is in exponential models.
## 4 Cosmic Microwave Background Tests of Inflation
Photons from overdense regions at the surface of last scatter are redder since they must climb out of deeper potential wells (the Sachs-Wolfe effect ). However, this is really only one of a number of physical mechanisms that give rise to temperature perturbations. We have also mentioned that if primordial perturbations are adiabatic, then the gas in deeper potential wells is hotter, and this partially offsets the reddening due to the depth of the potential. Density perturbations induce peculiar velocities, and these also produce temperature perturbations via Doppler shifts. Growth of the gravitational potential near the CMB surface of last scatter can produce temperature anisotropies \[the early-time integrated Sachs-Wolfe (ISW) effect\], and so can the growth of the gravitational potential at late times in a flat cosmological-constant or open Universe (the late-time ISW effect).
Modern calculations of the CMB power spectra (the $`C_l`$) take into account all of these effects. The cosmological perturbation theory underlying these calculations has been reviewed , and solution of the Boltzmann equations for the observed angular distribution of CMB photons is discussed elsewhere . Such calculations for the CMB power spectra from density perturbations were developed in a series of pioneering papers from 1970 until the late 1980s , and these calculations have been refined extensively in the post-COBE era. Similar calculations can also be carried out for the CMB power spectra from gravitational waves .
The calculations for both scalar and tensor power spectra require solution of a series of several thousand coupled differential equations for the perturbations to the spacetime metric, densities and velocities of baryons and cold dark matter, and the moments of the CMB photon and neutrino distributions. A code for carrying out these calculations (that required several hours for each model) was made publicly available . Hu & Sugiyama came up with useful semianalytic fits to the numerical calculations that provided some physical intuition into the numerical results. More recently, Seljak & Zaldarriaga developed a line-of-sight approach that speeded up the numerical calculations by several orders of magnitude. A code (CMBFAST) was made publicly available and has become widely used.
Given the values of the classical cosmological parameters (e.g. the nonrelativistic matter density $`\mathrm{\Omega }_0`$, cosmological constant $`\mathrm{\Omega }_\mathrm{\Lambda }`$, and baryon density $`\mathrm{\Omega }_b`$, all in units of the critical density, and the Hubble parameter $`h`$ in units of 100 km sec<sup>-1</sup> Mpc<sup>-1</sup>), and primordial scalar and tensor power spectra, $`P_s(k)`$ and $`P_t(k)`$, it is straightforward to calculate the $`C_l`$ with the machinery described above. Figure 3 shows results of such calculations for models with a Peebles-Harrison-Zeldovich (i.e. $`n_s=1`$) power spectrum of primordial adiabatic perturbations. Each panel shows the effect of independent variation of one of the cosmological parameters. As illustrated, the height, width, and spacing of the acoustic peaks in the angular spectrum depend on these (and other) cosmological parameters.
The wiggles<sup>6</sup><sup>6</sup>6These are sometimes referred to inaccurately as “Doppler” peaks, but are more accurately referred to as acoustic peaks. They are sometimes called Sakharov oscillations in honor of the scientist who first postulated the existence of photon-baryon oscillations in the primordial plasma . The existence of these peaks in the CMB power spectrum was, to the best of our knowledge, first identified by Sunyaev & Zeldovich and Peebles & Yu . come from oscillations in the photon-baryon fluid at the surface of last scatter. Consider an individual Fourier mode of an initial adiabatic density perturbation. Because the density perturbation is nonzero initially, this mode begins at its maximum amplitude. The amplitude remains fixed initially when the wavelength of the mode is larger than the Hubble radius. When the Universe has expanded enough that the Hubble radius becomes larger than the wavelength of this particular mode, then causal physics can act, and the amplitude of this Fourier mode begins to oscillate as a standing acoustic wave . Since modes with smaller wavelengths come within the horizon earlier and oscillate more rapidly, they have at any given time undergone more oscillations than longer-wavelength modes have. The peaks evident in Figure 3 arise because modes of different wavelength are at different points of their oscillation cycles . The first peak corresponds to the mode that has had just enough time to come within the horizon and compress once. The second peak corresponds to the mode that is at its maximum amplitude after the first compression, and so forth.
### 4.1 Determination of the Geometry
These acoustic peaks in the CMB temperature power spectrum can be used to determine the geometry of the Universe . The angle subtended by the horizon at the surface of last scatter is $`\theta _H\mathrm{\Omega }^{1/2}\mathrm{\hspace{0.33em}1}^{}`$, where $`\mathrm{\Omega }=\mathrm{\Omega }_0+\mathrm{\Omega }_\mathrm{\Lambda }`$ is the total density (objects appear to be larger in a closed Universe than they would in a flat Universe, and smaller in an open Universe than they would in an flat Universe). Moreover, the peaks in the CMB spectrum are due to causal processes at the surface of last scatter. Therefore, the angles (or values of $`l`$) at which the peaks occur determine the geometry of the Universe. This is illustrated in the top left panel of Figure 3, where the CMB spectra for several values of $`\mathrm{\Omega }`$ are shown. As illustrated in the other panels, the angular position of the first peak is relatively insensitive to the values of other undetermined (or still imprecisely determined) cosmological parameters such as the baryon density, the Hubble constant, and the cosmological constant.
Small changes to the spectral index $`n_s`$ tilt the entire spectrum slightly to smaller (larger) $`l`$ for $`n_s<1`$ ($`n_s>1`$), and the location of the first peak is only weakly affected. Gravitational waves would only add to the temperature power spectrum at $`l200`$ (as discussed below in Section 4.4). Therefore, although gravitational waves could affect the height of the peaks relative to the normalization at small $`l`$, the locations would not be affected. It is plausible that an early generation of star formation released a sufficient flux of ionizing radiation to at least partially reionize the Universe, and if so, these ionized electrons would rescatter some fraction $`\tau `$ of the CMB photons. A variety of theoretical arguments suggest that a fraction $`\tau =𝒪(0.1)`$ of CMB photons were rescattered (and the amplitude of anisotropy at degree scales observed already supports this). Although reionization would damp the peaks by a factor $`e^{2\tau }`$, as indicated by the curve labeled “reion” in the top left panel of Figure 3 (but note that the figure assumes $`\tau =1`$), the location of the peaks would remain unchanged. Therefore, if peak structure is observed in the CMB power spectrum, determination of the location of the first peak will provide a robust determination of the geometry of the Universe .
#### 4.1.1 Open inflation
The most distinctive signature of an open-inflation model would be a low-$`\mathrm{\Omega }_0`$ CMB power spectrum from adiabatic perturbations such as one of those shown in the top left panel of Figure 3, in which the first peak is shifted to larger $`l`$. Open inflation would also produce an increase in large-angle anisotropy from the integrated Sachs-Wolfe effect , but different open-inflation models make different predictions about the large-angle anisotropy. Moreover, determination of the CMB power spectrum at these large angular scales is cosmic-variance limited, so it is unlikely that large-angle CMB anisotropies alone will be able to provide a robust test of open-inflation models. The ISW effect may alternatively be identified by cross-correlation of the CMB with some tracer of the mass density along the line of sight , such as the X-ray background or possibly weak-lensing maps (as discussed further in Section 6.3).
### 4.2 Adiabatic Versus Isocurvature Modes
The physics described above yields a distinctive pattern in the peak structure of the CMB power spectrum, and this leads to an important test of inflation. If primordial perturbations are isocurvature rather than adiabatic, then when a given Fourier mode comes within the horizon and begins to oscillate, it begins to oscillate from its minimum (rather than maximum) amplitude. Thus, the phase of its oscillation differs by $`\pi /2`$ from what it would be if the perturbation were adiabatic. As a result, the locations of the peaks in the CMB power spectrum in isocurvature models differ in phase from what they would be in adiabatic models , as shown in Figure 4. It has also been shown that the relative locations of the higher peaks differ in adiabatic and isocurvature models, independent of the shift in the locations of the peaks due to the geometry .
#### 4.2.1 Axion inflation
When the matter power spectrum is normalized to the amplitude of galaxy clustering, isocurvature models with nearly scale-invariant primordial power spectra (e.g. axion isocurvature or “axion inflation” models) produce roughly six times the CMB anisotropy seen by COBE (since there is no cancellation between the effects of the intrinsic temperature and the potential-well depth at the surface of last scatter) and are thus ruled out.
### 4.3 Coherent Perturbations and Polarization
Each Fourier component of the density field induces a Fourier component of the peculiar-velocity field, and the oscillations of these peculiar velocities are out of phase with the oscillations in the density perturbation (just as the velocity and position of a harmonic oscillator are out of phase). These peculiar velocities induce temperature anisotropies (via the Doppler effect) that are thus out of phase with those from density perturbations. This Doppler effect fills in the troughs in the $`C_l^{\mathrm{TT}}`$, which would otherwise fall to zero.
The CMB polarization is related to the peculiar velocity at the surface of last scatter , so the peaks in the polarization power spectrum (from density perturbations) are out of phase from those in the temperature power spectrum and fall close to zero (Figure 5). This relative positioning of the temperature and polarization peaks is a signature of coherent perturbations (rather than those produced, for example, by the action of topological defects, as discussed below) . Zaldarriaga & Spergel argue that the location of the first peak in the polarization power spectrum provides a test of primordial perturbations as the origin of structure and thus of inflation. If some causal mechanism (such as topological defects) produced large-scale structure, the the first peak would have to occur at smaller angular scales in order to be within the horizon at the surface of last scatter (see below). A peak so close to the causal horizon could only occur with super-horizon-sized primordial perturbations, for which inflation is the only causal mechanism.
### 4.4 Polarization and Gravitational Waves
Gravitational waves are usually detected by observation of the motion they induce in test masses. The photon-baryon fluid at the surface of last scatter acts as a set of test masses for detection of gravitational waves with wavelengths comparable to the horizon, such as those predicted by inflation. These motions are imprinted onto the temperature and polarization of the CMB. The top left panel of Figure 5 (solid curve) shows the temperature power spectrum for a COBE-normalized flat scale-invariant ($`n_t=0`$) spectrum of gravitational waves. It is flat and relatively featureless for $`l\begin{array}{c}<\hfill \\ \hfill \end{array}70`$. The dropoff at $`l\begin{array}{c}>\hfill \\ \hfill \end{array}70`$ is due to the fact that the amplitudes of gravitational-wave modes that enter the horizon before the epoch of last scatter have decayed with the expansion of the Universe. Unfortunately, cosmic variance from scalar perturbations provides a fundamental limit to the sensitivity of CMB temperature maps to tensor perturbations . Even if all other cosmological parameters are somehow fixed, a perfect temperature map can never detect an inflaton-potential height smaller than one-tenth the upper limit provided by COBE . More realistically, the effects of gravitational waves and reionization on the temperature power spectrum are similar and difficult to disentangle, so improvements to the current COBE sensitivity to gravitational waves is unlikely with a temperature map alone.
However, with a polarization map of the CMB, the scalar and tensor contributions to CMB polarization can be geometrically decomposed in a model-independent fashion, and the cosmic-variance limit present in temperature maps can thereby be circumvented . Scalar perturbations have no handedness, so they cannot give rise to a curl component. On the other hand, tensor perturbations do have a handedness, so they induce a curl component. Therefore, if any curl coefficient, $`a_{(lm)}^\mathrm{C}`$, is found to be nonzero, it suggests the presence of gravitational waves.<sup>7</sup><sup>7</sup>7A curl component may also be due to vector perturbations. Although topological-defect models may excite such modes, they do not arise in inflationary models.
To illustrate, Figure 5 shows the four nonzero temperature-polarization power spectra. The dotted curves correspond to a COBE-normalized inflationary model with no gravitational waves. The solid curves show the spectra for a COBE-normalized stochastic gravitational-wave background.
#### 4.4.1 Detectability of gravitational waves: curl component only
Consider a mapping experiment that measures the polarization on the entire sky with a temperature sensitivity $`s`$ (which has units $`\mu `$$`\sqrt{\mathrm{sec}}`$) for a time $`t_{\mathrm{yr}}`$ years. If only the curl component of the polarization is used to detect tensor perturbations, then such an experiment can distinguish a tensor signal from a null result at the $`2\sigma `$ level if the inflaton potential height is
$$V\begin{array}{c}>\hfill \\ \hfill \end{array}(4\times 10^{15}\mathrm{GeV})^4t_{\mathrm{yr}}^1(s/\mu \mathrm{K}\sqrt{\mathrm{sec}})^2.$$
(27)
Equation 27 indicates that to access an inflaton-potential height not already excluded by COBE requires a detector sensitivity $`s\begin{array}{c}<\hfill \\ \hfill \end{array}35t_{\mathrm{yr}}^{1/2}\mu `$K$`\sqrt{\mathrm{sec}}`$. To compare this with realistic values, the effective sensitivity of MAP is $`s150t_{\mathrm{yr}}^{1/2}\mu `$K$`\sqrt{\mathrm{sec}}`$ and that for Planck is about $`s35t_{\mathrm{yr}}^{1/2}\mu `$K$`\sqrt{\mathrm{sec}}`$, and technological developments have improved the detector sensitivity roughly an order of magnitude per decade for the past several decades. Even better sensitivities may be possible with deep integration on a smaller region of the sky.
#### 4.4.2 Reionization
In some sense, Equation 27 is conservative because even a small amount of reionization will significantly increase the polarization signal at low $`l`$ (indicated by the dashed curve in the CC panel of Figure 5 ). If $`\tau =0.1`$, then the sensitivity to tensor modes is increased by roughly a factor of five .
#### 4.4.3 Full polarization and temperature spectra
Although searching only for the curl component provides a model-independent probe of tensor modes, a stochastic gravitational-wave background leads to specific predictions for all four nonzero temperature-polarization power spectra (Figure 5). Fitting an inflationary model to the entire set of temperature and polarization power spectra can improve tensor detectability, especially at poorer sensitivities, albeit with the introduction of some model dependence. For detector sensitivities $`s\begin{array}{c}>\hfill \\ \hfill \end{array}15t_{\mathrm{yr}}^{1/2}\mu `$K$`\sqrt{\mathrm{sec}}`$, the sensitivity to a tensor signal is improved by factors of a few or so , depending on the cosmological model, whereas for detector sensitivities $`s\begin{array}{c}<\hfill \\ \hfill \end{array}15t_{\mathrm{yr}}^{1/2}\mu `$K$`\sqrt{\mathrm{sec}}`$, the sensitivity is attributable almost entirely to the CC power spectrum and approaches the limit in Equation 27.
#### 4.4.4 Measurement of inflationary observables
Several authors have addressed the question of how precisely the inflationary observables can be reconstructed in the case a positive detection of the stochastic gravitational-wave background with only a temperature map and with a polarization map as well . We follow the discussion of Ref. .
Figure 6 shows the $`2\sigma `$ error ellipses that would be obtained by the Planck Surveyor using the temperature only (i.e. the cosmic-variance limit) and with the polarization, assuming a gravitational-wave background with $`r0.01`$ and $`n_s0.9`$. (A larger gravitational-wave amplitude would be detectable, as shown in Figures 3–6 in Reference .) The $`2\sigma `$ cosmic-variance limit from a temperature map is shown as is the $`2\sigma `$ constraint to the parameter space expected for Planck (with polarization). Although such a tensor amplitude cannot be distinguished from a null result, the figure shows (the dark shaded region) that a hypothetical experiment with three times the Planck polarization sensitivity could discriminate between a such a tensor signal and a null result. It would also discriminate between a single small-field model and a hybrid model. Of course, the sensitivity to tensor modes can be improved as the instrumental sensitivity is improved, as indicated by Equation 27. For example, the thin horizontal line at $`r=0.001`$ shows the smallest value of $`r`$ that could be distinguished from a null result by a hypothetical one-year experiment with an instrumental sensitivity $`s=\mu `$K$`\sqrt{\mathrm{sec}}`$, roughly 30 times that of Planck . A null result from such an experiment would suggest that if inflation occurred, it would have required a small-field model.
### 4.5 Gaussianity
The prediction of primordial Gaussianity or of some specific small deviations from Gaussianity can be probed with the CMB angular bispectrum or higher $`n`$-point correlation functions discussed above. A nonzero large-angle CMB bispectrum may arise from the integrated Sachs-Wolfe effect if there is a cosmological constant . Such a bispectrum, as well as that probed by the Sunyaev-Zeldovich effect, may be discernible via cross-correlation between gravitational-lensing maps and the CMB . A more powerful test of inflation models may arise from probing the bispectrum induced by nonlinear evolution at the surface of last scattering (S Winitzki, A Kosowsky, DN Spergel, manuscript in preparation).
Ferreira et al and Pando et al recently claim to have already found some signature of non-Gaussianity in the COBE maps. In particular, Ferreira et al find $`I_l^31`$ for $`l16`$ (although it is still not clear if the effect is real ). If this result is correct, then the simplest slow-roll inflation models are not viable (see Equation 25).
## 5 Topological-Defect Models
The leading alternative to structure-formation models based on inflation have been those based on topological defects, particularly cosmic strings , global monopoles , domain walls , and textures (for reviews, see ). Defect models postulate a phase transition in the early Universe that leads to a vacuum manifold with nontrivial topology; the type of defect depends on the specific topology (see for a review). Since defect formation is a process governed by causal physics, the vacuum state of the field must be uncorrelated on scales larger than the horizon at the time of the phase transition, guaranteeing the formation of defects with a characteristic length scale of the horizon (the “Kibble mechanism” ).
The simplest defects are domain walls, which arise in theories with a discrete symmetry. Domain-wall models are not viable because their energy densities are large enough to produce larger CMB temperature fluctuations than those observed .
Cosmic strings are stable defects that arise in gauge models with a $`U(1)`$ symmetry. They produce density perturbations by their gravitational interactions with ordinary matter. Global-monopole and texture models are unstable defects that arise in models with a perfect global symmetry. They provide two mechanisms for structure formation: (a) the energy-density provided by misalignment of scalar fields as causally disconnected regions of the Universe come into causal contact, and (b) the explosive events that occur when the topological defects unwind. Non-topological texture models postulate an even higher global symmetry and seed structure via scalar-field alignment even though no topological defects are formed. Generically, one expects quantum gravity to violate global symmetries to the level that would render global-monopole, texture, and scalar-field-alignment models unworkable . If it could be shown that such models do seed large-scale structure, valuable information on Planck-scale physics would thus be provided.
### 5.1 Cosmic-Microwave-Background Power Spectra in Defect Models
In contrast to inflationary models, which lay down an initial spectrum of density perturbations, defect models produce perturbations actively throughout the history of the Universe. This generally leads to loss of coherence in the perturbations and a corresponding smoothing of the acoustic peaks . Defect-model perturbations are also causal, being generated by physical processes inside the horizon . And finally, primordial perturbations in defect models more closely resemble those in primordial-isocurvature rather than those in adiabatic models . Moreover, the action of topological defects generically produces vector and tensor perturbations which increase the anisotropy on small angular scales , further suppressing any peak structure (although it may produce some characteristic C polarization ).
Until recently, different groups obtained different results about the extent to which acoustic peaks exist in defect models, and under what circumstances . Calculations of CMB power spectra based on large simulations of a variety of defect sources have now been performed (see Figure 7 for some). The numerical results indicate that the acoustic peaks are washed out.
At this point, it appears that the simplest defect models are inconsistent with the observed CMB fluctuations and the large-scale structure traced by galaxy surveys. Although this could have been inferred from the generic arguments discussed in the Introduction , it has been supported by these more recent precise calculations . The question now is whether any more complicated (or “sophisticated”) defect models be viable. Albrecht et al have suggested that a cosmological constant might help improve concordance with current data. However, suppose the CMB temperature power spectrum continues to look increasingly like that caused by inflation (i.e. with identifiable acoustic peaks), as new data seem to suggest. If so, can any defect model reproduce such a power spectrum? Turok produced a power spectrum with a phenomenological defect model that closely mimicked an inflation power spectrum, and Hu has invented a similar isocurvature model. But it is hard to see how to position the acoustic peaks in isocurvature-like models at the same angular scales as in adiabatic models without some rather artificial initial conditions . It is also difficult to simultaneously account for the fluctuation amplitude in the CMB and galaxy surveys, unless there is a breaking of scale invariance (possibly from some finite breaking of the global symmetry ). It may, in fact, be possible to construct some causal models that produce peaks in the CMB power spectrum , but it is unclear whether fluctuations that mimic a specific inflationary model can be produced, particularly when additional constraints from polarization are taken into account. Finally, it should be noted that hybrid models with both primordial adiabatic perturbations and defects have been entertained .
### 5.2 Non-Gaussianity
Topological-defect models may also be distinguished by the non-Gaussian signatures they produce in the CMB. Because the evolution of topological defects is nonlinear, they generically produce non-Gaussian structures in the CMB. Put another way, the production of defects via the Kibble mechanism is a Poisson process; the number of defects within any volume in the Universe is Poisson distributed. The central-limit theorem guarantees that as the number density of defects becomes large, the distribution should become increasingly Gaussian. Thus cosmic-string models should look more like Gaussian perturbations than textures should, since the Kibble mechanism produces roughly one texture per 25 Hubble volumes as opposed to roughly one cosmic string per Hubble volume . In the large-$`N`$ limit of the $`O(N)`$ sigma model, the clearest signature of non-Gaussianity from scalar-field alignment is at large angular scales ; on small distance scales, the theory looks roughly Gaussian. Constraints to non-Gaussianity from the galaxy distribution have already posed problems for scalar-field-alignment models for several years. Since defects are coherent structures, they can produce corresponding coherent structures in the CMB temperature anisotropy. For example, a cosmic string can produce a linear discontinuity in the CMB temperature , which can be searched for most efficiently through statistics tailored to match this particular signal . Textures might form large hot spots .
## 6 Dark Matter
The CMB can potentially provide a wealth of information about the dark matter known to dominate the mass of the Universe. The smallness of the amplitude of CMB temperature fluctuations has for a long time provided some of the strongest evidence for the existence of dark matter. In a low-density Universe, density perturbations grow when the Universe becomes matter-dominated and end when it becomes curvature-dominated. If the luminous matter ($`\mathrm{\Omega }_{\mathrm{lum}}10^3`$) were all the mass in the Universe, then the epoch of structure formation would be too short to allow density perturbations to grow from their early-Universe amplitude, fixed by COBE, to the amplitude observed today in galaxy surveys.
More precise measurements of the CMB power spectrum hold the promise of providing much more detailed information about the properties and distribution of dark matter. There are currently several very plausible dark-matter candidates that arise from new particle physics, and some evidence has already been claimed for the existence of several of these. For example, some observational evidence points to the existence of a cosmological constant , and the LSND experiment suggests that massive neutrinos may constitute a significant fraction of the mass of the Universe . Moreover, there are good arguments that a significant fraction of the mass in galactic halos is made of some type of cold-dark-matter particle, e.g. weakly interacting massive particles (WIMPs) or axions .
### 6.1 Cold Dark Matter
A number of dynamical measurements suggest that the nonrelativistic-matter density is $`\mathrm{\Omega }_0\begin{array}{c}>\hfill \\ \hfill \end{array}0.1`$, whereas big-bang nucleosynthesis suggests a baryon density of $`\mathrm{\Omega }_b\begin{array}{c}<\hfill \\ \hfill \end{array}0.1`$. Observations of X-ray emission from galaxy clusters suggest that the nonrelativistic matter in clusters outweighs the baryonic matter by a factor of three or more , and weak lensing of background galaxies by clusters directly reveals large amounts of dark matter . This evidence strongly indicates the existence of some nonbaryonic dark matter. By fitting the power spectra from MAP and Planck to theoretical predictions, one should simultaneously be able to determine both $`\mathrm{\Omega }_0h^2`$ and $`\mathrm{\Omega }_bh^2`$ to far better precision than that obtained by current observations . If a substantial fraction of the mass in the Universe is in fact made of nonbaryonic dark matter (e.g. WIMPS or axions), then it will become evident after MAP and Planck. Unfortunately, there is no way to discriminate between WIMPs and axions with the CMB.
### 6.2 Neutrinos
One of the primary goals of experimental particle physics is pursuit of a nonzero neutrino mass. Some recent (still controversial) experimental results suggest that one of the neutrinos may have a mass of $`𝒪(5\mathrm{eV})`$ . There have been some arguments (again, still controversial) that such a neutrino mass is exactly what is required to explain apparent discrepancies between large-scale-structure observations and the simplest inflation-inspired standard-CDM model .
If the neutrino does indeed have a mass of $`𝒪(5\mathrm{eV})`$, then roughly 30% of the mass in the Universe is in the form of light neutrinos. These neutrinos will affect the growth of gravitational-potential wells near the epoch of last scatter, thus leaving an imprint on the CMB angular power spectrum . The effect of a light neutrino on the power spectrum is small, so other cosmological parameters that might affect the shape of the power spectrum at larger $`l`$ must be known well. Eisenstein et al argue that by combining measurements of the CMB power spectrum with those of the mass power spectrum measured by, for instance, the Sloan Digital Sky Survey, a neutrino mass of $`𝒪(5\mathrm{eV})`$ can be identified. The CMB may constrain the number of noninteracting relativistic degrees of freedom in the early Universe . Although weaker than the bound from big-bang nucleosynthesis , the CMB probes a different epoch ($`T`$eV rather than $`T`$MeV) and may thus be viewed as complementary.
### 6.3 Cosmological Constant
Some recent evidence seems to point to the existence of an accelerating expansion, possibly due to a nonzero cosmological constant (; for a review of the cosmological constant, see ). The CMB may help probe the existence of a cosmological constant in a number of ways. As discussed above, if adiabatic perturbations are responsible for large-scale structure, then the position of the first acoustic peak in the CMB power spectrum provides a model-independent probe of the total density, $`\mathrm{\Omega }=\mathrm{\Omega }_0+\mathrm{\Omega }_\mathrm{\Lambda }`$ . In contrast, the supernova measurements of the Hubble diagram at large redshifts determine primarily the deceleration parameter $`q_0=\mathrm{\Omega }_0/2\mathrm{\Omega }_\mathrm{\Lambda }`$, so the two measurements together can give tight limits on both $`\mathrm{\Omega }_0`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }`$ individually .
As the bottom panels of Figure 3 show, variations to $`\mathrm{\Omega }_0`$ and $`h`$ affect the the height and width of the first acoustic peak; the dependence is more precisely on the quantity $`\mathrm{\Omega }_0h^2`$. Thus, if the Hubble constant is known, then the CMB can determine $`\mathrm{\Omega }_0`$ and $`\mathrm{\Omega }`$ (from the peak location) and therefore the cosmological constant $`\mathrm{\Omega }_\mathrm{\Lambda }`$.
A cosmological constant may also be distinguished from the CMB via the additional large-angle anisotropy it produces via the ISW effect from density perturbations at redshifts $`z\begin{array}{c}<\hfill \\ \hfill \end{array}`$few. If there is a cosmological constant, there should be a cross-correlation between the CMB temperature and some tracer of the mass distribution, e.g. the extragalactic X-ray background or weak lensing , at these redshifts (the same also occurs in an open Universe ). An experimental upper limit to the amplitude of this cross-correlation can already be used to constrain $`\mathrm{\Omega }_0`$, with some assumptions about the bias of sources that give rise to the extragalactic X-ray background. If $`\mathrm{\Omega }_00.3`$ (either in an open or a flat cosmological-constant model), then these X-ray sources can be no more than weakly biased tracers of the mass distribution .
### 6.4 Rolling Scalar Fields
The supernova evidence for an accelerating expansion has engendered a burst of theoretical activity on exotic forms of matter with an equation of state $`p<\rho /3`$ (i.e. the equation of state needed for $`q_0<0`$). The simplest possibility is of course a cosmological constant. However, as explained in Section 3.1, a rolling scalar field may also provide such an equation of state, provided the scalar field is not rolling too quickly. An almost endless variety of equations of state—and expansion histories—are possible in principle, given the freedom to choose the scalar-field potential and the initial conditions. This idea is variously referred to in the literature as rolling-scalar-field, variable-cosmological-constant, x-matter, generalized-dark-matter, loitering-Universe, and/or quintessence models . Additional work has explored attractor solutions based on exponential potentials or “tracker-field” solutions that attempt to explain why the matter density would be comparable to a scalar-field energy density today.
Because the expansion rate at decoupling in such models is the same as in cosmological-constant models with the same $`\mathrm{\Omega }_0`$, the peak structure in the CMB is virtually indistinguishable from that in cosmological-constant models . However, perturbations in the scalar field track perturbations to the matter density on large scales in such a way that the large-angle ISW effect that appears in cosmological-constant models is canceled by the effect of scalar-field perturbations . Data from cosmological observations, particularly supernova measurements of the expansion history and measurements of the power spectrum through large galaxy surveys, may in principle be used to break these degeneracies .
## 7 Other Constraints on Particle Physics
### 7.1 Decaying Particles
As discussed in Section 2.1, FIRAS limits to $`\mu `$ and $`y`$ distortions limit the injection of energy into the early Universe and can thus be used to constrain the mass-lifetime plane of particles that decay to electromagnetically interacting particles (as shown in Figure 8) . The CMB power spectrum can also constrain decaying particles. For example, a neutrino of mass $`\begin{array}{c}>\hfill \\ \hfill \end{array}10`$ eV that decays to relativistic particles with a lifetime $`\tau 10^{1317}`$ alters the expansion rate of the Universe between recombination and today and thus produces large-angle anisotropy (via the ISW effect) in disagreement with observations .
### 7.2 Time Variation of Fundamental Constants
A number of ideas for new physics postulate that some of the fundamental constants of nature, such as the fine-structure constant $`\alpha `$, may actually be varying (for a review, see ). Such a variation could be caused by the cosmological evolution of compact spatial dimensions in string theory or Kaluza-Klein theories or through scalar fields coupled to electromagnetism . Limits of $`|\mathrm{\Delta }\alpha /\alpha |\begin{array}{c}<\hfill \\ \hfill \end{array}10^7`$ were provided by the natural nuclear reactor at Oklo , and observations of atomic- and molecular-line positions at high redshifts provide limits of $`|\mathrm{\Delta }\alpha /\alpha |<3\times 10^6`$ at redshifts less than 1 and $`|\mathrm{\Delta }\alpha /\alpha |\begin{array}{c}<\hfill \\ \hfill \end{array}3\times 10^4`$ at redshifts of 3 . In fact, a detection of $`\mathrm{\Delta }\alpha /\alpha =1.9\pm 0.5\times 10^5`$ has been claimed on the basis of absorption lines at redshifts greater than 1 , but there are some potential problems with this result . Primordial nucleosynthesis can also provide a less useful model-dependent limit .
A change in $`\alpha `$ would affect the recombination rate of hydrogen and thus alter the redshift of last scatter. This effect on the CMB can potentially lead to upper limits on $`|\mathrm{\Delta }\alpha /\alpha |`$ between 0.01 and 0.001 out to redshifts $`z1100`$, much larger than those probed by quasar absorption spectra.
### 7.3 Topology of the Universe
The fundamental cosmological assumptions of homogeneity and isotropy require the Universe to be either the open, closed, or flat Friedmann-Robertson-Walker model. However, if the assumption of isotropy is incorrect, then the Universe may have some nontrivial topology (see for a review). The open and flat FRW models have infinite volume, but a Universe with either zero or negative curvature can have finite volume if the Universe has nontrivial topology. A number of (somewhat imprecise) theoretical arguments suggest that a finite Universe is easier to explain than an open Universe or could be used to explain the homogeneity of the Universe .
If the volume of such a Universe is comparable to or less than that observable today, then there may be signatures in the CMB. Consider the simplest nontrivial topology (for a flat Universe), that of a toroid. If the Universe is a three-dimensional toroid, then two different directions on the sky will point to the same point in space, and there should be observable correlations between the CMB temperature at distant locations on the sky. Such models have essentially been ruled out by COBE .
Interest in negative-curvature models with nontrivial topology has reawakened recently because evidence seems to suggest $`\mathrm{\Omega }_00.3<1`$, and thus possibly an open Universe. If the Universe is negatively curved (hyperbolic), the spacetime volume element increases rapidly with distance, so that even if the volume of the Universe is close to the horizon volume, many copies of the Universe may still fit inside the horizon volume. Thus, none of the flat-Universe limits on topology apply to hyperbolic Universes . Furthermore, if the total density of the Universe is $`\mathrm{\Omega }0.3`$, the curvature scale is small enough so that a huge number of topologies exist that have proper volumes significantly smaller than the proper Hubble volume .
Because the surface of last scatter is spherical, matched pairs of temperature circles would appear in a negatively-curved Universe with nontrivial topology provided that the topology radius were smaller than the current horizon . Levin et al propose searching for specific correlations between a given pixel and all others in a map. A null search for such correlations in the COBE maps ruled out a particular horn topology . Souradeep et al claim that the COBE maps already rule out most hyperbolic Universes through this technique, although details have not been presented.
### 7.4 Primordial Magnetic Fields
Magnetic fields of strength $`10^6`$ G are ubiquitous in our Galaxy and in distant clusters of galaxies. All mechanisms for the origin of these magnetic fields postulate that they grew via some mechanism (e.g. dynamo or adiabatic compression) from small primordial magnetic fields. However, the origin of these primordial seed fields remains a mystery. Many of the most intriguing hypotheses the origin of these fields come from new ideas in particle theory. Proposed generation mechanisms include inflation , the electroweak or QCD phase transitions , a ferromagnetic Yang-Mills vacuum state , charge asymmetry , and dilaton evolution .
Magnetic fields have several potentially measurable effects: Faraday rotation (A Mack, A Kosowsky, manuscript in preparation) and and associated depolarization of the original CMB polarization; magnetosonic waves that modify the acoustic oscillation frequencies ; and Alfven waves, which can amplify vector perturbations and induce additional correlations , and for which diffusion damping is decreased, thereby increasing CMB power at small scales . The Faraday rotation signals can be detected through the CC, TC, and GC power spectra they induce (although these power spectra are frequency dependent). A recent analysis of the COBE maps has placed a limit on a homogeneous primordial field strength corresponding to $`B_0<3.4\times 10^9(\mathrm{\Omega }_0h_{50}^2)^{1/2}`$ G by searching for the temperature pattern of a Bianchi type VII anisotropic spacetime .
### 7.5 Large-Scale Parity Violation
It is usually assumed that gravity is parity-invariant. However, weak interactions are parity-violating , and we surmise that the electroweak interactions are united with gravity at the Planck scale by some fundamental unified theory. Are there any remnants of parity-violating new physics in the early Universe? As discussed in Section 2.2.2, if either of the temperature-polarization cross-correlation moments $`C_l^{\mathrm{TC}}`$ or $`C_l^{\mathrm{TG}}`$ is nonzero, it signals cosmological parity breaking. Lue et al and Lepora discuss how a parity-violating term, $`\varphi F_{\mu \nu }\stackrel{~}{F}^{\mu \nu }`$ , that couples a scalar field $`\varphi `$ to the pseudoscalar $`\stackrel{}{E}\stackrel{}{B}`$ of electromagnetism could yield nonzero $`C_l^{\mathrm{TC}}`$ and $`C_l^{\mathrm{GC}}`$. Lue et al also discuss a parity-violating term in the Lagrangian for gravitation that would yield an asymmetry between the density of right- and left-handed gravitational waves produced during inflation; such an asymmetry would also give rise to nonzero $`C_l^{\mathrm{TC}}`$ and $`C_l^{\mathrm{GC}}`$. These parity-breaking effects would produce frequency-independent $`C_l^{\mathrm{TC}}`$ and $`C_l^{\mathrm{GC}}`$, unlike the frequency-dependent effect of Faraday rotation.
### 7.6 Baryon Asymmetry
There are very good theoretical and observational reasons to believe that our entire observable Universe is made of baryons and no antibaryons. But suppose momentarily that the observable Universe consisted of some domains with antibaryons rather than baryons (see e.g. ). If so, then particle-antiparticle annihilations at the interfaces of the matter and antimatter regions would release a significant amount of energy in $`\gamma `$-rays, thus heating the region and causing a $`y`$-distortion of the CMB spectrum of order $`y10^6`$ . These distortions would appear in thin strips on the sky which could potentially be identified. However, the point is moot because limits on the diffuse extragalactic $`\gamma `$-ray background limit the size of our matter domain to be essentially as large as the horizon .
### 7.7 Alternative Gravity Models
We now know through a variety of experiments that general relativity provides an accurate accounting of observed gravitational phenomena. On the other hand, string theories generically predict at least some small deviation from general relativity, often in the form of scalar-tensor theories of gravity . The simplest of these is Jordan-Fierz-Brans-Dicke (more commonly, Brans-Dicke) theory . An inflation theory (“extended inflation”) based on Brans-Dicke gravity was ruled out by the isotropy of the CMB , although models based on more complicated scalar-tensor theories (“hyperextended inflation”) have also been considered .
Brans-Dicke theory includes a scalar field $`\mathrm{\Phi }`$ and a new parameter $`\omega `$. As $`\omega \mathrm{}`$, the theory recovers general relativity (in some sense). Solar-system constraints from Viking spacecraft data limit $`\omega 500`$ (for a review, see ) and recent Very-Long Baseline Interferometry measurements of time delays of millisecond pulsars may further raise this limit . In cosmological models based on Brans-Dicke theories, general relativity is an attractor solution , so gravity could have conceivably differed from general relativity in the early Universe even if it resembles general relativity today.
Because the expansion rate and growth of gravitational-potential perturbations are different in alternative-gravity theories, the precise predictions for CMB power spectra should be different in these models. The epoch of matter-radiation equality is altered in Brans-Dicke theories, and this may produce an observable signal in forthcoming precise CMB maps . Cosmological perturbation theory in scalar-tensor theories has been worked out and the CMB power spectra calculated (X Chen, M Kamionkowski, manuscript in preparation). If the scalar-field time derivative $`\dot{\mathrm{\Phi }}`$ is fixed to be small enough to be consistent with big-bang-nucleosynthesis constraints and $`\omega >500`$, then the differences between the general-relativistic and Brans-Dicke predictions is small, although conceivably detectable with the Planck Surveyor. Of course, the scalar-field evolution may be significantly different in more sophisticated scalar-tensor theories, but predictions for these models have yet to be carried out.
### 7.8 Cosmic Rays
We close this tour of the CMB/particle intersection with possibly the oldest and most venerable connection between these two topics. Soon after the initial discovery of the CMB, it was realized that cosmic rays with energy $`E\begin{array}{c}>\hfill \\ \hfill \end{array}5\times 10^{19}`$ eV can scatter from CMB photons and produce pions. If a cosmic ray is produced with an energy above $`5\times 10^{19}`$ eV, repeated scatterings will reduce its energy to below this threshold within a distance of about 50 Mpc (the Greisen-Zatsepin-Kuzmin bound). These constraint have become increasingly intriguing recently, as several cosmic rays with energies $`>10^{20}`$ eV have been observed , and they do not appear to be coming from any identifiable astrophysical sources (e.g. radio galaxies or quasars ) as near as 50 Mpc . So where are these cosmic rays coming from? Some possibilities are exotic production mechanisms such as topological defects or supermassive unstable particles (; see for a review). If a recently claimed alignment of the highest-energy events with very distant radio quasars is confirmed by larger numbers of events, then it may be that these cosmic rays are exotic particles that interact with baryons but not photons , e.g. supersymmetric $`S_0`$ baryons . In the absence of any compelling traditional astrophysical origin, it seems that the simultaneous existence of the CMB and these cosmic rays may be pointing to some intriguing new particle physics.
## 8 Summary, Current Results, and Future Prospects
The primary cosmological observables pursued by CMB experiments are the frequency spectrum of the CMB, parameterized by $`\mu `$ and $`y`$ distortions, and the angular temperature and polarization power spectra, $`C_l^{\mathrm{TT}}`$, $`C_l^{\mathrm{GG}}`$, $`C_l^{\mathrm{CC}}`$, $`C_l^{\mathrm{TG}}`$, $`C_l^{\mathrm{TC}}`$, and $`C_l^{\mathrm{GC}}`$. There are additional observables, such as higher-order correlation functions or cross-correlation of the CMB temperature/polarization with other diffuse extragalactic backgrounds. Rough estimates of the $`C_l^{\mathrm{TT}}`$ at degree angular scales were obtained from the first generation of ground-based and balloon-borne CMB experiments. Forthcoming experiments will require far more sophisticated techniques for disentangling the CMB from foregrounds, and for recovering the power spectra from noisy data and from maps that cover only a fraction of the sky. A large literature is now devoted to these important issues, which we cannot review here.
Progress in CMB experiments is so rapid at the time of writing that any current data we might review would almost certainly become obsolete by the time of publication. We therefore refrain from showing any experimental results in detail and instead describe the current observations qualitatively. First, there is the isotropy of the CMB, which has long been explained only by inflation. Among the numerous pre-COBE models for the origin of large-scale structure, those based on a nearly scale-free spectrum of primordial adiabatic perturbations seem to account most easily for the amplitudes of both the large-angle CMB anisotropy measured by COBE and the amplitude of clustering in galaxy surveys. The galaxy distribution seems to be consistent with primordial Gaussianity.
Moreover, data from a large number of CMB experiments that probe the angular power spectrum at degree angular scales have now found (fairly convincingly) that there is significantly more power at degree angular separations ($`l200`$) than at COBE scales, as one would expect if the acoustic peaks do exist, but in apparent conflict with most theorists’ expectations for the degree-scale anisotropy in topological-defect models. The existence of this small-scale anisotropy further suggests no more than a small level of reionization (i.e. $`\tau 1`$). At the time of writing, the measurements are not precise enough to discern either the first or any higher peaks in the temperature power spectrum (some recent data are shown in Reference and are usually updated at Reference ). Some experiments have claimed to see the outline of a first acoustic peak at $`l200`$ (which would indicate a flat Universe). Moreover, some maximum-likelihood analyses of combined results of all experiments claim that the data indicate a flat Universe . However, these results are not yet robust.
Thus, although inflation is by no means yet in the clear, observations do seem to be pointing increasingly toward inflation. MAP and the Planck Surveyor will soon make far more precise tests of inflation (see for simulated data from MAP and Planck). First of all, the predictions of primordial adiabatic perturbations will be tested with unprecedented precision by the peak structure in the CMB temperature power spectrum. If the peaks do appear, then MAP and the Planck Surveyor should be able to measure the total density $`\mathrm{\Omega }`$ to a few percent or better by determining the location of the first acoustic peak . Moreover, by fitting MAP and Planck satellite data to theoretical curves, such as those shown in Figure 3, precise information on the values of other classical cosmological parameters can also be obtained . If nonrelativistic matter outweighs baryons, then it should become evident with MAP and Planck. The existence of a cosmological constant will further be tested, and some of the tests of gravity, decaying particles, etc, that we have reviewed will become possible.
If MAP and Planck confirm that the Universe is flat and that structure grew from primordial adiabatic perturbations, then the next step will be to search for the gravitational-wave background predicted by inflation. Such a gravitational-wave background could be isolated uniquely with the curl component of the polarization. If the inflaton-potential height is $`V^{1/4}10^{15}`$ GeV, then the gravitational-wave background will be unobservably small. However, if inflation had something to do with grand unification (i.e. $`V^{1/4}10^{1516}`$ GeV, as many theorists surmise), then the curl component of the polarization is conceivably detectable with the Planck Surveyor or with a realistic next-generation dedicated polarization satellite experiment. If detected, the curl component would provide a “smoking-gun” signature of inflation and indicate unambiguously that inflation occurred at $`T10^{1516}`$ GeV. Although an observable signature is by no means guaranteed, even if inflation did occur, the prospects for peering directly back to $`10^{40}`$ sec after the big bang are so tantalizing that a vigorous pursuit is certainly warranted.
Acknowledgments
We thank R Caldwell, A Liddle, and L Wang for very useful comments. MK was supported by a DOE Outstanding Junior Investigator Award, DE-FG02-92ER40699, NASA Astrophysics Theory Program grant NAG5-3091, and the Alfred P. Sloan Foundation. AK was supported by NASA Astrophysics Theory Program grant NAG5-7015 and acknowledges the kind hospitality of the Institute for Advanced Study.
|
no-problem/9904/astro-ph9904164.html
|
ar5iv
|
text
|
# Invariance violation extends the cosmic ray horizon ?
## 1 Introduction
Phenomena strikingly deviating from the conventional understanding can take place if the speed of photons is allowed to vary from the constant light velocity $`c`$. Coleman and Glashow (1997) argued that the velocity $`c_0`$ of light in the vacuum in a frame moving relative to the rest frame of the universe can differ from the maximal attainable speed $`c`$ of a material body. If $`c_0>c`$, the photon 4-momentum $`(\stackrel{}{E}/c_0,E)`$ is time-like, allowing a sufficiently energetic photon to decay into an electron-positron pair. Alternatively, if $`c_0<c`$, a charged particle will lose energy via vacuum Čerenkov radiation because the particle speed approaching $`c`$ at high energies can exceed the light speed $`c_0`$. Thus, the distance that energetic $`\gamma `$-rays or charged particles can travel through is shortened for either $`c_0>c`$ or $`c_0<c`$. However, the violation of relativity may also have a possibility of increasing the travelling distance of high energy radiation rather than decreasing.
The speed of photons can vary due to modifications of the vacuum (Latorre et al. 1995). Different energies excite the vacuum differently, and with quantum gravity, Amelino-Camelia et al. (1997) speculated that the energy-momentum relation for photons is modified to violate Lorentz invariance and the anomalous effect increases as photon energy approaches the Planck energy scale ($`10^{19}`$ GeV). Amelino-Camelia et al. (1998) has proposed a test for the modified speed by measuring the arrival time structure of photons from objects at cosmological distances. The propagation speed is $`dE/dk=c(1\xi _\gamma E/E_0)`$ from the modified relation of energy $`E`$ and momentum $`k`$,
$$c^2k^2=(E+\xi _\gamma \frac{E^2}{2E_0})^2=E^2+\xi _\gamma \frac{E^3}{E_0}$$
(1)
with $`\xi _\gamma =\pm 1`$ and $`E_o10^{19}`$ GeV, according to Amelino-Camelia et al. 1998. By assuming the existence of a preferred frame for Lorenz transformation but preserving other kinds of invariance, Coleman and Glashow 1998 gives a relation, which is similar to (1) and equivalent to $`c^2k^2=E^2m_i^2c^4+(c^2c_i^2)E^2`$ for $`ckEm_ic^2`$, where $`c_i`$ is the maximal attainable speed of particle species $`i`$ with mass $`m_i`$. Our study in what follows will be in an explicit form based on (1), but the results are, to some extent, applicable also to the other case by putting $`c^2c_i^2`$ in the place where the term $`\xi _iE/E_0`$ appears.
The electromagnetic radiation of the shortest wavelength so far detected is TeV $`\gamma `$-rays (see for example, Weekes et al. 1997; Ong 1998), but even for these high energy photons, the effect of velocity variation remains as small as a fraction $`10^{15}`$ of the light velocity. However, a higher degree of anomalous effect is expected for the absorption process, in which $`\gamma `$-rays are converted into an electron–positron pair when they collide with soft photons of the universal radiation field. Photons of TeV energies react with infrared photons, and PeV $`\gamma `$-rays with the 2.7K CMB (cosmic microwave background). From the number density of the photons of the universal infrared and CMB, the collision mean free path is usually estimated as about 100 Mpc and 10 kpc (e.g. Strong et al. 1974; Stecker et al. 1992), respectively for TeV and PeV $`\gamma `$-rays. The largest distance we can see with the high energy radiations is thus much shorter than those of X-rays and optical photons of longer wavelengths.
## 2 Kinematical constraints on the photon-photon collision process
The threshold energy is determined by a function of the total energy $`W`$ and momentum $`Q`$ of the initial state, $`W^2c^2Q^2`$, which must exceed $`(2mc^2)^2`$ for the pair creation, where $`m`$ is the electron mass. In the initial state of a $`\gamma `$-ray obeying the relation (1) and soft photons of energy $`\epsilon `$, the square of the energy $`E`$ and the momentum $`E/c`$ in $`W^2c^2Q^2`$ almost cancel each other to leave $`4\epsilon E\xi _\gamma E^3/E_0`$ (the two photons are assumed to be from opposite directions). The remaining two terms are, when $`\epsilon /E`$ is as small as $`\xi _\gamma E/E_0`$, of similar magnitude to each other, incurring a considerable degree of anomalous effect. In addition, the two terms become as large as $`m^2c^4`$ when $`E10^{12}\xi _\gamma ^{1/3}10^{12}\epsilon ^1`$ eV, and the threshold condition is considerably affected by the violation effect for $`E>10^{13}\xi _\gamma ^{1/3}`$ eV, which can be much lower than the Planck energy scale.
Let us examine the above speculation by calculating in our rest frame the conservation of momentum of the reaction. By putting $`c`$ = 1 hereafter, the momentum $`E\epsilon +\xi _\gamma E^2/(2E_0)`$ of the initial state (we assume for simplicity, but not to lose the generality, $`\mathrm{`}`$head-on’ collision of $`\gamma `$-ray and soft photon) must be equal to $`p_1cos\alpha +p_2cos\beta `$ of the final state, where ($`E_1`$, $`\stackrel{}{p}_1`$) and ($`E_2`$, $`\stackrel{}{p}_2`$) is energy and momentum of electron and positron and $`E+\epsilon =E_1+E_2`$. The $`\alpha `$ and $`\beta `$ are the angles $`\stackrel{}{p}_1`$ and $`\stackrel{}{p}_2`$ make against the direction of the initial gamma ray and are in the order of $`m/E_i`$ ($`i=1`$ and 2). By using the approximation $`p_i=E_im^2/(2E_i)`$ and the condition on the transverse momenta $`\alpha p_1=\beta p_2`$, we obtain
$$2\epsilon +\frac{\xi _\gamma E^2}{2E_0}=\frac{E+\epsilon }{2}(\frac{m^2}{E_1E_2}+\alpha \beta ).$$
(2)
We use a parameter $`x`$ to replace $`E_1`$ and $`E_2`$ by $`E_{1,\mathrm{\hspace{0.17em}2}}=(1\pm x)(E+\epsilon )/2`$ with the condition $`|x|12m/(E+\epsilon )`$. The equation (2) is then written as
$$\epsilon \frac{m^2}{E}\frac{1}{1x^2}+\frac{\alpha \beta E}{4}+\xi _\gamma \frac{E^2}{4E_0}.$$
(3)
The first and second terms are of the same order of magnitude, the latter term approaching zero at the threshold of the reaction. With increasing $`E`$, softer photons can serve as the target photons, shortening the absorption mean free path. This conventional view is altered as $`E`$ further increases, by the presence of the term $`\xi _\gamma E^2/(4E_0)`$. However, electron and positron may also obey a modified relation similar to (1). Let us put, similarly to (1), the square of the modified momentum $`q`$ to be equal to
$$q^2=p^2+\xi _e\frac{E^3}{E_0}=E^2m^2+\xi _e\frac{E^3}{E_0}.$$
(4)
We grant $`|\xi _e|1`$ for allowing electrons to have a different degree of the invariance violation. By replacing $`p`$ by $`q`$, a relation similar to (3),
$$\epsilon =\frac{m^2}{E}\frac{1}{1x^2}+\frac{\alpha \beta E}{4}+\frac{E^2}{4E_0}(\xi _\gamma \xi _e\frac{1+x^2}{2}),$$
(5)
is obtained. Since energy $`E`$ is divided almost equally into two particles (when $`x0`$) and the anomalous term has $`E^2`$ dependence, the factor 2 appears in the denominator of $`\xi _e(1+x^2)/2`$, thus reducing the effect in the final state. The fact might seem to suggest that the single $`\mathrm{`}`$leading’ particle taking almost all the total energy ($`|x|1`$) is capable of eliminating the anomalous effect. However, the substitution of $`|x|=12m/(E+\epsilon )`$ into (5) increases the energy $`\epsilon `$ to $`m1`$ MeV which is much higher than $`m^2/E`$, and then leads to a very large travelling distance, resulting in no less anomalous consequence. Only if $`\xi _e`$ is tuned to be $`2\xi _\gamma `$ for unknown reason, we can recover the ordinary relation for $`\epsilon `$.
## 3 Comparison with observations
Energy $`\epsilon `$ of the target photon is plotted against $`\gamma `$-ray energy $`E`$ in Fig. 1. A minimum value of $`\epsilon `$ is taken, if $`\xi =2\xi _\gamma \xi _e>0`$, at the critical energy $`E_c=(4m^2E_0/\xi )^{1/3}10^{13}`$ eV for $`\xi =1`$. The minimum value is about 0.03 eV and $`\epsilon `$ must increase to about 30 eV at $`E=10^{15}`$ eV. Thus, CMB of $`10^3`$ eV is never given a chance of contributing to the pair creation process. The travelling length has the shortest value of $``$10 Mpc at $`E=E_c`$, and increases again for $`E>E_c`$. Since $`30`$eV photons are very sparse, $`\gamma `$-rays would travel through distances $`>`$ 1 Gpc. When $`\xi <0`$, $`\epsilon `$ must rapidly decrease as $`E`$ approaches $`E_c`$. Before taking negative value, CMB photons become available as target. The energetic $`\gamma `$-rays are then efficiently absorbed with a mean free path of $`10`$ kpc, which is shorter by orders of magnitude than the ordinary case.
Among the TeV $`\gamma `$-ray sources are two active galaxies, Mrk 421 and Mrk 501 (Punch et al. 1992; Aharoninan et al. 1997; Catanese et al. 1997). The objects are at distances of $``$ 100 Mpc, and the detection is consistent with the absorption mean free path determined by infrared intensity in the extra-galactic space. The energy spectrum of the $`\gamma `$-rays is found to extend up to 10 - 20 TeV. If the spectrum continues to extend to even higher energies, $`\xi <0`$ would be excluded. Because, if $`\xi <0`$, a very sharp cut-off at the common energy $`E_c`$ is expected in the energy spectrum for any extra-galactic objects. The negative $`\epsilon `$ for $`E>E_c`$ in the case of $`\xi <0`$ implies that $`\gamma `$-rays does not need target photons for pair creation. The threshold of $`\gamma `$-ray decaying into an electron-positron pair (Coleman and Glashow 1997) is given by $`|E_c|`$. By putting $`|E_c|=E_c=(4m^2E_0/(\xi ))^{1/3}>20`$ TeV, we obtain a limit on $`\xi `$ to be $`\xi >1.3`$ (by putting $`E_0=1\times 10^{28}`$ eV). On the other hand, when $`\xi `$ is positive and $`1`$, the travelling distance turns over to increase above about 10 TeV and to become larger than 1 Gpc for the photons around 100 TeV. Thus, 100 TeV photons would arrive from extra-galactic sources, implying that evidence of detection would be strongly in favour of the $`\xi >0`$ case. Extra-galactic diffuse $`\gamma `$-rays are most likely from distant unresolved active galaxies. The energy spectrum is observed by Sreekumar et al. 1997 to extend to about 50 GeV with a hard power index of $`2.3`$. The spectrum is expected to become steeper at higher energies due to the intenser infrared photons and then CMB at longer wavelengths. However, in the case of $`\xi >0`$, the energy spectrum may have a turn-over of showing a harder slope above $`E=E_c`$. Electrons lose their energy quickly and have difficulty in producing such high energy $`\gamma `$-rays. However, cosmic ray protons are known to have energy spectrum extending to $`10^{20}`$ eV, and are able to produce $`\gamma `$-rays as high as or larger than 100 TeV. The increase of travelling distance suggests more contribution from more distant sources and thus we may expect considerably intense, diffuse $`\gamma `$-rays in the higher energy region. From the Crab nebula, 50 TeV photons have been detected (Tanimori et al. 1997) by using imaging Čerenkov telescope at large zenith angles. Detection of 100 TeV $`\gamma `$-rays is not unrealistic by the extensive application of such methods.
## 4 Case of cosmic ray protons
Our phenomenological consideration can be extended to other processes of soft photons acting as the reaction target. Extremely high energy cosmic ray protons (e.g. Hill and Schramm 1985; Axford 1994; O’Halloran et al. 1998; Takeda et al. 1998) are believed to lose energy by pion production in collision with CMB. The number density of CMB and the photo-pion cross section determine the collision mean free path to be $`10`$ Mpc. The energy spectrum thus suffers from the GZK cutoff (Greisen 1966; Zatsepin and Kuzmin 1966) at about $`10^{20}`$ eV, if the protons are from the distances farther than the mean free path. However, the relation (4) assumed for proton and pion alters the threshold of the energy $`\epsilon `$ of target photons from $`\mu M/(2E)`$ to
$$\epsilon =\frac{\mu M}{2E}+\mathrm{\Xi }\frac{E^2}{4E_0},\mathrm{where}\mathrm{\Xi }=\xi _p\xi _p(\frac{M}{M+\mu })^2\xi _\pi (\frac{\mu }{M+\mu })^2.$$
(6)
The energy $`E`$ is for the initial proton, $`M`$ and $`\mu `$ are the proton and pion mass, and $`\xi _p`$ and $`\xi _\pi `$ for protons and pions, respectively. We have put the parameter $`x`$ to be $`(M\mu )/(M+\mu )`$, corresponding to the case in which the ratio of proton to pion energy in the final state is $`(1+x)/(1x)=M/\mu `$ near the threshold of the process. If $`\mathrm{\Xi }`$ is positive and not very far from the order of one, the second term becomes comparable with $`\mu M/(2E)`$ at the critical energy $`E_{p\pi }=(M\mu E_0/\mathrm{\Xi })^{1/3}10^{15}`$ eV. At $`E10^{20}`$ eV, the energy $`\epsilon `$ needs to be, as dominated by $`\mathrm{\Xi }E^2/(4E_0)`$, as large as $`10^{12}`$ eV, and CMB can not play the role of target photons. Cosmic ray protons at $`10^{20}`$ eV can then travel without the attenuation from cosmological distances farther than $`10`$ Mpc. As a result, the energy spectrum of cosmic rays would extend beyond $`10^{20}`$ eV, provided that the acceleration to such high energies takes place in celestial objects. There is speculation by Vietri 1995 and Waxman 1995 that GRB ($`\gamma `$-ray bursts) provides the origin of the highest energy cosmic rays. If so, GRBs at cosmological distances would produce a cosmic ray spectrum without GZK cutoff. If $`\mathrm{\Xi }<0`$, $`\epsilon `$ can be $`10^3`$ eV when $`E`$ approaches $`|E_{p\pi }|=E_{p\pi }10^{15}`$ eV, implying that the target photons of CMB gives a characteristic attenuation length of $``$10 Mpc to $`10^{15}`$ eV cosmic rays. For $`E>|E_{p\pi }|`$, the kinematical condition does not require the target photon for the pion production to occur and the processes of $`pp+\pi ^0`$ and $`pn+\pi ^+`$ will cause protons to rapidly lose energy. The energy loss due to the spontaneous production of pion is in contradiction with the evidence of cosmic ray protons up to $`10^{20}`$ eV. Thus, we limit $`|E_{p\pi }|`$ longer than $`10^{20}`$ eV, and obtain a bound of $`\mathrm{\Xi }>10^{17}`$.
## 5 Discussions
Collision processes of high energy radiations with soft photons enable us to test the consequences of violation of Lorentz invariance. In particular, the effect of quantum gravity which Amelino-Camelia et al. (1998) suggest can be made detectable by the reaction process in the energy region much lower than the Planck energy scale, but in the sufficiently high energy region to which the validity of Lorentz transformation has not been examined yet by accelerator experiments. The violation suggests an interesting possibility of increasing the distance we can reach by observing high energy $`\gamma `$-rays and cosmic rays. Absence of the GZK cutoff would not necessarily mean nearby origin of the extremely high energy cosmic rays. Photons of energies larger than 100 TeV from extra-galactic objects would provide a clear test because it is hard to explain by other mechanisms than the violation of invariance. It seems impossible, once we admit the relation (1) for photons, to totally eliminate the anomalous effect, without a specially designed tuning among the invariance violating terms for various particles of different masses. Other ways have been proposed to alter the conventional view on the relativity (Gonzares-Mestres 1997; Sato 1998). Absence of GZK cutoff can be explained, according to Sato 1998, without abandoning the Lorenz invariance, but by giving a special status to $`\mathrm{`}\mathrm{`}`$the universe frame” in which CMB is isotropic.
Protons are expected to lose energy via vacuum Čerenkov radiation (Coleman and Glashow 1997), when $`\xi _\gamma \xi _p>0`$. The threshold energy, $`E_{th}=(M^2E_0/(\xi _\gamma \xi _p))^{1/3}`$, of the radiation is calculated from equating the speed of photon to that of proton. The detection of the highest energy $`10^{20}`$ eV of cosmic rays sets a constraint $`E_{th}>10^{20}`$ eV and limits the $`\xi `$-parameters to be $`\xi _\gamma \xi _p<10^{15}`$ or rather $`\xi _\gamma \xi _p`$ for practical use. The Čerenkov radiation from electron yields a less tight constraint because of the dependence of $`\xi `$ on $`E_{th}`$, i.e. $`\xi =\xi _\gamma \xi _eE_{th}^3`$. The highest energy $`2\times 10^{12}`$ eV of cosmic ray electrons ever detected (Nishiumura et al. 1997) must be less than $`E_{th}`$ and then we obtain $`\xi _\gamma \xi _e<10^3`$. If we choose, for the bound on $`E_{th}`$, $`100`$ TeV electrons which are responsible for TeV $`\gamma `$-rays and which are considered to extend over a spatial scale of at least $``$ 1 pc, the constraint is made somewhat tighter as $`\xi _\gamma \xi _e<10^2`$. We may argue about allowed region of $`(\xi _\gamma ,\xi _e)`$ values by combining the condition $`1.3<2\xi _\gamma \xi _e`$ from the decay process $`\gamma e^++e^{}`$. The region is, to some extent, in disfavor with negative values. The time structure of the outbursts of TeV $`\gamma `$-rays from Mrk 421 and Mrk 501 sets constraint on $`\xi _\gamma `$. The delay time of 10 TeV $`\gamma `$-rays relative to photons of lower energies is $`(d/c)\xi _\gamma (10^{13}\mathrm{eV}/E_0)10\xi _\gamma s`$ for distance $`d=100`$ Mpc. Duration of the outbursts which has been observed typically as a few hours is considered to be longer than the delay time, providing a constraint of $`\xi _\gamma <10^3`$. We need to know the outburst structure in time scales as short as $`10s`$ to constrain $`|\xi |`$ less than $`1`$. In order to restrict the $`|\xi |`$ parameter more tightly, It is very useful to observe photons of higher energies from more distant objects. However, we can not expect that such $`\gamma `$-rays survive against conversion into an electron-positron pair and reach us, unless we premise that the Lorenz invariance is violated.
We may argue that the extensive air shower events of $`10^{20}`$ eV are in truth not due to proton. Such a claim is relevant to $`\mathrm{`}`$correct’ understanding of the complex phenomena of the cascade shower process, which still leaves a room for a speculation of $`\mathrm{`}`$anomalous interaction’ of cosmic rays. Thus, it is interesting to indicate that the modified relation of energy and momentum can affect detection of the high energy radiations. Observation of $`\gamma `$-rays for example is based on the pair creation process in the detector material (satellite experiments) or in the atmosphere (ground-based observation for TeV photons and cosmic rays). The recoil momentum $`\delta `$ of atomic nucleus in the material is calculated as twice the right hand side of the expression (5). The ordinary recoil momentum is $`\delta _0m^2E/(2E_1E_2)=0.1`$ eV/c at $`E10^{13}`$ eV, which corresponds to quite a small energy transfer to the recoiled atomic nucleus of a large mass. When $`\xi _\gamma =1`$ and $`E>E_c`$, $`\delta `$ must be larger than $`\delta _0`$ but remains reasonably small so that the process would not be largely affected. When $`\xi _\gamma =1`$ and $`E>E_c`$, $`\delta `$ appears negative, the recoil being in the opposite direction to the initial $`\gamma `$-ray. There are generally several processes which compete the pair creation and for which the invariance violation is of different strengths. The consequence may appear as observation of an anomalous behaviour of very high energy radiations.
The author thanks the anonymous referee for the comments and advice that have brought the revised form and content of the present paper.
|
no-problem/9904/hep-ph9904466.html
|
ar5iv
|
text
|
# TRANSVERSE MOMENTUM DEPENDENCE OF CUMULATIVE PIONS This work is supported by the Russian Foundation for Fundamental Research, Grant No. 97-02-18123
## 1 Introduction
A few years ago we have proposed a quark-parton model of cumulative phenomena in the interactions with nuclei based on perturbative QCD calculations of the corresponding quark diagrams near the thresholds, at which other quarks (”donors”) in the nuclear flucton transfer all their longitudinal momenta to the distinguished active quark and become soft.
Consider the scattering of a hadronic projectile off a nucleus with the momentum $`P`$ in the c.m. system. At high energies the momentum $`K`$ of the produced pion belongs to the cumulative region if $`K_z>P_z/A`$. As a reasonable first approximation, we treat the nucleus as a collection of $`N=3A`$ valence quarks, which, on the average, carry each longitudinal momentum $`x_0P_z/A`$ with $`x_0=1/3`$. In our approach the cumulative pion production proceeds in two steps. First a valence quark with a scaling variable $`x>1`$ is created. Afterwards it decays into the observed hadron with its scaling variable $`x`$ close to the initial cumulative quark’s one. This second step is described by the well-known quark fragmentation functions and will not be discussed here.
The produced cumulative (”active”) quark acquires the momentum much greater than $`x_0P_z/A`$ only if this quark has interacted by means of gluon exchanges with other $`p`$ quarks of flucton (”donors”) and has taken some of their longitudinal momenta (see Fig.1). If this active quark accumulates all longitudinal momentum of these $`p`$ quarks then $`K_z=(p+1)x_0P_z/A`$ and the donors become soft. It is well-known that interactions which make the longitudinal momentum of one of the quark equal to zero may be treated by perturbation theory . This allows to calculate the part of Fig. 1 responsible for the creation of a cumulative quark explicitly. This was done in , to which papers we refer the reader for all the details. As a result we were able to explain the exponential fall-off of the production rate in the cumulative region.
Since with the rise of $`x`$ the active quark has to interact with a greater number of donors, one expects that its average transverse momentum also grows with $`x`$. Roughly one expects that $`K_{}^2`$ is proportional to the number of interactions, that is, to $`x`$. In this point was not studied: we have limited ourselves with the inclusive cross-section integrated over the transverse momenta, which lead to some simplifications. The aim of the present paper is to find the pion production rate dependence on the transverse momentum and the mean value of the latter as a function of $`x`$ in the cumulative region. This dependence and also the magnitude of $`K_{}^2`$ have been studied experimentally. The comparison of our predictions with the data allows to obtain further support for our model and fix one of the two its parameters (the infrared cutoff).
## 2 The $`K_{}`$ dependence
Repeating the calculations of the diagram in Fig.1 described in but not limiting ourselves with the inclusive cross-section integrated over the transverse momentum, we readily find that all dependence upon the transverse momentum $`K_{}`$ of the produced particle is concentrated in a factor:
$$J(K_{})=\rho _A(\underset{p+1}{\underset{}{r,\mathrm{},r}}|\underset{p+1}{\underset{}{\overline{r},\mathrm{},\overline{r}}})G(c_1,\mathrm{},c_p)\underset{i=1}{\overset{p}{}}\lambda (c_ir)\lambda (c_i\overline{r})d^2c_ie^{i(\overline{r}r)K_{}}d^2rd^2\overline{r}$$
(1)
Here $`\rho _A`$ is the (translationally invariant) quark density matrix of the nucleus:
$$\rho _A(r_i|\overline{r}_i)\psi _A(r_i,r_m)\psi _A^{}(\overline{r}_i,r_m)\underset{m=p+2}{\overset{N}{}}d^2r_m$$
(2)
where $`\psi _A`$ is the transverse part of the nuclear quark wave function. The propgation of soft donor quarks is decribed by
$$\lambda (c)=\frac{K_0(m|c|)}{2\pi }$$
(3)
where $`m`$ is the constituent quark mass and $`K_0`$ is the modified Bessel function (the Mac-Donald function). The interaction with the projectile contributes a factor
$$G(c_1,\mathrm{},c_p)=\underset{i=1}{\overset{p}{}}\sigma _{qq}(c_ib_i)\eta _H(b_1,\mathrm{},b_p)d^2b_i$$
(4)
where $`\sigma _{qq}(c)`$ is the quark-quark cross-section at a given value of impact parameter $`c`$ and
$$\eta _H(b_1,\mathrm{},b_p)=\underset{Lp}{}\frac{L!}{(Lp)!}|\psi _H(b_i)|^2\delta ^{(2)}(\frac{1}{L}\underset{i=1}{\overset{L}{}}b_i)d^2b_{p+1}\mathrm{}d^2b_L$$
(5)
is a multiparton distribution in the projectile, expressed via the transverse part of its partonic wave function $`\psi _H`$ . If we integrate $`J(K_{})`$ over $`K_{}`$ we come back to our old result (Eq. (33) in ):
$$J(K_{})\frac{d^2K_{}}{(2\pi )^2}=\rho _A(\underset{p+1}{\underset{}{0,\mathrm{},0}}|\underset{p+1}{\underset{}{0,\mathrm{},0}})G(c_1,\mathrm{},c_p)\underset{i=1}{\overset{p}{}}\lambda ^2(c_ir)d^2c_id^2r$$
If one assumes factorization of the multiparton distribution $`\eta _H(b_1,\mathrm{},b_p)`$ then
$`G(c_1,\mathrm{},c_p)`$ also factorizes:
$$G(c_1,\mathrm{},c_p)=\underset{i=1}{\overset{p}{}}G_0(c_i)$$
(6)
Following we use the quasi-eikonal approximation for $`\eta _H`$:
$$\eta _H(b_1,\mathrm{},b_p)=\xi ^{(p1)/2}\nu _H^p\underset{i=1}{\overset{p}{}}\eta _H(b_i)$$
where $`\xi `$ is the quasi-eikonal diffraction factor, $`\nu _H`$ is the mean number of partons in the projectile hadron and the single parton distribution $`\eta _H(b)`$ is normalized to unity. In a Gaussian approximation for $`\sigma (c)`$ and $`\eta _H(b)`$ we find:
$$G_0(c)=\xi ^{\frac{1}{2}\frac{1}{2p}}\frac{\nu _H\sigma _{qq}}{\pi r_{0H}^2}e^{\frac{c^2}{r_{0H}^2}}$$
where $`\sigma _{qq}`$ is the total quark-quark cross-section, $`r_{0H}^2=r_0^2+r_H^2`$, $`r_0`$ and $`r_H`$ are the widths of $`\sigma (c)`$ and $`\eta _H(b)`$ respectively.
With the factorised $`G(c_1,\mathrm{},c_p)`$ (6) we have
$$J(K_{})=\rho _A(\underset{p+1}{\underset{}{0,\mathrm{},0}}|\underset{p+1}{\underset{}{\overline{r}r,\mathrm{},\overline{r}r}})j^p(r,\overline{r})e^{i(\overline{r}r)K_{}}d^2rd^2\overline{r}$$
where
$$j(r,\overline{r})=d^2cG_0(c)\lambda (cr)\lambda (c\overline{r})$$
We also have used the translational invariance of the $`\rho `$-matrix. Note that near the real threshold we have no spectators and
$$\rho _A(\underset{p+1}{\underset{}{0,\mathrm{},0}}|\underset{p+1}{\underset{}{\overline{r}r,\mathrm{},\overline{r}r}})=\rho _A(\underset{p+1}{\underset{}{0,\mathrm{},0}}|\underset{p+1}{\underset{}{0,\mathrm{},0}})$$
In any case large $`K_{}`$ corresponds to small $`\overline{r}r`$ so we factor $`\rho _A`$ out of the integral in zero point. In the rest integral we pass to the variables
$$B=\frac{r+\overline{r}}{2},b=\overline{r}r$$
and shift the integration variable $`c`$, then
$$J(K_{})=\rho _A(\underset{p+1}{\underset{}{0,\mathrm{},0}}|\underset{p+1}{\underset{}{0,\mathrm{},0}})j^p(B,b)e^{ibK_{}}d^2bd^2B$$
(7)
where
$$j(B,b)=G_0(B+c)\lambda (\frac{b}{2}c)\lambda (\frac{b}{2}+c)d^2c$$
(8)
## 3 The calculation of $`|K_{}|`$
Now we would like to find the width of the distribution on $`K_{}`$ as a function of $`p`$ or what is the same of the cumulative number $`x=(p+1)/3`$. From the mathematical point of view it is simpler to calculate the mean squared width of the distribution $`K_{}^2`$. Unfortunately in our case this quantity is logarithmically divergent at large $`K_{}`$. This divergency results from the behavior of $`j(B,b)`$ at small $`b`$. This behavior is determined by the behavior of the $`\lambda (b)=K_0(m|b|)/(2\pi )`$ (3), which has a logarithmical singularity at $`|b|=0`$. Smooth $`G_0(B+c)`$ does not affect this behavior.
For this reason we shall rather calculate $`|K_{}|`$:
$$|K_{}|=\frac{1}{J_N}j^p(B,b)|K_{}|e^{ibK_{}}d^2bd^2B\frac{d^2K_{}}{(2\pi )^2}$$
(9)
where $`J_N`$ is the same integral as in the numerator but without $`|K_{}|`$. Presenting $`|K_{}|`$ as $`K_{}^2/|K_{}|`$ and $`K_{}^2`$ as the Laplacian $`\mathrm{\Delta }_b`$ applied to the exponent we find
$$|K_{}|=\frac{1}{J_N}j^p(B,b)\mathrm{\Delta }_be^{ibK_{}}d^2bd^2B\frac{d^2K_{}}{|K_{}|(2\pi )^2}$$
Twice integrating by parts and using the formula
$$\frac{d^2K_{}}{|K_{}|}e^{ibK_{}}=\frac{2\pi }{|b|}$$
we find
$$|K_{}|=\frac{1}{2\pi J_N}\frac{1}{|b|}\mathrm{\Delta }_bj^p(B,b)d^2bd^2B$$
Now we again integrate by parts once to find
$$|K_{}|=\frac{1}{2\pi J_N}d^2B\frac{d^2b}{|b|^2}(n_b_b)j^p(B,b)$$
where $`n_b=b/|b|`$. This leads to our final formula
$$|K_{}|=\frac{p}{2\pi J_N}d^2B\frac{d^2b}{|b|^2}j^{p1}(B,b)(n_b_b)j(B,b)$$
(10)
where $`j(B,b)`$ is given by (8), $`\lambda (b)`$ is given by (3) and
$$J_N=d^2Bj^p(B,b=0)$$
## 4 Approximations
To simplify numerical calculations we make some additional approximations, which are not very essential but are well supported by the comparison with exact calulations at a few sample points.
As follows from the the asymptotics of $`K_0(z)`$ at large $`z`$
$$K_0(z)\sqrt{\frac{\pi }{2z}}e^z$$
the width of $`\lambda (b)`$ (3) is of the order $`m^1`$. The function $`G_0`$ is smooth in the vicinity of the origin and its width $`r_{0H}=\sqrt{r_0^2+r_H^2}`$ is substancially larger than the width of $`\lambda `$. For this reason we factor $`G_0(B+c)`$ out of the integral (8) over $`c`$ at the point $`B`$:
$$j(B,b)=G_0(B)\mathrm{\Lambda }(b),\mathrm{\Lambda }(b)\lambda (c)\lambda (cb)d^2c=\frac{|b|}{4\pi m}K_1(m|b|)$$
(11)
Then we find that the integrals over $`B`$ and $`b`$ decouple
$$J(K_{})=\rho _A(\underset{p+1}{\underset{}{0,\mathrm{},0}}|\underset{p+1}{\underset{}{0,\mathrm{},0}})G_0^p(B)d^2B\mathrm{\Lambda }^p(b)e^{ibK_{}}d^2b$$
(12)
In this approximation we find that $`|K_{}|`$ depends only on one parameter - the constituent quark mass $`m`$, which in our approach plays the role of an infrared cutoff:
$$|K_{}|=pm_0^{\mathrm{}}𝑑zK_0(z)(zK_1(z))^{p1}$$
(13)
This allows to relate $`m`$ directly to the experimental data on the transverse momentum dependence.
## 5 Comparison with the data and discussion
The integral in (13) can be easy calculated numerically. For values of $`p=1,\mathrm{},12`$ it is very well approximated by a power dependence (see Fig.2), so that we obtain
$$|K_{}|/m=1.594p^{0.625}$$
(14)
As we observe, the rise of $`|K_{}|`$ turns out to be even faster than expected on naive physical grounds mentioned in the Introduction ($`\sqrt{p}`$). The resulting plots for $`|K_{}|^2`$ as a function of the cumulative number $`x=(p+1)/3`$ at different values of parameter $`m`$ are shown in Fig.3 together with avaiable experimental data from on $`K_{}^2`$ for pion production obtained in experiments - with 10 $`GeV`$ protons and with 8.94 $`GeV`$ protons.
Note that earlier publications of the first group reported a much stronger increase of $`K_{}^2`$ with $`x`$, up to value 2 $`(GeV/c)^2`$ at $`x=3`$ for pion production. In our approach such an increase would require the quark mass to be as high as $`m225MeV`$. In a more recent publication the rise of $`K_{}^2`$ is substancially weaker (it corresponds to $`m175MeV`$ in our approach). The authors of explain this by new experimental data obtained and by a cutoff $`K_{max}`$ introduced in calculations of $`K_{}^2`$ in . The introduction of this cutoff considerably (approximatly two times) decreases the experimental value of $`K_{}^2`$ at $`x=3`$. In our opinion this is a confirmation that the cumulative pion production rate only weakly decreases with $`K_{}`$ in the cumulative region so that the the integral over $`K_{}^2`$ which enters the definition of $`K_{}^2`$ is weakly convergent or even divergent, as in our approach. Undoubtedly presentation of the experimental data in terms of the mean value $`|K_{}|^2`$, rather than $`K_{}^2`$ should reduce the dependence on the cutoff $`K_{max}`$ and make the results more informative.
One of the ideas behind the investigations of the cumulative phenomena is that they may be a manifestation of a cold quark-gluon plasma formed when several nucleons overlap in the nuclear matter. In we pointed out that our model does not correspond to this picture. It implies coherent interactions of the active quark with donors and, as a result, strong correlations between the longitudinal and transverse motion. Predictions for the dependence of $`|K_{}|`$ on $`x`$ are also different. From the cold quark-gluon plasma model one expects $`|K_{}|`$ to behave as $`x^{1/3}`$, since the Fermi momentum of the quarks inside the overlap volume is proportional to the cubic root of the quark density. Our model predicts a much faster increase, with a power twice larger. The experimental data seem to support our predictions.
## 6 Acknowledgments
The authors are greatly thankful to Prof. P.Hoyer who attracted their attention to the problem.
This work is supported by the Russian Foundation for Fundamental Research, Grant No. 97-02-18123.
Figure captions
The diagram for the production of a cumulative quark with the momentum $`K`$ in the scattering of a projectile hadron with the momentum $`H`$ off a nucleus $`A`$ with the momentum $`P`$. Dashed and chain lines show gluon and pomeron exchanges, respectively.
The $`|K_{}|/m`$ as a function of $`p`$. The points are the results of calculations on (13). The line is the best power fit.
The $`|K_{}|^2`$ as a function of the cumulative number $`x=(p+1)/3`$. The lines are the results of calculations on (13) at different values of parameter $`m`$. The points ($``$) are the experimental data from on $`K_{}^2`$ for pion production with a cutoff (see the text) obtained in experiments - on a bombardment of nuclei by 10 $`GeV`$ and 9 $`GeV`$ protons. The points ($``$) are the data from the earlier publications of the group without a cutoff.
M.Braun and V.Vechernin
Fig. 1
M.Braun and V.Vechernin
Fig. 2
M.Braun and V.Vechernin
Fig. 3
M.Braun and V.Vechernin
|
no-problem/9904/adap-org9904004.html
|
ar5iv
|
text
|
# Evolution dynamics in terraced NK landscapes
## Abstract
We consider populations of agents evolving in the fitness landscape of an extended $`\mathrm{𝐍𝐊}`$ model with a tunable amount of neutrality. We study the statistics of the jumps in mean population fitness which occur in the ‘punctuated equilibrium’ regime and show that, for a wide range of landscapes parameters, the number of events in time $`t`$ is Poisson distributed, with the time parameter replaced by the logarithm of time. This simple log-Poisson statistics likewise describes the number of records in any sequence of $`t`$ independently generated random numbers. The implications of such behavior for evolution dynamics are discussed.
PACS numbers: 87.10.+e, 87.15.Aa, 05.40.-a
Introduction. Evolutionary dynamics can be described as a stochastic process unfolding in a ‘fitness landscape’, a process which can be simulated by means of genetic algorithms . The dynamical behavior of such algorithms is controlled by a number of parameters as e.g. the population size, the rate of mutation and the strenght of selection.
A pervasive dynamical regime is the so-called ‘punctuated equilibrium’ or ‘epochal behavior’, where relevant measures of evolution, as e.g. the mean fitness, remain constants for long periods during which the fitness distributions of the individuals is strongly peaked. Occasionally, a fitter mutant appears and quickly ‘takes over’ the population (see e.g. Fig. 1). Real experiments performed on bacterial colonies evolving in a controlled environment have shown that the fitness and cell size increase at a decelerating rate . A similar slowing down is discussed by Kauffman in the ‘long jump’ dynamics of the $`\mathrm{𝐍𝐊}`$ model, while Aranson et al. find a logarithmic growth of the average fitness for quasi-species evolving in a rugged fitness landscape.
In a macroevolutionary context, Raup and Sepkoski suggested that the noticeable decay of the extinction rate might stem from the properties of an underlying optimization process. This idea was taken up in the ‘reset’ model, where jumps in the average fitness of populations are linked to fitness record achieved during evolution. Such a link between small and grand scale evolution is rather controversial: If macroevolutionary events are mainly driven extrinsically, e.g. by meteorite impacts, any patterns in the fossil record must ultimately stem from the mechanics of celestial bodies in the solar system. Conversely, if, as recently proposed by several authors, the fossil record is mainly shaped by complex interactions within the biotic system, macro-evolutionary patterns must emerge from population dynamics.
Our main interest lies in the statistical properties of the evolutionary jumps underlying fitness changes. Since the role of neutral mutations for evolution is well established, and since punctuated equilibria exist even in the absence of local fitness maxima, we chose to study evolution on ‘terraced’ landscapes with a tunable degree of neutrality.
We find that, on average, the number of jumps taking place in time $`t`$ grows proportionate to $`\mathrm{log}t`$, and that the rate of events consequently decays as $`1/t`$. Secondly and most importantly, we show that the record dynamics provides a reasonable description of microevolution, with some limitations which are also outlined. Thirdly, we emphasize that power-law decays generically characterize the correlation functions of time series generated by a record driven dynamics.
Method. Each ‘genome’ in a population constitutes a point in an abstract configuration space usually called a fitness landscape . To construct such a landscape we use an elegant prescription due to Kauffman, the widely known $`\mathrm{𝐍𝐊}`$ model: We represent genomes as strings of $`𝐍`$ bits $`𝐱=(x_1,x_2,\mathrm{}x_N)`$, each being either $`0`$ or $`1`$. The fitness $`F(𝐱)`$ of configuration $`𝐱`$ is defined as
$$F(𝐱)=\frac{1}{𝐍}\underset{i=1}{\overset{𝐍}{}}f_i,$$
(1)
where the contribution from site $`i`$, $`f_i`$, is a random function depending on $`x_i`$ and $`𝐊`$ other $`x_l`$’s. More precisely, $`f_i`$ is a random function of $`2^{𝐊+1}`$ arguments with values uniformly distributed in $`(0,1]`$. We let $`\mu _f`$ and $`\sigma _f`$ be the average and spread of the distribution from which the $`f_i`$ values are generated.
If $`𝐊`$ is zero, the change in fitness due to the change of one $`x_i`$ (a point mutation) is of order $`1/N`$, and the landscape may be regarded as smooth. By way of contrast, when $`𝐊=𝐍\mathrm{𝟏}`$ a single point mutation changes all the $`f_i`$’s, and the landscape becomes ‘rugged’. Intermediate cases correspond, of course, to intermediate $`𝐊`$ values.
Our version of the $`\mathrm{𝐍𝐊}`$ model is modified in two respects. Firstly, the sum in Eq.1 is shifted by $`\mu _f`$ and scaled by $`\sqrt{𝐍}`$ rather than $`N`$. This ensures that the distribution of fitness values keeps the same variance for any value of $`N`$. More importantly, we introduce tunable neutrality by discretizing the fitness values into ‘terraces’, according to the formula:
$$F^{}(𝐱)=\frac{\mathrm{nint}(j\sqrt{𝐍}[F(𝐱)\mu _f]+j\mu _f)}{j}.$$
(2)
Here $`\mathrm{nint}`$ stands for the nearest integer function, and $`j`$ denotes the number of terraces. For small $`j`$ there are few broad terraces, while for $`j\mathrm{}`$ the fitness values approach a continuum and the original $`\mathrm{𝐍𝐊}`$ model is regained. Accordingly, one expects that, as seen e.g. in Fig. 2, the effect of varying $`j`$ should be stronger for small values of $`j`$.
In the simulations, a configuration $`𝐱`$ is cloned with a probability $`p\mathrm{exp}(\beta F(𝐱))`$. With probability $`u<<1`$, the cloned string undergoes a one-point mutation at a randomly chosen locus. Finally, a random and fitness independent deletion mechanisms is applied, which keeps the population size fluctuating around a fixed average $`n`$. Both the generation $`/`$ mutation part and the deletion part of the algorithm are performed sequentially. Subsequently, the information about the new fitness distribution is incorporated in the cloning probabilities. This entire process counts as one update and defines the unit of time.
Punctuated equilibrium dynamics requires that a fitter mutant be able to survive and spread in the population on a time scale short compared to the inverse mutation frequency. Therefore, $`u`$ should not be too high and $`\beta `$ should not be too low. Within these constraints $`u`$ should be as high as possible, in order to have a good statistics within the time window of the computation. Also, too high a $`\beta `$ value quenches the dynamics completely. These design consideration lead to the values of $`u`$ and $`\beta `$ used throughout the calculations.
A concise description of the dynamics is provided by the average $`\overline{f}(t)`$ of the distribution of fitness values through the population. Such a trajectory is shown in the main panel of Fig. 1 to consist of a number of flat plateaus separated by rather well defined jumps. The number $`m(t)`$ of jumps occurring in the interval $`[0,t)`$ is a stochastic process whose distribution can be sampled by repeating the simulations or, equivalently, by considering an ensemble of landscapes, where different trajectories are generated by independently updating each system. The ensemble average and variance of $`m`$ are denoted by $`E(m)`$ and $`\sigma ^2(m)`$ respectively, while $`a_m(t)`$ and $`v_m(t)`$ are the corresponding estimators. The notation describing the inputs and results of our simulations is summarized in Table 1.
Results. The basic qualitative features of the data are expressed by Fig. 1: Its left panel shows the mean fitness $`\overline{f}`$ of a single population on a semilogarithmic scale. The right panel details the behavior close to the evolutionary jump at $`t3\times 10^4`$ by depicting the distribution of fitness values right before, during and right after the jump. Nearly all strings have the same fitness in the initial and final situations, while the transition stage features two different fitness values. Punctuated equilibrium behavior and a very peaked fitness distribution are widely found in previous studies as well as in our simulations.
For a more quantitative data analysis, the statistical properties of the jumps and their associated waiting times must be studied. We let $`m_i(t)`$ be the number of jumps occurred at time $`t`$ in trajectory $`i`$, and consider the sample average:
$$a_m(t)=\frac{_{i=1}^sm_i(t)}{s},$$
(3)
and the sample variance:
$$v_m(t)=\frac{_{i=1}^s(m_i(t)a_m(t))^2}{s1}$$
(4)
as functions of time. To calculate error bars on $`a_m`$ and $`v_m`$ we need the corresponding standard deviations, which are $`\sigma (m)/\sqrt{s}`$ and $`\sqrt{(\sigma ^2(m^2)+8E^2(m)E(m^2)4E(m)(E^3(m)+E(m^3)))}/\sqrt{s}+𝒪(1/s)`$. The latter relation results from straightforward but rather tedious algebra. To lowest order in $`s^{1/2}`$ one may now replace the moments of $`m`$ with their sample estimators. In general, the relative errors on various quantities of interest are of order $`1/\sqrt{s}10\%`$.
Fig. 2 shows the average number of jumps, for a number of different parameters values, as a function of time. $`E(m)`$ is seen to grow logarithmically, with a strong $`𝐊`$ dependence slope, for ‘short’ log-times. The leveling off noticed at large times stems from the fact that, as the fitness increases, fitness improvements become progressively smaller. Eventually, they get lost in the noise, rather than triggering a jump. At this point record statistics and fitness evolution must part company.
For large $`t`$, the probability of $`n`$ records in a sequence of $`t`$ independently drawn random numbers is given by:
$$P_t(n)=\frac{(\mathrm{log}\lambda t)^n}{n!}t^\lambda .$$
(5)
This is a Poisson distribution with $`\mathrm{log}t`$ replacing the time argument. The strength parameter $`\lambda `$ describes the possibility that many searches for records take place independently and in parallel and/or the situation where records remain undetected. In our systems we observed that several records were indeed lost in the noise, and did not trigger any evolutionary event.
A mathematically equivalent description of record dynamics is provided by the distribution of the variables
$$\mathrm{\Delta }_k=\mathrm{log}t_k\mathrm{log}t_{k1}=\mathrm{log}(t_k/t_{k1}).$$
(6)
It follows from standard arguments that in a (log) Poisson process these $`\mathrm{\Delta }_k`$ are independent and have the common distribution:
$$P(\mathrm{\Delta }>x)=\mathrm{exp}(\lambda x).$$
(7)
The empirical distribution of the $`\mathrm{\Delta }_k`$’s is shown on a semilogarithmic scale in Fig. 3 for $`𝐊=31`$ and for differing degrees of terracing. The decay of $`P(\mathrm{\Delta }>x)`$ seems quite well described by an exponential for $`x5`$, with the tail of the distribution falling off more rapidly, likely due to the inevitably poor sampling of large $`x`$ values in a finite time simulation. Banning the effect of the deviations from pure record statistics, $`\lambda `$ is, by Eqs. 5 and 7, the slope of $`a_m(t)`$ vs. $`\mathrm{log}t`$ as well as minus the slope of $`\mathrm{log}P(\mathrm{\Delta }>x)`$ vs $`x`$. We calculated these slopes from the data in both ways (cutting off the tails of the data), and obtained the following results. $`j=10^2,\lambda =0.54(0.34)`$; $`j=10^3,\lambda =0.49(0.39)`$; $`j=10^4,\lambda =0.58(0.44)`$; $`j=10^5,\lambda =0.54(0.35)`$; $`j=10^6,\lambda =0.62(0.47)`$; and $`j=10^9,\lambda =0.63(0.39)`$. The figures in parentheses stem from Fig. 3. There is a rough agreement, but certainly also considerable scatter in these data, with the log-wait time type of analysis yielding systematically lower figures. At this stage it is unclear whether the non-monotonic dependence of $`\lambda `$ on $`j`$ seen in Fig. 3 is a real effect - or just due to a combination of statistical fluctuations and systematic deviations from the ideal log-Poisson behavior.
Summarizing the results from Figs. 2 and 3, it appears that the value of $`𝐊`$ very strongly affects the average slope of the curves (i.e. the value of $`\lambda `$). The degree of terracing might also have an effect on the slope, albeit a minor one.
The independence of the different $`\mathrm{\Delta }_k`$’s implied by the record statistics was tested by calculating the correlation coefficients $`C_{ij}`$ between $`\mathrm{\Delta }_i`$ and $`\mathrm{\Delta }_j`$. In practice, we checked for $`C_{12}`$, $`C_{23}`$ and $`C_{13}`$. As expected, the highest degree of correlation was found for the relatively smooth landscape with $`𝐊=7`$. In this case, the $`C`$ values were close to $`0.4`$. For $`𝐊=31`$ the correlation coefficients were of the order of $`0.1`$, i.e. of the same order as the statistical sampling error.
To conclude the description of our data we plot in Fig. 4 the estimated variance $`v_m(t)`$ versus the estimated average $`a_m`$, for $`𝐊=31`$ and a number of terrace values. For a perfect agreement with the log-Poisson distribution the points should lie on a straight line of slope $`1`$. This is close to the observed behavior, except for the highest values (where, on the other hand, the statistics is poorest). A similar plot for $`𝐊=15`$ and $`𝐊=7`$ shows a systematic deviation from a straight line, with considerably less variance than in a purely random case. Additional details on the simulations and on the genetic algorithm utilized to produce them can be found in Ref. .
In summary, the dynamics of our evolutionary model is time inhomogeneous stochastic process, with a rate of events falling off as $`1/t`$. The log-Poisson statistics describes the data best for landscapes with large $`𝐊`$, with or without terraces. In this respect terraces have a minor effect on the dynamics.
Discussion. While the shape of the fossil record certainly reflects many different factors, including e.g. biogeography and external perturbations of the abiotic environment, the event statistics demonstrated here should have rather general implications for all models which do not completely dismiss the influence of population dynamics on macroevolution. If the evolutionary ‘jumps’ are the elementary events in any dynamics, possibly involving interactions between evolving species, using $`\mathrm{log}t`$ as the independent variable makes this dynamics appear as a stationary, markovian process, described e.g. by a master or Fokker-Planck equation. The eigenvalues (and eigenvectors) of the evolution equation describe the relaxation of any average of interest. Since the observational time window is usually narrow in $`\mathrm{log}`$ time, one is restricted to observing the decay of a single (or a few) relaxational mode(s). As an exponential function of $`\mathrm{log}`$ is a power-law, the above mechanism offers a generic explanation for the power-law like behavior found in several evolutionary patterns.
Aknowledgments. Both authors would like to thank Mark Newman and Richard Palmer for inspiring conversations and exchanges of ideas at the Santa Fe Institute of Complex Studies and at the Telluride Summer Research Institute. Parts of this work were commenced during a visit by A. P. at Duke University. A. P. owes a special thank to Richard Palmer for the kind hospitality extended to him and for the guidance he received during his stay. This project was partly supported by a block grant from Statens Naturvidenskabelige Forskningsråd.
|
no-problem/9904/cond-mat9904373.html
|
ar5iv
|
text
|
# Molecular weight effects on chain pull-out fracture of reinforced polymeric interfaces
## Abstract
Using Brownian dynamics, we simulate the fracture of polymer interfaces reinforced by diblock connector chains. We find that for short chains the interface fracture toughness depends linearly on the degree of polymerization $`N`$ of the connector chains, while for longer chains the dependence becomes $`N^{3/2}`$. Based on the geometry of initial chain configuration, we propose a scaling argument that accounts for both short and long chain limits and crossover between them.
Creating advanced materials often means mixing different homo-polymers to produce systems with desired combined properties. However, most polymer blends are immiscible: They form macroscopically phase-separated mixtures with only interfacial van der Waals forces keeping domains of different phases together. The fracture toughness of such blends is limited by one of the interfaces: In ideal defectless conditions, it is equal to the work of adhesion $`W=\gamma _a+\gamma _b\gamma _{ab}`$ between two homo-polymer phases. Here, $`\gamma _a`$ and $`\gamma _b`$ are respectively the surface energy of homo-polymer A and homo-polymer B and $`\gamma _{ab}`$ is the interfacial free energy. Reinforcement of these weak polymeric interfaces is often achieved by the addition of A-B diblock copolymers which compatibilize the blend and strengthen the interface . The strengthening can be attributed to the miscibility of each block with one of the homopolymers. This causes the block copolymer to expand and entangle with homo-polymer phases on either side of the interface. The interfacial tension (energy) is reduced, the interfacial width is increased, and the adhesion thereby improved. Fracture toughness and failure mechanisms of such reinforced polymeric interfaces have been investigated extensively by, for example, experiments on different incompatible systems of polymer glasses and cross-linked networks (elastomers) .
Several theoretical models have been proposed to explain the reinforcing effect of connector chains in both elastomers and glassy polymers. A “failure mechanism map” has been developed which relates the mechanism of interface failure to the polymerization index $`N`$, surface density $`\sigma `$ of connector chains, and the time scale on which the deformation occurs. According to the failure map there are three major mechanisms: (i) chain scission, which happens whenever the stress along the connector chain becomes larger than the strength of the covalent bond between segments of the chain, (ii) pull-out of the connector chain as a result of disentanglement from homo-polymer phase, and (iii) failure by craze formation, followed by chain scission or chain pull-out, which take place when a large stress is transfered to the bulk of the homo-polymer phases.
In this paper we direct our attention to the case where interface failure is due to chain pull-out, and specifically focus on the effect of the polymerization index $`N`$ of the connector chain. We consider a “mushroom regime” where connector chains are grafted with low surface density, $`\sigma 1/Nl^2`$ ($`N`$ is the number of monomers per grafted chain and $`l`$ is the monomer size). In this low-density regime, the equilibrium shape of the connector chain in the homo-polymer phase is mushroom or plume-like. Furthermore, mutual entanglement between different connectors is negligible. Hence we need consider only the behavior of a single chain. Experimental data on the dependence of $`G`$ on $`N`$ is scattered, typically assuming a form
$$GN^\alpha ,$$
(1)
where estimates give variously $`1\alpha 2`$ . Both linear and quadratic dependences of $`G`$ on $`N`$ have been predicted, utilizing a tube picture . Different constitutive equations have been proposed which relate the local stress in planar cohesive zone near the crack tip to various phenomena, such as, the penetration depth of the chains, their surface density, and the pulling rate . These studies have been focused mainly on the pull-out fracture in tensile mode, when the applied force is normal to the interface. An alternative mode of interface failure is shear fracture, or the resistance of the interface against slip. In fracture mechanics of bulk materials generally only the tensile or opening mode is important, since cracks normally travel in a direction that maximizes the opening mode. However, for the interface between two different materials, the situation can be more complex: The crack is often constrained to follow the interface, giving rise to the possibility of crack propagation involving a combination of tensile and shear modes . Herein we study the dynamics of chain pull-out fracture in both tensile and shear mode separately and examine the dependence of $`G`$ on $`N`$.
We use a Brownian dynamics method, due to Picket, Jasnow, and Balazs , to simulate the pull-out of a single connector chain of length $`N`$ from a two-dimensional homo-polymer phase. To quantify the interface toughness for different chain length, we calculate the work that is required to pull out the chain with constant velocity $`𝐯_\mathrm{𝟎}`$. The homo-polymer phase is modeled by a two-dimensional semi-infinite square lattice of obstacles. Each obstacle represents an entanglement or cross linked point, depending on the glassy or elastomeric structure of the homo-polymer phase, and provides lateral constraint on the movement of the connector chain. The connector is represented by a freely jointed chain with $`N`$ links wherein there is no self-interaction. Hence individual monomers can freely pass over each other, provided links have a constant length. The initial configuration is created by putting the first monomer at the boundary and then continuing the chain as a random walk in dimension $`d=2`$, with a reflecting boundary at the interface. The random walk is restricted by requiring, as we shall see below, that monomers are repelled by the obstacles. The pull-out dynamics is simulated by pulling the chain by the first monomer at a constant velocity in either tensile mode (perpendicular to the interface) or shear mode (parallel to the interface). This is done conveniently by moving the obstacle matrix at velocity $`𝐯_\mathrm{𝟎}`$ while keeping the first monomer fixed. That is, if $`𝐫_𝐢`$ is the position of the ith monomer, $`d𝐫_\mathrm{𝟏}/dt0`$.
The movement of the other $`i=2,\mathrm{},N`$ monomers is governed by the following over-damped Langevin equation.
$$\nu \left(\frac{d𝐫_𝐢}{dt}𝐯_\mathrm{𝟎}\right)=\tau _i(𝐫_{𝐢+\mathrm{𝟏}}𝐫_𝐢)\tau _{i1}(𝐫_𝐢𝐫_{𝐢\mathrm{𝟏}})+𝐅_𝐢+\eta _𝐢,$$
(2)
where $`𝐫_{𝐍+\mathrm{𝟏}}𝐫_𝐍`$. Here $`\tau _i`$ is the amplitude of tension in the segment connecting monomers $`i`$ and $`i+1`$, $`\eta _i`$ is the random Gaussian noise representing the effect of thermal fluctuations on each monomer, and $`\nu `$ is the viscous friction coefficient per monomer. The strength of the random noise is related to the monomeric friction by the fluctuation-dissipation theorem $`\eta _i(t)\eta _j(t^{^{}})=2\nu k_BT\delta _{ij}\delta (tt^{^{}})`$, where $`k_B`$ is Boltzmann’s constant and $`T`$ is temperature. $`𝐅_i(\mathrm{\Delta }𝐫_𝐢)`$ is a short range monomer-obstacle repulsive force, where $`\mathrm{\Delta }𝐫_𝐢`$ is the distance between the ith monomer and the closest obstacle. For $`|\mathrm{\Delta }𝐫_𝐢|<r_c`$, there is hard-core repulsion. For $`r_c<|\mathrm{\Delta }𝐫_𝐢|<2r_c`$, there is soft-core repulsion, where the force obeys
$$𝐅_𝐢=\sigma \mathrm{\Delta }𝐫_𝐢\left(\frac{1}{|\mathrm{\Delta }𝐫_𝐢|^2r_c^2}\frac{1}{3r_c^2}\right)^2.$$
(3)
The parameter $`\sigma `$ controls the strength of the force . For larger separations, $`|\mathrm{\Delta }𝐫_𝐢|>2r_c`$, there is no force acting on the monomer. In our simulations we have used $`r_c=l/2`$, where $`l`$ is the link length of the connector chain.
To solve Eq. (2), the tensions $`\tau _i`$ in each segment of the connector chain have to be determined. This is done by enforcing the constant segmental length constraint:
$$|𝐫_{𝐢+\mathrm{𝟏}}𝐫_𝐢|^2=l^2,$$
(4)
for $`1<i<N1`$. This results in a tridiagonal matrix equation for the $`\tau _i`$’s :
$`{\displaystyle \frac{d(𝐫_{𝐢+\mathrm{𝟏}}𝐫_𝐢)^2}{2dt}}`$ $`=`$ $`\tau _{i+1}(𝐫_{𝐢+\mathrm{𝟐}}𝐫_{𝐢+\mathrm{𝟏}})(𝐫_{𝐢+\mathrm{𝟏}}𝐫_𝐢)2\tau _il^2+`$ (5)
$`+`$ $`\tau _{i1}(𝐫_𝐢𝐫_{𝐢\mathrm{𝟏}})(𝐫_{𝐢+\mathrm{𝟏}}𝐫_𝐢)+`$ (6)
$`+`$ $`(𝐅_{𝐢+\mathrm{𝟏}}+\eta _{𝐢+\mathrm{𝟏}}𝐅_𝐢\eta _𝐢)(𝐫_{𝐢+\mathrm{𝟏}}𝐫_𝐢).`$ (7)
Ideally, by applying the constant segmental length constraint, the left-hand side of Eq. (4) should be zero. However, after several updates of the simulation, as a result of accumulation of roundoff errors, the distances between some of the adjacent monomers differ slightly from $`l`$. To improve the stability of the algorithm, a correction, restoring the original segment length, is introduced:
$$\frac{d(𝐫_{𝐢+\mathrm{𝟏}}𝐫_𝐢)^2}{2dt}=|𝐫_{𝐢+\mathrm{𝟏}}𝐫_𝐢|\frac{(|𝐫_{𝐢+\mathrm{𝟏}}𝐫_𝐢|l)}{dt}.$$
(8)
Eq. (2) is solved using a fourth-order Runga-Kutta algorithm.
We consider reduced units in which $`\nu `$, $`v_0`$, $`\sigma `$, $`k_BT`$, and $`D`$, the distance between obstacles, are set equal to unity. The distance between two adjacent monomer in connector chain $`l`$ is chosen to be 0.4. The time step of the simulation, typically $`10^4`$ to $`10^5`$, is adjusted so that the average difference in segment length from $`l`$ in each run is less than 0.1 percent. The results were averaged over $`20100`$ independent realizations of the initial conditions.
Fig. 1 shows a snapshot of the simulation for both tensile and shear pull out. We quantify the fracture toughness of the interface by determining the work $`G`$ required to remove the connector chain completely from the obstacle lattice. The work is $`G=P𝑑t`$, where the power $`P`$ is obtained via $`P=v_0\tau `$, $`v_0`$ is the constant pulling velocity, and $`\tau `$ is the instantaneous tension in the segment crossing the boundary of the obstacle lattice in the pulling direction at each time step.
Fig. 2 shows the fracture toughness $`G`$ as a function of the polymerization index $`N`$ of the connector chain for tensile and shear modes. Naturally, the fracture toughness grows as the degree of polymerization of the connector chain increases. This growth is more significant for larger $`N`$ due to the fact that longer connectors can penetrate well beyond the neighborhood of the interface. They entangle efficiently with the bulk polymers of the compatible phase. It also shows that the effect of reinforcement on the interface is higher in tensile mode, by a factor of about five for the present model . One can notice the existence of two scaling regimes for $`G`$ vs. $`N`$, corresponding to short connectors, $`N<N_c`$, and long connectors, $`N>N_c`$ where $`N_c`$ is the crossover length.
Figure 3 shows the power-law dependence of $`G`$ on $`N`$. For short connectors $`G`$ scales linearly with $`N`$, $`GN`$, while for long connectors it scales as $`GN^{3/2}`$ .
These regimes can be understood as follows. For small $`N`$, the chain is entangled with one or two layers of obstacles, even if the free radius of gyration $`R_gl\sqrt{N}`$ is of order or smaller than the obstacle lattice spacing $`D`$. Since the the initial configuration is repelled by the obstacles, the available phase space for a chain is quite restricted. For the first few chain segments, the reflecting boundary condition and the the area taken up by the obstacles, results in the chains going straight into the lattice. After passing the first row or two of obstacles, the chain, if it is long enough, usually bends, with tight loops around obstacle cores being highly unprobable. When the chain is pulled out, the dominant deterministic part of the tension equals the total viscous drag force. This force is roughly proportional to the number of monomers $`N_{move}`$ simultaneously in motion. Since pulling tension cannot propagate through a loop unless it is tightened around an obstacle core, for short chains $`N_{move}`$ is of order of the number of monomers stretched in one obstacle lattice spacing $`D`$, that is $`D/l`$. Hence for small $`N`$, the total work to pull the chain out can be estimated as
$$Gv_{0}^{}{}_{}{}^{2}\nu \frac{D}{l}_0^{Nl/v_0}𝑑t=v_0\nu ND,$$
(9)
which is consistent with the linear scaling regime observed in simulations for short connectors.
For large $`N`$, the penetration depth increases. Indeed, for sufficiently large $`N`$, the chain center of mass is located at a distance of order of the radius of gyration $`R_g`$ from interface. Then, the number of monomers that are simultaneously in motion becomes proportional to this distance, $`N_{move}R_gN^{1/2}`$, which we have observed directly. Consequently, the pull-out fracture energy becomes
$$G=v_0_0^{Nl/v_0}F_{drag}𝑑tv_0\nu N^{3/2}l.$$
(10)
This is roughly analogous to what happens when one pulls on a wet garden hose of length $`N`$, left on the ground by a gardener who has performed a random walk among a grove of trees. Upon pulling, only the currently stretched segment starts to move, if friction in the self-intersections of the hose is negligible. If the hose configuration was created as a result of random walk, the length of such a segment scales as a typical size of such a walk, $`\sqrt{N}`$, where $`N`$ is the total length of the hose. The hose configuration may include a few loops around the tree trunks, which, upon tightening, also start to move. The total length of these tightened loops has no important contribution: That length scales as a winding angle of a random walk of the length $`N`$, i.e., as $`\mathrm{ln}N`$ (see, for example, Ref. ). As a result, the drag force is proportional to the length of the hose that is constantly in motion, $`\sqrt{N}`$.
Although our numerical work is in two dimensions, these physical arguments also follow in three dimensions, and are applicable to experimental systems . The small $`N`$ regime crosses over to a large $`N`$ regime when the radius of gyration $`R_g`$ becomes of order of the obstacle spacing $`D`$. That gives $`N_{cross}[D/l]^2`$, and we expect
$$G=Nf(N/N_{cross}),$$
(11)
where the crossover scaling function obeys $`f(x0)=\mathrm{const}`$, and $`f(x\mathrm{})=x^{1/2}`$.
As mentioned earlier, different forms of the power-low dependence of $`G`$ on $`N`$ have been proposed in previous theoretical studies: Both $`GN`$ and $`GN^2`$ have been predicted. The results from experiments are somewhat ambiguous. The linear dependence of $`G`$ on $`N`$ is predicted for very slow crack propagation velocities, and therefore corresponds to $`GG_0`$, where $`G_0`$ is the fracture toughness threshold or minimal energy required to break the interface when $`v_00`$. On the other hand, $`GN^2`$ is predicted for pull-out fractures where the crack propagation speed is high or at least larger than a critical velocity. In other words, these two predicted regimes differ by the pull-out velocity. The crossover between the two scaling regimes $`GN`$ and $`GN^{3/2}`$ obtained in our simulations takes place in the latter high velocity regime: The pulling velocity used in our simulations $`v_0=k_BT/(\nu D)`$ corresponds to relatively fast pulling rate , which in turn results in high crack propagation speeds. Our work is in the same regime, and indeed extends the numerical work and theoretical arguments of Picket, Jasnow, and Balazs , wherein this model was introduced. For large $`N`$, Picket et al. argued $`GN^2`$ in contrast to our result of $`GN^{3/2}`$. This was based on an analogy of the polymer to the motion of a rope in “block and tackle” pulleys. In that case, the average drag force to pull the connector chain out of the matrix is $`f_{drag}\nu v_0N`$. Hence, it follows that the number of monomers instantaneously in motion are $`N_{move}N`$. In fact, the correct analogy is not to pulleys, but to the motion of a garden hose lying on grass, as described above. Then $`f_{drag}N^{1/2}`$, giving rise to the $`3/2`$ exponent. Our numerical results, which are more extensive than those of the earlier work, support this picture.
To conclude, we studied the chain pull-out fracture of a reinforced polymeric interface in tensile and shear modes using molecular dynamics algorithm. Our results confirm the nonlinear dependence of the fracture toughness of the interface on the length of the connector chain observed in experiments. We found that, depending on the length of connector chain, the fracture toughness of the interface shows different scaling dependences: For short chains, $`G`$ scales linearly with $`N`$, while for long connectors, we observed a crossover to a new scaling regime $`GN^{3/2}`$. Our results can be tested experimentally on elastomeric networks for long connector chains in the low coverage mushroom regime .
This work was supported by the Natural Sciences and Engineering Research Council of Canada, and le Fonds pour la Formation de Chercheurs et l’Aide à la Recherche de Québec.
|
no-problem/9904/cond-mat9904315.html
|
ar5iv
|
text
|
# Impurity-induced spin polarization and NMR line broadening in underdoped cuprates
## I Introduction
The normal state of underdoped cuprates exhibits unusual magnetic properties which are believed to be intimately related to the mechanism of high-$`T_c`$ superconductivity. Most peculiar in this respect are the simultaneous occurrence of a magnetic pseudogap and the persistence of antiferromagnetic (AF) correlations as holes are doped into the antiferromagnetic insulator and the system becomes metallic. It is one of the most challenging theoretical problems in the physics of high-$`T_c`$ cuprates to reconcile the gaplike features reminiscent of a spin liquid with the presence of antiferromagnetic correlations signaling the closeness of the system to a spin-ordered Néel state. Experimentally, insight into the nature of these anomalous features can be gained by introducing impurities into the magnetically active Cu sites. A subsequent NMR probe on nuclei coupled to the CuO<sub>2</sub> planes yields information on the local magnetic structure. In this paper we present a microscopic theory of the impurity-induced local spin polarization of CuO<sub>2</sub> planes and its impact on the NMR linewidth.
Introducing magnetically active or inert impurities into underdoped cuprates leads, in both cases, to the formation of local magnetic moments. Specifically, Cu ($`d^9`$) with an effective in-plane spin $`S=\frac{1}{2}`$ can be replaced by Ni ($`d^8`$) with $`S=1`$ or Zn ($`d^{10}`$) with $`S=0`$. Superconducting quantum interference device (SQUID) measurements of the macroscopic susceptibility reveal an almost perfect $`1/T`$ Curie behavior. Recently, Bobroff et al. presented NMR measurements on <sup>17</sup>O for the underdoped compound $`\text{YBa}_2(\text{Cu}_{1x}M_x)_3\text{O}_{6.6}`$, with $`M`$ = Zn or Ni. The polarization of Cu spins in the presence of impurities leads to a broadening of the NMR line. In contrast to the aforementioned SQUID measurement, the linewidth displays a marked non-Curie behavior, indicating an inherent temperature dependence of the polarizability of CuO<sub>2</sub> planes. This was suggested by Morr et al. to be a clear indication for a temperature dependence of the AF correlation length. Still another interesting observation can be made by comparing the two experiments: While the NMR study shows nonmagnetic Zn to have a more pronounced effect on the linewidth than Ni, measurements of the macroscopic susceptibility reveal a reversed effect. Since only the NMR experiment is sensitive to a spatial variation of the spin polarization, a very different shape of the spin density induced by the two types of impurities can be inferred.
In the following, we present a microscopic theory of moments induced by magnetic and nonmagnetic impurities in the spin gap phase of underdoped cuprates. We analyze the different nature of coupling between Cu and impurity spins and derive expressions for the local spin polarization of CuO<sub>2</sub> planes. The presence of the spin gap and of short-range AF correlations is shown to strongly modify the conventional Ruderman-Kittel-Kasiya-Yosida (RKKY) picture. Finally, we derive expressions for the NMR line broadening which account well for the peculiarities of the experimental data.
## II Impurity Model
The relevant physics of the CuO<sub>2</sub> planes of high-$`T_c`$ cuprates is believed to be described by the large-$`U`$ Hubbard or $`t`$-$`J`$ model. The dualism between itinerant charge motion and local electron interaction that is inherent to these models can, in an approximate way, be captured by introducing separate quasiparticles for spin and charge degrees of freedom. Within this picture, the normal state of underdoped cuprates is viewed as a phase in which spins form singlet pairs while coherence between holes that would eventually lead to superconductivity has not been established. We follow this line of thinking but restrict ourselves to the magnetic sector of the Hilbert space. Our starting point is the spin-$`\frac{1}{2}`$ AF Heisenberg model $`H_J=J_{ij}𝒔_i𝒔_j`$. Keeping in mind the presence of itinerant holes which prevent the system from developing long-range magnetic order, we treat this Hamiltonian within resonance valence bond (RVB) mean-field theory – this accounts well for the spin-liquid features of cuprates. The mean-field Hamiltonian is
$$H_{\text{RVB}}=\underset{ij\sigma }{}\left(\mathrm{\Delta }_{ij}f_{i\sigma }^{}f_{j\sigma }+\text{H.c.}\right).$$
(1)
Original spin operators $`𝒔_i`$ have been expressed in terms of fermionic operators by $`𝒔_i=\frac{1}{2}_{\sigma \sigma ^{}}𝝉_{\sigma \sigma ^{}}f_{i\sigma }^{}f_{i\sigma ^{}}`$ with Pauli matrix vector $`𝝉=(\tau ^x,\tau ^y,\tau ^z)`$. The local constraint prohibiting a double occupancy of sites has been relaxed to a global one. The mean-field bond parameter is $`\mathrm{\Delta }_{ij}=\mathrm{\Delta }_𝜹=J_\sigma f_{i+\delta ,\sigma }^{}f_{i\sigma }^0`$, where $`\mathrm{}^0`$ is the expectation value that corresponds to Hamiltonian (1). The phase of this mean-field parameter is yet undetermined and has to be chosen such as to resemble the experimental situation most closely. An appropriate choice for the spin gap regime is the flux phase $`\mathrm{\Delta }_{\pm x}=i\mathrm{\Delta }_{\pm y}\mathrm{\Delta }`$. Dividing the lattice into two sublattices $`A`$ and $`B`$ and going to the momentum representation, Hamiltonian (1) can be diagonalized
$$H_{\text{RVB}}=\underset{𝒌\nu }{}\xi _𝒌^\nu f_{𝒌\nu }^{}f_{𝒌\nu },$$
(2)
with index $`\nu =\pm `$. The spectrum of spin excitations or spinons is
$`\xi _𝒌^\pm =\pm 2\mathrm{\Delta }\left(\mathrm{cos}^2k_x+\mathrm{cos}^2k_y\right)^{1/2}.`$
It has nodes at $`(\pm \pi /2,\pm \pi /2)`$, yielding a V-shaped pseudogap in the density of states centered at the spinon chemical potential $`\mu _s=0`$: $`\rho ^{(0)}(\omega )=|\omega |/D^2`$ (defined per spin up/down state), where $`D=2\sqrt{\pi }\mathrm{\Delta }`$ is the spinon half-band-width.
To simulate first a nonmagnetic Zn impurity we introduce into Hamiltonian (1) a local chemical potential $`\lambda `$ acting on site $`𝑹=0`$, which by convention lies on sublattice $`A`$. In the limit $`\lambda \mathrm{}`$ spinons are expelled from that site, creating a vacancy. The Hamiltonian is then
$$H_{\text{Zn}}=H_{\text{RVB}}+\lambda \underset{\sigma }{}f_{0\sigma }^{}f_{0\sigma }|_\lambda \mathrm{}.$$
(3)
To describe a magnetic Ni impurity, into the empty site we insert an impurity spin $`𝑺_0`$ with $`S=1`$ which is coupled antiferromagnetically to the surrounding Cu spins $`𝒔_𝜹`$. The corresponding Hamiltonian is
$$H_{\text{Ni}}=H_{\text{RVB}}+\lambda \underset{\sigma }{}f_{0\sigma }^{}f_{0\sigma }|_\lambda \mathrm{}+H_{\text{imp}}$$
(4)
with the exchange interaction term
$`H_{\text{imp}}=J^{}{\displaystyle \underset{𝜹}{}}𝑺_0𝒔_𝜹.`$
Formally, Hamiltonian (4) differs from Eq. (3) only in the presence of an additional term $`H_{\text{imp}}J^{}`$. In the following, we put emphasis on the case of a magnetic impurity with $`J^{}>0`$. A nonmagnetic impurity can be simulated by setting $`J^{}=0`$, which decouples the impurity site from the rest of the system. The $`S=1`$ impurity spin is then free and can easily be disregarded. We discuss this limit in the following, but only shortly. More detailed treatments on nonmagnetic impurities are given in Ref. as well as in Refs. -.
## III Local Magnetic Moments
We analyze an impurity spin $`S=1`$ embedded in a spin gap system as described by Hamiltonian (4). Spinons stemming from the initial Cu spin at site $`𝑹=0`$ are ejected by the local potential $`\lambda `$. The impurity spin, which is placed in the vacant site, is conveniently represented by two spins $`\frac{1}{2}`$, i.e., $`𝑺=𝑺_a+𝑺_b`$. An infinitely strong ferromagnetic interaction $`H_c=J_c𝑺_a𝑺_b`$ between these two spins is assumed. Expressing $`𝑺_a`$ and $`𝑺_b`$ in terms of fermionic operators $`a_\sigma `$ and $`b_\sigma `$, respectively, a mean-field decoupling can be performed:
$$H_{\text{imp}}=\underset{𝜹\sigma }{}\left(\mathrm{\Delta }_𝜹^{}\frac{a_\sigma ^{}+b_\sigma ^{}}{\sqrt{2}}f_{𝜹\sigma }+\text{H.c.}\right)J_c𝑺_a𝑺_b.$$
(5)
Introducing operators $`f_{0\sigma }=\left(a_\sigma +b_\sigma \right)/\sqrt{2}`$ and $`d_\sigma =\left(a_\sigma b_\sigma \right)/\sqrt{2}`$, one obtains
$$H_{\text{imp}}=\underset{𝜹\sigma }{}\left(\mathrm{\Delta }_𝜹^{}f_{0\sigma }^{}f_{𝜹\sigma }+\text{H.c.}\right)J_c𝑺_{\text{eff}}𝒔_0.$$
(6)
The impurity spin has thus been decomposed into two $`S=\frac{1}{2}`$ effective spins $`𝑺_{\text{eff}}`$ and $`𝒔_0`$. The former is represented by operators $`d_\sigma `$, the latter by operators $`f_{0\sigma }`$. Due to the first term in Eq. (6), the $`f`$ spinons on the impurity site hybridize with the ones on adjacent Cu sites. This process is controlled by the local mean-field parameter $`\mathrm{\Delta }_𝜹^{}=J^{}_\sigma f_{𝜹\sigma }^{}f_{0\sigma }`$ replacing $`\mathrm{\Delta }_𝜹`$ on bonds connecting to the impurity. A system of itinerant spinons extending over the whole lattice including the impurity site is formed. These itinerant spinons couple ferromagnetically to the localized spin $`𝑺_{\text{eff}}`$. In the presence of a magnetic field this coupling is responsible for a polarization of the spinon system to be discussed in Sec. IV. The $`T`$ matrix that describes scattering of spinons on the localized spin vanishes as $`T(\omega )\omega \mathrm{ln}|\omega |`$ in the flux phase. This means that the effective local spin $`𝑺_{\text{eff}}`$ becomes asymptotically free in the limit of low energies. In the remainder of the present section we analyze this low-energy fixed point, emphasizing the role of bond parameters $`\mathrm{\Delta }_𝜹^{}`$ that induce an inhomogeneity in the spinon sector.
First we consider the special case of equal exchange integrals $`J^{}=J`$. Regarding the spinon sector, the impurity site becomes indistinguishable from the rest of the system as $`\mathrm{\Delta }_𝜹^{}=\mathrm{\Delta }_𝜹`$. The spin $`𝒔_0`$ takes the role of the original Cu spin at $`𝑹=0`$, and a homogeneous spin liquid, as described by $`H_{\text{RVB}}`$, Eq. (1), is formed. Generally, the two exchange integrals differ, $`J^{}<J`$, and translational invariance of the spinon system is broken. The bond parameter then acquires an additional spatial dependence which has to be treated self-consistently. To simplify the discussion, however, we distinguish only between bonds that do and do not connect to the impurity (see Fig. 1), respectively:
$`\mathrm{\Delta }_{ij}=\{\begin{array}{c}\mathrm{\Delta }_𝜹^{}\text{for}i=0\text{or}j=0\hfill \\ \mathrm{\Delta }_𝜹\text{for}i,j0,\hfill \end{array}`$
where $`\mathrm{\Delta }_𝜹`$ is the mean-field parameter of the impurity-free system.
The two parameters $`\mathrm{\Delta }_𝜹^{}`$ and $`\mathrm{\Delta }_𝜹`$ are assumed to exhibit the same phase relation, but in general their amplitudes differ. As a result, spinons scatter on the impurity bonds. To study this effect we write the spinon part of Hamiltonian (4) as
$$H_{\text{Ni}}^{\text{sp}}=H_{\text{RVB}}+(1x)\underset{𝜹\sigma }{}\left(\mathrm{\Delta }_𝜹f_{0\sigma }^{}f_{𝜹\sigma }+\text{H.c.}\right),$$
(7)
where $`H_{\text{RVB}}`$ represents the impurity-free system. The scattering amplitude $`(1x)`$ with $`x=|\mathrm{\Delta }_𝜹^{}/\mathrm{\Delta }_𝜹|`$ is controlled by the ratio of $`J^{}`$ to $`J`$. It vanishes for $`J^{}=J`$, and has to be treated self-consistently for $`J^{}<J`$. Approximately, we find $`x=J^{}/J`$. At this point, we introduce spinon propagators $`g_\lambda ^{(0)}(i\omega )=T_\tau f_\lambda (\tau )f_\lambda ^{}(0)_{i\omega }^0=(i\omega \xi _\lambda )^1`$ and $`g_{\lambda \lambda ^{}}(i\omega )=T_\tau f_\lambda (\tau )f_\lambda ^{}^{}(0)_{i\omega }`$ for the pure and impurity-doped system. These can be related by a scattering matrix $`T_{\lambda \lambda ^{}}(i\omega )`$:
$$g_{\lambda \lambda ^{}}(i\omega )=g_\lambda ^{(0)}(i\omega )\delta _{\lambda \lambda ^{}}+g_\lambda ^{(0)}(i\omega )T_{\lambda \lambda ^{}}(i\omega )g_\lambda ^{}^{(0)}(i\omega ).$$
(8)
A simplified notation $`\lambda =(𝒌,\nu )`$ and Matsubara frequencies $`i\omega =i(2n+1)\pi T`$, where $`T`$ denotes temperature and $`n`$ integer numbers, are employed. The $`T_{\lambda \lambda ^{}}`$ matrix in Eq. (8) describes scattering of spinons on the four bonds that connect the impurity site to its nearest neighbors. We find it to be given by the expression
$$T_{\lambda \lambda ^{}}(i\omega )=\frac{t_{\lambda \lambda ^{}}(i\omega )}{i\omega G^{(0)}(i\omega )+p^2},$$
(9)
with
$`t_{\lambda \lambda ^{}}(i\omega )`$ $`=`$ $`{\displaystyle \frac{1x}{1+x}}G^{(0)}(i\omega )(i\omega \xi _\lambda )(i\omega \xi _\lambda ^{})`$
$`+{\displaystyle \frac{x}{1+x}}(2i\omega \xi _\lambda \xi _\lambda ^{})i\omega .`$
Here $`G^{(0)}(i\omega )=_\lambda g_\lambda ^{(0)}(i\omega )=(2i\omega /D^2)\mathrm{ln}(D/|\omega |)`$, and $`p^2=x^2/(1x^2)`$. The important point is that in the flux phase the scattering matrix of Eq. (9) has two poles that are determined by the roots of
$$\omega G^{(0)}(i\omega \omega +i0^+)+p^2=0.$$
(10)
One of the poles lies below the spinon chemical potential which signals the formation of a spinon bound state. This can be interpreted as follows: Due to impurity substitution, one Cu spin loses its RVB singlet partner. In a spin gap system in which short-range spin-singlet correlations dominate, this unpaired spin does not dissolve into the RVB ground state but rather forms a local moment distributed over Cu sites in the proximity of the impurity. At finite coupling $`J^{}`$ this moment forms a local singlet with the impurity-site spinon $`f_{0\sigma }`$. The characteristic binding energy $`\omega _K`$ and lifetime $`\delta _K`$ of the resulting bound state are given by the real and imaginary part of the pole, respectively. For $`J^{}J`$, one obtains
$$\omega _K=\frac{\pi }{4}\frac{J^{}}{\mathrm{ln}D/J^{}},\delta _K=\frac{\pi }{4}\frac{\omega _K}{\mathrm{ln}D/\omega _K}.$$
(11)
In the following, two different energy scales are distinguished: $`\omega <\omega _K`$ and $`\omega >\omega _K`$. These control the physics at large and short distance from the impurity as compared to $`R_K=D/\omega _K`$, respectively, where $`R_K`$ is measured in units of lattice spacing.
First we analyze the low-energy fixed point of the system with a magnetic impurity for which $`J^{}`$ and hence $`\omega _K`$ are finite. It is determined by the regime $`\omega <\omega _K`$ and applies to distances $`R>R_K`$ from the impurity site. We calculate the impurity contribution $`\delta \rho (\omega )`$ to the density of states from the Green’s function $`\delta G(i\omega )=_{\lambda \lambda ^{}}g_\lambda ^{(0)}(i\omega )T_{\lambda \lambda ^{}}(i\omega )g_\lambda ^{}^{(0)}(i\omega )=(/i\omega )\mathrm{ln}[i\omega G^{(0)}(i\omega )+p^2]`$. For $`\omega D`$, the latter is
$$\delta G(i\omega )=\frac{2G^{(0)}(i\omega )}{i\omega G^{(0)}(i\omega )+p^2},$$
(12)
which yields
$$\delta \rho (\omega )=\frac{2}{\pi }\omega _K\delta _K\frac{|\omega |}{(\omega ^2\omega _K^2)^2+(2\omega _K\delta _K)^2}.$$
(13)
Figure 2(a) schematically shows the spinon density
of states $`\rho ^{(0)}(\omega )`$ and $`\rho (\omega )=\rho ^{(0)}(\omega )+\delta \rho (\omega )`$ for the pure and impurity-doped system. The very existence of a magnetic pseudogap is found to be unaffected by the presence of the impurity – $`\rho ^{(0)}(\omega )`$ as well as $`\rho (\omega )`$ vanish linearly in the limit $`\omega 0`$. As a consequence, the static spin susceptibility, which is related to the spinon density of states by
$$\chi (T)=\frac{1}{4T}_{\mathrm{}}^{\mathrm{}}𝑑x\frac{\rho (x)}{\mathrm{cosh}^2(x/2T)},$$
(14)
vanishes as $`T`$ at low temperatures. This indicates that in the low-energy limit all spins (except $`𝑺_{\text{eff}}`$ which is not part of the spinon system) participate in the formation of singlets. The spinon bound state discussed above hence partially screens the impurity spin by forming a Kondo singlet with $`𝒔_0`$. An effective $`S=\frac{1}{2}`$ impurity spin $`𝑺_{\text{eff}}`$ remains. In this underscreened Kondo problem, the spinon binding energy $`\omega _K`$ of Eq. (11) plays the role of the Kondo temperature: $`T_K=\omega _K`$. For temperatures $`TT_K`$, the susceptibility associated with the spinon bound state is that of a free spin $`\frac{1}{2}`$, i.e., $`\chi (T)=1/(4T)`$; simultaniously, the original $`S=1`$ impurity spin is recovered. We note that the Kondo temperature exhibits an unconventional power-law dependence on the coupling parameter $`J^{}`$, contrasting the conventional exponential behavior. This peculiarity is ascribed to the fact that the impurity spin couples to bound spinons which are predomenantly in localized rather than bandlike states. Finally, we shortly discuss how the presence of a Kondo singlet affects the properties of the spinon system at $`TT_K`$. Although the impurity does not fill the magnetic pseudogap, it nevertheless renormalizes its slope. The leading term in a low-energy expansion of Eq. (13) is related to the density of states of the pure system by
$$\delta \rho (\omega )=\frac{1}{p^2}\rho ^{(0)}(\omega ),$$
(15)
valid for $`J^{}J`$. At low energy and large distance from the impurity, the spinon system hence behaves qualitatively as in the impurity-free case.
To finish the discussion of magnetic moments, we turn to the case of a nonmagnetic impurity. The relevant physics is modelled by decoupling the spinon sector from the impurity site, setting $`J^{}=0`$, and by discarding contributions stemming from the impurity spin which is now free. Since $`\omega _K`$ consequently vanishes, one is always in the regime $`\omega >\omega _K`$. A Kondo singlet cannot form even in the zero-energy limit as the impurity carries no inherent spin. The spinon bound state induced by the impurity lies at the spinon chemical potential in the center of the pseudogap \[see Fig. 2(b)\]:
$$\delta \rho (\omega )=\delta (\omega ).$$
(16)
This is associated with the magnetic susceptibility $`1/(4T)`$ of a free spin $`\frac{1}{2}`$ which holds down to zero temperature. We note that the impurity-induced moment is broadly distributed over planar Cu sites on sublattice $`B`$ that does not contain the impurity site, its density falling off as $`R^2`$ with distance from the impurity.
To summarize, magnetic Ni and nonmagnetic Zn impurities are both associated with $`S=\frac{1}{2}`$ magnetic moments. These are, however, of very different natures (see Fig. 3):
In the former case, the spinon bound state partially screens the original $`S=1`$ impurity spin. One is left with an effective impurity spin $`\frac{1}{2}`$ ferromagnetically coupled to an ensemble of inherent spinons that, in the absence of a magnetic field, behaves qualitatively the same as an impurity-free system. In the latter case, the moment is carried by the spinon bound state itself, and is broadly distributed over Cu sites.
## IV Spin Polarization
The effective impurity moments discussed in Sec. III can be polarized by applying an external magnetic field. In this section we analyze the incidental local response of planar Cu spins. In the case of a magnetic impurity, the applied field acts on a localized impurity spin $`\frac{1}{2}`$ ferromagnetically coupled to the spin liquid. Cu spins respond via a RKKY-type interaction. In the case of a nonmagnetic impurity, the moment itself resides on Cu sites. Applying a magnetic field therefore directly polarizes the Cu spins.
We first discuss the situation of a magnetic impurity. The static polarizability is defined by $`K_{\text{Ni}}(T,𝑹)=T_\tau s_𝑹^z(\tau )S_{\text{eff}}^z(0)_{\omega =0}`$, where $`s_𝑹^z`$ and $`S_{\text{eff}}^z`$ denote the $`z`$ components of a given Cu spin at site $`𝑹`$ and of the effective impurity spin, respectively. It is expressed in terms of Green’s functions as (see Fig. 4)
$$K_{\text{Ni}}(T,𝑹)=T^2\underset{\epsilon ,\epsilon ^{}}{}\mathrm{\Pi }_d(i\epsilon ^{})J_c(i\epsilon i\epsilon ^{})\mathrm{\Pi }_f(i\epsilon ,𝑹)$$
(17)
with particle-hole convolution functions
$`\mathrm{\Pi }_d(i\epsilon )`$ $`=`$ $`D^2(i\epsilon ),`$
$`\mathrm{\Pi }_f(i\epsilon ,𝑹)`$ $`=`$ $`G(i\epsilon ,𝑹)G(i\epsilon ,𝑹).`$
Here the impurity Green’s function is $`D(i\omega )=T_\tau d_\sigma (\tau )d_\sigma ^{}(0)_{i\omega }=1/(i\omega )`$, and the intersite spinon Green’s function $`G(i\omega ,𝑹)=T_\tau f_{0\sigma }(\tau )f_{𝑹\sigma }^{}(0)_{i\omega }`$. Operators $`d`$ and $`f`$ act on separated sectors of the Hilbert space. At site $`𝑹=0`$, however, $`f`$ spinons are polarized by the local spin $`𝑺_{\text{eff}}`$ due to the ferromagnetic interaction of bare strength $`J_c`$. This coupling is accounted for by the vertex function $`J_c(i\omega )`$. Employing a ladder approximation it is
$$J_c(i\omega )=\frac{J_c}{1+J_c\mathrm{\Pi }_c^{i\omega }}=\frac{1}{\mathrm{\Pi }_c^{i\omega }},$$
(18)
with
$`\mathrm{\Pi }_c^{i\omega }=T{\displaystyle \underset{\epsilon }{}}D(i\epsilon +i\omega )G(i\epsilon ,𝑹=0).`$
The second equality in Eq. (18) holds due to $`J_c`$ being infinitely large. Replacing the vertex function by its zero-frequency limit, $`J(i\omega )J(0)`$, the polarizability of Eq. (17) can be factorized. Within this approximation, which is valid at low temperatures, one obtains
$$K_{\text{Ni}}(T,𝑹)=\chi _{\text{eff}}(T)J_c(0)\chi _{\text{pl}}(T,𝑹).$$
(19)
The polarizability has thus been decomposed into the magnetic susceptibility of the effectively free $`\frac{1}{2}`$ impurity spin, $`\chi _{\text{eff}}(T)=1/(4T)`$, the nonlocal magnetic susceptibility of CuO<sub>2</sub> planes, $`\chi _{\text{pl}}(T,𝑹)`$, and an effective coupling parameter $`J_c(0)`$. The susceptibilities are defined as $`\chi _{\text{eff}}(T)=T_\tau S_{\text{eff}}^z(\tau )S_{\text{eff}}^z(0)_{\omega =0}=T_\epsilon \mathrm{\Pi }_d(i\epsilon )`$ and $`\chi _{\text{pl}}(T,𝑹)=T_\tau s_𝑹^z(\tau )s_0^z(0)_{\omega =0}=T_\epsilon \mathrm{\Pi }_f(i\epsilon ,𝑹)`$.
To further analyze the polarizability in Eq. (19), $`J_c(0)`$ and $`\chi _{\text{pl}}(𝑹)`$ have to be evaluated. This requires the on-site and intersite spinon Green’s functions
$$G(i\omega ,𝑹)=\{\begin{array}{c}\left(\frac{p}{x}\right)^2\frac{G^{(0)}(i\omega )}{i\omega G^{(0)}(i\omega )+p^2}\text{for}R=0\hfill \\ \frac{1}{x}G^{(0)}(i\omega ,𝑹)\text{for}R>R_K,\hfill \end{array}$$
(20)
where $`G^{(0)}(i\omega ,𝑹)`$ is defined for the impurity-free system
$$G^{(0)}(i\omega ,𝑹)=\frac{2i|\omega |}{D^2}\phi (𝑹)K_1\left(\frac{R|\omega |}{D}\right),$$
(21)
with a modified Bessel function of the second kind, $`K_\nu (x)`$. Equation (21) holds for sites on sublattice $`B`$; contributions from sublattice $`A`$ containing the impurity are found to be negligible. The angular dependence is determined by the phase factor
$$\phi (𝑹)=\frac{1}{2}\left(\stackrel{~}{R}^+e^{i\pi R^+/2}+\stackrel{~}{R}^{}e^{i\pi R^{}/2}\right)$$
(22)
with $`R^\pm =R_x\pm R_y`$ and $`\stackrel{~}{R}^\pm =(R_x\pm iR_y)/R`$. We are now in the position to calculate the effective coupling parameter from the zero-frequency limit of Eq. (18),
$$J_c(0)=\{\begin{array}{c}D\text{for}J^{}=J\hfill \\ 2\omega _K\text{for}J^{}J,\hfill \end{array}$$
(23)
and the nonlocal spin susceptibility of CuO<sub>2</sub> planes in the presence of the impurity,
$$\chi _{\text{pl}}(𝑹)=\frac{3}{4\pi }\frac{1}{Jx^2}\frac{\mathrm{\Phi }(𝑹)}{R^3},$$
(24)
the latter being valid for $`𝑹B`$ with $`R>R_K`$. The phase factor in Eq. (24) is defined by $`\mathrm{\Phi }(𝑹)=|\phi (𝑹)|^2`$. Finally, combining these results, we obtain
$$K_{\text{Ni}}(T,𝑹)=\frac{3}{16\pi }\frac{J_c(0)}{Jx^2}\frac{\mathrm{\Phi }(𝑹)}{R^3}\frac{1}{T},$$
(25)
which describes the polarizability of a Cu spin at site $`𝑹B`$ responding to a magnetic field that acts on the effective impurity spin $`𝑺_{\text{eff}}`$; contributions from sublattice $`A`$ are found to be small. We note that the $`T^1`$ Curie behavior displayed by Eq. (25) stems solely from the susceptibility $`\chi _{\text{eff}}(T)`$ of the effective impurity spin. Within the present mean-field treatment, the planar susceptibility is independent of temperature: $`\chi _{\text{pl}}(T,𝑹)=\chi _{\text{pl}}(𝑹)`$.
We now briefly review the result for a nonmagnetic impurity which was derived in Ref. . Here, the Cu spins carry the impurity-induced moment, and can therefore be directly polarized by the magnetic field. The polarizability is given by the local susceptibility of the impurity-induced moment, $`K_{\text{Zn}}(T,𝑹)=\delta \chi _{\text{pl}}(T,𝑹)=_𝑹^{}\left(T_\tau s_𝑹^z(\tau )s_𝑹^{}^z(0)_{\omega =0}T_\tau s_𝑹^z(\tau )s_𝑹^{}^z(0)_{\omega =0}^0\right)`$, yielding
$$K_{\text{Zn}}(T,𝑹)=\frac{1}{2\pi }\frac{\mathrm{\Phi }(𝑹)}{R^2}\frac{1}{T\mathrm{ln}D/T}.$$
(26)
Equation (26) is valid for $`𝑹B`$, while contributions from $`𝑹A`$ are again negligible. The polarizability is found to decay slowly as $`R^2`$ with distance from the impurity which compares to a $`R^3`$ behavior in the case of Ni, reflecting the delocalized nature of the moment induced by a Zn impurity. Further, a logarithmic correction to the Curie-like temperature behavior is to be marked.
In deriving Eqs. (25) and (26) for the polarizability of Cu spins, we have, up to this point, built upon RVB mean-field theory. This picture accounts well for the spin liquid features of underdoped cuprates including the presence of a magnetic pseudogap. Its strength lies on the description of long-range properties controlled by low-energy excitations. The mean-field treatment does, however, severely underestimate local AF correlations which reflect the proximity of a critical instability towards AF spin ordering. As a consequence, the above expressions contain no reference to the AF correlation length which was suggested to introduce a temperature dependence beyond the Curie behavior of free moments. Furthermore, mean-field theory yields a polarizability of Cu spins on one sublattice only, undervaluing the staggered magnetization of spins on the opposite sublattice. This is in disaccord with NMR measurements that yield no overall shift of the <sup>17</sup>O line, as would be expected from the polarization of only one sublattice as well as with numerical studies.
To compensate for these deficiencies of the mean-field treatment, we simulate the closeness of the spin system towards an antiferromagnetically ordered state by performing a random-phase approximation (RPA) in the magnetic susceptibility. In the momentum representation, the susceptibility of planar Cu spins then becomes
$$\chi _{\text{pl}}^{\text{RPA}}(T,𝒒)=\chi _{\text{pl}}(𝒒)S(T,𝒒)$$
(27)
with the Stoner enhancement factor
$$S(T,𝒒)=\frac{1}{1+J_𝒒\chi _{\text{pl}}(T,𝒒)},$$
(28)
where $`J_𝒒=2J(\mathrm{cos}q_x+\mathrm{cos}q_y)`$. We closely follow the theory of a nearly AF Fermi liquid, which maps Eq. (27) onto a phenomenological expression involving the AF correlation length $`\xi (T)`$. Within this picture, $`\chi _{\text{pl}}^{\text{RPA}}(T,𝒒)`$ is assumed to be controlled solely by the momentum region close to the AF wave vector $`𝑸=(\pi ,\pi )`$. However, we do take a slightly different point of view in this respect: The momentum dependence of the bare susceptibility $`\chi _{\text{pl}}(𝒒)`$ in Eq. (27), which describes the long-range characteristics of spin correlations in the presence of a magnetic pseudogap, is explicitly kept. Only the scaling function $`S(T,𝒒)`$, which controls short-range AF correlations, is expanded around $`𝑸=(\pi ,\pi )`$. Identifying $`J\chi _{\text{pl}}(T,𝑸)/[14J\chi _{\text{pl}}(T,𝑸)]=\xi ^2(T)`$ and $`1/[J\chi _{\text{pl}}(T,𝑸)]=\alpha `$, Eq. (28) can be written in phenomenological form
$$S(T,𝒒)=\frac{\alpha \xi ^2(T)}{1+(𝒒𝑸)^2\xi ^2(T)},$$
(29)
where $`\alpha 1`$ on a mean-field level. We note that the explicit form of $`\xi (T)`$ lies beyond the accessibility of a mean-field treatment, and has to be chosen according to general physical considerations.
Turning back to real space, the nonlocal susceptibility is
$$\chi _{\text{pl}}^{\text{RPA}}(T,𝑹)=\underset{𝑹^{}B}{}\chi _{\text{pl}}(𝑹^{})S(T,𝑹𝑹^{}).$$
(30)
For distances $`R\xi (T)`$, it can be approximated by $`\chi _{\text{pl}}^{\text{RPA}}(T,𝑹)=\chi _{\text{pl}}(𝑹)\xi ^2(T)/2`$ with interpolation formula for the $`A`$ sublattice $`\chi _{\text{pl}}(𝑹A)=(1/z)_𝜹\chi _{\text{pl}}(𝑹+𝜹)`$. Analogous expressions are obtained for the local susceptibility $`\delta \chi _{\text{pl}}(T,𝑹)`$ induced by a nonmagnetic impurity. Combining these results with Eqs. (25) and (26) and performing an angular average over phase factors $`\mathrm{\Phi }(𝑹)`$, one finally arrives at the following expressions for the polarizability of Cu spins in the impurity-doped system:
$`K_{\text{Ni}}(T,𝑹)`$ $`=`$ $`\mathrm{cos}(𝑸𝑹){\displaystyle \frac{3}{64\pi }}{\displaystyle \frac{J_c(0)}{Jx^2}}{\displaystyle \frac{1}{R^3}}{\displaystyle \frac{\xi ^2(T)}{T}},`$ (31)
$`K_{\text{Zn}}(T,𝑹)`$ $`=`$ $`\mathrm{cos}(𝑸𝑹){\displaystyle \frac{1}{8\pi }}{\displaystyle \frac{1}{R^2}}{\displaystyle \frac{\xi ^2(T)}{T\mathrm{ln}D/T}}.`$ (32)
These equations now hold for both sublattices, $`𝑹\{A,B\}`$, the staggered nature of spin correlations being manifested in the alternating sign implied by $`\mathrm{cos}(𝑸𝑹)`$. Further, the dependence upon the AF correlation length $`\xi (T)`$ is now explicitly accounted for.
## V NMR Line Broadening
The impurity-induced polarization of Cu spins in a magnetic field affects the energy levels of nuclear spins via supertransferred hyperfine interaction. The coupling of a given nuclear spin $`𝑰`$ to electron spins $`𝒔_i`$ on close by Cu sites is described by
$$H_{\text{hf}}=\gamma _n\gamma _eC_{\text{hf}}\underset{i}{}𝒔_i𝑰,$$
(33)
where $`\gamma _n`$ and $`\gamma _e`$ denote the nuclear and electron gyromagnetic ratios, respectively, and $`C_{\text{hf}}`$ is the supertransferred hyperfine coupling constant. In the following, we restrict ourselves to NMR measurements on <sup>17</sup>O nuclei ($`I=\frac{5}{2}`$). On a mean-field level, $`𝒔_i`$ can be replaced by its average value $`𝒔_i=K(T,𝑹_i)𝑯_0`$ with external magnetic field $`𝑯_0`$ and polarizability $`K(T,𝑹_i)`$ given by either one of Eqs. (31) and (32) for the two types of impurities. Since each <sup>17</sup>O nucleus lies symmetrically in between two Cu sites that belong to different sublattices with spins polarized in opposite directions, the impurity-induced energy shift partially cancels (see Fig. 5).
At large enough distance from the impurity, the shift is then effectively determined by the spatial derivative of the polarizability,
$$\omega (𝑹)=\kappa \frac{|K(T,𝑹)|}{R}\mathrm{cos}\varphi ,$$
(34)
where $`\varphi `$ denotes the angle enclosed by $`𝑹`$ and the $`x`$ or $`y`$ axis and $`\kappa =\gamma _n\gamma _eC_{\text{hf}}H_0`$.
In a system with randomly distributed impurities of concentration $`c`$, the superposition of energy shifts induced by different impurities leads to a broadening of the NMR line. We calculate the line shape that follows from Eq. (34), employing the formalism of Ref. . The line shape function $`g(\nu )`$ is defined as the Fourier transform of the characteristic or free-induction function
$$f(t)=\mathrm{exp}\left[c\underset{𝑹}{}\left(1e^{i\omega (𝑹)t}\right)\right].$$
(35)
Integrating over lattice sites, for Ni and Zn, respectively, one obtains
$$\mathrm{ln}f(t)=\{\begin{array}{c}\mathrm{\Lambda }_{\text{Ni}}|t|^{1/2}\hfill \\ \mathrm{\Lambda }_{\text{Zn}}|t|^{2/3},\hfill \end{array}$$
(36)
with
$`\mathrm{\Lambda }_{\text{Ni}}`$ $`=`$ $`{\displaystyle \frac{2\sqrt{6}\pi \mathrm{\Gamma }(3/4)}{\mathrm{\Gamma }(1/4)}}\left[\kappa {\displaystyle \frac{3}{64\pi }}{\displaystyle \frac{J_c(0)}{Jx^2}}{\displaystyle \frac{\xi ^2(T)}{T}}\right]^{1/2}c,`$
$`\mathrm{\Lambda }_{\text{Zn}}`$ $`=`$ $`{\displaystyle \frac{2\sqrt{3}\pi ^2}{\mathrm{\Gamma }^2(1/3)}}\left[\kappa {\displaystyle \frac{1}{8\pi }}{\displaystyle \frac{\xi ^2(T)}{T\mathrm{ln}D/T}}\right]^{2/3}c.`$
Figure 6 shows the different line shapes induced by Ni and Zn impurities as obtained by performing a Fourier transformation on $`f(t)`$.
Comparing to the Lorentzian shape that results from $`\mathrm{ln}f(t)|t|`$ in conventional RKKY theory, one finds a marked difference in both shape and width. Using the numerical values shown in the inset of Fig. 6, we finally arrive at the following expressions for the full linewidth induced by magnetic and nonmagnetic impurities:
$`\mathrm{\Delta }\nu _{\text{Ni}}`$ $`=`$ $`2\times 0.22\left(\mathrm{\Lambda }_{\text{Ni}}\right)^2,`$ (37)
$`\mathrm{\Delta }\nu _{\text{Zn}}`$ $`=`$ $`2\times 0.51\left(\mathrm{\Lambda }_{\text{Zn}}\right)^{2/3}.`$ (38)
## VI Comparison with Experiment
In this section, we compare the impurity-induced <sup>17</sup>O NMR line broadening as described by Eqs. (37) and (38) with experimental data of Bobroff et al. obtained on $`\text{YBa}_2(\text{Cu}_{1x}M_x)_3\text{O}_{6.6}`$ with $`M`$ = Zn or Ni.
The following constants are chosen: The superexchange parameters are specified by $`J=0.13`$ eV for Cu-Cu interaction and $`J^{}=J/2`$ for Cu-Ni. A self-consistent treatment yields a scattering amplitude $`(1x)=0.5`$, where $`\mathrm{\Delta }=\frac{1}{4}`$ has been used. The Kondo temperature is obtained by numerically solving Eq. (10) which gives $`T_K=560`$ K. Below this temperature, the Ni spin is partially screened and behaves as a spin $`\frac{1}{2}`$ ferromagnetically coupled to the CuO<sub>2</sub> plane. The effective coupling constant of this interaction given by Eq. (23) is $`J_c(0)=0.1`$ eV. The hyperfine coupling constant between <sup>17</sup>O nuclear and Cu electron spins is $`C_{\text{hf}}=3.3`$ T/$`\mu _B`$. The magnetic-field strength used in the experiment is $`H_0=7.5`$ T, and the concentration of Ni and Zn impurities is $`1\%`$. The effective impurity concentration within CuO<sub>2</sub> planes, which is larger by a factor of $`\frac{3}{2}`$, is finally $`c=1.5\%`$.
Next, an expression for the AF correlation length $`\xi (T)`$ has to be specified. It is argued in Ref. that below a critical temperature $`T_{\text{cr}}`$ specified by $`\xi (T_{\text{cr}})2`$, the correlation length assumes the form
$$\xi (T)=\frac{1}{a+bT},$$
(39)
where $`a`$ and $`b`$ are fitting constants of the theory. Saturation of $`\xi (T)`$ at low temperatures is neglected here.
Figure 7 shows the impurity-induced line broadening $`\mathrm{\Delta }\nu _{\text{imp}}`$ scaled with temperature.
The curves are fitted to the experimental data by setting $`a=0.07`$ and $`b=0.0007`$, which correspond to an AF correlation length of $`\xi =4.8`$ in units of lattice spacings at $`T=200`$ K. This compares well to $`\xi =5.9`$ obtained in Ref. . No further fitting parameters are needed.
The theory correctly accounts for the peculiar experimental observation of Zn having a more pronounced effect on the NMR signal than Ni – this seems to be in contradiction to SQUID measurements on the macroscopic susceptibility. We are able to ascribe this behavior to the different spatial dependence of the polarizability: $`K(T,𝑹)`$ decays as $`R^3`$ in the case of Ni, but only as $`R^2`$ in the case of Zn. Averaging over all impurity site, this leads to an enhanced line-broadening effect of Zn (see Fig. 6). Our theory further correctly describes the anomalous non-Curie temperature dependence exhibited by the NMR linewidth – this seems to be in disaccord with an almost perfect $`T^1`$ behavior exhibited by the macroscopic susceptibility. One can resolve this disagreement by assuming a temperature dependence of the AF correlation length $`\xi (T)`$ which enters the polarizability of Cu spins. Good agreement with experiment is obtained by employing $`\xi (T)`$ of the form given in Eq. (39).
## VII Conclusion
In summary, we have studied local moments induced in underdoped cuprates by doping with magnetic ($`S=1`$) Ni and nonmagnetic Zn impurities. In the presence of a spin gap, both types of impurities are associated with $`S=\frac{1}{2}`$ magnetic moments in the CuO<sub>2</sub> planes. These are, however, of very different natures. Ni as well as Zn disturb the spin liquid formed by planar Cu spins, resulting in a magnetic moment residing on Cu sites in the proximity of the impurity. In the case of Ni, this moment partially shields the impurity spin below a critical temperature $`T_K`$ in what resembles an underscreened Kondo model; an effective impurity spin $`\frac{1}{2}`$ results. Since predominantly localized rather than bandlike states are involved in the screening of the impurity spin, the Kondo temperature exhibits an unconventional power-law dependence on the coupling constant. In the case of Zn, on the other hand, one deals with a $`S=\frac{1}{2}`$ moment broadly distributed over Cu sites. We have further investigated the RKKY-type response of Cu spins in a magnetic field. The spin polarization is found to decay as $`R^3`$ with distance from a Ni impurity, but only as $`R^2`$ in the case of Zn. This different behavior reflects the delocalized character of Zn moments, and explains why Zn has a stronger impact on the NMR linewidth than Ni. Further, accounting for the presence of temperature-dependent AF correlations in underdoped cuprates, we can successfully describe the non-Curie behavior of the impurity effect on the NMR linewidth. In general, it can be concluded that the anomalous impurity properties of underdoped cuprates are a clear manifestation of the peculiar mixture of spin-singlet and antiferromagnetic correlations present in these compounds.
|
no-problem/9904/astro-ph9904349.html
|
ar5iv
|
text
|
# A BeppoSAX observation of the massive X-ray binary 4U 1700-37
## 1 Introduction
4U 1700-37 is a massive eclipsing X-ray binary consisting of a compact accreting object embedded in the wind of HD153919, a supergiant O star (Jones et al. 1973). Orbiting every 3.4 days, the compact object has an X-ray spectrum reminiscent of an accreting highly magnetized neutron star system (Haberl et al. 1989), although no X-ray pulsations have ever been confirmed. The X-ray flux is significantly less than expected on the basis of standard wind accretion theory.
The most recent estimates of the mass of HD 153919 and its companion are from Heap & Corcoran (1992), and Rubin et al. (1996). Heap & Corcoran (1992) propose a mass for HD 153919 of $`50\pm 2\mathrm{M}_{}`$, with a companion mass of $`1.8\pm 0.4\mathrm{M}_{}`$. Rubin et al. (1996) reanalyzed the system parameters using Monte Carlo methods, finding masses of 30$`{}_{7}{}^{}{}_{}{}^{+11}`$ M and 2.6 $`{}_{1.4}{}^{}{}_{}{}^{+2.3}`$ M for HD 153919 and its companion, respectively. Both estimates are, at their lower ranges, consistent with the compact object being a neutron star, but a black hole cannot be excluded.
Since pulsations have never been detected, further evidence for the object’s nature must be sought from its X-ray spectrum. In general, X-ray pulsar spectra require an exponential cutoff to the continuum between $``$10–20 keV, unlike black hole spectra, which require either a single power-law or an ultra-soft component plus hard power-law tail (e.g., Tanaka & Lewin 1995). Additionally, about half of the sample of 23 X-ray pulsars observed by Ginga show cyclotron features in their spectra, usually at energies between 10–60 keV (Makishima & Mihara 1992; Mihara 1995), indicating the presence of strong magnetic fields.
## 2 Previous X-ray spectroscopy of 4U 1700-37
The X-ray spectrum of 4U 1700-37 has been studied using many different satellites, including HEAO-1 and Einstein (White et al. 1983), and EXOSAT (Haberl et al. 1989). The spectral shape is highly reminiscent of other X-ray pulsars (White et al. 1983), being well described by a power-law with photon index, $`\alpha `$, of $``$0.15, modified by a high-energy cutoff above 6–7 keV. Haberl & Day (1992) confirmed the above results with Ginga observations, also finding that $`\alpha `$ varies between $``$0.5 and 1.0. Haberl et al. (1994) show that the 0.1–2.4 keV ROSAT Position Sensitive Proportional Counter (PSPC) spectrum is consistent with a thermal bremsstrahlung model with a temperature, kT, fixed at 0.74 keV, as originally found by Ginga (Haberl & Day 1992), plus a hydrogen column density of $`3.6\times 10^{22}`$ atom cm<sup>-2</sup>. Haberl & Day (1992) also found that the best-fit values of $`\alpha `$ varied depending on the photoelectric column density. Since this is physically unrealistic, the authors show that the spectral variations could also be modeled using a two-component scattering model, consisting of two power-laws with the same slopes but variable absorption, one of which represents a less absorbed scattered component.
The spectrum above 20 keV was examined using BATSE data by Rubin et al. (1996, and references therein) who confirmed earlier findings that it can be represented by a thermal bremsstrahlung model with kT $``$25 keV, out to 120 keV. There have been no reported changes in the shape of the high energy spectrum due to orbital phase or source intensity.
While the above observations are individually useful, the launch of BeppoSAX, with its very wide spectral bandpass, offers the first opportunity to study the entire X-ray spectrum of 4U 1700-37 simultaneously. BeppoSAX is proving to be a highly capable mission for discovering and studying cyclotron lines (e.g., Dal Fiume et al. 1998). For the first time, therefore, we present a single analysis of the entire spectrum, obtained at one epoch, spanning 0.5–200 keV.
## 3 Observations
The X-ray astronomy satellite BeppoSAX (Boella et al. 1997a) contains four coaligned Narrow Field Instruments, or NFI. Results from the Low-Energy Concentrator Spectrometer (LECS; 0.1–10 keV; Parmar et al. 1997), Medium-Energy Concentrator Spectrometer (MECS; 1.3–10 keV; Boella et al. 1997b), High Pressure Gas Scintillation Proportional Counter (HPGSPC; 5–120 keV; Manzo et al. 1997) and the Phoswich Detection System (PDS; 15–300 keV; Frontera et al. 1997) are presented here. The MECS consisted of three identical grazing incidence telescopes with imaging gas scintillation proportional counters in their focal planes. The LECS uses an identical concentrator system as the MECS, but utilizes an ultra-thin entrance window and a driftless configuration to extend the low-energy response to 0.1 keV. The non-imaging HPGSPC consists of a single unit with a collimator that is alternatively rocked on- and off-source. The non-imaging PDS consists of four independent units arranged in pairs each having a separate collimator. Each collimator can be alternatively rocked on- and off-source.
4U 1700-37 was observed by BeppoSAX between 1997 April 1 11:21 and 22:15 UTC. This interval corresponds to orbital phases 0.44–0.58, where mid-eclipse of the X-ray source occurs at phase 0.0, using the ephemeris of Rubin et al. (1996). Good data were selected from intervals when the elevation angle above the Earth’s limb was $`>`$$`4^{}`$ and when the instrument configurations were nominal, using the SAXDAS 1.3.0 data analysis package. The standard collimator dwell time of 96 s for each on- and off-source position was used, together with rocking angles of 180′ and 210′ for the HPGSPC and PDS, respectively. The exposures in the LECS, MECS, HPGSPC, and PDS instruments are 12.2 ks, 23.7 ks, 11.2 ks, and 10.7 ks, respectively. LECS and MECS data were extracted centered on the position of 4U 1700-37 using radii of 8′ and 4′, respectively. Background subtraction in the imaging instruments was performed using standard files, but is not critical for such a bright source. Background subtraction in the non-imaging instruments was carried out using data from the offset intervals. The background subtracted count rates in the LECS, MECS, HPGSPC and PDS were 5.3, 18.5, 40.5, and 27.0 s<sup>-1</sup>, respectively.
### 3.1 Spectral fits
The 4U 1700-37 spectrum was investigated by simultaneously fitting data from all the NFI. The LECS and MECS spectra were rebinned to oversample the full width half maximum of the energy resolution by a factor 3 and to have additionally a minimum of 20 counts per bin to allow use of the $`\chi ^2`$ statistic. Data was selected in the energy ranges 0.5–5.0 keV (LECS) and 1.8–10.0 keV (MECS) where the instrument responses are well determined. The HPGSPC and PDS data were rebinned using standard procedures in the energy ranges 7–40 keV and 15–200 keV, respectively. The photoelectric absorption cross sections of Morrison & McCammon (1983) and the solar abundances of Anders & Grevesse (1989) are used throughout.
The spectrum was first fit with an absorbed power-law model, including an iron line, but with no cutoff or soft component. The iron line energy, $`\mathrm{E}_{\mathrm{Fe}}`$, was fixed at 6.5 keV, but the width, $`\sigma _{\mathrm{Fe}}`$, and normalization (and all other parameters) were allowed to vary. Unsurprisingly, given the simplicity of the model, only a very poor fit with a reduced $`\chi ^2`$ of 47.0 for 280 degrees of freedom (dof) was obtained. Matters are significantly improved if a high energy cutoff and a low energy thermal bremsstrahlung component are included, resulting in a reduced $`\chi ^2`$ of 2.00 for 276 dof. The best-fit thermal bremsstrahlung component (see Haberl et al. 1994) has a kT of $`0.2\pm 0.1`$ keV. Both continuum components suffer low-energy absorption, $`\mathrm{N}_\mathrm{H}`$, of $`(5.1\pm 0.2)\times 10^{20}`$ atom cm<sup>-2</sup>. The line width was consistent with a narrow line and so was fixed at a value (0.1 keV) much smaller than the instrumental resolution. This gives an equivalent width, EW, of $`120\pm 20`$ eV for the iron line. The functional form of the cutoff above energy $`\mathrm{E}_{\mathrm{cut}}`$ is $`\mathrm{exp}[(\mathrm{E}_{\mathrm{cut}}\mathrm{E})/\mathrm{E}_{\mathrm{fold}}]`$, where $`\mathrm{E}_{\mathrm{cut}}`$ and $`\mathrm{E}_{\mathrm{fold}}`$ are the cutoff and folding energies, respectively. The count rate spectrum for this model is shown in the left panels of Fig. 1, together with the contributions to $`\chi `$<sup>2</sup>.
A significant contribution to the existing poor fit quality is residual structure in the PDS spectrum at $``$37 keV (see the left hand panels of Fig. 1). Note that the same structure is not evident in the HPGSPC spectrum. Since this feature is reminiscent of a cyclotron resonance line, we added a single cyclotron component (the cyclabs model in xspec) to the model and re-fitted the data. The fit quality improves sharply, with a reduced $`\chi `$<sup>2</sup> of 1.42 for 273 dof. This improved fit is shown in the right hand panels of Fig. 1. The best-fit values of $`\mathrm{E}_{\mathrm{cut}}`$ and $`\mathrm{E}_{\mathrm{fold}}`$ are $`5.9\pm 0.2`$ keV and $`23.9\pm 0.5`$ keV, respectively. These values are consistent with those of $`6.6\pm 0.7`$ keV and $`21.1_{3.3}^{+4.3}`$ keV found by Haberl et al. (1989) using EXOSAT data. Haberl & Day (1992) determined a slightly higher cutoff energy using Ginga, measuring $`\mathrm{E}_{\mathrm{cut}}`$ = 7.6 keV, and $`\mathrm{E}_{\mathrm{fold}}`$ = 19.5 keV, with no uncertainties quoted. Nonetheless, given that these observations were made with different missions at different epochs, the broad agreement is gratifying and shows that the cutoff is a necessary component in the spectrum.
The best-fit energy, $`\mathrm{E}_{\mathrm{cyc}}`$, and width, $`\sigma _{\mathrm{cyc}}`$, of the feature are $`36.6\pm 1.0`$ keV and $`11_3^{+5}`$ keV, respectively. This width is a factor $``$2.5 broader then expected from the correlation between these parameters observed by Dal Fiume et al. (1998) for five other massive X-ray binaries. This suggests that the feature may arise through incorrect modeling of the continuum. This is supported by the lack of detection in the HPGSPC. Variations in PDS performance, which could produce a similar spectral feature if the background and source spectra had slightly different gains, are however excluded. The feature is still present if the continuum is modeled with a broken power-law together with a high-energy cutoff, as suggested by Dal Fiume et al. (1998). Following Mihara (1995) and Dal Fiume et al. (1998), we also generated a Crab ratio spectrum, by dividing the PDS data by those obtained from a PDS observation of the Crab Nebula, which is time-invariant and has a smoothly varying E<sup>-2.1</sup> continuum in this energy range. This technique is useful in minimizing instrumental and model dependent effects. While the Crab ratio spectrum indicates some structure at $``$37 keV, it does not show the pronounced deficit which is the signature of the cyclotron features discussed in Dal Fiume et al. (1998). We conclude that while the suggestion of a cyclotron line is intriguing, and results in a significantly lower value of $`\chi ^2`$, we cannot exclude the possibility that the feature is an artefact of the fitting process.
Further improvements in fit quality require a better description of the spectral region covered by the LECS, since the main remaining contribution to $`\chi ^2`$ is a feature at $``$1 keV (see Fig. 1). This feature has not been seen in broad-band spectra of 4U 1700-37 before, since the Ginga observations did not cover the energy region below 1.5 keV, but it may also be evident in the residuals of the ROSAT PSPC spectrum presented in Haberl et al. (1994). Since Rubin et al. (1996) show that the BATSE spectrum above 20 keV is consistent with a thermal bremsstrahlung model, we examined the PDS data separately to check for consistency with this earlier study. The result of Rubin et al. (1996) is confirmed, with the PDS spectrum being described by a thermal bremsstrahlung with kT = $`25.9\pm 0.3`$ keV, in agreement with the temperature of 25 keV derived from the BATSE data, for a reduced $`\chi ^2`$ of 2.16 for 46 dof. The addition of a broad cyclotron line at $``$40 keV improves the fit significantly giving a reduced $`\chi ^2`$ of 1.03 for 43 dof.
### 3.2 The nature of the compact object in 4U 1700-37
The broadband BeppoSAX spectrum presented here is very similar to the X-ray spectrum of an accreting pulsar. The underlying shape is an absorbed power-law sharply modified $`\text{ }>`$6.0 keV by a cutoff. Although the photon index of $``$1.0 is rather higher than the average reported by previous authors, it lies within the range of observed variation exhibited by 4U 1700-37, and is at the upper range of the values for various pulsars summarized in White et al. (1983).
In what respects does the 4U 1700-37 spectrum differ from, or resemble that, of a black hole? Four diagnostics of black hole candidates (none of them individually conclusive) are listed in Tanaka & Lewin (1995). These are: (1) ultrasoft spectra when luminous, (2) high-energy power-law tails, (3) bimodal spectral states, and (4) millisecond variability and flickering in the hard state. 4U 1700-37 possesses none of these attributes. In particular, steep, cutoff spectra ($`\mathrm{E}_{\mathrm{cut}}\text{ }<`$20 keV) are a feature of neutron stars rather than black holes (White et al. 1988), while black holes (at least in outburst) generally possess hard tails, emitting significant flux $`\text{ }>`$100 keV (see e.g., Ballet et al. 1994). 4U 1700-37 resembles neither the ultrasoft “high-state”, nor the hard power-law “low-state” spectrum of the canonical black hole candidate Cyg X-1.
## 4 Discussion
We have demonstrated that the broadband X-ray spectrum of 4U 1700-37 is qualitatively similar to that of other accreting neutron star X-ray pulsars. A feature at $``$37 keV may be modeled as a broad cyclotron absorption line, but its presence is uncertain due to uncertainties in continuum modeling and instrument performance. Such features have only been reported in about half of the known X-ray pulsars – and in some cases only at marginal significance. The lack of pulsations remains puzzling, and may suggest (as has been argued by previous authors) that the magnetic field is intrinsically weak or aligned with the rotation axis. The neutron star’s progenitor may have been very massive, suggesting that there may be a range of masses over which progenitors form both black holes and neutron stars, or that black holes are formed only through a limited range of progenitor masses, above which the remnants are again neutron stars.
###### Acknowledgements.
The BeppoSAX satellite is a joint Italian-Dutch programme. We thank Fabio Favata, Matteo Guainazzi, and Tim Oosterbroek for helpful discussions. We also thank the referee, Frank Haberl, for suggestions which helped improve the paper.
|
no-problem/9904/cond-mat9904338.html
|
ar5iv
|
text
|
# Coiling of Cylindrical Membrane Stacks with Anchored Polymers
\[
## Abstract
We study experimentally a coiling instability of cylindrical multilamellar stacks of phospholipid membranes, induced by polymers with hydrophobic anchors grafted along their hydrophilic backbone. We interpret our experimental results in terms of a model in which local membrane curvature and polymer concentration are coupled. The model predicts the occurrence of maximally tight coils above a threshold polymer concentration. Indeed, only maximally tight coils are observed experimentally. Our system is unique in that coils form in the absence of twist and adhesion.
PACS numbers: 87.16.Dg, 68.10.-m \]
The coil motif is ubiquitous in a wide range of natural contexts. One-dimensional filaments of mutant bacteria , supercoiled DNA molecules , and tendrils of climbing plants all exhibit a writhing instability as a result of forcing or interaction with an external agent. Such systems are dominated by elastic properties and the appearance of coils is a result of the relief of twist. In this paper we show that coiling can also be effected in cylindrical multilamellar tubes of phospholipid bilayers, by anchoring hydrophilic polymers with hydrophobic side groups grafted along the backbone. This system is unique in that, in contrast with the above examples, fluid membranes cannot support any twist. Yet coils are formed in the system, and are stable for a very long time.
Our system is representative of a wide class of systems of membranes with embedded inclusions, such as biological cells with membrane-associated proteins. Other examples include erythrocyte ghosts incorporating amphipathic drugs and liposomes with covalently attached polymers used for drug delivery (for additional examples see ). Our experiments give us a unique opportunity to study a wide variety of phenomena induced by such inclusions in a relatively simple and controlled environment.
The cylindrical multilamellar tubes, called myelin figures, consist of a large number of bilayers reaching almost to the core . Adjacent bilayers are separated by thin hydration layers. Coiling of myelin figures has already been observed in phospholipid binary mixtures when the constituents undergo a phase separation process triggered by the addition of Ca<sup>++</sup>. Coiled myelin figures of egg-yolk phosphatidylcholine (egg-pc) have also been reported . In both cases coiling was attributed to surface adhesion. The novelty of our work is that coiling occurs in the absence of both adhesion and twist. Our experiments clearly show that surface adhesion is negligible in our system.
We believe that in our system coiling results from a coupling between the polymer concentration and local membrane curvature, induced by anchoring of the polymer in the membrane. We show below that a high enough polymer density together with the constraints imposed by the cylindrical geometry of the tube destabilizes the straight tube. Coiling as a manifestation of this instability is proposed here for the first time.
The coupling between membrane curvature and polymers has been considered both theoretically and experimentally . The theoretical studies emphasize the polymer backbone and its effect on the elastic properties of the membrane. They do not consider the mobility of anchors embedded in the bilayers. We believe this effect is important in our system, and attempt to capture it in the model presented below.
In our experiments, tubular membrane stacks were made of stearoyl-oleoyl-phosphatidylcholine (SOPC), with $`C_{18}`$ alkyl chains. The polymer we used is hydrophilic dextran (MW 162,000 g/mol) functionalized both with $`C_{16}`$ alkyl chains and dodecanoic NBD chains as fluorescent markers. The hydrophobic anchors, distributed statistically along the backbone (about 1 alkyl chains for 25 glucose units) are $`C_{16}`$ long. On average there are 4 persistence lengths between consecutive anchors. Therefore, the extension of each polymer molecule on the two-dimensional membrane is much larger than its extension into the third dimension. Samples were prepared by drying a 0.5-1.0 $`\mu `$l droplet of SOPC dissolved in a 4:1 chloroform-methanol solution (7.35 mg/ml) on a glass slide. The sample was then closed, and hydration was effected by injecting a polymer solution of known concentration, $`c_p`$, into the cell. The development of myelin structures and their coiling were followed using phase contrast microscopy and recorded on video. Our experiments were conducted at room temperature, well above the solid-liquid transition for SOPC.
For small values of $`c_p`$ we observe myelin figures, which display a clear tendency to straighten over lengths many times larger than their diameter. As $`c_p`$ is increased, myelin figures become more floppy and curved. For large enough values of $`c_p`$, a writhing instability sets in and tubes bend, forming irregular structures, single coils (Fig. 1) and double helices (Fig. 2). We emphasize that all the coiled structures we observe are maximally tight as they form and do not tighten up gradually (see Fig. 2); no loose coils have been found (unlike Sakurai et al. ). In quantitative terms, this means that the curvature of the tube central line, $`C`$, is $`C\frac{1}{r_0}`$, where $`r_0`$ is the radius of the tube. Fig. 2 also demonstrates that there is no adhesion between membrane surfaces in our system; tube segments that are in contact in Fig. 2b are separated in Fig. 2c.
We stress that while the polymer concentration in solution, $`c_p`$, is known, we do not control the surface concentration on the bilayers. The slow evolution of some of the structures we observe is consistent with a possible variation of this concentration over time. As a first step towards a theoretical understanding of this system, we neglect this slow evolution.
We now make the following assumptions: (i) anchors penetrate the membrane to a depth of about half a bilayer. This is because the anchor length is comparable to that of a lipid, and the large hydrophilic backbone to which anchors are attached cannot penetrate a bilayer. We assume that these anchors and the polymer backbone induce a local spontaneous curvature, $`H_0`$ . (ii) Polymer molecules are present everywhere in the system, including inner regions of the myelin figures. We have carried out fluorescence microscopy experiments and found that the polymer is present between layers of the stack in significant concentrations (additional details of the experiments will be published elsewhere). (iii) Given that bilayers are in a liquid-like state, polymer molecules can diffuse on the membrane. Thus, the energy of the system is lowered if they migrate to regions where the mean surface curvature $`H`$ is closer to $`H_0`$. (iv) Bilayers in a myelin figure maintain a constant area and are strongly constrained by the geometry of the stack, which prevents them from buckling. We therefore regard the myelin figures as flexible cylinders having a fixed circular cross section everywhere along their axis. These assumptions are strongly supported by our experimental observations. Note that the central line of the tube can bend. However, as a consequence of the above assumptions, its length remains fixed.
Based on these assumptions, we present a simple model which captures the essential physical features of the experimental system, and accounts for many of the experimental observations. We represent each bilayer as two square lattices (in the spirit of lattice-gas models), corresponding to the outer and inner monolayers. We associate two degrees of freedom with each lattice site, which corresponds to a membrane patch of area $`a^2`$. The first is the local mean surface curvature, $`H`$. The second is a binary occupation variable, which takes the values one or zero when the site is occupied or unoccupied respectively by an anchored molecule. This molecule induces a local spontaneous curvature, $`H_0`$. The area of a site, $`a^2`$, and the value of $`H_0`$ depend on the specific mechanism responsible for the spontaneous curvature. Thus, if $`H_0`$ is induced by individual anchors, the area of a site is microscopic ($`a^260\text{Å}^2`$). If, on the other hand, the spontaneous curvature is induced by the polymer backbone, the area of a site is mesoscopic ($`a^2610^5\text{Å}^2`$).
By convention, the curvatures of the inner and outer layers have opposite signs at the same position. Within the model, the energy of the system is a sum of the curvature energies of the individual area patches: $`2\kappa H^2`$ for a vacant site, and $`2\kappa ^{^{}}(HH_0)^2`$ for an occupied site. $`\kappa `$ and $`\kappa ^{}`$ are the local bending rigidities of a single layer without and with an anchored molecule, respectively. We suppose $`\kappa ^{^{}}>\kappa `$; this is consistent with models of composite membranes (although the systems these models describe are different from ours).
In order to find the equilibrium state of a tube, we have to calculate its free energy. This free energy depends on the curvature of its central line, $`C`$, and on $`\rho `$, the average of the occupation variable ($`\rho a^2`$ is the average density of the anchored molecules). If the spontaneous curvature, $`H_0`$, is large enough, the free energy of a bent tube is lower than that of a straight one. To show this we evaluate separately the energy and the entropy of the system.
Consider one cylindrical bilayer of length $`l`$ and circular cross section of radius $`r`$, with the same average occupancy, $`\rho `$, on both sides. Let us calculate the energy cost of bending the bilayer into a portion of a coil with central line curvature $`C`$ in two steps. First, the energy of a bent cylindrical bilayer with a homogeneous distribution of anchored molecules is $`E_{hom}=E_{hom}^{out}+E_{hom}^{in}`$, where $`E_{hom}^{out,in}`$ are the energies of the outer and inner monolayers. According to our model
$`E_{hom}^{out,in}(C)`$ $`=`$ $`2\rho \kappa ^{}{\displaystyle 𝑑A\left[H(C)H_0\right]^2}`$ (1)
$`+`$ $`2(1\rho )\kappa {\displaystyle 𝑑A\left[H(C)\right]^2},`$ (2)
where $`H(C)`$ is the local membrane curvature, and is known for a cylindrical geometry. The expressions for $`E_{hom}^{out}`$ and $`E_{hom}^{in}`$ are different since the curvatures of the inner and outer monolayers have opposite signs. Note that when $`C0`$, $`H(C)`$ varies around the bent cylinder. For our geometry the total mean curvature $`𝑑AH(C)=2\pi l`$ for the outer monolayer, while for the inner one $`𝑑AH(C)=2\pi l`$, independent of the central line curvature $`C`$. Thus the cost of bending the cylindrical membrane, keeping the distribution of anchored molecules homogeneous is $`\mathrm{\Delta }E_{hom}=4[\rho \kappa ^{}+(1\rho )\kappa ]𝑑A\left[H(C)\right]^2`$, independent of the value of $`H_0`$. Note that the term containing $`H_0^2`$ is independent of $`C`$ and therefore does not contribute to the cost of bending.
Next, we take into account inhomogeneities in the distribution of anchored molecules around the tube. Such inhomogeneities reduce the energy if these molecules move to regions of membrane curvature closer to $`H_0`$ in both the outer and inner monolayers. The full calculation shows that the energy gain, $`\mathrm{\Delta }E_{inhom}(C,H_0)`$, depends linearly on the spontaneous curvature and can become arbitrarily large for large values of $`H_0`$.
As for the entropy of the system, we assume that the dominant contribution is the entropy of mixing of occupied and vacant sites. This entropy is larger when the distribution of anchored molecules around the cylindrical bilayer is homogeneous, favoring a straight tube. However, it does not depend on the spontaneous curvature. Therefore, if $`H_0`$ is large enough, the energy gain due to $`\mathrm{\Delta }E_{inhom}(C,H_0)`$ is larger than the free energy cost coming from $`\mathrm{\Delta }E_{hom}`$ and the entropy of mixing. In this case, the tube is bent at equilibrium. It remains to be shown that such an equilibrium state can occur for reasonable and physical values of the model parameters.
We have carried out the full calculation of the free energy as a function of the central line curvature, the average occupancy and the spontaneous curvature (details will be published elsewhere). In the limit of a thick tube (a realistic case), the calculation can be done analytically, and the free energy of the entire tube (summing over all the bilayers) reads:
$$F(C,\rho )=\frac{2l\kappa _{tube}(\rho )}{r_0}\mathrm{ln}\left[\frac{2}{1+\sqrt{1(Cr_0)^2}}\right],$$
(3)
where $`\kappa _{tube}(\rho )`$ is the effective bending rigidity of the entire tube, and $`l`$ is its length.
Fig. 3 shows the typical dependence of $`\kappa _{tube}`$ on $`\rho `$ for large enough values of $`H_0`$ (solid line). When $`\rho <\rho _{}`$, $`\kappa _{tube}`$ is positive and decreases with $`\rho `$. In this regime the minimum of the free energy is at $`C=0`$. Therefore, the tube is predicted to be straight on the average, but with enhanced fluctuations due to the smaller bending modulus. Although anchored molecules increase the local bending rigidity of the membrane ($`\kappa ^{}>\kappa `$), their mobility makes it easier to bend the tube.
For $`\rho _{}<\rho <\rho _+`$, $`\kappa _{tube}`$ is negative, and the free energy decreases upon bending with its minimum at the maximally possible central line curvature $`C=1/r_0`$ (in agreement with the qualitative argument outlined above). Hence, $`\rho _{}`$ is a threshold occupancy above which straight tubes are unstable and form maximally tight coiled structures. Above $`\rho _+`$ straight tubes become stable again; however, this regime is probably unreachable in our experiments, since too large a polymer concentration destroys the bilayers.
A salient feature of our model is that it precludes loose coiled structures; i.e. tubes are either uncoiled ($`\rho <\rho _{}`$ and $`\rho >\rho _+`$) or maximally tight and coiled ($`\rho _{}<\rho <\rho _+`$). This is consistent with our experimental observations. In addition, according to the model all the coils with $`C=1/r_0`$ are equally probable. One should therefore expect to see irregular coils without a well defined chirality as well as regular helices, all of which we indeed observe.
A somewhat similar instability of a flat membrane due to coupling between membrane shape and local spontaneous curvature has been discussed by Leibler and by Safran . There is, however, a fundamental difference between these models and ours. They considered an unconstrained flat membrane, whereas our membranes are severely constrained by the cylindrical geometry of the tube. These geometrical constraints play a crucial role in determining the shapes of the observed structures, and the final state of the system.
We believe coiling occurs mainly due to the spontaneous curvature induced by the anchored molecules as well as their mobility. Our model emphasizes these aspects, and neglects others such as interactions between polymer molecules. We do not expect these effects to change the qualitative behavior of the system. So far, we have not been able to identify the precise mechanism by which the polymer induces spontaneous curvature. We intend to study polymer length effects on the coiling phenomenon in order to partially address this issue.
We acknowledge useful exchanges with D. Bensimon, R. Granek, R. Lipowsky, E. Moses and S. Safran. This research was supported by The Israel Science Foundation administered by the Israel Academy of Sciences and Humanities - Recanati and IDB Group Foundation. V.F. gratefully acknowledges support from The Research Council of Norway (NFR). D.K. is the incumbent of the Ruth Epstein Recu Career Development Chair.
|
no-problem/9904/gr-qc9904038.html
|
ar5iv
|
text
|
# Curvature conditions for the occurrence of a class of spacetime singularities
## I Introduction
Recently, one of us has shown that, under certain physically reasonable conditions, a generic gravitational collapse developing from a regular initial state cannot lead to the formation of a final state resembling the Kerr solution with $`a^2>m^2`$—i.e., of a naked singularity accompanied by closed timelike curves. This result supports the validity of Penrose’s cosmic censorship hypothesis and suggests that there may exist some deeper connection between cosmic censorship and the chronology protection conjecture put forward by Hawking . An important role in this result plays the so-called inextendibility condition (see Sec. II), which is assumed to be satisfied for certain incomplete null geodesics. This condition enables one to rule out artificial naked singularities that could easily be created by simply removing points from otherwise well-behaved spacetimes. The inextendibility condition is based on the idea that physically essential singularities should always be associated with large curvature strengths, which are in turn usually associated with the focusing of Jacobi fields along null geodesics.
It is easily seen that the inextendibility condition will always hold for null geodesics terminating at the so-called strong curvature singularities defined by Tipler (see below). Singularities of this type are sometimes considered to be the only physically reasonable singularities (cf., e.g., ). However, strong curvature singularities can exist only if the curvature in their neighborhood diverges strong enough , while it is not unlikely that some singularities occurring in generic collapse situations will involve a weaker divergence of the curvature. In fact, one cannot a priori exclude the existence of some “real” singularities near which the curvature would remain even bounded (such singularities occur, for example, in Taub-NUT space). Accordingly, since we still have no fully accepted necessary condition on the behavior of the curvature near generic singularities, one should try to prove any cosmic censorship result under as weak a curvature condition as possible. It would be therefore of interest, in view of the mentioned censorship result , to know what are curvature conditions for the occurrence of singularities corresponding to the inextendibility condition. Furthermore, the inextendibility condition has also been used in proving some other recent results that restrict a class of possible causality violations in classical general relativity.
In this paper, we formulate and prove a theorem that establishes some relations between the inextendibility condition and the rate of growth of the Ricci curvature along incomplete null geodesics. This theorem shows that the inextendibility condition may hold for a much more general class of possible singularities than only those of the strong curvature type. Our theorem will be stated in Sec. II of the paper. In Sec. III we present a proof of the theorem; our main mathematical tool in this proof is a Sturm-type comparison lemma for nonoscillatory solutions of second-order differential equations. In Sec. IV we give a few concluding remarks; in particular, we argue that some earlier cosmic censorship results obtained for strong curvature singularities can be extended to singularities corresponding to the inextendibility condition.
## II The theorem
To begin with, we clearly need to recall the precise formulation of the inextendibility condition. Let $`\eta (t)`$ be an affinely parametrized null geodesic, and let $`Z_1`$ and $`Z_2`$ be two linearly independent spacelike vorticity-free Jacobi fields along $`\eta (t)`$. The exterior product of these Jacobi fields defines a spacelike area element, whose magnitude at affine parameter value $`t`$ we denote by $`A(t)`$. If we now introduce the function $`z(t)`$ defined by $`A(t)z^2(t)`$, then one can show that $`z(t)`$ satisfies the following equation:
$$\frac{d^2z}{dt^2}+\frac{1}{2}(R_{ab}K^aK^b+2\sigma ^2)z=0,$$
(1)
where $`K^a`$ is the tangent vector to $`\eta (t)`$ and $`\sigma ^2`$ is a non-negative function of $`t`$ defined as follows: $`2\sigma ^2\sigma _{mn}\sigma ^{mn}`$ $`(m,n=1,2)`$. Here $`\sigma _{mn}`$ is the shear tensor (see , p. 88) that satisfies the equation :
$$\frac{d}{dt}\sigma _{mn}=C_{manb}K^aK^b\frac{2}{z}\frac{dz}{dt}\sigma _{mn}.$$
(2)
In the following, by $`M`$ we shall denote a spacetime, i.e., a smooth, boundaryless, connected, four-dimensional Hausdorff manifold with a globally defined $`C^2`$ Lorentz metric.
Definition: Let $`\eta :(0,a]M`$ be an affinely parametrized, incomplete null geodesic. Assume also that $`\eta (t)`$ generates an achronal set, i.e., a set such that no two points of it can be joined by a timelike curve. Then $`\eta (t)`$ is said to satisfy the inextendibility condition if for some affine parameter value $`t_1(0,a)`$ there exists a solution $`z(t)`$ of Eq. (1) along $`\eta (t)`$ such that $`z(t_1)=0`$, $`dz/dt|_{t_1}0`$ and $`lim_{t0}z(t)=0`$.
The key idea behind the inextendibility condition is based on the fact that any two zeros of any solution of Eq. (1), which is not identically zero along a given null geodesic, correspond to a pair of conjugate points along the geodesic (see ). From Proposition 4.5.12 of Ref. it follows that incomplete null geodesics generating achronal sets cannot contain any pairs of conjugate points. One can thus easily show that if a geodesic $`\eta :(0,a]M`$ satisfies the inextendibility condition, then there is $`no`$ extension of the spacetime $`M`$, preserving all the above mentioned properties of $`M`$, in which $`\eta (t)`$ could be extended beyond a point $`\eta (0)`$. This means, according to the standard interpretation, that $`\eta (t)`$ should then approach a genuine singularity of the spacetime $`M`$ at affine parameter value 0. \[Formally, this singularity has the same status as those predicted by the familiar singularity theorems , because these theorems predict in fact the existence of incomplete causal (usually null) geodesics in maximally extended spacetimes satisfying just the same topological and smoothness conditions as those imposed on $`M`$.\]
Let us now compare the inextendibility condition with the concept of a strong curvature singularity . Consider a null geodesic $`\lambda :(0,a]M`$ that terminates in a strong curvature singularity at affine parameter value 0. This means that every solution $`z(t)`$ of Eq. (1) along $`\lambda (t)`$, which vanishes for at most finitely many points in $`(0,a]`$, satisfies $`lim_{t0}z(t)=0`$ (cf. Ref. , p. 160). Suppose now that $`\lambda (t)`$ generates an achronal set; then any solution of Eq. (1), which is not identically zero along $`\lambda (t)`$, cannot vanish for any two points in $`(0,a]`$ by the argument with conjugate points mentioned above. Thus, for all $`t_1(0,a]`$ and for all solutions $`z(t)`$ of Eq. (1) along $`\lambda (t)`$ with initial conditions $`z(t_1)=0`$ we will have $`lim_{t0}z(t)=0`$. It is thus clear that any null geodesic terminating in Tipler’s strong curvature singularity and generating an achronal set must always satisfy the inextendibility condition. Notice also that the terms “all” emphasized above imply, via Eqs. (1) and (2), that $`\lambda (t)`$ can terminate in the strong curvature singularity only if the curvature diverges strong enough along $`\lambda (t)`$ as $`t0`$, while the inextendibility condition could actually be satisfied for $`\lambda (t)`$ even if the curvature along it would remain bounded. Indeed, the theorem stated below makes it clear \[see condition (i)\] that the curvature need not necessarily diverge along geodesics satisfying the inextendibility condition.
Theorem:Let $`\eta :(0,a]M`$ be an affinely parametrized, incomplete null geodesic generating an achronal set. Suppose that the Ricci tensor term $`r(t)R_{ab}K^aK^b`$ along $`\eta (t)`$, where $`t`$ is the affine parameter and $`K^a`$ is the tangent vector to $`\eta (t)`$, obeys at least one of the following conditions.
(i) There exists an affine parameter value $`b(0,a)`$ such that $`inf\{r(t)|\mathrm{\hspace{0.17em}0}<t<b\}2(\pi /b)^2`$.
(ii) There exist an affine parameter value $`c(0,a)`$ and a constant $`\mu (0,2)`$ such that $`r(t)\kappa t^\mu `$ for all $`t(0,c]`$,
where $`\kappa =(2/3)(3326\mu +5\mu ^2)c^{\mu 2}`$.
Then $`\eta (t)`$ satisfies the inextendibility condition.
Remark 1: From the proof of this theorem, which is given below, it may be seen that the parameter values $`b`$ and $`c`$ mentioned above in conditions (i) and (ii) correspond to the parameter value $`t_1`$ occurring in the definition of the inextendibility condition.
Remark 2: Since in the theorem $`\eta (t)`$ is assumed to be a generator of an achronal set, $`\eta (t)`$ cannot contain any pair of conjugate points, and so one can expect that there should exist an upper limit on the rate of growth of the curvature along $`\eta (t)`$. Indeed, from Theorems (3) and (4) of Ref. it follows immediately that the Ricci tensor term $`r(t)`$ along $`\eta (t)`$ must satisfy the following two conditions: (1) there is no affine parameter value $`b^{}(0,a]`$ such that $`inf\{r(t)|\mathrm{\hspace{0.17em}0}<t<b^{}\}>8(\pi /b^{})^2`$; and (2) if $`r(t)0`$ on $`\eta (t)`$, then $`lim_{t0}inft^2r(t)1/2`$. Similar restrictions on the growth of the Weyl part of the curvature along $`\eta (t)`$ can be obtained from Proposition 2.2 of Ref. .
In the context of our theorem, it is worth recalling the analogous results obtained by Clarke and Królak for singularities of the strong curvature type. They have been obtained for two definitions of a strong curvature singularity: the original one formulated by Tipler and its modification proposed by Królak . According to these results, if a null geodesic $`\eta :(0,a]M`$ terminates at affine parameter value 0 in a strong curvature singularity defined by Tipler (resp., by Królak), then there must exist some affine parameter value $`c(0,a]`$ such that $`R_{ab}K^aK^b>At^2`$ (resp., $`R_{ab}K^aK^b>At^1`$) on $`(0,c]`$, where $`K^a`$ is the tangent vector to $`\eta (t)`$, $`t`$ is the affine parameter, and $`A`$ is some fixed positive constant. \[Or very similar conditions on the rate of growth of the Weyl part of the curvature along $`\eta (t)`$ must be satisfied; see Corollary 2 of Ref. 7.\] Comparing these results with condition (ii) of our theorem we see that singularities of the strong curvature type involve a considerably stronger divergence of the Ricci tensor term $`R_{ab}K^aK^b`$ than singularities corresponding to the inextendibility condition. There may thus exist a large class of curvature singularities that are not strong in the sense of the definition of Tipler or Królak, but they may still satisfy the inextendibility condition. Note also that the above conditions for strong curvature singularities are the necessary ones, whereas conditions (i) and (ii) of our theorem are only sufficient to ensure that the inextendibility condition does hold for a given geodesic. This implies that the inextendibility condition might be satisfied in more general situations than only those characterized by conditions (i) and (ii).
## III Proof of the theorem
Now we shall prove the theorem; our main tool in this proof will be the following comparison lemma.
Lemma (The comparison lemma): Suppose that $`u(s)`$ is a solution of the equation
$$\frac{d^2u}{ds^2}+F(s)u(s)=0$$
on an interval $`(a,b]`$ with initial conditions: $`u(b)=0`$ and $`du/ds|_b0`$. Let $`v(s)`$ be a solution of
$$\frac{d^2v}{ds^2}+G(s)v(s)=0$$
on $`(a,b]`$ such that $`v(b)=0`$, $`dv/ds|_b=du/ds|_b`$ and $`v(s)>0`$ on $`(a,b)`$. Assume also that $`F(s)`$ and $`G(s)`$ are piecewise continuous on $`(a,b]`$, and let $`G(s)F(s)`$ on $`(a,b]`$. Then $`u(s)v(s)`$ on $`(a,b]`$.
$`Proof:`$ The proof of this lemma is based essentially on Theorem 1.2 of Ref. , p. 210. To apply this theorem in its original form, it is convenient to reparametrize both of the equations in the lemma introducing the parameter $`t=s`$ instead of $`s`$. Note that this reparametrization does not change the form of the equations. Clearly, we shall now have established the lemma if we show that for any $`c(a,b)`$, $`u(t)v(t)`$ on $`[b,c]`$.
Consider the ratio $`u(t)/v(t)`$. Since $`v(t)>0`$ on $`(b,a)`$, it is well defined on $`(b,c]`$. Using l’Hospital’s rule, we get
$$\underset{tb}{lim}\frac{u(t)}{v(t)}=1.$$
Therefore, as $`v(t)>0`$ on $`(b,c]`$, to show that $`u(t)v(t)`$ on $`[b,c]`$, it suffices to show that
$$\frac{d}{dt}\left[\frac{u(t)}{v(t)}\right]0$$
on $`(b,c]`$. It is easy to see that this inequality holds if
$$\frac{v(t)}{\dot{v}(t)}\frac{u(t)}{\dot{u}(t)}$$
(3)
on $`(b,c]`$, where the overdot denotes the first derivative with respect to $`t`$. Since $`F(t)`$ and $`G(t)`$ are piecewise continuous on $`[b,c]`$, by Theorem 1.2 of Ref. 13, p. 210, we have
$$\mathrm{tan}^1\left[\frac{v(t)}{\dot{v}(t)}\right]\mathrm{tan}^1\left[\frac{u(t)}{\dot{u}(t)}\right]$$
for all $`t[b,c]`$. Thus, as $`\mathrm{tan}^1`$ is an increasing function, the inequality (3) does hold as it is desirable. $`\mathrm{}`$
Proof of the theorem:
(Part I) Suppose the condition (i) is satisfied. Let $`z_0(t)`$ be a solution of Eq. (1) along $`\eta (t)`$ such that $`z_0(t)`$ is not identically zero on $`(0,b]`$ and $`z_0(b)=0`$, where $`b`$ is the parameter value mentioned in condition (i). Clearly, such a solution will always exist. Since $`\eta (t)`$ generates an achronal set, $`z_0(t)`$ can vanish nowhere in $`(0,b)`$; otherwise $`\eta (t)`$ would have a pair of conjugate points in $`(0,b]`$ (see Ref. ), which would contradict, by Proposition 4.5.12 of Ref. , the achronality of $`\eta (t)`$. Notice also that Eq. (1) is linear, and so the function $`z_0(t)`$ will be a solution of Eq. (1) as well. Thus, as $`z_0(t)0`$ on $`(0,b)`$, without loss of generality we can assume that $`z_0(t)>0`$ on $`(0,b)`$. This implies, as $`z_0(b)=0`$, that $`dz_0/dt|_b0`$. Since $`z_0(t)>0`$ on $`(0,b)`$, and condition (i) holds, from Eq. (1) we see at once that $`z_0(t)`$ must be a concave function on $`(0,b]`$. This makes it obvious that $`dz_0/dt|_b0`$, and so we must have $`dz_0/dt|_b=\alpha <0`$. Let us now define the function $`z_1(t)(1/\alpha )z_0(t)`$. As Eq. (1) is linear, it is clear that $`z_1(t)`$ will be a solution of Eq. (1) along $`\eta (t)`$; notice also that $`z_1(t)>0`$ on $`(0,b)`$, $`z_1(b)=0`$ and $`dz_1/dt|_b=1`$.
Consider now the equation
$$\frac{d^2x}{dt^2}+\omega x(t)=0,$$
(4)
where $`\omega =\frac{1}{2}inf\{r(t)|\mathrm{\hspace{0.17em}0}<tb\}`$ and $`r(t)`$ is the function defined in the theorem. Notice that $`\omega >0`$ by condition (i). Let $`x_1(t)`$ be a solution of Eq. (4) on $`(0,b]`$ with initial conditions: $`x_1(b)=0`$ and $`dx_1/dt|_b=1`$. It is a simple matter to see that $`x_1(t)=\omega ^{1/2}\mathrm{sin}[\omega ^{1/2}(bt)]`$. Let us now apply the comparison lemma to the equations (1) and (4) and their solutions $`z_1(t)`$ and $`x_1(t)`$. Since $`\omega \frac{1}{2}r(t)`$ on $`(0,b]`$, by the comparison lemma we must have $`x_1(t)z_1(t)`$ on $`(0,b]`$. Consequently, as $`z_1(t)>0`$ on $`(0,b)`$, we obtain $`x_1(t)>0`$ on $`(0,b)`$. This implies, by the above form of $`x_1(t)`$, that $`\omega (\pi /b)^2`$. But $`\omega (\pi /b)^2`$ by condition (1). We must thus have $`\omega =(\pi /b)^2`$, which gives $`lim_{t0}x_1(t)=0`$. Therefore $`lim_{t0}z_1(t)=0`$ since $`x_1(t)z_1(t)>0`$ on $`(0,b)`$. This means that $`\eta (t)`$ does satisfy the inextendibility condition.
(Part II) The task is now to prove the theorem in the case when condition (ii) holds. For this purpose, let us consider the following equation
$$\frac{d^2y}{dt^2}+Bt^\mu y(t)=0$$
(5)
on $`(0,c]`$, where $`B=\kappa /2`$, and $`\kappa `$, $`\mu `$ and $`c`$ are some fixed constants mentioned in the condition (ii). Let $`y_1(t)`$ be a solution of this equation with initial conditions: $`y_1(c)=0`$ and $`dy_1/dt|_c=1`$. Let $`z_2(t)`$ be a solution of Eq. (1) along $`\eta (t)`$ such that $`z_2(c)=0`$ and $`dz_2/dt|_c=1`$. \[There is no loss of generality in assuming $`z_2(t)`$ to exist; the existence of $`z_2(t)`$ can be established in the same manner as the existence of the analogous solution $`z_1(t)`$ considered in the first part of the proof.\] Clearly, the solution $`z_2(t)`$, just as $`z_1(t)`$, can vanish nowhere in $`(0,c)`$ by the argument with conjugate points. Therefore, as $`dz_2/dt|_c=1`$, we must have $`z_2(t)>0`$ on $`(0,c)`$. Let us now apply the comparison lemma to the equations (1) and (5) and their solutions $`z_2(t)`$ and $`y_1(t)`$. By condition (ii) we have $`r(t)\kappa t^\mu `$ on $`(0,c]`$. Thus by the comparison lemma, we must have $`y_1(t)z_2(t)`$ on $`(0,c]`$. Of course, in order to prove the theorem, it suffices to show that $`lim_{t0}z_2(t)=0`$. Thus, as $`y_1(t)z_2(t)>0`$ on $`(0,c)`$, to complete the proof it suffices to show that $`lim_{t0}y_1(t)=0`$. We shall show below that $`y_1(t)`$ does possess this property.
To this end, let us first find the general solution of Eq. (5). It is easy to check that if one puts $`x=t`$, $`\alpha =1/2`$, $`\beta =2\sqrt{B}(2\mu )^1`$, $`\gamma =(2\mu )/2`$ and $`n=(2\mu )^1`$ into the equation (4.1) of Ref. , p. 138, then this equation reduces to our equation (5). Thus, according to the solution (4.3) of Eq. (4.1) of Ref. , our equation (5) has the following general solution
$$y(t)=t^{1/2}[C_1J_n(\beta t^\gamma )+C_2Y_n(\beta t^\gamma )],$$
(6)
where $`C_1`$ and $`C_2`$ are arbitrary constants of integration, and $`J_n(\beta t^\gamma )`$ and $`Y_n(\beta t^\gamma )`$ are the Bessel functions of order $`n`$, of the first and second kind, respectively. Since $`\mu (0,2)`$, from the above relations it follows that $`1/2<n<\mathrm{}`$, $`\sqrt{B}<\beta <\mathrm{}`$ and $`0<\gamma <1`$.
Let us recall that any Bessel function of the first kind has infinitely many positive zeros (cf., e.g., , p. 29). Let $`j_{n,1}`$ be the first positive zero of the function $`J_n(\beta t^\gamma )`$, i.e., $`J_n(j_{n,1})=0`$ and $`J_n(\beta t^\gamma )0`$ as long as $`0<\beta t^\gamma <j_{n,1}`$. Since $`n>1/2`$, $`j_{n,1}`$ must satisfy the following relation (see Eq. (2) of Ref. , p. 29):
$$j_{n,1}<2[(n+1)(n+5)/3]^{1/2}.$$
(7)
For $`J_n(\beta t^\gamma )`$ we now define $`L`$ to be the number such that $`j_{n,1}=L\beta c^\gamma `$. Putting this into (7), and taking into account the fact that $`\beta =(2\kappa )^{1/2}(2\mu )^1`$, $`\kappa =3^1(6652\mu +10\mu ^2)c^{\mu 2}`$, $`\gamma =(2\mu )/2`$ and $`n=(2\mu )^1`$, we readily find that $`L^2<1`$.
Consider now equation (5) with $`B`$ replaced by $`B^{}=L^2B`$. Let $`y_2(t)`$ be a solution of this equation on $`(0,c]`$ with initial conditions: $`y_2(c)=0`$ and $`dy_2/dt|_c=1`$. The general form of this solution is given by (6), where $`\beta `$ should be replaced by $`\beta ^{}=2\sqrt{B^{}}(2\mu )^1`$ (notice that $`\beta ^{}=L\beta `$). Let us now insert the initial conditions for $`y_2(t)`$ into this general solution in order to determine for $`y_2(t)`$ the constants $`C_1`$ and $`C_2`$ occurring in (6). To find the first derivative of the general solution (6), we use the following recurrence formula
$$\frac{dJ_n(x)}{dx}=J_{n+1}(x)+\frac{n}{x}J_n(x),$$
which is also valid for $`Y_n(x)`$ (see , p. 197). We can now easily calculate the constants $`C_1`$ and $`C_2`$; the result is as follows
$$C_1=\frac{Y_n(\beta ^{}c^\gamma )}{\beta ^{}\gamma c^{\gamma 1/2}\left[Y_n(\beta ^{}c^\gamma )J_{n+1}(\beta ^{}c^\gamma )Y_{n+1}(\beta ^{}c^\gamma )J_n(\beta ^{}c^\gamma )\right]}$$
(8)
and
$$C_2=\frac{J_n(\beta ^{}c^\gamma )}{\beta ^{}\gamma c^{\gamma 1/2}\left[Y_n(\beta ^{}c^\gamma )J_{n+1}(\beta ^{}c^\gamma )Y_{n+1}(\beta ^{}c^\gamma )J_n(\beta ^{}c^\gamma )\right]}.$$
(9)
As $`\beta ^{}=L\beta `$, from the above definition of $`L`$ it is clear that $`\beta ^{}c^\gamma =j_{n,1}`$. Thus $`J_n(\beta ^{}c^\gamma )=0`$ and the numerator in (9) must vanish. As $`J_n(\beta ^{}c^\gamma )=0`$, the denominator in (9) can vanish only if $`Y_n(\beta ^{}c^\gamma )J_{n+1}(\beta ^{}c^\gamma )=0`$. But the Bessel functions $`J_{n+1}`$ and $`Y_n`$ cannot have any common zeros with the Bessel function $`J_n`$ (see , pp. 29-32), and so the denominator in (9) cannot vanish. We thus have $`C_2=0`$ and, by (6) and (8), the solution $`y_2(t)`$ can be written as follows
$$y_2(t)=C_1t^{1/2}J_n(\beta ^{}t^\gamma ),$$
(10)
where $`C_1=[\beta ^{}\gamma c^{\gamma 1/2}J_{n+1}(\beta ^{}c^\gamma )]^1`$.
Let us now compare the solutions $`y_1(t)`$ and $`y_2(t)`$ by means of the comparison lemma. Recall that $`y_1(t)`$ is a solution of equation (5) with $`B=\kappa /2`$, while $`y_2(t)`$ is a solution of the same equation with $`B`$ replaced by $`B^{}=L^2\kappa /2`$. Since $`L^2<1`$, by the comparison lemma we must have $`y_2(t)y_1(t)`$ for all $`t(0,c]`$. We recall that any Bessel function $`J_k(x)`$ of the first kind with real $`x`$ and $`k>0`$ is continuous at $`x=0`$ (cf. , p. 182). Thus, as $`n>1/2`$ and $`0<\gamma <1`$, from (10) it follows immediately that $`lim_{t0}y_2(t)=0`$. Therefore, as $`y_2(t)y_1(t)>0`$ on $`(0,c)`$, we obtain $`lim_{t0}y_1(t)=0`$, which completes the proof. $`\mathrm{}`$
## IV Concluding remarks
We have been concerned in this paper with the problem of determining what are curvature conditions for the occurrence of singularities corresponding to the inextendibility condition. We have found two such sufficient conditions concerning the behavior of the Ricci tensor term $`R_{ab}K^aK^b`$ along incomplete null geodesics—these are conditions (i) and (ii) of the theorem stated in Sec. II. This theorem shows that the inextendibility condition may hold for a considerably larger class of possible singularities than only those of the strong curvature type. In particular, condition (i) of the theorem shows that the inextendibility condition may hold even if the curvature along incomplete geodesics would remain bounded. In this context, it is worth recalling that singularities predicted by the famous singularity theorems can be interpreted as regions of the universe at which the normal classical spacetime picture and/or certain energy conditions break down, and this may occur in regions where the curvature, though extremely large, still remains finite. Accordingly, if one attempts to establish, for example, whether or not these singular regions will conform to any cosmic censorship principle, it would be well to try to characterize, if necessary, incomplete geodesics terminating in these regions by a condition that may hold even if the curvature along the geodesics would remain bounded. One possible candidate for such a condition may thus be the inextendibility condition.
It should also be stressed here that some earlier cosmic censorship theorems proved for strong curvature singularities can be extended to singularities corresponding to the inextendibility condition. To see this, let us first recall that these theorems show, briefly, that under certain restrictions imposed on the causal structure, strong curvature singularities are censored (see Refs. for details). Proofs of these theorems are, in essence, alike. In a brief outline, they run as follows. First, one shows that if the theorem under consideration were false, then there would have to exist a sequence $`\{\mu _i\}`$ of future endless, future complete null geodesics converging to a null geodesic $`\mu `$ that terminates in the future at a strong curvature singularity. One also shows that $`\mu `$ and all the $`\mu _i`$ must be generators of achronal sets. As all $`\mu _i`$ are achronal, none of them can have a pair of conjugate points, and so any irrotational congruence of Jacobi fields along any $`\mu _i`$ cannot be refocused. As $`\{\mu _i\}`$ converges to $`\mu `$, this must then imply, by continuity, that any irrotational congruence of Jacobi fields along $`\mu `$ cannot be refocused as well. However, as $`\mu `$ terminates in a strong curvature singularity, all irrotational congruences of Jacobi fields along $`\mu `$ should be refocused. This gives the required contradiction. It is not difficult to see, however, that this contradiction can equally well be obtained if $`\mu `$ would be assumed to satisfy the inextendibility condition, for this condition holds if at least one irrotational congruence of Jacobi fields along a given geodesic is refocused. It is thus clear that the censorship theorems given in Refs. are unnecessarily restricted to strong curvature singularities and they can be extended to singularities corresponding to the inextendibility condition.
## ACKNOWLEDGMENTS
This research was supported in part by the Polish State Committee for Scientific Research (KBN) under Grant No. 2 P03B 073 15.
|
no-problem/9904/cond-mat9904012.html
|
ar5iv
|
text
|
# Penetration and transformations of vortices in bulk current-carrying superconductors
## I INTRODUCTION
Not long ago a variety of complex magnetic structures formed by many strongly curved and entangled vortices was discovered in bulk superconductors . The origin of these structures can not be explained if treat the motion of vortices like that of stick-like objects. It is necessary to consider the evolution of three-dimensional magnetic flux lines with potentially arbitrary shape.
For the first, it is useful to investigate the motion and shaping of a single vortex but interacting with a surface supercurrent which represents either transport current or Meissner current induced by external field. As far as we know, even this simple task previously was not under careful consideration.
Of course, a vortex never lives alone, without interactions with other vortices, and no stable many-vortex structure could exist without mutual repulsion of vortices. However, one can suppose that the scenario of magnetic flux penetration into a current-carrying bulk superconductor should be dominated only by vortex interaction with the surface current distribution, i.e. eventually by geometry of superconducting sample, not by the vortex-vortex interaction. The latter can not seriously affect this scenario, merely because it itself is unable to ensure a deep penetration at all.
To prove this statement, let us imagine the steady flow of vortices which arise at the flat current-carrying boundary of half-infinite superconductor and then move deep into the sample due to their repulsion. Clearly, because of viscous character of vortex motion, such the flow needs in nonzero gradient of concentration of vortices. As a consequence, both the concentration and the local drift velocity of vortices must be decreasing functions of the depth. Hence, their product is not constant, that is the magnetic flux conservation can not be satisfied. This discrepancy means that no steady flow could be supported by the inter-vortex forces only. In particular, it is impossible to realize the stationary lasting penetration of vortices from infinite flat boundary parallel to external magnetic field.
Therefore, the only force what can push a vortex through the sample interior is nothing but self-action of vortex caused by its distorsion. But in order to involve this force into the evolution, the vortex must feel the shape of the sample boundary.
Hence, the true picture looks as follows. After nucleation in a surface layer with thickness of order of London penetration depth $`\lambda `$ , a vortex firstly expands over the sample boundary remaining in this layer. At this stage only the end fragments of vortex are factually moving. The ends slide along the boundary, and the resulting shape of vortex core reflects that of boundary. This process lasts until the curvature of main middle part of the core becomes sufficiently strong in order to cause the deepening of vortex as a whole.
In view of these reasonings, the geometry of steady transport of magnetic flux into a bulk supeconductor looks rather insensible to inter-vortex interactions and thus can be testified in terms of a single vortex, at least if not consider details of vortex nucleation and processes like annihilation and reconnections of vortices which take place deep inside the sample.
For example, many aspects of resistivity in supercurrent-carrying wires can be described as evolution of ring-like vortices as if thats instantly arise near the boundary, then contract independently one on another and finally self-annihilate . However, more correct consideration should include the first stage when vortex transforms from small nucleus into a ring. We shall see that in fact this stage may result also in a non-ring penetration geometry, and more detailed theory can predict what the scenario realizes under given transport current value and sample dimensions.
Though a lot of works were published previously touching upon a role of vortex distorsions, for instance, under a pinning by randomly distributed centers , always some preliminary restrictions of the vortex geometry were attracted. In the present work the general equations of evolution of arbitrarily curved vortex lines in isotropic superconductors are formulated and analysed. We shall especially discuss the true formulation of boundary conditions for these equations.
It will be shown that in a sample whose dimensions noticably exceed $`\lambda `$ the vortex can penetrate either as flexible stick or as elastic (similarly to a rubber thread attracted by its ends through water). of surface supercurrent. The latter case occurs only if surface supercurrent exceeds $`H_{c1}c/4\pi `$ (in CGS units) and is characterized by giant stretching of the vortex core along the sample boundary in the direction parallel to drift of the ends. The stretching is accompanied by decrease of both the vortex energy and viscous dissipation per unit drift velocity, and results in strong increase of the vortex drift velocity under given transport current. In the framework of this scenario, the vortex core firstly tranforms into a ring-like curve winding round the wire cross-section (or into a spiral, if there is an external magnetic field parallel to current), and only later the vortex begins to cut the wire and enter deep into its interior. This general picture is in agreement with the known simplific model of magnetic flux penetration into round wires. Additionally, our approach allows to scope very different stages of vortex evolution in unified manner and obtain quantitative estimates for each stage.
## II LONDON APPROXIMATION
We shall confine ourselves by the London approximation. Of course, it would have no sense if one could not apply it to actually moving vortices. But in any case the requirement must be satisfied that characteristic velocity $`u_0`$ of viscous vortex motion influenced by magnetic fields comparable with low critical magnetic field $`H_{c1}`$ , must be significantly smaller than the speed of electromagnetic waves.
The velocity scale $`u_0`$ can be naturally estimated as
$`u_0\mu \epsilon /\lambda `$where
$`\epsilon =\mathrm{\Phi }_0H_{c1}/4\pi `$is the self-energy per unit length of long straight-line vortex, $`\epsilon /\lambda `$ is the characteristic scale of Lorentz force also related to unit length, and $`\mu `$ is mobility of the vortex core. Below it will be seen that so defined $`u_0`$ really serves as the velocity unit. If combine this definition and the known relations
$`{\displaystyle \frac{c^2}{\mathrm{\Phi }_0\mu }}\sigma _nH_{c2},\sigma _n{\displaystyle \frac{\mathrm{}}{\mathrm{\Delta }}}({\displaystyle \frac{c}{\lambda }})^2`$with standard notations, $`\sigma _n`$ being the normal conductivity and $`\mathrm{\Delta }2k_BT_c`$ being the order parameter, one obtains
$`u_0{\displaystyle \frac{\lambda k_BT_c}{2\pi \mathrm{}}}{\displaystyle \frac{H_{c1}}{H_{c2}}}`$
As a typical example, at $`T_c100K`$ , $`\lambda 310^5cm`$ and $`H_{c2}/H_{c1}500`$ , one gets the estimate $`u_010^5cm/s`$ . This value looks small enough to allow for applicability of quazi-static London approximation. In fact, such an approach was used in large number of works on motion of separate vortices as well as vortex lattices. The obvious exception is very dense lattice, with small inter-vortex distancies of order of coherence length. But our present subjects of interest are far from such complications.
## III EVOLUTION EQUATIONS
In the framework of London approximation, the free energy $`E`$ of vortex, placed into a given surroundings, is completely determined by the shape of its core, $`R(p)=\{X(p),Y(p),Z(p)\}`$ , with $`X,Y,Z`$ being coordinates of the core points and $`p`$ being a scalar parameter. In accordance with the principles of mechanics and nonequilibrium thermodynamics, the simplest equation of a massless viscous evolution of the core line looks as
$$\mu ^1R/t=f(R)$$
(1)
with Lorentz force on the right-hand side and friction force on the left, both being related to unit core length.
By its sense, the parameter $`\mu ^1`$ is the effective drag coefficient which is determined by all the dissipative energy losses conjugated with the core motion. Generally, there are at least two sorts of dissipative processes accompanying the motion (see, for example, the review ), namely, relaxation of the order parameter and normal currents induced by time-dependent own magnetic field of the vortex. A concrete expression for $`\mu `$ can be derived from more detailed theory, for instance, from the Ginzburg-Landau functional approach, under its reduction to London approximation . The reduction is possible because normal self-current of moving vortex and corresponding dissipation are located mainly in a close vicinity of the core line, at distance comparable with coherence length. After the transition to London’s description, the effect of normal currents becomes hidden in $`\mu `$ , but these currents give no contribution to the Lorentz force . Therefore, the reduction results in the identity whose meaning is balance of friction force and Lorentz force, as it is stated by the Eq.1, with $`f(R)`$ being determined only by supercurrents.
To write $`f(R)`$ , one has not to evaluate the supercurrent distribution. Instead, as in general in mechanics and statistical thermodynamics, $`f(R)`$ can be expressed by means of $`E`$’s variation under a small displacement of a local core fragment, that is as the functional derivative $`\delta E/\delta R(p)`$ . However, the latter itself is not invariant with respect to arbitrary (non-degenerated) transformations of the parametrization $`R(p)`$ and to physical dimensionality of $`p`$ . In case of isotropic media, the only true invariant expression for the Lorentz force is
$$f=\frac{dp}{dL}\frac{\delta E}{\delta R(p)}=\left|\frac{R}{p}\right|^1\frac{\delta E}{\delta R(p)}=\frac{\mathrm{\Phi }_0}{c}\left[J\times N\right]$$
(2)
Here the vector $`R/pR^{}`$ is locally parallel to the core, $`dL=|R^{}|dp`$ is the differential of the core length, $`NR^{}\left|R^{}\right|^1=R/L`$ , and $`J`$ is the density of full effective supercurrent which streams around core and pushes a given core fragment.
The energy $`E=E\{R(p)\}`$ includes self-interaction of vortex and its interaction with surroundings, in particular, with other vortices. Correspondingly, in general $`J`$ consists of external currents and self-current of vortex determined by its distorsion. The Eqs.1 and 2 could be directly extended to a number of interacting vortices. Besides, in principle, one may add into $`E`$ also interactions with pinning potentials. However, below we are interested only in motion of separate vortex in absence of pinning.
The parameter $`p`$ enumerates strictly the core points. But in practice it is preferable to use another kind of parametrization, concretely, to introduce the parameter $`q`$ which enumerates some suitable continuum of surfaces $`Q(r)=q`$ , $`r=\{x,y,z\}`$ , each possessing only one intersection with core line. The connection between $`p`$ and new parameter $`q`$ is implied by the obvious relation $`Q(R(p(q,t),t))=q`$ , and simple algebraic manipulations lead to the modified form of the evolution equations,
$$\frac{R}{t}=\mu \left[1\frac{R}{q}\frac{Q(R)}{R}\right]f(R),f(R)=\left|\frac{R}{q}\right|^1\frac{\delta E}{\delta R(q)}$$
(3)
where the symbol $``$ denotes the tensor product of two vectors.
These equations describe how the intersection points marked by $`q`$ move along the corresponding surfaces $`Q(r)=q`$ . Clearly, this is factually two-dimensional motion. This feature becomes quite obvious if it is possible to identify $`q`$ as one of cartezian coordinates, that is to use parallel planes as the marking surfaces. For instance, if thats are XY-planes, $`q=Z`$ and $`Q(r)=z`$ , then the Eqs.3 reduces to the equation
$$\frac{}{dt}\left(\begin{array}{c}X\\ Y\end{array}\right)=\frac{\mu }{\sqrt{1+X^2+Y^2}}\left(\begin{array}{cc}1+X^2& X^{}Y^{}\\ Y^{}X^{}& 1+Y^2\end{array}\right)\left(\begin{array}{c}\delta E/\delta X(Z)\\ \delta E/\delta Y(Z)\end{array}\right)$$
(4)
with shortened notations $`X^{}X/Z`$ , $`Y^{}Y/Z`$ . The Eq.4 describes the time evolution of $`X`$ and $`Y`$ coordinates of the core points marked with their $`Z`$-coordinate.
## IV BOUNDARY CONDITIONS AND LOCAL APPROXIMATION
In absence of pinning and more vortices, the vortex energy $`E=E_s+E_i`$ consists of two parts: the energy $`E_i`$ of the vortex interaction with transport or Meissner supercurent and the self-energy $`E_s`$. Therefore, the current in the Eq.2 also can be devided into two parts, $`J=J_s+J_i`$ .
Formally, $`E_s`$ is a complicated spatially non-local functional depending on both the core configuration $`R(p)`$ and the shape of sample. Among other factors, $`E_s`$ includes the vortex interaction with the sample boundary what can be interpreted as attraction of the end fragments of the core to their mirror images placed outside superconductor.
But, if the curvature radius of the core everywhere is not too small as compared with $`\lambda `$ , and besides, if the core nowhere is too close to itself, then the so-called local approximation is possible,
$$E_s\epsilon L=\epsilon \left|dR(p)\right|$$
(5)
where $`L`$ is the core length. This well known approximation was argued and used as long ago as in 1968 by Galaiko , and later by many other authors (in particular, in \[2-5\]). Our own computer simulations showed that the relative error of evaluation of self-action force by means of local approximation does not exceed a few percents even if the curvature radius is as small as $`0.1\lambda `$ .
In the local approximation the Eqs.1 and 2 take the form
$$\frac{R}{t}=\mu \frac{\mathrm{\Phi }_0}{c}\left[(J_s+J_i)\times N\right],J_s=\frac{cH_{c1}}{4\pi }\left[N\times \frac{^2R}{p^2}\right]\left|\frac{R}{p}\right|^2$$
(6)
Here $`J_s`$ is the self-current what flows through the very core. Its absolute value is inversely proportional to the local curvature radius of core.
However, the local approximation needs to be accompanied by correct boundary conditions. The true conditions should take into account the vortex interaction with superconductor boundary. There are two ways to show that this interaction results in the orthogonality of the end fragments of core to the boundary. Thought these conditions are known at least since , sometimes thats are neglected, so it is desirable to present more argumentation.
First, let us note that the force vector $`f(R)`$ is always perpendicular to the local core direction. Indeed, any variarion $`\delta R`$ parallel to this core direction, $`\delta RR/p`$ , merely is identical to a change of parametrization, without factual change of the shape, so it has no physical meaning and should result in $`\delta E=0`$ (therefore the last expression in (2) always is consistent with previous ones). The same is seen from (6). As a consequense, any core point displaces perpendicularly to the core, in particular, the end points do which are placed just on the boundary. Hence, we must conclude that the end fragments always are oriented to be orthogonal with respect to the boundary.
Secondly, the non-orthogonality would mean that the contour formed by core and its mirror image is broken at the end point, i.e. has infinitely small curvature radius here. From the point of view of exact $`E_s`$ , if such a sharp ”knee” occured it would cause infinitely strong Lorentz self-action force and consequently would be immediately straightened thus restoring the orthogonality.
But, we must to underline that the principal conclusions to be deduced do not refer to the local approximation and can be derived from general non-local Eqs.1 and 2 only.
## V STICK-ELASTIC VORTEX TRANSFORMATION <br>IN CURRENT-CARRYING PLATE
To avoid a complicated mathematics, we confine ourselves by simplific superconductor geometry. Consider the vortex evolution in an infinitely wide plate, $`D<Z<D`$ , without pinning but in presence of transport surface supercurrent uniformly distributed over the boundary planes and obeying the London equation. If this current flows along Y-axis then
$`J_x=J_z=0,J_y={\displaystyle \frac{c}{4\pi \lambda }}H_{c1}j(Z),j(Z)h{\displaystyle \frac{\mathrm{cosh}(Z/\lambda )}{\mathrm{cosh}(D/\lambda )}}`$with $`h`$ being the dimensionless measure of current density.
Let initially the vortex pierces the plate in Z-direction being described with $`R=\{X(Z,0)=0,0,Z\}`$ . It has similar orientation soon after nucleation near the edge of a real finite plate. Then, due to obvious spatial symmetry, the vortex will remain inside the XZ-plane $`Y=0`$ and keep only one intersection with any of XY-planes. In this situation the Eq.4 can be applied and, besides, reduced to only equation for X-coordinate, $`X=X(Z,t)`$ , as a function of time and Z -coordinate:
$$\frac{X}{t}=\mu \sqrt{1+X^2}\frac{\delta E}{\delta X(Z)}$$
(7)
The energy can be expressed as
$$E=E_s+E_i=E_s\frac{\epsilon }{\lambda }X(Z,t)j(Z)𝑑Z$$
(8)
where the integral represents the energy $`E_i`$ of vortex interaction with transport current (this expression differs only by some constant from the general $`E_i`$ representation ). In the local approximation (5), the Eq.7 looks as strongly nonlinear diffusion-type equation
$$\frac{X}{t}=u_0\left[\lambda \frac{X^{\prime \prime }}{1+X^2}+\sqrt{1+X^2}j(Z)\right]$$
(9)
with notation $`X^{\prime \prime }^2X/Z^2`$ and characteristic velocity $`u_0`$ introduced in Sec.2.
Here the left side is responsible for the friction, and two terms on the right-hand side represent the self-action force and transport current-induced Lorenz force, respectively. Clearly, because of the latter force both the vortex ends will forwardly move in one and the same direction (to opposite edge of the plate), while the middle of vortex will be more or less backward, and the larger is transport current the longer should be the distance $`\mathrm{\Delta }X=X(\pm D,t)X(0,t)`$ (below termed vortex stretching).
Some predictions of further vortex behaviour can be deduced merely from the energy expression (8). Just after start the middle is still in rest. As the Eq.8 shows, in thick plate ($`D>>\lambda `$) the unit displacement of every end leads to the $`E_i`$ ’s decrease by $`hϵ`$ . At the same time, the corresponding lengthening of each of two symmetrical core branches results in the $`E_s`$ ’s increase by $`ϵ`$ per unit length. Consequently, if $`h>1`$ then the total energy decreases, and the vortex stretchening along the drift direction becomes profitable. The lengthening process should last until the curvature of the most backward central part of the core becomes so large that the self-action force makes this part moving as quickly as the ends do.
Thus, at $`h>1`$ the vortex gets over the friction like elastic in water. Oppositely, at $`h<1`$ the stretchening is energetically unprofitable, and the vortex should move as deformed flexible stick. The transition from this stick-like behaviour to elastic-like one, when transport current increases from $`h<1`$ to $`h>1`$ , is the example of so-called ”nonequilibrium phase transitions”.
Let us consider the steady drift of vortex as a whole, without change of shape. The corresponding solution on the Eqs. 7 or 9 can be written as $`X(Z,t)=ut+X(Z)`$ . The stationary shape $`X(Z)`$ and the drift velocity $`u=u(h,D)`$ should be obtained from (7) or (9) with the help of above discussed orthogonality boundary conditions
$`{\displaystyle \frac{dX}{dZ}}(\pm D)=0`$Besides, due to the mirror symmetry, the condition $`X^{}(\pm 0)=0`$ should be satisfied.
In this steady nonequilibrium state the self-energy $`E_s`$ is constant, therefore, the work $`M_j`$ produced by transport current per unit time,
$`M_j={\displaystyle \frac{dE_i}{dt}}={\displaystyle \frac{u\mathrm{\Phi }_0}{c}}{\displaystyle _D^D}J_y(Z)𝑑Z=2uh\epsilon \mathrm{tanh}(D/\lambda )`$coinsides with the energy dissipation per unit time $`M_d`$ . In accordance with (1) and (2),
$`W_d={\displaystyle \frac{1}{\mu }}{\displaystyle \left|\frac{R}{t}\right|^2𝑑L}={\displaystyle \frac{1}{\mu }}{\displaystyle \left(\frac{u}{Q}\right)^2Q𝑑Z}=2\delta u^2/\mu `$where the notations
$`Q\sqrt{1+X^2}={\displaystyle \frac{dL}{dZ}},\delta {\displaystyle \underset{0}{\overset{D}{}}}{\displaystyle \frac{dZ}{Q}}`$are introduced. We took into account that actual displacement of the core always is locally perpendicular to its orientation. Only such displacements are physically meaningful and really cause the friction. Therefore, the drift velocity and the local core velocity are connected by the relation
$`\left|{\displaystyle \frac{R}{t}}\right|={\displaystyle \frac{u}{Q}}`$Evidently, the factor $`Q`$ determines at one and the same time local orientation of the core and degree of its stretching.
Hence, the equality $`M_d=M_j`$ yields
$$U\frac{u}{u_0}=\frac{h\lambda }{\delta }\mathrm{tanh}(D/\lambda )\frac{h\lambda }{\delta }$$
(10)
In view of above reasonings, at $`h<1`$ the vortex stretchening is weak, therefore, $`X^2`$ is comparable with unit, $`Q1`$, $`\mathrm{\Delta }XD`$ and $`\delta D`$. Then the Eq.10 shows that in this stick-like regime $`Uh\lambda /\delta h\lambda /D<<h`$ , i.e. the drift velocity is inversely proportional to the plate thickness. This is quite natural, because the surface current-induced Lorentz force acts only on the ends, while the friction almost equally acts on any core fragment.
In general, the parameter $`\delta `$ serves as the effective plate half-thickness. Obviously, always $`\delta <D`$ . In the stretched elastic-like regime in thick plate anywhere at $`D|Z|>>\lambda `$ the inequalities $`\left|X^{}\right|>>1`$ and $`Q>>1`$ take place. Hence, $`\delta <<D`$ and what is essential it becomes almost insensitive to thickness. As a consequence, both the drift velocity and mobility $`u/h`$ strongly increase as compared with stick-like regime and both become independent on thickness (below we shall see that $`\delta \lambda `$ and $`Uh`$ , i.e. $`U`$ becomes approximately $`D/\lambda `$ times larger). According to the $`M_d`$’s expression, the matter is that thought the energy dissipation $`Q`$ times increases due to the core lengthening this effect is overpowered by its $`Q^2`$ decrease because of $`Q`$ times decrease of the factual core velocity $`\left|\frac{R}{t}\right|`$ . As the result, the vortex stretching leads to smaller friction and smaller entropy production, under fixed vortex velocity, and to larger velocity ander fixed transport current. The picture looks as if most part of core slides along itself, but this process does not mean a real motion of core and so does not cause a friction and dissipation.
## VI DRIFT OF THE VORTEX ENDS
To be convinced in what was said, let us consider vortex shape in the stretched elastic-like regime. Because at $`\mathrm{\Delta }X>>D`$ most part of the core inevitably has a small curvature, it can be considered with neglecting the self-action. Then any of the Eqs.7 and 9 reduces to
$$U\sqrt{1+X^2}j(Z)$$
(11)
Here from the characteristic exponential asymptotics does follow,
$$X(Z)\frac{\lambda U}{h}\left[\mathrm{exp}\left(\frac{DZ}{\lambda }\right)1\right]$$
(12)
Here $`Z>0`$ , $`X(Z)=X(Z)`$ , and for definitness the position $`X=0`$ is prescribed to the end fragments. It is easy to verify that corresponding self-action contribution in the Eq.9 indeed is negligibly small as compared with the current-induced force.
We can get a rough estimate of the stretching if put on $`Z=0`$ in (12) and take into account that $`U>h\lambda /D`$. Then the Eq.12 yields
$`\mathrm{\Delta }X/\lambda {\displaystyle \frac{U}{h}}\mathrm{exp}\left({\displaystyle \frac{D}{\lambda }}\right)>{\displaystyle \frac{\lambda }{D}}\mathrm{exp}\left({\displaystyle \frac{D}{\lambda }}\right)`$Hence, $`\mathrm{\Delta }X/\lambda `$ possesses exponentially strong dependence on $`D/\lambda `$ , and it can be giantly large if $`D`$ exceeds $`\lambda `$ by an order of value or more.
In view of this circumstance, the ratio $`\mathrm{\Delta }X/W`$ with $`W`$ being the width of a real finite plate, becomes of principal importance. Clearly, if $`\mathrm{\Delta }X>>W`$ then the steady drift of the vortex as a whole is impossible: the ends of vortex will achieve the opposite edge before the displacement of its backward central part will be comparable with $`W`$ (all the more, before the velocity of this part becomes equal to that of the ends).
Consider the drift of the ends in such a non-stationary situation. Because the vortex lengthening is profitable, this drift can do independently on the motion of deepened backward part, as if thickness was infinitely large ($`D/\lambda \mathrm{}`$) . To estimate the drift velocity, let us multiply the Eq.7 or 9 (with $`\frac{X}{t}u`$ ) by $`Q^1`$ and integrate over variable $`z=DZ`$ from zero to infinity, with the condition $`X^{}(z\mathrm{})=\mathrm{}`$ which evidently corresponds to infinitely far backward center. Then both the Eqs.7 and 9 result in
$`U={\displaystyle \frac{\lambda }{\delta }}(h1),\delta ={\displaystyle _0^{\mathrm{}}}{\displaystyle \frac{dz}{Q}}`$
To evaluate this integral, note that in accordance with the orthogonality boundary conditions the shape of the end fragments of the core is parabolic, for instance, at upper end $`X(Z)=X(D)(ZD)^2/2\rho `$ , with $`\rho `$ being the curvature radius at the end point. It follows from the Eq.9 that $`\lambda /\rho =hU`$ . In this parabolic region the integration divergers but becomes cut after transition to exponential asymptotics (12). The estimate of the integral leads to approximate equation
$$Uh(h1)/\{h1+\mathrm{ln}[2z_0(hU)/\lambda ]\}$$
(13)
where $`z_0`$ is the depth of the crossover point, $`z_04\lambda `$ . The Eq.12 helps to estimate the end drift velocity in thick plate. Obviously, it turns into zero at $`h0`$, in agreement with $`D\mathrm{}`$ limit of the estimate for stick-like regime. It can be shown that velocity of the steady drift of the vortex as a whole is only slightly smaller differing by a multiplier of order of unit.
## VII GIANT VORTEX STRETCHING
The exponentially large vortex stretching is the most significant possibility of vortex evolution in thick plate, as well as in bulk samples in general. Therefore it would be useful to more correctly justify the above simplific estimate of $`\mathrm{\Delta }X`$ .
Note that $`\mathrm{\Delta }X>L/2D`$ . Divide both sides of (9) by $`j(Z)`$ and integrate from zero to $`D`$ . This results in
$`L/2=BA,BU{\displaystyle \frac{dZ}{j(Z)}},A{\displaystyle \mathrm{arctan}(X^{})\left|\frac{d}{dZ}\frac{\lambda }{j(Z)}\right|𝑑Z}`$It is easy to notice that
$`A<A_0{\displaystyle \frac{\pi \lambda }{2}}\left[{\displaystyle \frac{1}{j(0)}}{\displaystyle \frac{1}{j(D)}}\right]`$so $`\mathrm{\Delta }X>BA_0D`$. The calculation of integral $`B`$ gives
$`B={\displaystyle \frac{2U\lambda }{h}}\mathrm{cosh}(D/\lambda )\{\mathrm{arctan}[\mathrm{exp}(D/\lambda )]{\displaystyle \frac{\pi }{4}}\}`$Then, after simplifications possible due to $`D>>\lambda `$, one finally obtains
$$\mathrm{\Delta }X/\lambda >\frac{\pi (Uk)}{4h}\mathrm{exp}(D/\lambda )$$
(14)
with $`k<1`$ . Because $`u`$ is monotonously growing function of $`h`$ , the coefficient in front of exponent is positive if $`h`$ exceeds some level larger than unit, for example, if $`h>2`$ . Hence, ate least at $`h>2`$ the vortex stretching is exponentially strong.
This estimate is obtained in the framework of local approximation. More correct estimate should give a lesser value, because of self-attraction of the core in middle part of the plate where two symmetrical exponential tails described by (12) meet one another and form an arc. Such a non-local effect is most essential just under the specific plate geometry. However, the non-local correction can not change the shape of the front vortex part where the non-local interaction is weak as compared with other forces. It is not hard to show that the latter requirement is satisfied if $`|Z|>Z_0`$, where $`Z_0`$ is the solution on equation
$`h\mathrm{exp}[(D\left|Z\right|)/\lambda ]\sqrt{\lambda /2\pi \left|Z\right|}\mathrm{exp}(2\left|Z\right|/\lambda )`$If take into account that the more is $`h`$ the less is $`Z_0`$, then this equation yields $`Z_0<D/3`$ . Hence, at $`|Z|>D/3`$ the mutual attraction of two core branches can be neglected, and the asymptotics (12) remains valid. This means that the maximally possible effect of non-locality is the replacing $`D`$ in the exponent by $`\alpha D`$ with $`\alpha >2/3`$ . Consequently, the lower bound for the stretching with confidence can be estimated as
$`\mathrm{\Delta }X/\lambda >{\displaystyle \frac{\lambda }{D}}\mathrm{exp}(2D/3\lambda )`$Thus, even in the worst case the non-local effects do not abolish the exponential character of stretching.
For example, if $`\lambda 310^5cm`$ and $`h`$ equals to a few units, then even at $`D20\lambda <10^3cm`$ one gets $`\mathrm{\Delta }X>m\lambda \mathrm{exp}(2D/3\lambda )`$ , with $`m1`$ , i.e. $`\mathrm{\Delta }X>1cm`$ what exceeds a width of any realistic sample. Thus at first the vortex should form a ring whose shape approximately copies that of the sample cross-section. During this process the velocity of backward deepened core part is primarity determined by its distorsion which is created in the beginning of stretchening and thus has curvature radius of order of $`D`$ . Hence, this velocity is of order of $`u_0\lambda /D`$ , and at the moment when the ends will meet one another the displacement of most backward point will be yet as small as $`\lambda W/Dh<<W`$ .
## VIII DISCUSSION AND RESUME
It seems clear that both the above conclusions can be extended to bulk current-carrying superconductors with another geometry, for instance, to round wires, if treat $`2D`$ and $`W`$ as minimal and maximal diameters of cross-section of the wire, respectively.
Due to possibility of giant deformation and stretching of vortices, the thermodynamically nonequilibrium process of vortex penetration can promote formation of complicated many-vortex dynamical configurations which seem rather strange and unprofitable from the point of view of equilibrium thermodynamics. The presence of an external magnetic field parallel to transport current should lead to formation of spiral-like configuration instead of ring-like one and thus especially ensure the entangling of vortices. The Eq.4 enables us to describe this scenario in details, if choose Z-axis to be directed along the wire. Besides, the presence of weak pinning should amplifier the stretching of vortex and additionly complicate its shaping, because the motion of deepened part of vortex is characterized by relatively small forces of order of $`ϵ\lambda /D`$ (much smaller than forces $`ϵ`$ what act on the end fragments) and so may be easily held back by pinning centers.
We would like to underline the role of orthogonality boundary conditions. In the work the equation was under use similar to our Eq.9, but boundary conditions was formulated in terms of the tension of core line. One can see from that such conditions make it impossible to consider the case of high surface transport current $`>H_{c1}`$ corresponding to the elastic-like regime.
To resume, we formulated the invariant equations of viscous motion of arbitrarily shaped 3D vortex lines, and applied them to careful analysis of the scenario of vortex penetration into a thick superconducting sample. As it was argued, the vortex-vortex interaction does not significantly affect the penetration process. But, of course, a full description of resistive state leads to more complicated tasks about vortex-vortex interactions deep inside the sample.
ACKNOWLEDGEMENTS
I would like to thank Dr. M.Indenbom, Dr. Yu.Genenko and Dr. A.Radievskiy for usefull discussions.
REFERENCES
1. M.V.Indenbom, C.J.van der Beek, V.Berseth, W.Benoit, G.D’Anna,
A.Erb, E.Walker and R.Flukiger, Nature, 1997, Feb.20 .
2. Yu.A.Genenko, Phys.Rev., B 49, 1994, 6950.
3. Chao Tang, Shechao Feng and L.Golubovich, Phys.Rev.Lett., 472, 1994, 1264.
4. L.P.Gorkov and N.B.Kopnin, Uspekhi fizicheskikh nauk, 116, 1975, 411 (transl. in English in Sov.Phys.-Usp., 1975).
5. Yu.E.Kuzovlev, Physica, C 292, 1997, 117.
6. V.P.Galaiko, Zh.Teor.Eksp.Fiz., 50 , 1966, 1322.
|
no-problem/9904/cond-mat9904383.html
|
ar5iv
|
text
|
# Charge stripes seen with x-rays in 𝐋𝐚_1.45𝐍𝐝_0.4𝐒𝐫_0.15𝐂𝐮𝐎_𝟒
## I Introduction
The ordering of holes and spins into stripes in the CuO<sub>2</sub>-planes of Nd-doped La<sub>2-x</sub>Sr<sub>x</sub>CuO<sub>4</sub> (LSCO) has attracted much attention, because charge separation and antiferromagnetic spin fluctuations are believed to be important for understanding the mechanism of high temperature superconductivity. The idea of pinned stripes in the low temperature tetragonal (LTT) phase of Nd-doped LSCO or La<sub>2-x</sub>Ba<sub>x</sub>CuO<sub>4</sub> provides an explanation for the anomaly observed at $`x\frac{1}{8}`$, where superconductivity is destroyed, or at least strongly suppressed . The tilting of the oxygen octahedra along the $`[\mathrm{1\hspace{0.17em}0\hspace{0.17em}0}]`$ and $`[\mathrm{0\hspace{0.17em}1\hspace{0.17em}0}]`$ directions (i.e., parallel to Cu–O bonds) in the LTT phase introduces a pinning potential for horizontal and vertical stripes, whereas in the low temperature orthorhombic (LTO) phase it is absent due to the rotation of the tilting axis into the $`[\mathrm{1\hspace{0.17em}1\hspace{0.17em}0}]`$ direction.
In a single crystal of La<sub>1.48</sub>Nd<sub>0.4</sub>Sr<sub>0.12</sub>CuO<sub>4</sub>, neutron diffraction allowed the observation of both magnetic and charge-ordering superstructure reflections . Consistent with the idea of pinning by the LTT lattice modulation, static stripe ordering within the CuO<sub>2</sub> planes appears at the transition temperature from the LTO to the LTT phase, which is $`68\mathrm{K}`$. From the positions of the superlattice peaks and the nominal hole concentration it follows that stripes of holes are approximately half-filled, and act as antiphase domain walls with respect to the antiferromagnetically ordered Cu spins. Thus, a Sr doping level of $`\frac{1}{8}`$ yields a spacing between stripes of $`4a`$, where $`a`$ is the lattice constant. High energy x-ray studies have been succesful in confirming the results for the charge stripe ordering in an $`x=0.12`$ sample .
It is still an open question as to whether charge stripes are limited to hole concentrations near $`\frac{1}{8}`$ in Nd-doped LSCO or whether they influence the physics of cuprate superconductivity in general, but there is accumulating evidence for the latter. Inelastic neutron scattering experiments indicate the possibility of moving, fluctuating stripes in LSCO and even in YBCO . Local charge ordering in La<sub>2-x</sub>Sr<sub>x</sub>CuO<sub>4</sub> with $`x\frac{1}{8}`$ is suggested by a recent nuclear quadrupole resonance (NQR) study . It is interesting to imagine that dynamic stripe correlations might be necessary for superconductivity , whereas pinning the stripes leads to a strongly reduced critical temperature $`\mathrm{T}_\mathrm{c}`$, but the experimental evidence for such a scenario is still incomplete. So far the Nd-doped LSCO system offers a unique opportunity to study pinned stripe patterns in diffraction experiments, giving information about the nature of charge and spin ordering.
The superstructure reflections due to the ordering of holes are shifted by $`2ϵ`$ in $`h`$\- or $`k`$-direction relative to fundamental reflections, whereas the magnetic peaks are located around the antiferromagnetic peak position $`(\frac{1}{2},\frac{1}{2},0)`$, shifted by $`ϵ`$ also in $`h`$\- or $`k`$-direction. In neutron scattering experiments, the splitting $`ϵ`$ has been observed to increase slightly as the Sr content increases from 0.12 to 0.15 and 0.20, implying a decrease in the average stripe spacing; however, it has only been practical to study the superstructure reflections due to the antiferromagnetic ordering . It is now imperative to directly characterize the charge order at Sr concentrations away from $`\frac{1}{8}`$.
In this work we present x-ray studies with 120 keV photons on $`\mathrm{La}_{1.45}\mathrm{Nd}_{0.4}\mathrm{Sr}_{0.15}\mathrm{CuO}_4`$. Our results are in complete agreement with the stripe model and complementary to the experimental evidence of antiferromagnetic stripe ordering in this sample . As observed in the $`x=0.12`$ case , the superstructure reflections due to the ordering of the holes sets in at a higher temperature compared to the magnetic signal, indicating that the transition into the stripe phase is driven by the charge separation .
## II Experiment
The experiments have been performed on the triple-axis diffractometer designed for the use of $`100`$ keV photons at the high-field wiggler beamline BW5 at HASYLAB, Hamburg . X-ray diffraction in this energy range has proven to be very successful in studying charge ordering in cuprates, nickelates and manganates . As in neutron scattering experiments, the large penetration depth ($`1`$ mm) allows one to probe the bulk of the sample, enabling a direct comparison of x-ray and neutron diffraction data. In contrast to former experiments on $`\mathrm{La}_{1.48}\mathrm{Nd}_{0.4}\mathrm{Sr}_{0.12}\mathrm{CuO}_4`$ , the Si/TaSi<sub>2</sub> monochromator and analyzer crystals have been replaced by the new Si<sub>1-x</sub>Ge<sub>x</sub> gradient crystal material. These crystals show very high reflectivity values of 96% (not corrected for absorption) and variable widths of the rocking curves depending on the Ge content . By use of this new material, the scattered intensity at the stripe peak positions in Nd doped LSCO is about four times higher compared to the results obtained previously with the utilization of $`\mathrm{Si}/\mathrm{TaSi}_2`$ crystals as monochromator and analyzer . Figure 1 shows a comparison of a superstructure reflection measured with the $`\mathrm{Si}/\mathrm{TaSi}_2`$ and Si<sub>1-x</sub>Ge<sub>x</sub> gradient crystals. The count rate collected with the gradient crystals is nearly two times higher, and a slightly better signal to background ratio could be reached by reducing the beam spot on the sample, i.e. by probing only the center of the crystal. The lower curve has been measured with the $`\mathrm{Si}/\mathrm{TaSi}_2`$ crystals, illuminating a two times larger sample volume.
In this experiment 120 keV photons have been employed, with a monochromatic beam intensity of $`1.2\times 10^{11}`$ photons/$`\mathrm{mm}^2`$. The resolution (FWHM) at a (2,0,0) reflection of $`\mathrm{La}_{1.45}\mathrm{Nd}_{0.4}\mathrm{Sr}_{0.15}\mathrm{CuO}_4`$ was 0.020 Å<sup>-1</sup> in the longitudinal and 0.0014 Å<sup>-1</sup> in the transverse direction, the latter being limited by the sample mosaicity. At smaller diffraction angles the longitudinal resolution improves because the diffraction geometry becomes less dispersive, e.g. the FWHM of the (1,1,0) reflection is 0.011 Å<sup>-1</sup>. A closed-cycle cryostat has been utilised and temperatures between 9 K and 300 K could be reached at the sample position. The studied crystal ($`2\times 2\times 4`$ mm<sup>3</sup>) is a piece of a cylindrical rod that was grown by the travelling-solvent floating-zone method.
## III Results
Previous hard x-ray diffraction experiments on a $`\mathrm{La}_{1.48}\mathrm{Nd}_{0.4}\mathrm{Sr}_{0.12}\mathrm{CuO}_4`$ crystal have shown that the stripe peaks are displaced not only within the $`(h,k,0)`$ plane, but also in the $`\mathrm{}`$ direction . Due to the better resolution in reciprocal space compared to neutron scattering experiments it has been possible to find a modulated intensity of the stripe peaks along the $`\mathrm{}`$-direction with maxima at $`\mathrm{}=\pm 0.5`$, indicating a correlation of the charge stripes in next-nearest-neighbor layers along the $`c`$-axis. The highest scattering intensities for the stripe peaks are expected at the positions $`(22ϵ,0,\pm 0.5)`$ and $`(2+2ϵ,0,\pm 0.5)`$. Figure 2 clarifies the positions of the stripe peaks in reciprocal space relative to the CuO<sub>2</sub> planes. We have used the tetragonal unit cell with $`a=b=3.775`$ Å, $`c=13.10`$ Å.
Longitudinal scans along $`(h,0,0.5)`$ are shown in Fig. 3. As anticipated, small superstructure reflections are observed at $`h=2\pm 2ϵ`$ with $`2ϵ=0.256(1)`$. The peaks that are present at $`T=9`$ K have disappeared after raising the temperature to 70 K. The signal to background ratio is $`0.2`$ for the $`22ϵ`$ reflection and $`0.1`$ at the $`2+2ϵ`$ position. In both scans the background rises due to the vicinity of the (2,0,0) reflection. At this fundamental Bragg reflection the count rate in the peak maximum is about $`10^8`$ times larger than in the stripe peaks.
The curves through the data points in Fig. 3 are least-squares fits. In both (a) and (b) the charge-order peaks are modelled with a Gaussian. The background in (a) is approximated by the tail of a second Gaussian centered near $`h=2`$ plus a linear contribution, while only a linear background is used in (b). Usually, the tail of a fundamental reflection is Lorentzian shaped, but nevertheless it is possible that locally other functions are a better approximation to a non-linear background, since the shape of the background is related to the sample quality. The studied $`\mathrm{La}_2\mathrm{CuO}_4`$ crystal with Sr and Nd as dopants on the La-site certainly incorporates defects and strain which are responsible for the rather high background, which is $`150`$ photons per second at the superlattice positions.
In the inset of Fig. 3(a), two more low-temperature longitudinal scans, slightly shifted in the $`k`$-direction, are displayed. The absence of any peaks in these scans indicates that the peak found at $`(22ϵ,0,0.5)`$ is narrow in $`k`$. This conclusion is confirmed by the 9 K transverse scan shown in Fig. 4. The background in the transverse scans is linear, and a fit of the stripe peak with a Lorentzian is slightly more successful than utilizing a Gaussian. From the present data, it is difficult to determine the true peak shape, given the small signal-to-background ratio.
Figure 5(a) shows the temperature variation of the amplitudes of the $`h`$ and $`k`$ scans at the $`(22ϵ,0,0.5)`$ position. With rising temperature the amplitudes decrease rather linearly; the peak vanishes at 62(5) K, which is $`8`$ K below the structural transition from the LTT into the LTO phase. The intensity of the (3,0,0) reflection, also shown in Fig. 5(a), is a measure of this transition because it only occurs in the LTT phase. From the longitudinal and transverse scans a FWHM of 0.028(5) r.l.u. can be inferred, resulting in a correlation length of 43(8) Å. The change of the peak widths with temperature can be seen in Fig. 5(b). Up to $`55`$ K it is rather constant, but above this temperature the width of the peak in $`h`$\- and $`k`$-directions increases considerably. Note that the increasing width as well as the vanishing peak intensity are not directly connected to the structural transition from the LTT into the LTO phase, but occur at a remarkably lower temperature.
## IV Discussion
Our x-ray results provide direct evidence for charge-stripe order in $`\mathrm{La}_{1.45}\mathrm{Nd}_{0.4}\mathrm{Sr}_{0.15}\mathrm{CuO}_4`$. Such charge order had previously been inferred from the magnetic order observed by neutron scattering . This is an important result, because high-field magnetization and muon-spin-rotation ($`\mu `$SR) studies have indicated that bulk superconductivity also exists in this sample. Another conclusion of the $`\mu `$SR study is that static magnetic order is present throughout essentially the entire sample volume. Thus, it appears that stripe order and superconductivity coexist intimately at $`x=0.15`$.
Compared to the $`x=0.12`$ composition, both the charge ordering temperature and the superstructure peak intensities are reduced. The superlattice intensities \[normalized to the (200) peak\] are roughly 30% weaker in $`x=0.15`$ relative to $`x=0.12`$. A similar reduction in magnetic ordering temperature and intensity have been observed previously. At both Sr concentrations, charge ordering appears at a higher temperature than magnetic order.
Within the error bars the correlation length of the stripe order in the $`x=0.15`$ sample is the same as in the $`x=0.12`$ sample, and in both samples the correlation length of the stripes is constant over a large temperature range. In contrast to this finding, it is important to note that neutron scattering and muon-spin rotation experiments revealed that the spins are only quasi-static, with a decreasing correlation length, in the temperature range above 30 K in the $`x=0.12`$ sample . This continuous loss of correlation is only observed in the magnetic signal, whereas the correlation length for the charge order shows a decrease only very close to the disordering temperature. The temperature range in which a residual signal with an increased width is observed can be interpreted with fluctuating stripes, since the detected signal in our x-ray diffraction experiment is a sum of elastic and inelastic contributions.
One might conclude that fluctuations of stripes set in if the pinning potential in the LTT phase is destroyed at the structural transition from LTT to LTO, as observed in the x = 0.12 sample, or if rising temperature competes with a pinning potential via thermal activation. In the sample with $`x=0.15`$, the second possibility holds, since the suppression of the stripe ordering is not connected to the structural transition. The pinning potential is smaller in the $`x=0.15`$ sample than in the $`x=0.12`$ sample because the tilting angle of the oxygen octahedra is reduced with increasing Sr content in Nd doped LSCO .
On the other hand, the idea that the LTT structure is required for charge-stripe order is challenged by the interpretation of the Cu NQR results in LSCO by Hunt et al. . That work suggests that static stripes can occur even within the LTO phase, and that the ordering temperature continues to increase as $`x`$ decreases below 0.12. As an initial test of this picture, we intend to study charge order in a Nd-doped crystal with $`x=0.10`$ in the near future.
To summarize, we have been successful in validating the existence of stripe order in $`\mathrm{La}_{1.45}\mathrm{Nd}_{0.4}\mathrm{Sr}_{0.15}\mathrm{CuO}_4`$, a sample which is superconducting below $`10`$ K. Charge-order peaks due to stripe ordering have been observed at $`(22ϵ,0,\pm 0.5)`$, $`(2+2ϵ,0,0.5)`$, and at $`(0,22ϵ,\pm 0.5)`$. The $`2ϵ`$ value of 0.256(1) is in very good agreement with neutron scattering results for the magnetic peaks. This is a further step in establishing the picture of static stripes in Nd doped LSCO. The intensity of the stripe signal decreases almost linearly with rising temperature and vanishes at 62(5) K, $`8`$ K below the structural transition from the LTT to the LTO phase. The formation of the stripe pattern in $`\mathrm{La}_{1.45}\mathrm{Nd}_{0.4}\mathrm{Sr}_{0.15}\mathrm{CuO}_4`$ is less pronounced than in $`\mathrm{La}_{1.48}\mathrm{Nd}_{0.4}\mathrm{Sr}_{0.12}\mathrm{CuO}_4`$.
Acknowledgements:
Work at Brookhaven is supported by Contract No. DE-AC02-98CH10886, Division of Materials Sciences, U.S. Department of Energy.
|
no-problem/9904/cond-mat9904399.html
|
ar5iv
|
text
|
# Excitation of a Dipole Topological Mode in a Strongly Coupled Two-Component Bose-Einstein Condensate
## I Introduction
An intriguing aspect of Bose-Einstein condensation (BEC) in a dilute atomic gas is that the internal atomic state of the condensate can be manipulated to produce quite novel systems. A number of interesting experiments have produced uncoupled multi-component condensates, in which two or more internal states of the condensate exist together in a magnetic or optical trap . These experimental studies, along with their theoretical counterparts, have investigated various topics such as the ground state of the system , the elementary excitations , and the nonlinear dynamics of component separation .
Many fascinating properties can be studied by applying an external electromagnetic field that coherently couples the internal atomic states of the condensate . In the experiment described in , the relative phase between two hyperfine components was measured using a technique based on Ramsey’s method of separated oscillating fields and these results have motivated further theoretical investigation . In this system, several key parameters can be varied over a wide range of values, such as the coupling-field intensity and frequency, the confining potentials, the total number of atoms, and the temperature, making this a very rich system to explore.
Several theoretical papers have investigated the weak coupling limit of this system, where the intensity of the external driving field is very low but is turned on for a time long compared to the period of oscillation in the magnetic trap. A clear analogy exists between this system and the Josephson junction . In the Josephson junction, identical particles in spatially separated condensates are coupled via the tunneling mechanism . In this two-component system, however, two distinct internal states of the condensate are coupled by an applied field. By adjusting the magnetic fields that confine the atoms, the degree of spatial overlap between the two components can be controlled.
Recently, there have been both experimental and theoretical studies on the dressed states of a driven two-component condensate, drawing an analogy to the dressed states of a driven two-level atom in quantum optics. Due to the interplay between the internal and external degrees of freedom, the condensate dressed states have spatial structure that depends on the trap parameters, the mean-field interaction, and the frequency and intensity of the driving field . In the experiment reported in , the dressed states were created via an adiabatic passage by sweeping the detuning.
In this paper, we focus on the limit of a very strong and sustained coupling between hyperfine states, which is the situation achieved experimentally in . In that experiment, a BEC of about $`8\times 10^5`$ atoms was produced in the $`|F=1,M_F=1`$ hyperfine state of <sup>87</sup>Rb, at a temperature close to zero, $`T0`$. The atoms were confined in a time-averaged, orbiting potential (TOP) magnetic trap by a harmonic potential with axial symmetry along the vertical axis. An external field was then applied that coupled the $`|1,1`$ state to the $`|2,1`$ hyperfine state via a two-photon transition. The Rabi frequency was five to ten times larger than the vertical trap frequency and the detuning could be adjusted arbitrarily. Due to their different magnetic moments and the force of gravity, the two hyperfine states sit in shifted traps offset along the vertical axis . The degree of separation could be controlled by adjusting the magnetic trapping fields.
The subsequent behavior of the system described in was quite unexpected: after the coupling field was turned on, the Rabi oscillations between the states appeared to collapse and revive on a time scale which was long compared to the Rabi period of 3 ms. An example of this behavior taken from is shown in Fig. 1. It was observed that the period of this modulation increased with decreasing detuning. The behavior of the system also depended critically on the separation between the traps for each state. As the separation was taken to zero, the effect went away. The evolution of the density of each component was also very interesting. Each component cycled between a density profile with one peak and a profile with two peaks. The two-peaked structure was most clearly visible around the collapse time, which is $`t20`$ ms for the case shown in Figure 1.
A numerical calculation of the three-dimensional, coupled Gross-Pitaevskii (GP) equations Eq. (1) (below) describing the system in the zero temperature limit agrees, at least qualitatively, with the outcome of the experiment, as shown in Figure 1. Of course, the agreement between the numerical integration of Eq. (1) and the laboratory data does not in itself provide an intuitive explanation of the underlying physical mechanism responsible for the collapse-revival behavior. In this paper, we present a detailed analysis of this problem, and arrive at a rather simple model that explains the major features of the system’s behavior.
Before presenting the details of our analysis, it is useful to first give an overview of the results. There are two main concepts that play key roles in obtaining an intuitive understanding of this problem. First, there is a clear separation of time scales: the period of Rabi oscillations between internal states is much shorter than the period of the trap. That is, the internal dynamics occur on a much shorter time scale than the motional dynamics of the system. It is therefore useful to go to a frame rotating at the effective Rabi frequency. In this rotating frame, we show that there exists a weak coupling between the low lying motional states which is proportional to the offset between the two traps. This weak coupling has the effect of modulating the amplitude of the fast Rabi oscillations in the lab frame.
The second key point is to understand exactly which motional states are excited. They are not the linear response collective excitations (normal modes) that have been studied frequently in the BEC literature . Instead, they are many-particle topological modes determined by the self-consistent solutions to the two-component GP equations Eq. (25). The well known vortex mode is one example of such an excitation in which phase continuity requires quantized circulation around a vortex core. The related excitation which plays a key role in this paper does not exhibit circulation but has a node in the wavefunction amplitude and exhibits odd-parity behavior characteristic of such a dipole mode. For a single component in the limit of the uniform gas, the exact solution of this mode is known as a dark soliton . In that case the scale of the density perturbation around the node is the healing length and is determined by a balance of kinetic and mean-field interaction energies. In the problem we consider here, however, it is necessary to account for the mean field of the the remaining population in the condensate ground state. Consequently the two modes—the ground state and the dipole mode—are inextricably linked and must be determined self consistently.
We first present a detailed theoretical analysis in Section II. After making several reasonable approximations, we arrive in Section II E at the two-mode model – a simplified description that encapsulates the essential properties of the system. In Section III we present results of numerical calculations that illustrate the behavior of the system and we compare our model with the exact numerical solution of the coupled GP equations. We finally summarize our work in Section IV and suggest further studies based on our understanding of this phenomenon.
## II Theoretical Description
The following theoretical development was motivated by the experiment described in . Therefore, we have not tried to keep our calculations general, but instead have made several assumptions based on that particular situation. However, our approach could easily be extended to treat a broader class of systems. We give a brief discussion in the conclusion of the paper about possible extensions of this work to other interesting systems.
We begin this section by writing down the coupled mean-field equations, valid for zero temperature, that describe this driven, two-component BEC. In Section II B we rewrite the mean-field equation in a direct-product representation that clearly separates out the external and internal dynamics. We then go to a frame rotating at the effective Rabi frequency in Section II C in order to focus on the slower motional dynamics of the system. After making some approximations in Section II D, we finally arrive at the main result of our study in Section II E, the two-mode model.
### A Coupled Mean-field Equations
A mean-field description of this many-body system that includes the atom-field interaction has been developed, which generalizes the standard Gross-Pitaevskii (GP) equation to treat systems with internal state coupling . The resulting time-dependent GP equation describing the driven, two-component condensate is
$$\begin{array}{ccc}i(\begin{array}{c}\dot{\psi }_1\\ \dot{\psi }_2\end{array})& =& (\begin{array}{cc}H_1^0+H_1^{\mathrm{MF}}+\delta /2& \mathrm{\Omega }/2\\ \mathrm{\Omega }/2& H_2^0+H_2^{\mathrm{MF}}\delta /2\end{array})(\begin{array}{c}\psi _1\\ \psi _2\end{array}).\end{array}$$
(1)
The Hamiltonians describe the evolution in the trap $`H_i^0`$ and the mean field interaction $`H_i^{\mathrm{MF}}`$ for each component
$`H_i^0`$ $`=`$ $`{\displaystyle \frac{1}{2}}^2+{\displaystyle \frac{1}{2}}[({\displaystyle \frac{\rho }{\alpha }})^2+(z+\gamma _iz_0)^2]`$ (2)
$`H_i^{\mathrm{MF}}`$ $`=`$ $`N(\lambda _{ii}|\psi _i|^2+\lambda _{ij}|\psi _j|^2),`$ (3)
where $`\gamma _1=1`$ and $`\gamma _2=1`$, and $`z_0`$ is the shift of each trap from the origin along the vertical axis. The factor $`\alpha =\omega _z/\omega _\rho `$ is the ratio of axial and radial trap frequencies. In the experiment reported in , $`\alpha >1`$. The mean-field strength is characterized by $`\lambda _{ij}=4\pi a_{ij}/z_{\mathrm{sho}}`$, which depends on the scattering length $`a_{ij}`$ of the collision. In general there will be three different values, one for each type of collision in this two-component gas: $`a_{11}`$, $`a_{22}`$, $`a_{12}`$. The detuning between the driving field and the hyperfine transition is given by $`\delta `$ while $`\mathrm{\Omega }`$ denotes the strength of the coupling. We work in dimensionless units: time is in units of $`1/\omega _z`$, energy is in terms of the trap level spacing $`\mathrm{}\omega _z`$, and position is in units of the harmonic oscillator length $`z_{\mathrm{sho}}=\sqrt{\mathrm{}/m_{\mathrm{Rb}}\omega _z}`$. The complex functions $`\psi _i(𝐫,t)`$ are the mean-field amplitudes of each component, where $`i=\{1,2\}`$. They obey the normalization condition $`(|\psi _i|^2+|\psi _j|^2)d^3x=1`$. The total population is $`N`$.
The coupled mean-field equations Eq. (1) can be solved numerically using a finite-difference Crank-Nicholson algorithm . We show results of such calculations in Fig. 1 for the three-dimensional solution and in Section III for a one-dimensional version. However, in order to gain a more intuitive understanding of the behavior shown in Fig. 1, we formulate a simplified description of the system in the following section.
### B External $``$ Internal Representation
The coupled mean-field equations Eq. (1) can be rewritten in a more illuminating form by making a clear separation of the external and internal degrees of freedom. The system exists in a direct-product Hilbert space $`=H_{\mathrm{ex}}H_{\mathrm{in}}`$, where $`_{\mathrm{ex}}`$ is the infinite-dimensional Hilbert space describing the motional state of the system in the trap and $`_{\mathrm{in}}`$ is the two-dimensional Hilbert space describing the spin of the system. A general operator in $``$ can be written as a sum over the direct-product of operators from $`_{\mathrm{ex}}`$ and $`_{\mathrm{in}}`$. We rewrite Eq. (1) in this representation as
$$i\frac{}{t}|\psi (t)=[\widehat{H}_0\widehat{1}+\widehat{1}(\frac{\mathrm{\Omega }}{2}\widehat{\sigma }_x+\frac{\delta }{2}\widehat{\sigma }_z)+\widehat{H}_z\widehat{\sigma }_z]|\psi (t)$$
(4)
where $`\{\widehat{1},\widehat{\sigma }_x,\widehat{\sigma }_y,\widehat{\sigma }_z\}`$ are the standard Pauli spin matrices. The state of the system $`|\psi (t)`$ in general has a nonzero projection on the internal states $`|1`$ and $`|2`$, represented by $`\psi _i(𝐫,t)=𝐫|i|\psi (t)`$, where $`i=\{1,2\}`$. The position representations of $`\widehat{H}_0`$ and $`\widehat{H}_z`$ are local, i.e. $`𝐫|\widehat{H}_0|𝐫^{}=H_0(𝐫)\delta (𝐫𝐫^{})`$ and $`𝐫|\widehat{H}_z|𝐫^{}=H_z(𝐫)\delta (𝐫𝐫^{})`$, where $`H_0(𝐫)`$ and $`H_z(𝐫)`$ are given by
$`H_0(𝐫)`$ $`=`$ $`{\displaystyle \frac{1}{2}}^2+{\displaystyle \frac{1}{2}}[({\displaystyle \frac{\rho }{\alpha }})^2+z^2]+\psi (t)|\widehat{P}_𝐫\widehat{\lambda }_+|\psi (t),`$ (5)
$`H_z(𝐫)`$ $`=`$ $`z_0z+\psi (t)|\widehat{P}_𝐫\widehat{\lambda }_{}|\psi (t).`$ (6)
The operator $`\widehat{P}_𝐫`$ is the projector onto the position eigenstates $`\widehat{P}_𝐫=|𝐫𝐫|`$, and the matrix representations of $`\widehat{\lambda }_+`$ and $`\widehat{\lambda }_{}`$ are given as
$`\widehat{\lambda }_+`$ $`=`$ $`{\displaystyle \frac{N}{2}}\left(\begin{array}{cc}\lambda _1+\lambda _{12}& 0\\ 0& \lambda _2+\lambda _{12}\end{array}\right),`$ (9)
$`\widehat{\lambda }_{}`$ $`=`$ $`{\displaystyle \frac{N}{2}}\left(\begin{array}{cc}\lambda _1\lambda _{12}& 0\\ 0& \lambda _{12}\lambda _2\end{array}\right).`$ (12)
Note that the harmonic potential in $`\widehat{H}_0`$ is centered at the origin. The mean-field interaction has been rewritten in terms of a part that acts identically on both components $`\psi |\widehat{P}_𝐫\widehat{\lambda }_+|\psi \widehat{1}`$, and a part that acts with the opposite sign on each state $`\psi |\widehat{P}_𝐫\widehat{\lambda }_{}|\psi \widehat{\sigma }_z`$.
The first two terms in Eq. (4) separately describe the external and internal dynamics of the system, respectively. The third term in Eq. (4), however, couples the internal state evolution to the condensate dynamics in the trap and can lead to interesting behavior. If the term $`\widehat{H}_z`$ were identically zero, then the problem would be completely separable in terms of the external and internal degrees-of-freedom. The term $`\widehat{H}_z`$ would be zero if the trap separation $`z_0`$ were zero and if the scattering lengths were all exactly equal. In fact, for <sup>87</sup>Rb the three scattering lengths are nearly degenerate, so the main effect of $`\widehat{H}_z`$ comes from the term $`z_0z`$, which is the difference in the shifted traps. It causes there to be a spatially varying detuning across the condensate.
### C Rotating Frame
As previously stated, we are concentrating on the situation where the coupling strength is large, so that the frequency of the Rabi oscillations $`\mathrm{\Omega }`$ is significantly larger than the trap frequency $`\nu _z`$. In this case, the internal spin dynamics and the motion of the condensate in the trap occur on two different time scales. Therefore, it is useful to go to a rotating frame that eliminates the second term in Eq. (4) describing the fast Rabi oscillations between the two internal states. In the rotating frame, we will be able to understand more clearly how the third term in Eq. (4), which couples the motional and spin dynamics of the condensate, effects the system on a time scale much longer than the period of Rabi oscillation.
We go to the rotating frame, or interaction picture, by making a unitary transformation using the operator
$$U_I(t)=e^{i\widehat{1}(\frac{\mathrm{\Omega }}{2}\widehat{\sigma }_x+\frac{\delta }{2}\widehat{\sigma }_z)t}.$$
(13)
This can be rewritten in the equivalent form
$$U_I(t)=\widehat{1}(\mathrm{cos}(\mathrm{\Omega }_{\mathrm{eff}}/2t)\widehat{1}\frac{i}{\mathrm{\Omega }_{\mathrm{eff}}}\mathrm{sin}(\mathrm{\Omega }_{\mathrm{eff}}/2t)[\mathrm{\Omega }\widehat{\sigma }_x+\delta \widehat{\sigma }_z]),$$
(14)
where $`\mathrm{\Omega }_{\mathrm{eff}}=\sqrt{\mathrm{\Omega }^2+\delta ^2}`$. The state vector $`|\psi ^{(I)}(t)`$ in the rotating frame is related to the state vector in the lab frame $`|\psi (t)`$ by
$$|\psi ^{(I)}(t)=U_I^{}|\psi (t).$$
(15)
In the rotating frame, the system evolves according to
$$i\frac{}{t}|\psi ^{(I)}(t)=\widehat{H}^{(I)}(t)|\psi ^{(I)}(t),$$
(16)
where $`\widehat{H}^{(I)}(t)`$ is the interaction Hamiltonian
$$\widehat{H}^{(I)}(t)=\widehat{H}_0\widehat{1}+\widehat{H}_z(\alpha _\mathrm{x}(t)\widehat{\sigma }_\mathrm{x}+\alpha _\mathrm{y}(t)\widehat{\sigma }_\mathrm{y}+\alpha _\mathrm{z}(t)\widehat{\sigma }_\mathrm{z}).$$
(17)
Note that $`\widehat{H}_0`$ and $`\widehat{H}_z`$ are unaffected by the unitary transformation to the rotating frame. The time-varying coefficients $`\alpha _\mathrm{x}(t)`$, $`\alpha _\mathrm{y}(t)`$, and $`\alpha _\mathrm{z}(t)`$ are
$`\alpha _\mathrm{x}(t)`$ $`=`$ $`{\displaystyle \frac{\mathrm{\Omega }}{\mathrm{\Omega }_{\mathrm{eff}}}}{\displaystyle \frac{\delta }{\mathrm{\Omega }_{\mathrm{eff}}}}[1\mathrm{cos}(\mathrm{\Omega }_{\mathrm{eff}}t)]`$ (18)
$`\alpha _\mathrm{y}(t)`$ $`=`$ $`{\displaystyle \frac{\mathrm{\Omega }}{\mathrm{\Omega }_{\mathrm{eff}}}}\mathrm{sin}(\mathrm{\Omega }_{\mathrm{eff}}t)`$ (19)
$`\alpha _\mathrm{z}(t)`$ $`=`$ $`{\displaystyle \frac{\delta ^2}{\mathrm{\Omega }_{\mathrm{eff}}^2}}+{\displaystyle \frac{\mathrm{\Omega }^2}{\mathrm{\Omega }_{\mathrm{eff}}^2}}\mathrm{cos}(\mathrm{\Omega }_{\mathrm{eff}}t).`$ (20)
### D Approximations
We now make three simplifications in order to extract out the dominant behavior of the system. We first note that the coefficients $`\alpha _i(t)`$ given in Eq. (20) oscillate rapidly at the Rabi frequency. We expect the system in the rotating frame $`|\psi ^{(I)}(t)`$ to evolve on a much slower time scale than the period of Rabi oscillation. We can utilize this fact in order to simplify the interaction Hamiltonian $`\widehat{H}^{(I)}(t)`$ given in Eq. (17) by taking the average values of the coefficients $`\alpha _i(t)`$ – this is equivalent to coarse graining Eq (16). The coefficients in Eq. (20) become time independent and reduce to: $`\alpha _\mathrm{x}=\delta \mathrm{\Omega }/\mathrm{\Omega }_{\mathrm{eff}}^2`$, $`\alpha _\mathrm{y}=0`$, and $`\alpha _\mathrm{z}=\delta ^2/\mathrm{\Omega }_{\mathrm{eff}}^2`$.
We make the additional assumption that the system is being driven close to resonance, so that $`\delta /\mathrm{\Omega }_{\mathrm{eff}}<<1`$. We therefore set $`\alpha _\mathrm{z}=0`$ since $`\alpha _\mathrm{z}`$ depends quadratically on this small parameter.
Finally, we take advantage of the fact that the scattering lengths for <sup>87</sup>Rb are nearly degenerate, with the ratios between inter- and intra-species scattering lengths given by $`\{a_2:a_{12}:a_1\}=\{0.97:1:1.03\}`$ . This allows us to simplify the two mean-field terms appearing in Eq. (6). We first make the approximation $`\widehat{\lambda }_+\lambda N\widehat{1}`$ by assuming equal scattering lengths, or $`\lambda _1\lambda _2\lambda _{12}\lambda `$. We can also simplify the other term $`\widehat{\lambda }_{}`$ by assuming that its predominant effect is to shift the levels slightly. Instead of neglecting it altogether, we simply replace it by a mean-field shift of the levels $`\psi |\widehat{P}_𝐫\widehat{\lambda }_{}|\psi \delta _{\mathrm{MF}}`$, where the shift is given by $`\delta _{\mathrm{MF}}=\mathrm{\Delta }\lambda n^2(𝐫,0)d^3x/N`$. Here $`n(𝐫,0)`$ is the total density at $`t=0`$ and $`\mathrm{\Delta }\lambda =(\lambda _1\lambda _{12})=(\lambda _{12}\lambda _2)`$. We can absorb it into the detuning $`\delta `$ by defining an effective detuning $`\delta ^{}`$ that includes this mean-field shift $`\delta \delta ^{}=\delta +\delta _{\mathrm{MF}}`$.
After making the above approximations, we can now write the interaction Hamiltonian $`\widehat{H}^{(I)}`$ from Eq. (17) in a much simpler form
$$\widehat{H}^{(I)}=\widehat{H}_0^{}\widehat{1}+\widehat{H}_z^{}\widehat{\sigma }_\mathrm{x},$$
(21)
where the position representations of $`\widehat{H}_0^{}`$ and $`\widehat{H}_z^{}`$ are local, i.e. $`𝐫|\widehat{H}_0^{}|𝐫^{}=H_0^{}(𝐫)\delta (𝐫𝐫^{})`$ and $`𝐫|\widehat{H}_z^{}|𝐫^{}=H_z^{}(𝐫)\delta (𝐫𝐫^{})`$, where $`H_0^{}(𝐫)`$ and $`H_z^{}(𝐫)`$ are given by
$`\widehat{H}_0^{}(𝐫)`$ $`=`$ $`{\displaystyle \frac{1}{2}}^2+{\displaystyle \frac{1}{2}}[({\displaystyle \frac{\rho }{\alpha }})^2+z^2]+\lambda n(𝐫,t),`$ (22)
$`\widehat{H}_z^{}(𝐫)`$ $`=`$ $`\beta z.`$ (23)
The total density is $`n(𝐫,t)=N\psi (t)|\widehat{P}_𝐫\widehat{1}|\psi (t)`$, and $`\beta =z_0\delta \mathrm{\Omega }/\mathrm{\Omega }_{\mathrm{eff}}^2`$. For the typical values of the parameters in the experiment, $`\mathrm{\Omega }=400`$ Hz, $`\delta ^{}=100`$ Hz, $`z_0/z_{\mathrm{sho}}=0.1`$, this factor is rather small $`\beta 0.02`$ in harmonic oscillator units. Note that $`\widehat{H}_0^{}`$ still varies slowly in time through the nonlinear mean-field term, which depends on the density. We refer to this result Eq. (21) as the coarse-grained, small detuning (CGSD) model to distinguish it from the two-mode model presented below, which makes further assumptions.
We have managed to greatly simplify the description of the system by going to the rotating frame. The first term in Eq. (21) contains the kinetic energy, a harmonic potential centered at the origin, and a mean-field interaction term depending on the slowly varying density $`n(𝐫,t)`$. The second term in Eq. (21) represents a very weak coupling between the two internal states $`|1`$ and $`|2`$, and between motional states $`|\varphi _n`$ and $`|\varphi _m`$ via the dipole operator $`z`$. The states $`|\varphi _n`$ and $`|\varphi _m`$ are the instantaneous self-consistent eigenmodes of $`\widehat{H}_0^{}`$. In the next subsection we present a model that assumes only two motional states are coupled, the self-consistent ground state $`|\varphi _0`$ and the self-consistent first-excited state $`|\varphi _1`$, which has odd-parity along the z-axis.
### E Two-mode model
It is useful to define a basis of motional states with which to describe the system in the rotating frame. A natural choice is the set of instantaneous eigenstates of $`\widehat{H}_0^{}`$, which satisfy
$`({\displaystyle \frac{1}{2}}^2+{\displaystyle \frac{1}{2}}[({\displaystyle \frac{\rho }{\alpha }})^2+z^2]+\lambda n(𝐫,t))\varphi _i(𝐫)`$ $`=`$ $`ϵ_i\varphi _i(𝐫)`$ (24)
$`{\displaystyle _{\mathrm{}}^{\mathrm{}}}\varphi _i(𝐫)\varphi _j(𝐫)d^3r`$ $`=`$ $`\delta _{i,j},`$ (25)
where the index $`i`$ refers to all of the relevant quantum numbers that uniquely specify each eigenstate, $`i=\{n_z,n_\rho ,n_\varphi \}`$, given the cylindrical symmetry of the system. In general, many modes can be occupied and the state vector is written
$$|\psi ^{(I)}(t)=\underset{i}{}[c_i(t)|\varphi _i|1+d_i(t)|\varphi _i|2],$$
(26)
where $`\varphi _i(𝐫)=𝐫|\varphi _i`$. The density appearing in Eq. (25) is then
$$n(𝐫,t)=N(|\underset{i}{}c_i(t)\varphi _i(𝐫)|^2+|\underset{i}{}d_i(t)\varphi _i(𝐫)|^2).$$
(27)
It is clear that the set of coupled eigenvalue equations given in Eq. (25) is nonlinear and requires a numerical procedure that will converge upon the solution in a self-consistent manner. The eigenstates $`\varphi _i(𝐫)`$ and eigenenergies $`ϵ_i`$ depend on time implicitly through the coefficients $`c_i(t)`$ and $`d_i(t)`$, however we do not show this time dependence in order to simplify the notation. We assume that the eigenbasis evolves slowly in time so that the adiabatic condition is satisfied .
Based on the experiment reported in the initial motional state of the system is $`\psi _1^{(\mathrm{I})}(𝐫,0)=\varphi _0(𝐫z_0\widehat{z})`$; the system is in the ground state of $`\widehat{H}_0^{}`$, but displaced from the origin along the vertical axis by $`z_0`$. This displacement is small compared to the width $`w_z`$ of the condensate $`z_0/w_z0.01`$. We therefore approximate the initial state of the system as $`|\psi ^{(I)}(t)=|\varphi _0|1`$.
The system in the rotating frame evolves according to the Hamiltonian described by Eq. (21) and Eq. (23). The term $`\widehat{H}_z^{}\widehat{\sigma }_\mathrm{x}`$ couples the internal states $`|1`$ and $`|2`$ via $`\widehat{\sigma }_\mathrm{x}`$. It also drives transitions between motional states via the dipole operator $`\widehat{z}`$. The dipole matrix element $`z_{ij}=\varphi _i|\widehat{z}|\varphi _j`$ is the largest between neighboring states and falls off quickly as $`|ij|`$ increases. For a small coupling parameter $`\beta `$, we expect the coupling to the first excited state $`|\varphi _1`$ to dominate the other transitions, making the evolution of the system predominantly a two state evolution. We therefore make the approximation that the system occupies only two modes
$$|\psi ^{(I)}(t)=c_0(t)|\varphi _0|1+d_1(t)|\varphi _1|2,$$
(28)
where $`|\varphi _0`$ is the ground state $`i=\{0,0,0\}`$ and $`|\varphi _1`$ is the first excited state with odd parity along the z-axis $`i=\{1,0,0\}`$.
If we substitute this ansatz into Eq. (16), using the Hamiltonian described by Eq. (21) and Eq. (23), we get the equation of motion for the coefficients $`c_0(t)`$ and $`d_1(t)`$
$$\begin{array}{ccc}i(\begin{array}{c}\dot{c}_0\\ \dot{d}_1\end{array})& =& (\begin{array}{cc}ϵ_0& \beta z_{01}\\ \beta z_{01}& ϵ_1\end{array})(\begin{array}{c}c_0\\ d_1\end{array})\end{array},$$
(29)
where we have neglected the time rate-of-change of the slowly varying adiabatic eigenbasis. This coupled pair of equations must be solved numerically by updating the energies $`ϵ_i`$ and the dipole matrix element $`z_{01}`$ from solving Eq. (25) at each time step. However, in order to see how the behavior depends on the various physical parameters, one can obtain a simple estimate of the solution by fixing $`ϵ_i`$ and $`z_{01}`$ to their initial values. In this case the solution of Eq. (29) is trivial and is given by $`c_0(t)=\mathrm{cos}(\mathrm{\Omega }_{01}/2t)i(\mathrm{\Delta }ϵ_{01}/\mathrm{\Omega }_{01})\mathrm{sin}(\mathrm{\Omega }_{01}/2t)`$ and $`d_1(t)=i(2\beta z/\mathrm{\Omega }_{01})\mathrm{sin}(\mathrm{\Omega }_{01}/2t)`$, where $`\mathrm{\Delta }ϵ_{01}=ϵ_1ϵ_0`$ and $`\mathrm{\Omega }_{01}=\sqrt{4\beta ^2z^2+\mathrm{\Delta }ϵ_{01}^2}`$. In the rotating frame, the system oscillates between the two states at a frequency of $`\mathrm{\Omega }_{01}`$, which is much slower than the effective Rabi frequency $`\mathrm{\Omega }_{\mathrm{eff}}`$.
The oscillation frequency $`\mathrm{\Omega }_{01}`$ increases with increasing detuning $`\delta ^{}`$ and increasing trap separation $`z_0`$ through the coupling parameter $`\beta `$. The amplitude of oscillation depends on the energy spacing between modes $`\mathrm{\Delta }ϵ_{01}`$. Based on numerical calculations, we have found that this effect is enhanced by the mean-field interaction because $`\mathrm{\Delta }ϵ_{01}`$ decreases with increasing population $`N`$. Also, the dipole matrix element $`z`$ increases with increasing $`N`$, since the width of the condensate increases with increasing population.
The solution in the lab frame can be obtained by applying $`U_\mathrm{I}(t)`$ from Eq (14) to $`|\psi ^{(\mathrm{I})}`$ in Eq. (28) to yield
$`|\psi (t)`$ $`=`$ $`(\alpha _1(t)c_0(t)|\varphi _0+\alpha _2(t)d_1(t)|\varphi _1)|1`$ (30)
$`+`$ $`(\alpha _2(t)c_0(t)|\varphi _0+\alpha _1^{}(t)d_1(t)|\varphi _1)|2,`$ (31)
where $`\alpha _1(t)=\mathrm{cos}(\mathrm{\Omega }_{\mathrm{eff}}/2t)i(\delta ^{}/\mathrm{\Omega }_{\mathrm{eff}})\mathrm{sin}(\mathrm{\Omega }_{\mathrm{eff}}/2t)`$ and $`\alpha _2(t)=i(\mathrm{\Omega }/\mathrm{\Omega }_{\mathrm{eff}})\mathrm{sin}(\mathrm{\Omega }_{\mathrm{eff}}/2t)`$. Eq. (31) is the main result of our study, with which we can explain the essential properties of the system. During the first few Rabi cycles $`t1/\mathrm{\Omega }_{\mathrm{eff}}`$, the coefficient $`d_1(t)0`$, so that the solution for short times is $`|\psi (t)=(\alpha _1(t)|1+\alpha _2(t)|2)|\varphi _0`$. That is, for short times, the internal and external degrees of freedom appear to be decoupled and the system simply oscillates rapidly between internal states. However, for longer times, the coefficient $`d_1(t)`$ grows in magnitude as $`c_0(t)`$ correspondingly decreases. This results in a modulation of the Rabi oscillations. Furthermore, a two-peaked structure in the density appears, associated with the first-excited state $`|\varphi _1`$.
## III Results
The main goal of this section is to illustrate the behavior of the system by showing results of numerical calculations. For this purpose, it is useful to treat the system in only one dimension—along the vertical axis . We also assume equal scattering lengths throughout this section, so that $`\delta _{\mathrm{MF}}=0`$. Values of most of the physical parameters are given in Table 1. Values of the remaining parameters are stated for each case considered in the text.
### A Understanding the dual dynamics
In Figure 2 we plot the fractional population of state $`|1`$, given by $`N_1(t)=|z|1|\psi (t)|^2𝑑z`$, for the case of $`\mathrm{\Omega }=700`$ Hz and $`\delta =100`$ Hz. This is a numerical solution of Eq. (1). The population is cycling rapidly at the effective Rabi frequency $`\mathrm{\Omega }_{\mathrm{eff}}=707`$ Hz, while simultaneously being modulated at a much lower frequency of about 11 Hz.
In order to visualize how the spin and motional dynamics become entangled over a time long compared to the Rabi period, we show snapshots of the density of each state in Figure 3. Three different sets of snapshots are shown, corresponding to the three circled numbers in Figure 2. A full Rabi cycle is shown for each set. The first set begins at $`t=0`$ with all of the atoms in the $`|1`$ internal state and in the mean-field ground state of the trap $`|\varphi _0`$. During this first Rabi cycle, the shape of the density profile for each internal state does not change much—only the height changes. That is, the motional state remains the ground state while population cycles rapidly between internal states, as discussed below Eq. (31).
The second set of snapshots in Figure 3 is taken at around $`t=45`$ ms, which is halfway through the modulation. The density profiles for each spin state cycle rapidly between a single-peaked and a double-peaked structure. For example, in the first snapshot, the $`|1`$ state is in the single-peaked structure, while the $`|2`$ state is in the double-peaked structure, but halfway through the Rabi cycle the situation is reversed, as shown in the third and fourth snapshots. Finally, at about $`t=90`$ ms when the amplitude of the Rabi oscillations has revived, the third set shows that the motional and spin degrees of freedom appear to be decoupled again, with the density profile of each spin state appearing as it did during the first Rabi cycle.
As outlined in Section II, this peculiar behavior is most easily understood by going to the rotating frame. In Figure 4, we plot the fractional population in the $`|1`$ state in the rotating frame $`N_1^{(I)}(t)=|z|1|\psi ^{(I)}(t)|^2𝑑z`$. The solid line corresponds to the CGSD model presented in Section IID. In the rotating frame, population is slowly transferred out of the $`|1`$ state due to the coupling from $`\widehat{H}_z^{}\widehat{\sigma }_x`$ in Eq. (21).
In the rotating frame, the system is being excited out of the ground state $`|\varphi _0`$ due to the dipole coupling $`H_z^{}`$. This can be seen in the top strip of snapshots in Figure 5, where the density of each spin state in the rotating frame is shown, corresponding to the solid line in Figure 4. Initially, all of the atoms are in the $`|1`$ internal state and the mean-field ground state of the trap $`|\varphi _0`$. Due to the dipole coupling, population is transferred out of the ground state.
The strongest coupling is between the ground $`|\varphi _0`$ and the first excited $`|\varphi _1`$ modes. These eigenmodes are shown in the bottom strip of Figure 5. They evolve slowly in time as the coefficients $`c_0(t)`$ and $`d_1(t)`$ change. For example, initially the ground state is just the Thomas-Fermi-like ground state, since all of the population is in that state. However, at $`t=45`$ ms, about one-third of the population is in the first excited mode, which pinches the ground state due to the mean-field interaction term in Eq. (23). That is why the self-consistent ground state at $`t=45`$ ms is narrower than at $`t=0`$.
It is clear from Figure 4 that the low-frequency modulation of the rapid Rabi oscillations in the lab frame is just the frequency of oscillation in the rotating frame between $`|\varphi _0|1`$ and $`|\varphi _1|2`$. This is reflected in the two-mode solution given by Eq. (31), which also helps explain the peculiar behavior of the densities shown in Figure 3. In the lab frame the system is cycling rapidly between the two modes shown in Figure 5. The initial values of the energies are $`ϵ_0=13.6\mathrm{}\omega _z`$ and $`ϵ_1=13.7\mathrm{}\omega _z`$, which makes $`\mathrm{\Delta }_{01}=0.1\mathrm{}\omega _z`$. This small energy splitting is due to the effect of the mean field, since in the limit $`N1`$ these energies move apart by a factor of ten, which greatly reduces the coupling between the modes and thus greatly reduces the modulation effect.
If we make the two-mode ansatz and solve Eq. (29), we get the dot-dashed line in Figure 4. The discrepancy from the solid line arises due to a weak coupling between the first $`|\varphi _1`$ and second $`|\varphi _2`$ excited modes. If we extend our two-state model to include this third mode, we get the dashed line in Figure 4, which nearly sits on top of the solid line. In this case, the second excited mode $`|\varphi _2`$ gains less than $`5\%`$ of the total population.
### B Dependence on detuning
In Figure 6, we show how the behavior of the system depends on the detuning $`\delta `$. The Rabi frequency $`\mathrm{\Omega }=700`$ Hz was held fixed for each plot while the detuning was varied from zero at the top $`\delta =0`$ to $`\delta =200`$ Hz in the bottom plot. As predicted by the coupling parameter $`\beta =z_0\delta \mathrm{\Omega }/\mathrm{\Omega }_{\mathrm{eff}}^2`$ in the CGSD model, no coupling between motional states occurs if $`\delta =0`$, and thus the Rabi oscillations experience no modulation. As $`\delta `$ is increased the motional-state coupling becomes stronger and we expect the modulation frequency to increase. The amplitude of modulation also increases as the detuning is increased.
We show the dependence of the period of modulation on detuning more explicitly in Figure 7. The dashed line is the numerical solution of the full problem given by Eq. (1), while the solid line is the numerical solution of the two-mode model given by Eq. (29).
### C Dependence on trap displacement
In Figure 8, we show how the behavior of the system depends on the trap displacement $`z_0`$. The Rabi frequency $`\mathrm{\Omega }=700`$ Hz and the detuning $`\delta =100`$ Hz were held fixed, while the trap displacement was varied from zero $`z_0=0`$ in the top plot to $`z_0=1\mu `$m in the bottom plot. Again, the coupling parameter $`\beta `$ predicts no modulation if $`z_0=0`$. As $`z_0`$ is increased, the frequency of modulation increases as the system is driven harder. However, for the large separation in the bottom plot, the modulation becomes highly irregular and the two-mode model most certainly breaks down. This behavior may be chaotic and warrants further investigation.
## IV Conclusions
The gross features predicted by our model, such as double-peaked structure in the density distribution, and the presence of collapses and revivals in the relative population dynamics, are supported by experimental observation . Experiment-theory agreement on finer points is only fair. The theory tends to underestimate the contrast ratio of the collapses, for instance. Moreover, to match the detuning trends shown in Fig. 6 and Fig. 7 one needs to add by hand an unexplained overall detuning offset. This is most likely due to there actually being a spatial dependence of the bare Rabi frequency due to the influence of gravity on the untrapped intermediate state of the two-photon transition. To model the experimental situation in more detail one would have to include this effect as well as inelastic loss processes and finite-temperature effects neglected here. It may be also that treating the TOP trap potential as purely static may be an oversimplication.
In this paper we have demonstrated the possibility for quantum state engineering of topological excitations through the interplay between the internal and spatial degrees of freedom in a Bose condensed gas. Due to the symmetry of the system we have analyzed, the excitation in our case was the odd-parity dipole mode. The intriguing possibility of exciting modes with alternative symmetries, such as a vortex mode , would require a different trap geometry, but is a straight-forward extension of the analysis presented here. Although we have focussed in this work on a particular parameter regime, the system is a rich one for study and exhibits complex and perhaps chaotic dynamics under strong excitation conditions.
## V Acknowledgments
We would like to thank Howard Carmichael for highlighting the analogies between this system and the bichromatically driven two-level atoms , and also Allan Griffin and Eugene Zaremba for insightful discussions. Finally, we would like to thank David Hall, Mike Matthews, and Paul Haljan for working in parallel on the experimental side of this project and for sharing the results of their observations in the laboratory . This work was supported by the National Science Foundation. E.C. would also like to thank the Office of Naval Research and the National Institute for Standards and Technology for funding support.
|
no-problem/9904/astro-ph9904158.html
|
ar5iv
|
text
|
# Hidden Bars and Boxy Bulges
## 1 Introduction
For some considerable time, N-Body simulations have predicted the existence of a vertical instability in galactic bars (Combes & Sanders cs81 (1981), Combes et al. cdfp90 (1990), Raha et al. rea91 (1991)). These simulations show that a bar forming in a flat disk will not remain thin, but will quickly buckle and form a thickened structure perpendicular to the plane of the disk. Viewed edge-on, such fat bars have a characteristic shape with very boxy isophotes. In the most extreme cases, the structure appears double-lobed, like a peanut in its shell. These simulations are intriguing because observations of real edge-on galaxies reveal that a significant fraction have central bulges with boxy or peanut-shaped isophotes (Jarvis j86 (1986), Shaw s87 (1987), de Souza & dos Anjos dd87 (1987)). It is therefore tempting to associate such bulge morphologies with edge-on bars, and even draw more general conclusions regarding the formation of all bulges.
Unfortunately, establishing the link between bars and boxy bulges observationally has proved difficult. Bars are recognizable only in fairly face-on galaxies, while the vertical structure of a bulge can only be deduced from an edge-on view. Thus, it has proved impossible to connect the two phenomena unequivocally using photometric data.
A few years ago, we suggested that bars could be detected in edge-on galaxies from their kinematic signature (Kuijken & Merrifield km95 (1995)). The orbits followed by material in a barred potential will be non-circular, and the arrangement of the orbits changes abruptly near resonances. As a consequence of this complexity, the observable line-of-sight velocities of material as a function of position in an edge-on barred galaxy will also display complex structure, with multiple components and gaps associated with the resonances. The existence of this complexity, which we originally investigated using simple perturbation theory applied to closed orbits, has subsequently been confirmed by full hydrodynamical gas simulations (Athanassoula & Bureau ba99 (1999)).
In a pilot spectral study, we found such kinematic structure in both the stellar and gaseous components of two edge-on disk galaxies (Kuijken & Merrifield km95 (1995)). Since these galaxies were selected because they contained peanut-shaped bulges, this discovery provided some evidence that bars and boxy bulges are the same phenomenon viewed from different directions. However, the sample size was very small, and lacked comparison data from galaxies without boxy bulges.
We have therefore now carried out a larger spectral survey of edge-on galaxies. The sample was taken from the largest early-type disk galaxies observable from the northern hemisphere in the RC3 catalog (de Vaucouleurs et al. dvea91 (1991)), from which we selected a subset of 10 galaxies designed to span a complete range in bulge morphology, from elliptical to peanut-shaped. Section 2 describes how the shapes of these bulges have been quantified, while Sect. 3 presents the spectral data. The results and their implications are discussed in Sect. 4.
## 2 Bulge shapes
Previous studies of bulge shapes (e.g. Jarvis j86 (1986), Shaw s87 (1987), de Souza & dos Anjos dd87 (1987)) have been based on the visual impression of galaxies in sky survey plates. Unfortunately, this subjective approach cannot be used reliably when comparing the bulge shapes to other properties of galaxies such as their kinematics. If, for example, we were to detect complex kinematics in a galaxy, there is a significant risk that we would then reinforce out initial prejudice by convincing ourselves that there were signs of boxiness in the galaxy’s isophotes. What we require, therefore, is some more objective approach.
In the case of elliptical galaxies, shapes are relatively easy to classify by measuring the minor departures of the isophotes from ellipses (e.g. Bender, Döbereiner & Möllenhoff bdm88 (1988)). However, the analysis of an edge-on disk galaxy is less straightforward, as the contribution to the total light from the disk, as well as dust absorption in the disk plane somewhat confuse the bulge isophotal shapes. Nevertheless, after some experiments we have found that it is possible to obtain robust measures of the bulge isophotes using the techniques developed for elliptical galaxies.
Images of most of our sample galaxies’ bulges were obtained in the I band at the William Herschel Telescope along with the spectral data described below. Where this was not possible, we have searched the La Palma archive, or failing this, used the Digitized Sky Survey (DSS). We have measured the bulge isophotes with the ELLIPSE task in the STSDAS analysis package of IRAF (Jedrjezewski j87 (1987)). We masked out a wedge-shaped region of each image within 12 degrees of the disk major axis to minimize the disk influence, and, in order to minimize the effects of extinction by the disk, only fitted on the side of the galaxy where the disk projects behind the bulge. The masking process leaves less than half of the isophote available for fitting, but by fixing the centroid and position angle of the isophotes to coincide with those of the disk stable results can still be obtained.
Having measured the isophote shapes at a range of surface brightnesses, we then classified a galaxy’s boxiness on the basis of the most extreme value (positive or negative) of the $`a_4`$ isophote shape parameter. Images of the half of the bulges to which the fit was applied, ordered by the value of this parameter, are presented in Fig. 1. Reassuringly, this objective ordering process arranges the galaxies in almost exactly the same sequence as one would have done by eye, but without the dangers of a posteriori bias that are inherent in any subjective classification.
## 3 Spectral data
Using the ISIS spectrograph on the William Herschel $`4.2\mathrm{m}`$ telescope, we obtained long-slit spectra for all the galaxies in the sample. In each case, the spectrograph slit was aligned along the major axis of the galaxy (if necessary just avoiding the dust lane). The spectra were obtained using a 1200-line grating, giving a resolution of 1Å (FWHM); the spectral range was centred on H$`\alpha `$ (6563Å).
We have previously found that the clearest signature of bar-induced peculiar kinematics comes from the emission lines in the spectra (Kuijken & Merrifield km95 (1995)). The strongest emission line in this spectral region is H$`\alpha `$ itself, but it lies on top of the H$`\alpha `$ absorption line associated with the stellar continuum, so its signature is somewhat confused. A clearer signal comes from the neighbouring \[N ii\] line at 6584Å, which does not lie on top of any significant absorption features. We have therefore analyzed the spectra in the vicinity of this line by first subtracting a low-order fit to the stellar continuum, then subtracting sky spectra from the ends of the slit so as to remove any night sky lines. The resulting two-dimensional spectra – intensity of the \[N ii\] line as a function of position along the slit and wavelength – are shown in Fig. 1. The wavelength scales in this figure have been adjusted so as to present all the galaxies with similar integrated line widths, thus rendering the kinematics from different galaxies more directly comparable.
## 4 Discussion
From inspection of Fig. 1, it is apparent that there is, indeed, a link between bulge morphology and the complexity of the gas kinematics – generally speaking, galaxies with non-boxy bulges have a simple kinematic structure, while boxier bulges seem to be associated with complex multiple-component emission line kinematics. Bureau and Freeman (bf99 (1999), b98 (1998)) reach a similar conclusion on the basis of a sample of southern galaxies. The structure apparent in the more complex emission-line kinematics shown in Fig. 1 – X-shaped two-component systems, distorted central parallelogram structures, and skewed figures-of-eight – are exactly the classes of feature that are generated by a non-axisymmetric barred potential when viewed from a variety of angles (Merrifield m96 (1996), Bureau & Athanassoula ba99 (1999), Athanassoula & Bureau ab99 (1999)). It is also notable that the radial scale over which the complex kinematics occurs is comparable in extent to the bulges of the host systems, again suggesting that the two phenomena are linked.
However, before we conclude from the association of boxy bulges with complex kinematics that these systems are bars viewed edge-on, we should consider other possible causes of complexity in the observed kinematics. One such possibility is that the gaps in the emission-line profiles result from differential extinction by dust down the line of sight. However, in axisymmetric galaxies, the line-of-sight velocity of the gas must be a smooth function of distance down the line of sight, so it is difficult to see how partial obscuration in an unbarred galaxy could result in multiple-components in the kinematics.
A second possibility is that the structure in the kinematics arises from rings of gas in the galaxy, as are seen in a number of more face-on disk systems (Buta & Combes bc96 (1996)). In a two-dimensional spectrum of an edge-on galaxy, an axisymmetric ring of emission will appear as an inclined straight line. Superficially, some of the structures in the spectra in Fig. 1 conform to this pattern. However, there are several crucial differences. First, rings are edge-brightened when seen in projection, yet most of the linear features in Fig. 1 do not get brighter towards their ends. Second, axisymmetric rings project to straight lines that pass through the systemic velocity of the galaxy at its centre, whereas many of the linear features in Fig. 1 are not quite straight, and have non-zero velocities at the centres of their galaxies. Such features cannot occur in an axisymmetric potential, so we are once again forced to conclude that these galaxies are barred.
A third possibility, finally, is that we are seeing variations in ionization structure of the gas, and not inhomogeneities in the distribution of the gas itself. In fact, in barred galaxies we also expect such variations, induced by the shocks; and they are indeed observed as systematic variations in N\[II\]/H$`\alpha `$ ratio over the $`(R,v)`$ diagrams (see also Bureau b98 (1998)). A detailed analysis of the line ratio lies beyond the scope of this paper. However, as far as can be ascertained, the H$`\alpha `$ emission line shows identical structure to the \[NII\] line, providing further evidence that the structure cannot be attributed to the details of the gas’ ionization state.
In summary, this spectral study of edge-on galaxies quite firmly establishes the link between boxy bulges and galactic bars. However, we have only just begun to tap into the wealth of information that the spectra provide. Modelling the full complexity of spectral data such as those shown in Fig. 1 should yield a wealth of information about barred galaxies, allowing us to map out their complete three-dimensional structure for the first time.
###### Acknowledgements.
The Digitized Sky Survey was produced by STScI under U.S. Government grant NAG W-2166, and is subject to a variety of copyrights (http://stdatu.stsci.edu/dss/dss\_copyright.html). The William Herschel Telescope is operated on the island of La Palma by the Isaac Newton Group in the Spanish Observatorio del Roque de los Muchachos of the Instituto de Astrofisica de Canarias. Much of the analysis in this paper was performed using iraf, which is distributed by NOAO, operated by AURA under cooperative agreement with the NSF. MRM has been supported by a PPARC Advanced Fellowship (B/94/AF/1840)
|
no-problem/9904/hep-ph9904314.html
|
ar5iv
|
text
|
# NEW RESULTS FROM EXPERIMENT WA102: A PARTIAL WAVE ANALYSIS OF THE CENTRALLY PRODUCED 𝐾𝐾̄ AND 𝜋𝜋 SYSTEMS
## 1 Introduction
There is considerable current interest in trying to isolate the lightest glueball. Several experiments have been performed using glue-rich production mechanisms. One such mechanism is Double Pomeron Exchange (DPE) where the Pomeron is thought to be a multi-gluonic object. Consequently it has been anticipated that production of glueballs may be especially favoured in this process.
The WA102 experiment at the CERN Omega Spectrometer studies centrally produced exclusive final states formed in the reaction
$$ppp_fX^0p_s,$$
(1)
where the subscripts $`f`$ and $`s`$ refer to the fastest and slowest particles in the laboratory frame respectively and $`X^0`$ represents the central system.
## 2 A partial wave analysis of the $`K\overline{K}`$ system
The isolation of the reaction
$$ppp_f(K^+K^{})p_s$$
(2)
has been described in detail in a previous publication. A Partial Wave Analysis (PWA) of the centrally produced $`K^+K^{}`$ system has been performed, using the reflectivity basis, in 40 MeV intervals of the $`K^+K^{}`$ mass spectrum using an event-by-event maximum likelihood method. The $`S_0^{}`$ and $`D_0^{}`$-Waves from the physical solution are shown in fig. 1.
The $`S_0^{}`$-wave shows a threshold enhancement; the peaks at 1.5 GeV and 1.7 GeV are interpreted as being due to the $`f_0(1500)`$ and $`f_J(1710)`$ with J = 0. A fit has been performed to the $`S_0^{}`$ wave using three interfering Breit-Wigners to describe the $`f_0(980)`$, $`f_0(1500)`$ and $`f_J(1710)`$ and a background of the form $`a(mm_{th})^bexp(cmdm^2)`$, where $`m`$ is the $`K^+K^{}`$ mass, $`m_{th}`$ is the $`K^+K^{}`$ threshold mass and a, b, c, d are fit parameters. The resulting fit is shown in fig. 1 and gives for the $`f_0(980)`$ M = 985 $`\pm `$ 10 MeV, $`\mathrm{\Gamma }`$ = 65 $`\pm `$ 20 MeV, for the $`f_0(1500)`$ M = 1497 $`\pm `$ 10 MeV, $`\mathrm{\Gamma }`$ = 104 $`\pm `$ 25 MeV and for the $`f_0(1710)`$ M = 1730 $`\pm `$ 15 MeV, $`\mathrm{\Gamma }`$ = 100 $`\pm `$ 25 MeV parameters which are consistent with the PDG values for these resonances.
The $`D_0^{}`$-wave shows peaks in the 1.3 and 1.5 GeV regions, presumably due to the $`f_2(1270)/a_2(1320)`$ and $`f_2^{}(1525)`$ and a wide structure above 2 GeV. There is no evidence for any significant structure in the D-wave in the region of the $`f_J(1710)`$. In addition, there are no statistically significant structures in any of the other waves. A fit has been performed to the $`D_0^{}`$ wave above 1.2 GeV using three incoherent relativistic spin 2 Breit-Wigners to describe the $`f_2(1270)/a_2(1320)`$, $`f_2^{}(1525)`$ and the peak at 2.2 GeV and a background of the form described above. The resulting fit is shown in fig. 1 and gives for the $`f_2(1270)/a_2(1320)`$ M = 1305 $`\pm `$ 20 MeV, $`\mathrm{\Gamma }`$ = 132 $`\pm `$ 25 MeV, for the $`f_2^{}(1525)`$ M = 1515 $`\pm `$ 15 MeV, $`\mathrm{\Gamma }`$ = 70 $`\pm `$ 25 MeV and for the $`f_2(2150)`$ M = 2130 $`\pm `$ 35 MeV, $`\mathrm{\Gamma }`$ = 270 $`\pm `$ 50 MeV.
A study has also been made of the centrally produced $`K_S^0K_S^0`$ channel. This channel has lower statistics than the $`K^+K^{}`$ channel but has the advantage that only even spins can contribute, which also means that there are only two ambiguous solutions to the PWA. The physical solution resulting from the PWA is the same as for the $`K^+K^{}`$ final state; namely that the $`S_0^{}`$-wave shows a threshold enhancement and peaks at 1.5 GeV and 1.7 GeV interpreted as being due to the $`f_0(1500)`$ and $`f_J(1710)`$ with J = 0.
## 3 A partial wave analysis of the $`\pi \pi `$ system
The isolation of the reaction
$$ppp_f(\pi ^+\pi ^{})p_s$$
(3)
has been described in detail in a previous publication. The resulting centrally produced $`\pi ^+\pi ^{}`$ system consists of 2.87 million events. A PWA of the centrally produced $`\pi ^+\pi ^{}`$ system has been performed, using the reflectivity basis, in 20 MeV intervals of the $`\pi ^+\pi ^{}`$ mass spectrum using an event-by-event maximum likelihood method. The $`S_0^{}`$, $`P_0^{}`$ and $`D_0^{}`$-Waves from the physical solution are shown in fig. 2. The $`S_0^{}`$-wave spectrum shows a clear threshold enhancement followed by a sharp drop at 1 GeV. There is clear evidence for the $`\rho (770)`$ in the $`P_0^{}`$ wave and for the $`f_2(1270)`$ in the $`D_0^{}`$ wave.
An interesting feature of the $`D_0^{}`$ wave is the presence of a structure below 1 GeV. In order to see if this effect is due to acceptance problems or problems due to non-central events, we have reanalysed the data using a series of different cuts but after acceptance correction no cut has been found that can remove the low mass structure. In order to investigate any systematic effects we have also analysed the central $`\pi ^0\pi ^0`$ data and a similar structure is also found. This structure does indeed seem to be a real effect which is present in other centrally produced $`\pi \pi `$ systems.
In order to obtain a satisfactory fit to the $`S_0^{}`$ wave from threshold to 2 GeV it has been found to be necessary to use three interfering Breit-Wigners to describe the $`f_0(980)`$, $`f_0(1300)`$ and $`f_0(1500)`$ and a background of the form $`a(mm_{th})^bexp(cmdm^2)`$, where $`m`$ is the $`\pi ^+\pi ^{}`$ mass, $`m_{th}`$ is the $`\pi ^+\pi ^{}`$ threshold mass and a, b, c, d are fit parameters. The fit is shown in fig. 2a) for the entire mass range and in fig. 2b) for masses above 1 GeV. The resulting parameters are for the $`f_0(980)`$ M = 982 $`\pm `$ 3 MeV, $`\mathrm{\Gamma }`$ = 80 $`\pm `$ 10 MeV, for the $`f_0(1300)`$ M = 1308 $`\pm `$ 10 MeV, $`\mathrm{\Gamma }`$ = 222 $`\pm `$ 20 MeV and for the $`f_0(1500)`$ M = 1502 $`\pm `$ 10 MeV, $`\mathrm{\Gamma }`$ = 131 $`\pm `$ 15 MeV which are consistent with the PDG values for these resonances. As can be seen, the fit describes the data well for masses below 1 GeV. It was not possible to describe the data above 1 GeV without the addition of both the $`f_0(1300)`$ and $`f_0(1500)`$ resonances. However, even with this fit using three Breit-Wigners it can be seen that the fit does not describe well the 1.7 GeV region. This could be due to a $`\pi ^+\pi ^{}`$ decay mode of the $`f_J(1710)`$ with J = 0. Including a fourth Breit-Wigner in this mass region decreases the $`\chi ^2`$ from 256 to 203 and yields for the $`f_J(1710)`$ M = 1750 $`\pm `$ 20 MeV and $`\mathrm{\Gamma }`$ = 160 $`\pm `$ 30 MeV parameters which are consistent with the PDG values for the $`f_J(1710)`$. The fit is shown in fig. 2c) for masses above 1 GeV.
## 4 A Glueball-$`q\overline{q}`$ filter in central production ?
The WA102 experiment studies mesons produced in double exchange processes. However, even in the case of pure DPE the exchanged particles still have to couple to a final state meson. The coupling of the two exchanged particles can either be by gluon exchange or quark exchange. Assuming the Pomeron is a colour singlet gluonic system if a gluon is exchanged then a gluonic state is produced, whereas if a quark is exchanged then a $`q\overline{q}`$ state is produced. In order to describe the data in terms of a physical model, Close and Kirk, have proposed that the data be analysed in terms of the difference in transverse momentum ($`dP_T`$) between the particles exchanged from the fast and slow vertices. The idea being that for small differences in transverse momentum between the two exchanged particles an enhancement in the production of glueballs relative to $`q\overline{q}`$ states may occur.
The contribution of each resonance as a function of $`dP_T`$ has been calculated. Figure 3 shows the ratio of the number of events for $`dP_T`$ $`<`$ 0.2 GeV to the number of events for $`dP_T`$ $`>`$ 0.5 GeV for each resonance considered. It can be observed that all the undisputed $`q\overline{q}`$ states which can be produced in DPE, namely those with positive G parity and $`I=0`$, have a very small value for this ratio ($`0.1`$). Some of the states with $`I=1`$ or G parity negative, which can not be produced by DPE, have a slightly higher value ($`0.25`$). However, all of these states are suppressed relative to the the glueball candidates the $`f_0(1500)`$, $`f_J(1710)`$, and $`f_2(1930)`$, together with the enigmatic $`f_0(980)`$, which have a large value for this ratio.
## 5 The azimuthal angle between the outgoing protons
The azimuthal angle ($`\varphi `$) is defined as the angle between the $`p_T`$ vectors of the two protons. Naively it may be expected that this angle would be flat irrespective of the resonances produced. Fig. 4 shows the $`\varphi `$ dependence for two $`J^{PC}`$ = $`0^+`$ final states (the $`\eta `$ and $`\eta ^{}`$), two $`J^{PC}`$ = $`1^{++}`$ final states (the $`f_1(1285)`$ and $`f_1(1420)`$) and two $`J^{PC}`$ = $`2^{++}`$ final states (the $`\varphi \varphi `$ and $`K^{}(892)\overline{K}^{}(892)`$ systems). The $`\varphi `$ dependence is clearly not flat and considerable variation is observed between final states with different $`J^{PC}`$s.
## 6 Summary
In conclusion, a partial wave analysis of the centrally produced $`K\overline{K}`$ system has been performed. The striking feature is the observation of peaks in the $`S_0^{}`$-wave corresponding to the $`f_0(1500)`$ and $`f_J(1710)`$ with J = 0. In addition, a partial wave analysis of a high statistics sample of centrally produced $`\pi ^+\pi ^{}`$ events shows that the $`S_0^{}`$-wave is composed of a broad enhancement at threshold, a sharp drop at 1 GeV due to the interference between the $`f_0(980)`$ and the S-wave background, the $`f_0(1300)`$, the $`f_0(1500)`$ and the $`f_J(1710)`$ with J = 0.
A study of centrally produced pp interactions show that there is the possibility of a glueball-$`q\overline{q}`$ filter mechanism ($`dP_T`$). All the undisputed $`q\overline{q}`$ states are observed to be suppressed at small $`dP_T`$, but the glueball candidates $`f_0(1500)`$, $`f_J(1710)`$, and $`f_2(1930)`$ , together with the enigmatic $`f_0(980)`$, survive. In addition, the production cross section for different resonances depends strongly on the azimuthal angle between the two outgoing protons.
|
no-problem/9904/hep-ph9904375.html
|
ar5iv
|
text
|
# TRANSVERSE SPIN / TRANSVERSE MOMENTUM ANALOGY AND GAUGE INVARIANCE OF DEEPLY VIRTUAL COMPTON SCATTERING
## References
|
no-problem/9904/cond-mat9904082.html
|
ar5iv
|
text
|
# Blume-Emery-Griffiths Model in a Random Crystal Field[1]
## I Introduction
The Blume-Emery-Griffiths (BEG) model is a spin-1 Ising model, originally proposed to study <sup>3</sup>He-<sup>4</sup>He mixtures. Later, it was used in the description of a variety of different physical phenomena: multicomponent fluids , microemulsions , semiconductor alloys , electronic conduction models , etc. Its hamiltonian reads:
$$=J\underset{<i,j>}{}S_iS_jK\underset{<i,j>}{}S_i^2S_j^2+\underset{i}{}\mathrm{\Delta }_iS_i^2,$$
(1)
where the first two sums are over all nearest-neighbor pairs on a lattice, the last one is over all sites and $`S_i=\pm 1,0`$. $`J`$ is the exchange parameter, $`K`$ is the biquadratic interaction and $`\mathrm{\Delta }_i`$ is a site-dependent crystal field ($`\mathrm{\Delta }_i=\mathrm{\Delta }`$ for the pure model). The phase diagram of the model presents first-order and continuous phase transitions and, for $`K<0`$, a rich variety of multicritical points .
Nevertheless, some systems were found to be better described by a disordered BEG model, as, for instance, conventional shape memory alloys and fluid mixtures on disorder materials (like aerogel, for example) . From the theoretical point of view, on the other hand, it has been argued that randomness may have drastic consequences on multicritical behavior . In two dimensions, for instance, any infinitesimal amount of disorder supresses non-symmetry-breaking first-order phase transitions and replaces symmetry-breaking first-order phase transitions by continuous ones. The effect of disorder on three-dimensional systems is different: first-order phase transitions only disappear at a finite amount of randomness . This behavior has been observed in some models .
In order to study the effects of disorder on its phase-diagram, we study the BEG model in a random crystal field (henceforth called RBEG model) given by the probability distribution:
$$𝒫(\mathrm{\Delta }_i)=r\delta (\mathrm{\Delta }_i+\mathrm{\Delta })+(1r)\delta (\mathrm{\Delta }_i\mathrm{\Delta })$$
(2)
It is worthy stressing that the exact form of the disorder is not relevant to the overall qualitative consequences on the phase diagram. If randomness is chosen to be in the interactions $`J`$ or $`K`$, the qualitative effects will be the same (in what concerns first-order phase transitions). This is due to the fact that, even if the initial disorder is on the bonds (interactions $`J`$ or $`K`$), a scale transformation will propagate this disorder to the crystal field term, which will act just like field randomness on the coexistence boundary. Moreover, the exact form of the probability distribution is not relevant, either; we have performed calculations with other distributions and they lead to the same qualitative picture as the one found in this work.
Finally, we would like to mention that, to the best of our knowledge, the BEG model in a random crystal field has not been studied so far. Previous studies concentrated on the random Blume-Capel model , which has a simpler phase-diagram than the BEG model’s.
The remainder of this paper is organized as follows. In section II we outline the mean-field approximation we use and discuss the results, in section III we present the real-space renormalization-group (RSRG) calculation (expected to hold for two-dimensional systems), and in the last section we summarize our main conclusions and comment on the influence of a degeneracy parameter $`p`$ on the critical behavior.
## II Mean-field Calculation
We chose an ordinary mean-field approximation to study the three-dimensional system. The procedure is rather usual and we refer the reader to Ref. , where a detailed discussion of the method is done.
However, we would like to stress that the mean-field approximation we use is equivalent to a model where the interaction is of infinite-range, i.e., each spin interacts with every other spin in the system. This will have explicit consequences on the phase diagram and we will return to this point later.
Most of the information about the phase diagram is numerically calculated but some analytical results can be obtained. Among them, we can find the ground state for any values of $`J`$, $`K`$, $`\mathrm{\Delta }`$ and $`r`$.
It is possible to show that the ground state magnetization, $`m_0`$, for $`d\mathrm{\Delta }/zJ>0`$, where $`z`$ is the coordination number of the lattice, is given by \[results for $`\mathrm{\Delta }<0`$ can be inferred from the mapping $`(r,\mathrm{\Delta })(1r,\mathrm{\Delta })`$\]:
$$m_0=1(1r)\theta \left[d\left(k+1\right)\left(\frac{1+r}{2}\right)\right],$$
(3)
where $`kK/J`$ and $`\theta [x]`$ is the step function, such that $`\theta [x]=0`$ or $`1`$ for $`x<0`$ or $`x>0`$ respectively. The ferromagnetic phase $`O_1`$ (see figures in this subsection), with $`m_0=1`$, is stable for $`dd_c=\left(k+1\right)\left(\frac{1+r}{2}\right)`$, while for $`dd_c`$ the ground state is such that $`m_0=r`$ (denoted $`O_2`$ in our figures). Note that, except for $`r=0`$, the ground state is always ordered; this is a consequence of the simple mean-field approximation we used (we will return to this point below).
One can obtain the continuous transition line exactly, by expanding $`\mathrm{\Phi }_{min}`$ in powers of the magnetization $`m`$ and taking the coefficient of $`m^2`$ equal to zero:
$$t_c=2\left(\frac{1r}{2+e^ke^{d/t_c}}+\frac{r}{2+e^ke^{d/t_c}}\right),$$
(4)
where $`t_ck_BT_c/zJ`$. More specifically, note that, for $`d1`$, the value of the critical temperature is $`t_c=r`$. So, for any value of $`r0`$, the critical line between the paramagnet and the $`O_2`$ phases extends to $`d=\mathrm{}`$ (see figures in this subsection). This is not the expected behavior for a cubic lattice, for the following reason. The RBEG model for $`d=\mathrm{}`$ is equivalent to the site-diluted spin-1/2 Ising model, since for $`\mathrm{\Delta }=\mathrm{}`$, a $`+\mathrm{\Delta }`$ crystal field acting on a given site forces that site to be in the $`S=0`$ state (absent), while a $`\mathrm{\Delta }`$ field forces the site to be either in the state $`S=1`$ or in the state $`S=1`$ (both represent a present site). Thus, only for high enough $`r`$ an infinite cluster of $`S=\pm 1`$ states will form and will be able to sustain order. Exactly at $`r=r_c`$, there is such an infinite cluster but its critical temperature is zero. Therefore, the critical parameter $`d_c`$ should only reach infinity for $`rr_c`$. However, the simple mean-field analysis we made leads to $`r_c=0`$, since it is equivalent to a model with infinite-range interactions. In some cases , more elaborated mean-field-like procedures were applied to the Blume-Capel model in a random crystal field. Briefly, the consequence of these approaches is that the transition line between $`O_2`$ and $`D`$ phases does not extend to $`d=\mathrm{}`$ for all values of $`r`$. All other results are similar to the ones obtained with our simple mean-field approximation. We note in advance that the approach we used for the two-dimenisonal model leads to a finite value of $`r_c`$, as expected on physical grounds.
We have already pointed out that $`t_c(\mathrm{\Delta }=\mathrm{})`$ does not depend on $`K`$; this comes from the mapping between the RDBEG model and the site-diluted spin-1/2 Ising model. The $`S=0`$ states (absent sites) play no role in the dynamics of the model and the present sites can only be in the states $`S=1`$ or $`S=1`$; thus, the biquadractic interaction, $`K`$, is irrelevant in this limit. If, for instance, the probability distribution $`𝒫(\mathrm{\Delta }_i)=r\delta (\mathrm{\Delta }_i)+(1r)\delta (\mathrm{\Delta }_i\mathrm{\Delta })`$ is used, the $`\mathrm{\Delta }=\mathrm{}`$ limit will be equivalent to the site-diluted spin-1 Ising model; then, $`t_c(\mathrm{\Delta }=\mathrm{})`$ will depend on $`K`$. Note that the discussion in this paragraph applies to the two-dimensional case as well.
We now turn to the discussion of the $`k_BT/zJ\times \mathrm{\Delta }/zJ`$ phase diagrams. In Figs. 1, 2, 3 and 4 we depict sections of constant $`K/J=5`$, for many values of $`r`$. The phase diagram for $`r=0`$ (pure BEG model) is qualitatively the same as for $`r=0.1`$ (Fig. 1), except that the $`O_2`$ phase is not present.
Note that the size of the ordered phases increases with $`r`$. This is expected, since $`r`$ is the fraction of sites which feel a $`\mathrm{\Delta }`$ crystal field (we have already commented on the “tail” which separates the $`O_2`$ and $`D_2`$ phases, given by $`t_c=r`$). Another important feature is the presence of a first-order line between two disordered phases, for $`r=0.1`$ and $`r=0.3`$. In both of these phases $`m=0`$ but $`q>1/2`$ for $`D_1`$, while $`q<1/2`$ for $`D_2`$. We would like to call attention for the phase diagram for $`r=0.3`$ (Fig. 2); this type of diagram is not present in the Blume-Capel model.
The corresponding phase diagrams for $`K/J=3`$ show only three types of critical behavior: for $`r`$ near zero, they are similar to the phase diagram for $`K/J=5`$ and $`r=0.3`$ (Fig. 2); for intermediate values of $`r`$, the behavior is the same as for $`K/J=5`$ and $`r=0.5`$ (Fig. 3); and for $`r`$ close to one, the equivalence is with the diagrams for $`K/J=5`$ and $`r=0.7`$ (Fig. 4).
The Blume-Capel model ($`K/J=0`$) has already been studied within mean-field approximations , although for different probability distributions; the results we find in this case are in qualitative agreement with those of Refs. and and we shall not depict all of them here. The only exception is the diagram for $`r=0.1`$ (Fig. 5), which is not present for higher values of $`K/J`$.
On general grounds, one should note that the mean-field approximation we employed suggests that the random crystal field does not destroy the first order transitions between disordered phases and between an ordered and a disordered phase. Even first order lines between ordered phases (like the one in Fig. 5) survive the introduction of randomness.
## III Two-dimensional Renormalization-group
It is well known that mean-field-like approximations are not suitable to describe low-dimensional systems. We have then to resort to a different technique, in order to study the RBEG model in two dimensions. RSRG procedures, on the other hand, have been successfully applied to two-dimensional systems. Note, however, that RSRG approximations, in general, do not lead to results as precise as those obtained with Monte Carlo simulations, phenomenological renormalization or conformal invariance techniques. Nevertheless, they allow for a correct description of universality classes, order of the transitions, crossover phenomena, etc.
The procedure is the same as the one thouroughly discussed in Ref. . There is just one technical point we would like to stress. Although we start with a uniform distribution for $`J`$ and $`K`$, the renormalization procedure will introduce randomness in all renormalized quantities ($`J^{},K^{}`$ and $`\mathrm{\Delta }^{}`$). A possible approach is to follow the successive renormalized distributions of these parameters in order to study the phase diagram. We adopted an alternative way, which forces the renormalized distributions to be the same as the initial ones, but with renormalized parameters, namely, $`𝒫_{ap}^{}(J)=\delta (JJ^{})`$, $`𝒫_{ap}^{}(K)=\delta (KK^{})`$ and $`𝒫_{ap}^{}(\mathrm{\Delta }_i)=r^{}\delta (\mathrm{\Delta }_i+\mathrm{\Delta }^{})+(1r^{})\delta (\mathrm{\Delta }_i\mathrm{\Delta }^{})`$. The values of $`J^{}`$ and $`K^{}`$ are obtained imposing that the first moment of the actual distributions for $`J`$ and $`K`$ and of $`𝒫_{ap}^{}(J)`$ and $`𝒫_{ap}^{}(K)`$ are equal, respectively. The values $`r^{}`$ and $`\mathrm{\Delta }^{}`$ are calculated imposing that the two lowest moments of $`𝒫_{ap}^{}(\mathrm{\Delta })`$ match those of the real distribution. This procedure has to be used with some care: in certain systems, where the random-field mechanism is important and the initial randomness is on the interaction ($`J`$, for instance), forcing the field back into a uniform distribution leads to incorrect results. In Ref. , for instance, the crystal field probability distribution is maintained uniform throughout the renormalization procedure. Consequently, the critical behavior of the random model is characteristic of a high-dimensional system: the critical temperature of the tricritical point diminishes as randomness is increased but only reaches the zero temperature axis at a finite value of the disorder. As discussed in Ref. , the mechanism responsible for the lack of first-order phase transitions in two-dimensional random systems is the disorder in the crystal field, which is not taken into account by approximations such as the one used in Ref. . In the model we study in this paper, however, the disorder in the field is not approximated away by our RSRG procedure.
Our results for $`K/J=2`$ are presented in Fig. 6, where we depict the $`kT/zJ\times \mathrm{\Delta }/zJ`$ phase diagram for $`r=0`$ (pure BEG model), $`r=0.2`$, $`r=0.45`$, and $`r=0.5`$.
Let us first comment on the pure BEG model (curve (a) of Fig. 6). As for $`K/J=5`$ in three dimensions, there are two types of disordered phases: both have $`m=0`$ but $`q>1/2`$ for phase $`D_1`$ and $`q<1/2`$ for phase $`D_2`$. The continuous line between phases $`O`$ and $`D_1`$ belongs to the universality class of the Ising model: this line is attracted to the Ising fixed point, $`C^{}\left(J=0.4407,K=0.07308,\mathrm{\Delta }=\mathrm{}\right)`$. The dashed line between phases $`O`$ and $`D_2`$ is attracted to the fixed point $`F_1\left(J=\mathrm{},K=\mathrm{},\mathrm{\Delta }=2(J+K)\right)`$, which represents a first-order transition in both $`m`$ and $`q`$, i.e., the largest eigenvalue of the even and the odd sectors of the linearized RGT matrix are equal to $`b^d`$ (see Ref. ). On the other hand, the dashed line between phases $`D_1`$ and $`D_2`$ is attracted to the fixed point $`F_2\left(J=0,K=\mathrm{},\mathrm{\Delta }=2K+\mathrm{ln}2\right)`$; in this fixed point only the largest eigenvalue of the even sector of the linearized RGT matrix is equal to $`b^d`$; this is a sign of a discontinuity in $`q`$ (but not in $`m`$) when the line is crossed .
In curves $`(b)`$, $`(c)`$, and $`(d)`$ of Fig. 6 we depict the $`kT/zJ\times \mathrm{\Delta }/zJ`$ phase diagram for $`r0`$. We note that the first-order line is either replaced by a line of continuous transitions (between $`O`$ and $`D_2`$ phases) or is eliminated (between $`D_1`$ and $`D_2`$ phases), for any infinitesimal amount of randomness. In fact, the first-order fixed point attractors, $`F_1`$ and $`F_2`$, are unstable along the $`r`$ direction. There is still a line separating the two disordered phases (not depicted in Fig. 6), $`D_1`$ and $`D_2`$, for $`r0`$, but this line is attracted to the $`\left(r=1/2,J=0,K=0,\mathrm{\Delta }=\mathrm{}\right)`$ fixed point. This point represents a model with independent spins, in which no phase transition can take place. We note that our results are in accordance with general arguments on the effects of randomness on multicritical phase diagrams , although, to the best of our knowledge, some features of these arguments have never been tested so far.
On the other hand, the whole line of continuous transitions for $`r0`$ belongs to the pure Ising model universality class, i.e., $`C^{}`$ is a stable fixed-point along the $`r`$ direction. This is expected, since, for the hierarchical lattice used in this work, the specific heat critical exponent of the pure Ising model, $`\alpha `$, is negative and disorder is irrelevant, according to the Harris criterion . For the corresponding model on a two-dimensional Bravais lattice, where $`\alpha =0`$, the Harris criterion is inconclusive. The accepted behavior, when disorder is present, is the following: critical exponents of the random model retain the same values as their pure conterparts but logarithmic corrections are introduced by randomness . Experimental results also indicate the same critical exponents for pure and random two-dimensional Ising model .
We would like to call attention to the behavior of the critical point which separates the $`O`$ and $`D_2`$ phases at $`T=0`$. For $`r<0.5`$, the transition at zero temperature takes place at a finite value of $`\mathrm{\Delta }/zJ`$. For $`r0.5`$, the critical line between the ordered and the disordered phases extends to $`\mathrm{\Delta }/zJ=\mathrm{}`$ in the diagram. In fact, for $`\mathrm{\Delta }/zJ=\mathrm{}`$ the RBEG model is equivalent to the site-dilute spin-1/2 Ising model, as discussed above. Thus, only for high enough $`r`$ an infinite cluster of $`S=\pm 1`$ states will form and will be able to sustain order. There is a critical value of $`r`$, $`r_c`$, such that the critical line between the ordered and disordered phases only reaches $`\mathrm{\Delta }/zJ=\mathrm{}`$ for $`rr_c`$. Our evaluation of $`r_c`$ is 1/2, while the accepted value for the site percolation critical probability on the square lattice is $`r_c=0.5927`$ . This difference is due to the small cell we use in this work; nevertheless, the correct qualitative picture is obtained, i.e., a finite value of $`r_c`$.
Finally, we would like to stress that there are only two types of phase diagrams for the BEG model; for high values of $`K/J`$ these diagrams have the same structure as for $`K/J=2`$. For small values of $`K/J`$, the structure is the same as for the Blume-Capel ($`K=0`$) model. As this model has been studied elsewhere , we will not discuss it here.
## IV Summary
We studied the BEG model in two and three dimensions within a RSRG framework and a mean-field approximation, respectively. The disorder is on the crystal field term, which follows a probability distribution given by: $`𝒫(\mathrm{\Delta }_i)=r\delta (\mathrm{\Delta }_i+\mathrm{\Delta })+(1r)\delta (\mathrm{\Delta }_i\mathrm{\Delta })`$.
For the mean-field approximation (expected to represent the qualitative behavior of three-dimensional systems), the presence of randomness increases the ordered phase and brings qualitative changes to the $`kT/zJ\times \mathrm{\Delta }/zJ`$ phase diagram. More specifically, first-order transitions are present in the disordered model, but new multicritical points emerge, depending on the value of $`r`$.
In two dimensions, the RSRG approach we use shows that randomness has a drastic effect on critical behavior: it supresses non-symmetry-breaking first-order transitions and replaces symmetry-breaking discontinuous transitions by continuous ones. These results are in accordance with general arguments concerning the effects of quenched impurities on multicritical behavior (but, to the best of our knowledge, the disappearance of the first-order line between disordered phases or between ordered phases has never been seen in an actual calculation). The line of continuous transition, present for the disordered $`(r0)`$ model, belongs to the universality class of the two-dimensional pure Ising model; this results agrees with the Harris criterion, since the specific heat critical exponent, $`\alpha `$, is negative for the hierarchical lattice used in this work. It has been conjectured that a new unstable critical point, at finite temperatures, might be present for the disordered system ; we found no evidence of this point, for any value of $`K`$.
We have also studied the so-called degenerate Blume-Emery-Griffiths (DBEG) model, introduced in the study of martensitic transitions . In the DBEG model, the $`S=0`$ states are assumed to have a degeneracy $`p`$, which mimics the effects of vibrational degrees of freedom. It has been shown in Ref. that the effect of increasing $`p`$ is to shrink the ordered phase and to increase the region where the transition is of first-order. Using the same probability distribution for the crystal field as in the RBEG model, we were able to show that the parameter $`p`$ may bring only quantitative changes to the phase diagrams, for any $`K/J`$, $`r`$, and $`p`$. This is expected, since the DBEG model is equivalent to the usual $`BEG`$ model with all crystal fields displaced by $`\mathrm{ln}(p)`$ . In particular, any infinitesimal amount of randomness in two dimensions destroys the first order transitions, no matter the value of $`p`$.
Finally, we would like to stress that our approximation does not allow for a study of the BEG model with negative $`K`$, where new and interesting critical behavior emerges . Work is now being made to discuss this model in the presence of a random crystal field.
## V Acknowledgments
We would like to thank Prof. J. F. Stilck for a critical reading of the manuscript and Prof. Anna Chame for calling our attention to Ref. while this work was in progress.
|
no-problem/9904/quant-ph9904022.html
|
ar5iv
|
text
|
# Quantum information and physics: some future directionsCALT-68-2219
## I Introduction
With the discovery of an apparent separation between the classical and quantum classifications of computational complexity , and of fault-tolerant schemes for quantum computation , quantum information theory has earned a lasting and prominent place at the foundations of computer science. But at present this discipline seems rather isolated from most of the rest of physics. Will this change in the future? How might it change?
One view is that thinking about information theory will lead us to a deeper understanding of the foundations of quantum mechanics. This vision has been vividly expressed by John Wheeler ; Bill Wootters and Chris Fuchs have been among its particularly eloquent spokespersons. But I am not convinced in my heart that we are supposed to understand the foundations of quantum mechanics much better than we currently do. So I prefer to look in a different direction to anticipate where quantum information may have an impact on physics.
What I tend to find most exciting in science are ideas that can build bridges across the traditional boundaries between disciplines. Perhaps that is why I find quantum computation appealing — it has established an unprecedentedly deep link between the foundations of computer science and the foundations of physics. Truly great ideas in science tend to have broad consequences that can’t be anticipated easily.
Now the quantum information community is sitting atop two ideas with potential for greatness: quantum computation and quantum error correction. I’d like to suggest two directions in which quantum information theory might evolve in the future that could lead to broad and exciting consequences for other subfields of physics. These are:
1. Precision measurement.
Our deepening understanding of quantum information may lead to new strategies for pushing back the boundaries of quantum-limited measurements. Quantum entanglement, quantum error correction, and quantum information processing might all be exploited to improve the information-gathering capability of physics experiments.
2. Many-body quantum entanglement.
The most challenging and interesting problems in quantum dynamics involve understanding the behavior of strongly-coupled many-body systems — systems with many degrees of freedom that undergo large quantum fluctuations. Better ways of characterizing and classifying the features of many-particle entanglement may lead to new and more effective methods for understanding the dynamical behavior of complex quantum systems.
## II Quantum information theory and precision measurement
The connections between quantum information and precision measurement are explored in a separate article , which I will only summarize here.
My own interest in the quantum limitations on precision measurement has been spurred in part by Caltech’s heavy involvement in the LIGO project, the Laser Interferometer Gravitational-Wave Observatory . LIGO is scheduled to begin collecting data in 2002, and a major upgrade is planned for two years later, which will boost the optical power in the interferometer and improve the sensitivity. In its most sensitive frequency band, the LIGO II observatory will actually be operating at the standard quantum limit (SQL) for detection of a weak classical force by monitoring a free mass. (In this case, the SQL corresponds to a force that nudges an 11 kg mass by about $`10^{17}`$ cm at a frequency of 100 Hz.)
Then within another 4 years (by 2008), another upgrade is expected, which will boost the sensitivity in the most critical frequency band beyond the SQL. Even an improvement by a factor of two can have a very significant payoff, for a factor of two in sensitivity means a factor of 8 in event rate. But the design of the LIGO III detection system is still largely undecided — clever innovations will be needed. So Big Science will meet quantum measurement in the first decade of the new century, and ideas from quantum information theory may steer the subsequent developments in detection of gravitational waves and other weak forces.
I learned the right way to think about the quantum limits on measurement sensitivity from Hideo Mabuchi — in a quantum measurement, a classical signal is conveyed over a quantum channel.<sup>*</sup><sup>*</sup>*Of course, connections between quantum information theory and precision measurement have been recognized by many authors. Especially relevant is the work by Wootters , by Braunstein, and by Braunstein and Caves on state distinguishability and parameter estimation, and by Braginsky and others on quantum nondemolition measurement. Nature sends us a message, a weak classical force, that can be regarded as a classical parameter appearing in the Hamiltonian of the apparatus (or more properly, if there is noise, a master equation). The apparatus undergoes a quantum operation $`\$(a)`$, and we are to extract as much information as we can about the parameter(s) $`a`$ by choosing an initial preparation of the apparatus, and a POVM to read it out. Quantum information theory should be able to provide a theory of the distinguishability of superoperators, a measure of how much information we can extract that distinguishes one superoperator from another, given some specified resources that are available for the purpose. This distinguishability measure would characterize the inviolable limits on measurement precision that can be achieved with fixed resources.
I don’t know exactly what shape this nascent theory of the distinguishability of superoperators should take, but there are already some highly suggestive hints that progress in quantum information processing can promote the development of new strategies for performing high-precision measurements.
### A Superdense coding: improved distinguishability through entanglement
A watchword of quantum information theory is: “Entanglement is a Useful Resource.” It should not be a surprise if entanglement can extend the capabilities of the laboratory physicist.
For example, the phenomenon of superdense coding illustrates that shared entanglement can enhance classical communication between two parties . The same strategy can sometimes be used to exploit entanglement to improve the distinguishability among Hamiltonians (an idea suggested by Chris Fuchs ). Suppose I wish to observe the precession of spin-$`1/2`$ objects to determine the value of an unknown magnetic field. If two spins are available, one way to estimate the value of the unknown field is to allow both spins to precess in the field independently, and then measure them separately. An alternative method is to prepare an entangled Bell pair, expose one of the two spins to the magnetic field while the other is carefully shielded from the field, and finally carry out a collective Bell measurement on the pair. It turns out that in many cases (for example when we have no a priori knowledge about the field direction), the entangled strategy extracts more information about the unknown field than the strategy in which uncorrelated spins are measured one at a time . This separation still holds even if we allow the unentangled strategy to be adaptive; that is, even if the outcome of the measurement of the first spin is permitted to influence the choice of the measurement that is performed on the second spin.
### B Grover’s database search: improved distinguishability through driving
An important paradigm emerging from the recent studies of quantum algorithms is Grover’s method for rapidly searching an unsorted database . Farhi and Gutmann observed that Grover’s algorithm may be interpreted as a method for improving the distinguishability of a set of Hamiltonians by adding a controlled driving term.
In the formulation they suggested, the Hamiltonian acting in an $`N`$-dimensional Hilbert space is known to be one of the operators
$$H_x=E|xx|,$$
(1)
where $`\{|x,x=0,1,\mathrm{},N1\}`$ is an orthonormal basis. We may gain information about the value of $`x`$ by preparing states, allowing them to evolve under $`H_x`$ for a while, and then measuring suitable observables. But determining the value of $`x`$ by this strategy requires a total time of order $`N`$. A more effective strategy is to modify the Hamiltonian by adding a controlled driving term
$$H_D=E|ss|,$$
(2)
where $`|s=N^{1/2}_{y=0}^{N1}|y`$, so that the full Hamiltonian becomes $`H_x^{}=H_x+H_D`$. If the initial state $`|s`$ is prepared and allowed to evolve under $`H_x^{}`$ for a time $`T=\pi \sqrt{N}/2E`$, then an orthogonal measurement in the $`\{|x\}`$ basis will reveal the true identity of the Hamiltonian. The time required is of order $`\sqrt{N}`$; this is Grover’s quadratic speed-up.
In this Grover-Farhi-Gutmann problem, there is a sense in which an optimal measurement procedure is known: Just as the Grover iteration allows one to identify a marked state with a minimum number of queries to the oracle , the Grover perturbation allows us to identify the actual Hamiltonian in the minimal elapsed time (asymptotically for large $`N`$).
Grover’s algorithm presumes the existence of a quantum oracle that can reply to coherent queries. In an algorithmic setting, the oracle may be regarded as a quantum circuit that can be executed repeatedly. In experimental physics, the quantum oracle is Nature, whose secrets we are eager to expose. The experimenter is challenged to find the most effective (and practical!) way to query Nature and learn her Truths.
### C Semiclassical quantum Fourier transform as adaptive phase measurement
Shor’s quantum factoring algorithm , which apparently achieves an exponential speed-up relative to classical algorithms, is based on the efficient quantum Fourier transform (QFT). Fourier analysis is a versatile tool in the laboratory, so we might expect the fast QFT to have important applications to physics.
One example could be the high-precision measurement of a frequency, like the energy splitting between the ground state and an excited state of an atom . As Cleve et al. have emphasized, the QFT can be viewed as a procedure for estimating an unknown phase. With a quantum computer, we could execute the quantum Fourier transform on $`n`$ two-level atoms, and then read out a result by measuring the internal state of each atom. If losses are negligible, the measurement outcomes provide an estimate of the frequency to an accuracy of order $`2^n`$. This procedure makes optimal use of an essential resource (the number of atoms measured), in that about one bit of information about the frequency is acquired in each binary measurement.
In fact, the complexity of the quantum information processing needed to execute this protocol is modest. In its “semiclassical” implementation proposed by Griffiths and Niu , the QFT is an adaptive procedure for phase estimation that makes use of the information collected in previous measurements to extract the best possible information from subsequent measurements. Less significant bits of the phase are measured first, and the measurement results are used to determine what single-qubit phase rotations should be applied to other qubits to extract the more significant bits more reliably. In conventional Ramsey spectroscopy, these single-qubit transformations are applied simply by prescribing the proper time interval between the Ramsey pulses.
These and other related examples give strong hints that ideas emerging from the theory of quantum information and computation are destined to profoundly influence the experimental physics techniques of the future.
## III Many-body entanglement and strongly-coupled quantum physics
### A Some signposts in Hilbert space
The most challenging and interesting problems in quantum mechanics concern many-body systems with strong quantum fluctuations. An important goal is to understand the dynamics of such systems, but it is not easy. Indeed, it is largely because strongly-coupled quantum dynamics is so difficult to understand that we want so badly to build a quantum computer !
I expect that, short of building a full-blown quantum simulator, there are many possible theoretical advances that potentially could enhance our understanding of strongly-coupled systems, including advances that could emerge from the theory of quantum information. A central task of quantum information theory has been to characterize and quantify the entanglement of multipartite systems. Up until now, most attention has focused on systems divided into a small number of parts (like twoSee , for example.), but also of great importance are the properties of $`n`$-body entanglement in the limit of large $`n`$. Studies of these properties may give us some guidance concerning what quantum simulation problems are genuinely computationally difficult, and may suggest to experimenters what kinds of systems are most likely to exhibit qualitatively new phenomena.
Hilbert space is a big place , and so far we have become familiar with only a tiny part of it. In its unexplored vastness, there is sure to be exciting new physics to discover. But much of Hilbert space is bound to be very boring indeed, so we will need some clear signposts to show the way to the exotic new phenomena.
It is truism (but still profoundly true!) that More is Different . So many of the collective phenomena exhibited by many-body systems (crystals, phase transitions, superconductivity, fractional quantum Hall effect, $`\mathrm{}`$) would be exceedingly hard to predict from first principles. That’s good news for experimenters — marvelous things could happen in many-body systems that we have been unable to anticipate. But it is easier to find something new when theory can provide some guidance.
### B Quantum error-correcting codes
A prototype for many-body entanglement has been developed in the past few years: the quantum error-correcting codes . For example, a (nondegenerate) code that can correct any $`t`$ errors in a block of $`n`$ qubits has the property that no information resides in any set of $`2t`$ qubits chosen from the block – the density matrix of the $`2t`$ qubits is completely random. Information can be encoded in the block, but the encoded information has a global character; there is no way to access any information at all by looking at only a few qubits at a time.
For example, associated with the familiar five-qubit code that can protect a single encoded qubit from an error afflicting any of the five qubits in the code block, there is a maximally entangled six-qubit pure state. This state has the property that if we trace over any three of the qubits, the density matrix of the remaining three is a multiple of the identity. It has been recognized only rather recently how unusual this state is : there exist no $`2n`$-qubit states with $`n`$ larger than three such that tracing over half of the qubits leaves the other half in a completely random state.But there are such maximally entangled states with more than six parts if each part is a higher-dimensional system rather than a qubit .
Asymptotically, we don’t know precisely “how entangled” an $`n`$-qubit state can be, but there are useful upper and lower bounds. For large $`n`$, the number $`s`$ of qubits such that the density matrix for any $`s`$ of the $`n`$ is random, must satisfy $`s/n<1/3`$ . On the other hand, states with this property are known to exist for $`s/n.1893\mathrm{}`$ . Somewhere between $`1/3`$ and $`.1893`$, there is a critical value that has not yet been pinned down. These upper and lower bounds are instructive examples of interesting results regarding multi-body entanglement that have emerged from the study of quantum error-correcting codes.
### C Classes of entangled states
This kind of global encoding of information is actually found in some systems that can be realized in the laboratory, such as systems that exhibit the fractional quantum Hall effect , or certain kinds of frustrated antiferromagnets. These systems have in common that the microscopic degrees of freedom are locally “frustrated” – that is, they are unable to find a configuration that satisfactorily minimizes the local energy density. In response, the system seeks an unusual collective state that relieves the frustration, a state such that the microscopic degrees of freedom are profoundly entangled. Condensed matter physicists have found useful ways to characterize the global properties of the entanglement that results.
For example, in the case of a two-dimensional system, we may consider how the ground state degeneracy of the system behaves on a topologically nontrivial surface in the thermodynamic limit. As Wen emphasized, in fractional quantum Hall systems the degeneracy increases with the genus (number of handles) of the surface as
$$\mathrm{ground}\mathrm{state}\mathrm{degeneracy}(A)^{\mathrm{genus}}.$$
(3)
This dependence arises from the “winding” of entanglement around the handles of the surface, and the value of $`A`$ distinguishes qualitatively different types of entangled states that must be separated from one another by phase boundaries. Just such a topological degeneracy is exploited in the ingenious quantum error-correcting codes constructed by Alexei Kitaev . A closely related observation is that in a two-dimensional system with a boundary, there can be excitations confined to the boundary, and the properties of these edge excitations reflect the nature of the entanglement in the bulk system .
I am hopeful that quantum information theory may lead to other as yet unknown ways to characterize the entangled many-body ground states of condensed matter systems, which may suggest new types of collective phenomena. We should also advance our understanding of how the profoundly entangled systems that Nature already provides might be exploited for stable storage of quantum information.
### D Information and renormalization group flow
The renormalization group (RG), one of the most profound ideas in science, is another topic that might be profoundly elucidated by an information-theoretic approach. Especially in the hands of Ken Wilson , the RG spawned one of the central unifying insights of modern physics, that of universality — physics at long distances can be quite insensitive to the details of physics at much shorter distances. Indeed, for the purpose of describing the long-distance physics, all of the short-distance physics can be absorbed into the values of the parameters of an effective field theory, where the number of parameters needed is modest if we are content with predictions to some specified accuracy.
So it is that physics is possible at all. Fortunately, it is not necessary to grasp all the subtleties of quantum gravity at the Planck scale to understand (say) the spectrum of the hydrogen atom in great detail!
The renormalization group describes how a quantum field theory “flows” as we “integrate out” short distance physics, obtaining a new theory with a smaller value of the ultraviolet momentum cutoff $`\mathrm{\Lambda }`$. “Universal” features are associated with the “fixed” points in the space of theories where the flow is stationary. In the neighborhood of each fixed point are a finite number of independent directions in theory space along which the flow is repelled by the fixed point, the “relevant” directions of flow.
Each fixed point provides a potential description of physics in the far infrared, with the number of free (“renormalized”) parameters in the description given by the number of relevant directions of flow away from the fixed point. Infrared theories with more parameters are less generic, in the sense that more “bare” parameters in the microscopic Hamiltonian of the system need to be carefully tuned in order for the flow to avoid all relevant directions and hence carry the theory to the vicinity of the fixed point. Typically, RG flow will carry a theory from the vicinity of a less generic fixed point toward the vicinity of a more generic fixed point.
Now there is at least a heuristic sense in which information is lost as a theory flows along an RG trajectory — the infrared theory “forgets” about its ultraviolet origins. One of the most intriguing challenges at the interface of physics and information is to make this connection more concrete.<sup>§</sup><sup>§</sup>§For an interesting recent attempt, see . Can we quantify how much information is discarded when a theory flows from the vicinity of one fixed point to the vicinity of another?
The proposal that an effective theory forgets more and more about its microscopic origins under RG flow leads to a robust expectation. RG flow should be a gradient flow: it always runs downhill (toward “less information”), and never uphill (toward “more information”). Indeed, this property does hold for translationally invariant and relativistically invariant quantum field theories in one spatial dimension. Zamalodchikov’s $`c`$-theorem identifies a function $`C`$ of the parameters in the Hamiltonian that can extracted from the two-point correlation function of the conserved energy-momentum tensor, and shows that $`C`$ is non-increasing along an RG trajectory. At a fixed point, the quantity $`C`$ coincides with the central charge $`c`$ that characterizes the representation of the conformal algebra according to which the fields of the fixed-point theory transform. Last year, an extension of this result to higher even-dimensional spacetimes was reported (following a suggestion by Cardy ).
It seems natural that the Zamolodchikov $`C`$-function should have a sharp interpretation relating it to loss of information along the flow, but none such is known (at least to me). A more precise information-theoretic interpretation of RG flow might guide the way to more general formulations of the $`c`$-theorem, applicable for example to theories in odd-dimensional spacetimes and to theories with less symmetry. And it might enrich our understanding of the classification of fixed-point theories and the general structure of renormalization group flow.
### E Bulk-boundary interactions
If the information-theoretic foundations underlying the $`c`$-theorem continue to prove elusive, there is another related problem that might turn out to be more tractable. It is known that a one-dimensional system with a boundary (like a semi-infinite antiferromagnetic spin chain) can sometimes exhibit an anomalous zero-temperature entropy. The entropy has a piece proportional to the length of the chain that vanishes as $`T0`$, but there is also a length-independent contribution that is nonvanishing at zero temperature (discovered by Cardy and by Affleck and Ludwig ). Ordinarily, we expect that zero-temperature entropy has an interpretation in terms of ground-state degeneracy, but in these systems (which have no mass gap, so that the ground-state degeneracy becomes a subtle concept in the thermodynamic limit), $`g=e^{S(T=0)}`$ is not an integer; hence the interpretation of the entropy is obscure.
A fascinating feature is that the “ground-state degeneracy” $`g`$ is a universal property — in the vicinity of an RG fixed point, its value is insensitive to the ultraviolet details (the microscopic interactions among the spins in the chain). Furthermore, there is evidence for a $`g`$-theorem; $`g`$ has a smaller value at more generic fixed points and a larger value at less generic fixed points .
The $`g`$-theorem, like the $`c`$-theorem, invites an interpretation in terms of loss of information along an RG trajectory. But I am hopeful that the information-theoretic origin of the $`g`$-theorem may turn out to be easier to understand. Upon hearing of entropy at zero temperature, a quantum information theorist’s ears prick up – it sounds like entanglement. It is tempting to interpret the entropy as arising from entanglement of degrees of freedom isolated at the boundary of the chain with degrees of freedom that reside in the bulk. So far, I have been unable to find a precise interpretation of this sort, but I still suspect that it could be possible.
### F Holographic universe
While on the subject of bulk-boundary interactions, I should mention the most grandiose such interaction of all. A new view of the quantum mechanics of spacetime is emerging from recent work in string theory, according to which the quantum information encoded in a spatial volume can be read completely on the surface that bounds the volume (“the holographic principle”) . This too has a whiff of entanglement – for we have seen that in a profoundly entangled state the amount of information stored locally in the microscopic degrees of freedom can be far less than we would naively expect. (Think of a quantum error-correcting code, in which the encoded information may occupy a small “global” subspace of a much larger Hilbert space.) The holographic viewpoint is particularly powerful in the case of the quantum behavior of a black hole. The information that disappears behind the event horizon can be completely encoded on the horizon, and so can be transferred to the outgoing Hawking radiation that is emitted as the black hole evaporates. This way, the evaporation process need not destroy any quantum information.
As the evidence supporting the holographic principle mounts , an unsettling question becomes more deeply puzzling: If quantum information can be encoded completely on the boundary, why does physics seem to be local? It’s strange that I imagine that I can reach out and embrace you, when we are both just shadows projected on the wall. Perhaps as the tools for analyzing many-body entanglement grow more powerful, we can begin to grasp the origin of the persistent illusion that physics is founded on the locality of spacetime.A different possible connection between quantum error-correcting codes and the black-hole information puzzle was suggested in .
## IV Conclusions
In the future, I expect quantum information to solidify its central position at the foundations of computer science, and also to erect bridges that connect with precision measurement, condensed matter physics, quantum field theory, quantum gravity, and other fields that we can only guess at today. I have identified two general areas in which I feel such connections may prove to be particularly enlightening. Progress in quantum information processing may guide the development of new ideas for improving the information-gathering capabilities of physics experiments. And a richer classification of the phases exhibited by highly entangled many-body systems may deepen our appreciation of the wealth of phenomena that can be realized by strongly-coupled quantum systems.
## ACKNOWLEDGMENTS
My work on the applications of quantum information theory to quantum-limited measurements has been in collaboration with Andrew Childs and Joe Renes . I’m very grateful to Hideo Mabuchi for stimulating my interest in that subject, and to Dave Beckman and Chris Fuchs for their helpful suggestions. I have also benefitted from discussions about precision measurement with Constantin Brif, Jon Dowling, Steven van Enk, Jeff Kimble, Alesha Kitaev, and Kip Thorne. I thank Michael Nielsen for emphasizing the relevance of quantum information in quantum critical phenomena, Ian Affleck for enlightening correspondence about conformal field theory, Anton Kapustin for a discussion about Ref. , Dorje Brody for informing me about Ref. , and Curt Callan for encouragement. Finally, I am indebted to Ike Chuang for challenging me to speculate about the future of quantum information theory. This work has been supported in part by the Department of Energy under Grant No. DE-FG03-92-ER40701, and by DARPA through the Quantum Information and Computation (QUIC) project administered by the Army Research Office under Grant No. DAAH04-96-1-0386.
|
no-problem/9904/astro-ph9904302.html
|
ar5iv
|
text
|
# Abstract
### Abstract
We have detected extraplanar cold dust at distances out to $`>10`$kpc, situated in the halo of the interacting galaxy NGC 4631. The dust emission disk is much thinner than the warped Hi disk and new structures emerge. In particular, a giant arc has been found that is linked to anomalies in the kinematical structure of the atomic gas. Most of the extraplanar dust is closely associated with Hi spurs that have been found earlier \[Weliachew, L., Sancisi, R. & Guélin, M. (1978) Astron. Astrophys. 65, 37-45; Rand, R.J. (1994) Astron. Astrophys. 285, 833-856\]. These spurs obviously are traces of the interaction \[Combes, F. (1978) Astron. Astrophys. 65, 47-55\]. The dust emission within the plane reaches the border of the optical disk.
The activity of the disk of NGC 4631 is moderately enhanced by the interaction, but no gas moving in the $`z`$-direction could be found \[Rand, R.J., Kulkarni, S.R. & Hester, J.J. (1992) Astrophys. J. 396, 97-103; Golla, G., Dettmar, R.-J. & Domgörgen, H. (1996) Astron. Astrophys. 313, 439-447\]. Hence, it seems unlikely that strong winds have deposited the high-$`z`$ dust. Instead, the coincidence with the Hi features suggests that we see a track left behind by the interaction. In addition, the Hi shows a supershell formed by an impact \[Rand, R.J. & Stone, J.M. (1996) Astron. J. 111, 190-196\] in the zone where the dust trail crosses the disk. This region is also characterized by disturbances in the distribution of the H$`\alpha `$ light. The masses associated with the dust can be estimated only very roughly on the basis of the existing data; they are of the order of a few $`10^9`$ M of gas.
## 1 Introduction
Cold dust has come into focus only recently because it had to await the development of sensitive millimeter/sub-millimeter bolometer arrays to be detectable unambiguously. The IRAS survey could provide only hints at its existence because it was blinded by the strong emission from the small percentage of warmer dust that is radiating far more brightly. The large amount of cold dust ($`T_d25`$ K) can be detected only at (sub)mm wavelengths where the radiation of the warmer components has vanished. To give an example: The peak brightness of a blackbody at 30 K is 30$`\times `$ higher than that of a 15-K object, all other parameters being equal. On the other hand, the radiation of a blackbody at 15 K peaks at $`200\mu `$m and remains more than one order of magnitude brighter at $`\lambda `$ 1.2 mm than that of a blackbody at 30 K with the same peak brightness. Now, the emissivity of interstellar dust is roughly proportional to $`T_d^6`$ – for a blackbody $`B(T)=\sigma T^4`$ – so even a very large amount of cold dust emits only weakly. Because of this $`T^6`$ dependence of the emission, a very large energy input is needed to heat dust, and the majority of it remains at lower temperatures. This cold component thus is an important tracer, and, indeed, it may represent $`>90`$% of the interstellar dust (cf. refs. 1 and 2).
In itself, the contribution of the dust to the total mass of a galaxy is $`<1`$% of the gas mass, but there are indications that the dust-to-gas ratio is relatively constant, independent of the type of gas (atomic or molecular). Indeed, the dust grains are believed to play a crucial rôle in the formation of molecules. The study of their properties thus also offers an independent means of studying the molecular gas content of galaxies. This is important because the standard practice of observing the CO molecule and deriving, thereby, the properties of the H<sub>2</sub> has substantial uncertainties particularly concerning the derived masses. On the other hand, investigating the cold dust is technically difficult and cannot provide any information about the kinematics because it is based on broadband continuum observations.
In several runs, the IRAM (Institute for Radio Astronomy in the Millimeter domain) 30-m telescope equipped with MPIfR (Max-Planck-Institute for Radioastronomy) bolometer arrays has been used to map nearby galaxies in the $`\lambda `$ 1.2-mm continuum emission. The first maps led to the impression that it is well correlated with the CO emission, and drops off similarly steeply with increasing distance from the center. This behavior was shown, for example, for galaxies NGC 891 (3), M 51 (4), and NGC 4631 (5). It soon became evident, however, that this is not generally the case. The Sb galaxy NGC 4565 is significantly more extended in the emission of $`\lambda `$ 1.2-mm continuum than in that of the CO line (6). The cold dust is even detected in the warped outermost rim of the disk. As an intermediate case, NGC 5907 also shows an extended dust disk (7). The sensitivity needed to detect this extended emission has been achieved only recently, however, and the sample is still small. So it is not yet clear what determines the extent of the cold dust – the profile of the cold dust along the major axis in NGC 891 remains very close to the rapidly vanishing CO, even when studied with much higher sensitivity than previously published (R. Zylka, personal communication).
## 2 The observations and the object: NGC 4631
### 2.1 Observational details
All recent maps were obtained with bolometer arrays consisting of 19 elements whose sensitivity is in practice about a factor of 2 better than that of the 7-element detector used before. The 19-element bolometer array has a bandwidth of $`80`$ GHz centered at $``$ 230 GHz. The individual elements are arranged in a closely packed hexagonal pattern. The beam size at the 30-m telescope is $`11^{\prime \prime }`$ and the spacing between the beams $`20^{\prime \prime }`$. The observations were made in March 1997 during a period of stable weather with zenith opacities typically $`<0.2`$. We monitored the sky opacity before and after each subimage and mapped Mars every night to determine the absolute flux scale. To obtain a map, the object is scanned in azimuthal direction including parts of blank sky on both sides to define a proper zero level. In addition, the sub-reflector of the telescope is oscillating at a frequency of 2 Hz which makes the beam switch between two positions separated by $`45^{\prime \prime }`$ in the orientation of the scanning. This yields an “on-off” measurement that cancels atmospheric variations at short time scales. The whole area of NGC 4631 was covered with a mosaic of 19 individual fields. Their distribution is rather uniform along the whole disk and, hence, the sensitivity drops only at the outer edges of the final map. Furthermore, we kept the scanning orientation close to the minor axis of the galaxy by carefully chosing the hour angles of the individual observations. This minimizes spurious contributions and the noise in the map. The field presented here is about $`15^{}\times 8^{}`$ after the cutoff of the edges with lower sensitivity. In the central part, the noise is $`2`$ mJy/beam for the data smoothed to an angular resolution of $`20^{\prime \prime }`$ and rises to $`3.5`$ mJy/beam at the edges.
### 2.2 NGC 4631
It is obviously best to choose edge-on galaxies for studies of weak phenomena, because the lines of sight are long – remember that the dust emission in the mm regime is optically thin and thus the whole disk contributes to the detectable flux. The 19-element bolometer made it possible to map large areas, so we decided to re-observe NGC 4631. This moderately active galaxy has long been a favorite candidate for an interacting system. It is relatively nearby, at 7.5 megaparsecs (Mpc) (ref. 8; $`1^{}`$ corresponds to $`2`$ kpc), and two obvious companions are close by. The dwarf elliptical NGC 4627 is situated $`3^{}`$ northwest of the nucleus and $`30^{}`$ to the southeast the distorted spiral NGC 4656 can be found. The whole group has been extensively studied in Hi (8,9) in order to understand the traces of the interaction (cf. Fig. 4). According to a modeling of the encounter (10) the prominent streamers of atomic gas can be explained as being pulled out of the members of this group during the interaction. Presumably also as a result of the interaction, the disk of NGC 4631 has a disturbed appearance in the optical continuum and H$`\alpha `$ line emission. At two positions in the disk highly energetic “supershells” have been found (11) of which one is described as being caused by the impact of a high velocity object (12).
Almost simultaneously with the early high-resolution Hi observations a large radio halo was detected (13). It is of nonthermal origin and one of the most prominent radio halos known. Detailed investigations (14) have subsequently shown that the magnetic field lines of the galactic disk open into the halo – in stark contrast to most spiral galaxies, where the field is more or less confined in the disk (15). This magnetic field structure allows electrons, cosmic ray particles and hot gas to escape from the active disk into the halo. ROSAT (Roentgen Satellite) detections (16,17) show a large X-ray envelope that is a natural consequence of this configuration.
The investigation of the molecular gas indicates rather normal conditions, however. The central region of $`2^{}\times 1^{}`$ was completely mapped in the (1-0) and (2-1) transitions of CO (18). In addition, a major axis strip of $`7^{}`$ length was obtained with a higher sensitivity of $`20`$ mK. Some additional spectra at other locations did not reveal significant emission.
## 3 The distribution of the cold dust
### 3.1 The dusty disk
At a first glance, the $`\lambda `$ 1.2-mm emission of NGC 4631 (Fig. 1) is characterized by a narrow, extended disk, with a double-peaked central region. About three times weaker than the brightest peaks, at a level of $`30`$ mJy/beam, the disk is stretched out to a distance of $`13`$ kpc on either side of the nucleus, gently decreasing in the west, and at a relatively constant level for some 10 kpc in the east – see Fig. 2. This distribution is similar to that found in NGC 4565 (6): The correlation between dust and CO is restricted to the nuclear region, whereas the dust in the outer parts of the disk seems to follow the Hi. The CO emission drops in a similarly steep manner as the centimetric radio continuum: At a radius of $`2.5^{}`$ ($`=5.5`$kpc) it has decreased to a tenth of the peak value, and a bit further out, no CO emission could be detected at all. At this radius, the dust emission is still at about one-third of the peak level, however, and it actually stretches out at least to the edge of the optical disk. In fact, the limits are set by the border of the map, not by the vanishing emission. The old bolometer map (5) is limited to the innermost part due to its restricted coverage and sensitivity.
The narrow main emission ridge is somewhat surprising given the fact that NGC 4631 does not even show a dust lane. Its thickness is similar to that of the undisturbed edge-on galaxies NGC 4565 (6) or NGC 891 (3). This suggests that the dust is concentrated in the midplane of a galaxy in any case (cf. ref. 19), and the absence of an optical dust lane may just reflect a high clumpiness with large “holes”.
A closer inspection shows some radial variations in comparison with the H$`\alpha `$ map (Fig. 3). Near the center, the optically bright regions surround the dust emission peaks. Here, the bulk of the optical emission is produced in the central region and later absorbed on its way through the disk. In contrast, in the outer parts of the disk several dust maxima coincide with optically bright spots. It is unlikely that all of these places are just accidental line-of-sight coincidences. The source of the radiation could be young stars in their dusty birthplaces (e.g. in a spiral arm tangent). In any case, the material along the light path should be rather transparent. Such large variations of the opacity are not surprising for spiral galaxies, however (see ref. 20 for a compilation).
### 3.2 The extraplanar dust
Completely new is the detection of significant dust emission out to $`z`$-distances of at least 10 kpc. The distribution of this intergalactic dust strongly suggests that it has been brought there by the same mechanism that formed the four Hi streamers (8,9). All three Hi spurs that are touched by our map are connected with the thin optical disk by corresponding dust features (see Fig. 4). The disk formed by the atomic gas is much thicker than the optical or dust emission disk, however, so that most of the $`\lambda `$ 1.2-\]mm emission still lies within their boundaries. On the other hand, CO emission could only be detected out to $`z1`$kpc. So either there is hidden molecular gas in those outer regions or the dust is associated with atomic gas.
Although these streamers follow the already known Hi features rather closely, north of the center of NGC 4631 a structure has been unveiled that was invisible in the thick atomic gas disk: A giant arc spans over the central region with its footpoints about 4 kpc east and west of the central region. The eastern footpoint is situated opposite the onset of the southern streamer 2, and the western part of the arc is blending into spur 4. The thickness of the disk emission seems to be reduced in the central region, but this might be an artifact of the data-reducing technique. The question arises regarding whether there is a common origin for these extraplanar structures.
## 4 Origin of the extraplanar dust
Unfortunately, we do not have any velocity information from the $`\lambda `$ 1.2-mm continuum observations, and hence have to rely on indirect arguments for the determination of the history of this dust distribution. To some degree, it seems reasonable to assume that the Hi velocities are a good indicator also for the dust kinematics. Indeed, there are at least two clear coincidences between Hi velocity anomalies and the arc. The velocity gradient along the major axis is much steeper in the northern part of the disk than in the southern. A possible interpretation suggests that the disk is somewhat inclined and warped along the line of sight, so we would be looking at different portions of the galaxy (8) – north of the center at the inner disk with a steeper gradient and south of it at the near part of the outer disk.
If we compare the velocity field with the dust emission (see Fig. 5), a different explanation arises. In addition to the steeper gradient, the northern disk shows anomalous velocity components at two places – and they coincide perfectly with the footpoints of the arc. These anomalous components are visible as regions with almost closed isovelocity contours in Fig. 5 (marked as A1 and A2). Although we are not able to determine the precise location of the extragalactic dust with respect to the disk, it seems clear that there is a local interaction. The additional velocity components point towards us in the west and away from us in the east.
In addition there are two structural indications, both of them suggesting that material has followed the arc in an counterclockwise direction and hit the disk in the east: (i) The appearance of the disk in the light of the H$`\alpha `$ line is much more disturbed at the eastern footpoint of the arc (cf. Fig. 3: In broadband red light the eastern part seems truncated. The disturbed region lies between the arc and the streamer 2, whereas the western side seems unaffected by the arc or streamers 3 and 4. (ii) In the Hi emission of NGC 4631 two supershells have been found near the midplane and subsequently have been modeled. Shell 2 (west of the nucleus at Right Ascension 12<sup>h</sup>39<sup>m</sup>30<sup>s</sup>) seems to be an expanding bubble, whereas shell 1 (at 12<sup>h</sup>39<sup>m</sup>56<sup>s</sup>) is most probably caused by an impact of a cloud with a mass of $`10^7M_{}`$ at a speed of 200 km/s coming from the north (12). The geometry of the collision is very well constrained by its kinematical signature in the Hi data and indicates that he material not only came from the north but, moreover, must have had a velocity component away from us. The location of this shell is between arc and spur 2. This fits to the other indications of the trajectory of the arc (the shells and the anomalous components mentioned above are distinct features).
If we try to describe all extraplanar dust features as a single trail, its path could be as follows: starting from spur 3, it runs through the western disk, follows it to the western footpoint of the arc, sweeps along the arc, penetrates the eastern disk and leaves it via spur 2. In the way, some material is forced away: for example, to follow streamer 4. Of course, such a concatenation is purely speculative and, at this point, merely meant to summarize the structure of the dust distribution. In principle, the concept of a continuous trail is a natural explanation within the framework of an interaction, however. In any case, we have to answer the question regarding which of the three involved galaxies has left the dust traces.
The interaction has been modeled rather early on the basis of the first high-resolution Hi data (10). This description suggests that the bridge 1 and the streamer 4 had been parts of the disk of NGC 4631. The two perpendicular features 2 and 3 are made from material pulled out of the now dwarf elliptical NGC 4627 which might have been of a different type before. However, the Hi emission of structures 2 and 3 is only weakly connected with the disk of NGC 4631 and not at all with the dwarf. It has to be stressed that this model is by no means unique – three involved galaxies open a vast parameter space for an interaction scenario.
Today, the location of part of the Hi gas can be determined even in the third dimension with the help of X-ray data: The soft band is very sensitive to absorption by atomic hydrogen, and it can be safely concluded from the distributions that the southern part of the disk is at the near side (16,17). Looking a bit more in detail, the onset of spur 4 seems to be located in front of the X-ray halo (cf. Figure 6 of ref. 17) but, higher up, the X-ray emission becomes stronger again, suggesting that this streamer is pointing away from us. Such additional information will help to constrain the parameters of the interaction much better than it was possible before.
## 5 Discussion and outlook
### 5.1 Origin of the mm emission
NGC 4631 is characterized not only by the interaction with two other galaxies but also by a unusually large radio halo. In particular the enhanced star formation might be responsible for the uncommon magnetic field configuration and thus for the radio and X-ray halo – could this possibly produce extraordinary continuum emission at $`\lambda `$ 1.2 mm as well?
In addition to radiation from dust, there are three candidates for the source of radiation: free-free emission in ionized clouds, nonthermal emission and line radiation within the bandpass of the detector. The free-free emission should be correlated with H$`\alpha `$ – so a contribution is only expected in the disk since the sparse high-$`z`$ H$`\alpha `$ emission (21) is not correlated with dust features. Within the disk, even in the brightest spots the emission measure reaches only a relatively low value of 1000 pc$``$cm<sup>-6</sup> (22). So the total flux due to free-free radiation is of the order of 10 mJy only.
The synchrotron emission is very extended around NGC 4631 at cm wavelengths (14). From the fluxes and the determined spectral index (23) we can estimate its contribution in our mm band. In the center, the spectral index is rather flat at about $`\alpha 0.65`$. The peak flux here is 60 mJy/$`84^{\prime \prime }`$ beam at 10.55 GHz, this translates into 0.3 mJy/$`20^{\prime \prime }`$ beam at 230 GHz. Outside the disk the spectral index is even steeper, so this contribution is smaller than a tenth of the noise level. The only significant addition thus comes from the molecular line emission: The eastern peak reaches 70 K$``$km s<sup>-1</sup> of <sup>12</sup>CO(2-1) emission. Within our $`20^{\prime \prime }`$ beam, this contributes to a flux density of $`20`$ mJy/beam (cf. ref. 4), about a fifth of the total value. In the outer disk, however, the line emission adds $`<3`$ mJy/beam – just the figure of the noise level in the map.
So, the observed mm continuum radiation is pure thermal dust emission almost everywhere, but how did the dust reach such enormous $`z`$-heights? Could it simply be blown out of the active plane by large-scale winds? Outflows are known for many galaxies with strong starbursts like, for example, M 82 (see, e.g., refs. 24 and 25). But the disk of NGC 4631 is forming stars at an only moderately enhanced rate (18). H$`\alpha `$ kinematic data as a more direct tool do not show signs of a gas outflow from the disk (22) – in contrast to M 82 which has a similar orientation (24). Peculiar velocities have been found, but they are more likely explained by the direct influence of the interaction.
### 5.2 Temperatures and Masses
The dust around the disk of NGC 4631 seems to be cold – the spatially resolved IRAS CPC data as an indicator for warmer dust show at most a somewhat thicker disk near the center (26). Its width is of the order of $`90^{\prime \prime }`$ at $`100\mu `$m and $`45^{\prime \prime }`$ at $`50\mu `$m wavelength using the published beam sizes of $`95^{\prime \prime }`$ and $`80^{\prime \prime }`$, respectively. No significant emission can be seen further out. Unfortunately, the CPC instrument could not be calibrated properly, and because of the uncertainties of the order of $`\pm 60\%`$ we could not derive dust temperatures from these data.
In view of the moderate star forming activity the presence of very warm material does not seem very likely. Presumably the temperature of the coldest dust component is rather low even in the disk – similarly to other spirals in which it could be measured so far (2,6). We expect, therefore, temperatures in the range of 15 to 20 K for the disk, and even lower values for the extraplanar dust. To determine such temperatures, observations in the sub-mm range between 1 mm and 100 $`\mu `$m are needed, but, here, the opaque atmosphere renders them very difficult. This is why only few data of spiral galaxies have been published in this range (e.g. refs. 2 and 27). We will do more observations at $`\lambda \lambda `$ 450 $`\mu `$m and 850 $`\mu `$m to complement the existing data. Together, they should provide the spectral information needed for the determination of the temperatures.
Until then, we have to postpone the proper determination of masses and energies from the dust emission. In principle, the dust mass is directly proportional to the detected flux $`S_\lambda `$. With some additional assumptions about the dust and gas properties, we can derive the hydrogen column density. In the millimeter/sub-millimeter regime, the observed flux density per beam (of half power beam width $`\theta `$) produced by dust of temperature $`T_d`$ is given by $`S_\lambda =1.13\theta ^2(1e^{\tau _\lambda })B(\lambda ,T_d)`$. Here, $`\tau _\lambda =\sigma _\lambda ^\mathrm{H}N_\mathrm{H}`$ gives the dust absorption cross section per hydrogen atom, and $`B(\lambda ,T_d)`$ is the Planck law for the radiation of a black body (see ref. 1 for a derivation, and ref. 6 for a determination of a possibly typical value). Such a calculation yields about $`210^9`$ M of hydrogen for the gas associated with a dust component of 21.5 K in the central region (5). As already stated above, the warmer dust radiates much more efficiently, and, hence, a second dust component at 55 K is associated with $`<1`$% of this mass (5).
These values are poorly constrained, however, and we face large uncertainties in the temperatures of the coldest component. If we nevertheless assume 15 K for the extraplanar dust the total gas mass in the arc should be of the order of $`1.710^9`$ M, somewhat less than the mass in the central region. A similar value is found for its presumed continuation, spur 2, so that the total gas mass in this structure would be about half of the atomic hydrogen mass of the whole galaxy (8).
### 5.3 Conclusion
Obviously, the interaction is the dominant event in the history of NGC 4631. It has caused a partial disruption of the disk, triggered the star formation activity, provoked an upturned magnetic field configuration, and, last but not least, left trails of atomic gas and dust all around the disk. It is therefore crucial to know exactly how the interaction took place. The existing model (10) was obtained by using a “trial-and-error” method for the parameters. It is virtually impossible to find the proper solution that way; the number of possible parameter sets for an interaction in such a group with three members is simply too large. Moreover, the now existing data (including the distribution of the cold dust) give new and more detailed information to check the model against. The problem is how to explore the parameter space to find the best set. Now there are new, powerful techniques available like so-called ‘genetic codes’ (28), so we hope to obtain a much better view of the interaction, which in turn should help to fix the roots of the related phenomena.
In any case, the detection of cold dust outside of a galactic disk has unveiled the existence of hitherto invisible baryonic matter in the halo region. This will not solve the dark matter problem, even using a favorable estimation of the associated mass, but it is a step towards a more “normal” neighborhood of galaxies – containing less exotic material, probably simpler to understand, but in any case easier to investigate.
Acknowledgments We thank G. Golla for the H$`\alpha `$ picture and the CO data set, R. Rand for several sets of Hi data, A. Vogler for the ROSAT data and J. Kerp for help with their interpretation. F. Combes supplied the trajectories for her model so that we got an idea of the time evolution. Part of this work was supported by the Deutsche Forschungsgemeinschaft within the frame of SFB301.
#### References
1. Cox, P. & Mezger, P.G. (1989) Astron. Astrophys. Rev 1, 49-83
2. Krügel, E., Siebenmorgen, R., Zota, V. & Chini, R. (1998), Astron. Astrophys. 331, L9-L12
3. Guélin, M., Zylka, R., Mezger, P.G., Haslam, C.G.T., Kreysa, E., Lemke, R. & Sievers, A. (1993) Astron. Astrophys. 279, L37-L40
4. Guélin, M., Zylka, R., Mezger, P.G., Haslam, C.G.T. & Kreysa, E. (1995) Astron. Astrophys. 298, L29-L32
5. Braine, J., Krügel, E., Sievers, A. & Wielebinski, R. (1995) Astron. Astrophys. 295, L55-L58
6. Neininger, N., Guélin, M., García-Burillo, S., Zylka, R. & Wielebinski, R. (1996) Astron. Astrophys. 310, 725-736
7. Dumke, M., Braine, J., Krause, M., Zylka, R., Wielebinski, R. & Guélin, M. (1997) Astron. Astrophys. 325, 124-134
8. Rand, R.J. (1994) Astron. Astrophys. 285, 833-856
9. Weliachew, L., Sancisi, R. & Guélin, M. (1978) Astron. Astrophys. 65, 37-45
10. Combes, F. (1978) Astron. Astrophys. 65, 47-55
11. Rand, R.J. & van der Hulst, J.M. (1993) Astron. J. 105, 2098-2106
12. Rand, R.J. & Stone, J.M. (1996) Astron. J. 111, 190-196
13. Ekers, R.D. & Sancisi, R. (1977) Astron. Astrophys. 54, 973-974
14. Golla, G. & Hummel, E. (1994) Astron. Astrophys. 284, 777-792
15. Dumke, M., Krause, M., Wielebinski, R. & Klein, U. (1995) Astron. Astrophys. 302, 691-703
16. Wang, Q.D., Walterbos, R.A.M., Steakley, M.F., Norman, C.A. & Braun, R. (1995) Astrophys. J. 439, 176-184
17. Vogler, A. & Pietsch, W. (1996) Astron. Astrophys. 311, 35-48
18. Golla, G. & Wielebinski, R. (1994) Astron. Astrophys. 286, 733-747
19. Xilouris, E.M., Byun, Y.I., Kylafis, N.D., Paleologou, E.V. & Papamastorakis, J. (1999) Astron. Astrophys., 344, 868-878
20. Davies, J.I. & Burstein, D. (eds.), (1995) The Opacity of Spiral Disks (Kluwer, Dordrecht), Vol. 469
21. Rand, R.J., Kulkarni, S.R. & Hester, J.J. (1992) Astrophys. J. 396, 97-103
22. Golla, G., Dettmar, R.-J. & Domgörgen H. (1996) Astron. Astrophys. 313, 439-447
23. Golla, G. (1993) PhD Thesis, University of Bonn
24. McKeith, C.D., Greve, A., Downes, D. & Prada, F. (1995) Astron. Astrophys. 293, 703-709
25. Bregman, J.N., Schulman, E. & Tomisaka K. (1995) Astrophys. J. 439, 155-162
26. van Driel, W., de Graauw, Th., de Jong, T. & Wesselius, P.R. (1993) Astron. Astrophys. Sup. 101, 207-252
27. Alton, P.B., Bianchi, S., Rand, R.J., Xilouris, E.M., Davies, J.I. & Trewhella, M. Astrophys. J. 507, L125-L129
28. Theis, Ch. (1999) Reviews in Modern Astronomy 12 (in press)
|
no-problem/9904/astro-ph9904418.html
|
ar5iv
|
text
|
# Weak weak lensing: correcting weak shear measurements accurately for PSF anisotropy
## 1 Introduction
Gravitational lensing is one of the most powerful and direct methods for studying the gravitational potentials of massive objects in the universe. An important type of lensing study is ‘weak lensing.’ It is the study of mild systematic distortions of background sources as their light rays are perturbed by gravitational fields on their way to us. Weak lensing has already provided important results in the study of galaxy clusters (e.g., Tyson et al. tvw (1990); Bonnet et al. q2345 (1993); Fahlman et al. fahlman (1994); Squires et al. a2163 (1997); Fischer et al. q0957 (1997); Clowe et al. clowe (1998); Hoekstra et al. hoekstra (1998)), halos of individual galaxies (e.g., Brainerd et al. brainerd (1996)), and large-scale structure (Schneider et al. schneider98 (1998)). As the techniques are becoming better understood, research is progressing to the search for weaker and weaker distortions, which would enable the outer regions of galaxy clusters and galaxy halos, as well as lensing signals from large-scale structure (e.g., Jain & Seljak jain97 (1997); Kaiser kaiser98 (1998)), to be studied.
To be able to detect such very weak signals, it is important to accurately remove the dominant systematic effect affecting weak lensing measurements: anisotropy of, and smearing by, the point-spread function (PSF). In this paper we will first investigate the limits of the most commonly-used technique for weak lensing analysis, devised by Kaiser et al. (ksb (1995), henceforth KSB). We will show that after PSF anisotropy correction, residual effects on the order of 1% shear are difficult to avoid with this method, even for moderately elongated PSF’s. Since the ability to detect percent signals is important for a variety of scientific questions, we have therefore devised a new method which does not have such residuals, but which nevertheless has noise properties comparable to those of the KSB method.
There are several other methods for PSF anisotropy correction in the literature. The Autocorrelation Function method of Van Waerbeke et al. (waerbeke (1997)) is a variant of the KSB method in which not individual galaxy images, but the autocorrelation function of many of them, is analyzed. The Bonnet & Mellier (bm (1995)) method uses a different aperture weighting function from KSB, and treats the PSF convolution as a shear term. Fisher & Tyson (ft97 (1997)) convolve the image with a kernel constructed to make the PSF rounder again. A more sophisticated such kernel has recently been presented by Kaiser (k99 (1999)).
## 2 Limitations of the KSB formalism for PSF anisotropy correction
We first examine the Kaiser et al. (ksb (1995)) method, following our earlier limited investigation in the context of analysis of Hubble Space Telescope images (Hoekstra et al. hoekstra (1998), Appendix D).
### 2.1 The KSB method
The technique of weak (or statistical) lensing involves measuring the systematic, gravitationally induced, distortion of background images behind a gravitational lens. In the weak lensing regime small background images are distorted by a shear $`(\gamma _1,\gamma _2)`$ and a convergence $`\kappa `$, whose combined effect is represented by the mapping
$`\left(\begin{array}{c}x\\ y\end{array}\right)`$ $`\left(\begin{array}{cc}1\kappa \gamma _1& \gamma _2\\ \gamma _2& 1\kappa +\gamma _1\end{array}\right)\left(\begin{array}{c}x\\ y\end{array}\right)`$ (1)
$`(1\kappa )\left(\begin{array}{cc}1g_1& g_2\\ g_2& 1+g_1\end{array}\right)\left(\begin{array}{c}x\\ y\end{array}\right)`$ (2)
where $`g_i=\gamma _i/(1\kappa )`$. For simplicity, in what follows we neglect $`\kappa `$ as it is small in the weak lensing regime, and pretend we are deriving the true shear $`\gamma `$ instead of the reduced shear $`g`$. Thus, our results on shape measurements are valid, but their interpretation as a lensing signal may require consideration of the $`(1\kappa )`$ factor. Our analysis makes no assumptions on the smallness of $`g_i`$, though.
Kaiser, Squires and Broadhurst (1995, henceforth KSB) describe a method for recovering $`(\gamma _1,\gamma _2)`$ from images of distant galaxies. Essentially, they derive galaxy ellipticities from weighted second moments of the observed images, and then correct these for the effects of the weight function and of smearing by the point spread function (PSF). By averaging over many galaxies, which are assumed to be intrinsically randomly oriented, the effect of individual galaxy ellipticities should average out, leaving the systematic lensing signal. The KSB method has proved to be very effective, especially in the study of galaxy cluster potentials.
KSB define various “polarizabilities”, which express the ratio between an input distortion (gravitational shear or PSF anisotropy) and the measured polarization
$$e=(\frac{I_{xx}^wI_{yy}^w}{I_{xx}^w+I_{yy}^w},\frac{2I_{xy}^w}{I_{xx}^w+I_{yy}^w})$$
(3)
of an image $`f(x,y)`$. These polarizations are derived from weighted second moments
$$I_{xx}^w=f(x,y)x^2W(r)𝑑x𝑑y,\text{etc.,}$$
(4)
of the image intensities, where $`W`$ is a weight function which goes to zero at large radii. The weight function is required as otherwise the sky noise in the outer parts of the image dominates the measured moment. The significance of the measurement is optimized by taking the weight function to be relatively compact, of a size comparable to the image itself.
Details of the method can be found in KSB, and in Hoekstra et al. (1998, henceforth HFKS), where a few small errors in the formulae of KSB were corrected. For the purposes of the present paper, it is sufficient to know that in the KSB formalism, the “smear polarizability” $`P^{\mathrm{sm}}`$ defines the ratio between the PSF anisotropy $`p=(I_{xx}I_{yy},2I_{xy})`$, constructed from the unweighted second moments $`I_{ij}`$ of the (normalized) PSF, and the resulting change in image polarization $`e`$. The “shear polarizability” $`P^{\mathrm{sh}}`$ is the ratio between the applied shear $`(\gamma _1,\gamma _2)`$ and the resulting change in the image polarization. KSB show how the polarizabilities can be derived from higher weighted moments of the observed PSF and galaxy images.
### 2.2 How accurate is KSB?
In the context of ground-based cluster weak lensing, the KSB method works well. Nevertheless, it does involve some approximations. Now that weaker and weaker signals are of interest, it is therefore important to understand the limitations of the method. As already discussed by HFKS, for strongly non-Gaussian PSF’s the KSB method does not completely correct PSF anisotropy. This is particularly true when analyzing small galaxies in deep HST images, where it turns out that the choice of weight function in eq. 4 is important.
A simple PSF model can be used to illustrate why such residuals are, at some level, unavoidable. Consider the following PSF:
$$P(x,y)=G(1+\delta ,1)+G(4\delta ,4),$$
(5)
where $`G(a^2,b^2)`$ is a unit-area Gaussian of $`x`$\- and $`y`$-dispersions $`a`$ and $`b`$. $`\delta `$ is a small parameter. The case $`\delta =0.3`$ is plotted in Figure 1.
The PSF of equation 5 has exactly zero anisotropy $`p`$: the second moments in $`x`$ and in $`y`$ are equal. However, the ellipticity of the PSF varies with radius, which means that the weighted second moments are not equal: weighting the central parts more will enhance the $`x`$-moment preferentially. In fact, it is easy to show that the polarization constructed with weighted moments is $`O(\delta )`$. The precise result for a Gaussian weight function $`W=\mathrm{exp}(\frac{1}{2}r^2/w^2)`$ is
$$e_1=\frac{9w^2\left(7+5w^2+w^4\right)\delta }{2\left(1+w^2\right)\left(4+w^2\right)\left(20+16w^2+5w^4\right)}+\mathrm{O}(\delta ^2).$$
(6)
The polarization of the $`\delta =0.3`$ PSF plotted in Figure 1 is plotted in Figure 2. It has a value of 0.03 near $`w=2`$, roughly the radius of maximum significance which should be used to minimize photon noise in the polarization measurement.
The KSB polarizabilities are derived assuming that the PSF can be written as the convolution of a compact anisotropic part with an extended circular part (KSB eq. A1). This assumption allows the anisotropy to be characterized in terms of $`p`$ only. However, our example shows that this assumption may be too restrictive: it effectively couples the radial intensity profile of the PSF with its ellipticity profile. For example, a single Gaussian with constant ellipticity can be written as such a convolution, but a sum of two elliptical Gaussian such as the PSF of eq. 5 cannot. The systematic errors that arise are the result of this.
## 3 A new method
Here we present a new method, with which the PSF effects can be corrected for with greater accuracy. The essence of the method is not to work with the moments of the observed images; instead each image is fit directly as a PSF-convolved, sheared circular source of unknown radial profile.
Assume for the moment that we have managed to sum the images of many galaxies into an ‘average galaxy’ image $`\overline{g}(x,y)`$. Analysing a stacked galaxy image is similar to the approach discussed by Lombardi & Bertin (lb (1998)), who average image second moments before corrections are applied. It differs from methods such as KSB or Bonnet & Mellier (bm (1995)) in which galaxies are individually corrected for PSF effects before they are combined to produce a shear estimate.
Intrinsically, $`\overline{g}`$ is circular if the galaxies are randomly oriented, but the image we observe has been distorted first by gravitational lensing shear, then by the atmospheric seeing, and finally by the camera optics. The observed $`\overline{g}`$ is therefore a sheared circular source, convolved with a (known) PSF. We therefore fit $`\overline{g}`$ directly to such a model, with the minimum of further assumptions. This approach addresses the apparent difficulty in the KSB methodology in the case of radially changing ellipticity profiles: a sheared circular source has constant ellipticity at all radii, and so after convolution with the PSF only a subset of “allowed” ellipticity profiles remain.
Assuming that the PSF is known, e.g., from analysis of star images in the field, the model for $`\overline{g}`$ is specified by an unknown radial brightness profile, and by the shear parameters $`(\gamma _1,\gamma _2)`$ that we are interested in. In practice we model the radial profile as the superposition of several Gaussians of different fixed widths, and unknown amplitude. We have found that the following recipe for assigning the basis functions gives good results: (i) determine the best-fit circular Gaussian radii to the observed PSF and galaxy images, $`r_{\mathrm{PSF}}`$ and $`r_{\mathrm{GAL}}`$. (ii) Take $`r=(r_{\mathrm{GAL}}^2r_{\mathrm{PSF}}^2)^{1/2}`$ as an estimate for the intrinsic radius of $`\overline{g}`$. (iii) Use four components to describe the radial profile of $`\overline{g}`$, with Gaussian radii $`(0.5,1,2,4)\times r`$.
The algorithm is laid out in Figure 3. We now describe the results of tests to verify the accuracy of the PSF anisotropy correction, and to investigate how well it fares in the presence of noise in the images.
### 3.1 Simulations in the absence of noise
We tested how well PSF anisotropy can be corrected for by considering the case where there is no gravitational shear, only a range of PSF shapes of varying anisotropy. An accurate analysis should yield zero shear after correction for the PSF. On a large number of model images, described below, we compared the results of the algorithm of Figure 3 with those from the KSB algorithm as described in HFKS (implying in particular that the same weight function is used in the derivation of polarizations and polarizabilities of galaxy and PSF images). The weight function was taken to be the best-fit circular Gaussian to the post-seeing galaxy image.
The algorithm described in this paper directly yields an estimate for the shear. In the comparisons, the KSB galaxy polarization after seeing anisotropy correction was divided by the “pre-seeing shear polarizability” $`P^\gamma `$, for which we use the expression given by Luppino & Kaiser (1997).
#### 3.1.1 Double-Gaussian images and PSF
In most of our simulations, we modeled the round average galaxy images as
$$\overline{g}=G(r_g)+G(k_gr_g)$$
(7)
where $`G(\sigma )`$ is a unit-integral Gaussian of dispersion $`\sigma `$ (a double-Gaussian PSF was also considered by ???REF???). The parameter $`k_g`$ is unity for a Gaussian profile, and is larger for more radially extended profiles. A reasonable, though admittedly crude, approximation to an exponential profile is given by setting $`k_g=2`$, while $`k_g=3`$ gives a reasonable approximation to a de vaucouleurs profile (Figure 4).
The PSF’s were modeled in a similar way, but with anisotropy. Writing now $`G(a,b)`$ for a Gaussian with $`x`$\- and $`y`$-dispersions $`a`$ and $`b`$, we have
$$PSF=G(r_p,(1ϵ_1)r_p)+G(k_pr_p,(1ϵ_2)k_pr_p).$$
(8)
Again we introduced a shape parameter $`k_p`$, but we also included ellipticities $`ϵ_i`$ for the two components. We considered three kinds of PSF ellipticity profile: we either set $`ϵ_1=ϵ_2`$ (constant ellipticity with radius), or set one of the $`ϵ`$’s to zero, to give a radial increase or decrease of the PSF anisotropy. These three possibilities, though by no means exhaustive, form a representative set of PSF’s.
The advantage of the multiple-Gaussian formulation is that the PSF convolution can be done analytically. We thus constructed a large number of PSF-smeared galaxy images, calculated the polarizations and polarizabilities following HFKS, subtracted the PSF anisotropy correction
$$\delta e=P^{\mathrm{sm}g}(P^{\mathrm{sm}})^1e^{}$$
to the galaxy polarization, and divided the result by Luppino and Kaiser (1997)’s pre-seeing shear polarizability $`P^\gamma `$. We then compared these results with the results of our implementation of the new fitting algorithm of Figure 3.
The results of the simulations are presented in figures 5, 6 and 7. They show that the KSB method can suffer from systematic residuals around the 0.01 shear level once the PSF ellipticity exceeds 0.2 or so, whereas this is not so for the new method developed here. The systematic effects are strongest for small galaxies, for PSF profiles with long tails, and for radially increasing PSF ellipticity. The effect is clearly driven by the PSF shape, not by the galaxy brightness profile.
Notice that in the constant-ellipticity case (Figure 5), with a Gaussian PSF ($`k_{\mathrm{psf}}=1`$) the residuals left by the KSB method are high order in PSF ellipticity, but that for non-Gaussian PSF’s a low-order residual dominates. (We have verified this result analytically using symbolic mathematics.) This is a consequence of the fact that only the single elliptical Gaussian PSF can be written as a convolution of a compact anisotropic function with a round extended one, as assumed in the KSB derivation. It is clearly important to test algorithms not only for single-Gaussian PSF’s!
#### 3.1.2 A WFPC-2 PSF
In order to test whether our results are specific to the double-Gaussian formulation of the PSF, a test was also performed with a model PSF for the WFPC-2 camera on the Hubble Space Telescope. The model was generated with the Tiny TIM software package, provided on-line at STScI by J. Krist. An oversampled PSF was calculated for a position near the corner of CCD#4, and convolved with a Gaussian circular galaxy of FWHM 0.25arcsec. This ‘galaxy’ and the PSF (Figure 8) were then binned to a resolution of half a WFPC-2 pixel to avoid under-resolving the PSF, and analyzed as above. The results are summarized in table 1, and confirm the results obtained from the large number of double-Gaussian simulations described earlier.
### 3.2 Noise properties
#### 3.2.1 Analytic estimate
The error on the estimated shear due to photon noise can be estimated as follows. Let the 1-$`\sigma `$ error on each pixel of $`\overline{g}`$ be $`\sigma `$ (for simplicity we take this to be the same on every pixel, appropriate for background-limited work). Then the fit involves finding the minimum of
$$\chi ^2=\underset{k}{}\left[\overline{g}_k(Pf(𝐱𝚪^2𝐱))_k\right]^2/\sigma ^2$$
(9)
where $`P(x,y)`$ is the PSF, $`f(r^2)`$ is the intrinsic radial profile of the average galaxy, $``$ denotes convolution, $`𝐱_k`$ is the position of the $`k`$th pixel, and $`𝚪`$ is the distortion matrix of equation 2. If the fit parameters $`\gamma _i`$ are uncorrelated with the radial profile, their inverse variances are given by $`\frac{1}{2}^2\chi ^2/\gamma _i^2`$. For example, at the best fit
$`{\displaystyle \frac{1}{2}}{\displaystyle \frac{^2\chi ^2}{\gamma _1^2}}=`$ $`{\displaystyle \underset{k}{}}\left({\displaystyle \frac{}{\gamma _1}}(Pf(𝐱𝚪^2𝐱))_k\right)^2/\sigma ^2,`$ (10)
$`=`$ $`{\displaystyle \underset{k}{}}\left(P[4f^{}(r^2)(y^2x^2)]\right)^2/\sigma ^2.`$ (11)
The right-hand side of equation 11 can be estimated assuming that the PSF and observed average galaxy are Gaussians with dispersions $`r_{\mathrm{PSF}}`$ and $`r_{\mathrm{GAL}}`$ pixels of integral 1 and $`F`$, respectively. Then the 1-$`\sigma `$ error on $`\gamma _1`$ evaluates to
$`\sigma (\gamma _1)=`$ $`\left({\displaystyle \frac{1}{2}}{\displaystyle \frac{^2\chi ^2}{\gamma _1^2}}\right)^{1/2}={\displaystyle \frac{2\pi ^{1/2}r_{\mathrm{GAL}}^3\sigma }{(r_{\mathrm{GAL}}^2r_{\mathrm{PSF}}^2)F}}`$ (12)
$`=`$ $`{\displaystyle \frac{r_{\mathrm{GAL}}^2}{(r_{\mathrm{GAL}}^2r_{\mathrm{PSF}}^2)}}{\displaystyle \frac{\delta F}{F}},`$ (13)
where we have used the results that the PSF-fitting error on $`F`$ for a Gaussian source is $`\delta F=2\pi ^{1/2}r_{\mathrm{GAL}}\sigma `$. (The error on $`\gamma _2`$ is the same.) We have verified this formula by means of simulations, similar to those described below. Equation 13 shows the expected increase in noise for small objects, as well as the limit, even for large objects, of
$$\sigma (\gamma _i)\frac{\delta F}{F}.$$
(14)
#### 3.2.2 Simulations
We have checked the sensitivity to noise in the images by Monte Carlo simulation. Many realizations of random Gaussian noise superimposed on a PSF-smeared, intrinsically round galaxy image were analyzed with both algorithms, and the distributions of the resulting $`(\gamma _1,\gamma _2)`$ estimates compared. Selected results are shown in Table 2. Interestingly, the dispersions in the shears derived with both methods are very similar over a range of galaxy sizes. As we have already seen, the small bias in the results from KSB is present in the simulations with non-circular PSF’s, but not in the method advocated in this paper.
A possible way to avoid the systematic residuals of the KSB method is to increase the radius of the weight function $`W`$ in eq. 4, since the problems arise from the imperfect way in which the polarizabilities represent the effect of $`W`$. However, the primary function of $`W`$ is to control the noise in the images. Doubling the Gaussian radius of $`W`$ does in fact improve the anisotropy correction in the mean, but at the cost of almost doubling the noise on the result (see Table 2).
A by-product of our algorithm is an estimate of the intrinsic radial profile of $`\overline{g}`$. In practice, this estimate appears to be rather sensitive to the noise, especially for small images—not surprising given that this is effectively a deconvolution, albeit a constrained one. Nevertheless, it may be possible to use the information in the best-fit radial profile in several ways. If a suitable prior for the intrinsic radial profile of the average galaxy selected can be formulated (e.g., by combining results over a wide field, or from deeper, higher resolution images), this information might help to refine the best-fit shear solution further. Alternatively, the width of $`\overline{g}`$ might be used to attempt to derive the lensing convergence $`\kappa `$ directly, since in principle it is a direct measure of the magnification of faint galaxies. This possibility is yet to be explored in detail, but is likely to be difficult in practice.
#### 3.2.3 The effect of centroiding errors
The centroid of an image can be determined in different ways, each of them susceptible to errors due to photon noise. The effect of centroiding errors on the summed galaxy image will be a convolution with the distribution of centroid errors. Thus, the PSF needs to be convolved with this distribution before analysis of $`\overline{g}`$, so that the effect of the centroiding error can be compensated.
## 4 Galaxy-by-galaxy application
The method as described so far involves analysing the average galaxy $`\overline{g}`$. Very accurate shear measurements require $`\overline{g}`$ to be the average of a large number of galaxies ($`1000`$ for a 1-$`\sigma `$ shear accuracy of 0.01), otherwise intrinsic ellipticity scatter will dominate. However, constructing such a $`\overline{g}`$ is only possible if the shear and the PSF are constant over a large part of the image. Often this is not the case.
To cope with this limitation, we have therefore experimented with the algorithm in ‘galaxy-by-galaxy’ mode, where the algorithm is applied to individual galaxies and the resulting shear estimates averaged. Mathematically this approach is not perfect, because it involves fitting a constant-ellipticity model to individual galaxies even though this is not necessarily appropriate. Nevertheless it turns out to work better than might have been expected, and better than existing methods.
We tested this approach on various model galaxies, of differing axis ratios. To simulate typical galaxies, we include a round, central ‘bulge’ component, and an outer ‘disk’ of axis ratio between 0.1 and 1. (Simulations with different bulge axis ratios yielded the same results.) These were placed at all orientations, smeared with an elliptical PSF, and analysed with the algorithm described above. The best-fit $`(\gamma _1,\gamma _2)`$ values thus derived for each galaxy are then averaged to give an estimate of the shear.
As may be seen in Figure 9, the algorithm performs very well, essentially correcting all PSF anisotropy signal in the measured shear. By comparison the slightly biased answer returned by the KSB algorithm is apparant as before. Residual systematics of the new method are at the level of a few tenths of a percent.
## 5 Summary
In this paper we have studied possible systematic errors arising from the correction for anisotropic point-spread functions in weak lensing analyses based on the well-known Kaiser et al. (1995) method. While such effects are small, generally below a few percent in the deduced gravitational shear components $`(\gamma _1,\gamma _2)`$, they are at a level that is important for studies such as galaxy-galaxy lensing, lensing by large-scale structure or cluster lensing at large radii. A range of simulations shows that modelling the PSF as a convolution of a compact anisotropic function with a more extended, circular function, which underlies the KSB formulation, is not sufficiently general to describe many PSFs, and leads to these systematic residuals.
We have presented a new algorithm with which to carry out the PSF-correction in a single fitting step, and show with simulated images that the low-level residuals left in the KSB analysis can thus be avoided. The whole image is used in the fitting, so that not just the lowest moments are used to characterize the image shapes. We have also shown that the noise properties of this algorithm compare well with those of KSB. While mathematically the algorithm requires an intrinsically circular source, as may be constructed by stacking many observed galaxy images, in practice nearly unbiased results can also be obtained when the algorithm is used to correct individual galaxy shapes for the PSF. This galaxy-by-galaxy application of the algorithm allows observations with spatially varying PSF and/or shear fields to be handled.
Weak lensing is a unique technique with which to study gravitational potentials at large radii in galaxy clusters, galaxy halos and in the field. The present method holds the promise of allowing a little more information to be extracted from the large volumes of data that will be gathered with the coming generation of wide-field imagers.
###### Acknowledgements.
I would like to thank Peter Schneider, Marijn Franx, Henk Hoekstra and the referee for critical readings of the manuscript and for suggesting several improvements.
|
no-problem/9904/hep-ph9904412.html
|
ar5iv
|
text
|
# Higgs Boson Bounds in Three and Four Generation Scenarios
## I Bounds on $`M_H`$
We now determine a lower bound on the Higgs boson mass in the SM . We first alert the reader to our phenomenologically viable assumption that the physical vacuum corresponds to a global, not merely a local, minimum of the effective potential. This assumption is consistent with our intention to accept the SM as a truly valid theory and compute the consequences, i.e. to zeroth order there is no motivation to consider the physical vacuum to be anything other than the true vacuum. If one considers the possibility that the physical vacuum is a metastable vacuum with a lifetime longer than the age of the universe, that there exist deeper minima of the potential, then the SM lower bounds on the Higgs boson mass become less stringent in general for certain choices of $`\mathrm{\Lambda }`$ and $`M_{top}`$, where $`\mathrm{\Lambda }`$ is the cutoff beyond which the SM is no longer valid . But, for $`M_{top}177`$ GeV and $`\mathrm{\Lambda }=10^{19}`$ GeV, the SM3 absolute stability lower bound is relaxed by only $`O(5)`$ GeV when one only imposes metastability requirements, and this small effect only becomes diminished with the inclusion of a fourth generation. We obtain lower limits on the SM Higgs boson mass by requiring stability of this observed vacuum. It is well known that lower values of $`\mathrm{\Lambda }`$ relax the SM lower bounds , but we note that the lower bounds on the SM Higgs boson mass are insensitive to the precise value of $`\mathrm{\Lambda }`$ for large $`\mathrm{\Lambda }`$, i.e. for $`10^{11}`$ GeV $`<\mathrm{\Lambda }<10^{19}`$ GeV.
Working with the two-loop RGE requires the imposition of one-loop boundary conditions on the running parameters . As pointed out by Casas et al. , the necessary condition for vacuum stability is derived from requiring that the effective coupling $`\stackrel{~}{\lambda }(\mu )>`$ 0 rather than $`\lambda >0`$ for $`\mu (t)<\mathrm{\Lambda }`$, where $`\mathrm{\Lambda }`$ is the cut-off beyond which the SM is no longer valid. The effective coupling $`\stackrel{~}{\lambda }`$ in the SM4 is defined as:
$$\stackrel{~}{\lambda }=\frac{\lambda }{3}\frac{1}{16\pi ^2}\left\{\underset{i=1}{\overset{5}{}}2\kappa _ih_i^4\left[\mathrm{ln}\frac{h_i^2}{2}1\right]\right\}$$
where the three generation case is simply the same as the above expression without the fourth generation Yukawa coupling contributions. Choosing $`\mathrm{\Lambda }=10^{19}`$ GeV and $`M_{top}=172`$ GeV, we arrive at a vacuum stability lower bound on $`M_h`$ of $``$ 134 GeV for the SM with three generations. Allowing $`M_{top}`$ to be as large as 179 GeV increases the lower bound on $`M_H`$ to $``$ 150 GeV.
To compute the MSSM upper bound on $`M_H`$, we assume that all of the sparticles have masses $`O(M_{susy})`$ or greater and that of the two Higgs isodoublets of the MSSM, one linear combination is massive, also with a mass of $`O(M_{susy})`$ or greater, while the other linear combination, orthogonal to the first, has a mass of the order of weak-scale symmetry breaking. With these two assumptions, it is clear that below the supersymmetry breaking scale $`M_{susy}`$, the effective theory is the SM. This fact enables us to use the SM effective potential for the Higgs boson when we treat the lightest Higgs boson in the MSSM.
In the MSSM(3,4), the boundary condition for $`\lambda `$ at $`M_{susy}`$ is
$$\frac{\lambda }{3}(M_{susy})=\frac{1}{4}\left[g_1^2(M_{susy})+g_2^2(M_{susy})\right]\mathrm{cos}^2(2\beta )+\frac{\kappa _ih_i^4(M_{susy})}{16\pi ^2}\left(2\frac{X_i}{M_{susy}^2}\frac{X_i^4}{6M_{susy}^4}\right)$$
where $`\kappa _i`$ = 3 for $`i=(t,T,B)`$ and $`\kappa _i`$ = 1 for $`i=(N,E)`$ and $`X_i`$ is the supersymmetric mixing parameter for the ith fermion. Zero threshold corrections correspond to $`X_i`$ = 0. Maximum threshold corrections occur for $`X_i=6M_{susy}^2`$.
(1) (2)
> Figure 1: The lightest Higgs boson mass $`M_H`$ as a function of $`\mathrm{cos}^2(2\beta )`$. The bottom two curves correspond to MSSM upper bounds with no threshold corrections, for $`M_{top}`$ = 172 GeV and 179 GeV, respectively. The two upper curves correspond to MSSM upper bounds with maximum threshold corrections, for $`M_{top}`$ = 172 GeV and 179 GeV, respectively. The two horizontal lines are the $`\mathrm{cos}^2(2\beta )`$-independent SM3 vacuum stability bounds. The lower horizontal line corresponds to $`M_{top}`$ = 172 GeV, while the other horizontal line was computed with $`M_{top}`$ = 179 GeV. Figure 2: Same as Figure 1, but now the MSSM bounds correspond to the minimal threshold corrections consistent with the experimental lower limit on $`M_H`$
In Fig. (1) we present our numerical two-loop results for the lightest Higgs boson mass bounds in the SM and the MSSM3 as a function of the supersymmetric parameter $`\mathrm{cos}^2(2\beta )`$. The bottom two curves correspond to the MSSM3 upper bound for the two cases $`M_{top}=172`$ GeV and the slightly greater upper bound that results when $`M_{top}`$ = 179 GeV and with no threshold corrections. When the case of maximum threshold corrections is considered, these two curves are translated upwards by $``$ 55 GeV - 60 GeV, illustrating the strong dependence of the upper bound on the precise value of the threshold corrections. Yet even with such a dramatic increase in the upper bounds with increasing threshold corrections, we observe that the SM lower bound exceeds the MSSM upper bound for $`M_{top}=172`$ GeV and $`0<\mathrm{cos}^2(2\beta )<.2`$ for all values of the threshold correction contribution. Similarly, for $`M_{top}=179`$ GeV, the troublesome situation is only exacerbated, as the SM lower bound exceeds the MSSM upper bound for $`0<\mathrm{cos}^2(2\beta )<.38`$ independent of the threshold corrections.
In Fig.(2) we present the problem more clearly. Taking into account the present experimental lower limit on $`M_H`$ of $``$ 90 GeV at 95$`\%`$ CL, we find the value of the threshold correction that gives a smallest upper bound consistent with the experimental lower limit. Clearly, for this phenomenologically determined lower limit of the threshold contributions, there is a large area in $`M_H\times \mathrm{cos}^2(2\beta )`$ space that is inconsistent with both the SM and the MSSM. For $`M_{top}=172`$ GeV, the region 92 GeV $`<M_H<`$ 134 GeV invalidates both theories independent of $`\mathrm{cos}^2(2\beta )`$, while for $`M_{top}=179`$ GeV, the range of mutual invalidiation is 92 GeV $`<M_H<`$ 150 GeV.
## II Fourth Generation
To resolve the above conundrum, one would like to either raise the MSSM upper bounds, lower the SM lower bounds, or both. Upon adding a fourth generation, the SM4 lower bounds exceed the SM3 lower bounds and are an increasing function of the fourth generation masses. If a Higgs is detected in the region of mutual invalidation of both the SM and the MSSM, consideration of SM4 vacuum stability lower bounds only exacerbates the problem. It is readily apparent that the way out of the area of inconsistency is to consider the MSSM4 and see if the additional matter of the MSSM4 results in MSSM4 upper bounds that exceed the SM3 lower bounds.
We now discuss restrictions on the possible fourth generation fermion masses . The close agreement betweeen the direct measurements of the top quark at the Tevatron and its indirect determination from the global fits of precision electroweak data including radiative corrections within the framework of the SM imply that there is no significant violation of the isospin symmetry for the extra generation. Thus the masses of the fourth generation isopartners must be very close to degenerate ; i.e.
$$\frac{M_T^2M_B^2}{M_Z^2}1,\frac{M_E^2M_N^2}{M_Z^2}1$$
Recently, the limit on the masses of the extra neutral and charged lepton masses, $`M_N`$ and $`M_E`$, has been improved by LEP1.5 to $`M_N>59`$ GeV and $`M_E>62`$ GeV. Also, CDF has yielded a lower bound on $`M_B`$ of $``$ 140 GeV.
In our previous work, we considered a completely degenerate fourth generation of fermions with mass $`m_4`$. We derived an upper bound on $`m_4`$ in the MSSM4 by demanding pertubative validity of all the couplings out to the GUT scale . This constraint led to an upper bound on $`m_4`$ of $``$ 110 GeV. The above experimental lower limit on $`M_B`$ naturally forces us to now a consider a fourth generation where degeneracy only holds among the isodoublets seperately. We therefore consider a fourth generation with masses $`M_L`$ and $`M_Q`$.
In Fig.(3), we present the SM lower bound, the MSSM4 upper bound with the fourth generation masses at their experimental lower limits and with fourth generation masses large enough to remove the problem area for all values of $`\mathrm{cos}^2(2\beta )`$. The MSSM bounds were calculated with no threshold corrections, and $`M_{top}`$ is fixed at 172 GeV. Fig.(4) shows the same information for $`M_{top}`$ = 179 GeV. The MSSM4 upper bounds are much more sensitive to $`M_Q`$ than they are to $`M_L`$. This qualitative behaviour is readily understood from inspection of the equation for $`m_\varphi ^2`$. For this reason, it is necessary to increase $`M_Q`$ appropriately in order to generate a MSSM4 upper bound that is greater than the SM lower bound for all values of $`\mathrm{cos}^2(2\beta )`$. In fact, keeping $`M_Q`$ at 146 GeV and allowing $`M_L`$ to be 110 GeV does not resolve the problem. But increasing both $`M_Q`$ and $`M_L`$ as indicated in the figures does remove the problem. Because all of the bounds increase as $`M_L`$ and $`M_Q`$ increase, and because the upper bounds on $`m_4`$ from the previous work are saturated when the masses of the fourth generation reach some critical values from below, we can conclude that $`M_L`$ must still be $`<110`$ GeV. This conclusion follows because it is $`h_N`$ that violates pertubative validity, so in the non-degenerate case, it is $`M_L`$ that must still respect this upper bound if gauge coupling unification is still to be achieved in the MSSM4.
(3) (4)
> Figure 3: Plots of the physical Higgs boson mass as a function of $`\mathrm{cos}^2(2\beta )`$. The $`\mathrm{cos}^2(2\beta )`$-independent flat line is the MSSM3 vacuum stability lower bound for $`M_{top}`$ = 172 GeV. The lower curve is the MSSM4 upper bound for the same value of $`M_{top}`$, no threshold corrections and the indicated values for $`M_L`$ and $`M_Q`$. Similarly for the upper curve. Figure 4: Same as Figure 3, but with $`M_{top}`$ = 179 GeV.
## III CONCLUDING REMARKS
In conclusion, we have studied the upper bounds on the lightest Higgs boson mass $`M_H`$ in the MSSM with four generations by solving the two-loop RGE’s and using the one-loop EP. We find that if the Higgs boson is discovered with a mass $`M_H<`$134 GeV (150 GeV) for $`M_{top}=`$ 172 GeV (179 GeV), then there is a demand for the introduction of new physics. This mass range for $`M_H`$ will be explored shortly and thus an explanation of what new physics could be consistent with such a Higgs mass measurement is desirable. We propose that such a measurement could be taken as indirect evidence for a fourth generation of fermions. Considering a fourth generation where degeneracy only holds within the isodoublets individually, we find that a measurement of $`M_H`$ in the above range is consistent with the MSSM4 upper bounds on $`M_H`$. In addition, the possibility of gauge coupling unification remains intact for 60 GeV $`<M_L<`$ 110 GeV and $`M_Q\stackrel{>}{_{}}`$ 170 GeV. Therefore, if $`M_H`$ is measured to be below the SM3 lower bound, we suggest a search for fourth generation fermions with 60 GeV $`<M_L<`$ 110 GeV and $`M_Q\stackrel{>}{_{}}`$ 170 GeV.
## IV ACKNOWLEDGEMENTS
We wish to thank M. Machacek and M. Vaughn for helpful discussions concerning the RGE used in this investigation. Support for this work was provided in part by U.S. Dept. of Energy Contract DE-FG-02-91ER40688-Task A.
|
no-problem/9904/cond-mat9904050.html
|
ar5iv
|
text
|
# A re-examination of the electronic structure of 𝐵𝑖₂𝑆𝑟₂𝐶𝑎𝐶𝑢₂𝑂_{8+𝛿} and 𝐵𝑖₂𝑆𝑟₂𝐶𝑢₁𝑂_{6+𝛿} - An electron-like Fermi Surface and the absence of flat bands at 𝐸_𝐹.
## Abstract
We present a re-examination of the electronic structure and Fermi Surface (FS) of Bi-Sr-Ca-Cu-O (BSCCO) as obtained from angle-resolved photoemission experiments. By applying a stricter set of FS crossing criteria as well as by varying the incident photon energy outside the usual range, we have found very different behavior from that previously observed. In particular we have found an electron-like FS centered around the $`\mathrm{\Gamma }`$ point, and the flat bands at $`E_F`$ near the $`\overline{M}`$ point of the zone are absent. These results are robust over a large range of dopings and from single to double layer samples.
Angle-Resolved Photoemission Spectroscopy (ARPES) has emerged as one of the most powerful tools for unearthing the electronic structure and physics of the high temperature superconductors (HTSC) and other correlated electron systems since it allows one to directly probe the $`\stackrel{}{k}`$ space information of the electronic structure. Major discoveries obtained from ARPES experiments on the HTSC’s have included the observation of flat bands (an extended van Hove singularity) at or very near the Fermi level, a superconducting gap with d-wave symmetry, and an anomalous pseudogap above $`T_c`$. Such advances are all predicated on a thorough knowledge of the normal-state FS topology of these superconductors, which has been almost universally accepted to be a hole-like pocket centered around the Brillouin zone corners (($`\pi ,\pi `$) or X, Y points). In this Letter we argue that this most fundamental assumption of the FS topology is incorrect or at least greatly oversimplified. This finding should have a major impact on many previous studies and theories, including all those concentrating on the flat bands at $`E_F`$, the superconducting gap, and on the normal state pseudogap.
The measurements we report on here were taken on high quality single crystals of the BiSrCaCuO (BSCCO) family of cuprate superconductors. These are nearly ideal materials for surface sensitive ARPES studies because of the beautiful cleaved surfaces that are obtainable. Thus the ARPES information is usually interpreted as being representative of the bulk physical properties of BSCCO.
To date ARPES data from BSCCO have been mostly limited to a narrow photon energy range between $`19`$ and $`25eV`$. This has been largely a matter of convenience as well as the general expectation that the physics observed at other photon energy ranges would be essentially unchanged. We have measured BSCCO at photon energies well outside this range and have found very different physics from that observed between $`19`$ and $`25eV`$. Most of the data presented in this paper was taken with a photon energy of $`33eV`$, at which we find qualitatively different but much clearer behavior than that observed between $`19`$ and $`25eV`$. Namely, the data shows with high clarity the existence of an electron-pocket Fermi surface as well as the absence of flat bands at $`E_F`$.
The experiments were mostly performed at the Synchrotron Radiation Center in Wisconsin with a few backup experiments performed at the Stanford Synchrotron Radiation Laboratory in California. At both labs we used VSW 50mm hemispherical energy analyzers mounted on two-axis goniometers. The total experimental energy resolution was about $`50meV`$ FWHM and the angular resolution was $`\pm 1^o`$. All data shown in this paper was taken at or near 100K, comfortably in the normal state. Most of the $`Bi_2Sr_2CaCu_2O_{8+\delta }`$ (Bi2212) samples for this study came from CRIEPI, with a portion coming from ETL. The $`Bi_2Sr_2Cu_1O_{6+\delta }`$ (Bi2201) sample studied came from the University of Tokyo.
Fig. 1(a) shows Ding et al’s version of the generally accepted hole-like FS topology of $`Bi_2Sr_2CaCu_2O_{8+\delta }`$ (Bi2212), from one of the most complete and heavily referenced data sets available. The thick lines represent the FS due to the main $`CuO_2`$ band. The thin lines are FS replicas due to the superstructure modulation with a $`Q=(0.2\pi ,0.2\pi )`$. Support for such a topology is displayed in Fig. 1(c) and (d). Here we have Energy Distribution Curves (EDCs) from Ding taken at hv=19 eV along the high symmetry cuts $`\mathrm{\Gamma }\overline{M}`$ and $`\overline{M}X`$, where $`\mathrm{\Gamma }=(0,0)`$ and $`\overline{M}=(\pi ,0)`$. From these two panels, the authors claim there is no main band FS crossing along $`\mathrm{\Gamma }\overline{M}`$ but that there is one along $`\overline{M}X`$, as evidenced by the relatively rapid loss of peak weight in this region.
Figure 1(e) presents a small portion of our new normal-state ARPES data at $`33eV`$ obtained from an overdoped Bi2212 sample along the high symmetry cut $`\mathrm{\Gamma }\overline{M}Z`$ ($`Z=(2\pi ,0)`$). The energy scale for these spectra is the same as in Ding’s data in panels (c) and (d). Our peaks are sharper and better resolved than Ding’s, showing the very high quality of the data set. Clear dispersion is observed with the peak reaching $`E_F`$ near $`.81\pi `$ (red curve). At this same k-value, the peak rapidly loses weight, indicating a FS crossing. The peak reappears in the second zone and disperses back towards higher binding energy (BE). Although only a single cut, this data already indicates that there is a fundamental difference between our data at the new photon energy of $`33eV`$, and the previous data of Ding et al. Our data indicates main-band FS crossings along $`\mathrm{\Gamma }\overline{M}Z`$, while they are not expected according to the accepted hole-like FS topology. In addition, the flat bands at $`E_F`$ at $`\overline{M}`$ (extended van-Hove singularity) observed near $`20eV`$ have been replaced at $`33eV`$ by a strongly dispersive band which crosses $`E_F`$.
To analyze the data in more detail we have made $`\stackrel{}{k}`$-space plots of the integrated spectral intensity $`n(\stackrel{}{k})`$ as well as the weight right at $`E_F`$ which we term $`E_F(\stackrel{}{k})`$. We first illustrate how $`n(\stackrel{}{k})`$ and $`E_F(\stackrel{}{k})`$ are expected to behave for a simple band-like state crossing the FS. Fig. 2(a) shows the zero-temperature $`n(\stackrel{}{k})`$ and $`E_F(\stackrel{}{k})`$ for a non-interacting system. In this case there is only weight at $`E=E_F`$ when $`\stackrel{}{k}=\stackrel{}{k}_F`$. In Fig. 2(b) we introduce interactions, which within the Fermi Liquid theory framework will reduce both the weight of the delta function peak in $`E_f(\stackrel{}{k})`$ and the step in $`n(\stackrel{}{k})`$ to the value Z (quasiparticle weight). Finite experimental energy and momentum resolution will broaden both curves, as illustrated in Fig. 2(c). Finally, polarization and matrix element effects will slowly alter $`n(\stackrel{}{k})`$, as shown in Fig. 2(d). It is clear from the above plots that at a true FS crossing the following criteria should be obeyed: (1) $`E_F(\stackrel{}{k})`$ should be maximal. (2) $`n(\stackrel{}{k})`$ should be at 50% of it’s maximal value, or equivalently at the maximal gradient point. Also, at $`E_F`$ we expect (3) the peak dispersion to extrapolate to zero energy, and (4) the midpoint of the leading edge of the spectrum to be at or even beyond (on the unoccupied side of) $`E_F`$.
Panels (e)-(i) of figure 2 show real $`n(\stackrel{}{k})`$ and $`E_F(\stackrel{}{k})`$ plots from our new data on the BSCCO family. To our knowledge, this is the first time that $`n(\stackrel{}{k})`$ and $`E_F(\stackrel{}{k})`$ plots have been analyzed together, which turns out to be a powerful new tool to obtain FS crossings. To determine $`n(\stackrel{}{k})`$ we integrated the ARPES spectral weight from $`500meV`$ to $`+100meV`$ so as to span the full energy width of the peak, and for $`E_F(\stackrel{}{k})`$ we integrated over a $`50meV`$ wide window centered at $`E_F`$. All plots were normalized so that the maximum weight along $`\mathrm{\Gamma }\overline{M}Z`$ was set to $`1`$. According to criteria (1) and (2), a FS crossing should occur when $`n(\stackrel{}{k})`$ loses half of its maximum value (excluding the background) and $`E_F(\stackrel{}{k})`$ simultaneously peaks. These points are indicated in the figure by the dashed green lines. Panel (e) shows $`n(\stackrel{}{k})`$ and $`E_F(\stackrel{}{k})`$ obtained from the raw data of Fig. 1 (e). As expected, this way of analyzing the data (criteria 1 and 2) gives identical FS crossings as obtained by studying the peak dispersion (criteria 3 and 4). Importantly, the drastic drop in $`n(\stackrel{}{k})`$ at $`\overline{M}`$ to a value comparable to that at $`\mathrm{\Gamma }`$ or $`Z`$ indicates that this crossing must be due to the main band and not due to the crossing of a weak superstructure band. Also, the drop in $`n(\stackrel{}{k})`$ at the $`Z`$ point to a level equivalent to that at the $`\mathrm{\Gamma }`$ point can not be explained by photon polarization or orbital symmetry arguments. Although not central to this paper, this observation includes new physics which warrants much further experimental and theoretical attention.
Panels (f) and (g) show data taken at hv=$`22eV`$ from a similar sample used to make the $`33eV`$ data of panel (e). Our raw $`22eV`$ data looks qualitatively similar to Ding’s data of figure 1(b) or to other previously published data including the presence of flat bands at $`E_F`$ near $`\overline{M}`$. The data of panels (f) and (g) show that the peak in $`E_F(\stackrel{}{k})`$ and the $`50\%`$ point of $`n(\stackrel{}{k})`$ do not coincide, but rather the peaks of each coincide. This behavior is extremely unusual. Furthermore, the data of panel (f) is highly asymmetric about $`\overline{M}`$ (in the 2d approximation $`\mathrm{\Gamma }`$ would be equivalent to $`Z`$ and $`\overline{M}`$ would be a real high symmetry point), in contrast to the much more symmetric behavior observed at $`33eV`$. This strange behavior makes us question the nature of the states probed near $`\overline{M}`$ at photon energies near $`22`$ eV.
The interesting behavior of the data at $`33eV`$ calls for a complete mapping of the FS topology at this new photon energy. We have taken many cuts over the Brillouin zone on a slightly overdoped Bi2212 sample (different from that used in figures (1) and (2)). Within experimental resolution, each of the FS crossing criteria gave identical crossing locations for each cut. The crossing points indicated from these cuts are shown in Fig. 3. We have connected these points with thick lines indicating the main FS (stronger ARPES peaks) and thin lines indicating the superstructure-derived FS (ARPES peaks of about $`30\%`$ the intensity of the main peaks). The superstructure FS’s are seen to be replicas of the main FS’s, but shifted by $`\pm 0.2(\pi ,\pi )`$. We note that the FS topology measured here is closed around the $`\mathrm{\Gamma }`$ point, i.e. it is electron-like. This is manifestly different from the “accepted” hole-like FS topology of Bi2212.
Despite this discrepancy, there is data in the literature that confirms our new FS topology. Figure 3 shows an overlay of our new FS with an $`E_F`$ intensity plot from an optimally doped Bi2212 crystal at 33 eV measured by Saini et al. The $`E_F`$ intensity is analogous to our $`E_F(\stackrel{}{k})`$), i.e. the maximum intensity locations should correspond to FS crossings. The overlay shows a striking similarity between these two independently obtained results. However, instead of interpreting their data as an electron-like FS, Saini et al argued that it was still representative of a hole-like FS centered around $`X`$ and $`Y`$. They attributed the loss of weight at $`\overline{M}`$ to a pseudogap in the spectral function, and further claimed that the high intensity cusps orthogonal to the $`\mathrm{\Gamma }Y`$ direction were due to the formation of one-dimensional charge stripes. Our experiments and analysis attribute the weight loss around $`\overline{M}`$ to a FS crossing. Indeed our data of figure 1 (e) shows no evidence for a pseudogap in these samples (the signature of a pseudogap in ARPES is that at $`k=k_F`$ the leading edge of the spectrum is depressed from $`E_F`$). Figure 1(e) shows that the midpoint of our leading edge is fully up to or even past $`E_F`$. Our data also indicates that the periodic arrangement of high intensity cusps are due to the superstructure bands. Mesot et al have also argued that superstructure bands and not stripes are important for understanding Saini’s data. However, in their picture the data is analyzed starting from the standard hole-like FS and so they are not capable of explaining the near-complete weight loss observed at $`\overline{M}`$.
Figs. 2 (h) and (i) show that the new behavior detailed here is robust as a function of sample type. These plots show $`n(\stackrel{}{k})`$ and $`E_F(\stackrel{}{k})`$ along $`\mathrm{\Gamma }\overline{M}Z`$ at hv’s near 33 eV for a heavily underdoped Bi2212 sample and for an overdoped sample of the single-layer compound Bi2201. In both cases the data indicates a main band crossing between $`\mathrm{\Gamma }`$ and $`\overline{M}`$, indicating that the FS topology should also be electron-like for these samples (closed around $`\mathrm{\Gamma }`$) (the crossing is more washed out for the underdoped sample, but appears to be qualitatively similar in other respects). This indicates that the new physics shown here is not peculiar to one doping level or sample type. Rather, we argue that it is a product of the new photon energy range used for our measurements.
Changing the photon energy in an ARPES experiment can have a number of effects. For a fixed $`k_{}`$, the most obvious effect is that $`k_{}`$ will change since the magnitude of the total momentum must change. In this way, variations in the electronic structure vs. $`k_{}`$ may be mapped out, i.e. we can map out the full three-dimensional electronic structure. We have taken data at many more photon energies (not shown here) to check whether the Fermi surface topology oscillates as a function of $`k_{}`$. We did not observe any clear effects with a periodicity in $`k_{}`$ of $`2\pi /c`$ where $`c`$ is the c-axis lattice constant. Thus we conclude that the differences in the data as a function of photon energy are not naturally linkable to a coherent three-dimensionality of the band structure. This is consistent with the huge in-plane vs. out-of-plane transport anisotropy of these materials.
Figure 4a shows the generally accepted $`E`$ vs. $`\stackrel{}{k}`$ relation for near-optimal BSCCO, with a key feature being the large $`\stackrel{}{k}`$-space region near $`\overline{M}`$ which has flat bands just below $`E_F`$. Our figure 2g indicates that the crossing along $`\overline{M}Y`$ is not robust, so we conclude that figure 4a is not a good representation of the physics. Rather, we propose the scenario in figures 4b and 4c. The main band dispersion shown by the red lines is as we have measured at $`33eV`$, and is electron-like with a small saddle-point at $`\overline{M}`$ above $`E_F`$. Additionally, there is a large region of flat bands at $`\overline{M}`$ which is observed at $`22eV`$ but not $`33eV`$. At 22 eV these states act to mask the true crossing behavior of the bands observed at $`33eV`$, giving the impression of a hole-like FS. An important goal for future studies will be to elucidate the origin of these additional states, including whether or not they are intrinsic to the basic electronic structure of BSCCO. Possible origins include quantum confinement due to stripe formation, a contribution from the BiO states, indirect transitions , photoelectron diffraction effects , final state (i.e. matrix element) effects , or possibly a new type of correlated electron state due for instance to magnetism.
Considering that electron transport should be dominated by the dispersive states we have observed at $`33eV`$, we are left with the puzzling result that the FS topology looks electron-like, while Hall effect measurements indicate that the carriers should be hole-like. A similar disagreement has been reported for the n-type superconductor $`Nd_{2x}Ce_xCuO_4`$ \- Hall effect measurements have indicated an electron-like FS while ARPES results have indicated a hole-like FS. These results indicate a non-simple relationship between the Hall resistance and the electronic structure in these materials. This again highlights the importance of physics beyond the band structure (i.e. correlation effects) in the cuprates.
In conclusion, by invoking a more complete set of FS crossing criteria and by going to a non-traditional photon energy range we have discovered an electron-like FS topology and the absence of flat bands at $`E_F`$ in the BSCCO family of hole-doped superconductors. The data is robust as a function of doping and is clearer than the previous data which was interpreted as indicating a hole-like topology. The main dissimilarity in the data sets occurs near $`\overline{M}`$, which is a critical location in the Brillouin zone - it is where both the superconducting gap and normal state pseudogap were found to reach maximum amplitude. Clearly this calls for more experimental and theoretical works considering this new FS topology as well as the effects of varying photon energy.
We acknowledge support from the Office of Naval Research’s Young Investigator Program. The SRC is supported by the National Science Foundation, and SSRL is operated by the Department of Energy, Office of Basic Energy Sciences. We thank Aharon Kapitulnik for the use of one Bi2212 sample.
|
no-problem/9904/cond-mat9904177.html
|
ar5iv
|
text
|
# Probability Distribution of the Shortest Path on the Percolation Cluster, its Backbone and Skeleton
## Abstract
We consider the mean distribution functions $`\mathrm{\Phi }(r|\mathrm{})`$, $`\mathrm{\Phi }_\mathrm{B}(r|\mathrm{})`$, and $`\mathrm{\Phi }_\mathrm{S}(r|\mathrm{})`$, giving the probability that two sites on the incipient percolation cluster, on its backbone and on its skeleton, respectively, connected by a shortest path of length $`\mathrm{}`$ are separated by an Euclidean distance $`r`$. Following a scaling argument due to de Gennes for self-avoiding walks, we derive analytical expressions for the exponents $`g_1=d_f+d_{\mathrm{min}}d`$ and $`g_1^\mathrm{B}=g_1^\mathrm{S}=3d_{\mathrm{min}}d`$, which determine the scaling behavior of the distribution functions in the limit $`xr/\mathrm{}^{\stackrel{~}{\nu }}1`$, i.e. $`\mathrm{\Phi }(r|\mathrm{})\mathrm{}^{\stackrel{~}{\nu }d}x^{g_1}`$, $`\mathrm{\Phi }_\mathrm{B}(r|\mathrm{})\mathrm{}^{\stackrel{~}{\nu }d}x^{g_1^\mathrm{B}}`$, and $`\mathrm{\Phi }_\mathrm{S}(r|\mathrm{})\mathrm{}^{\stackrel{~}{\nu }d}x^{g_1^\mathrm{S}}`$, with $`\stackrel{~}{\nu }1/d_{\mathrm{min}}`$, where $`d_f`$ and $`d_{\mathrm{min}}`$ are the fractal dimensions of the percolation cluster and the shortest path, respectively. The theoretical predictions for $`g_1`$, $`g_1^\mathrm{B}`$ and $`g_1^\mathrm{S}`$ are in very good agreement with our numerical results.
Percolation constitutes a useful model for a variety of disordered systems in many fields of science displaying both structural disorder and self-similarity (i.e. fractal behavior) within some range of length scales . In many circumstances, the knowledge of the internal structure of percolation clusters is required, as for instance in the study of transport processes near the percolation threshold $`p_\mathrm{c}`$, where the complex topology of the available conducting paths play a crucial role .
It is known that at the percolation threshold $`p_\mathrm{c}`$, the incipient infinite cluster displays fractal behavior over all length scales, i.e. its mass $`s`$ contained within a distance $`r`$ from a given cluster site chosen as the origin, averaged over many origins, scales as $`sr^{d_f}`$, where $`d_f=91/48`$ in two dimensions, $`d_f=2.524\pm 0.008`$ in three dimensions, and $`d_f=4`$ above the critical dimension, i.e. when $`dd_\mathrm{c}=6`$ . A second, useful metric is the “chemical” distance $`\mathrm{}`$ between two cluster sites , defined as the length of the shortest path connecting them. It is found that the mean distance $`r`$ between two cluster sites, averaged over many pairs of sites, behaves as a function of $`\mathrm{}`$ as $`r\mathrm{}^{1/d_{\mathrm{min}}}`$, where $`d_{\mathrm{min}}=1.130\pm 0.004`$ in $`d=2`$ , $`d_{\mathrm{min}}=1.374\pm 0.004`$ in $`d=3`$ , and $`d_{\mathrm{min}}=2`$ when $`dd_\mathrm{c}`$, is the so-called fractal dimension of the shortest path. From the above scaling relations follow that in “chemical” space, the mass of the cluster scales with distance $`\mathrm{}`$ as $`s\mathrm{}^d_{\mathrm{}}`$, where $`d_{\mathrm{}}=d_f/d_{\mathrm{min}}`$, with $`d_{\mathrm{}}=2`$ when $`dd_\mathrm{c}`$ .
The incipient infinite cluster exhibits a variety of substructures that are self-similar as well . A prominent example is the backbone of the cluster, defined as the subset of cluster sites that can carry a current when a potential difference is applied between two sites (see and references therein). Thus, the structure of the backbone alone determines the conductivity of the whole percolation network between two sites. The structural and dynamical properties of the backbone of the incipient cluster have been studied recently . A second cluster substructure, denoted as the skeleton (a subset of the backbone, also called the “elastic” backbone) is defined as the union of all shortest paths between the two cluster sites.
In this Rapid Communication, we extend our previous studies of the structural properties of the incipient infinite cluster and its backbone in two and three dimensions. We consider the structural distribution function $`\mathrm{\Phi }(r|\mathrm{})`$ for the incipient infinite cluster, where $`\mathrm{\Phi }(r|\mathrm{})\mathrm{d}r`$ is the probability that two cluster sites connected by a shortest path of length $`\mathrm{}`$ are at Euclidean distance between $`r`$ and $`r+\mathrm{d}r`$ from each other in space. The probability distribution $`\mathrm{\Phi }(r|\mathrm{})`$ is normalized according to $`r^{d1}\mathrm{\Phi }(r|\mathrm{})dr=1`$, and is found to obey an scaling behavior with the variable $`xr/\mathrm{}^{\stackrel{~}{\nu }}`$ of the form $`\mathrm{\Phi }(r|\mathrm{})=\mathrm{}^{\stackrel{~}{\nu }d}f(x)`$ (see e.g. ), where $`\stackrel{~}{\nu }1/d_{\mathrm{min}}`$. Here, we draw our attention to the limit $`x1`$, where the scaling function $`f(x)`$ follows a simple power law, $`f(x)x^{g_1}`$, i.e.
$$\mathrm{\Phi }(r|\mathrm{})\frac{1}{\mathrm{}^{\stackrel{~}{\nu }d}}\left(\frac{r}{\mathrm{}^{\stackrel{~}{\nu }}}\right)^{g_1},\text{for }r/\mathrm{}^{\stackrel{~}{\nu }}1.$$
(1)
Similar scaling forms for the substructural distribution functions as a function of $`xr/\mathrm{}^{\stackrel{~}{\nu }}`$, $`\mathrm{\Phi }_\mathrm{B}(r|\mathrm{})=\mathrm{}^{\stackrel{~}{\nu }d}f_\mathrm{B}(x)`$ for the backbone and $`\mathrm{\Phi }_\mathrm{S}(r|\mathrm{})=\mathrm{}^{\stackrel{~}{\nu }d}f_\mathrm{S}(x)`$ for the skeleton, are expected . In the case $`x1`$, the corresponding scaling functions, $`f_\mathrm{B}(x)`$ and $`f_\mathrm{S}(x)`$, are found to behave as $`f_\mathrm{B}(x)x^{g_1^\mathrm{B}}`$ and $`f_\mathrm{S}(x)x^{g_1^\mathrm{S}}`$, respectively, yielding
$$\mathrm{\Phi }_\mathrm{B}(r|\mathrm{})\frac{1}{\mathrm{}^{\stackrel{~}{\nu }d}}\left(\frac{r}{\mathrm{}^{\stackrel{~}{\nu }}}\right)^{g_1^\mathrm{B}}\text{and}\mathrm{\Phi }_\mathrm{S}(r|\mathrm{})\frac{1}{\mathrm{}^{\stackrel{~}{\nu }d}}\left(\frac{r}{\mathrm{}^{\stackrel{~}{\nu }}}\right)^{g_1^\mathrm{S}},\text{for }r/\mathrm{}^{\stackrel{~}{\nu }}1.$$
(2)
Numerical results (see Refs. and below) indicate that $`g_1<g_1^\mathrm{B}g_1^\mathrm{S}`$ in both two and three dimensions. For $`dd_\mathrm{c}`$, one expects the mean field (MF) values $`g_1=g_1^\mathrm{B}=g_1^\mathrm{S}=0`$, since percolation clusters behave similarly to simple random walks above the critical dimension $`d_\mathrm{c}`$ .
We first study the above defined distribution functions numerically, both in two and three dimensions. To this end, we generate large percolation cluster at $`p_\mathrm{c}`$ on square and s.c. lattices, respectively, using the well-known Leath algorithm . To identify the backbone and skeleton of the cluster, we apply an improved version of the well-known burning algorithm . We perform averages over more than $`10^5`$ clusters, which are grown until they reach a maximum of chemical shells $`\mathrm{}_{\mathrm{max}}=2000`$ in $`d=2`$ and $`\mathrm{}_{\mathrm{max}}=1000`$ in $`d=3`$. The results for $`\mathrm{\Phi }(r|\mathrm{})`$, $`\mathrm{\Phi }_\mathrm{B}(r|\mathrm{})`$, and $`\mathrm{\Phi }_\mathrm{S}(r|\mathrm{})`$ are shown in Figs. 1, 2, and 3, respectively. For the incipient infinite cluster we obtain $`g_1=1.04\pm 0.05`$ in $`d=2`$ and $`g_1=0.88\pm 0.05`$ in $`d=3`$ (see also ). For the backbone, we find $`g_1^\mathrm{B}=1.34\pm 0.10`$ in $`d=2`$ and $`g_1^\mathrm{B}=1.08\pm 0.10`$ in $`d=3`$ (see also ). In addition, our results suggest that $`\mathrm{\Phi }_\mathrm{B}(r|\mathrm{})`$ and $`\mathrm{\Phi }_\mathrm{S}(r|\mathrm{})`$ coincide, within the accuracy of the present data, and as a result, the values of $`g_1^\mathrm{S}`$ for the skeleton are indistinguishable from those of the backbone, i.e. $`g_1^\mathrm{S}g_1^\mathrm{B}`$. These results are summarized in Table I.
To estimate values for the exponents $`g_1`$, $`g_1^\mathrm{B}`$ and $`g_1^\mathrm{S}`$ analytically, we follow a method similar to the one discussed by de Gennes for determining the structure of self-avoiding walks (SAW) of $`N`$ steps. The latter is described by the probability distribution function $`P_{\mathrm{SAW}}(r|N)=N^{\nu d}f_{\mathrm{SAW}}(y)`$, with $`yr/N^\nu `$, where $`P_{\mathrm{SAW}}(r|N)\mathrm{d}r`$ gives the probability that the two end points of a SAW of fixed length $`N`$ (i.e. the first and the $`N+1`$ monomers) are at a distance between $`r`$ and $`r+\mathrm{d}r`$. Here, $`\nu `$ is the Flory exponent, $`\nu (d+2)/3`$ for $`d4`$ and $`\nu =\nu _{\mathrm{MF}}=\frac{1}{2}`$ for $`d4`$, and $`f_{\mathrm{SAW}}(y)`$ is the scaling function, with $`f_{\mathrm{SAW}}(y)y^g`$ when $`y1`$. For SAW defined on the lattice, de Gennes argues that the behavior of $`f_{\mathrm{SAW}}(y)`$ for $`y1`$ can be obtained by considering the probability $`P_{\mathrm{SAW}}(r1|N)`$ that a SAW of $`N1`$ steps returns close to its starting point (origin), which can be written as
$$P_{\mathrm{SAW}}(r1|N)\frac{N_{\mathrm{SAW}}^{r1}(N)}{N_{\mathrm{SAW}}(N)},\text{for }N1,$$
(3)
where $`N_{\mathrm{SAW}}^{r1}(N)N^{\nu d}\overline{z}^N`$ is the number of SAW of length $`N`$ returning close to the origin and $`N_{\mathrm{SAW}}(N)N^{\gamma 1}\overline{z}^N`$ is the total number of SAW of length $`N`$. Here, $`\overline{z}`$ is the effective coordination number of the lattice, and $`\gamma `$ is the enhancement exponent, with $`\gamma =\gamma _{\mathrm{MF}}=1`$ for $`d4`$.
As noted by de Gennes , the enhancement factor $`N^{\gamma 1}`$ occurs only in the denominator of the ratio $`N_{\mathrm{SAW}}^{r1}(N)/N_{\mathrm{SAW}}(N)`$, but not in the numerator, indicating the “difficulty” for a SAW to return near to its starting point. Note that this missing enhancement factor in the numerator can be viewed as corresponding to its mean-field value, $`N^{\gamma _{\mathrm{MF}}1}1`$, and one can write equivalently
$$P_{\mathrm{SAW}}(r1|N)\frac{1}{N^{\nu d}}\frac{N^{\gamma _{\mathrm{MF}}1}}{N^{\gamma 1}},\text{for }N1,$$
(4)
corresponding to the behavior $`f_{\mathrm{SAW}}(y)y^g`$, with $`g=(\gamma 1)/\nu `$. This observation suggested us a procedure for describing the structural function of the incipient percolation cluster and its substructures analytically, in the case $`r/\mathrm{}^{\stackrel{~}{\nu }}1`$. We consider the incipient percolation cluster first.
Let us generalize Eq. (4) to percolation clusters by writing the distribution function $`\mathrm{\Phi }(r|\mathrm{})`$, for a chemical distance $`\mathrm{}1`$ and Euclidean distance $`r1`$, as
$$\mathrm{\Phi }(r1|\mathrm{})\frac{1}{\mathrm{}^{\stackrel{~}{\nu }d}}\frac{\mathrm{\Pi }_{\mathrm{MF}}(\mathrm{})}{\mathrm{\Pi }(\mathrm{})},\text{for }\mathrm{}1,$$
(5)
where $`\mathrm{\Pi }(\mathrm{})`$ plays the role of the function $`N^{\gamma 1}`$ in Eq. (4), and $`\mathrm{\Pi }_{\mathrm{MF}}(\mathrm{})`$ denotes its mean field value. Here we argue that, to a first approximation, $`\mathrm{\Pi }(\mathrm{})`$ is given by the probability that the two chosen sites are on a cluster of chemical size $`\mathrm{}`$. Therefore we relate $`\mathrm{\Pi }(\mathrm{})`$ to the probability distribution of cluster sizes $`sn(s)`$, which is known to behave as $`sn(s)s^{(\tau 1)}`$, with $`\tau =1+d/d_f`$ for $`dd_\mathrm{c}`$, and $`\tau _{\mathrm{MF}}=5/2`$ . Hence, $`\mathrm{\Pi }(\mathrm{})`$ is given by $`\mathrm{\Pi }(\mathrm{})sn(s)\mathrm{d}s/\mathrm{d}\mathrm{}`$, and noting that $`s\mathrm{}^d_{\mathrm{}}`$, we find $`\mathrm{\Pi }(\mathrm{})\mathrm{}^{d_{\mathrm{}}(\tau 2)1}`$ for $`dd_\mathrm{c}`$, and $`\mathrm{\Pi }_{\mathrm{MF}}(\mathrm{})\mathrm{}^2`$. Thus, Eq. (5) becomes
$$\mathrm{\Phi }(r1|\mathrm{})\frac{1}{\mathrm{}^{\stackrel{~}{\nu }d}}\frac{\mathrm{}^2}{\mathrm{}^{d_{\mathrm{}}(\tau 2)1}}\frac{1}{\mathrm{}^{\stackrel{~}{\nu }d}}\mathrm{}^{d_{\mathrm{}}(\tau 2)1},\text{for }\mathrm{}1.$$
(6)
Comparing this result with the one obtained from Eq. (1) in the limit $`r1`$, yields $`\stackrel{~}{\nu }g_1=d_{\mathrm{}}(\tau 2)1`$, i.e.
$$g_1=d_f+d_{\mathrm{min}}d,$$
(7)
which predicts $`g_1=1.026\pm 0.004`$ for $`d=2`$ and $`g_1=0.898\pm 0.008`$ for $`d=3`$. These theoretical values for $`g_1`$ are in very good agreement with our numerical results (cf. Fig. 1 and Table I). Note that Eq. (7) yields by construction $`g_1=0`$ for $`dd_\mathrm{c}`$, as required.
The above argument can be applied straightforwardly to the backbone and the skeleton of the incipient cluster, where now analogous equations to Eq. (5) can be written for $`\mathrm{\Phi }_\mathrm{B}(r1|\mathrm{})`$ and $`\mathrm{\Phi }_\mathrm{S}(r1|\mathrm{})`$, with $`\mathrm{\Pi }(\mathrm{})`$ replaced by $`\mathrm{\Pi }_\mathrm{B}(\mathrm{})`$ and $`\mathrm{\Pi }_\mathrm{S}(\mathrm{})`$, respectively. In the case of the backbone, we argue that $`\mathrm{\Pi }_\mathrm{B}(\mathrm{})n(s)\mathrm{d}s/\mathrm{d}\mathrm{}`$, with $`n(s)s^\tau `$, and $`s\mathrm{}^d_{\mathrm{}}`$ as for the incipient cluster. Note the absence of the factor $`s`$ in the expression for $`\mathrm{\Pi }_\mathrm{B}(\mathrm{})`$, reflecting the fact that the backbone represents a subset of the incipient cluster having a vanishing measure when $`s\mathrm{}`$ . Since the same argument applies to the skeleton, we have that $`\mathrm{\Pi }_\mathrm{S}(\mathrm{})\mathrm{\Pi }_\mathrm{B}(\mathrm{})`$, yielding
$$\mathrm{\Phi }_\mathrm{S}(r1|\mathrm{})\mathrm{\Phi }_\mathrm{B}(r1|\mathrm{}),\text{for }\mathrm{}1,$$
(8)
in agreement with the numerical results shown in Figs. 2 and 3. In terms of $`\mathrm{\Pi }_\mathrm{B}(\mathrm{})`$, $`\mathrm{\Phi }_\mathrm{B}(r|\mathrm{})`$ in the limit $`r1`$ is given by
$$\mathrm{\Phi }_\mathrm{B}(r1|\mathrm{})\frac{1}{\mathrm{}^{\stackrel{~}{\nu }d}}\frac{\mathrm{\Pi }_{\mathrm{B},\mathrm{MF}}(\mathrm{})}{\mathrm{\Pi }_\mathrm{B}(\mathrm{})},\text{for }\mathrm{}1,$$
(9)
and with $`\mathrm{\Pi }_\mathrm{B}(\mathrm{})\mathrm{}^{d_{\mathrm{}}(\tau 1)1}`$
$$\mathrm{\Phi }_\mathrm{B}(r1|\mathrm{})\frac{1}{\mathrm{}^{\stackrel{~}{\nu }d}}\frac{\mathrm{}^4}{\mathrm{}^{d_{\mathrm{}}(\tau 1)1}}\frac{1}{\mathrm{}^{\stackrel{~}{\nu }d}}\mathrm{}^{d_{\mathrm{}}(\tau 1)3},\text{for }\mathrm{}1.$$
(10)
Comparing this result with the scaling form for $`\mathrm{\Phi }_\mathrm{B}(r|\mathrm{})`$ given in Eq. (2) in the limit $`r1`$, yields $`\stackrel{~}{\nu }g_1^\mathrm{B}=d_{\mathrm{}}(\tau 1)3`$, i.e.
$$g_1^\mathrm{B}=3d_{\mathrm{min}}d,$$
(11)
predicting $`g_1^\mathrm{B}=1.390\pm 0.004`$ in $`d=2`$ and $`g_1^\mathrm{B}=1.122\pm 0.004`$ in $`d=3`$, with $`g_1^\mathrm{S}=g_1^\mathrm{B}`$, in remarkable agreement with the numerical results (cf. Figs. 2 and 3, and Table I). Note also that from Eqs. (11) and (8) one obtains $`g_1^\mathrm{B}=g_1^\mathrm{S}=0`$ for $`dd_\mathrm{c}`$, as expected.
In summary, we derive the analytical expressions $`g_1=d_f+d_{\mathrm{min}}d`$ and $`g_1^\mathrm{B}=g_1^\mathrm{S}=3d_{\mathrm{min}}d`$ describing the scaling behavior of the structural distribution functions, $`\mathrm{\Phi }(r|\mathrm{})\mathrm{}^{\stackrel{~}{\nu }d}x^{g_1}`$, $`\mathrm{\Phi }_\mathrm{B}(r|\mathrm{})\mathrm{}^{\stackrel{~}{\nu }d}x^{g_1^\mathrm{B}}`$, and $`\mathrm{\Phi }_\mathrm{S}(r|\mathrm{})\mathrm{}^{\stackrel{~}{\nu }d}x^{g_1^\mathrm{S}}`$, of the incipient percolation cluster, its backbone and skeleton, respectively, at the critical concentration $`p_\mathrm{c}`$ in the limit $`xr/\mathrm{}^{\stackrel{~}{\nu }}1`$. Here, $`\stackrel{~}{\nu }1/d_{\mathrm{min}}`$, and $`d_f`$ and $`d_{\mathrm{min}}`$ are the fractal dimensions of the incipient percolation cluster and the shortest path, respectively. Note that from the above expressions for the exponents $`g_1`$, $`g_1^\mathrm{B}`$, and $`g_1^\mathrm{S}`$ follow that the corresponding distribution functions for $`\mathrm{}1`$, in the limit $`r1`$, scale as $`\mathrm{\Phi }(r1|\mathrm{})\mathrm{}^{(d_{\mathrm{}}+1)}`$ and $`\mathrm{\Phi }_\mathrm{B}(r1|\mathrm{})\mathrm{\Phi }_\mathrm{S}(r1|\mathrm{})\mathrm{}^3`$, the latter being independent of the lattice dimension $`d`$. We note that the result $`\mathrm{\Phi }(r1|\mathrm{})\mathrm{}^{(d_{\mathrm{}}+1)}`$ for $`\mathrm{}1`$, based on numerical simulations, was also suggested for two other variants of percolation, invasion percolation with as well as without trapping , and seems therefore to be more general.
We acknowledge useful discussions with I. Webman, J. Dräger and A. Ordemann. This work has been supported by the Minerva Center for the Physics of Mesoscopics, Fractals and Neural Network; the German-Israeli Foundation; the Alexander von Humboldt Foundation and the Deutsche Forschungsgemeinschaft.
|
no-problem/9904/gr-qc9904064.html
|
ar5iv
|
text
|
# I Introduction
## I Introduction
The study involved in the attempt to detect gravitational waves represents one of the most exciting field in modern physics. The detection will provide a decisive test of general relativity and open a new window for astrophysical observation. Unfortunately, in spite a great deal of effort, no one has yet succeeded in the detection of the gravitational waves. However ground-based laser interferometers, such as LIGO, VIRGO, TAMA and GEO600, are being constructed and will soon enter a stage of data taking. Strong sources of gravitational waves should be regions where gravity is general relativistic and where the velocities of bulk motion are near the speed of light. Many researchers have theoretically investigated coalescing and colliding black holes and neutron stars as candidates for detection. In addition, gravitational waves emitted during stellar collapse to black holes have been studied by perturbative calculations of a spherically symmetric background. Because of the existence of event horizon, the emitted gravitational waves are dominated by the quasi-normal modes of the Schwarzschild black hole in the asymptotic region. Some other realistic candidates exist, e.g., supernovae, rapidly rotating neutron stars, relic gravitational waves of the early universe, stochastic gravitational waves from cosmic strings, and so on.
In this paper we attempt to investigate whether a naked singularity, if such exists, is a strong source of gravitational radiation, and, moreover, to understand the dynamics and observational meaning of naked singularity formation. Several researchers have shown that the final fate of gravitational collapse is not always a singularity covered by an event horizon. For example, in the Lemaître-Tolman-Bondi (LTB) spacetime, a naked shell-focusing singularity appears from generic initial data for spherically symmetric configurations of the rest mass density and a specific energy of the dust fluid. The initial functions in the most general expandable form have been considered. The matter content in such a spacetime may satisfy even the dominant energy condition. In this case with a small disturbance of the spacetime, very short wavelength gravitational waves, which are created in the high density region around a singularity, may propagate to the observer outside the dust cloud because of the absence of an event horizon. If this is true, extremely high energy phenomena which cannot be realized by any high energy experiment on Earth can be observed. In this case information regarding the physics of so-called ‘quantum gravity’ may be obtained. Also, these waves may be so intense that they destroy the Cauchy horizon. In this paper the generation of gravitational waves during the collapse of spherical dust ball with small rotational motion is considered.
Nakamura, Shibata and Nakao have suggested that a naked singularity may emit considerable gravitational wave radiation. This was proposed from the estimate of gravitational radiation from a spindle-like naked singularity. They modeled the spindle-like naked singularity formation in gravitational collapse by the Newtonian prolate dust collapse for a sequence of general relativistic, momentarily static initial data. It should be noted that the system they considered is different from that considered in this article and that their result is controversial. There are numerical analyses that both support and do not support the results of Nakamura, Shibata and Nakao for prolate collapse and for cylindrical collapse.
Due to the non-linear nature of the problem, it is difficult to analytically solve the Einstein equation. Therefore, numerical methods will provide the final tool. However, its singular behavior makes accurate numerical analysis very difficult at some stage. In this article, we investigate odd-parity linear gravitational waves from the collapse of an inhomogeneous spherically symmetric dust cloud. Even for the linearized Einstein equation, we must perform the numerical integration. However, in contrast to the numerical simulation of the full Einstein equation, high precision is guaranteed for the numerical integration of the linearized Einstein equation, even in the region with extremely large spacetime curvature. Furthermore, the linear stability of known examples of naked singularity formation is necessary as a first step to understand the general dynamics near the “naked singularity formation.”
Recently, Iguchi, Nakao and Harada (INH) studied odd-parity metric perturbations around a naked singularity in the LTB spacetime. In INH, it was found that the propagation of odd-parity gravitational waves is not affected by collapse of a dust cloud, even if there appears a central naked singularity. When we consider the generation of gravitational waves from the dust collapse, we should analyze the perturbations including their matter part. Here we investigate the evolution equation of the odd-parity quadrupole mode for metric and matter perturbations. This matter perturbation relates to small rotational motion, and it produces gravitational waves during the collapse. We follow the derivation of the evolution equations of the perturbations and the numerical method to integrate these equations of INH. We investigate the time evolutions of the gauge invariant metric variables at the symmetric center, where a naked singularity appears, and at a constant circumferential radius $`R`$. We show that the gauge invariant variable diverges only at the center and it does not propagate to the outside.
As we know, the LTB spacetime is one candidate as a counterexample of the cosmic censorship hypothesis (CCH), which was introduced by Penrose. The CCH is a very helpful assumption, because various theorems on the properties of black holes were proved under this assumption. Here the precise formulation and validity of cosmic censorship is not our concern. We only consider the situation in which the extremely high-density and large-curvature region can be seen by some observer from the gravitational collapse. Such a situation can be regarded as a naked singularity in a practical sense since we are not yet able to predict phenomena beyond the Planck scale. However, here it should be noted that even from this practical point of view, the stability of the Cauchy horizon with respect to gravitational-wave perturbations still has an important physical meaning. Since in the system considered here, the mode propagating with the speed of light is the gravitational waves only, the instability of the Cauchy horizon implies that the gravitational wave introduces an extremely large spacetime curvature along the null hypersurface near the Cauchy horizon from the region near the central singularity. Hence, if the Cauchy horizon in this system is unstable, naked singularity formation of this type might be a strong source of gravitational radiation.
Here we comment on the problem motivated in INH, i.e., “nakedness of the naked singularity.” Odd-parity matter perturbations are produced by rotational motion of the dust cloud. Therefore the growth of the matter perturbation would cause the centrifugal force to dominate the radial motion of the cloud. Hence, an angular momentum bounce would occur. If this occurs at the center, the central singularity will disappear. This situation seems to be inevitable for the dipole mode of odd-parity matter perturbations. In the case of a spherically symmetric system composed of counter-rotating particles, the angular-momentum bounce can actually prevent the naked singularity formation. However, for a system without spherical symmetry, it is still an open question how the non-linear asphericity works in the final stage of singularity formation with rotational matter perturbations because of the difficulty of both the analytic and numerical treatment. Further, if the initial matter perturbation is sufficiently small, it might still be possible that the radius of the spacetime curvature becomes the Planck length. In this case, the behavior of other modes of perturbations is crucial to understand the classical dynamics in the region of the Planck scale. As will be shown below, since there is no gravitational wave of the dipole mode and the quadrupole mode generated in the dust cloud does not propagate to the outside, the Cauchy horizon is marginally stable against odd-parity perturbations originating in the aspherical rotational motion of matter.
The paper is organized as follows: In §II, the basic equations are developed; in §III, the numerical results are presented; these results are discussed in §IV; in §V, we summarize our results. We adopt the geometrized units, in which $`c=G=1`$. The signature of the metric tensor and sign convention of the Riemann tensor follow Ref. .
## II Basic equations
We consider the evolution of odd-parity perturbations of the LTB spacetime up to linear order. The background spacetime and perturbation method are described in INH. In this section we briefly review them.
Using the synchronous coordinate system, the line element of the background LTB spacetime can be expressed in the form
$$d\overline{s}^2=\overline{g}_{\mu \nu }dx^\mu dx^\nu dt^2+A^2(t,r)dr^2+R^2(t,r)(d\theta ^2+\mathrm{sin}^2\theta d\varphi ^2).$$
(1)
The energy-momentum tensor for the dust fluid is
$$\overline{T}^{\mu \nu }=\overline{\rho }(t,r)\overline{u}^\mu \overline{u}^\nu ,$$
(2)
where $`\overline{\rho }(t,r)`$ is the rest mass density and $`\overline{u}^\mu `$ is the 4-velocity of the dust fluid. In the synchronous coordinate system, the unit vector field normal to the spacelike hypersurfaces is a geodesic, and there is a freedom of which timelike geodesic field is adopted as the hypersurface unit normal. Using this freedom, we can always set $`\overline{u}^\mu =\delta _0^\mu `$, since the 4-velocity of the spherically symmetric dust fluid is tangent to an irrotational timelike geodesic field.
Then the Einstein equations and the equation of motion for the dust fluid reduce to the simple equations
$`A`$ $`=`$ $`{\displaystyle \frac{_rR}{\sqrt{1+f(r)}}},`$ (3)
$`\overline{\rho }(t,r)`$ $`=`$ $`{\displaystyle \frac{1}{8\pi }}{\displaystyle \frac{1}{R^2_rR}}{\displaystyle \frac{dF(r)}{dr}},`$ (4)
$`(_tR)^2{\displaystyle \frac{F(r)}{R}}`$ $`=`$ $`f(r),`$ (5)
where $`f(r)`$ and $`F(r)`$ are arbitrary functions of the radial coordinate, $`r`$. From Eq. (4), $`F(r)`$ is related to the Misner-Sharp mass function, $`m(r)`$, of the dust cloud in the manner
$$m(r)=4\pi _0^{R(t,r)}\overline{\rho }(t,r)R^2𝑑R=4\pi _0^r\overline{\rho }(t,r)R^2_rRdr=\frac{F(r)}{2}.$$
(6)
Hence Eq. (5) might be regarded as the energy equation per unit mass. This means that the other arbitrary function, $`f(r)`$, is recognized as the specific energy of the dust fluid. The motion of the dust cloud is completely specified by the function $`F(r)`$ (or equivalently, the initial distribution of the rest mass density, $`\overline{\rho }`$) and the specific energy, $`f(r)`$. When we restrict our calculation to the case that the symmetric center, $`r=0`$, is initially regular, the central shell focusing singularity is naked if and only if $`_r^2\overline{\rho }|_{r=0}<0`$ is initially satisfied for the marginally bound collapse, $`f(r)=0`$. For a collapse that is not marginally bound, there exists a similar condition as an inequality for a value depending on the functional forms of $`F(r)`$ and $`f(r)`$.
Let us now derive the perturbation equations. The perturbed metric tensor is expressed in the form
$$g_{\mu \nu }=\overline{g}_{\mu \nu }+h_{\mu \nu },$$
(7)
where $`\overline{g}_{\mu \nu }`$ is a metric tensor of a spherically symmetric background spacetime and $`h_{\mu \nu }`$ is a perturbation. The energy-momentum tensor is written in the form
$$T_{\mu \nu }=\overline{T}_{\mu \nu }+\delta T_{\mu \nu },$$
(8)
where $`\overline{T}_{\mu \nu }`$ is a background quantity and $`\delta T_{\mu \nu }`$ is a perturbation. By virtue of the spherical symmetry of the background spacetime, $`\overline{T}_{\mu \nu }`$ can be expressed in the form
$`\overline{T}_{\mu \nu }dx^\mu dx^\nu `$ $`=`$ $`\overline{T}_{ab}dx^adx^b+{\displaystyle \frac{1}{2}}\overline{T}_B^BR^2(t,r)d\mathrm{\Omega }^2,`$ (9)
where the subscripts and superscripts $`a,b,\mathrm{}`$ represent $`t`$ and $`r`$, while $`A,B,\mathrm{}`$ represent $`\theta `$ and $`\varphi `$. The odd-parity perturbations of $`h_{\mu \nu }`$ and $`\delta T_{\mu \nu }`$ are expressed in the form
$`h_{\mu \nu }`$ $`=`$ $`\left(\begin{array}{ccc}0& 0& h_0(t,r)\mathrm{\Phi }_{lB}^m\\ & 0& h_1(t,r)\mathrm{\Phi }_{lB}^m\\ \text{sym}& & h_2(t,r)\chi _{lAB}^m\end{array}\right),`$ (13)
$`\delta T_{\mu \nu }`$ $`=`$ $`\left(\begin{array}{ccc}0& 0& t_0(t,r)\mathrm{\Phi }_{lB}^m\\ & 0& t_1(t,r)\mathrm{\Phi }_{lB}^m\\ \text{sym}& & t_2(t,r)\chi _{lAB}^m\end{array}\right),`$ (17)
where $`\mathrm{\Phi }_{lB}^m`$ and $`\chi _{lAB}^m`$ are odd-parity vector and tensor harmonics associated with the spherical symmetry of the background spacetime. We set all the arbitrary constants in the definitions of harmonics to unity. We then introduce the gauge-invariant variables defined by Gerlach and Sengupta. The metric variables are given by
$$k_a=h_a\frac{1}{2}R^2_a\left(\frac{h_2}{R^2}\right).$$
(18)
The matter variables are given by the combinations
$`L_a`$ $`=`$ $`t_a{\displaystyle \frac{1}{2}}\overline{T}_B^Bh_a,`$ (19)
$`L`$ $`=`$ $`t_2{\displaystyle \frac{1}{2}}\overline{T}_B^Bh_2.`$ (20)
In the LTB case, the odd-parity gauge-invariant matter variables become
$`L_0=\overline{\rho }(t,r)U(t,r)\mathrm{and}L_1=L=0,`$ (21)
where $`U(t,r)`$ represents the perturbation of the 4-velocity as $`\delta u_\mu =(0,0,U(t,r)\mathrm{\Phi }_{lB}^m)`$. The evolution equation for the matter variable (Eq. (3$``$19) in INH),
$$_t\left(AR^2L_0\right)=0,$$
(22)
is easily integrated, and we obtain
$$L_0=\frac{1}{AR^2}\frac{dJ(r)}{dr},$$
(23)
where $`J(r)`$ is an arbitrary function depending only on $`r`$. From Eqs. (4), (21), and (23), we obtain the relation
$`U(t,r)`$ $`=`$ $`8\pi \sqrt{1+f(r)}{\displaystyle \frac{dJ(r)/dr}{dF(r)/dr}}`$ (24)
$``$ $`U(r),`$ (25)
so $`U(t,r)`$ is independent of the time coordinate $`t`$. We introduce a gauge-invariant variable for the metric as
$$\psi _s\frac{1}{A}\left[_t\left(\frac{k_1}{R^2}\right)_r\left(\frac{k_0}{R^2}\right)\right].$$
(26)
The metric perturbation variables, $`k_0`$ and $`k_1`$, are reconstructed from the linearized Einstein equations,
$`_r\left(R^4\psi _s\right)+A\left(l1\right)\left(l+2\right)k_0`$ $`=`$ $`16\pi AR^2L_0,`$ (27)
$`_t\left(R^4\psi _s\right)+{\displaystyle \frac{1}{A}}\left(l1\right)\left(l+2\right)k_1`$ $`=`$ $`0,`$ (28)
by substituting $`\psi _s`$. It was shown in INH that $`\psi _s`$ is closely connected to the tetrad components of the magnetic part of the Weyl tensor. These components diverge where and only where $`\psi _s`$ diverges. From the linearized Einstein equations we obtain the linearized evolution equation for the odd-parity perturbation as
$`_t\left({\displaystyle \frac{A}{R^2}}_t\left(R^4\psi _s\right)\right)`$ $``$ $`_r\left({\displaystyle \frac{1}{AR^2}}_r\left(R^4\psi _s\right)\right)+\left(l1\right)\left(l+2\right)A\psi _s`$ (29)
$`=`$ $`16\pi _r\left({\displaystyle \frac{1}{AR^2}}{\displaystyle \frac{dJ}{dr}}\right).`$ (30)
The regularity conditions at the center are also considered in INH. The result for the matter perturbation $`L_0`$ is given by
$$L_0L_c(t)r^{l+1}+O(r^{l+3}).$$
(31)
Therefore the matter perturbation variables vanish at the regular center independent of the value of $`l`$. The metric variable $`\psi _s`$ behaves near the center as
$`\psi _s`$ $``$ $`\psi _{sc}(t)r^{l2}+O(r^l)\text{for }l2,`$ (32)
$`\psi _s`$ $``$ $`\psi _{sc}(t)r+O(r^3)\text{for }l=1.`$ (33)
From the above equations, it is shown that only the quadrupole mode, $`l=2`$, of $`\psi _s`$ does not vanish at the regular center.
## III Results
We numerically solve the wave equation (29) in the case of marginally bound collapse, $`f(r)=0`$, and the quadrupole mode, $`l=2`$. We follow numerical techniques employed in INH to integrate the time evolution of the perturbations. The numerical code is essentially the same as that in INH, which is tested by comparison with the analytic solution for the Minkowski background.
By virtue of the relation $`f(r)=0`$, we can easily integrate Eq. (5) and obtain
$$R(t,r)=\left(\frac{9F}{4}\right)^{1/3}[t_0(r)t]^{2/3},$$
(34)
where $`t_0(r)`$ is an arbitrary function of $`r`$. The naked singularity formation time is $`t_0=t_0(0)`$. Using the freedom for the scaling of $`r`$, we choose $`R(0,r)=r`$. This scaling of $`r`$ corresponds to the following choice of $`t_0(r)`$:
$$t_0(r)=\frac{2}{3\sqrt{F}}r^{3/2}.$$
(35)
Here note that, from Eq. (3), the background metric variable, $`A`$, is equal to $`_rR`$. Then, the wave equation (29) becomes
$`{\displaystyle \frac{^2\psi _s}{t^2}}{\displaystyle \frac{1}{(_rR)^2}}{\displaystyle \frac{^2\psi _s}{r^2}}`$ $`=`$ $`{\displaystyle \frac{1}{(_rR)^2}}\left(6{\displaystyle \frac{_rR}{R}}{\displaystyle \frac{_r^2R}{_rR}}\right){\displaystyle \frac{\psi _s}{r}}\left(6{\displaystyle \frac{_tR}{R}}+{\displaystyle \frac{_t_rR}{_rR}}\right){\displaystyle \frac{\psi _s}{t}}`$ (38)
$`4\left[\left({\displaystyle \frac{_t_rR}{_rR}}\right){\displaystyle \frac{_tR}{R}}+{\displaystyle \frac{1}{2}}\left({\displaystyle \frac{_tR}{R}}\right)^2\right]\psi _s`$
$`{\displaystyle \frac{16\pi }{(_rR)R^2}}_r\left({\displaystyle \frac{r^2\rho (r)U(r)}{(_rR)R^2}}\right),`$
where $`\rho (r)=\overline{\rho }(0,r)`$ is the density profile at $`t=0`$. We solve this partial differential equation numerically.
Before getting into the detailed explanation of the numerical techniques, we comment on the behavior of the matter perturbation variable $`L_0`$ around a naked singularity on the slice $`t=t_0`$. The regularity conditions of $`L_0`$ and $`\overline{\rho }`$ determine the behavior of $`U(r)`$ near the center as
$$U(r)r^{l+1}.$$
(39)
This property does not change even if a central singularity appears. However, the $`r`$ dependence of $`R`$ and $`A`$ near the center changes at that time. Assuming a rest mass density profile of the form
$$\rho (r)=\rho _0+\rho _nr^n+\mathrm{},$$
(40)
we obtain the relation
$$t_0(r)t_0+t_nr^n$$
(41)
from Eqs. (4) and (35), where $`n`$ is a positive even integer. After substituting this relation into Eq. (34), the lowest order term is absent from the square brackets of it. Then we obtain the behavior of $`R`$ and $`A`$ around the central singularity as
$$R(t_0,r)r^{1+\frac{2}{3}n},$$
(42)
and
$$A(t_0,r)r^{\frac{2}{3}n}$$
(43)
on the slice $`t=t_0`$. As a result, we obtain the $`r`$ dependence of $`L_0`$ around the center when the naked singularity appears as
$$L_0(t_0,r)r^{l2n+1}.$$
(44)
For example, if $`l=2`$ and $`n=2`$, then $`L_0`$ is inversely proportional to $`r`$ and diverges toward the central naked singularity. Therefore the source term of the wave equation is expected to have a large magnitude around the naked singularity. Thus the metric perturbation variable $`\psi _s`$ as well as matter variable $`L_0`$ may diverge toward the naked singularity.
Instead of the $`(t,r)`$ coordinate system, we introduce a single-null coordinate system, $`(u,r^{})`$, where $`u`$ is an outgoing null coordinate and chosen so that it agrees with $`t`$ at the symmetric center and we choose $`r^{}=r`$. We perform the numerical integration along two characteristic directions. Therefore we use a double null grid in the numerical calculation. A detailed explanation of these coordinates is given in INH. By using this new coordinate system, $`(u,r^{})`$, Eq. (38) is expressed in the form
$`{\displaystyle \frac{d\varphi _s}{du}}`$ $`=`$ $`{\displaystyle \frac{\alpha }{R}}\left[3_rR+{\displaystyle \frac{1}{2}}R(_tR)_t_rR{\displaystyle \frac{5}{4}}(_tR)^2_rR\right]\psi _s`$ (46)
$`{\displaystyle \frac{\alpha }{2}}\left[{\displaystyle \frac{_r^2R}{(_rR)^2}}{\displaystyle \frac{2}{R}}\left(1_tR\right)\right]\varphi _s{\displaystyle \frac{8\pi \alpha }{R}}_r\left({\displaystyle \frac{r^2\rho (r)U(r)}{(_rR)R^2}}\right),`$
$`_r^{}\psi _s`$ $`=`$ $`{\displaystyle \frac{1}{R}}\varphi _s3{\displaystyle \frac{_rR}{R}}\left(1+_tR\right)\psi _s,`$ (47)
where the ordinary derivative on the left-hand side of Eq. (46) and the partial derivative on the left-hand side of Eq. (47) are given by
$`{\displaystyle \frac{d}{du}}`$ $`=`$ $`_u+{\displaystyle \frac{dr^{}}{du}}_r^{}=_u{\displaystyle \frac{\alpha }{2_rR}}_r^{}`$ (48)
$`=`$ $`{\displaystyle \frac{\alpha }{2}}_t{\displaystyle \frac{\alpha }{2_rR}}_r,`$ (49)
$`_r^{}`$ $`=`$ $`{\displaystyle \frac{(_ru)_t}{(_tu)_r}}_t+_r=(_rR)_t+_r,`$ (50)
respectively. Also, $`\varphi _s`$ is defined by Eq. (47) and $`\alpha `$ is given by
$$\alpha \frac{1}{(_tu)_r}.$$
(51)
We integrate Eq. (46) using the scheme of an explicit first order difference equation. We use the trapezoidal rule,
$$\psi _{sj+1}=\psi _{sj}+\frac{\mathrm{\Delta }r^{}}{2}\left((_r^{}\psi _s)_j+(_r^{}\psi _s)_{j+1}\right),$$
(52)
to integrate Eq. (47).
For the boundary condition at the center we demand that $`\psi _s`$ behaves as $`\psi _{sc}(t)+\psi _{s2}(t)r^2`$ on a surface of $`t=\text{const}`$. We numerically realize this condition by two-step interpolation. First the values of $`\psi _s`$ are derived at two points on the surface of $`t=\text{const}`$ from the interpolation on the slices of $`u=\text{const}`$. Next, using these two values, the central value of $`\psi _s`$ is derived from the interpolation on the slice of $`t=\text{const}`$. Another way to determine the central value of $`\psi _s`$ is as follows. We first obtain the central value of $`\varphi _s`$ from Eq. (46). From Eq. (47) and the boundary conditions, the relation of $`\psi _s`$ and $`\varphi _s`$ at the center is given by
$$\varphi _s=3_r^{}R\psi _s.$$
(53)
Using this relation the central value of $`\psi _s`$ is obtained. In our numerical analyses the results of these two methods agree well.
We assume $`\psi _s`$ vanishes on the initial null hypersurface. Therefore, there exist initial ingoing waves which offset the waves produced by the source term on the initial null hypersurface. In INH, it is confirmed that this type of the initial ingoing waves propagate through the dust cloud without net amplification even when they pass through the cloud just before the appearance of the naked singularity. Therefore those parts of the metric perturbations would not diverge at the center.
We adopt the initial rest mass density profile
$$\rho (r)=\rho _0\frac{1+\mathrm{exp}\left(\frac{1}{2}\frac{r_1}{r_2}\right)}{1+\mathrm{exp}\left(\frac{r^nr_1^n}{2r_1^{n1}r_2}\right)},$$
(54)
where $`\rho _0`$, $`r_1`$ and $`r_2`$ are positive constants and $`n`$ is a positive even integer. As a result the dust fluid spreads all over the space. However, if $`rr_1,r_2`$, then $`\rho (r)`$ decreases exponentially, so that the dust cloud is divided into the core part and the envelope which would be considered as the vacuum region essentially. We define a core radius as
$$r_{\text{core}}=r_1+\frac{r_2}{2}.$$
(55)
If we set $`n=2`$, there appears a central naked singularity. This singularity becomes locally or globally naked depending on the parameters ($`\rho _0,r_1,r_2`$). However, if the integer $`n`$ is greater than $`2`$, the final state of the dust cloud is a black hole independently of the parameters. Then we consider three different density profiles connected with three types of the final state of the dust cloud, globally and locally naked singularities and a black hole. The outgoing null coordinate $`u`$ is chosen so that it agrees with the proper time at the symmetric center. Therefore, even if the black hole background is considered, we can analyze the inside of the event horizon. Corresponding parameters are given in Table I. Using this density profile, we numerically calculate the total gravitational mass of the dust cloud $`M`$. In our calculation we adopt the total mass $`M`$ as the unit of the variables.
The source term of Eq. (46),
$$S(t,r)=\frac{8\pi \alpha }{R}_r\left(\frac{r^2\rho (r)U(r)}{(_rR)R^2}\right),$$
(56)
is determined by $`U(r)`$. As mentioned above, the constraints on the functional form of $`U(r)`$ are given by the regularity condition of $`L_0`$. From Eq. (39), $`U(r)`$ should be proportional to $`r^{l+1}`$ toward the center. We localize the matter perturbation near the center to diminish the effects of the initial ingoing waves. Therefore we define $`U(r)`$ such that
$$r^2\rho (r)U(r)=\{\begin{array}{ccc}U_0\left(\frac{r}{r_b}\right)^5\left(1\left(\frac{r}{r_b}\right)^2\right)^5& \text{for}& 0rr_b,\\ 0& \text{for}& r>r_b,\end{array}$$
(57)
where $`U_0`$ and $`r_b`$ are arbitrary constants. In our numerical calculation we chose $`r_b`$ as $`r_{\text{core}}/2`$. This choice of $`r_b`$ has no special meaning, and the results of our numerical calculations are not sensitive to it.
First we observe the behavior of $`\psi _s`$ at the center. The results are plotted in Fig. 1. The initial oscillations correspond to the initial ingoing waves. After these oscillations, $`\psi _s`$ grows proportional to $`(t_0t)^\delta `$ for the naked singularity cases near the formation epoch of the naked singularity. For the case of black hole formation, $`\psi _s`$ exhibits power-law growth in the early part. Later its slope gradually changes but it grows faster than in the case of the naked singularity. For the naked singularity cases, the power-law indices $`\delta `$ are determined by $`(t_0t)\dot{\psi _s}/\psi _s`$ locally. The results are shown in Fig. 2. From this figure we read the final indices as 5/3 for both naked cases. Therefore the metric perturbations diverge at the central naked singularity.
We also observe the wave form of $`\psi _s`$ along the line of a constant circumferential radius outside the dust cloud. The results are shown in Figs. 35. Figure 3 displays the wave form of the globally naked case (a), Fig. 4 displays the wave form of the locally naked case (b), and Fig. 5 displays the wave form of the black hole case (c). The initial oscillations correspond to the initial ingoing waves. In the case of a locally naked singularity and black hole formation, damped oscillations dominate the gravitational waves. We read the frequencies and damping rates of these damped oscillations from Figs. 4 and 5 and give them in terms of complex frequencies as $`0.37+0.089i`$ for locally naked and black hole cases. These agree well with the fundamental quasi-normal frequency of the quadrupole mode $`(2M\omega =0.74734+0.17792i)`$ of a Schwarzschild black hole given by Chandrasekhar and Detweiler. In the globally naked singularity case (a), we did not see this damped oscillation because of the existence of the Cauchy horizon. In all cases the gravitational waves generated by matter perturbations are at most quasi-normal modes of a black hole, which is generated outside the dust cloud. Therefore intense odd-parity gravitational waves would not be produced by the inhomogeneous dust cloud collapse. We should not expect that the central extremely high density region can be observed by this mode of gravitational waves.
We can calculate the radiated power of the gravitational waves and thereby grasp the physical meaning of the gauge-invariant quantities. To relate the perturbation of the metric to the radiated gravitational power, it is useful to specialize to the radiation gauge, in which the tetrad components $`h_{(\theta )(\theta )}h_{(\varphi )(\varphi )}`$ and $`h_{(\theta )(\varphi )}`$ fall off as $`O(1/R)`$, and all other tetrad components fall off as $`O(1/R^2)`$ or faster. Note that, in vacuum at large distance, the spherically symmetric background metric is given by the Schwarzschild solution, where hereafter we adopt the Schwarzschild coordinates,
$$ds^2=\left(1\frac{2M}{R}\right)d\tau ^2+\left(1\frac{2M}{R}\right)^1dR^2+R^2\left(d\theta ^2+\mathrm{sin}^2\theta d\varphi ^2\right).$$
(58)
The relation between the line elements Eq. (1) and Eq. (58) is given by the transfer matrix
$`d\tau `$ $`=`$ $`{\displaystyle \frac{1}{1(_tR)^2}}(dt+_rR_tRdr),`$ (59)
$`dR`$ $`=`$ $`_tRdt+_rRdr.`$ (60)
In this gauge, the metric perturbations in Eq. (13) behave as
$`h_0,h_1`$ $`=`$ $`O\left({\displaystyle \frac{1}{R}}\right),`$ (61)
$`h_2`$ $`=`$ $`w(\tau R_{})+O(1),`$ (62)
where
$$R_{}=R+2M\mathrm{ln}\left(\frac{R}{2M}1\right)+\text{const}.$$
(63)
Then, the gauge-invariant metric perturbations (18) are calculated as
$`k_0`$ $`=`$ $`{\displaystyle \frac{1}{2}}w^{(1)}R+O(1),`$ (64)
$`k_1`$ $`=`$ $`{\displaystyle \frac{1}{2}}w^{(1)}R+O(1),`$ (65)
where $`w^{(1)}`$ denotes the first derivative of $`w`$ with respect to its argument.
In this radiation gauge, the radiated power $`P`$ per unit solid angle is given by the formula which was derived by Landau and Lifshitz from their stress-energy pseudo-tensor:
$$\frac{dP}{d\mathrm{\Omega }}=\frac{R^2}{16\pi }\left[\left(\frac{h_{(\theta )(\varphi )}}{\tau }\right)^2+\frac{1}{4}\left(\frac{h_{(\theta )(\theta )}}{\tau }\frac{h_{(\varphi )(\varphi )}}{\tau }\right)^2\right].$$
(66)
For the axisymmetric mode, i.e., $`m=0`$, the above formula is reduced to
$$\frac{dP}{d\mathrm{\Omega }}=\frac{1}{64\pi }(w^{(1)})^2A_l(\theta ),$$
(67)
where
$$A_l(\theta )\frac{2l+1}{4\pi }\mathrm{sin}^4\theta \left(\frac{d^2P_l(\mathrm{cos}\theta )}{(d\mathrm{cos}\theta )^2}\right)^2.$$
(68)
Then, by using the gauge-invariant quantities and integrating over the all solid angles, the formula for the power of gravitational radiation is obtained in the following form:
$`{\displaystyle \frac{dP}{d\mathrm{\Omega }}}`$ $`=`$ $`{\displaystyle \frac{1}{16\pi }}{\displaystyle \frac{k_0^2}{R^2}}A_l(\theta )={\displaystyle \frac{1}{16\pi }}{\displaystyle \frac{k_1^2}{R^2}}A_l(\theta ),`$ (69)
$`P`$ $`=`$ $`{\displaystyle \frac{1}{16\pi }}B_l{\displaystyle \frac{k_0^2}{R^2}}={\displaystyle \frac{1}{16\pi }}B_l{\displaystyle \frac{k_1^2}{R^2}},`$ (70)
where
$$B_l\frac{(l+2)!}{(l2)!}.$$
(71)
Using Eq. (28), the radiated power $`P`$ of the quadrupole mode is given by
$$P=\frac{3}{32\pi }R^2\left[_\tau \left(R^3\psi _s\right)\right]^2.$$
(72)
Figure 6 displays the time evolution of the radiated power $`P`$. The radiated power also has a finite value at the Cauchy horizon. The total energy radiated by odd-parity quadrupole gravitational waves during the dust collapse should not diverge.
## IV Discussion
First we consider the behavior of the source term $`S(t,r)`$ around the naked singularity. From the regularity conditions and Eqs. (39), (42), and (43), the asymptotic behavior of the source term is obtained as
$$S(t,r)r^{l1}$$
(73)
for $`t<t_0`$ and
$$S(t,r)r^{l\frac{8}{3}n1},$$
(74)
at $`t=t_0`$. For example, in the case $`l=2`$ and $`n=2`$, the souce term behaves on $`t=t_0`$ as
$$S(t,r)r^{13/3},$$
(75)
and then it diverges at the center. Thus the divergency of $`\psi _s`$ at the center originates from the source term. To confirm this, we numerically integrate the source term along the ingoing null lines with respect to $`u`$ and estimate the central value of $`\varphi _s`$. We define this ‘estimated’ value as
$$\mathrm{\Phi }_sS(t,r)𝑑u.$$
(76)
Using Eq. (53) we can define the estimated value of $`\psi _s`$ as
$$\mathrm{\Psi }_s\frac{\mathrm{\Phi }_s}{3_rR}.$$
(77)
We plot it in Fig. 7 together with the corresponding $`\psi _s`$. The estimated value has the same power-law index of $`\psi _s`$. We conclude that the behavior of $`\psi _s`$ is determined by the source term in the dust cloud.
We next consider the stability of the Cauchy horizon. We found that the metric perturbation produced by the source term does not propagate outside the dust cloud, except for quasi-normal ringing. The source term, which controls $`\psi _s`$, does not diverge at the Cauchy horizon. Therefore $`\psi _s`$ should not diverge at the Cauchy horizon and should not destroy it. Then, even if odd-parity perturbations are considered, it will not be the case that the LTB spacetime loses its character as a counterexample to CCH due to Cauchy horizon instability. Also, it does not seem that such collapse is a strong source of gravitational waves.
In this paper, we have dealt with the marginally bound case. For the case of non-marginally bound collapse, the condition of the appearance of the central naked singularity is slightly different from that in the above case and hence there is the possibility that the behavior of $`\psi _s`$ in this case is different from that in the marginally bound case. However, it is well known that the limiting behavior of the metric as $`tt_0(r)`$ is common for all the cases:
$$R\left(\frac{9F}{4}\right)^{1/3}\left(t_0t\right)^{2/3},A\left(\frac{2F}{3}\right)^{1/3}\frac{t_0^{}}{\sqrt{1+f}}\left(t_0t\right)^{1/3}.$$
(78)
We conjecture that the results of the perturbed analysis for the non-marginal collapse would be similar to the results for the marginal bound.
## V Summary
We have studied the behavior of the odd-parity perturbation in the LTB spacetime including the matter perturbation. For the quadrupole mode, where gravitational waves exist, we have numerically investigated the wave equation. For the case of naked singularity formation, the gauge-invariant metric variable, $`\psi _s`$, diverges according to a power law with power index 5/3 at the center. This power index is closely related to the behavior of the matter perturbation around the center. We have also observed $`\psi _s`$ at a constant circumferential radius. For the globally naked case, we cannot see intense gravitational waves propagated from the center just before the crossing of the Cauchy horizon. For the locally naked case, we have confirmed that there exist quasi-normal oscillations. As a result, we conclude that the type of singularity changes due to the odd-parity perturbation because $`\psi _s`$ diverges at the center. However, the Cauchy horizon is marginally stable against odd-parity perturbations.
At the final stage of the collapse, the effects of the rotational motion are important and the centrifugal force might dominate the radial motion. If this is true, the central singularity would disappear when an odd-parity matter perturbation is introduced. For the dipole mode, such a situation seems to be inevitable. However, we should note that it is a non-trivial and open question how non-linear asphericity affects the final fate of the singularity-formation process. Further, in the case of initially sufficiently small aspherical perturbations, the radius of spacetime curvature at the center might reach the Planck length, and hence there is still the possibility that the naked singularity is formed there in a practical sense. However, as our present analysis has revealed, since the Cauchy horizon is stable with respect to odd-parity linear perturbations, there is little possibility that this collapse is a strong source of odd-parity gravitational waves.
There remains important related works to be completed. The first problem is to consider the even-parity mode in which the metric and matter perturbations are essentially coupled with each other. We are now investigating this problem. Finally, we should consider the non-linear effects to complete this analysis. This problem will be analyzed in the future.
## Acknowledgements
We would like to thank T. Nakamura for helpful and useful discussions and N. Sugiyama for careful reading of the manuscript. We are also grateful to H. Sato and colleagues in the theoretical astrophysics group at Kyoto University for useful comments and encouragement. This work was partly supported by Grants-in-Aid for Scientific Research (No. 9204) and for Creative Basic Research (No. 09NP0801) from the Japanese Ministry of Education, Science, Sports and Culture.
|
no-problem/9904/gr-qc9904004.html
|
ar5iv
|
text
|
# Conformal classes of Brans-Dicke gravity
## Abstract
A classification of Brans-Dicke theories of gravitation, based on the behaviour of the dimensionless gravitational coupling constant, is given. It is noted that the discussion takes place in the current literature, about which of the two distinguished conformal frames in which scalar-tensor theories of gravity can be formulated: the Jordan frame and the Einstein frame, is the physical one, may, in most cases, be meaningless for both frames may belong to the same conformal class. It is also noted that the Jordan frame formulation of Brans-Dicke gravity with ordinary matter nonminimally coupled, that is shown to be just the Jordan frame formulation of general relativity, is scale-invariant, unlike the situation with the Jordan frame representation of Brans-Dicke gravity with matter minimally coupled (the original formulation of Brans-Dicke theory), where the presence of nonzero mass ordinary matter breaks the scale-invariance of the theory.
Among all conformal mappings relating different scalar-tensor gravity (STG) theories, two frames are distinguished : the Jordan frame (JF) and the Einstein frame (EF).
Some arguments have been raised against the physical equivalence of both formulations of STG theories (see Ref. and references therein), so it is of prime interest to know which frame is the physical one. This issue is being intensively discussed in the literature (see the review ).
The most undesirable feature of EF STG theories is that, in this frame, the ordinary matter is nonminimally coupled to the scalar field so, in particular, test particles don’t follow the geodesics of the geometry. For his part, the strongest objection against JF STG theories is that this formulation of STG leads to a negative definite, or indefinite kinetic energy for the scalar field, this implies that the theory does not have a stable ground state .
However, as it has been carefully noted in , the terms with the second covariant derivatives of the scalar field on the right hand side of the JF field equation, contain the connection, and hence a part of the dynamical description of gravity. Santiago and Silbergleit introduced a new connection that removes the gravitational dynamical terms from the right of the Einstein field equation, leading to the correct energy-momentum tensor for the scalar field. The scalar field energy density can now be made nonnegative by imposing some constraints to the coupling function and to the potential.
Yet another source of uncertainty respecting this subject was introduced by Magnano and Sokolowski when they studied the possibility of adding matter nonminimally coupled in the JF, such as to have the matter minimally coupled in the EF.
Our letter is aimed precisely at the study of this possibility for Brans-Dicke (BD) gravity (the prototype of STG theory) and, in connection with it, the classification of BD theories attending the different kind of behaviour of the dimensionless gravitational coupling constant $`Gm^2(\mathrm{}=c=1)`$ that can arise.
We shall begin with the JF formulation of BD gravity due to Brans and Dicke , that is based upon the Lagrangian density:
$$L^I[g,\varphi ]=\frac{\sqrt{g}}{16\pi }(\varphi R\frac{\omega }{\varphi }g^{nm}_n\varphi _m\varphi )+L_{matter}[g]$$
$`(1)`$
where $`R`$ is the curvature scalar, $`\varphi `$ is the scalar BD field, $`\omega `$ is the BD coupling constant, and $`L_{matter}[g]`$ is the Lagrangian density of the ordinary matter minimally coupled to the scalar BD field.
The Lagrangian density conformally dual to (1) gives rise to the EF formulation of BD gravity :
$$L^I[\widehat{g},\widehat{\varphi }]=\frac{\sqrt{\widehat{g}}}{16\pi }(\widehat{R}(\omega +\frac{3}{2})\widehat{g}^{nm}\widehat{}_n\widehat{\varphi }\widehat{}_m\widehat{\varphi })+\widehat{L}_{matter}[\widehat{g},\widehat{\varphi }]$$
$`(2)`$
where $`\widehat{R}`$ is the curvature scalar in the EF metric $`\widehat{𝐠}`$, conformally dual to $`𝐠`$:
$$\widehat{g}_{ab}=\varphi g_{ab}$$
$`(3)`$
The scalar function $`\widehat{\varphi }\mathrm{ln}\varphi `$ is the EF BD scalar field, and $`\widehat{L}_{matter}[\widehat{g},\widehat{\varphi }]`$ is the Lagrangian density for the ordinary matter nonminimally coupled to the scalar field.
Magnano and Sokolowski studied the possibility of changing the coupling in (1); $`L_{matter}[g]L_{matter}[g,\varphi ]`$, while keeping intact the gravitational part:
$$L^{II}[g,\varphi ]=\frac{\sqrt{g}}{16\pi }(\varphi R\frac{\omega }{\varphi }g^{nm}_n\varphi _m\varphi )+L_{matter}[g,\varphi ]$$
$`(4)`$
In this case, unlike the original JF BD theory as formulated by Brans and Dicke , the ordinary matter is nonminimally coupled to the scalar field in the JF.
The EF of the theory based on (4) is given by the Lagrangian density conformally dual to it:
$$L^{II}[\widehat{g},\widehat{\varphi }]=\frac{\sqrt{\widehat{g}}}{16\pi }(\widehat{R}(\omega +\frac{3}{2})\widehat{g}^{nm}\widehat{}_n\widehat{\varphi }\widehat{}_m\widehat{\varphi })+\widehat{L}_{matter}[\widehat{g}]$$
$`(5)`$
In this frame, test particles follow the geodesics of the geometry, so the inertial mass $`m`$ of a given material particle is constant over the spacetime manifold $`M`$. At the same time, the scalar field $`\widehat{\varphi }`$ is minimally coupled to the curvature (it is a distinctive feature of the EF), so the dimensional gravitational constant $`G`$ is a real constant, this means that the dimensionless gravitational coupling constant $`Gm^2(\mathrm{}=c=1)`$ is constant on $`M`$ as well, i.e.; the strong equivalence principle (SEP) holds in the EF of the theory. Besides, since $`Gm^2`$ is a dimensionless constant, it is not affected by the conformal rescaling of the metric (eq.(3))<sup>*</sup><sup>*</sup>*In fact it is a conformal transformation of the units of measurement ., so it is a constant in the JF of the theory as well.
This way, the distinctive feature of the theory given by $`L^{II}[g,\varphi ]`$ (and its conformally dual given by $`L^{II}[\widehat{g},\widehat{\varphi }]`$) is that the dimensionless gravitational coupling constant $`Gm^2`$ is a real constant over $`M`$. The class given by the pair {$`L^{II}[g,\varphi ],L^{II}[\widehat{g},\widehat{\varphi }]`$} we shall call Class II BD gravity, while the class given by the pair {$`L^I[g,\varphi ],L^I[\widehat{g},\widehat{\varphi }]`$} we shall call Class I BD gravity. The distinctive feature of this last class is that the dimensionless gravitational coupling constant changes from point to point in spacetime: $`Gm^2\varphi ^1`$ . Such as in this case both the weak equivalence principle (WEP) and SEP hold, we can realize that the Class II BD theory is just general relativity (GR), so (4) leads to the JF GR.
The different formulations inside of a class are dual to each other in the sense that experiment can not differentiate between them. Actually, both the JF and the EF of the theory are connected by the conformal transformation of units (3) , and then the observables of the theory, that are always dimensionless (as assumed in and more recently remarked in the introductory part of references ), are not affected by this transformation that affects only dimensional quantities. In this sense the comparison between the members inside of a class is a non well-posed comparison A well-posed comparison is one that can be resolved by physical experimentation. so, the discussion that takes place in the current literature about which frame is the physical one, may, in most cases, be meaningless. Only theories that belong to different classes can be consistently compared.
Another important aspect of the BD gravity we shall refer in this letter is connected with the conformal symmetry. As it has been shown in Ref. , the graviational part of the JF BD Lagrangian density $`L^I[g,\varphi ]`$ (eq.(1) without ordinary matter) is invariant in form under the conformal rescaling of the spacetime metric:
$$g_{ab}\stackrel{~}{g}_{ab}=\varphi ^{2\alpha }g_{ab}$$
$`(6)`$
the field redefinition:
$$\stackrel{~}{\varphi }=\varphi ^{12\alpha }$$
$`(7)`$
and the coupling constant redefinition:
$$\stackrel{~}{\omega }=\frac{\omega 6\alpha (\alpha 1)}{(12\alpha )^2}$$
$`(8)`$
with $`\alpha \frac{1}{2}`$. However, when one considers the full Lagrangian density (eq.(1)) the conformal invariance is broken, unless the matter stress-energy tensor is symmetric ($`T_{ab}=T_{ba}`$) and it has a vanishing trace: $`TT_n^n=0`$. In this last case the conservation equation: $`^nT_{na}=0`$ (which contains the dynamics of matter) is conformally invariant.
For his part, when one considers the JF of Class II BD gravity; i.e., the JF GR given by (4), the full theory is invariant in respect to the transformations (6-8), so the conformal symmetry is preserved even in the presence of ordinary matter. In this case, the dynamical equation of matter: $`^nT_{na}=\frac{1}{2}\varphi ^1_a\varphi T`$ is invariant in form respect to (6-8). In particular the free-motion equation for an uncharged, spinless mass point given by:
$$\frac{d^2x^a}{ds^2}=\mathrm{\Gamma }_{nm}^a\frac{dx^n}{ds}\frac{dx^m}{ds}\frac{1}{2}\varphi ^1_n\varphi (\frac{dx^n}{ds}\frac{dx^a}{ds}g^{an})$$
$`(9)`$
is preserved under (6-8). This amounts saying that, unlike the JF of Class I BD gravity, the presence of ordinary matter in the JF of Class II BD gravity does not break the scale-invariance of the theory. Full derivation of the conformal invariance of (4) will be given elsewhere.
Summing up: BD gravity theories can be grouped into two classes attending the behaviour of the dimensionless gravitational coupling constant in these theories. JF and EF formulations in each class are experimentally indistinguishable; only theories in different classes can be consistently compared. When comparing the JF formulation of Class I BD gravity (just BD theory) with the JF representation of Class II BD theory (JF GR), one finds that in the second case the presence of ordinary matter does not break the conformal invariance of the theory.
We thank David I. Santiago for pointing out a crucial error about a consideration on WEP violation in the original version of this letter, and MES of Cuba for financing.
|
no-problem/9904/hep-ph9904280.html
|
ar5iv
|
text
|
# 1 *
## Acknowledgments
C. M. thanks the Theory Group at SLAC for their kind hospitality and Prof. C. Pajares of the University of Santiago de Compostela, the Director of the research project which partially financed this work. We would also like to thank Markus Diehl for conversations.
|
no-problem/9904/hep-ph9904485.html
|
ar5iv
|
text
|
# 𝐷-branes, String Cosmology and Large Extra Dimensions
\[
## Abstract
$`D`$-branes are fundamental in all scenarios where there are large extra dimensions and the string scale is much smaller than the four-dimensional Planck mass. We show that this current picture leads to a new approach to string cosmology where inflation on our brane is driven by the large extra dimensions and the issue of the graceful exit becomes inextricably linked to the problem of the stabilization of the extra dimensions, suggesting the possibility of a common solution. We also show that branes may violently fluctuate along their transverse directions in curved spacetime, possibly leading to a period of brane-driven inflation. This phenomenon plays also a crucial role in many other cosmological issues, such as the smoothing out of the cosmological singularities and the generation of the baryon asymmetry on our three brane.
\]
$`D`$-branes are classical solutions which are known to exist in many string theories . A $`Dp`$-brane denotes a configuration which extends along $`p`$ spatial directions and is localized in all other spatial transverse directions. From the string point of view, a $`D`$-brane is a soliton on which open string endpoints may live and whose mass (or tension) is proportional to the inverse of the string coupling $`g_s`$. This means that $`D`$-branes become light in those regions of the moduli space where the string coupling is large. This simple property has led to a new interpretation for singularities of various fixed spacetime backgrounds in which strings propagate and provided a generic and physically intuitive mechanism for smoothing away cosmological singularities in the strong coupling/large curvature regime of the evolution of the early Universe .
The presence of branes in string theory has also given rise to the possibility that the Standard Model (SM) gauge fields live on branes rather than in the bulk of spacetime. This scenario corresponds to a novel compactification in which gravity lives in the bulk of the ten dimensional spacetime, but the other observed fields are confined to a brane of lower dimension . In particular, one might imagine the situation where the SM particles live on a three brane and the directions transverse to the three brane being compact. There might be also other branes separated from ours in the extra directions. Since the usual gauge and matter fields live on the brane, the only long-range interaction which sees the extra dimensions is gravity.
In the weakly coupled heterotic string there is a fundamental relation between the string scale $`M_s`$, the four-dimensional Planck mass $`M_p`$, the value of the dilaton field $`\varphi `$ and the gauge coupling constant $`\alpha _g`$, $`(M_s/M_p)^2\alpha _ge^\varphi `$ and the string scale is fixed to be about $`10^{18}`$ GeV. However, in other regions of the moduli space the string scale can be much smaller and it is tempting to imagine that the string scale is not far from the TeV region. If this is the case, the small ratio $`M_s/M_p`$ could be a consequence of large compactified dimensions through a scaling relation like
$$M_p^2=M_s^{n+2}V_n,$$
(1)
where $`V_n`$ is the volume of the $`n`$ large extra dimensions. For $`M_s`$ TeV and $`n2`$, the radius of the large extra dimensions may be as large as the millimeter. Large extra dimensions are only allowed if the SM fields are confined to live on a membrane orthogonal to the large extra dimensions and this hypothesis fits rather nicely with the current picture of string theory and $`D`$-branes.
Not surprisingly, one expects many interesting cosmological implications of this scenario (some of them have already been discussed in ) and the possibility of large extra dimensions is potentially exciting for several reasons.
In the presence of some matter with energy density $`\rho `$, the evolution of the Universe is controlled by the equation $`H^2(\dot{a}/a)^2=(8\pi \rho /3M_p^2)`$, where $`a`$ is the scale factor and $`H`$ is the Hubble rate. This implies that for an expanding Universe, $`\rho `$ and $`H^2`$ are proportional, they increase or decrease together in time. This is true only if $`M_p`$ is constant. Now, in the weakly coupled string theory where all compactified dimensions have size $`M_s^1`$, $`M_p`$ is controlled by the dilaton and it is perfectly possible to have growing $`H`$ while $`\rho `$ decreases, provided the dilaton $`\varphi `$ is also growing. On this peculiar feature is based the whole idea of Pre Big-Bang (PBB) cosmology where there exist superinflationary solutions for the scale factor $`a`$ having a growing – rather than a constant – Hubble parameter and a growing dilaton field . The suggestion from weakly coupled string theory is therefore rather compelling, since it is possible to generate a scenario containing a dilaton-driven PPB inflation.
These simple considerations together with the relation (1) holding in the current picture of string theory suggest a simple way of getting inflation on our three brane and a fresh approach to string cosmology: since $`M_p`$ is now governed by the size $`b`$ of the extra dimensions, it is possible to obtain superinflationary solutions provided that the extra dimensions change in time, that is a radion-driven PPB cosmology. The Einstein action in $`4+n`$ dimensions and metric $`g_{MN}=(1,a^2\delta _{ij},b^2\delta _{mn})`$ reads
$$Sd^4xa^3b^n\left[R2n\frac{\ddot{b}}{b}6n\frac{\dot{a}}{a}\frac{\dot{b}}{b}n(n1)\frac{\dot{b}^2}{b^2}\right],$$
(2)
where $`R`$ is the usual four-dimensional Ricci scalar. Setting $`b^n=e^{c\sigma }`$, with $`c=\left(n/n1\right)^{1/2}`$, and integrating by parts the term $`\ddot{b}`$ in (2), the action for the radion field $`\sigma `$ becomes
$$Sd^4xa^3e^{c\sigma }\left[R+_\mu \sigma ^\mu \sigma \right].$$
(3)
Were $`c=1`$, string cosmology experts would recognize the action of the dilaton field coupled to gravity (with $`\varphi =\sigma `$). For us $`c=1`$ in the limit of $`n\mathrm{}`$ and in this regime we can recover the more familiar results of dilaton-driven PPB cosmology. The evolution of the radion is governed by the equation obtained by varying the lapse function
$$\dot{\sigma }=ϵ_\pm H,ϵ_\pm =3c\pm \sqrt{9c^26},$$
(4)
while the equation for the Hubble parameter becomes
$$\dot{H}=\alpha _\pm H^2,\alpha _\pm =3+cϵ_\pm .$$
(5)
For each value of $`n`$, $`\alpha _{}`$ is negative and it generates a superinflationary solution, that is $`H`$ grows until a singularity is reached, say at $`t=0`$. In terms of the scale factors $`a`$ and $`b`$, the solutions are given by $`a(t)=a_0\left(t/t_0\right)^{\gamma _1}`$ and $`b(t)=b_0\left(t/t_0\right)^{\gamma _2}`$ where $`t`$ and $`t_0`$ are negative and $`\gamma _1=|1/\alpha _{}|`$, $`\gamma _2=|cϵ_{}/(n\alpha _{})|`$. Since $`ϵ_{}<0`$, the volume of the internal dimensions is contracting. Suppose that $`a`$ and $`b`$ start with comparable values at low curvature at $`tt_0`$, say $`abM_s^1`$; the subsequent evolution is such that the spatial dimensions of our three brane superinflate, whereas the extra dimensions evolve to values smaller than $`M_s^1`$. In string theory, due to the presence of winding states in the closed string (gravitational) sector, the low-energy physics is equivalent as if there was a radius $`b_I=(bM_s^2)^1`$ which is much larger than the string length and the roles of winding and Kaluza-Klein momenta are interchanged (the evolution in terms of $`b_I`$ is also characterized by a nontrivial dynamics for the dilaton field). On the other hand, in this T-dual description, the open string states – which give rise to the ordinary non gravitational matter on the three brane – have only heavy winding modes. As a result, in the dual theory the extra dimensions appear to be large. The dynamics of the radion field $`b`$ coupled to the four-dimensional gravity leads automatically to an inflationary stage on our three brane and explains the huge hierarchy between the sizes of our observed Universe and of the extra dimensions if $`|\gamma _1||\gamma _2|`$. The cosmological solution stops being valid when the Hubble rate becomes of the order of $`M_s`$. However, differently from the dilaton-driven PPB cosmology where one has to face a regime of strong coupling when approaching the big-bang singularity at $`t0`$, in the radion-driven PPB scenario the singularity may take place well inside the pertubative regime and be understood as the process of decompactifying the extra dimensions. The issue of the graceful exit in string cosmology becomes therefore inextricably linked to the problem of the stabilization of the extra dimensions, suggesting the possibility of a common solution.
There is – however – another feature of $`D`$-branes that may help in solving the radion-driven PPB singularity problem as well as give new insight into other fundamental cosmological issues: this is the peculiar behaviour of fluctuations transverse to the brane in curved spacetime. The action governing the dynamics of a generic $`Dp`$-brane (neglecting the gauge fields) is $`S=T_pd^{p+1}\xi e^\varphi \sqrt{\mathrm{det}G_{\alpha \beta }}`$, where $`T_p^1`$ is the brane tension, $`\xi ^\alpha `$ ($`\alpha =0,\mathrm{},p`$) parametrize the $`D`$-brane world-volume and $`G_{\alpha \beta }=g_{\mu \nu }(X)_\alpha X^\mu _\beta X^\nu `$ is the induced metric. Given a brane configuration $`\overline{X}^\mu (\xi )`$ in a spacetime of $`D`$-dimensions, arbitrary fluctuations transverse to this configuration (the would-be Goldstone bosons of the broken translational invariance along the directions perpendiclar to the brane) are parametrized by $`(Dp)`$ world-volume scalars $`y^a(\xi )`$ as $`X^\mu (\xi )=\overline{X}^\mu (\xi )+\overline{n}_a^\mu (\xi )y^a(\xi )`$, where $`\overline{n}_a^\mu (\xi )`$ ($`a=1,\mathrm{},(Dp)`$) are the $`(Dp)`$ vectors normal to the brane at the point $`\overline{X}(\xi )`$. The expansion of the metric in Riemann coordinates is
$$g_{\mu \nu }(X)=\overline{g}_{\mu \nu }\frac{1}{3}y^ay^b\overline{R}_{\mu \rho \nu \sigma }\overline{n}_a^\rho \overline{n}_b^\sigma +𝒪(y^3),$$
(6)
where the overbar indicates that the quantity is evaluated at $`\overline{X}`$. If the extrinsic curvature of the membrane is small, the normal vectors are $`\overline{n}_a^\rho \delta _a^\rho `$, and therefore $`g_{\mu \nu }(X)=\overline{g}_{\mu \nu }\frac{1}{3}y^ay^b\overline{R}_{\mu a\nu b}+𝒪(y^3)`$ .
Let us begin, for simplicity, by considering a single $`Dp`$-brane in an expanding accelerating ($`\ddot{a}>0`$) background where the dimensions orthogonal to the brane are static. We shall consider for definiteness a background where the common scale factor $`a`$ of the spatial coordinates $`\stackrel{}{\xi }`$ is inflating, $`a(\eta )(\eta )^q`$, where $`0<q1`$ and $`d\eta =(dt/a)`$ is the conformal time, $`\mathrm{}<\eta <0`$. For $`q1`$ the Hubble rate $`H`$ grows with time till a singularity is reached and the De Sitter background with constant $`H`$ is recovered for $`q=1`$. The equation of motion for small transverse oscillations $`y_a(\xi )`$ ($`y_a`$ is the comoving displacement) is
$$\ddot{y_a}+(p1)(\dot{a}/a)\dot{y_a}^2y_a=0,$$
(7)
where $`\dot{y_a}y_a/\eta `$. The canonical conjugate momentum is $`\pi _a=\tau _pa^{p1}\dot{y}_a`$, where $`\tau =T_p/g_s`$; imposing the canonical commutation relation $`[y_a(\eta ,\stackrel{}{\xi }),\pi _a(\eta ,\stackrel{}{\xi }^{})]=i\delta ^{(p)}(\stackrel{}{\xi }\stackrel{}{\xi }^{})`$, we can expand the solution as
$$y^a(\eta ,\stackrel{}{\xi })=\frac{1}{\tau ^{\frac{1}{p+1}}}\frac{d^pk}{(2\pi )^{p/2}}\left[a_kh_k^a(\eta )e^{i\stackrel{}{k}\stackrel{}{\xi }}+a_k^{}h_k^a(\eta )e^{+i\stackrel{}{k}\stackrel{}{\xi }}\right],$$
(8)
where $`a_k`$ and $`a_k^{}`$ are the annihilation and creation operators, respectively. The solution is very simple, $`h_k^a(\eta )=(\eta )^\beta \left[c_1H_\beta ^1(k\eta )+c_2H_\beta ^1(k\eta )\right]`$, where $`\beta =(q(p1)+1)/2`$. If we quantize the modes and define the vacuum state by $`a_k|0=0`$, then different choices of vacuum correspond to different choices of $`c_1`$ and $`c_2`$. The adiabatic vacuum corresponds to $`c_2=0`$ and this (Heisenberg) state is the state we will assume the brane is in.
The behaviour of a mode $`h_k^a(\eta )`$ is very simple. Suppose from now on to consider the case $`p=3`$, i.e. we consider a three brane. As long as the physical wavelength $`ak^1`$ of a mode is inside the Hubble radius, $`h_k^a(\eta )`$ oscillates with constant physical amplitude $`ah_k^a(\eta )`$. As the physical wavelength grows it crosses the Hubble radius and the comoving amplitude $`h_k^a(\eta )`$ becomes frozen, so that the physical amplitude grows like $`a`$.
In this state one can calculate the mean square transverse displacement
$`𝐲^2`$ $``$ $`{\displaystyle \frac{d^3\xi }{V_3}0|𝐲^2(\eta ,\stackrel{}{\xi })|0}`$ (9)
$`=`$ $`{\displaystyle \frac{D4}{\tau ^{1/2}}}{\displaystyle \frac{d^3k}{(2\pi )^3}\left|h_k^a\right|^2},`$ (10)
where $`𝐲^2_ay_a^2`$ and $`V_3=d^3\xi `$. The physical displacement $`a𝐲`$ has the usual ultraviolet divergence as in flat space: physical wavelengths with $`kaH`$ are always well inside the horizon at a given time and their amplitude is unaffected by the expansion. Their contribution leads to the usual $`d^3kk^1`$ divergence. We can subtract this divergence. Modes with $`ka_iH_i`$, where $`a_i`$ and $`H_i`$ are the values of the scalar factor and Hubble rate at the beginning of the inflationary stage, are always well outside the horizon and simply match on adiabatically to the modes before and after inflation. One therefore finds at $`\eta \eta _f0`$ (for $`q1`$)
$`a^2𝐲^2`$ $``$ $`a^2{\displaystyle \frac{D4}{\tau ^{1/2}}}{\displaystyle _{a_iH_i}^{a_fH_f}}H{\displaystyle \frac{d^3k}{(2\pi )^3}}\left|h_k^a\right|^2`$ (11)
$``$ $`{\displaystyle \frac{D4}{M_s^4\eta _f^2}},`$ (12)
where we have set $`\tau M_s^4`$. One can picture this result by saying that the modes with wavelength of the order of the Hubble radius have a physical width $`𝐲_p^2M_s^2`$ which gets amplified by the expansion after they cross the Hubble radius. Transverse fluctuations grow in time. In the special case of De Sitter with constant rate of expansion $`H`$, one finds that $`𝐲^2M_s^4H^2\mathrm{ln}(EH)M_s^4H^3t`$, where $`E`$ is the total number of e-foldings.
More interestingly, one can calculate the energy acquired by each mode in this process. The energy is given by $`EM_s^4d^3\xi a^3\left[1+1/2\dot{𝐲}^2/a^2+1/2\left(𝐲\right)^2/a^2\right]`$, where the first term is just the classical stretching. Just as in (11), we find
$$\frac{\left(𝐲\right)^2}{a^2}\left(\frac{H}{M_s}\right)^4.$$
(13)
The term $`\dot{𝐲}^2`$ gives no contribution after the flat spacetime substraction. Thus we deduce that the fractional energy in the transverse perturbations grows with time as $`H`$ increases with time and their contribution to the energy density becomes sizeable when $`HM_s`$. Physical transverse fluctuations of the three brane – even though initially oscillating – will eventually grow and give rise to an instability. Therefore, as the three brane is stretched by the expansion of the Universe, transverse fluctuations grow, the thickness of the brane gets larger and larger and the brane cannot be considered as a static object any longer. This growth is accompanied by a huge production of light pseudo-Goldstone bosons of the broken translation invariance. Let us try to feed back the effect of this production into the expansion of the Universe. As the system evolves towards the singularity, more and more energy in transferred into transverse fluctuations of the brane till – eventually – they dominate the energy density of the Universe. Since $`\left(𝐲\right)^2\dot{𝐲}^2`$, the gas of pseudo-Goldstone bosons is characterized by an energy density $`\rho _y3p_y`$, where $`p_y`$ is the pressure of the gas. We thus obtain an effective negative pressure. Assuming a flat Universe, the equation for the acceleration $`\ddot{a}`$ alone becomes
$$\ddot{a}/a(\rho _y+3p_y)=0$$
(14)
and we discover that the back-reaction of the gas of pseudo-Goldstone bosons onto the evolution of the Universe is to halt the period of acceleration and – therefore – to smooth the singularity away when $`H`$ becomes of the order of the string scale! It should be stressed that this result is only valid when the the stress tensor is of the perfect fluid type, i.e. when it becomes possible to neglect the viscosity terms due to mutual and self-interactions. These terms are not negligible when the approximation of small transverse fluctuations breaks down.
The role played by the transverse fluctuations may be even more dramatic, though. Remember that in the considerations made so far we have supposed that the directions transverse to our three brane were static. Assume now that these directions are suffering a period of decelerating contraction, while the spatial dimensions of the brane are accelerating. This is – for instance – what happens during the stage of radion-driven inflation previously described. From Eq. (6) the fluctuations $`y_a`$ transverse to the brane get a tachyonic mass $`m_y^2=1/3\left(\ddot{b}/b+3\dot{b}/b\dot{a}/a\right)`$, where now the dots stand for derivatives with respect to the cosmic time $`t`$. The situation is very similar to the thermal tachyon which occurs at the Hagedorn temperature in string theory . In that case a condensate of winding states forms when the periodic imaginary times becomes too small. In the present case the tachyon forms because of the nontrivial geometry of the spacetime. As a given physical wavelength $`ak^1`$ of a transverse fluctuation $`y_a`$ crosses the horizon, the comoving amplitude $`h_k^a`$ does not freeze out; instead the transverse physical amplitude grows faster than the scale factor $`a`$ and therefore faster than the physical coordinates defining the brane. In fact the brane tends to loose its own identity as the thickness of the brane grow faster than the physical coordinates defining the three brane itself; the brane violently fluctuate along the transverse directions. This phenomenon holds for branes of any dimensionality. If one goes beyond the approximation of small transverse fluctuations and computes the higher order terms in the tachyon effective potential, one can show that – along with these huge fluctuations – branes become almost tensionless . This is hardly surprising since the brane can fluctuate along the transverse directions paying no price in energy. When the tension becomes negative, branes become unstable and – at least in the regime of strong coupling – it becomes easier and easier for the gravitational background to produce further massless branes. On the other hand, when the transverse sizes of the branes become larger than the horizon, the branes presumably break down and decay, but branes are continuously created. The rate of brane creation due to these quantum effects may be so fast to balance the dilution of the brane density due to the expansion. One might be therefore led into a phase of constant brane density and exponentially expanding Universe, phase of brane-driven inflation reminescent of the old idea of string-driven inflation .
The huge transverse fluctuations of branes in curved spacetime may be also relevant to the issue of baryon asymmetry production on our three brane if we accept the idea that there might be other branes separated from ours in the extra directions. Even though on our brane global charges, such as the baryon and the lepton number, are conserved with a great accuracy, this is not necessarily the case on other branes. This breaking might be communicated to our brane by messenger fields that leave in the bulk . However, this communication is suppressed by the enormous distance separating the branes in the bulk and this is the reason why today the baryon number is conserved in SM interactions. Nevertheless, at early epochs in the evolution of the Universe the thickness of a brane is highly fluctuating. This makes it possible for two branes separated in the bulk to overlap. One can also envisage various other dynamical phenomena like the melting of two or more branes . If our brane was considerably overlapping with another brane where the breaking on the baryon number is $`𝒪(1)`$, a significant amount of baryon number might have been deposited on our three brane, thus explaining the observed baryon asymmery of our Universe .
The considerations reported here are very preliminary and certainly leave many questions unanswered. For instance, what are the initial conditions for the Universe (or perhaps just for our region of the Universe) right before the onset of inflationary PBB stage driven by the radion field? What is the subsequent dynamics of the gas of pseudo-Goldstone bosons which is generated at $`HM_s`$ and whose presence may help avoiding the singularity? How does the system get into the familiar radiation-dominated stage? At this point open strings attached to the branes have to play a role since the expansion time $`H^1`$ is of the order of the typical scale of the stringy network $`M_s^1`$. It seems safe therefore to assume that this gas of strings attached to the branes and pseudo-Goldstone bosons reaches thermal equilibrium. The fact that fundamental strings have a limiting temperature $`M_s`$ may be turn out to be crucial since it implies that – if radiation is created and is in thermal contact with the strings –it cannot attain a density higher than $`M_s^4`$, and the Universe necessarily is dominated by strings.
We are grateful to T. Banks, R. Brustein, E. Copeland, K. Dienes, M. Dine, S. Dimopoulos, T. Gherghetta, G.F. Giudice, D. Lyth, R. Rattazzi, G. Veneziano and especially to M. Maggiore for many useful discussions.
|
no-problem/9904/cond-mat9904345.html
|
ar5iv
|
text
|
# Polaronic Excitons in 𝐙𝐧_𝐱𝐂𝐝_{𝟏-𝐱}𝐒𝐞/𝐙𝐧𝐒𝐞 Quantum Wells
## Abstract
We present a detailed investigation of excitonic absorption in $`Zn_{0.69}Cd_{0.31}Se/ZnSe`$ quantum wells under the application of a perpendicular magnetic field. The large energy separation between heavy- and light-hole excitons allows us to clearly resolve and identify magneto-excitonic absorption resonant with the continuum edge of the 1S heavy-hole exciton. Experimental values of the exciton binding energy are compared with results of a theoretical model that includes the exciton-phonon interaction. The remarkable agreemeent found unambiguously indicates the predominant polaronic character of excitons in $`ZnSe`$-based heterostructures.
In recent years wide-gap II-VI semiconductor based heterostructures have attracted much attention mostly in light of their potential for the development of opto-electronic devices operating in the blue-green spectral region. Owing to their relatively large exciton binding energies ($`E_b`$), they also offer the possibility of realizing quantum wells (QWs) in which excitonic recombination is dominant even at room temperature. These systems also present other unique properties such as a strong exciton-phonon interaction which makes them ideal candidates for the investigation of these kind of interactions (from here on referred to *polaronic effects*) on the optical properties.
Polaronic effects have been extensively studied in the past, and significant polaron-related modifications of fundamental optical properties were demonstrated in bulk ionic semiconductors. Much less attention, on the contrary, was paid to this issue in systems with reduced dimensionality, e.g. QWs. In fact in the numerous experimental analyses on excitonic effects in II-VI QWs , the role of polarons in determining the exciton binding energy was not adequately addressed, and was actually often overlooked. This was probably due to the large body of studies focused on GaAs-based heterostructures where polaronic effects are indeed negligible owing to the low-ionicity of the atomic bonds.
In this communication we address this issue by analyzing the magneto-absorption properties of $`Zn_xCd_{1x}Se/ZnSe`$ QWs. Our results clearly demonstrate the important role played by polaronic effects in determining excitonic optical properties. We shall show that by *including polaronic effects* using an effective potential first derived by *Aldhrich and Bajaj* within an envelope-function formalism it is possible to reproduce quantitatively our experimental results with no adjustable material parameters.
Samples studied were grown by solid source molecular-beam epitaxy on $`GaAs(001)`$ substrates. They consist of a 0.5-$`\mu `$m $`GaAs`$ buffer layer grown at 580 <sup>o</sup>C followed by 1.5-$`\mu `$m-thick $`ZnSe`$ layer grown at 290<sup>o</sup> C. Ten $`Zn_xCd_{1x}Se/ZnSe`$ QW’s were then grown at 250<sup>o</sup> C with a 30-s interruption at each interface. A 500-nm-thick cap layer concluded the growth. $`ZnSe`$ barrier width is 30 nm, 2-, 3-, 4- and 5-nm-thick QWs were examined. In order to detect transmission signals, circular regions of about $`6\times 10^4cm^2`$ were selectively removed using standard photolithographic and wet-etching techniques. Further details about these structures can be found in Ref. . The separation between the energy positions of 1S heavy- and light-hole excitons caused by the compressive strain in the QWs, allowed us to resolve the absorption associated to the continuum edges, even at zero magnetic field.
It should be noted that despite the large Cd content the 1S heavy-hole exciton absorption peak at T=1.6K displays a full width at half maximum of about 9 meV indicating good sample quality. This was also confirmed by the small Stokes shift of the photoluminescence signal ($``$1 meV). Samples were mounted on a variable-temperature insert and magnetic fields $`B`$ between 0 and 10T were applied parallel to the growth direction. Magneto-absorption was studied using a 100W xenon lamp as the source. Light was focused and collected along the growth direction by optical fibers. Transmitted intensity was normalized to the incident light in order to detect absorption changes.
Figure 1 shows typical spectra at T=1.6K for the sample with QW thickness $`L_w=5`$ nm without (dashed line) and with an applied magnetic field (10 T, solid line). The two peaks at 2528 meV and 2610 meV observed in the B=0 spectrum are associated to the 1S (hh1-e1) heavy- and (lh1-e1) light-hole exciton, respectively. The absorption of what is supposed to be the heavy-hole continuum edge is observed at around 2560 meV. Thanks to the high confinement potential, an additional higher-energy peak due to the 1S (hh2-e2) heavy-hole exciton is also observed.
Upon application of the magnetic field, the 1S heavy- and light-hole exciton peaks display a small blue-shift and no appreciable oscillator strength enhancement. This behavior is immediately understood in light of their large binding energy in comparison to the cyclotron energy ($`E_c=`$5.2 meV at 10 T). The light-hole exciton, however, displays a peculiar non-monotonic behavior which may be linked to valence-band mixing effects . The heavy-hole continuum edge, on the contrary, is dramatically altered, and a sharp peak on its low-energy side gradually emerges with increasing magnetic fields (see the right panel of Fig. 1 where absorption spectra in the region of the continuum edge are displayed at three different magnetic fields). This feature can be unambiguously associated to the heavy-hole 2S excitonic level because of its diamagnetic behavior in the range $`6TB10`$ T (see Fig. 2). The diamagnetic shift of that level ($`\delta E4meV`$) indicates that the Coulomb interaction is still significant in the investigated range of magnetic fields. A further peak, visible from 8.5 T is identified with the 3S excitonic level. In this magnetic field range, however, this exciton state acquires a Landau-like magnetic-field dependence that, contrary to the 2S exciton case, indicates the predominance of the free-carrier cyclotron energy $`E_c`$ over the exciton binding energy. This is in qualitatively agreement with theory developed in Ref.
Figure 2 shows the peak energy positions of heavy- and light-hole excitonic states (solid circles) as a function of B for the sample with $`L_w=5`$nm (a) and $`L_w=4`$ nm (b). ¿From the expected values of 1S and 2S position to B=0 we can determine the energy difference bitween the binding energis of 1S and 2S states, adding the calculated values of $`E_b^{2S}`$ we can determine the binding energy of the 1S state of the $`hh`$ exciton.
Our theoretical analysis based on variational calculation of 1S exciton binding energy versus quantum well width demonstrates the importance of polaronic effects in $`Zn_xCd_{1x}Se/ZnSe`$ QW’s. A complete account of the polaronic effects leads to an effective Hamiltonian for the exciton in which not only electron and hole band masses are replaced by the corresponding electron and hole polaron effective masses but also the effective potential differs from the Coulomb potential screened by the static dielectric constant $`ϵ_0`$ .
The Hamiltonian of the exciton (apart from the electron and hole self-energy terms) can be expressed as a sum of three terms:
$$H=H_1+H_2+H_{ex}$$
(1)
where $`H_1`$ and $`H_2`$ describe the motion of electron polaron and hole polaron along the growth axis (z):
$$H_i=\frac{\mathrm{}^2}{2m_i^{}}\frac{^2}{z^2}+V_i(z_i)$$
(2)
where $`V_1(z_1)`$ and $`V_2(z_2)`$ are the confining potentials for the electron polaron and the hole polaron respectively, and $`m_1^{}`$ and $`m_2^{}`$ are the electron and hole polaron masses along the z axis.
The term $`H_{ex}`$ describes the internal motion of the exciton in the x-y plane:
$$H_{ex}=\frac{\mathrm{}^2}{2\mu ^{}}\left(\frac{1}{\rho }\frac{}{\rho }\left(\rho \frac{}{\rho }\right)+\frac{1}{\rho ^2}\frac{^2}{\varphi ^2}\right)+V_{int}(r)$$
(3)
where $`\mu ^{}`$ is the reduced electron and hole polaron mass in the x-y plane, $`\rho `$ and $`\varphi `$ are polar coordinates in the x-y plane and $`r^2=(z_1z_2)^2+\rho ^2`$. Hole masses are expressed in terms of the Luttinger parameter in these formulas:
For the effective interaction potential $`V_{int}`$, we use the following two different forms. First we consider the Coulomb potential screened by the static dielectric constant i.e.
$$V_C=\frac{e^2}{ϵ_0r}$$
(4)
and second, we consider the potential $`V_{AB}`$ derived by Aldrich and Bajaj that takes into account the polaronic effects.
$`V_{AB}={\displaystyle \frac{e^2}{ϵ_0r}}{\displaystyle \frac{e^2}{2ϵ^{}r}}(exp(\beta _1r)+exp(\beta _ir))`$ (5)
$`+{\displaystyle \underset{i=1}{\overset{2}{}}}{\displaystyle \frac{e\beta _i}{2ϵ^{}}}{\displaystyle \frac{exp(\beta _ir)}{1+\alpha _i/12+\alpha _i/(4+\alpha _i/3)}}`$ (6)
where:
$$\beta _i=(\frac{2m_i\omega }{\mathrm{}})^{1/2}$$
(7)
$$\alpha _i=\frac{e^2\beta _i}{2ϵ^{}\mathrm{}\omega }$$
(8)
$$\frac{1}{ϵ^{}}=\frac{1}{ϵ_{\mathrm{}}}\frac{1}{ϵ_0}$$
(9)
$$m_\alpha =m_0\frac{1\alpha /12}{1+\alpha /12}$$
(10)
where $`\omega `$ is the longitudinal optical phonon frequency, $`ϵ_{\mathrm{}}`$ is high frequency dielectric constant, and $`m_i`$ is electron (hole) band mass.
In order to calculate the binding energy of 1S state of an exciton as a function of well width we follow a variational approach and use the following form of the trial wave function:
$$\mathrm{\Psi }(z_1,z_2,\rho ,\varphi )=F_1(z_1)F_2(z_2)exp(\lambda r\nu \rho ^2\eta (z_1z_2)^2)$$
(11)
where $`F_1(z_1)`$ and $`F_2(z_2)`$ are ground state eigenfunctions of the electron polaron and hole polaron Hamiltonians $`H_1`$ and $`H_2`$ respectively.
The binding energy $`E_B`$ of the 1S state of the exciton is then defined as:
$$E_B=E_1+E_2min_{\lambda ,\nu ,\eta }(\frac{<\mathrm{\Psi }|H|\mathrm{\Psi }>}{<\mathrm{\Psi }|\mathrm{\Psi }>})$$
(12)
where $`E_1`$ and $`E_2`$ are ground state eigenvalues of the electron polaron and hole polaron Hamiltonians $`H_1`$ and $`H_2`$ respectively.
In our calculation we have used the following values of various physical parameters for the $`Zn_xCd_{1x}Se/ZnSe`$ QW’s. Electron band offset $`\mathrm{\Delta }E_e=0.230`$ eV, hole band offset $`\mathrm{\Delta }E_h=0.115`$ eV, electron polaron mass for the barrier material $`m_1^b=0.155m_0`$, electron polaron mass for the well material $`m_1^w=0.14m_0`$, static dielectric constant $`ϵ_0=8.7`$, high frequency dielectric constant $`ϵ_{\mathrm{}}=5.73`$, LO phonon energy $`E_{LO}=0.0317eV`$, Luttinger parameters $`\gamma _1=2.45`$ and $`\gamma _2=0.61`$.
The results of the calculation of binding energy versus well width for both $`V_C`$ and $`V_{AB}`$ are shown in Figure 3. The use of simple screened Coulomb potential leads to essential underestimation of the exciton binding energy in $`Zn_xCd_{1x}Se/ZnSe`$ QW’s. On the other hand, the values of binding energy obtaind with $`V_{AB}`$ are in a good agreement with the experimentally measured binding energies for the samples with $`L_w=5`$ nm and $`L_w=4`$ nm.
In conclusion, we have measured … Using variational approach, we have calculated the exciton binding energies for $`Zn_xCd_{1x}Se/ZnSe`$ QW’s vs. quantum well width using both Coulomb potential screened by static dielectric constant $`ϵ_0`$ and the potential $`V_{AB}`$ derived by Aldrich and Bajaj, that takes account of polaronic effects. The comparision of the calculated values of the exciton binding energy to the experimantally obtained binding energies for 4 nm and 5 nm quantum wells demonstrates that the potential $`V_{AB}`$ should be used in the calculations of binding energy in order to achieve a reasonable agreement with the experimantal data and that the polaronic effects essentially influence physical properties of excitons in $`ZnSe`$-based heterostructures. Detailed investigation of the diamagnetic shift will be presented elsewhere.
|
no-problem/9904/cond-mat9904365.html
|
ar5iv
|
text
|
# Transmission resonances on metallic gratings with very narrow slits
\[
## Abstract
In this letter we show how transmission metallic gratings with very narrow and deep enough slits can exhibit transmission resonances for wavelengths larger than the period of the grating. By using a transfer matrix formalism and a quasi-analytical model based on a modal expansion, we show that there are two possible ways of transferring light from the upper surface to the lower one: by the excitation of coupled surface plasmon polaritons on both surfaces of the metallic grating or by the coupling of incident plane waves with waveguide resonances located in the slits. Both mechanisms can lead to almost perfect transmittance for those particular resonances.
\]
Absorption anomalies in metallic gratings have attracted much attention since their discovery by Wood in 1902. One of these anomalies is only observed for p-polarized light (H parallel to the grating grooves) and appears as a minimum on the specular reflectance. Now it is well known that this anomaly stems from the excitation of surface plasmon polaritons (SPPs) by the incident electromagnetic radiation. The dependence of these SPP modes on the grating shape and the possible existence of modes localized in the grooves have been studied for a number of years and are still of interest . On the other hand, the activity of the last decade in the field of photonic crystals has originated a renewed interest in the properties of SPPs, as they can be viewed as surface electromagnetic modes propagating in 1D periodic dielectric media .
Besides, experimental evidence of the excitation of optical waveguide modes inside the narrow grooves of zero-order reflection gratings has been recently given . Both from fundamental and practical points of view it would be very interesting to analyze the transmission properties of these waveguide modes. Also very recently some experiments carried out in arrays of submicrometre cylindrical holes in metallic films have shown an extraordinary optical transmission at wavelengths up to ten times larger than the diameter of the holes . The similarities between this last structure and metal gratings suggest the possibility of equivalent resonant effects in transmission metal gratings of very narrow slits.
In this letter we test theoretically this possibility by analyzing the response of transmission metallic gratings to $`p`$-polarized electromagnetic radiation. We will show how, for very narrow slits, the coupling of the incident light with surface electromagnetic modes of the grating can lead to almost perfect transmission resonances appearing at wavelengths larger than the period of the grating and hence much larger than the lateral dimensions of the slits.
Inset of Fig. 1 shows a schematic view of the structures under study with the definition of the different parameters: the period of the grating ($`d`$), the width ($`a`$) and height ($`h`$) of the slits. The substrate is characterized by a dielectric constant, $`ϵ`$. Advances in material technology have allowed the production of transmission gratings with well controlled profiles, which have already been used in different interesting applications, such as polarizers or x-ray spectrometers . From the theoretical point of view, there have been some studies of these structures in the last few years . However, up to our knowledge, transmission properties of very narrow slits that are periodically structured remain unstudied.
In this letter we consider metal gratings made of gold and we use fixed values for the grating period ($`d=3.5\mu `$m) and the width of the slits ($`a=0.5\mu `$m), although the dependence of our results on $`a`$ is also addressed. The thickness of the metallic grating ($`h`$) will be varied between $`0`$ and $`4\mu `$m. The choice of these geometrical values is motivated by the experimental findings of waveguide resonances reported in Ref. for reflection metal gratings with the same set of parameters. Nevertheless, it should be pointed out that the effects discussed in this letter do appear for any other range provided $`a`$ is very small in comparison to $`d`$ and the frequency of the incident light is well below the plasma frequency of the metal. The dielectric function of gold is described using the tables reported in Ref. .
We have analyzed the electromagnetic properties of these gratings by means of a transfer matrix formalism . Within this formalism it is possible to calculate transmission and reflection coefficients for an incoming plane wave. Subsequently, the transmittance and reflectance of the grating as well as real-space electromagnetic fields can be calculated. Fig. 1 shows zero-order transmittance for normal incident radiation on metallic gratings in vacuum as a function of the wavelength of the incoming plane wave. The grating height ($`h`$) is varied in these calculations from $`0.2`$ to $`4\mu `$m. As can be seen in Fig. 1a, for deep enough gratings ($`h0.6\mu `$m) a remarkable transmission peak appears for a wavelength slightly larger than the grating period (in this case $`3.5\mu `$m). This transmission peak moves to larger wavelengths as the grating height increases whereas its linewidth is broadened. And, as illustrated in Fig. 1b, subsequent transmission peaks emerge for deeper gratings. The behavior of the transmittance spectrum as a function of the metal thickness seems to suggest that these peaks could be linked to the coupling of incident plane waves with waveguide resonances of the slits.
In order to analyze the physical origin of these transmission resonances, we have also developed an approximated modal method. We incorporate two main simplifications to the exact modal method reported in . First, as the frequency regime we are interested in is below the plasma frequency of the metal, surface-impedance boundary conditions (SIBC) are imposed on the metallic boundaries, except on the vertical walls of the slits which are treated as perfect metal surfaces. Second, we only consider the fundamental eigenmode in the modal expansion of the electric and magnetic fields inside the slits, which is justified in the limit where the wavelength of light is much larger than the width of the slits. The validity of these two approximations is confirmed by the good agreement between the results calculated by this simplified modal method and the numerical simulations performed with the transfer-matrix formalism. Within this single-mode approximation the two field amplitudes inside the slits (the one associated with the $`e^{ik_0z}`$ wave and the other with the $`e^{ik_0z}`$ one) are proportional to $`1/D`$, where the denominator $`D`$ is given by:
$`D=(1(1+\eta )\varphi )(1(1+\eta )\psi )e^{ık_0h}`$ (1)
$`(1+(1\eta )\varphi )(1+(1\eta )\psi )e^{ık_0h},`$ (2)
with $`k_0=2\pi /\lambda `$, $`\eta =ϵ_{metal}^{1/2}`$, and $`\psi `$ given by the sum:
$$\psi =\frac{a}{d}ϵ\underset{m=\mathrm{}}{\overset{\mathrm{}}{}}\frac{(sinc(\frac{k_0\gamma _ma}{2}))^2}{(ϵ\gamma _m^2)^{1/2}+ϵ\eta },$$
(3)
where $`ϵ`$ is the dielectric constant of the substrate, $`sinc(\xi )\mathrm{sin}(\xi )/\xi `$, and $`\gamma _m=\mathrm{sin}\theta +m\frac{\lambda }{d}`$ is associated with the $`m`$-th diffraction order. The quantity $`\varphi `$ is also given by Eq. (2) but with $`ϵ=1`$.
The zero-order transmittance spectrum of the grating is completely governed by the behavior of the denominator $`D`$. For example, for normal incidence and gratings in vacuum, zero-order transmittance simplifies to:
$$T_0=\frac{16(\frac{a}{d})^2}{D^21+\eta ^4}.$$
(4)
Moreover, we have found that there is a close correspondence between maxima of zero-order transmittance and spectral positions of the zeros of the imaginary part of $`D`$, $`\mathrm{}(D)`$. This result allows us to analyze the nature of the electromagnetic modes responsible for the transmission resonances shown in Fig.1 just by studying the zeros of $`\mathrm{}(D)`$ as given by Eq. (1). Also, by varying the angle of incidence $`\theta `$ we can calculate the photonic band structure, $`\omega (k_x)`$, of these surface excitations. In Fig. 2a we show the photonic band structure for the case $`h=0.6\mu `$m (black dots) and, for comparison, the energetic positions of the SPP excitation in the limit $`h0`$ (gray dots). Note that due to the range of photon energies we are analyzing, these SPP frequencies almost coincide with the energetic positions of the Rayleigh anomalies that are linked to zeros of transmittance . As can be seen in the inset of Fig. 2a, a very narrow band gap between the first and second bands appears in the spectrum. The lower branch at $`k_x=0`$ is associated with the transmission peak at $`\lambda `$ close to $`d`$ in Fig. 1a. Its close proximity to the energy of SPP bands suggests that this transmissive mode is associated with the excitation of a surface plasmon with SPP character in each surface of the grating. From now on, we name this kind of resonances as coupled SPPs. As $`h`$ is increased, new bands that are associated with waveguide modes of the slits appear in the spectrum. This can be seen in Fig. 2b which shows that for $`h=3\mu `$m a flat waveguide band is present at $`\omega =0.17`$ eV. This localized mode is responsible for the transmission peak located at $`\lambda =7.5\mu `$m (see Fig. 1b). The other transmission resonance obtained for $`h=3\mu `$m at $`\lambda 5\mu `$m corresponds to the lower branch of the first band gap. Differently from the $`h=0.6\mu `$m case, a strong hybridization between SPP bands and a waveguide mode of similar energy results in the opening of a very broad gap. Then, the transmissive mode appearing at $`\lambda 5\mu `$m will present a hybrid character between coupled SPPs and a waveguide resonance. Therefore, by looking at the photonic band structure of surface plasmons, we can conclude that transmission resonances appearing in Fig. 1 are mainly due to the excitation of two kinds of electromagnetic modes: coupled SPPs for $`\lambda d`$ and waveguide resonances for $`\lambda d`$. Fig. 2 can also give us additional information about the dependence of the transmission resonances on the angle of incidence, $`\theta `$. A totally different dispersion relation for the two kinds of transmissive modes is clearly seen in Fig. 2. Hence, transmittance associated with coupled SPPs will show a strong dependence on $`\theta `$ whereas for waveguide resonances transmission is almost independent of $`\theta `$.
Using the simplified modal method, we can also study in detail the behavior of transmission resonances as a function of the width of the slits, $`a`$. For this purpose, we show in Fig. 3 zero-order transmittance curves for gratings in vacuum of thickness (a) $`h=0.6\mu `$m and (b) $`h=3.0\mu `$m in the wavelength region where transmission peaks appear. The width of the slits is varied between $`0`$ and $`1.5\mu `$m. For the case of transmission resonances linked to coupled SPPs (Fig. 3a) a minimum value of $`a`$ is needed in order to couple SPPs of each surface of the grating. Above this threshold (whose wavelength depends on the depth of the slits and for $`h=0.6\mu `$m is around $`0.2\mu `$m), the resonance is extremely narrow and hence these structures could be used as filters of electromagnetic radiation for wavelengths close to the period of the grating. On the other hand, for waveguide resonances (Fig. 3b), even for extremely narrow slits the transmission peak could be close to 1. In this limit, the wavelength of the resonance tends to $`2h`$ (that corresponds to the first zero of $`\mathrm{sin}k_0h`$) and its linewidth goes to zero. As shown in Fig. 3, transmission resonances associated with coupled SPPs are much narrower than the ones linked to waveguide modes and in both cases their linewidths are rapidly broadened as the width of the slits is increased.
Finally, two questions remain to be answered: how is light transmitted from one side of the metallic grating to the other one by these electromagnetic modes? and what is the difference in the transmission process between the two mechanisms mentioned above? In order to answer these questions, we show in Fig. 4 detailed pictures of the E-field for two cases, both with $`a=0.5\mu `$m: (a) $`h=0.6\mu `$m and $`\lambda =3.6\mu `$m (that corresponds to coupled SPPs) and (b) $`h=3.0\mu `$m and $`\lambda =7.5\mu `$m (example of waveguide resonance). As can be seen in Fig. 4a, the normal incident plane wave is exciting first a SPP in the upper metal surface. Although metal thickness is much larger than the skin depth of the metal, this SPP couples with the corresponding SPP mode of the lower metal surface through a waveguide mode located in the slits. Then the SPP mode of the lower surface can match to an outgoing propagating plane wave of the same frequency and momentum as the incident one, leading to a large transmittance. Due to the nature of this process, these transmission resonances are very sensitive to the presence of a substrate in the lower surface: when energies of the two SPPs involved do not coincide, the coupling between them is less effective and transmittance is severely reduced. The transmission process associated with waveguide resonances is completely different to the one described above for coupled SPPs. As shown in Fig. 4b, for these electromagnetic modes only the metal walls of the slits play an active role in the process. Incident light induces current densities flowing parallel to the slits’ walls, having different signs on the two opposite surfaces of the slits. Therefore, and different from coupled SPPs, the transmittance associated with these waveguide resonances is not very sensitive to the refraction index of the substrate.
We believe that electromagnetic modes of a nature similar to coupled SPPs in transmission gratings are responsible for the extraordinary optical transmission reported in hole arrays . There are several facts that support this belief. First, the positions of the transmission peaks in both structures (hole arrays and transmission gratings) are mainly determined by the periodicity of the system and are almost independent of the diameter of the holes or the slits width and of the particular metal used. Besides, transmission resonances in both structures disperse significantly with the angle of incidence. However, hole arrays and gratings are two different geometries and the correspondence between both systems must be established with certain caveats .
In conclusion, transmission properties of metallic gratings with very narrow slits have been analyzed by means of a transfer matrix formalism and a quasi-analytical approach based on a modal expansion. We have shown how for deep enough gratings, resonances in the zero-order transmission spectra appear for wavelengths larger than the period of the grating. For these resonances, zero-order transmittance could be close to 1 besides the fact that the wavelength of the transmitted light is much larger than the lateral dimension of the slits. Two different transmission mechanisms have been described: excitation of SPPs on both surfaces of the metal grating and coupling of the incident light with waveguide resonances of the slits.
We are indebted to J. Sánchez-Dehesa and T. López-Ríos for many helpful discussions and L. Martín-Moreno and J.J. Greffet for a critical reading of the manuscript. We also acknowledge partial financial support from the Acciones Integradas Program under contract HB-1997-0032. J.A.P. acknowledges a postdoctoral grant from the Ministerio de Educación y Cultura of Spain.
|
no-problem/9904/hep-th9904130.html
|
ar5iv
|
text
|
# Space-time Symmetry Transformations of Elementary Particles realized in Optics Laboratories
## I Introduction
In our earlier papers , we have formulated the Jones vectors and Stokes parameters in terms of the two-by-two and four-by-four matrix representations of the six-parameter Lorentz group . It was seen there that, to every two-by-two transformation matrix for the Jones vector, there is a corresponding four-by-four matrix for the Stokes parameters. It was found also that the Stokes parameters are like the components of Minkowskian four-vectors, and two-component Jones vectors are like two-component spinors in the relativistic world. This enhances our capacity to approach polarization optics in terms of the kinematics of special relativity.
Indeed, we can now design specific experiments which will test some of the consequences derivable from the principles of special relativity. The most widely known example is the Wigner rotation. This has been extensively discussed in the literature in connection with the Thomas effect , Berry’s phase , and squeezed states of light .
In our earlier papers, we discussed an optical filter which will exhibit the matrix form of
$$\left(\begin{array}{cc}1& u\\ 0& 1\end{array}\right)$$
(1)
applicable to two transverse components of the light wave, where u is a controllable parameter. When applied to a two-component system, this matrix performs a superposition in the upper channel while leaving the low channel invariant. The question is whether it is possible to produce optical filters with this property.
In Ref. , we approached this problem in terms of the generators of the Lorentz group. It is very difficult, if not impossible, to manufacture optical devices performing the function of group generators. In the case of optical filters, this means an infinite number of layers of zero thickness. In the present paper, we deal with the same problem from the experimental point of view. We will present a specific design for optical filters performing this function. We will of course present our case in terms of a combination of three filters of finite thickness.
In order to achieve this goal, we use the fact that polarization optics and special relativity shares the same mathematics. This aspect was already noted in the literature for the case of the Wigner rotation . The concept of the Wigner rotation comes from the kinematics of special relativity, in which two successive non-collinear Lorentz boosts do not end up with a boost. The result is a boost followed or preceded by a rotation. Thus we can achieve a rotation from three non-collinear boosts starting from a particle at rest. Since each boost corresponds to an attenuation filter, it requires three attenuation filters to achieve a Wigner rotation in polarization optics.
While the Wigner rotation is based on Lorentz transformations of massive particles, there are similar transformations for massless particles. Here, two non-collinear Lorentz boosts do not result in one boost. They become one boost preceded or followed by a transformation which corresponds to a gauge transformation. In two-by-two formalism, the transformation takes the form of Eq.(1). We shall show in this paper that the filter possessing the property of Eq.(1) can be constructed from one rotation filter and one attenuation filter. In mathematics, this type of decomposition is called the Iwasawa decomposition .
While the primary purpose of this paper is to discuss filters and their combinations in polarization optics, we provide also concrete illustrative examples of Wigner’s little group . The little group is the maximal subgroup of the Lorentz group whose transformations leave the four-momentum of a given particle invariant, and has a long history . The Wigner rotation and the Iwasawa decomposition are transformations of the little groups for massive and massless particles respectively. It is interesting to note that these transformations can be also achieved in optics laboratories.
In Sec. II, we review the formalism for optical filters based on the Lorentz group and explain why filters are like Lorentz transformations. It is shown in Sec. III, that a rotation can be achieved by three non-collinear Lorentz boosts. In Sec. IV, we spell out in detail how the Iwasawa decomposition can be achieved from the combination of two optical filters.
## II Formulation of the Problem
In studying polarized light propagating along the $`z`$ direction, the traditional approach is to consider the $`x`$ and $`y`$ components of the electric fields. Their amplitude ratio and the phase difference determine the degree of polarization. Thus, we can change the polarization either by adjusting the amplitudes, by changing the relative phases, or both. For convenience, we call the optical device which changes amplitudes an “attenuator” and the device which changes the relative phase a “phase shifter.”
Let us write these electric fields as
$$\left(\begin{array}{c}E_x\\ E_y\end{array}\right)=\left(\begin{array}{c}A\mathrm{exp}\left\{i(kz\omega t+\varphi _1)\right\}\\ B\mathrm{exp}\left\{i(kz\omega t+\varphi _2)\right\}\end{array}\right).$$
(2)
where $`A`$ and $`B`$ are the amplitudes which are real and positive numbers, and $`\varphi _1`$ and $`\varphi _2`$ are the phases of the $`x`$ and $`y`$ components respectively. This column matrix is called the Jones vector. In dealing with light waves, we have to realize that the intensity is the quantity we measure. Then there arises the question of coherence and time average. We are thus led to consider the following parameters.
$`S_{11}`$ $`=`$ $`<E_x^{}E_x>,S_{22}=<E_y^{}E_y>,`$ (3)
$`S_{12}`$ $`=`$ $`<E_x^{}E_y>,S_{21}=<E_y^{}E_x>.`$ (4)
Then, we are naturally invited to write down the two-by-two matrix:
$$C=\left(\begin{array}{cc}<E_x^{}E_x>& <E_y^{}E_x>\\ <E_x^{}E_y>& <E_y^{}E_y>\end{array}\right),$$
(5)
where $`<E_i^{}E_j>`$ is the time average of $`E_i^{}E_j`$. The above form is called the coherency matrix .
It is sometimes more convenient to use the following combinations of parameters.
$`S_0=S_{11}+S_{22},`$ (6)
$`S_1=S_{11}S_{22},`$ (7)
$`S_2=S_{12}+S_{21},`$ (8)
$`S_3=i\left(S_{12}S_{21}\right).`$ (9)
These four parameters are called the Stokes parameters in the literature .
We have shown in our earlier papers that the Jones vectors and the Stokes parameters can be formulated in terms of the two-by-two spinor and four-by-four vector representations of the Lorentz group. This group theoretical formalism allows to discuss three different sets of physical quantities using one mathematical device. In our earlier publications, we used the concept of Lie groups extensively and used their generators based on infinitesimal generators.
In this paper, we avoid the Lie groups and work only with explicit transformation matrices. For this purpose, we start with the following two matrices.
$`B=\left(\begin{array}{cccc}\mathrm{cosh}\chi & \mathrm{sinh}\chi & 0& 0\\ \mathrm{sinh}\chi & \mathrm{cosh}\chi & 0& 0\\ 0& 0& 1& 0\\ 0& 0& 0& 1\end{array}\right),`$ (10)
$`R=\left(\begin{array}{cccc}1& 0& 0& 0\\ 0& \mathrm{cos}\varphi & \mathrm{sin}\varphi & 0\\ 0& \mathrm{sin}\varphi & \mathrm{cos}\varphi & 0\\ 0& 0& 0& 1\end{array}\right).`$ (11)
If the above matrices are applied to the Minkowskian space of $`(ct,z,x,y)`$, the matrix $`B`$ performs a Lorentz boost:
$`t^{}=(\mathrm{cosh}\chi )t+(\mathrm{sinh}\chi )z,`$ (12)
$`z^{}=(\mathrm{sinh}\chi )t+(\mathrm{cosh}\chi )z,`$ (13)
while $`R`$ leads to a rotation:
$`z^{}=(\mathrm{cos}\varphi )z(\mathrm{sin}\varphi )x,`$ (14)
$`x^{}=(\mathrm{sin}\varphi )z+(\mathrm{cos}\varphi )x.`$ (15)
In our previous paper, we discussed in detail what these matrices do when they are applied to the Stokes four-vectors.
In the two-component spinor space, the above transformation matrices take the form
$$\left(\begin{array}{cc}e^{\chi /2}& 0\\ 0& e^{\chi /2}\end{array}\right),\left(\begin{array}{cc}\mathrm{cos}(\varphi /2)& \mathrm{sin}(\varphi /2)\\ \mathrm{sin}(\varphi /2)& \mathrm{cos}(\varphi /2)\end{array}\right).$$
(16)
We discussed the effect of these matrices on the Jones spinors in our earlier publications.
In this paper, we discuss some of nontrivial consequences derivable from the algebra generated by these two sets of matrices. We shall study Wigner rotations and Iwasawa decompositions. The Wigner rotation has been discussed in optical science in connection with Berry’s phase, but the Iwasawa decomposition is a relatively new word in optics. We would like to emphasize here that both the Wigner rotation and Iwasawa decomposition come from the concept of subgroup of the Lorentz groups whose transformations leave the momentum of a given particle invariant.
## III Wigner Rotations
There are many different versions of the Wigner rotation in the literature. Basically, this rotation is a product of two non-collinear Lorentz boosts. The result of these two boosts is not a boost, but a boost preceded or followed by a rotation. This rotation is called the Wigner rotation.
In this paper, we approach the problem by using three boosts described in Fig. 1. Let us start with a particle at rest, with its four momentum
$$P_a=(m,0,0,0),$$
(17)
where we use the metric convention $`(ct,z,x,y)`$. Let us next boost this four-momentum along the $`z`$ direction using the matrix
$$B_1=\left(\begin{array}{cccc}\mathrm{cosh}\eta & \mathrm{sinh}\eta & 0& 0\\ \mathrm{sinh}\eta & \mathrm{cosh}\eta & 0& 0\\ 0& 0& 1& \\ 0& 0& 0& 1\end{array}\right),$$
(18)
resulting in the four-momentum
$$P_b=m(\mathrm{cosh}\eta ,\mathrm{sinh}\eta ,0,0).$$
(19)
Let us rotate this vector around the $`y`$ axis by an angle $`\theta `$. Then the resulting four-momentum is
$$P_c=m(\mathrm{cosh}\eta ,(\mathrm{sinh}\eta )\mathrm{cos}\theta ,(\mathrm{sinh}\eta )\mathrm{sin}\theta ,0).$$
(20)
Instead of this rotation, we propose to obtain this four-vector by boosting the four-momentum of Eq.(19). The boost matrix in this case is
$$B_2=\left(\begin{array}{cccc}1& 0& 0& 0\\ 0& \mathrm{cos}\psi & \mathrm{sin}\psi & 0\\ 0& \mathrm{sin}\psi & \mathrm{cos}\psi & 0\\ 0& 0& 0& 1\end{array}\right)\left(\begin{array}{cccc}\mathrm{cosh}\lambda & \mathrm{sinh}\lambda & 0& 0\\ \mathrm{sinh}\lambda & \mathrm{cosh}\lambda & 0& 0\\ 0& 0& 1& 0\\ 0& 0& 0& 1\end{array}\right)\left(\begin{array}{cccc}1& 0& 0& 0\\ 0& \mathrm{cos}\psi & \mathrm{sin}\psi & 0\\ 0& \mathrm{sin}\psi & \mathrm{cos}\psi & 0\\ 0& 0& 0& 1\end{array}\right),$$
(21)
with
$$\lambda =2\mathrm{tanh}^1\left\{[\mathrm{sin}(\theta /2)]\mathrm{tanh}\eta \right\},\psi =\frac{\theta }{2}+\frac{\pi }{2}.$$
(22)
If we carry out the matrix multiplication,
$$B_2=\left(\begin{array}{cccc}\mathrm{cosh}\lambda & \mathrm{sin}(\theta /2)\mathrm{sinh}\lambda & \mathrm{cos}(\theta /2)\mathrm{sinh}\lambda & 0\\ \mathrm{sin}(\theta /2)\mathrm{sinh}\lambda & 1+\mathrm{sin}^2(\theta /2)(\mathrm{cosh}\lambda 1)& \mathrm{sin}\theta \mathrm{sinh}^2(\lambda /2)& 0\\ \mathrm{cos}(\theta /2)\mathrm{sinh}\lambda & \mathrm{sin}\theta \mathrm{sinh}^2(\lambda /2)& 1+\mathrm{cos}^2(\theta /2)(\mathrm{cosh}\lambda 1)& 0\\ 0& 0& 0& 1\end{array}\right).$$
(23)
Next, we boost the four-momentum of Eq.(20) to that of Eq.(17). The particle is again at rest. The boost matrix is
$$B_3=\left(\begin{array}{cccc}1& 0& 0& 0\\ 0& \mathrm{cos}\theta & \mathrm{sin}\theta & 0\\ 0& \mathrm{sin}\theta & \mathrm{cos}\theta & 0\\ 0& 0& 0& 1\end{array}\right)\left(\begin{array}{cccc}\mathrm{cosh}\eta & \mathrm{sinh}\eta & 0& 0\\ \mathrm{sinh}\eta & \mathrm{cosh}\eta & 0& 0\\ 0& 0& 1& 0\\ 0& 0& 0& 1\end{array}\right)\left(\begin{array}{cccc}1& 0& 0& 0\\ 0& \mathrm{cos}\theta & \mathrm{sin}\theta & 0\\ 0& \mathrm{sin}\theta & \mathrm{cos}\theta & 0\\ 0& 0& 0& 1\end{array}\right).$$
(24)
After the matrix multiplication,
$$B_3=\left(\begin{array}{cccc}\mathrm{cosh}\eta & \mathrm{cos}\theta \mathrm{sinh}\eta & \mathrm{sin}\theta \mathrm{sinh}\eta & 0\\ \mathrm{cos}\theta \mathrm{sinh}\eta & 1+\mathrm{cos}^2\theta (\mathrm{cosh}\eta 1)& \mathrm{sin}\theta \mathrm{cos}\theta (\mathrm{cosh}\eta 1)& 0\\ \mathrm{sin}\theta \mathrm{sinh}\eta & \mathrm{sin}\theta \mathrm{cos}\theta (\mathrm{cosh}\eta 1)& 1+\mathrm{sin}^2\theta (\mathrm{cosh}\eta 1)& 0\\ 0& 0& 0& 1\end{array}\right).$$
(25)
The net result of these transformations is $`B_3B_2B_1`$. This leaves the initial four-momentum of Eq.(17) invariant. Is it going to be an identity matrix? The answer is No. The result of the matrix multiplications is
$$W=\left(\begin{array}{cccc}1& 0& 0& 0\\ 0& \mathrm{cos}\mathrm{\Omega }& \mathrm{sin}\mathrm{\Omega }& 0\\ 0& \mathrm{sin}\mathrm{\Omega }& \mathrm{cos}\mathrm{\Omega }& 0\\ 0& 0& 0& 1\end{array}\right),$$
(26)
with
$$\mathrm{\Omega }=2\mathrm{sin}^1\left\{\frac{(\mathrm{sin}\theta )\mathrm{sinh}^2(\eta /2)}{\sqrt{\mathrm{cosh}^2\eta \mathrm{sinh}^2\eta \mathrm{sin}^2(\theta /2)}}\right\}.$$
(27)
This matrix performs a rotation around the $`y`$ axis and leaves the four-momentum of Eq.(17) invariant. This rotation is an element of Wigner’s little group whose transformations leave the four-momentum invariant. This is precisely the Wigner rotation.
This relativistic effect manifests itself in atomic spectra as the Thomas precession. Otherwise, the experiments on Wigner rotation in special relativity is largely academic. On the other hand, as was noted in the literature, this effect could be tested in optics laboratories. As for the Stokes parameters, the above four-by-four matrices are directly applicable. Indeed, each four-by-four matrix corresponds to one optical filter applicable to polarized light.
In order to see this effect more clearly, let us use the Jones matrix formalism. The two-by-two squeeze matrix corresponding to the boost matrix $`B_1`$ of Eq.(18) is
$$S_1=\left(\begin{array}{cc}e^{\eta /2}& 0\\ 0& e^{\eta /2}\end{array}\right).$$
(28)
The two-by-two squeeze matrix corresponding to the boost matrix of Eq.(21) is now
$$S_2=\left(\begin{array}{cc}\mathrm{cos}(\psi /2)& \mathrm{sin}(\psi /2)\\ \mathrm{sin}(\psi /2)& \mathrm{cos}(\psi /2)\end{array}\right)\left(\begin{array}{cc}e^{\lambda /2}& 0\\ 0& e^{\lambda /2}\end{array}\right)\left(\begin{array}{cc}\mathrm{cos}(\psi /2)& \mathrm{sin}(\psi /2)\\ \mathrm{sin}(\psi /2)& \mathrm{cos}(\psi /2)\end{array}\right),$$
(29)
where the parameters $`\psi `$ and $`\lambda `$ are given in Eq.(22). After the matrix multiplication, $`S_2`$ becomes
$$S_2=\left(\begin{array}{cc}\mathrm{cosh}(\lambda /2)\mathrm{sin}(\theta /2)\mathrm{sinh}(\lambda /2)& \mathrm{cos}(\theta /2)\mathrm{sinh}(\lambda /2)\\ \mathrm{cos}(\theta /2)\mathrm{sinh}(\lambda /2)& \mathrm{cosh}(\lambda /2)+\mathrm{sin}(\theta /2)\mathrm{sinh}(\lambda /2)\end{array}\right).$$
(30)
This is a matrix which squeezes along the direction which makes the angle $`(\pi +\theta )/2`$ with the $`z`$ axis. The two-by-two squeeze matrix corresponding to $`B_3`$ of Eq.(24) is
$$S_3=\left(\begin{array}{cc}\mathrm{cosh}(\eta /2)\mathrm{cos}\theta \mathrm{sinh}(\eta /2)& \mathrm{sin}\theta \mathrm{sinh}(\eta /2)\\ \mathrm{sin}\theta \mathrm{sinh}(\eta /2)& \mathrm{cosh}(\eta /2)+\mathrm{cos}\theta \mathrm{sinh}(\eta /2)\end{array}\right).$$
(31)
Now the matrix multiplication $`S_3S_2S_1`$ corresponds to the closure of the kinematical triangle given in Fig. 1. The result is
$$S_3S_2S_1=\left(\begin{array}{cc}\mathrm{cos}(\mathrm{\Omega }/2)& \mathrm{sin}(\mathrm{\Omega }/2)\\ \mathrm{sin}(\mathrm{\Omega }/2)& \mathrm{cos}(\mathrm{\Omega }/2)\end{array}\right),$$
(32)
where $`\mathrm{\Omega }`$ is given in Eq.(27).
## IV Iwasawa Decompositions
In Sec. III, the Lorentz kinematics was based on a massive particle at rest. If the particle is massless, there are no Lorentz frames in which the particle is at rest. Thus, we start with a massless particle whose momentum is in the $`z`$ direction:
$$K_a=(k,k,0,0),$$
(33)
where $`k`$ is the magnitude of the momentum. We can rotate this four-vector to
$$K_b=(k,k\mathrm{sin}\alpha ,k\mathrm{cos}\alpha ,0)$$
(34)
by applying to $`K_a`$ the rotation matrix
$$R_+=\left(\begin{array}{cccc}1& 0& 0& 0\\ 0& \mathrm{cos}\alpha _+& \mathrm{sin}\alpha _+& 0\\ 0& \mathrm{sin}\alpha _+& \mathrm{cos}\alpha _+& 0\\ 0& 0& 0& 1\end{array}\right),$$
(35)
with $`\alpha _+=\alpha +\pi /2`$.
If we rotate $`K_b`$ around the $`y`$ axis by $`2\alpha `$, the resulting four-momentum will be
$$K_c=(k,k\mathrm{sin}\alpha ,k\mathrm{cos}\alpha ,0).$$
(36)
It is possible to transform $`K_b`$ to $`K_c`$ by applying to $`K_b`$ the boost matrix
$$B=\left(\begin{array}{cccc}\mathrm{cosh}\gamma & \mathrm{sinh}\gamma & 0& 0\\ \mathrm{cosh}\gamma & \mathrm{sinh}\gamma & 0& 0\\ 0& 0& 1& 0\\ 0& 0& 0& 1\end{array}\right).$$
(37)
with
$$\mathrm{sinh}\gamma =\frac{2\mathrm{sin}\alpha }{\mathrm{cos}^2\alpha },\mathrm{cosh}\gamma =\frac{1+\mathrm{sin}^2\alpha }{\mathrm{cos}^2\alpha }.$$
(38)
We can transform $`K_c`$ to $`K_a`$ by rotating it around the $`y`$ axis by $`(\alpha \pi /2)`$. The rotation matrix takes the form
$$R_{}=\left(\begin{array}{cccc}1& 0& 0& 0\\ 0& \mathrm{cos}\alpha _{}& \mathrm{sin}\alpha _{}& 0\\ 0& \mathrm{sin}\alpha _{}& \mathrm{cos}\alpha _{}& 0\\ 0& 0& 0& 1\end{array}\right),$$
(39)
with $`\alpha _{}=\alpha \pi /2`$. Thus, the multiplication of the three matrices, $`R_{}BR_+`$, gives
$$T=\left(\begin{array}{cccc}1+u^2/2& u^2/2& u& 0\\ u^2/2& 1u^2/2& u& 0\\ u& u& 1& 0\\ 0& 0& 0& 1\end{array}\right),$$
(40)
with
$$u=2\mathrm{tan}\alpha .$$
This $`T`$ matrix plays an important role in studying space-time symmetries of massless particles. If this matrix is applied to the four-momentum $`K_a`$ given in Eq. (33), the four-momentum remains invariant. If this matrix is applied to the electromagnetic four-potential for the plane wave propagating along the $`z`$ direction with the frequency $`k`$, the result is a gauge transformation.
Again, the above four-by-four matrices are directly applicable to the Stokes parameters. On the other hand, if we are interested in designing optical filters, we need two-by-two representations corresponding to the four-by-four matrices given so far. The two-by-two squeeze matrix corresponding to the boost matrix $`B`$ of Eq.(37) is
$$S=\left(\begin{array}{cc}e^{\gamma /2}& 0\\ 0& e^{\gamma /2}\end{array}\right),$$
(41)
while the two-by-two matrices corresponding to $`R_+`$ of Eq.(35) and $`R_{}`$ of (39) are
$$R_\pm =\left(\begin{array}{cc}\mathrm{cos}(\alpha _\pm /2)& \mathrm{sin}(\alpha _\pm /2)\\ \mathrm{sin}(\alpha _\pm /2)& \mathrm{cos}(\alpha _\pm /2)\end{array}\right),$$
(42)
where $`\alpha _+`$ and $`\alpha _{}`$ are given in Eq.(35) and Eq.(39) respectively. They satisfy the equations
$$\alpha _++\alpha _{}=2\alpha ,\alpha _+\alpha _{}=\pi .$$
The relation between $`\gamma `$ and $`\alpha `$ given in Eq.(38) can also be written as $`\mathrm{cosh}(\gamma /2)=1/\mathrm{cos}\alpha `$, which is more useful for carrying out the two-by-two matrix algebra.
The matrix multiplication $`R_{}SR_+`$ leads to
$$T=R_{}SR_+=\left(\begin{array}{cc}1& 2\mathrm{tan}\alpha \\ 0& 1\end{array}\right).$$
(43)
Conversely, we can write the
$$\left(\begin{array}{cc}1& 2\mathrm{tan}\alpha \\ 0& 1\end{array}\right)=R_{}SR_+.$$
(44)
The $`T`$ matrix can be decomposed into rotation and squeeze matrices. This possibility is called the Iwasawa decomposition. In the present case, $`T`$ of Eq.(43) can also be written as
$`T=R_{}S\left\{\left(R_{}\right)^1R_1\right\}R_+`$ (45)
$`=\left\{R_{}S\left(R_{}\right)^1\right\}\left(R_1R_+\right)`$ (46)
The matrix chain $`R_{}S\left(R_{}\right)^1`$ is one squeeze matrix whose squeeze axis is rotated by $`\alpha _{}/2`$, and the matrix product $`R_1R_+`$ becomes one rotation matrix. The result is
$$T=S(\alpha _{})R(2\alpha ),$$
(47)
with
$`S(\alpha _{})=\left(\begin{array}{cc}\mathrm{cosh}(\gamma /2)+\mathrm{cos}\alpha _{}\mathrm{sinh}(\gamma /2)& \mathrm{sin}\alpha _{}\mathrm{sinh}(\gamma /2)\\ \mathrm{sin}\alpha _{}\mathrm{sinh}(\gamma /2)& \mathrm{cosh}(\gamma /2)\mathrm{cos}\alpha _{}\mathrm{sinh}(\gamma /2)\end{array}\right),`$ (48)
$`R(2\alpha )=\left(\begin{array}{cc}\mathrm{cos}\alpha & \mathrm{sin}\alpha \\ \mathrm{sin}\alpha & \mathrm{cos}\alpha \end{array}\right).`$ (49)
It is indeed gratifying to note that the $`T`$ matrix can be decomposed into one rotation and one squeeze matrix. The squeeze is made along the direction which makes an angle of $`\alpha _{}/2`$ or $`(\pi /2\alpha )/2`$ with the $`z`$ axis. The angle $`\alpha `$ is smaller than $`\pi /2`$.
We have discussed in our earlier papers optical filters with the property given in Eq.(43). We said there that the filters with this property can be produced from an infinite number of infinitely thin filters. This argument was based on the theory of Lie groups where transformations are generated by infinitesimal generators. This may be possible these days, but the method presented in this paper is far more practical. We need only two filters .
We are able to achieve this improvement because we used here the analogy between polarization optics and Lorentz transformations who share the same mathematical framework.
## Concluding Remarks
In this paper, we noted first that both the Wigner rotation and the Iwasawa decomposition come from Wigner’s little group whose transformations leave the four-momentum of a given particle invariant. Since the Lorentz group is applicable also to the Jones vector and the Stokes parameters, it is possible to construct corresponding transformations in polarization optics. We have shown that both the Wigner rotation and the Iwasawa decomposition can be realized in optics laboratories.
The matrix of Eq.(1) performs a shear transformation when applied to a two-dimensional object, and has a long history in physics and engineering. It also has a history in mathematics. The fact that a shear can be decomposed into a squeeze and rotations is known as the Iwasawa decomposition .
Among the many interesting applications of shear transformations, there is a special class of squeezed states of photons or phonons having the symmetry of shear . The wave-packet spread can be formulated in terms of shear transformations .
As we can see from this paper, a set of shear transformations can be formulated as a subset of Lorentz transformations. This set plays an important role in understanding internal space-time symmetry of massless particles, such as gauge transformation and neutrino polarizations .
## Acknowledgments
We would like to thank A. E. Bak for bringing to our attention to his early works with C. S. Brown on applications of the Lorentz group to polarization optics. We are also grateful to S. Baskal for telling us about the recent paper by Simon and Mukunda on the Iwasawa decomposition .
|
no-problem/9904/math9904072.html
|
ar5iv
|
text
|
# On Cantor’s singular moments
(Date: April 14, 1999)
## Abstract.
We evaluate a constant explicitly, thereby answering a question raised in .
In problem 10621 of the American Mathematical Monthly, Cantor’s singular moments $`J_n`$ were to be computed. In the published answers they come out as
$$J_n=\frac{2}{3(n+1)}\underset{j=0}{\overset{n}{}}\left(\genfrac{}{}{0pt}{}{n+1}{j}\right)\frac{B_j}{32^{j1}1}\text{for}n1$$
and $`J_0=1`$, with Bernoulli numbers $`B_n`$.
The editor asked, whether it is possible to compute
$$J_1=\underset{n0}{}J_n$$
exactly.
The purpose of this note is to do that. In we considered a similar problem, and the gentle reader is invited to consult this paper for more background about the technique, as well as for more information about the Cantor distribution.
Following the method described in , we can write $`J_n`$ as a contour integral viz.
$`J_n`$ $`={\displaystyle \frac{2}{3(n+1)}}{\displaystyle \frac{1}{2\pi i}}{\displaystyle _{\frac{1}{2}i\mathrm{}}^{\frac{1}{2}+i\mathrm{}}}{\displaystyle \frac{\mathrm{\Gamma }(n+2)\mathrm{\Gamma }(1s)}{\mathrm{\Gamma }(n+2s)}}{\displaystyle \frac{\zeta (1s)}{32^{s1}1}}𝑑s`$
$`={\displaystyle \frac{2}{3}}{\displaystyle \frac{1}{2\pi i}}{\displaystyle _{\frac{1}{2}i\mathrm{}}^{\frac{1}{2}+i\mathrm{}}}{\displaystyle \frac{\mathrm{\Gamma }(n+1)\mathrm{\Gamma }(1s)}{\mathrm{\Gamma }(n+2s)}}{\displaystyle \frac{\zeta (1s)}{32^{s1}1}}𝑑s.`$
Therefore
$`{\displaystyle \underset{n=0}{\overset{N}{}}}J_n`$ $`=J_0+{\displaystyle \frac{2}{3}}{\displaystyle \frac{1}{2\pi i}}{\displaystyle _{\frac{1}{2}i\mathrm{}}^{\frac{1}{2}+i\mathrm{}}}{\displaystyle \frac{\mathrm{\Gamma }(N+2)\mathrm{\Gamma }(1s)}{\mathrm{\Gamma }(N+2s)s}}{\displaystyle \frac{\zeta (1s)}{32^{s1}1}}𝑑s`$
$`{\displaystyle \frac{2}{3}}{\displaystyle \frac{1}{2\pi i}}{\displaystyle _{\frac{1}{2}i\mathrm{}}^{\frac{1}{2}+i\mathrm{}}}{\displaystyle \frac{\mathrm{\Gamma }(2)\mathrm{\Gamma }(1s)}{\mathrm{\Gamma }(2s)s}}{\displaystyle \frac{\zeta (1s)}{32^{s1}1}}𝑑s.`$
From this form, one could even compute the asympotics as $`N\mathrm{}`$. However, here, we only have to note that the first integral is of order $`N^{1\mathrm{log}_23}`$, which means that it goes to zero. Consequently
$`{\displaystyle \underset{n0}{}}J_n`$ $`=1{\displaystyle \frac{2}{3}}{\displaystyle \frac{1}{2\pi i}}{\displaystyle _{\frac{1}{2}i\mathrm{}}^{\frac{1}{2}+i\mathrm{}}}{\displaystyle \frac{1}{(1s)s}}{\displaystyle \frac{\zeta (1s)}{32^{s1}1}}𝑑s`$
$`={\displaystyle \frac{4}{3}}+{\displaystyle \frac{2}{3}}{\displaystyle \frac{1}{2\pi i}}{\displaystyle _{\frac{3}{2}i\mathrm{}}^{\frac{3}{2}+i\mathrm{}}}{\displaystyle \frac{1}{s(s1)}}{\displaystyle \frac{\zeta (s)}{32^s1}}𝑑s`$
$`={\displaystyle \frac{4}{3}}+{\displaystyle \frac{2}{3}}{\displaystyle \underset{k,m1}{}}3^k{\displaystyle \frac{1}{2\pi i}}{\displaystyle _{\frac{3}{2}i\mathrm{}}^{\frac{3}{2}+i\mathrm{}}}{\displaystyle \frac{1}{s(s1)}}\left({\displaystyle \frac{2^k}{m}}\right)^s𝑑s.`$
The last step was by using the Dirichlet series for $`\zeta (s)`$ and the geometric series, both valid for $`\mathrm{}s=\frac{3}{2}`$. A simple application of residue calculus, as it is often used in the context of the Mellin–Perron summation formula (see ) evaluates the integrals inside the summation:
$$\frac{1}{2\pi i}_{\frac{3}{2}i\mathrm{}}^{\frac{3}{2}+i\mathrm{}}\frac{1}{s(s1)}t^s𝑑s=\{\begin{array}{cc}t1\hfill & \text{for }t1\hfill \\ 0\hfill & \text{for }t<1\hfill \end{array}.$$
Therefore
$`{\displaystyle \underset{n0}{}}J_n`$ $`=1+{\displaystyle \frac{2}{3}}{\displaystyle \underset{k1}{}}{\displaystyle \underset{1m2^k}{}}3^k\left({\displaystyle \frac{2^k}{m}}1\right)`$
$`=1+{\displaystyle \frac{2}{3}}{\displaystyle \underset{k1}{}}\left({\displaystyle \frac{2}{3}}\right)^kH_{2^k}{\displaystyle \frac{2}{3}}{\displaystyle \underset{k1}{}}\left({\displaystyle \frac{2}{3}}\right)^k`$
$`={\displaystyle \frac{1}{3}}+{\displaystyle \frac{2}{3}}{\displaystyle \underset{k1}{}}\left({\displaystyle \frac{2}{3}}\right)^kH_{2^k}=\mathrm{3.36465\hspace{0.33em}07281\hspace{0.33em}00925\hspace{0.33em}16083\hspace{0.33em}89349\hspace{0.33em}6289}\mathrm{},`$
with harmonic numbers $`H_n=_{1kn}\frac{1}{k}`$.
|
no-problem/9904/cond-mat9904394.html
|
ar5iv
|
text
|
# Mid-infrared absorption in YBa2Cu3O6: Failure of spin-wave theory in undoped cuprates?
\[
## Abstract
The optical conductivity $`\sigma (\omega )`$ of undoped YBa<sub>2</sub>Cu<sub>3</sub>O<sub>6</sub> is studied in detail in the mid-infrared range. Substitutions on all but the Ba site are used to identify the prominent absorption processes at 2800 and 3800 cm<sup>-1</sup>. Experimental evidence for bimagnon–plus–phonon absorption is collected. A more critical analysis of the lineshape and the spectral weight reveals the limits of this approach. Although phonon-2-magnon multiple scattering seems to reproduce the lineshape, the necessary coupling is unrealistically large. The strong increase of high frequency spectral weight with increasing temperature makes the failure of spin-wave theory even more evident.
\]
The undoped parent compounds of the high T<sub>c</sub> cuprates are regarded as an almost ideal realization of a two-dimensional (2D) spin 1/2 Heisenberg antiferromagnet. Despite the low dimensionality and the low spin the excitations are thought to be spin-waves with a well-defined dispersion, as opposed to e.g. 1D systems, where a spinon continuum is observed in neutron scattering. In the cuprates, a spin-wave dispersion has been extracted throughout the whole Brillouin zone from the maxima in neutron scattering intensities, but energies are rather high, large backgrounds are observed and the magnitude of quantum corrections is unclear. Moreover, the assumption that magnons are not well-defined particles at the Brillouin zone boundary was a keypoint in the successful description of the photoemission data of insulating Sr<sub>2</sub>CuO<sub>2</sub>Cl<sub>2</sub> in Ref. . Two-magnon (2M) Raman scattering shows several anomalies in the cuprates, in particular a very broad lineshape, spectral weight at high energies and a finite signal in $`A_{\mathit{1}g}`$ geometry. A large body of theoretical work has been dedicated to this problem, and the importance of resonance phenomena was emphasized. Other treatments include the interaction with phonons and extensions of the Heisenberg model. Certainly, the strong influence of the charge transfer (CT) resonance on the Raman spectra complicates the problem significantly.
Optical spectroscopy probes the magnetic excitations more directly. The main peak in mid-infrared absorption (MIR) spectra of La<sub>2</sub>CuO<sub>4</sub> and other single layer cuprates has been interpreted by Lorenzana and Sawatzky in terms of bimagnon–plus–phonon (BIMP) absorption. A similar feature was reported in the bilayer system YBa<sub>2</sub>Cu<sub>3</sub>O<sub>6</sub> (YBCO<sub>6</sub>), in 2D S=1 La<sub>2</sub>NiO<sub>4</sub> and in 1D S=1/2 Sr<sub>2</sub>CuO<sub>3</sub>. Good agreement is achieved in 1D because quantum fluctuations are included ab initio, and for the 2D S=1 nickelates because fluctuations beyond spin-wave theory are small. However, in the cuprates the estimate for the spectral weight of BIMPs is one order of magnitude too small. Moreover, a large amount of spectral weight is observed above the BIMP absorption. Interpretations in terms of multi-magnon–plus–phonon absorption, $`d`$-$`d`$ transitions and CT excitons have been proposed. We challenge these approaches and suggest that a full account of our MIR data in the undoped cuprates has to include quantum fluctuations beyond spin-wave theory. This might provide an important feedback to the Raman experiment and the basic picture of the undoped cuprates.
Specific ionic substitutions help to identify the prominent resonances in the MIR spectrum of YBCO<sub>6</sub>. In particular we use oxygen isotope substitution to distinguish vibrational from electronic degrees of freedom and to give direct experimental evidence for the existence of BIMP absorption. Substituting rare earth (RE) elements for Y increases the lattice parameter $`a`$ and thereby changes the exchange constant $`J`$. The dependence of the BIMP frequency on $`J(a)`$ is found to be similar to the one reported for 2M Raman scattering.
Single crystals of YBa<sub>2</sub>Cu<sub>3</sub>O<sub>7</sub> were grown in BaZrO<sub>3</sub> (BZO) crucibles. Crystals grown using this technique exhibit a superior purity ($`>`$ 99.995 at. %). The full exchange of the oxygen isotope was described elsewhere. For comparison, single crystals grown in Y<sub>2</sub>O<sub>3</sub> stabilized ZrO<sub>2</sub> (YSZ) crucibles were studied as well. A finite amount of Y in the single crystals of RE<sub>0.8</sub>Y<sub>0.2</sub>Ba<sub>2</sub>Cu<sub>3</sub>O<sub>6</sub> (RE=Pr, Gd) studied here is due to the Y<sub>2</sub>O<sub>3</sub> in the ZrO<sub>2</sub> crucibles. In the case of YBa<sub>2</sub>Cu<sub>3-y</sub>Zn<sub>y</sub>O<sub>6</sub>, we estimate $`0.05y0.07`$. The oxygen content of all crystals was fixed to a value O6 by annealing them for about a week in a flow of high purity Argon (99.998 %) at 750C or in ultra high vacuum at 700C. We calculated $`\sigma (\omega )`$ by inverting the Fresnel equations for the experimentally measured transmission and reflection data. Measurements were carried out with the electric field vector polarized parallel and perpendicular to the $`ab`$-plane. The small remnants of interference fringes in some of the calculated spectra of $`\sigma (\omega )`$ are artifacts caused by deviations of the measured data from the assumed ideal case of absolutely flat and plane parallel surfaces.
We first collect evidence for the applicability of a magnon-phonon interpretation of the main resonance. In Fig. 1 we display $`\sigma (\omega )`$ of YBCO<sub>6</sub> up to the onset of CT absorption for T=4 and 300 K. Note the very low values of $`\sigma (\omega )`$, which are two (four) orders of magnitude lower than for YBCO<sub>6.1</sub> (YBCO<sub>7</sub>) in this frequency range. Following the interpretation of the single layer compounds we ascribe the main peak at 2800 cm<sup>-1</sup> to BIMP absorption. A magnetic origin is also favored by recent measurements of the pressure dependence of the MIR and Raman spectrum of Sr<sub>2</sub>CuO<sub>2</sub>Cl<sub>2</sub>. In the cuprates bimagnon absorption is forbidden due to inversion symmetry. It only becomes weakly allowed by symmetry breaking effects like impurities or the combination with a phonon. Let us compare the experimental data with predictions of BIMP theory on a qualitative basis (Fig. 2). The two upper plots show spectra of two samples of YBCO<sub>6</sub> grown in different crucibles (top panel YSZ, below BZO). The strong temperature dependence of the BIMP peak frequency is most likely due to the reduction of spin stiffness with increasing temperature (left inset, open/full symbols YSZ/BZO). The width $`\gamma `$ of the BIMP resonance is smaller in the cleaner sample grown in a BZO crucible (right inset). We therefore expect that impurity scattering is not negligible in the determination of the correct lineshapes. Substituting Zn on Cu sites indeed broadens the 2800 cm<sup>-1</sup> peak drastically (panel 3 from top in Fig. 2). A finite phonon contribution to the BIMP peak is evident from the frequency shift induced by oxygen isotope substitution in YBCO<sub>6</sub> (see inset of same panel). The measured isotope shift of $`28\pm 8`$ cm<sup>-1</sup> is consistent with the BIMP interpretation, assuming that the longitudinal stretching phonon of approximately 550 – 600 cm<sup>-1</sup> is excited. Substitution of Y with Pr or Gd leads to a significant frequency shift of the main peak. Similar shifts were observed in 2M Raman scattering and were explained by the dependence of $`J`$ on the lattice parameter $`a`$. Similar to the case of Zn, the disorder on the Y site after substitution of 80% of Pr or Gd enhances $`\gamma `$. Finally, the 2800 cm<sup>-1</sup> peak is not observed if the electric field is applied parallel to the $`c`$ axis, in agreement with the single layer data. The spikes on top of the main resonance and the other sharp features have not yet been identified. A relation to BIMP absorption is unlikely, as their position is identical in the RE samples.
For a more critical analysis of peak frequency, lineshape and spectral weight we extend the BIMP theory to the bilayer case. In order to obtain the coupling to light we start from a Heisenberg Hamiltonian which takes into account a dependence of the in- and inter-plane exchange constants $`J`$ and $`J_{12}`$ on the external electric field E and the phonon coordinates:
$$H=\underset{L=1,2}{}\underset{<i,j>}{}J(𝐄,𝐮)𝐒_{L,i}𝐒_{L,j}+\underset{i}{}J_{12}(𝐄,𝐮)𝐒_{1,i}𝐒_{2,i}$$
where $`i`$ and $`j`$ label nearest neighbor Cu sites in a 2D square lattice, $`L`$ labels the two planes in a single bilayer, and $`𝐮`$ denotes the displacements of O ions. Only Einstein phonons are considered. The different phonons modulate the intersite hopping and the on-site energies on both Cu and O sites. Modulations are taken into account to second order since we also included phonon–2M multiple scattering processes in a refined approach (see below). In Ref. a high energy (HE) approximation was used which is inappropriate for zone center excitations. Our RPA results for two interacting magnons show that the HE approximation reproduces the line shape rather well. Both RPA and HE produce 2M bound states with a strong dispersion in momentum space. However, the sharp resonances at the low energy side of the BIMP reported in Ref. are removed in RPA.
Both the Raman and infrared 2M peak frequencies were calculated as a function of $`j=J_{12}/J`$ (left inset of Fig. 1). In the infrared case, the phonon frequency $`\mathrm{}\omega _{ph}`$ still has to be added. At T=4K the experimental BIMP and 2M Raman spectra peak at 2795 and 2720 $`\pm 10`$ cm<sup>-1</sup>, respectively. From these we can deduce the values of $`J`$ and $`\mathrm{}\omega _{ph}`$ for a given ratio $`J_{12}/J`$ (right inset of Fig. 1). We obtain $`J=790\pm 10`$ cm<sup>-1</sup> and $`j=0.08\pm 0.04`$. Neutron scattering suggests $`j`$=0.1 – 0.15 and $`\mathrm{}\omega _{ph}550`$ – 600 cm<sup>-1</sup> for the relevant longitudinal stretching phonon mode. However, a finite next-nearest neighbor coupling $`J^{}`$ will shift the values of $`J`$ and $`J_{12}`$ considerably.
Let us compare the calculated and measured lineshapes and oscillator strengths of the BIMP peak. The calculated BIMP absorption for $`J`$=780 cm<sup>-1</sup>, $`J_{12}`$=0.1$`J`$ and $`\mathrm{}\omega _{ph}=530`$ cm<sup>-1</sup> is plotted together with the experimental curve in Fig. 1 (dash-dotted line). An offset of 0.1 $`\mathrm{\Omega }^1`$cm<sup>-1</sup> has been used. In a bilayer we have to distinguish two contributions: a photon can flip two spins (i) in the same layer (dashed line, in-plane) or (ii) in adjacent layers (dotted line, inter-plane). For (i) a rough estimate of the 2M binding energy in the Ising limit is $`J`$. This implies that the 2M energy at the zone boundary is about $`J`$ below the upper cut-off of the 2M-spectrum, which approximately corresponds to the BIMP position in Fig. 1. Similarly, the binding energy is $`J_{12}`$ for case (ii) which explains the maximum of the calculated inter-plane conductivity just below the 2M cut-off. The estimated relative spectral weight of inter- and in-plane contributions is 0.06 – 0.3 for $`j`$=0.1. Due to this small value the spectral weight is similar in YBCO<sub>6</sub> and in La<sub>2</sub>CuO<sub>4</sub> both experimentally and theoretically. The perturbatively estimated spectral weight is a factor of 8 – 15 too small compared to experiment.
Regarding the lineshape the calculated curve is sharper than the experimental one, but the width of the BIMP peak is sample dependent, as stated above. As in the single layer cuprates, the real problem is obviously at higher frequencies: the strong peak at 3800 cm<sup>-1</sup> remains unexplained. It is likely that the high energy anomaly has the same origin in MIR and Raman spectra. This is substantiated by the absence of the anomaly in both spectroscopies in S=1 La<sub>2</sub>NiO<sub>4</sub>. In several cuprates, the frequency ratio of the two dominant MIR peaks is about 0.73 (Table 1), strongly suggesting a common magnetic origin. In other terms, the second peak in both single and bilayers is close to the 2M cut-off. One way of shifting spectral weight to the 2M cut-off is to consider a finite interaction between phonons and bimagnons. A dimensionless coupling constant for phonon-2M multiple scattering is defined as $`\lambda _{p2M}=\frac{1}{2J}\frac{d^2J}{du^2}u^2,`$ from which we estimate $`\lambda _{p2M}0.02\mathrm{}+0.01`$ for the stretching phonon mode. A negative value of $`\lambda _{p2M}`$ translates into a repulsive phonon-2M interaction and shifts spectral weight to higher frequencies. A way to test the reliability of our estimate of $`\lambda _{p2M}`$ is to compare the linear coupling $`dJ/du`$ with the experimental pressure dependence of $`J`$. There, our estimate is 1 – 2 times smaller. However, an excellent fit to the data (diamonds in top panel of Fig. 2) is obtained only if we assume $`\lambda _{p2M}=0.2`$ (dashed line) and add the BIMP contribution of the apical stretching phonon. The apical contribution is expected to have a 5 times smaller weight and a negligible phonon-2M coupling (dotted line). This large value of $`\lambda _{p2M}`$ makes such a scenario very unlikely. Phonon-magnon scattering processes have also been considered for the explanation of the width of the 2M Raman resonance. Contrary to our dynamic treatment, their adiabatic approach models static disorder, which enhances the width but does not result in a second resonance. Hence we conclude that phonon-magnon interaction processes cannot explain the magnetic MIR and Raman anomalies.
A severe constraint for any interpretation of the high frequency spectral weight is the observed increase by a factor of more than 2 from 4 to 300 K (triangles in inset of panel 2 of Fig. 2). A similar behavior can be detected in the temperature dependence of $`\sigma (\omega )`$ of Sr<sub>2</sub>CuO<sub>2</sub>Cl<sub>2</sub>. We propose that these findings support the notion of a strong local deviation from the Néel state which is even more pronounced for 300 K. Whereas the broken symmetry of the antiferromagnetic state will still support long wavelength spin-wave excitations, the character of the short wavelength magnetic excitations reflects the strong quantum fluctuations and consequently they are insufficiently represented by spin-waves. Both Raman and MIR are dominated by short wavelength magnetic excitations which makes their evaluation within spin-wave theory less reliable. We emphasize that this interpretation does not contradict the good agreement of neutron scattering results with spin-wave theory for small momenta. Note that even in 1D the inapplicability of a spin-wave picture to neutron data was not realized for many years.
Exact diagonalization should serve to identify the weight of magnetic excitations. However, only clusters of up to $`\sqrt{20}\times \sqrt{20}`$ were investigated and produced a minor contribution to MIR absorption at high frequencies. Since this cluster size is still comparable to the size of the considered high frequency excitations, a finite size scaling analysis would be a serious test whether the Heisenberg model or a 4-spin extension of it can generate the measured high frequency weight.
In the absence of doping, the only alternative to a magnetic origin of the anomaly are excitons. Perkins et al. suggested a $`d`$-$`d`$ exciton with sidebands. However, this exciton should not be IR active, and there is theoretical and experimental evidence that its actual energy is a factor of two to three higher. Wang et al. predict a CT exciton at 0.8 eV (6500 cm<sup>-1</sup>) from fits to EELS data between 2.5 and 4 eV. We consider the Coulomb attraction necessary to pull this exciton down to 3800 cm<sup>-1</sup> – far below the CT gap – as unrealistically large. Furthermore, a CT exciton should follow the strong redshift of the onset of CT absorption with increasing temperature (10500 to 9000 cm<sup>-1</sup>, see Fig. 1), which is not observed. Both exciton models fail to describe the increase of spectral weight with temperature. Hence the exciton interpretation is an unlikely scenario for the considered resonance. Direct excitation of 2 magnons in bilayers via spin-orbit coupling was suggested to reproduce the MIR lineshape in YBCO<sub>6</sub> up to 4000 cm<sup>-1</sup>. However, a spin exchange of $`J_{12}=0.5J`$ had to be assumed, which is not anymore consistent with neutron scattering.
We conclude that present day understanding of magnetic excitations in undoped cuprates is not sufficient to explain both MIR and Raman data. Only a more appropriate treatment of the short wavelength excitations will explain the observed anomalies.
It is a pleasure to acknowledge many stimulating discussions with G.A. Sawatzky. We also want to thank R. Eder and J. Brinckmann for helpful discussions. This project is supported by the Netherlands Foundation for Fundamental Research on Matter with financial aid from the Nederlandse Organisatie voor Wetenschappelijk Onderzoek and by the DFG and DFG-GK.
|
no-problem/9904/cond-mat9904270.html
|
ar5iv
|
text
|
# Ring-Pattern Dynamics in Smectic-C∗ and Smectic-C{^∗}{_A} Freely Suspended Liquid Crystal Films
## Abstract
Ring patterns of concentric 2$`\pi `$-solitons in molecular orientation, form in freely suspended chiral smectic-C films in response to an in-plane rotating electric field. We present measurements of the zero-field relaxation of ring patterns and of the driven dynamics of ring formation under conditions of synchronous winding, and a simple model which enables their quantitative description in low polarization DOBAMBC. In smectic C$`{}_{}{}^{}{}_{\mathrm{A}}{}^{}`$ TFMHPOBC we observe an odd-even layer number effect, with odd number layer films exhibiting order of magnitude slower relaxation rates than even layer films. We show that this rate difference is due to much larger spontaneous polarization in odd number layer films.
Smectic liquid crystal phases are one-dimensional (1d) crystals consisting of a periodic stack of two-dimensional (2d) liquid layers. This rigidity allows them to be drawn into stable freely suspended films, integer numbers of smectic layers thick. In tilted smectic phases, the rod-shaped molecules, their long axes on average pointing along $`𝐧`$, strongly prefer to make an angle $`\theta _0`$ with the layer normal $`𝐳`$, but have no energetic preference for the global azimuthal orientation of n about z. Consequently in these tilted phases, this azimuthal low energy degree of freedom can by described by a 2d unit director field, $`𝐜(x,y)`$, well correlated between layers and pointing along the projection of $`𝐧`$ onto the smectic layer, as shown in Fig. 1.
In chiral smectics, symmetry demands an in-plane polarization $`𝐏(x,y)`$ to be present, either along (longitudinal) or perpendicular (transverse) to $`𝐜(x,y)`$, depending on the details of the tilted smectic phase and the number of smectic layers in the film. Consequent coupling of the director c$`(x,y)`$ to an in-plane applied electric field has enabled several key optical experiments on freely suspended films ranging from light scattering and microscopy to the determination of ground state structure of novel liquid crystal phases.
In these experiments, in order to prevent flow instabilities, the direction of the in-plane aligning electric field must be reversed at a frequency of 0.1 to 10 Hz. This process, still not well understood, typically leads to the formation of beautiful concentric ring patterns consisting of 2$`\pi `$-walls in $`𝐜(x,y)`$, which can be imaged via Depolarized Reflected Light Microscopy (DRLM). A particular dark or light region of a DRLM image has a c-director $`𝐜(x,y)=[\mathrm{cos}(\phi (r)),\mathrm{sin}(\phi (r))]`$ of constant azimuthal orientation $`\phi `$. Thus, with $`\phi (r)`$ pinned at the outer radius of the film and monotonically increasing toward the center of the film, rings appear. These distinctive ring patterns in $`𝐜`$ were first studied quantitatively by Cladis et al., who generated them mechanically by rotating a needle inserted into the center of a smectic-C (SmC) film. Later it was reported that ring patterns could be produced in ferroelectric films by applying a rotating (rather than bipolar) in-plane electric field $`𝐄(t)`$, with the first systematic experimental study mapping out the ring-pattern formation phase diagram given in Ref.. Despite significant experimental attention there has only been qualitative theoretical understanding of ring-pattern formation and relaxation dynamics. Here we present a simple model, which enables quantitative understanding of E-field wound ring-pattern dynamics and subsequent zero field ring unwinding relaxation. We find excellent agreement with our experiments on the low polariztion ($`P3`$ nC/cm<sup>2</sup>) SmC\* DOBAMBC. However, our experiments on smectic-C$`{}_{}{}^{}{}_{\mathrm{A}}{}^{}`$ TFMHPOBC reveal a novel odd-even smectic layer effect, with order of magnitude slower relaxation times in the high polarization odd layer number films and large ring distortion, driven by bend/splay elastic anisotropy ($`P75/N`$ nC/cm<sup>2</sup>, where $`N`$ is an odd number of smectic layers). These latter observations lie beyond our model and their quantitative description most certainly requires understanding of screening-ion dynamics.
The free energy density of the 2d nematic orientation field c$`(x,y)`$ of a freely suspended film in an applied field is given by
$$f(x,y)=\frac{K_S}{2}(𝐜)^2+\frac{K_B}{2}(\widehat{𝐳}\times 𝐜)^2𝐄𝐏$$
(1)
where the first two terms are the energies of splay and bend of c with 2d elastic constants $`K_S`$ and $`K_B`$, respectively, originating from the 3d Frank free energy for $`𝐧`$. The last term is the electric field director aligning energy, acting through the polarization field density $`𝐏`$, which we take to be rigidly locked to the $`𝐜`$ director, and for purposes of this developement, orthogonal to c.
Assuming simple relaxational dynamics for $`𝐜`$, with a viscous damping coefficient $`\gamma `$, and taking $`K_SK_BK`$ (but see below), we obtain an equation of motion for $`\phi (x,y,t)`$,
$$\gamma \frac{\phi }{t}=K^2\phi PE\mathrm{sin}(\phi \omega _et).$$
(2)
In above, we have taken $`|𝐏|P`$ and $`|𝐄|E`$ to be constants and the electric field rotating at frequency $`\omega _e`$ in the $`x`$-$`y`$ plane. Eq. 2 quite clearly ignores the dipolar interaction and the rich 2d liquid hydrodynamics coupled to the $`𝐜`$ director, which can in principle become important under conditions of strong drive. Despite these shortcomings, as we will show below, most of our experimental data on ring patterns dynamics in the low $`P`$ liquid crystals is quantitatively described by this model. Its extension to treat observations in high polarization materials will be a subject of a future publication.
A detailed analysis of Eq. 2 predicts a ring formation phase diagram in the applied $`E`$-field strength and winding frequency $`\omega _e`$ space, which is consistent with our experimental observations. The dynamics is simplest in the regime of large $`EE_cU_s/P`$ and small $`\omega _e\omega _cPE/\gamma `$, in which the areal $`𝐜`$-director field alignment torque $`PE`$ is much larger than both the pinning energy (torque) $`U_s`$ of $`𝐜`$ at the outer boundary (radius $`R`$) of the film, and the frictional torque $`\gamma \omega _e`$. In this regime the director field $`𝐜`$ is uniform, synchronously following (at frequency $`\omega =\omega _e`$) the rotating field $`𝐄`$, with a constant phase lag $`\delta =\mathrm{sin}^1(\omega _e/\omega _c)`$, set by the balance between the $`E`$-field alignment and frictional torques. Clearly, for $`\omega _e>\omega _c`$ this synchronous metastable solution, corresponding to the areal alignment of P along E is unstable to a uniform asynchronous dynamical regime, in which the director $`𝐜`$ uniformly winds at a rate $`\omega `$ smaller than $`\omega _e`$ of the rotating $`E`$\- field. Away from the actual transition into this asynchronous regime, the dynamics can be explicitly worked out perturbatively in $`\omega _c/\omega _e`$ and we find that in addition to fast oscillatory dynamics at harmonics of the “washboard” frequency $`\omega _e`$, the spatially uniform phase $`\phi (t)`$ advances linearly in time, on average, with frequency $`\omega =\frac{1}{2}\omega _e(\omega _c/\omega _e)^2\omega _e`$, which, interestingly, decreases with $`\omega _e`$. Obviously, no rings are produced in these two high $`E`$-field regimes.
Ring winding regimes lie in the range of low applied $`E`$ fields, such that the pinning at the outer boundary (at $`R`$) is stronger than the areal alignment torque $`PE`$. To analyze the synchronous, $`\omega _e<\omega _c`$, ring winding dynamics, we take $`\phi (𝐫,t)\stackrel{~}{\phi }(𝐫,t)+\omega _et`$, and look for azimuthally symmetric traveling solution for $`\stackrel{~}{\phi }(𝐫,t)\vartheta (rvt)`$, which satisfies
$$\xi ^2\ddot{\vartheta }+\frac{v}{\omega _c}\dot{\vartheta }=\mathrm{sin}\vartheta +\frac{\omega _e}{\omega _c},$$
(3)
where “dot” indicates differentiation with respect to the argument $`rvt`$, $`\xi \sqrt{K/PE}`$, and we have, for now, neglected the term $`\xi ^2\dot{\vartheta }/r`$ that is subdominant for large rings. The boundary condition $`\phi (R,t)=\phi _0`$ translates into $`\vartheta =\phi _0\omega _et`$ and feeds in $`2\pi `$-solitons (winds rings) at a rate $`\omega _e`$ from the outer boundary $`r=R`$ of the film (see Fig. 2(a)).
Rings are traveling soliton solutions to the above equation, which can be found by noting the isomorphicism of the Eq. 3 with the Newtonian dynamics of a particle of mass $`\xi ^2`$, friction coefficient $`v/\omega _c`$ moving down a (unit strength) sinusoidal potential under an external force $`\omega _e/\omega _c`$. First we note that, without winding, a radial profile of an isolated ($`\omega _e=0`$) ring of radius $`r_0`$ can be determined in closed form and is given by a well-known soliton solution $`\vartheta (r)=4\mathrm{arctan}\left[e^{(rr_0)/\xi }\right]`$. It corresponds to the motion of a fictitious particle between two maxima (at $`0`$ and $`2\pi `$) of the potential $`V(\vartheta )=1\mathrm{cos}\vartheta `$ without external force and therefore a vanishing friction coefficient. Consequently, without winding, an isolated ring is stationary, $`v=0`$. This stationary $`\omega _e=0`$ solution can be easily extended to a concentric periodic array of rings $`d`$ apart, by choosing the “total energy” $`E`$ and therefore the initial “kinetic energy”, $`\xi ^2(\dot{\vartheta })^2/2`$, such that the effective particle can travel between maximas in “time” $`d`$. Although the solution can be expressed in terms of special functions, the only features of it that are important to us is its sigmoidal shape, the ring’s radial width $`\xi `$ and $`\dot{\vartheta }|_{2\pi n}2\pi /d`$.
For a finite synchronous winding rate, $`\omega _e0`$, the fictitious particle is under an external constant force $`\omega _e/\omega _c`$, leading to a tilted periodic potential. The mapping of the ring winding problem onto particle dynamics makes it immediately clear that in this case, a solution of a periodic array of self-similar concentric rings is only possible if the fictitious particle moves in the presence of a unique value of the frictional coefficient $`v/\omega _c`$. This value is determined by the condition that the “energy” $`2\pi \omega _e/\omega _c`$, gained by the particle from descending to a next local potential maximum is precisely the energy $`v/\omega _c_{\vartheta _0}^{\vartheta _0+2\pi }𝑑\vartheta \dot{\vartheta }`$ dissipated due to “friction”. This condition predicts that, even in the absence of tension, rings wound at rate $`\omega _e`$ must move toward the center with velocity $`v(\pi /4)\omega _e\xi `$, a result that can also be clearly seen from noting that the $`2\pi `$-soliton shifts by its width $`\xi `$ at frequency $`\omega _e`$, the rate of rotation of the $`𝐜`$-director (see Fig. 2(a)).
In the time $`\tau _e=2\pi /\omega _e`$ that it takes a new soliton ring to be created at the outer film edge $`R`$, rings created before it, move toward the center of the film a distance $`d=v\tau _e=(\pi ^2/2)\xi `$, predicting a steady state pattern of evenly spaced (by $`d`$), moving, concentric rings, as illustrated in Fig. 2(b, c). As shown in the inset of Fig. 2(c), our experiments indeed find $`d^21/E`$ in agreement with the above theoretical prediction for $`d`$ and $`\xi `$. We also, however, find that rings spacing increases toward the center of the film, inconsistent with the above $`r`$-independent prediction for $`d`$. It is easy to show that this deviation is due to the increased importance (at small $`r`$) of the ring line tension, contained in the $`\xi ^2\dot{\vartheta }/r`$ term, neglected in the Eq. 3. Line tension contributes an additional $`E`$-independent velocity $`\delta v_\tau =dr/dt=K/(\gamma r)`$ that must be superimposed on the velocity due to the rotating field, and predicts a parabolic spacing of solitons rings in the central region.
Contrary to the observations of Dascalu et al. experimentally we find both synchronous ($`\omega <\omega _c`$) and asynchronous ($`\omega >\omega _c`$) winding of ring patterns. However, in contrast to the well-defined soliton-like rings of width $`\xi `$ wound in the synchronous, $`\omega _e<\omega _c`$ regime discussed above, in the asynchronous, $`\omega _e>\omega _c`$ regime, the rings are not solitons and their width is roughly set by $`R/n(t)`$, decreasing as their number $`n(t)`$ grows with frequency $`\omega =\frac{1}{2}\omega _e(\omega _c/\omega _e)^2`$. We have confirmed experimentally our prediction of the linear $`E`$-field dependence of the critical frequency $`\omega _c`$, separating these two ring winding regimes.
We now turn our attention to ring unwinding dynamics at $`E=0`$ . We find that, in contrast to previous claims in the literature a general solution for the relaxation is given by
$$\phi (r,t)=\underset{n}{}A_nJ_0\left(\frac{a_n}{R}r\right)e^{t/\tau _n},$$
(4)
where $`\tau _n=(\gamma /K)(R/a_n)^2`$ is the time constant for $`n`$th mode, $`a_n`$ are the zeros of the zeroth order Bessel function $`J_0(r)`$, and $`A_n`$ are completely determined by the initial condition $`\phi (r,0)`$. Since higher order ($`n>1`$) terms relax with a larger time constant than lower order terms, $`\phi (r,t)`$ rapidly takes on the shape of the lowest $`n=1`$ term.
As a case study, we measured the relaxation of ring patterns in ferroelectric SmC\* DOBAMBC and antiferroelectric SmC$`{}_{}{}^{}{}_{\mathrm{A}}{}^{}`$ TFMHPOBC. Having determined $`A_n`$’s from the initial measured $`E=0`$ $`\phi (r,0)`$ director profile, the subsequent evolution of $`\phi (r,t)`$ is completely specified by Eq. 4. As we show in Fig. 3, our theoretical prediction for $`\phi (r,t)`$ is in excellent agreement with the experimentally determined evolution of the director profiles. These one parameter fits to experimental data allows us to quite accurately determine the important ratio $`\gamma /K`$. In a 5 layer film we find $`\gamma /K=1.997\times 10^5\mathrm{s}/\mathrm{cm}^2`$, which, surprisingly, is an order of magnitude larger than the previously reported value of $`1.4\times 10^4\mathrm{s}/\mathrm{cm}^2`$, for reasons that are still unclear to us.
Such quantitative measurements in SmC$`{}_{}{}^{}{}_{\mathrm{A}}{}^{}`$ TFMHPOBC revealed an intriguing odd-even dependence of $`\gamma /K`$ on layer number $`N`$. In these experiments the relaxation at the center of the pattern, $`\phi (0,t)`$ was recorded and fit to Eq. 4 and used to extract $`\gamma /K`$. The results shown in Fig. 4, reveal that $`N`$-odd films relax much more slowly than $`N`$-even films. The two main differences between $`N`$-odd and $`N`$-even films are that $`N`$-odd films have a large net transverse polarization, while $`N`$-even films have a significantly smaller net longitudinal polarization (in the tilt plane). To demonstrate that it is the difference in the magnitude of the polarization between the odd and even layer films that is responsible for this novel effect, we measured the $`\gamma /K`$ ratio via our ring relaxation technique in three-layer SmC$`{}_{}{}^{}{}_{\mathrm{A}}{}^{}`$ films of almost-racemic TFMHPOBC (a small amount of chiral TFMHPOBC was added to racemic TFMHPOBC so that $`N`$-odd films would have a small net polarization). These low polarization odd-layer films display large c-director fluctuations and ring pattern relaxation rates comparable to $`N=2`$ and $`N=4`$ films, and have an effective $`\gamma /K`$ an order of magnitude lower than the enantiamerically pure material. This strong dependence on the magnitude of the spontaneous polarization indicates, that while Eq. 4 captures the essence of ring pattern relaxation, it is unable to account for the difference in the relaxation rates between small and large polarization materials. We note that the dynamics of low-polarization $`N`$-even films are described much better by Eq. 4 than those of high-polarization $`N`$-odd films. The temperature dependence of $`\gamma /K`$, however, is similar in both $`N`$-even and $`N`$-odd as is shown in the inset of Fig. 4.
This work by was supported by NSF grants DMR96-14061, DMR89-20147, DMR-9809555, DMR-9625111, NASA grant NAG3-1846, and DARPA contract MDA972-90C-0037. L.R. also acknowledges support by the A.P. Sloan and David and Lucile Packard Foundations.
|
no-problem/9904/astro-ph9904004.html
|
ar5iv
|
text
|
# A Semi-Empirical Model of the Infra-Red Universe
## 1 Introduction
Recent observations of the infra-red (IR) and sub-millimeter (sub-mm) extra-galactic background (Puget et al 1996; Fixsen et al 1998; Burigana & Popa 1998; Schlegel, Finkbeiner & Davis 1998; Hauser et al 1998; Dwek et al 1998; Biller et al 1998) provide powerful constraints on models of galaxy evolution, since a large fraction ($`2/3`$) of stellar energy release, after reprocessing by dust, is emitted in this part of the spectrum. Various number count surveys are also being pursued across the waveband, from SCUBA at $`850\mu \mathrm{m}`$ (Blain et al 1999; Smail et al 1998), to ISO at $`175\mu \mathrm{m}`$ (Puget et al 1998) and $`15\mu \mathrm{m}`$ (Aussel et al 1998; Altieri at al 1998). All of the evidence points to strong evolution in the galaxy population as we look out to high redshift. We would like to try and understand this behavior.
Several groups have attempted to model the IR and sub-mm emission from galaxies, using forward evolution models (Blain & Longair 1993; Franceschini et al 1994; Guiderdoni et al 1998; Jimenez & Kashlinsky 1998; Trentham, Blain and Goldader 1999). Guiderdoni’s model is one of the most detailed. They model structure formation in the Cold Dark Matter (CDM) cosmological scenario, and introduce star formation with various recipes for “quiet-disk”, “burst” and “ultra-luminous” sources. The relative fractions of these populations and the timescales of star formation are free parameters in their model. By tuning these they are able to reproduce current observations, but there is a lack of physical motivation in many of their assumptions concerning the number and luminosity evolution of their populations.
In contrast to these models, Malkan & Stecker (1998) present a simple, empirically based calculation of the IR background. Using observed luminosity functions and spectral energy distributions, they predict reasonable values for the background flux by implementing various schemes of galaxy evolution. However, these prescriptions are still without a sound physical basis.
Pei, Fall & Hauser (1998) present yet a different approach. They consider the evolution of the global stellar, gaseous, chemical and radiation contents of the Universe. As inputs they use observations of the extra-galactic background, interstellar gas density history from damped Ly$`\alpha `$ surveys and the rest frame ultra-violet emissivity history from optical galaxy surveys. They derive cosmic histories of star formation, metallicity and radiation from stars and dust. Their global approach cannot distinguish between the different types of star formation processes in galaxies, such as in disks and starbursts, and thus comparison to galaxy number count surveys is difficult.
This work approaches the problem in a semi-empirical manner, somewhere in between the above extremes. The model makes use of reliable empirical results where they exist. For example, the spectra and luminosity functions of galaxies in the IR have been well determined by the IRAS satellite (Saunders et al 1990; Malkan & Stecker 1998). The star formation rate (SFR) history of the Milky Way’s disk is empirically derived from local observations, and then applied globally to all disks, assuming our galaxy is typical of these systems. We present a simple physical model relating SFR history to the IR luminosity, including dust, gas and spectral evolution. Apart from disks, observations of IR sources in the local universe, also reveal many high-luminosity, interacting systems, which appear to be undergoing an intense burst of star formation (Sanders & Mirabel 1996). These systems have been termed starbursts. We develop a physical model for the evolution of this population, assuming they result from mergers and strong tidal interactions between gas rich systems. The final theoretical input to the model is the evolution of the number density and average mass of the disk systems. This is obtained from the model of collision-induced galaxy formation (Balland, Silk & Schaeffer 1998, hereafter BSS98), which attributes galactic morphology to the number of collisions and tidal interactions suffered by a particular galaxy.
We evolve the present day populations of disks and starbursts backwards in time, out to high redshift. This approach results in relatively few free parameters. Without fine tuning, we predict number counts in various wavebands and the IR and sub-mm extra-galactic background, which agree remarkably well with observations. We present the details of the model in §2, our results in §3 and our conclusions in §4.
## 2 The Model
We model the evolution of two distinct populations which contribute significant flux in the IR and sub-mm: disk galaxies and starbursts. The latter we define to be systems undergoing a violent merger or strong tidal interaction, which leads to high rates of heavily obscured star formation and the creation of a spheroidal stellar system or component. An Einstein-de Sitter cosmology has been assumed with $`H_0=\mathrm{\hspace{0.25em}50}\mathrm{km}\mathrm{s}^1\mathrm{Mpc}^1`$. Thus $`t_0=13.0`$ Gyr. A summary of the model parameters, which are described below, is shown in Table 1.
### 2.1 Disk galaxies
We take the present day IR luminosity function of disk galaxies to be well described by the IRAS far-IR (FIR) luminosity function (Saunders et al 1990) below a certain threshold luminosity, $`L_{cut}`$:
$$\mathrm{\Phi }^{disk}(L)=\{\begin{array}{cc}0\hfill & L<L_{min}\hfill \\ C^{disk}(z)(L/L_{}^{disk}(z))^{1\alpha }\mathrm{exp}\left[\frac{1}{2\sigma ^2}\mathrm{log}_{10}^2\left(1+\frac{L}{L_{}^{disk}(z)}\right)\right]\hfill & L_{min}<L<L_{cut}\hfill \\ 0\hfill & L>L_{cut}\text{,}\hfill \end{array}$$
(1)
where $`C_0^{disk}=2.6\times 10^2h^3\mathrm{Mpc}^3`$, $`L_{}^{disk}=10^{8.77}h^2L_{}`$ and $`\alpha =1.09`$. We take $`L_{min}=10^5h^2L_{}`$. Observationally, the luminosity function is not well constrained at these low luminosities, but the extra-galactic background and, for the flux range of interest, the number counts are insensitive to the choice of $`L_{min}`$. We take the upper limit to be $`L_{cut}5.6\times 10^{10}h^2L_{}`$ (Sanders & Mirabel 1996). This defines the difference between disk-like systems and starbursts, which are modeled in §2.2. The effect of the precise choice of $`L_{cut}`$ is examined in §3. Note, the above numerical values of $`L_{min}`$, $`L_{}^{disk}`$ and $`L_{cut}`$ are defined for $`z=0`$.
Following Malkan & Stecker (1998), we divide the IRAS luminosity function into seven different spectral classes, ranging from $`10^3`$ to $`10^3L_{}`$. The spectra of these classes are derived empirically from their IRAS colors. These spectra are then analytically extended beyond 400 $`\mu \mathrm{m}`$ into the sub-mm regime, assuming a dust emissivity dependence of $`\nu ^\beta `$, with $`\beta =1.5`$ (Franceschini, Andreani, & Danese 1998; Roche & Chandler 1993). Results for $`\lambda 1000\mu \mathrm{m}`$ are quite insensitive to the precise choice of $`\beta `$, since most of the flux detected in this part of the spectrum is emitted from sources at $`z1`$, from their rest frame infra-red. At shorter wavelengths, the spectra average over the line and bump features seen around $`10\mu \mathrm{m}`$. This may be important for some of the details of the $`15\mu \mathrm{m}`$ ISO CAM source counts. As the luminosities of the galaxies evolve, the relative proportion of sources in each spectral class changes, and in this way we take account of spectral evolution.
Disk galaxies are observed to be undergoing relatively steady star formation over many Gyr. Their IR luminosity is due to dust heating from both young stars, which spend most of their lives in dusty star-forming regions, and older stars, which contribute to the general interstellar radiation field (ISRF). The present-day fraction of the total from young star heating is $`f_0^{young}0.65`$ (Mayya & Rengarajan 1997; Devereux et al 1994; Xu & Helou 1996; Walterbos & Greenawalt 1996). We consider the luminosity evolution due to these two heating sources separately.
The IR luminosity of disks due to young stars, $`L^{ir,young}`$, is expected to evolve with redshift in the following manner:
$$\frac{L^{ir,young}(z)}{L_0^{ir,young}}=\frac{\varphi (z)}{\varphi _0}\frac{(1e^{\tau ^{uv}(z)})}{(1e^{\tau _0^{uv}})}\frac{M^{gal}(z)}{M_0^{gal}},$$
(2)
where $`\varphi (z)`$ is the local star formation rate (SFR) per unit disk mass of disk galaxies, $`\tau ^{uv}(z)`$ is the disk optical depth to young stellar energy release and $`M^{gal}(z)`$ is the mean baryonic mass of disk galaxies. These are all redshift dependent.
We derive $`\varphi `$ from the metallicity distributions, $`dN/dZ`$, of local G-type stars (Rocha-Pinto & Maciel 1996; Wyse & Gilmore 1995), (Figure 1), and the age-metallicity relationship, $`dZ/dt`$, of local F-type stars (Edvardsson et al 1993), (Figure 2a), using
$$\varphi (t)\frac{dN}{dZ}\frac{dZ}{dt}.$$
(3)
$`\varphi (t)`$ is further constrained by having to produce a local present-day stellar disk surface density of $`40\mathrm{M}_{}\mathrm{pc}^2`$ (Sackett 1997), assuming the returned mass fraction is small. From a sample of nearby white dwarfs and consideration of their cooling curves, Oswald et al (1996) find the local disk of the Milky Way has an age of $`10\pm 1`$ Gyr. Using a similar method, Knox, Hawkins & Hambly (1999) derive an age of $`10_1^{+3}`$ Gyr. However, because we wish to describe the galaxy averaged evolution of disks, we take the mean age to be $`12`$ Gyr, since it is expected the inner disk will start forming at earlier times. For simplicity we use this formation time to complete the normalization of $`\varphi (t)`$, which we now take to approximate the evolution of the total Galactic SFR history. An age of 12 Gyr corresponds to a disk formation time, $`t_f1`$ Gyr, and redshift, $`z_f5.5`$. At these high redshifts, the model predictions of the source counts and backgrounds are quite insensitive to the precise choice of $`z_f`$. The major uncertainty in this procedure is in fitting the age-metallicity relation, because of the large scatter in the data. We fit the data for $`dN/dZ`$, which show less scatter, with a Gaussian. Only the fit at solar metallicities and below affects the derived SFR history. $`dZ/dt`$ is fit with a function which rises asymptotically from low metallicities at $`tt_f`$, and then levels off to solar metallicity at $`tt_0`$, with the additional constraint that the integrated SFR history (Figure 2b) matches the observed stellar disk surface density. The SFR rises rapidly at early times, peaks at a level roughly ten times that of today, and then decays exponentially. Assuming the Milky Way is typical of disk galaxies, we apply $`\varphi `$ to the entire population of disks. This method enables us to empirically probe the evolution of disk systems separately from the other components of the Universe.
With a redshift independent dust-to-metals mass ratio (Pei et al 1998, their Figure 1), $`\tau `$ is proportional to the metallicity, $`Z`$, and the local gas density in the disk, $`\rho ^{gas}`$. Thus,
$$\frac{\tau (z)}{\tau _0}=\frac{Z(z)}{Z_0}\frac{\rho ^{gas}(z)}{\rho _0^{gas}}=\frac{Z(z)}{Z_0}\left(\frac{\varphi (z)}{\varphi _0}\right)^{2/3},$$
(4)
where we have used the Schmidt law, $`\varphi (\rho ^{gas})^n`$, with $`n=1.5`$, to relate gas density to the rate of star formation. This relation is observed to hold over a large range of gas densities relevant to both disks and starbursts (Kennicutt 1998). The gas density history is shown in Figure 2c. We know the metallicity history directly from the age-metallicity relationship (Figure 2a). The optical depth history is shown in Figure 2d.
$`M^{gal}(z)`$ is derived from the model of collision-induced galaxy formation of BSS98 and is shown in Figure 2e. The mean baryonic mass in a galaxy is controlled by the competition between merging and cooling. It results from consideration of the Press-Schechter mass function, together with an applied cooling constraint, which gives an upper limit to galaxy masses. There is little evolution in the mean mass out to $`z2`$. By $`z5`$ it has fallen by a factor of a few.
The fraction of the baryonic mass bound to a galaxy that is in the form of gas, $`f^{gas}`$, can be calculated, given $`\varphi `$ and $`f_0^{gas}`$. This gas is not necessarily in the disk of the galaxy, but may be in the process of falling in from the halo. It is available to form stars in the event of a merger or close encounter with another galaxy, as tidal forces will channel the gas (or some portion of it) into the central starbursting regions. Thus knowledge of the evolution of $`f^{gas}`$ is necessary for modeling of the starbursts in §2.2. From observations of local disk galaxies, we take $`f_0^{gas}0.1`$ (Young & Scoville 1991; Young et al 1995). The effect of varying this choice is investigated in §3. Its evolution is shown in Figure 2f.
We describe the IR luminosity due to dust heating by the older stellar population as follows:
$$\frac{L^{ir,old}(z)}{L_0^{ir,old}}=\frac{_{t_f}^{t_z}\varphi (t)𝑑t}{_{t_f}^{t_0}\varphi (t)𝑑t}\frac{(1e^{\tau ^{opt}(z)})}{(1e^{\tau _0^{opt}})}\frac{M^{disk}(z)}{M_0^{disk}}.$$
(5)
For simplicity it has been assumed that the intensity of the ISRF of disks in the optical scales as the integrated SFR history. The combination of $`L^{ir,young}`$ and $`L^{ir,old}`$ is a simple representation of the true situation in which there is a continuously varying contribution to dust heating over the whole mass spectrum and environments of stars present in a galaxy.
From Wolfire et al (1999) we set $`\tau _0^{uv}0.7`$. We assume $`\tau _0^{opt}0.1`$, placing present-day disks in the optically thin limit, with respect to the general ISRF. The calculation of the IR and sub-mm number counts and background is relatively insensitive to the choices of $`\tau _0^{uv}`$ and $`\tau _0^{opt}`$, provided we are in the optically semi-thick and thin regimes respectively, since, in this model, it is the scaling of the luminosity function, with respect to today’s, which is important.
We implement luminosity evolution of the present-day characteristic luminosity, $`L_{}^{disk}`$, using equations (2) and (5), each contributing the appropriate fraction, $`f_0^{young}`$ and $`1f_0^{young}`$ respectively to the total. This is shown in Figure 3a.
We obtain number density evolution (Figure 3b) from the models of BSS98 for field (i.e. non-cluster) galaxies, and apply this scaling to the normalization constant, $`C^{disk}(z)`$, of the luminosity function. In their model disk galaxies form from clouds which experience few, if any, collisions between the formation time and the epoch under consideration. Strong collisions will tend to prevent the gas from settling into a disk, allow for tidal exchanges which average out angular momentum, and lead to the formation of ellipticals.
### 2.2 Starbursts
We model starburst galaxies as gas rich systems which are undergoing a merger or a strong tidal interaction. We assume the starburst number density scales with the rate of such interactions, while the mean luminosity scales as the average galactic mass, $`M^{gal}`$, and the gas fraction of baryonic matter, $`f^{gas}`$, available for star formation. We treat the star formation as being completely obscured by dust. Observations support this approximation (Sanders & Mirabel 1996), and it is imagined that even for metal free systems, the initial burst-induced star formation acts very quickly to pollute the ISM and so obscure the vast majority of the stellar energy release.
We model the local ($`z0.2`$) starburst population with the high luminosity end of the IRAS FIR luminosity function:
$$\mathrm{\Phi }^{sb}(L)=\{\begin{array}{cc}0\hfill & L<L_{cut}\hfill \\ C^{sb}(z)(L/L_{,0}^{sb}(z))^{1\alpha }\mathrm{exp}\left[\frac{1}{2\sigma ^2}\mathrm{log}_{10}^2\left(1+\frac{L}{L_{,0}^{sb}(z)}\right)\right]\hfill & L_{cut}<L<L_{max}\hfill \\ 0\hfill & L>L_{max}\text{,}\hfill \end{array}$$
(6)
where $`C_0^{sb}=2.6\times 10^2h^3\mathrm{Mpc}^3`$, $`L_{}^{sb}=10^{8.77}h^2L_{}`$, $`\alpha =1.09`$ (Saunders et al 1990) and $`L_{cut}5.6\times 10^{10}h^2L_{}`$ (Sanders & Mirabel 1996). We take $`L_{max}=10^{15}h^2L_{}`$. The results are insensitive to this choice because of the steepness of the luminosity function, equivalent to $`L^{2.35}`$, (Kim & Saunders 1998). Note, $`L_{}^{sb}`$ is not a characteristic starburst luminosity. It is simply used to parameterize the luminosity function.
We use the same spectral luminosity classes as for the disk model. Starbursts, with higher FIR luminosities than disks, have a stronger component of warm dust emission.
$`L_{}^{sb}`$ is modeled to evolve as follows:
$$\frac{L_{}^{sb}(z)}{L_{,0}^{sb}}=\frac{M^{gal}(z)}{M_0^{gal}}\frac{f^{gas}(z)}{f_0^{gas}}.$$
(7)
Since all the energy released by star formation is reradiated in the far-IR, $`L_{}^{sb}`$ is proportional to the average galactic mass and gas fraction of the merging systems.
To model the number density evolution, we assume $`C^{sb}`$ is proportional to the rate of collisions between gas rich systems, $`\mathrm{\Gamma }^{coll}`$. The collision rate of any given galaxy is
$$\mathrm{\Gamma }_1^{coll}=\frac{v}{\lambda _{mfp}}n^{gal}R^2vn_c^{gal}(1+z)^3M^{gal}\frac{v_0}{(1+z)^{1/2}},$$
(8)
where $`v`$ is the mean peculiar velocity of galaxies, $`\lambda _{mfp}`$ is the mean free path before a collision or strong interaction occurs, $`R`$ is the mean galactic linear size, $`n^{gal}`$ is the true number density of (gas rich) galaxies and $`n_c^{gal}`$ is the corresponding comoving number density. We assume that galaxies are disk-like so that $`M^{gal}=kR^2`$ for some constant $`k`$, which we take to be redshift independent. The redshift dependence of $`v=v_0(1+z)^{1/2}`$ results from consideration of structure formation in the linear regime. The total collision rate is thus:
$$\frac{\mathrm{\Gamma }^{coll}}{\mathrm{\Gamma }_0^{coll}}=\frac{C^{sb}}{C_0^{sb}}=(1+z)^{5/2}\left(\frac{n_c^{gal}}{n_{c,0}^{gal}}\right)^2\frac{M^{gal}}{M_0^{gal}}.$$
(9)
We obtain $`n_c^{gal}(z)`$ from BSS98, by summing their disk and irregular morphological types. These dominate the galaxy population at any particular epoch. The luminosity and number density evolutions of the starburst population are shown in Figure 4. The luminosity evolution depends on the value of $`f_0^{gas}`$ and the effect of its variation is shown.
## 3 Results
For our IR and sub-mm sources we predict number counts, redshift distributions, the intensity of the extra-galactic background and the global SFR history. The values of our model’s parameters are based directly on observations (Table 1) and so we do not attempt to fine tune them to obtain “perfect” fits to all the observations. Rather, we demonstrate that this simple model of galaxy evolution is consistent with all the available data, and examine the effects of varying our two most sensitive model parameters, $`L_{cut}`$ and $`f_0^{gas}`$.
The disk, starburst and total integrated number counts are shown in Figure 5. The model agrees well with observations made in the IRAS $`12\mu \mathrm{m}`$ (Rush et al 1993), ISO CAM $`15\mu \mathrm{m}`$ (Aussel et al 1998; Altieri et al 1998), IRAS $`60\mu \mathrm{m}`$ (Lonsdale et al 1990), ISO PHOT $`175\mu \mathrm{m}`$ (Puget et al 1998) and SCUBA $`850\mu \mathrm{m}`$ (Blain et al 1999; Smail et al 1998 and references therein) passbands. At 15 $`\mu \mathrm{m}`$ disks dominate the counts at high fluxes, while the starburst contribution becomes comparable between $`10^3`$ to $`10^4`$ Jy. At lower fluxes, the starburst counts flatten off, as we are probing the limit of the distribution. At 60 $`\mu \mathrm{m}`$ the disks again dominate at the high flux end, with starbursts becoming more important at around $`10^2`$ Jy. At 175 $`\mu \mathrm{m}`$ the starburst counts rise steeply to dominate the total between $`1`$ and $`10^2`$ Jy, which is the region probed by current ISO PHOT observations. At 850 $`\mu \mathrm{m}`$ starbursts completely dominate at all fluxes above $`10^3`$ Jy.
Figure 6 shows the total counts in expanded regions of the flux-number count diagrams relevant to the latest observations. The ISO 15 $`\mu \mathrm{m}`$, ISO 175 $`\mu \mathrm{m}`$ and SCUBA 850 $`\mu \mathrm{m}`$ plots also show the effect of varying $`f_0^{gas}`$ and $`L_{cut}`$ from the fiducial values. Varying $`f_0^{gas}`$ affects the starburst population while $`L_{cut}`$ dictates the relative contribution of disks to starbursts. The IRAS 60 $`\mu \mathrm{m}`$ data, sampling the low redshift population, are very insensitive to these parameters.
Figure 7 shows the predicted redshift distributions of the source populations observed by ISO CAM, IRAS, ISO PHOT and SCUBA. The discontinuities in the distributions are artifacts resulting from our simplistic method of dividing the luminosity function into disk and starburst sources and discrete spectral classes. The observed disks are at lower redshifts relative to the bulk of the observed starbursts. In the case of the SCUBA sources detected with $`S>0.63`$ mJy, about $`2/3`$ are predicted to be starbursting systems, and the rest normal disks. Most of these sources are predicted to be at redshifts greater than one, with many seen out to redshift five and beyond. This is due to the steep slope of the Rayleigh-Jeans portion of the modified blackbody spectra of the sources. As more redshift determinations are made of these source samples, it will be interesting to compare the inferred distributions, corrected for clustering effects and incompleteness, to the model.
The predictions of the extra-galactic background in the IR to sub-mm are shown in Figure 8. Recent independent estimates of the flux from the FIRAS residuals between 150 - 5000 $`\mu \mathrm{m}`$ (Puget et al 1996; Fixsen et al 1998) agree on the spectrum and amplitude of the background, although the sizes of the systematic errors are not well determined. Strong upper limit constraints are also being reported at shorter wavelengths from observations of TeV $`\gamma `$-rays (Stanev & Franceschini 1998; Biller et at 1998), while lower limits can be placed from the summed flux predicted by number count surveys. Our fiducial model is consistent with the existing data, although at longer wavelengths it predicts a flux about twice as high as the mean of the FIRAS amplitudes. This discrepancy is within the bounds of model and observational uncertainties, as illustrated in the lower panel of Figure 8.
Finally we derive the global SFR history (Figure 9) by assuming a conversion factor of $`2\times 10^{10}\mathrm{M}_{}\mathrm{yr}^1\mathrm{L}_{}^{\mathrm{FIR}}`$. This is in agreement with recent calibrations ($`13\times 10^{10}\mathrm{M}_{}\mathrm{yr}^1\mathrm{L}_{}^{\mathrm{FIR}}`$) from starburst synthesis models (Leitherer & Heckman 1995; Lehnert & Heckman 1996; Meurer et al 1997). This conversion factor is applied to the far-IR luminosity due to dust heating by young stars in disks, accounting for the evolving optical depth, as well as to the optically thick starburst population. The results for the disk contribution are thus relatively sensitive to our choice of $`\tau _0^{uv}0.7`$. However, with this fiducial value, the model predictions agree well with the low redshift observations, given the uncertainties in the SFR to far-IR luminosity conversion factor. At high redshifts the model is consistent with recent results which suggest a relatively flat SFR history out to $`z4`$ (Steidel et al 1998).
If we assume that our starbursts lead to the formation of spheroidal stellar systems, such as ellipticals and the bulges and halos of disk galaxies, then we can predict the mass densities of the disk and spheroidal components that exist in the present-day Universe. Neglecting mass returned to the inter-stellar and inter-galactic media, we find $`\mathrm{\Omega }_{disk}5.1\times 10^3`$ and $`\mathrm{\Omega }_{sph}4.4\times 10^3`$. These estimates are sensitive to $`L_{cut}`$ and are uncertain by factors of a few. Observationally, the spheroidal component is dominated by the bulges and halos of disk galaxies, due to the paucity of ellipticals in the field (Binggeli, Sandage & Tammann 1988). Our results suggest the mass in spheroids is comparable to the mass in the disks, consistent with the analysis of Schechter & Dressler (1987).
Our model indicates that there is a broad peak of spheroid formation at $`z3`$ (Figure 9). Recently, evolved ellipticals have been seen in deep NICMOS images (Benitez et al 1998), requiring a very high ($`z5`$) redshift of formation. Extrapolating the global SFR predictions of our model to these high redshifts, whilst uncertain, indicates that the SFR’s are high enough to account for some very early elliptical formation. Note, however, within the framework of our model, the contribution to the extra-galactic background and the number counts from these early times ($`t<1`$ Gyr) is negligible. In other words, almost all the star formation and associated energy release necessary to account for the observed extra-galactic background occurs at redshifts $`5`$ and is accessible to current observations.
## 4 Conclusions
We have presented a simple model for the evolution of disk galaxies and starbursts, tied as closely as possible to observations. Despite its simplicity, the model takes account of dust, gas and spectral evolution in a self-consistent manner, and is able to predict source counts and the extra-galactic background in the IR to sub-mm consistent with observations, without recourse to fine tuning of parameters. A disk-only model, with $`L_{cut}=L_{max}`$, gives only marginally greater counts and fluxes than the disk component of the fiducial model, thus failing to account for observations. This demonstrates the significant role of the starburst population. The fiducial model we present is based on the best estimates and observations of our various model parameters (Table 1). We have not attempted to vary these to obtain the best fit to the, often uncertain, high redshift IR and sub-mm data.
The predicted global SFR history agrees with recent observations, corrected for dust extinction, indicating that SFR’s remain high from $`z1`$ back to $`z4`$. At high redshifts the SFR is dominated by starbursts and is driven by galaxy-galaxy interactions. Over the history of the Universe, the total star formation occurring in starbursts is comparable to that in disks. This suggests the baryonic mass in bulges, formed from starbursts, is similar to that in disks. The bulges form at high redshift, and are then thought to act as the seeds for disk formation as gas infalls.
The energy release associated with the entire star formation history to $`z5`$ is enough to account for all of the observed extra-galactic background. This implies that the vast majority of star formation in the Universe occurs over this period.
The majority ($`2/3`$) of the sources recently detected with SCUBA are identified as starbursts. Typical redshifts are $`1z5`$. The direct effect of these sources, along with other foregrounds, on the angular power spectrum of the microwave background has been examined by Gawiser et al (1999). The effect of cluster-induced lensing of the sources on the background has been investigated by Scannapieco, Silk & Tan (1999) and found to be small.
Future possible improvements include a more sophisticated method of distinguishing disks and starbursts in the infra-red luminosity function, a more detailed treatment of the opacity of starbursts and the inclusion of active galactic nuclei. Extending the model to the ultra-violet, optical and radio, will allow additional observations to help constrain the galaxy evolution. Consideration of alternative cosmologies is left to a future date. The model will be considerably refined once results are available from future observations with FIRST, SIRTF and the NGST. Model results are available electronically at http://astro.berkeley.edu/ jt/irmodel.html.
We are grateful to Matt Malkan for kindly providing his data for the IR galaxy spectra in electronic form, and to Herve Aussel for his ISO CAM 15 $`\mu \mathrm{m}`$ number count data. We thank R. Bouwens, J. Puget, E. Scannapieco, A. Cumming and E. Gawiser for helpful discussions.
|
no-problem/9904/gr-qc9904009.html
|
ar5iv
|
text
|
# Non-linear Evolution of Rotating Relativistic Stars
## 1 Introduction
The numerical evolution of neutron stars in full General Relativity has been the focus of many research groups in recent years . So far, these studies have been limited to initially non-rotating stars. However, the numerical investigation of many interesting astrophysical applications, such as the rotational evolution of proto-neutron stars and merged neutron stars or the simulation of gravitational radiation from unstable pulsation modes, requires the ability of accurate long-term evolutions of rapidly rotating stars. We thus present here the first study of hydrodynamical evolutions of rotating neutron stars in the approximation of a static spacetime. This approximation allows us to evolve relativistic matter for a much longer time than present coupled spacetime plus hydrodynamical evolution codes. Since the pulsations of neutron stars are mainly a hydrodynamical process, the exclusion of the spacetime dynamics has only a limited effect and allows for qualitative conclusions to be drawn.
The rotational evolution of neutron stars can be affected by several instabilities (see for a recent review). If hot protoneutron stars are rapidly rotating, they can undergo a dynamical bar-mode instability . When the neutron star has cooled to about $`10^{10}`$K after its formation, it can be subject to the Chandrasekhar-Friedman-Schutz instability and it becomes an important source of gravitational waves. It was recently found that the $`l=m`$ $`r`$-mode has the shortest growth time of the instability and it can transform a rapidly rotating newly-born neutron star to a Crab-like slowly-rotating pulsar within about a year after its formation . In this model, there are two important questions still to be answered : What is the maximum amplitude that an unstable $`r`$-mode can reach (limited by nonlinear saturation) and is there any transfer of energy to other stable or unstable modes via non-linear couplings? Such questions cannot be answered by computations of normal modes of the linearized pulsation equations, but require non-linear effects to be taken into account. We therefore need to develop the capability of full non-linear numerical evolutions of rotating stars in General Relativity.
Our present 2-D (axisymmetric) code uses high-resolution shock-capturing (HRSC) finite-difference schemes for the numerical integration of the general relativistic hydrodynamic equations (see for a recent review of applications of HRSC schemes in relativistic hydrodynamics). In a similar context to the one presented here let us note that such schemes have been succesfully used before in the study of the numerical evolution and gravitational collapse of non-rotating neutron stars in 1-D . An alternative approach, based on pseudospectral methods, has been presented in .
Using our code in 1-D time-evolutions we can accurately identify specific normal modes of pulsation. In 2-D the code is suitable for the evolution of rotating stars, with the additional complication of having to pay special attention to an angular momentum-loss at the (non-spherical) surface of the star, as we will show below.
## 2 Initial Configurations
Our initial models are fully relativistic, stationary and axisymmetric configurations, rotating with uniform angular velocity $`\mathrm{\Omega }`$. The metric in quasi-isotropic coordinates is
$`ds^2=e^{2\nu }dt^2+B^2e^{2\nu }r^2\mathrm{sin}^2\theta (d\varphi \omega dt)^2+e^{2\alpha }(dr^2+r^2d\theta ^2),`$ (1)
where $`\nu `$, $`B`$, $`\alpha `$ and $`\omega `$ are metric functions (gravitational units are implied). In the non-rotating limit the above metric reduces to the metric of spherical relativistic stars in isotropic coordinates.
We assume a perfect fluid, zero-temperature equation of state (EOS), for which the energy density is a function of pressure only. The following relativistic generalization of the Newtonian polytropic EOS is chosen:
$`p`$ $`=`$ $`K\rho _0^{1+1/N}`$ (2)
$`ϵ`$ $`=`$ $`\rho _0+Np,`$ (3)
where $`p`$ is the pressure, $`ϵ`$ is the energy density, $`\rho _0`$ is the rest-mass density, $`K`$ is the polytropic constant and $`N`$ is the polytropic exponent.
The initial equilibrium models are computed using a numerical code by Stergioulas & Friedman which follows the Komatsu, Eriguchi & Hatchisu method (as modified in ) with some changes for improved accuracy (see for a comparison with other codes). The code is freely available and can be downloaded from the following URL address: http://www.gravity.phys.uwm.edu/Code/rns.
## 3 Relativistic Hydrodynamic Equations
The equations of (ideal) relativistic hydrodynamics are obtained from the local conservation laws of density current, $`J^\mu `$ and stress-energy, $`T^{\mu \nu }`$
$`_\mu J^\mu `$ $`=`$ $`0`$ (4)
$`_\mu T^{\mu \nu }`$ $`=`$ $`0`$ (5)
with
$`J^\mu `$ $`=`$ $`\rho _0u^\mu `$ (6)
$`T^{\mu \nu }`$ $`=`$ $`\rho _0hu^\mu u^\nu +pg^{\mu \nu },`$ (7)
for a general EOS $`p=p(\rho ,\epsilon )`$. This choice of the stress-energy tensor limits our study to perfect fluids.
In the previous expressions $`_\mu `$ is the covariant derivative, $`u^\mu `$ is the fluid 4-velocity and $`h`$ is the specific enthalpy
$`h=1+\epsilon +{\displaystyle \frac{p}{\rho _0}}`$ (8)
with $`\epsilon `$ being the specific internal energy, related to the energy density $`ϵ`$ by
$`\epsilon ={\displaystyle \frac{ϵ}{\rho _0}}1.`$ (9)
With an appropriate choice of matter fields the equations of relativistic hydrodynamics constitute a (non-strictly) hyperbolic system and can be written in a flux conservative form, as was first shown in for the one-dimensional case. The knowledge of the characteristic fields of the system allows the numerical integration to be performed by means of advanced high-resolution shock-capturing (HRSC) schemes,using approximate Riemann solvers (Godunov-type methods). The multidimensional case was studied in , within the framework of the 3+1 formulation. Further extensions of this work to account for dynamical spacetimes, described by the full set of Einstein’s non-vacuum equations, can be found in . Fully covariant formulations of the hydrodynamic equations (i.e., not restricted to spacelike approaches) and also adapted to Godunov-type methods, are presented in .
In the present work we use the hydrodynamic equations as formulated in . Specializing for the metric given by Eq. (1), the 3+1 quantities read
$`\stackrel{~}{\alpha }`$ $`=`$ $`e^\nu `$ (10)
$`\beta _\varphi `$ $`=`$ $`\omega B^2e^{2\nu }r^2\mathrm{sin}^2\theta `$ (11)
$`\gamma _{rr}`$ $`=`$ $`e^{2\alpha }`$ (12)
$`\gamma _{\theta \theta }`$ $`=`$ $`r^2e^{2\alpha }`$ (13)
$`\gamma _{\varphi \varphi }`$ $`=`$ $`B^2e^{2\nu }r^2\mathrm{sin}^2\theta `$ (14)
where $`\stackrel{~}{\alpha }`$ is the lapse function (the tilde is used to avoid confussion with the metric potential $`\alpha `$) and $`\beta _\varphi `$ is the azimuthal shift.
The hydrodynamic equations are written as a first-order flux conservative system of the form
$`{\displaystyle \frac{𝐮}{t}}+{\displaystyle \frac{\stackrel{~}{\alpha }𝐟^r}{r}}+{\displaystyle \frac{\stackrel{~}{\alpha }𝐟^\theta }{\theta }}=𝐬`$ (15)
where $`𝐮,𝐟^r,𝐟^\theta `$ and $`𝐬`$ are, respectively, the state vector of evolved quantities, the radial and polar fluxes and the source terms. More precisely, they take the form
$`𝐮`$ $`=`$ $`(D,S_r,S_\theta ,S_\varphi ,\tau )`$ (16)
$`𝐟^r`$ $`=`$ $`(Dv^r,S_rv^r+p,S_\theta v^r,S_\varphi v^r,(\tau +p)v^r)`$ (17)
$`𝐟^\theta `$ $`=`$ $`(Dv^\theta ,S_rv^\theta ,S_\theta v^\theta +p,S_\varphi v^\theta ,(\tau +p)v^\theta ).`$ (18)
The source terms can be decomposed in the following way
$`𝐬=\stackrel{~}{\alpha }𝐬^{}\stackrel{~}{\alpha }𝐟^r{\displaystyle \frac{\mathrm{log}\sqrt{\gamma }}{r}}\stackrel{~}{\alpha }𝐟^\theta {\displaystyle \frac{\mathrm{log}\sqrt{\gamma }}{\theta }}`$ (19)
with $`\gamma =det\gamma _{ij}`$ and
$`𝐬^{}=(0,T^{\mu \nu }\left[{\displaystyle \frac{g_{\nu j}}{x^\mu }}\mathrm{\Gamma }_{\mu \nu }^\delta g_{\delta j}\right],\stackrel{~}{\alpha }\left[T^{\mu t}{\displaystyle \frac{\mathrm{log}\stackrel{~}{\alpha }}{x^\mu }}T^{\mu \nu }\mathrm{\Gamma }_{\mu \nu }^t\right])`$ (20)
with $`j=r,\theta ,\varphi `$. The definitions of the evolved quantities in terms of the “primitive” variables ($`\rho ,v_j,\epsilon `$) are
$`D`$ $`=`$ $`\rho _0W`$ (21)
$`S_j`$ $`=`$ $`\rho _0hW^2v_j`$ (22)
$`\tau `$ $`=`$ $`\rho _0hW^2pD`$ (23)
where $`W`$ is the relativistic Lorentz factor
$`W\stackrel{~}{\alpha }u^t={\displaystyle \frac{1}{\sqrt{1v^2}}}`$ (24)
with $`v^2=\gamma _{ij}v^iv^j`$. The 3-velocity components are obtained from the spatial components of the 4-velocity in the following way
$`v^i={\displaystyle \frac{u^i}{W}}+{\displaystyle \frac{\beta ^i}{\stackrel{~}{\alpha }}}.`$ (25)
Explicit expressions for the non-vanishing Christoffel symbols for metric (1), appearing in the source terms of the hydrodynamic equations, are presented in .
## 4 Numerical Methods
As stated before, our numerical integration of system (15) is based on Godunov-type methods (also known as HRSC schemes). In a HRSC scheme, the knowledge of the characteristic fields (eigenvalues) of the equations, together with the corresponding eigenvectors, allows for accurate integrations, by means of either exact or approximate Riemann solvers, along the fluid characteristics. These solvers, which constitute the kernel of our numerical algorithm, compute, at every interface of the numerical grid, the solution of local Riemann problems (i.e., the simplest initial value problem with discontinuous initial data). Hence, HRSC schemes automatically guarantee that physical discontinuities appearing in the solution, e.g., shock waves, are treated consistently (the shock-capturing property). HRSC schemes are also known for giving stable and sharp discrete shock profiles. They have also a high order of accuracy, typically second order or more, in smooth regions of the solution.
We perform the time update of system (15) according to the following conservative algorithm:
$`𝐮_{i,j}^{n+1}=𝐮_{i,j}^n`$ $``$ $`{\displaystyle \frac{\mathrm{\Delta }t}{\mathrm{\Delta }r}}(\widehat{𝐟}_{i+1/2,j}\widehat{𝐟}_{i1/2,j})`$ (26)
$``$ $`{\displaystyle \frac{\mathrm{\Delta }t}{\mathrm{\Delta }\theta }}(\widehat{𝐠}_{i,j+1/2}\widehat{𝐠}_{i,j1/2})+\mathrm{\Delta }t𝐬_{i,j}.`$
Index $`n`$ represents the time level and the time (space) discretization interval is indicated by $`\mathrm{\Delta }t`$ ($`\mathrm{\Delta }r,\mathrm{\Delta }\theta `$). The “hat” in the fluxes is used to denote the so-called numerical fluxes which, in a HRSC scheme, are computed according to some generic flux-formula, of the following functional form (suppressing index $`j`$):
$`\widehat{𝐟}_{i\pm \frac{1}{2}}={\displaystyle \frac{1}{2}}\left(𝐟(𝐮_{i\pm \frac{1}{2}}^L)+𝐟(𝐮_{i\pm \frac{1}{2}}^R){\displaystyle \underset{\alpha =1}{\overset{5}{}}}\stackrel{~}{\lambda }_\alpha \mathrm{\Delta }\stackrel{~}{\omega }_\alpha \stackrel{~}{r}_\alpha \right).`$ (27)
Notice that the numerical flux is computed at cell interfaces ($`i\pm 1/2`$). Indices $`L`$ and $`R`$ indicate the left and right sides of a given interface. Quantities $`\lambda `$, $`\mathrm{\Delta }\omega `$ and $`r`$ denote the eigenvalues, the jump of the characteristic variables and the eigenvectors, respectively, computed at the cell interfaces according to some suitable average of the state vector variables. Generic expressions for the characteristic speeds and eigenfields can be found in . Our code has the ability of using different approximate Riemann solvers: the Roe solver , widely employed in fluid dynamics simulations, with arithmetically averaged states and the Marquina solver , which has been extended to Relativity in . The computations presented here were obtained using Marquina’s scheme.
A technical remark: the equilibrium star is supplemented by a low-density uniform atmosphere, which is necessary for computing non-singular solutions of the hydrodynamic equations everywhere in the computational domain. After each time-step we reset the atmosphere’s density and pressure to their initial values, avoiding unwanted accretion of matter onto the star. The influence of the atmosphere is thus restricted to the surface grid-cells.
## 5 Pulsations of Non-rotating Stars
Since our code uses spherical polar coordinates, it can also be employed to study the evolution of non-rotating stars in 1-D. In the evolution of initially static non-rotating stars, we observe the following properties (note that our numerical grid is Eulerian):
1. Small-amplitude radial pulsations are triggered by the truncation errors of the finite-differencing scheme.
2. The radial pulsations are dominated by a set of discrete frequencies, which correspond to the normal modes of pulsation of the star.
3. The numerical viscosity of the finite-difference scheme damps the pulsations and the damping is stronger for the higher frequency modes.
4. The presence of a constant density atmosphere affects the finite differencing at the surface grid-cells, which increases the numerical damping of pulsations and also causes a continuous but very small drifting of the density distribution.
The initial amplitude of the radial pulsations and the small drift in density converge to zero at a second order rate with increasing resolution. The value of the density in the atmosphere region has a large effect on the damping of the pulsations. If it is too large, the damping is strong. To minimize this effect, we typically set the density of the atmosphere equal to $`10^6`$ times the density of the last grid point inside the star.
### 5.1 1-D evolutions
We study the numerical evolution of a nonrotating $`N=1.5`$ relativistic polytrope with $`M/R=0.056`$. The star is immediately set into radial pulsation, triggered by the finite difference truncation errors. The time-evolution of the radial velocity $`v_r`$, computed at 25% of the star radius using a grid of 400 zones, is shown in Fig. 1. The vertical axis is dimensionless ($`c=G=M_{}=1`$). The radial velocity is initially a very complex function of time. As we will show next, the pulsation consists mainly of a superposition of normal modes of oscillation of the fluid. The high frequency normal modes are damped quickly and after 20ms the star pulsates mostly in lowest frequency modes. Because these oscillations are caused only by the truncation errors, the magnitude of the radial velocity is extremely small. As shown in Fig. 1 it is only a few times larger than the non-zero residual velocity around which the star is oscillating. This residual velocity converges to zero as second order with increased resolution.
The small-amplitude radial pulsations in the non-linear, fixed spacetime evolutions correspond to linear normal modes of pulsation in the relativistic Cowling approximation, in which perturbations of the spacetime are ignored. A Fourier transform of the density or radial velocity time-evolution can be used to identify the normal mode frequencies. Fig. 2 shows the Fourier transform of the radial velocity evolution shown in Fig. 1. The normal mode frequencies stand out as sharp peaks on a continuous background. The width of the peaks increases with frequency. The frequencies of radial pulsations identified from Fig. 2 are shown in Table 1.
To compare the obtained frequencies to linear normal mode frequencies, we use a different code that solves the linearized relativistic pulsation equations for the stellar fluid, in the Cowling approximation , as an eigenvalue problem. In Table 1 we present the results of this comparison. The typical agreement between frequencies computed by the two methods is better than 0.5% for the fundamental $`F`$-mode and the lowest frequency harmonics $`H_1H_4`$ and better than 0.8% for the the higher harmonics $`H_5H_9`$. This is a strong test for the accuracy of the evolution code and our results can be used as a testbed computation for other relativistic multi-dimensional evolution codes.
### 5.2 2-D evolutions
In a similar way, small-amplitude non-radial pulsations can be studied with the present evolution code and the obtained frequencies can be compared to perturbation results. We find that the truncation errors of the finite difference scheme do not excite non-radial pulsations to a sufficiently large amplitude compared to the amplitude of radial pulsations, so that one cannot identify them accurately in a Fourier transform. Instead, one has to perturb the initial configuration, using an appropriate eigenfunction for each nonradial angular index $`l`$. Such a perturbation can be constructed using the eigenfunctions of linear pulsation modes, computed with the perturbation code in the Cowling approximation. The frequencies of the non-radial modes are then found from a Fourier transform of the time-evolution of the velocity component $`v_\theta `$.
Table 2 shows a similar comparison as in Table 1 for the quadrupole ($`l=2`$) pulsations of the same $`N=1.5`$ relativistic polytrope. Since the non-radial modes have to be computed on a 2-D grid, we cannot use resolutions as high as in the 1-D computations. For a small grid-size of $`80\times 80`$ zones and a total evolution time of 6ms, the agreement between frequencies computed by the two methods is better than 1.4% for the fundamental $`f`$-mode and the $`p`$-modes $`p_1p_6`$. For this grid-size, frequencies higher than the $`p_6`$ mode could not be computed accurately, because the grid is to coarse to resolve their eigenfunctions (higher harmonic eigenfunctions have a larger number of nodes in the radial direction).
## 6 Rotating Stars
We now turn to the evolution of initially stationary, uniformly rotating neutron stars. In these evolutions, we observe the same qualitative properties as for non-rotating stars (section 5) and an additional important property: the angular momentum of the star is not conserved at the surface layer. This is due to the fact that the velocity component $`v_\varphi `$ of the fluid has a maximum at the surface, while the numerical scheme (although second-order accurate in smooth regions of the solution) is only first-order accurate at local extrema. Moreover, the code evolves the relativistic momenta, $`S_i`$, and the velocity components (as well as the rest of “primitive” variables) must be recovered through a root finding procedure which involves dividing by the density. At the surface of the star (where the density is very small) this contributes to obtaining less than second-order accuracy.
A representative example of the evolution of a rotating star is presented in Fig. 3, which shows the evolution, at different times, of the velocity component $`v_\varphi `$. The star is again a $`N=1.5`$ polytrope with the same central density as the non-rotating star presented in section 5 and rotating at 74% of the mass-shedding limit at same central density. The evolution was for one rotation period on a $`96\times 60`$ grid. The vertical axis is dimensionless ($`c=G=M_{}=1`$). The figure shows that the $`\varphi `$-velocity in the interior of the star remains close to its initial value, while it decreases as a function of time in the outer layers. We find that this is a generic property of the present numerical scheme for any rotation rate and for any grid-size. By comparing evolutions with different grid sizes, we verified that the loss of angular momentum at the surface improves as first-order with resolution, while the evolution of the $`\varphi `$-velocity in the interior is second-order accurate. However, as the evolution proceeds in time, the first-order surface effect gradually affects the interior of the star.
## 7 Quasi-radial Modes of Rotating Stars
As a first application of our code, we compute quasi-radial modes (i.e. modes that in the non-rotating limit reduce to radial modes) of rapidly rotating relativistic stars in the Cowling approximation. Previously, these modes have been computed for fully relativistic stars only in the slow-rotation limit (but without the assumption of a fixed spacetime) by Hartle & Friedman (see also ). We compute the three lowest-frequency quasi-radial modes for a sequence of rotating stars of same central density. The non-rotating member of the sequence is the non-rotating star of section 5. Table 3 and Fig. 4 show our results for a low resolution grid of $`100\times 80`$ zones (note that our computational grid assumes equatorial plane symmetry). For this resolution we estimate the accuracy of the frequencies to be of the order of $`12`$%.
For the sequence of stars considered here, the frequencies of the quasi-radial modes decrease with increasing rotation rate. This agrees with previous slow-rotation computations which predict a decrease as $`\mathrm{\Omega }^2`$, where $`\mathrm{\Omega }`$ is the angular velocity of the star. For fast rotation, the change in the frequencies of quasi-radial modes is affected by higher order terms in $`\mathrm{\Omega }`$, because of the large deformation of the equilibrium star. Also, for rapidly rotating stars the quasi-radial mode frequencies are more “closely packed” than in non-rotating stars.
## 8 Discussion
Our axisymmetric relativistic hydrodynamical code is capable to evolve rapidly rotating stars in a fixed spacetime. We find that, for non-rotating stars, small amplitude oscillations have frequencies that agree with linear normal mode frequencies in the Cowling approximation and we compute the quasi-radial modes of rapidly rotating stars.
Modern HRSC numerical schemes (as the ones used in our code), satisfying the “total variation diminishing” (TVD) property , are second-order accurate in smooth regions of the flow, but only first-order accurate at local extrema. In our rotating stars runs we find that this results in a loss of angular momentum of the surface layers of the star, which gradually also affects the interior of the star. This angular momentum loss only vanishes as first-order with incresing resolution and we thus conclude that for accurate long-term evolutions of rotating neutron stars it is essential to use rather fine grids. Furthermore, to reduce the computational cost, one could use surface-adapted coordinates or fixed-mesh refinement. It would also be interesting to see whether the loss of angular momentum per rotation period will be significantly smaller in a frame co-rotating with the star. An alternative solution to this problem, which we plan to investigate, could be the use of “essentially non-oscillatory” (ENO) schemes, which maintain high-order of accuracy even at local extrema .
All previous considerations are important for the study of the non-linear dynamics of unstable toroidal oscillations ($`r`$-modes) in 3-D, which have a long growth time and thus require highly accurate long-term evolutions.
## Acknowledgements
We thank John L. Friedman, Curt Cutler, Philippos Papadopoulos and Tom Goodale for helpful discussions. We also thank S. Yoshida for sending us for comparison unpublished results on quasi-radial modes of rotating stars in the Cowling approximation, computed with a linear perturbation code. J.A.F acknowledges financial support from a TMR grant from the European Union (contract nr. ERBFMBICT971902). K.D.K. is grateful to the Max-Planck-Institut für Gravitationsphysik (Albert-Einstein-Institut), Potsdam, for generous hospitality.
## References
|
no-problem/9904/hep-ph9904316.html
|
ar5iv
|
text
|
# Four-neutrino spectrum from oscillation data
## Abstract
It is shown that the Super-Kamiokande atmospheric up–down asymmetry, together with the results of all other neutrino oscillation experiments, allows to constraint the possible spectra of four massive neutrinos. The two schemes with two pairs of neutrinos with close masses separated by a gap of about 1 eV are favored by the data.
preprint: UWThPh-1999-24 DFTT 22/99 hep-ph/9904316
The recent observation of an up–down asymmetry of high-energy $`\mu `$-like events generated by atmospheric neutrinos in the Super-Kamiokande experiment represents a convincing model-independent evidence in favor of neutrino oscillations. Indications in favor of disappearance of atmospheric $`\nu _\mu `$’s have been obtained also in the Kamiokande and IMB experiments and in the recent Soudan 2 and MACRO experiments. Other indications in favor of neutrino oscillations have been obtained in solar neutrino experiments (Homestake, Kamiokande, GALLEX, SAGE, Super-Kamiokande) and in the LSND experiment. The flux of electron neutrinos measured in all five solar neutrino experiments is substantially smaller than the one predicted by the Standard Solar Model and a comparison of the data of different experiments indicate an energy dependence of the solar $`\nu _e`$ suppression, which represents a rather convincing evidence in favor of neutrino oscillations. The accelerator LSND experiment is the only one that claims the observation of neutrino oscillations in appearance channels, specifically $`\overline{\nu }_\mu \overline{\nu }_e`$ and $`\nu _\mu \nu _e`$.
The probabilities of neutrino oscillations depend on the elements of the neutrino mixing matrix $`U`$, that connects the flavor neutrino fields $`\nu _{\alpha L}`$ to the massive neutrino fields $`\nu _{kL}`$ through the relation $`\nu _{\alpha L}=_kU_{\alpha k}\nu _{kL}`$, and on the phases $`\mathrm{\Delta }m_{kj}^2L/E`$, where $`\mathrm{\Delta }m_{kj}^2m_k^2m_j^2`$ ($`m_k`$ is the mass of the neutrino field $`\nu _k`$), $`L`$ is the source-detector distance and $`E`$ is the neutrino energy. If $`\mathrm{\Delta }m_{kj}^2L/E1`$ neutrino flavor transitions cannot be observed and if $`\mathrm{\Delta }m_{kj}^2L/E1`$ only the averaged transition probability can be measured. Since a variation of the transition probability as a function of neutrino energy has been observed in all the experiments mentioned above and the range of $`L/E`$ probed by each type of experiment is different ($`L/E10^{10}\mathrm{eV}^2`$ for solar neutrino experiments, $`L/E10^210^3\mathrm{eV}^2`$ for atmospheric neutrino experiments and $`L/E1\mathrm{eV}^2`$ for the LSND experiment), it is clear that in order to explain all the observations with neutrino oscillations at least three $`\mathrm{\Delta }m^2`$’s with different scales are needed: $`\mathrm{\Delta }m_{\mathrm{sun}}^210^{10}\mathrm{eV}^2`$, $`\mathrm{\Delta }m_{\mathrm{atm}}^210^310^2\mathrm{eV}^2`$, $`\mathrm{\Delta }m_{\mathrm{LSND}}^21\mathrm{eV}^2`$ (if the MSW effect is responsible of solar neutrino transitions, $`\mathrm{\Delta }m_{\mathrm{sun}}^2`$ must be smaller than about $`10^4\mathrm{eV}^2`$ in order to have a resonance in the interior of the sun and is still at least one order of magnitude smaller than $`\mathrm{\Delta }m_{\mathrm{atm}}^2`$). This means that at least four light massive neutrinos must exist in nature. Here we consider the minimal possibility of four neutrinos, which implies that the three flavor neutrinos $`\nu _e`$, $`\nu _\mu `$, $`\nu _\tau `$ are accompanied by a sterile neutrino $`\nu _s`$ that does not take part in standard weak interactions.
The six types of four-neutrino mass spectra that can accommodate the hierarchy $`\mathrm{\Delta }m_{\mathrm{sun}}^2\mathrm{\Delta }m_{\mathrm{atm}}^2\mathrm{\Delta }m_{\mathrm{LSND}}^2`$ are shown in Fig. 1. In all these mass spectra there are two groups of close masses separated by the “LSND gap” of the order of 1 eV. In each scheme the smallest mass-squared difference corresponds to $`\mathrm{\Delta }m_{\mathrm{sun}}^2`$ ($`\mathrm{\Delta }m_{21}^2`$ in schemes I and B, $`\mathrm{\Delta }m_{32}^2`$ in schemes II and IV, $`\mathrm{\Delta }m_{43}^2`$ in schemes III and A), the intermediate one to $`\mathrm{\Delta }m_{\mathrm{atm}}^2`$ ($`\mathrm{\Delta }m_{31}^2`$ in schemes I and II, $`\mathrm{\Delta }m_{42}^2`$ in schemes III and IV, $`\mathrm{\Delta }m_{21}^2`$ in scheme A, $`\mathrm{\Delta }m_{43}^2`$ in scheme B) and the largest mass squared difference $`\mathrm{\Delta }m_{41}^2=\mathrm{\Delta }m_{\mathrm{LSND}}^2`$ is relevant for the oscillations observed in the LSND experiment.
It has been shown that the schemes I–IV are disfavored by the results of short-baseline accelerator and reactor disappearance neutrino oscillation experiments for all values of $`\mathrm{\Delta }m_{41}^2`$ in the LSND-allowed range 0.2 – 2 eV<sup>2</sup>, with the possible exception of the small interval from 0.2 to 0.3 eV<sup>2</sup> where there are no data from $`\nu _\mu `$ short-baseline disappearance experiments. Here we will show that this gap is closed by the inclusion in the analysis of the asymmetry of $`\mu `$-like high-energy events
$$𝒜=(UD)/(U+D)=0.311\pm 0.043\pm 0.01$$
(1)
measured in the Super-Kamiokande experiment. Here $`U`$ and $`D`$ are the number of events in the zenith angle intervals $`1<\mathrm{cos}\theta <0.2`$ and $`0.2<\mathrm{cos}\theta <1`$, respectively.
In the following we will consider only the scheme I with a mass hierarchy, but the results apply also to the schemes II, III and IV. Let us remark that in principle one could check which scheme is allowed by doing a combined fit of all data. However, at the moment it is not possible to perform such a fit because of the enormous complications due to the presence of many parameters (six mixing angles, etc.) and to the difficulties involved in a combined fit of the data of different experiments, which are usually analyzed by the experimental collaborations using different methods. Hence, we think that it is quite remarkable that one can exclude the schemes I–IV with the following relatively simple procedure.
The exclusion plots obtained in short-baseline $`\overline{\nu }_e`$ and $`\nu _\mu `$ disappearance experiments imply that $`|U_{\alpha 4}|^2a_\alpha ^0\text{or}|U_{\alpha 4}|^21a_\alpha ^0`$ for $`\alpha =e,\mu `$, with $`a_e^04\times 10^2`$ for $`\mathrm{\Delta }m_{41}^20.1\mathrm{eV}^2`$ and $`a_\mu ^00.2`$ for $`\mathrm{\Delta }m_{41}^20.4\mathrm{eV}^2`$. However, since the survival probability of solar $`\nu _e`$’s is bounded by $`P_{\nu _e\nu _e}^{\mathrm{sun}}|U_{e4}|^4`$, only the range
$$|U_{e4}|^2a_e^0$$
(2)
is acceptable. In a similar way, since the survival probability of atmospheric $`\nu _\mu `$’s and $`\overline{\nu }_\mu `$’s (all the following inequalities are valid both for neutrinos and antineutrinos) is bounded by $`P_{\nu _\mu \nu _\mu }^{\mathrm{atm}}|U_{\mu 4}|^4`$, it is clear that large values of $`|U_{\mu 4}|^2`$ are incompatible with the observed asymmetry (1).
Let us derive the upper bound for $`|U_{\mu 4}|^2`$ that follows from the asymmetry (1). Because of the small value of $`\mathrm{\Delta }m_{\mathrm{atm}}^2=\mathrm{\Delta }m_{31}^2`$, downward-going neutrinos do not oscillate with the atmospheric mass-squared difference and the survival probability of downward-going neutrinos given by
$$P_{\nu _\alpha \nu _\alpha }^D=|U_{\alpha 4}|^4+\left(1|U_{\alpha 4}|^2\right)^2.$$
(3)
The conservation of probability and Eq.(2) allow to deduce the upper bound
$$P_{\nu _e\nu _\mu }^D1P_{\nu _e\nu _e}^D=2|U_{e4}|^2\left(1|U_{e4}|^2\right)2a_e^0(1a_e^0).$$
(4)
From Eqs.(3) and (4) we obtain the upper bound
$$DN_\mu \left[|U_{\mu 4}|^4+\left(1|U_{\mu 4}|^2\right)^2\right]+2N_ea_e^0(1a_e^0),$$
(5)
where $`N_\mu `$ and $`N_e`$ are number of muon (electron) neutrinos and antineutrinos produced in the atmosphere. On the other hand, taking into account only the part of $`D`$ which is determined by the survival probability of $`\nu _\mu `$’s, we obtain the lower bound
$$DN_\mu \left[|U_{\mu 4}|^4+\left(1|U_{\mu 4}|^2\right)^2\right].$$
(6)
Furthermore, using the lower bound $`P_{\nu _\mu \nu _\mu }^{\mathrm{atm}}|U_{\mu 4}|^4`$, for upward-going neutrinos we have
$$UN_\mu |U_{\mu 4}|^4.$$
(7)
With the inequalities (5), (6) and (7), for the asymmetry (1) we obtain
$$𝒜\frac{\left(1|U_{\mu 4}|^2\right)^2+2a_e^0(1a_e^0)/r}{\left(1|U_{\mu 4}|^2\right)^2+2|U_{\mu 4}|^4},$$
(8)
where $`rN_\mu /N_e2.8`$. Solving the inequality (8) for $`|U_{\mu 4}|^2`$, we finally obtain the upper bound
$$|U_{\mu 4}|^2\frac{1+𝒜\sqrt{2\left[𝒜(1+𝒜)+(1+3𝒜)a_e^0(1a_e^0)/r\right]}}{1+3𝒜}a_\mu ^{\mathrm{SK}}.$$
(9)
Since the measured value (1) of $`𝒜`$ implies that $`𝒜0.254`$ at 90% CL, from the inequality (8) we obtain the upper bound depicted by the horizontal line in Fig. 2 (the vertically hatched area is excluded).
In Fig. 2 we have also shown the bound $`|U_{\mu 4}|^2a_\mu ^0`$ or $`|U_{\mu 4}|^21a_\mu ^0`$ obtained from the exclusion plot of the short-baseline CDHS $`\nu _\mu `$ disappearance experiment, which exclude the shadowed region.
The results of the LSND experiment imply a lower bound $`A_{\mu ;e}^{\mathrm{min}}`$ for the amplitude $`A_{\mu ;e}=4|U_{e4}|^2|U_{\mu 4}|^2`$ of $`\nu _\mu \nu _e`$ oscillations, from which we obtain the constraint
$$|U_{\mu 4}|^2A_{\mu ;e}^{\mathrm{min}}/4a_e^0.$$
(10)
This bound is represented by the curve in Fig. 2 labelled LSND + Bugey (the diagonally hatched area is excluded).
From Fig. 2 one can see that, in the framework of scheme I, there is no range of $`|U_{\mu 4}|^2`$ that is compatible with all the experimental data. Hence, the scheme with four neutrinos and a mass hierarchy is strongly disfavored.
The incompatibility of the experimental results with the mass spectrum I is shown also in Fig. 3, where we have plotted in the $`A_{\mu ;e}`$$`\mathrm{\Delta }m_{41}^2`$ plane the upper bound $`A_{\mu ;e}4a_e^0a_\mu ^0`$ for $`\mathrm{\Delta }m_{41}^2>0.26\mathrm{eV}^2`$ and $`A_{\mu ;e}4a_e^0a_\mu ^{\mathrm{SK}}`$ for $`\mathrm{\Delta }m_{41}^2<0.26\mathrm{eV}^2`$ (solid line, the region on the right is excluded). One can see that this constraint is incompatible with the LSND-allowed region (shadowed area).
The procedure presented above for the scheme I applies also to the schemes II, III and IV, in which there is a group of three close neutrino masses separated from the fourth mass by the LSND gap. Hence, we conclude that these schemes are disfavored. Only the four-neutrino schemes A and B in Fig. 1 are compatible with the results of all neutrino oscillation experiments.
|
no-problem/9904/astro-ph9904178.html
|
ar5iv
|
text
|
# The MAGIC Telescope - Prospects for GRB research
## 1 Introduction
The Major Atmospheric Gamma-ray Imaging Cherenkov (MAGIC) Telescope will be a large imaging air Cherenkov telescope for ground-based $`\gamma `$-ray observations above 10 GeV. Details of its design, the scientific motivation, the feasibility of the project, and other issues have been described elsewhere (Barrio et al. barrio (1998)). The telescope is expected to become operational by the middle of 2001.
The telescope is optimised to achieve the lowest energy threshold and highest flux sensitivity achievable with present technology. This makes it applicable to a large range of astrophysical research fields:
* Blazars (study of EGRET blazars, possible discovery of additional sources)
* Cosmology (measurements of the near-infrared background via $`\gamma `$-$`\gamma `$ absorption)
* Investigation of the pulsed $`\gamma `$-ray emission from pulsars
* Search for gamma-emission from Supernova remnants (and hence search for evidence of production and acceleration of cosmic rays)
* Search for decay/annihilation line-emission from WIMPs clustering at the galactic centre
* Identification of “unidentified EGRET sources” (position accuracy $``$ 1.2’)
* Search for high-energy counterparts of gamma-ray bursts
The last part of this scientific program will briefly be discussed in this article.
## 2 Fast follow-up observations
In order to make fast follow-up observations of GRBs possible, the MAGIC Telescope will have an interface (probably a socket connection) to the GRB Coordinate Network GCN (Barthelmy et al. barthelmy (1994) and these proceedings) and a secondary ground station for direct communication with HETE II (Ricker et al., these proceedings). HETE II is a dedicated GRB research satellite which is expected to be launched in 2000. It will provide approx. 30 burst positions per year with accuracies better than 10 arcmin. These notifications are expected to arrive with a delay of less than 5 s. The MAGIC Telescope is specially designed to have low inertia such that the telescope can be positioned on any point in the sky within less than 30 s.
In case of a notification, a fast check of the observability of the GRB location will be performed. If the decision is positive, the present observations will be stopped immediately and a special fast drive will position the telescope on the GRB which will then be observed for the remaining night-time and probably also the following night in order to detect possible delayed emission.
Taking into account a decision time of 5 s at our site, we expect the reaction time between the actual start of the burst and the start of the follow-up observation to be $`30\pm 10`$ s. The MAGIC Telescope will thus be able to perform observations of non-delayed emission of all bursts with durations above 30 s. This is the position of the rightmost peak in the BATSE burst duration ($`T_{90}`$) distribution (see e.g. Kouveliotou et al. kou (1995)) and corresponds to a fraction $``$ 33 % of all bursts which trigger BATSE on the 64 ms time scale.
The telescope will be built on the Canary Islands (Tenerife or La Palma). Cherenkov telescopes can only observe during the night. Observations will be possible up to zenith angles of $`80^{}`$, i.e. about 40 % of the total sky ($`4\pi `$) will be accessible. Assuming 30 % of the nights to have bad weather and taking into account that the presence of the moon can prevent observations of certain positions, we arrive at a duty cycle of $``$ 10 %.
The effective field of view of the photo-sensor camera of the MAGIC Telescope is 1.6 in diameter. We will therefore be able to safely observe any of the positions provided by HETE II and also some from the GCN and expect $``$ 5 serious immediate follow-up observations per year. For delayed emission (time-window of the order of an hour up to several days) this number will be larger.
## 3 Expected performance
The expected performance in terms of sensitivity is summarised in Figure 1. This is the performance we expect for the second phase of the project in which high quantum efficiency hybrid photo sensors will replace the photo-multipliers of the telescope’s camera. In the first phase our threshold will be $``$ 30 GeV, in the second phase 10 GeV. It is not yet clear how soon the second phase will follow the first. The sensitivity at energies $`>`$ 30 GeV is to a good approximation independent of these changes.
Given the diverse shape of GRB light-curves it is difficult to predict an average counting rate for gamma-rays in successful burst observations. We note instead that the MAGIC Telescope will have an effective collection area for primary gamma photons of $`10^8`$ cm<sup>2</sup> at the threshold rising to $`10^9`$ cm<sup>2</sup> at 100 GeV. For a hypothetical counting rate of 0.1 Hz for EGRET above 100 MeV, we expect (assuming a spectral index of 2.0) a counting rate of $``$ 100 Hz above 10 GeV. A rate of $``$ 6 Hz for 30 s will suffice for a 5 $`\sigma `$ detection with a moderate background rejection applied. For strong bursts such as GRB930131, gamma counting rates of the order of 1 kHz may occur. The data acquisition will therefore be prepared to sustain such rates without additional dead-time.
## 4 Conclusions
The MAGIC Telescope will be able to make a major contribution to GRB research by providing high sensitivity measurements in the (for GRBs) essentially unexplored regime above 10 GeV. The high counting rates in strong bursts will also permit to study the shape of the light-curve in more detail than previously possible (time resolutions of the order of 1 s). Search for small delays with respect to the low energy emission is thus possible. As an interesting side-result this may provide one of the best possible lower limits to the quantum gravity energy scale as pointed out by Amelino-Camelia et al. (amelino (1998)).
|
no-problem/9904/astro-ph9904334.html
|
ar5iv
|
text
|
# Untitled Document
A Java Calculator of Standard Big Bang Nucleosynthesis
Luis Mendoza and Craig J. Hogan, University of Washington
Abstract
A simple Java applet is presented which allows quick and easy computation and plotting of the predictions of light element abundances in SBBN, including errors in the predictions propagated from input reaction rates correlated and calibrated via Monte Carlo. The applet, which requires Java 1.1-compatible browsers, can be found at
http://www.astro.washington.edu/research/bbn/
Standard Big Bang Nucleosynthesis refers to an exceptionally simple and beautiful model of the early universe. Matter and radiation are spatially uniform and the expansion is controlled by known fields with negligible chemical potentials, so the dynamics of the expansion and density as a function of time and temperature is determined by “just physics” with no external parameters. The production of light elements depends on one parameter, the small baryon excess characterised by the baryon-to-photon ratio $`\eta `$. Numerical integration of the reaction network yields predictions for light element abundances with “theoretical errors” dominated by the experimental errors in the input physics such as nuclear reaction rates. The error matrix for SBBN has been computed by Fiorentini et al. (Phys. Rev. D58,1998; astro-ph/9803177) by Monte Carlo techniques.
For many years SBBN theory and observation were mainly compared by the theorists. Integrations using updated rates and techniques were used to make state-of-the-art predictions and these were then compared with a collection of disparate observational data. Many reviews of this kind successively refined the predictions and the arguments concerning the post-big-bang processing of the elements (see reviews by S. Sarkar, Rep. Prog. Phys. 59, 1493, 1996, and D. Schramm and M. S. Turner, Rev. Mod. Phys. (Colloquia) 70, 303, 1998). The theoretical errors are no longer negligible compared with the observational errors so correct statistical comparison requires handling all of the errors at the same time, forcing the theorists to interpret observers’ error budgets or vice versa. With theoretical predictions and errors now stable, it makes sense to empower the observers to do the comparison.
This applet provides easy access to the precise predictions of SBBN including calibrated errors, as described by Fiorentini et al. (1998). This version (1.1) includes the predictions and errors for deuterium, helium-3, helium-4, and lithium-7 abundances. It will be useful in using data on each abundance (with errors) to set statistical limits on $`\eta `$, or to make predictions for abundances from other estimates of $`\eta `$ (e.g., M. Fukugita, C. J. Hogan and P. J. E. Peebles, ApJ 503, 518, 1998). Future versions will allow statistical concordance tests of SBBN and estimates of $`\eta `$ using more than one abundance datum, as well as allowing changes in the input reaction rates and errors.
|
no-problem/9904/math9904125.html
|
ar5iv
|
text
|
# On endomorphisms of surface mapping class groups
## 1. Introduction
Let $`S`$ be a compact connected orientable surface. The mapping class group $`_S`$ of the surface $`S`$ is the group of isotopy classes of orientation preserving diffeomorphisms $`SS`$. The extended mapping class group $`_S^{}`$ of $`S`$ is the group of isotopy classes of all diffeomorphisms $`SS`$. Note that the isotopy classes of orientation reversing diffeomorphisms are also included in $`_S^{}`$, and hence $`_S`$ is a subgroup of $`_S^{}`$ of index two.
Recall that a group $`G`$ is called residually finite if for each $`x1`$ in $`G`$ there exists a homomorphism $`f`$ from $`G`$ onto some finite group such that $`f(x)`$ is nontrivial. Equivalently, there is some finite index normal subgroup of $`G`$ that does not contain $`x`$. $`G`$ is called hopfian if every surjective endomorphism of $`G`$ is an automorphism. It is well known that finitely generated residually finite groups are hopfian \[LS\]. $`G`$ is called cohopfian if every injective endomorphism of $`G`$ is an automorphism.
The mapping class group of an orientable surface is finitely generated \[L, B\] and residually finite \[G, I1\]. Hence it is hopfian. N.V. Ivanov and J.D. McCarthy \[IM\] proved that $`_S`$ is also cohopfian. Author and J.D. McCarthy \[KM\] proved that if $`\varphi :_S_S`$ is a homomorphism such that $`\varphi (_S)`$ is a normal subgroup and $`_S/\varphi (_S)`$ is abelian, then $`\varphi `$ is an automorphism.
In this paper, we prove further that if $`\varphi `$ is an endomorphism of the mapping class group $`_S`$ onto a finite index subgroup, then $`\varphi `$ is in fact an automorphism, with a few exceptions. The proof of this result relies on a result of R. Hirshon, which states that if $`\varphi `$ is an endomorphism of a finitely generated residually finite group $`G`$ such that $`\varphi (G)`$ is of finite index in $`G`$, then $`\varphi `$ restricted to $`\varphi ^n(G)`$ is an injection for some $`n`$.
D.T. Wise \[W\] gave an example of a finitely generated residually finite group $`G`$ and an endomorphism $`\mathrm{\Phi }`$ of $`G`$ such that the restriction of $`\mathrm{\Phi }`$ to $`\mathrm{\Phi }^n(G)`$ is not injective for any $`n`$, answering a question of R. Hirshon in negative. It might be interesting to consider the same question for mapping class groups of surfaces.
## 2. Endomorphisms of mapping class groups
Let $`S`$ be a compact connected oriented surface of genus $`g`$ with $`b`$ boundary components. For any simple closed curve $`a`$ on $`S`$, there is a well known diffeomorphism, called a right Dehn twist, supported in a regular neighborhood of $`a`$. We denote by $`t_a`$ the isotopy class of a right Dehn twist about $`a`$, also called a Dehn twist. Note that $`ft_af^1=t_{f(a)}`$ for any orientation preserving mapping class $`f`$.
The pure mapping class group $`𝒫_S`$ is the subgroup of $`_S`$ consisting of those orientation preserving mapping classes which preserve each boundary component.
For a group $`G`$ and a subgroup $`H`$ of it, we denote by $`C_G(H)`$ the centralizer of $`H`$ in $`G`$. The center of $`G`$ is denoted by $`C(G)`$.
###### Theorem 1.
Let $`G`$ be a finitely generated residually finite group, and let $`\varphi `$ be an endomorphism of $`G`$ onto a finite index subgroup. Then there exists an $`n`$ such that the restriction of $`\varphi `$ to $`\varphi ^n(G)`$ is an injection.
###### Theorem 2.
Let $`S`$ be a compact connected orientable surface of genus $`g`$ with $`b`$ boundary components. Suppose, in addition, that if $`g=0`$ then $`b5`$, if $`g=1`$ then $`b3`$, and if $`g=2`$ then $`b1`$. Then any isomorphism between two finite index subgroups of the extended mapping class group $`_S^{}`$ is the restriction of an inner automorphism of $`_S^{}`$.
Theorem 1 was proved by R. Hirshon (\[H\]), and Theorem 2 was proved by N.V. Ivanov \[I2\] for surfaces of genus at least two and by the author \[K\] for the remaining cases. Since the mapping class group $`_S`$ is normal in $`_S^{}`$, we deduce the following theorem.
###### Theorem 3.
Let $`S`$ be a compact connected orientable surface of genus $`g`$ with $`b`$ boundary components. Suppose, in addition, that if $`g=0`$ then $`b5`$, if $`g=1`$ then $`b3`$, and if $`g=2`$ then $`b1`$. Then any isomorphism between two finite index subgroups of the mapping class group $`_S`$ is the restriction of an automorphism of $`_S`$.
###### Lemma 4.
Let $`S`$ be a closed orientable surface of genus two and let $`\mathrm{\Gamma }`$ be a finite index subgroup of $`_S`$. Then the center $`C(\mathrm{\Gamma })`$ of $`\mathrm{\Gamma }`$ is equal to $`\mathrm{\Gamma }\sigma `$, where $`\sigma `$ is the hyperelliptic involution.
Proof: Since the subgroup $`\sigma =\{1,\sigma \}`$ is the center of $`_S`$, its intersection with $`\mathrm{\Gamma }`$ is contained in the center of $`\mathrm{\Gamma }`$.
Now let $`fC(\mathrm{\Gamma })`$ and let $`N`$ be the index of $`\mathrm{\Gamma }`$ in $`_S`$. Since $`t_a^N\mathrm{\Gamma }`$ for all simple closed curves $`a`$, we have $`t_{f(a)}^N=ft_a^Nf^1=t_a^N`$. It follows that $`f(a)=a`$ (cf. \[IM\]). Hence, $`ft_af^1=t_{f(a)}=t_a`$. Since $`_S`$ is generated by Dehn twists, $`fC(_S)=\sigma `$. $`\mathrm{}`$
We are now ready to state and prove the main result of this paper.
###### Theorem 5.
Let $`S`$ be a compact connected orientable surface of genus $`g`$ with $`b`$ boundary components. Suppose, in addition, that if $`g=0`$ then $`b2,3,4`$, and if $`g=1`$ then $`b2`$. If $`\varphi `$ is an endomorphism of $`_S`$ such that $`\varphi (_S)`$ is of finite index in $`_S`$, then $`\varphi `$ is an automorphism.
Proof: If $`S`$ is a (closed) sphere or a disk, then $`_S`$ is trivial. Clearly, the conclusion of the theorem holds.
Suppose first that $`S`$ is a torus with $`b1`$ boundary component. It is well known that $`_S`$ is isomorphic to $`SL_2(𝐙)`$. The commutator subgroup of $`SL_2(𝐙)`$ is a nonabelian free group of rank $`2`$ and its index in $`SL_2(𝐙)`$ is $`12`$. Let us denote it by $`F_2`$. $`\varphi (F_2)`$ is contained in $`F_2`$ as a finite index subgroup. If this index is $`k`$, $`\varphi (F_2)`$ is a free group of rank $`k+1`$. Since there is no homomorphism from $`F_2`$ onto a free group of rank $`3`$, it follows that $`k=1`$. That is, $`\varphi (F_2)=F_2`$. In particular, $`\varphi (SL_2(𝐙))`$ contains $`F_2`$. The fact that $`\varphi `$ is an automorphism in this case was proved in \[KM\].
Suppose now that $`S`$ is not one of the surface above and not a closed a surface of genus $`2`$. Let us orient $`S`$ arbitrarily. Since $`_S`$ is finitely generated and residually finite, there exists an $`n`$ such that the restriction of $`\varphi `$ to $`\varphi ^n(_S)`$ is an isomorphism onto $`\varphi ^{n+1}(_S)`$. Note that the subgroups $`\varphi ^n(_S)`$ and $`\varphi ^{n+1}(_S)`$ are of finite index in $`_S`$. Hence, there is an automorphisms $`\alpha `$ of $`_S`$ such that the restrictions of $`\alpha `$ and $`\varphi `$ to $`\varphi ^n(_S)`$ coincide.
Let $`N`$ be the index of $`\varphi ^n(_S)`$ in $`_S`$. For any simple closed curve $`a`$ on $`S`$, $`t_a^N`$ is contained in $`\varphi ^n(_S)`$. Hence, $`\alpha (t_a^N)=\varphi (t_a^N)`$. Let $`f_S`$ be any element. Then
$`t_{f(a)}^N`$ $`=`$ $`\alpha ^1(\varphi (t_{f(a)}^N))`$
$`=`$ $`\alpha ^1(\varphi (ft_a^Nf^1))`$
$`=`$ $`\alpha ^1(\varphi (f))\alpha ^1(\varphi (t_a^N))\alpha ^1(\varphi (f^1))`$
$`=`$ $`\alpha ^1(\varphi (f))t_a^N\alpha ^1(\varphi (f))^1`$
$`=`$ $`t_{\alpha ^1(\varphi (f))(a)}^N.`$
Hence, $`\alpha ^1(\varphi (f))(a)=f(a)`$ for all $`a`$ (cf. \[IM\]). It follows that $`f^1\alpha ^1(\varphi (f))`$ commutes with all Dehn twists. Since $`𝒫_S`$ is generated by Dehn twists, it is in $`C__S(𝒫_S)`$, the centralizer of $`𝒫_S`$ in $`_S`$. But $`C__S(𝒫_S)`$ is trivial \[IM\]. Hence, $`\alpha ^1(\varphi (f))=f`$. Therefore, $`\varphi =\alpha `$. In particular, $`\varphi `$ is an automorphism.
Suppose finally that $`S`$ is a closed surface of genus two. Let $`R`$ be a sphere with six holes. Then $`_R`$ is isomorphic to the quotient of $`_S`$ with its center $`\sigma `$, where $`\sigma `$ is the hyperelliptic involution (cf. \[BH\]) . Let us identify $`_R`$ and $`_s/\sigma `$, and let $`\pi :_S_R`$ be the quotient map. Since $`\varphi (\sigma )`$ is in the center of $`\varphi (_S)`$, either $`\varphi (\sigma )=\sigma `$ or $`\varphi (\sigma )=1`$ by the lemma above.
If $`\varphi (\sigma )=\sigma `$, then $`\varphi `$ induces an endomorphism $`\mathrm{\Phi }`$ of $`_R`$, such that $`\pi \varphi =\mathrm{\Phi }\pi `$. Then we have a diagram in which all squares are commutative:
$$\begin{array}{ccccccccc}1& & \sigma & & _S& \stackrel{\pi }{}& _R& & 1\\ & & \mathrm{I}& & \varphi & & \mathrm{\Phi }& & \\ 1& & \sigma & & _S& \stackrel{\pi }{}& _R& & 1\end{array}$$
where $`\mathrm{I}`$ is the identity homomorphism. Since the image $`\mathrm{\Phi }(_R)`$ of $`\mathrm{\Phi }`$ is of finite index, $`\mathrm{\Phi }`$ is an automorphism by the first part. By $`5`$-lemma, $`\varphi `$ is an automorphism.
If $`\varphi (\sigma )=1`$, then $`\varphi `$ induces a homomorphism $`\overline{\varphi }:_R_S`$ such that $`\overline{\varphi }\pi =\varphi `$. The image of the endomorphism $`\mathrm{\Phi }=\pi \overline{\varphi }`$ of $`_R`$ has finite index. Since $`R`$ is a sphere with six holes, $`\mathrm{\Phi }`$ is an automorphism by the first part. Then, $`\varphi `$ is an automorphism, and hence $`\sigma =1`$. This contradiction finishes the proof of our theorem. $`\mathrm{}`$
Remark: If $`S`$ is a sphere with two holes, then $`_S`$ is a group of order two, and if $`S`$ is a sphere with three holes, then $`_S`$ is isomorphic to the symmetric group on three letters. Hence, in these cases the trivial homomorphism is an endomorphism onto a finite index subgroup which is not automorphism. We do not know if the conclusion of Theorem 5 holds if a sphere with four holes and a torus with two holes.
Department of Mathematics, Middle East Technical University, 06531 Ankara, Turkey. E-mail: korkmaz`@`math.metu.edu.tr.
|
no-problem/9904/hep-th9904157.html
|
ar5iv
|
text
|
# The Complete Brane Solution in 𝐷-dimensional Coupled Gravity System
## Acknowledgments
We would like to thank Han-Ying Guo, Yi-hong Gao, Ke Wu, Ming Yu, Zhu-jun Zheng and Zhong-Yuan Zhu for discussions. This work is supported in part by funds from Chinese National Science Foundation and Pandeng Project.
|
no-problem/9904/cond-mat9904261.html
|
ar5iv
|
text
|
# Comment on the anisotropic impurity scattering in superconductors
## Abstract
We discuss a case of a strong anisotropic impurity scattering within the model introduced in our previous paper $`[`$Phys. Rev. B54, 15463 (1996)$`]`$ and clarify on our former statement about a possible enhancement of the critical temperature in this scattering regime. We show, that for the anisotropy of the impurity potential determined by the functions from the non identity irreducible representations of the crystal point group the critical temperature decreases with the impurity scattering rate which is consistent with a generally understood role of disorder in superconductors.
The anisotropy of the impurity potential has become an important issue in the interpretation of the critical temperature suppression data in the cuprates. Although still not confirmed by any experiments (a good test may be provided by the angle resolved density of states measurements below $`T_c`$) the scattering potentials assuming a momentum-dependent scattering probability can be very helpful in understanding the experimental data. In particular, it is possible to obtain within a phenomenological model which introduces anisotropy in the impurity scattering potential a good agreement with the electron irradiation data in $`YBa_2Cu_3O_{7\delta }`$ crystals. This model, however, has some interpretational difficulties. It has been mentioned that under certain conditions it could lead to an enhancement of $`T_c`$ which seems unphysical as superconductivity should be suppressed by an increasing amount of disorder in the system. In this letter we address this problem in more detail and show that the model really predicts a reduction of the critical temperature if the anisotropy of the impurity potential is represented by a certain class of functions. This class corresponds to the irreducible non identity representations of the crystal point group.
We discuss a superconducting state defined by the orbital part of the order parameter $`\mathrm{\Delta }\left(𝐤\right)=\mathrm{\Delta }e\left(𝐤\right)`$ where $`e\left(𝐤\right)`$ is a real basis function of a one-dimensional (1D) irreducible representation of an appropriate point group or a linear combination of such functions. For the sake of convenience $`e\left(𝐤\right)`$ is normalized as $`e^2=1`$, where $`<\mathrm{}>=_{FS}𝑑S_kn\left(𝐤\right)\left(\mathrm{}\right)`$ denotes the average value over the Fermi surface (FS), $`n\left(𝐤\right)`$ is the angle resolved FS density of states normalized to unity, i.e. $`_{FS}𝑑S_kn\left(𝐤\right)=1`$, and $`_{FS}𝑑S_k`$ represents the integration over the Fermi surface.
The impurity scattering potential appropriate for the second-order Born scattering limit is introduced as
$$|w\left(𝐤𝐤^{}\right)|^2=|w_0|^2+|w_1|^2f\left(𝐤\right)f\left(𝐤^{}\right)$$
(1)
with $`|w_0|`$ and $`|w_1|`$ representing isotropic and anisotropic scattering amplitude respectively. The momentum-dependent function $`f\left(𝐤\right)`$ determines the symmetry of the anisotropy of the impurity potential and is assumed to vanish after the integration over the Fermi surface, $`f=0`$. Therefore, the Fermi surface average of the scattering potential $`|w\left(𝐤𝐤^{}\right)|^2=|w_0|^2`$ is determined by the $`s`$-wave core ($`|w_0|`$) and the momentum-dependent part in Eq. (1) represents the deviations from the isotropic scattering. For the sake of simplicity we normalize $`f\left(𝐤\right)`$ according to $`f^2=1`$. Finally, the requirement of a non negative value of the squared potential in Eq. (1) yields $`|w_1|^2|w_0|^2`$ constraint on the potential amplitudes.
Following standard procedure the equation for the critical temperature of a superconductor with a separable pairing potential $`V(𝐤,𝐤^{})=V_0e\left(𝐤\right)e\left(𝐤^{}\right)`$ ($`V_0`$ is a positive potential amplitude) in the presence of anisotropic impurity scattering is obtained
$$\begin{array}{c}\mathrm{ln}\frac{T_c}{T_{c_0}}=\left(e^2+ef^21\right)\left[\psi \left(\frac{1}{2}+\frac{\mathrm{\Gamma }_0}{2\pi T_c}\right)\psi \left(\frac{1}{2}\right)\right]+\hfill \\ \\ ef^2\left[\psi \left(\frac{1}{2}\right)\psi \left(\frac{1}{2}+\frac{\mathrm{\Gamma }_0}{2\pi T_c}\left(1\frac{\mathrm{\Gamma }_1}{\mathrm{\Gamma }_0}\right)\right)\right]\hfill \end{array}$$
(2)
In writing above we have introduced the isotropic $`\mathrm{\Gamma }_0`$ and anisotropic $`\mathrm{\Gamma }_1`$ impurity scattering rates ($`\mathrm{\Gamma }_1\mathrm{\Gamma }_0`$)
$$\mathrm{\Gamma }_0=\pi N_0n_i|w_0|^2,\mathrm{\Gamma }_1=\pi N_0n_i|w_1|^2$$
(3)
where $`n_i`$ is impurity (defect) concentration. The model introduces two dimensionless parameters which characterize the anisotropy of the pair-breaking effect. First of them, $`ef^2=\left[_{FS}𝑑S_kn\left(𝐤\right)e\left(𝐤\right)f\left(𝐤\right)\right]^2`$, describes the interplay between the pair potential $`V(𝐤,𝐤^{})`$ and the anisotropic part of the scattering potential, $`|w\left(𝐤𝐤^{}\right)|^2`$ (Eq. (1)). This parameter is determined by the symmetry of the superconducting state, $`e\left(𝐤\right)`$, as well as that of the impurity scattering matrix element, $`f\left(𝐤\right)`$. According to the normalization of the order parameter, $`e^2=1`$, and the anisotropy function of the impurity potential, $`f^2=1`$, the parameter $`ef^2`$ takes values between 0 and 1. The second parameter in our model, ($`\mathrm{\Gamma }_1/\mathrm{\Gamma }_0`$), represents the amount of anisotropic scattering rate in impurity potential normalized by the isotropic scattering rate (Eq. (3)), and its value varies also from 0 to 1. Several limits of the above equation were examined in Ref. 1. Our analysis showed that the symmetry of the anisotropic potential is an important factor and a significant reduction in the pair-breaking strength appears for large values of $`ef^2`$ with an appropriate level of $`\mathrm{\Gamma }_1/\mathrm{\Gamma }_0`$.
Among the terms determining the critical temperature in Eq. (2) only one depends on $`\mathrm{\Gamma }_1/\mathrm{\Gamma }_0`$. Therefore, $`T_c`$ can be maximized with respect to this parameter by taking $`\mathrm{\Gamma }_1/\mathrm{\Gamma }_0`$ value maximizing the $`\mathrm{\Gamma }_1/\mathrm{\Gamma }_0`$-dependent term. This is achieved for $`\mathrm{\Gamma }_1/\mathrm{\Gamma }_0=1`$ that is, in the case of the strong anisotropic scattering. The critical temperature is then determined by
$$\mathrm{ln}\frac{T_c}{T_{c_0}}=\left(e^2+ef^21\right)\left(\psi \left(\frac{1}{2}+\frac{\mathrm{\Gamma }_0}{2\pi T_c}\right)\psi \left(\frac{1}{2}\right)\right)$$
(4)
It is easy to see that the critical temperature becomes very robust with respect to the impurity scattering rate $`\mathrm{\Gamma }_0`$, especially for a significant overlap between $`e\left(𝐤\right)`$ and $`f\left(𝐤\right)`$ functions, that is when $`ef^21`$. This case led to the earlier mentioned difficulty of the model. The critical temperature is determined now by a coefficient
$$\alpha =e^2+ef^21$$
(5)
As long as $`\alpha `$ is negative superconductivity is suppressed by the disorder. However, a positive value of $`\alpha `$ would mean an enhancement of the critical temperature due to impurity scattering. Below we show that the second option is impossible for a wide class of functions $`f\left(𝐤\right)`$ to which we postulate to restrict the model.
We assume that $`f\left(𝐤\right)`$ belongs to a non identity 1D representation of the crystal symmetry group or is given by a linear combination of such functions. It is worth observing, that the above choice of $`f\left(𝐤\right)`$ yields $`f=0`$ which is in agreement with the model. As the next step, we express the superconducting order parameter as a sum of a part belonging to the identity irreducible representation $`e_i\left(𝐤\right)`$ and $`e_n\left(𝐤\right)`$ given by a linear combination of functions transforming according to the non identity irreducible representations
$$e\left(𝐤\right)=e_i\left(𝐤\right)+e_n\left(𝐤\right)$$
(6)
In the case of a $`\left(d_{x^2y^2}+s\right)`$-wave superconductor for instance, $`e_n\left(𝐤\right)cos2\varphi `$ and $`e_i\left(𝐤\right)1`$. Using the property of orthogonality of different irreducible representations of a given symmetry group and taking into account that the FS average defines a scalar product in the space of functions defined on the Fermi sheet, we note that $`e^2=e_i^2`$ and $`ef^2=e_nf^2`$. Therefore, the coefficient $`\alpha `$ can be written as
$$\alpha =e_i^2+e_nf^21$$
(7)
Through the Schwarz inequality $`e_i^2e_i^2`$ we obtain an upper limit on $`\alpha `$
$$\alpha e_i^2+e_nf^21$$
(8)
It is worth noting, that the equality in the above relation holds only for $`e_i\left(𝐤\right)1`$ that is, for instance in the $`\left(d_{x^2y^2}+s\right)`$-wave or $`d_{x^2y^2}`$-wave superconducting state. A simple relation $`e^2=e_i^2+e_n^2`$ and the normalization condition $`e^2=1`$ yield $`e_i^2=1e_n^2`$ which leads to an equivalent to the relation (8) constraint on $`\alpha `$
$$\alpha e_nf^2e_n^2$$
(9)
Using a normalization $`f^2=1`$ we can rewrite the right-hand side of the inequality (9) as
$$e_nf^2e_n^2=e_nf^2e_n^2f^2$$
(10)
Finally, from the relations (9), (10) and by applying Schwarz inequality we get $`\alpha 0`$. Therefore, according to Eqs. (4) and (5) $`T_c`$ is a decreasing or a constant function of the impurity scattering rate $`\mathrm{\Gamma }_0`$. It is worth observing, that $`e_nf^2e_n^2f^2=0`$ only when $`f\left(𝐤\right)e_n\left(𝐤\right)`$ i.e. in a $`d_{x^2y^2}`$\- or $`\left(d_{x^2y^2}+s\right)`$-wave superconductor for $`f\left(𝐤\right)cos2\varphi `$. Thus, according to the model, these superconducting states will not be altered by anisotropic impurity scattering of the $`d_{x^2y^2}`$-wave symmetry provided the scattering strengths in the isotropic and anisotropic channels are equal ($`\mathrm{\Gamma }_1/\mathrm{\Gamma }_0=1`$). We believe, that it is rather a not plausible situation.
In conclusion, we have shown that the potential impurity scattering with the anisotropy introduced by Eq. (1) leads to a pair-breaking effect in superconductors for the anisotropy of the impurity potential, $`f\left(𝐤\right)`$, determined by a linear combination of functions from the non identity irreducible representations of the crystal point group. Therefore, in order to reflect a general feature of disorder induced superconductivity suppression, the anisotropy of the discussed model potential should be limited to this class of functions.
This work was supported by the Natural Sciences and Engineering Research Council of Canada.
|
no-problem/9904/cond-mat9904386.html
|
ar5iv
|
text
|
# 𝐪-dependence of low-frequency Raman scattering in silica glass
## Abstract
Accurate measurements of the dependence of low-frequency Raman scattering on the scattering angle were performed in two silica glasses. By a comparison of spectra measured at a large scattering angle (close to back scattering) and a small one (close to forward scattering), we for the first time observed a $`𝐪`$-dependence of the low-frequency light scattering in glasses. From the magnitude of the effect, the vibration correlation length is estimated and compared with results from the picosecond optical technique; a reasonable agreement is found.
Vibrational excitations in glasses show two peculiarities: an excess density of vibrational states with respect to the Debye one in the THz region (the boson peak) and a strong scattering of phonons at frequencies above around 100 GHz. These peculiarities appear at acoustic wavelengths of the order of nanometers and demonstrate themselves, e. g., in the anomalous thermal conductivity and specific heat at low temperatures (5$``$10 K) , and in inelastic neutron and light scattering spectra . Many works over the past two decades have been devoted to the study of these topics, but both the theoretical understanding and the experimental investigation of vibrational properties of glasses are far from being complete. New experimental techniques, like inelastic X-ray scattering (IXS) and small-angle inelastic neutron scattering (INS) open additional possibilities to study acoustic spectra in the Brillouin scattering regime and in the frequency range of the boson peak . Together with the picosecond optical technique (POT) , the results of these methods encourage further investigations of the dynamic and spatial properties of acoustic excitations in the range of 100$``$3000 GHz . However, the analysis of the results has been the subject of some controversy, that cannot be resolved at present due to the limited range of spectral resolution of the devices (see, e. g., Ref. ). Therefore, it would be very useful to obtain additional information about this subject from other experimental methods.
Traditionally, low-frequency Raman scattering is applied to study spectral properties of vibrational excitations in glasses. Usually the momentum dependence of the scattering ($`𝐪`$-dependence) is not considered , because the phonon mean free path in the corresponding frequency range is much less than the wavelength of the scattered light. As a result, any $`𝐪`$-dependence effect should be very small and difficult to observe experimentally. The $`𝐪`$-dependence of light scattering around the frequency of the boson peak has recently been investigated in SiO<sub>2</sub> glass at ambient conditions ; no $`𝐪`$-dependence was observed, but the accuracy of the measurement allowed an estimation of the upper limit of the localisation lenght of vibrations.
In the present Letter this approach is refined and extended to lower frequencies and lower temperatures. We present the first observation of a $`𝐪`$-dependence effect for low-frequency Raman scattering in glasses. In order to extend the available frequency range of the experiments and to demonstrate the reliability of our results, we use two different, complementary light scattering techniques: conventional Raman spectroscopy (employing a monochromator) and an interferometric technique (employing a tandem Fabry-Perot interferometer). In both cases we use spectrometers with the best elastic line suppression available in their class.
The Raman scattering experiment on a sample of glassy SiO<sub>2</sub> (Heralux, Heraeus, 130 ppm of OH<sup>-</sup>-groups) was performed using an Ar<sup>+</sup> laser (514.5 nm, 900 mW) and a five-grating monochromator Z40. The two scattering angles were 20 and 160; the respective wave vectors were $`q_1`$=0.62$`\times 10^2`$ nm<sup>-1</sup> and $`q_2`$=3.5$`\times 10^2`$ nm<sup>-1</sup> with the refractive index of the sample $`n`$=1.46. An aperture limiter was placed on the lens in order to collect the scattered light within an angle of $``$14. Polarized Raman spectra were measured using spectral slits of 60 GHz (2 cm<sup>-1</sup>) for the spectral interval 420$``$3180 GHz (14$``$106 cm<sup>-1</sup>) and 45 GHz (1.5 cm<sup>-1</sup>) for the spectral interval 150$``$660 GHz (5$``$22 cm<sup>-1</sup>). The position of the elastic line was checked for each spectrum with an accuracy of 3 GHz (0.1 cm<sup>-1</sup>).
One experiment was performed at ambient conditions. The temperature in the illuminated volume of the sample obtained from the Stokes/anti-Stokes ratio was 311.0$`\pm `$1.7 K, and within this precision was the same for experiments in both geometries. The temperature uncertainty leads to a spectral shape uncertainty lower than 3$`\times 10^3`$ over the measured spectral range. In order to obtain spectra with a high accuracy, for each scattering angle we recorded 120 and 36 spectra in the frequency ranges 420$``$2220 GHz and 150$``$660 GHz, respectively.
An Oxford optical helium cryostat was used for the low-temperature experiments. The incident laser beam entered the cryostat as in a conventional right angle experiment. We used a small mirror within the cryostat in order to have the necessary angle of the incident beam relative to the direction of the scattered light. With this optical scheme we exclude any Raman scattering contribution from cryostat windows that would be difficult to avoid when one uses the same window (or two parallel ones) for the incident laser beam and for the scattered light in the case of a monochromator with slits. The temperature in the illuminated volume of the sample was determined by the Stokes/anti-Stokes ratio to be 33.0$`\pm `$0.5 K. 72 and 40 spectra are accumulated for each scattering angle in the frequency ranges 150$``$660 GHz and 420$``$3180 GHz, respectively. We include low-frequency Raman scattering spectra of Heralux and Suprasil measured in previous experiments as right-angle Raman scattering, using the same laser and monochromator. The temperature of the illuminated part of the sample was 7 K in the case of Heralux and 4 K in the case of Suprasil as obtained from the Stokes/anti-Stokes ratio.
Inelastic light scattering spectra of a sample of Suprasil 300 (synthetic silica, Heraeus, $`<`$1 ppm of OH<sup>-</sup>-groups) were obtained at temperatures 300, 45 and 6 K (employing a CryoVac dynamic Helium cryostat) using a six-pass Sandercock tandem Fabry-Perot interferometer and an Ar<sup>+</sup> laser (514.5 nm, 600 mW). The scattered light was recorded with no selection of polarization; the polarization of the incident beam was perpendicular to the scattering plane leading to a domination of the polarized inelastic light scattering in this experiment due to the low depolarization ratio of silica glass ($``$0.3 at low frequencies). In this experiment either of two anti-parallel beams of equal intensity was focused onto the same volume of the sample. The direction of the scattered light corresponds to scattering angles of 7 and 173 for the respective beams ($`q_1`$=0.21$`\times 10^2`$ nm and $`q_2`$=3.6$`\times 10^2`$ nm), and was collected within an angle of $`16^{}`$. We used a free spectral range of 500 GHz over two spectral ranges on either side of the elastic line. The position of the elastic line was constantly kept aligned within about 4 GHz. The entrance and exit pinholes of the spectrometer were 450 and 700 $`\mu `$m, respectively, in order to suppress the transmission of higher orders of the tandem interferometer in combination with a prism (the tandem transmits every 20th order of a free spectral range; without suitable suppression this could significantly disturb a broad spectrum) . The experimentally determined finesse was better than 100.
To further validate the absence of possible contributions from the instrumental tail of the elastic line or from higher transmission orders of the tandem, we measured spectra for both geometries at a low temperature, $`T`$=6 K. At such a very low temperature the anti-Stokes part of the spectrum should be almost zero because the scattering intensity of the anti-Stokes spectral side is proportional to the Bose factor which is very small for this temperature (in the frequency range where this factor is not very small for this temperature, $`\nu `$$`<`$200 GHz, the signal itself is very low). Indeed, the anti-Stokes part of the spectrum at this temperature shows no deviations from the dark count level of 2.47 cts/s of our detector, demonstrating the absence of contributions of higher orders or from the elastic line. A quantitative estimation limits any systematic deviations to less than about 0.03 cts/s for the frequency range down to 100 GHz. This value is three orders of magnitude less than the level of the signal at $`T`$ = 300 K which was $``$ 50 cts/s. At $`T`$ = 45 K the level of the signal varies from 1.16 cts/s at 100 GHz to 10 cts/s at 1000 GHz. Thus, one can expect that a possible distortion of the experimental spectra is not higher than 0.3% for 1000 GHz and 3% for 200 GHz at 45 K.
A low-frequency light scattering spectrum of glasses consists of two contributions—the vibrational spectrum and a quasielastic one (QES); the latter increases faster than the Bose factor as temperature increases and dominates at low frequencies. Fig. 1 shows the reduced spectra (i. e., $`I/(n(\nu )+1)`$ of the Stokes side, where $`n(\nu )`$ is the Bose factor) of the Suprasil sample measured by the Sandercock tandem at $`T`$=300, 45 and 6 K in the forward scattering geometry; the right-angle monochromator spectrum recorded at 4 K is also included. At frequencies above some 600 GHz the vibrational contribution dominates and the reduced spectrum does not depend on temperature. At lower frequencies it increases with temperature, since QES becomes significant. At $`T`$=45 K the vibrational spectrum dominates at least down to 300 GHz.
The ratio of the back- and forward-scattering spectra, $`R`$, of Suprasil measured with the Sandercock tandem at $`T`$=300 K is shown in Fig. 2 for the Stokes and anti-Stokes sides. This ratio is a constant at high frequencies; we fix its value as 1 since it is constant up to the optical modes . The figure shows that the ratio deviates from a constant in the frequency range 100$``$300 GHz, which is a clear demonstration of the $`𝐪`$-dependence of the light scattering. To improve our statistics, in Fig. 3 (a) we show this ratio $`R`$ averaged over the Stokes and anti-Stokes sides and smoothed by adjacent averaging over 10 points. Circles in Fig. 3 (a) correspond to the result obtained for Heralux by Raman scattering. We note that in spite of a very weak $`𝐪`$-dependence (about a few percent) and different experimental techniques being used in these two cases, we find an excellent agreement between the two curves. This is a strong evidence for the reliability of the $`𝐪`$-dependence effect found in our experiments. Fig. 3 (b) shows the back- to forward scattering ratio $`R`$ obtained for Suprasil at $`T`$=45 K and for Heralux at $`T`$=33 K. Again a clear indication of the $`𝐪`$-dependence of the light scattering is found. The precision of these curves is lower than in Fig. 3 (a) due to a decrease of the signal at low temperatures; however, the magnitude of the $`𝐪`$-dependence effect is larger. Again, there is a good agreement between the results obtained by the different experimental techniques. A comparison of these results for the different temperatures with Fig. 1 leads to the conclusion that the $`𝐪`$-dependence effect increases with a decrease of the quasi-elastic contribution with respect to the vibrational contribution. The deviation of the $`𝐪`$-dependence from a constant is stronger for lower frequencies; this corresponds to an increase of the vibration correlation length with decreasing frequency, as will be explained in the following.
Let us turn to the interpretation of the experimental data. The $`𝐪`$-dependence of the inelastic light scattering intensity in the acoustic region is determined by the equation
$$I(q,\nu )F_\nu (q)\left|u_\nu (0)\right|^2g(\nu ),$$
(1)
where $`g(\nu )`$ is the vibrational density of states, $`u_\nu (r)`$ is the amplitude of a vibration of frequency $`\nu `$ and $`F_\nu (q)`$ is the spatial Fourier transform of the vibration correlation function
$$F(r)=u_\omega (\stackrel{}{r})u_\omega (0)/\left|u_\omega (0)\right|^2.$$
(2)
At small $`q`$, $`F_\nu (q)1b(ql_\nu )^2`$ where $`b`$ is a constant which depends on a particular form of $`F(r)`$ and $`l_\nu `$ is the vibration correlation length . To evaluate the $`𝐪`$-dependence of the experimental spectra, we use the ratio of $`I(q,\nu )`$ measured at the larger wavevector, $`q_2`$ (close to back scattering), to that at the small scattering angle corresponding to the wavevector $`q_1`$:
$$R(q_1,q_2,\nu )=I(q_2,\nu )/I(q_1,\nu )1b(q_2^2q_1^2)l_\nu ^2.$$
(3)
In Ref. it was shown that typical correlation functions (e. g., Gaussian or exponential) lead to a parameter $`b`$ in the range from 0.1 to 2. In the case of attenuated plane waves the respective expression for $`F_\nu (q)`$ is more complicated and will be considered in a forthcoming paper. For a rough estimate we follow Shuker and Gammon and approximate $`F(r)`$ by a step function: $`F(r)=𝚌onst`$ at $`r<l_\nu `$ and $`F(r)=0`$ at $`r>l_\nu `$. This leads to a value of $`b=0.1`$ (see, e. g., Ref. ).
The expression (3) corresponds to the vibrational part of the spectrum. At low frequencies (below 300 GHz at 45 K and 600 GHz at 300 K), the QES gives a significant contribution; its $`𝐪`$-dependence is not known and in principle may be different from that of vibrations as a function of frequency. So, we first consider the part of our spectra which is dominated by vibrations. Our data at 45 and 33 K show a clear $`𝐪`$-dependence in the region between 300 and 600 GHz. To estimate the correlation length $`l_\nu `$ from the observed $`𝐪`$-dependence by Eq. 3, we use our value of $`R`$ from Fig. 3 at, e. g., 320 GHz. This leads to $`l_\nu `$=27 nm at 33 K.
It is interesting to compare $`l_\nu `$ with the mean free path $`L_\nu `$ of the acoustic vibrations obtained by POT , because it is natural to expect that $`L_\nu `$ and $`l_\nu `$ are interrelated. In an estimate $`L_\nu `$=24 nm was obtained for $`L_\nu `$ in silica at 342 GHz and room temperature. Therefore we need to know $`l_\nu `$ at 300 K, where one should separate the vibrational contribution from that of QES in the data of $`R`$. We assume that within our accuracy the $`𝐪`$-dependence of QES is negligible. This is in accordance with the model of QES developed, e. g., in Ref. . In these papers it is argued that the $`𝐪`$-dependence of QES is the same as that of the boson peak. However, at the frequencies of the boson peak, as it was already shown in Ref. , a $`𝐪`$-dependence is absent with a precision of 0.3%, i. e. it is unobservable with our accuracy. The idea that QES is related to relaxational modes that are localized on a short range leads to the same conclusion. Therefore, assuming that $`R_{\mathrm{relax}}`$=1, it is easy to show that
$$R(q_1,q_2,\nu )1\eta b(q_2^2q_1^2)l_\nu ^2,$$
(4)
where $`\eta =I_{\mathrm{vib}}/I`$. Now, using Eq. (4), taking $`R`$ from Fig. 3 and $`\eta `$ from Fig. 1 (with the spectrum at $`T`$=4 K as the vibrational contribution $`I_{\mathrm{vib}}`$) one obtains $`l_\nu `$= 14 nm at $`T`$=300 K and $`\nu =`$ 320 GHz. This estimate is made for a particular choice of the parameter $`b`$. For different choices of $`b`$ the magnitude of $`l_\nu `$ will change as $`\sqrt{b}`$. The fact that the values of $`l_\nu `$ and $`L_\nu `$ have the same order of magnitude may be an additional evidence that the mechanism of the $`𝐪`$-dependence we use in this paper is correct.
We note, that at $`T`$=33 K the correlation length is by a factor of 2 bigger than at room temperature. Our result agrees with the conclusion of Ref. that at least up to frequencies of 400 GHz the vibrational mean free path decreases with increasing temperature up to 100 K. The conclusion of Ref. was based on a comparison of the data for the phonon mean free path from the POT , the thermal conductivity of SiO<sub>2</sub> and from the tunnel junction technique applied at different temperatures. Our result leads to the same conclusion for $`l_\nu `$ (independent of the choice of $`b`$), but is obtained by a single experimental technique. It is interesting to note that the ratio of $`l_\nu `$ at $`T`$=300 K to that at 33 K in our case (a factor of 2 for $`\nu `$=320 GHz) is similar to that obtained by Brillouin scattering at $`\nu `$=35 GHz ; this means that the dominating mechanism of the phonon attenuation may be the same at both frequencies at $`T`$$`>`$30 K.
In conclusion, we present the first observation of a $`𝐪`$-dependence of the low-frequency (0.1-1 THz) light scattering spectra of silica glasses. This effect provides information on the spatial properties of THz dynamics in glasses. We demonstrate that the estimate of the vibration correlation length found from the $`𝐪`$-dependence of light scattering is in a reasonable agreement with the vibrational mean free path found by POT.
Helpful discussions with Prof. A. P. Sokolov and Prof. V. K. Malinovsky are appreciated. N. V. S. thanks Université Lyon I and Universität Bayreuth for hospitality and the support from a grant for young scientists by the Siberian Branch of the Russian Academy of Sciences. This work has been supported by SFB 279 of the Deutsche Forschungsgemeinschaft.
|
no-problem/9904/cond-mat9904356.html
|
ar5iv
|
text
|
# Finite-size effects in the self-organized critical forest-fire model
## I Introduction
During the past years, systems which exhibit self–organized criticality (SOC) have attracted much attention, since they might explain part of the abundance of fractal structures in nature . Their common features are slow driving or energy input and rare dissipation events which are instantaneous on the time scale of driving. In the stationary state, the size distribution of dissipation events obeys a power law, irrespective of initial conditions and without the need to fine-tune parameters. Examples for such systems are the sandpile model , the self-organized critical forest fire model , the earthquake model by Olami, Feder, and Christensen , and the Bak-Sneppen evolution model . Numerical as well as analytical studies of those systems are usually based on the assumption that their critical behaviour can be described in similar terms as that of equilibrium critical systems. This assumption is given a basis in , where it is suggested that SOC systems can be mapped on conventional critical systems by interchanging control and order parameters. Thus, the Bak-Sneppen model can be mapped on a depinning problem . However, it has been shown in that the mapping suggested in for the SOC forest-fire model does not generate a system with a conventional critical point. Instead, the phase transition shows hysteresis effects and is discontinuous when approached from above. Other unconventional features have also been seen in the SOC forest-fire model, like the existence of more than one diverging length scale , the absence of a spanning cluster immediately beyond the critical point , and the dependence of the large-scale behaviour on details of the model rules . (For a review on the SOC forest-fire model, see .) Other SOC systems show also unconventional scaling behaviour. Thus, in the two-dimensional abelian sandpile model finite-size scaling is violated , and the critical exponents for the earthquake model by Olami, Feder, and Christensen appear to depend continuously on the parameters. There is substantial need to better understand the nature of the scaling behaviour of those systems.
It is the purpose of this paper to shed some light on the unconventional critical behaviour of the SOC forest-fire model by studying its finite-size effects. We choose a version of the model which is identical to the SOC forest-fire model for system sizes much larger than the correlation length, and we discuss the changes that occur in the model as the system size is decreased below the correlation length. We find that instead of displaying finite-size scaling, small systems undergo a rearrangement from a structure with patches of different density to a more homogeneous structure with large density fluctuations in time. We find also that, contrary to conventional critical systems, small systems and small parts of large systems differ in the probability distribution for the density and in the fire-size distribution. We suggest that these results can be explained by the fact that the system has two qualitatively different types of fires.
The outline of this paper is as follows: In section II, we define the model that we used for studying finite-size effects, and discuss briefly known results. Section III shows computer simulation results for the fire-size distribution and the tree density as the system size changes from values larger than the correlation length to values much smaller than it. In the conclusion, we summarize and discuss our findings.
## II The Model
The version of the SOC forest-fire model studied in this paper is defined on a square lattice with $`L^2`$ sites. Each site is either occupied (“tree”) or empty (“no tree”). At each time step, the system is updated according to the following rules: (i) “Burning”: A site in the system is chosen at random (“struck by lightning”). If the site is occupied, the whole cluster of occupied sites connected to this site (by nearest-neighbour coupling) is removed from the system (“burnt”), i.e., the occupied sites of that cluster turn to empty sites. If the chosen site is empty, nothing happens. (ii) “Tree growth”: We select randomly $`s_0pL^2`$ sites from the system and occupy those that are empty (possibly also including sites which have become empty due to the removal of the cluster). These sites are selected one after another, allowing for the same site being selected more than once during the same filling step. In principle, $`s_0`$ can therefore be larger than $`L^2`$, however, in our simulations we chose usually values smaller than $`L^2`$.
For fixed $`s_0`$ and very large system size $`L`$, these rules are equivalent to having a lightning probability $`f=1/L^2`$ per site and time step, and a tree growth probability $`p`$, and the model is identical to the original SOC forest-fire model . Because of this equivalence, which was first pointed out by Grassberger most numerical studies of the SOC forest-fire model up to now were performed using the above rules, which allow for fast and efficient computer simulations. With the above rules, finite-size effects can also be studied very efficiently, as was suggested in . However, one has to keep in mind that the results are somewhat different from those for the original model. While in the original model lightning can strike the system between the growth of any two trees, it can strike the system in the present model only after growth step (ii) is finished. This leads to density peaks in Figures 10 and 11 below that are not present in the original model. However, our main conclusions are not affected by the particular choice of the dynamical rules, as will be discussed further below.
Let us first summarize shortly the major numerical results for the case $`s_0L^2`$, as reported in the literature on the SOC forest-fire model. In this limit, only a small number of trees grow at each time step (compared to the total number of trees). After a transient time, a stationary state is reached where the tree density has only small fluctuations around some average value $`\overline{\rho }(s_0)`$ that does not depend on $`L`$. Throughout this paper, we study only stationary states and do not evaluate the initial transient behaviour. Since the mean number of trees $`\overline{s}`$ burnt during a fire must be identical to the mean number of trees growing between two fires, we have the relation
$$\overline{s}=s_0(1\overline{\rho })/\overline{\rho }.$$
(1)
The leading finite-size corrections to this equation are of order $`s_0/L^2`$ and can be neglected in the case $`s_0L^2`$ which we are considering in this paragraph. As $`s_0`$ increases, the mean fire size increases also, and we approach the critical point of the SOC forest-fire model, where the mean tree density is given by $`\overline{\rho }_c0.41`$. The correlation length $`\xi `$ is a measure for the radius of the largest tree cluster and is related to $`s_0`$ via $`\xi s_0^\nu `$, with $`\nu `$ 0.58 in $`d=2`$ dimensions. The size distribution of tree clusters near the critical point is well described by the scaling form
$$n(s)s^\tau 𝒞(s/s_{max}),$$
(2)
with a cutoff function $`𝒞`$ that is constant for small arguments and decays exponentially fast when the argument is considerably larger than 1. The cutoff cluster size $`s_{max}`$ is related to the correlation length $`\xi `$ via $`s_{max}\xi ^\mu `$, with $`\mu `$ being the fractal dimension of tree clusters, which is found to be 1.95 or 1.96 . The value of the exponent $`\tau `$ is approximately 2.14. The relation between $`s_{max}`$ and $`s_0`$ is $`s_{max}s_0^\lambda `$, with $`\lambda =\nu \mu 1.15`$ .
All these numerical findings agree well with conventional scaling assumptions based on a single diverging length scale. Analytical studies of the model, such as mean-field theories and renormalization group calculations are also based on conventional scaling assumptions. Therefore, the violation of finite-size scaling described in the following might appear surprising to many readers. However, one must keep in mind that the simulation data do not cover much more than one decade in the correlation length $`\xi `$. The observed scaling behaviour eq. (2), together with the measured values of the critical exponents, do not necessarily indicate an exact asymptotic scaling form, but may simply be a good approximation to more complicated scaling, which works well for the system sizes and parameter values studied in simulations. A similar phenomenon is known for the sandpile model, where good scaling collapses for the avalanche size distribution could be achieved in and older papers, although it has been recently shown that finite-size scaling is violated and that the simple scaling ansatz used for the data collapse is incorrect.
Figure 1 shows a snapshot of a system with a tree density $`\overline{\rho }`$ just below $`\overline{\rho }_c`$. One can distinguish regions of different densities with a rather homogeneous tree distribution within a region. These regions are obviously created by a fire that burns down a cluster of high tree density. After the fire, a burnt region is almost empty and becomes slowly filled with trees according to the law $`\dot{\rho }=p(1\rho )`$. We call these regions of homogeneous tree density “patches”, as we did in . If lightning strikes a tree in a patch of low density, it usually burns down a small tree cluster. If it strikes a patch of a density larger than the percolation threshold, it burns down a tree cluster as large as the patch itself. This observation indicates that there are two qualitatively different types of fires in the system: those that span an entire patch, and those that destroy a small percolation cluster within a patch of a tree density below the percolation threshold. As we will see below, this gives rise to the unusual finite-size properties of the model.
If the correlation length $`\xi `$ is of the same order as or larger than $`L`$, the behaviour sketched above is modified due to finite-size effects. For not too small values of $`s_0/L^2`$, the tree density increases by a noticeable amount between two fires, leading to large density fluctuations and to fires that span the entire system. If the SOC forest-fire model showed conventional critical behaviour, there would be a single diverging length scale, namely the correlation length $`\xi `$, which would be related to $`f/p`$ or, equivalently, to $`s_0`$, via $`\xi (f/p)^\nu `$ or $`\xi s_0^\nu `$. Finite-size effects would then manifest themselves in a scaling form
$$n(s)s^\tau 𝒞(s/L^\mu ),$$
(3)
for the size distribution of tree clusters. Furthermore, on scales smaller than $`L`$ and $`\xi `$, all measured quantities should be indistinguishable from those measured in a small section of an infinitely large critical system.
The following section presents simulation results that show that none of these finite-size scaling assumptions is satisfied for the SOC forest-fire model. In fact, the invalidity of the assumption of a single diverging length scale has already been shown in . The invalidity of the second assumption that measurements in small systems and in small sections of large systems should give identical results, can be understood by considering for instance the mean time interval between two fires. In a small subsystem of linear size $`l`$ of a large system with a correlation length $`\xi l`$, this is given by $`(p(1\overline{\rho }_c))^1`$. Just before fire reaches the subsystem, its tree density is far above the percolation threshold, and the spanning cluster of the subsystem is part of a large tree cluster that extends far beyond the limits of the subsystem. Lightning usually strikes this large cluster outside the subsystem, the time interval between two lightning strokes within the subsystem being $`L^2/l^2`$, which diverges as $`L`$ diverges. In contrast, fire cannot enter a small system from outside, but the tree density of a small system increases until lightning strikes a tree within the system. According to our rules, time is measured in units of the mean time interval between two lightning strokes. On this time scale, the time between two fires within a small system of linear size $`l`$ is finite. In contrast, the time interval between two fires within a subsystem of size $`l`$ of a much larger system is vanishingly small compared to the time interval between two lightning strokes within the subsystem. All these arguments are backed up and complemented by the numerical results reported in the following.
## III Results of Computer Simulations
In this section we will present and explain data obtained from about 300 runs of the model for various values of $`s_0`$ and $`L`$. Since many runs of the simulation were necessary, we chose a cluster of workstations rather than a ”supercomputer”. The system size $`L`$ varied between 10 and 2000 in these runs. We found that as finite-size effects become more important, the system shows a transition between two qualitatively different types of behaviour which we call critical behaviour and percolation-like behaviour. The critical behaviour is is characterized by a good scaling collapse of the fire size distribution and by large spatial variations in the local tree density. The percolation-like behaviour is characterized by large temporal fluctuations in the global tree density, with a rather homogeneous tree distribution within the system for any given time. Snapshots of the system therefore resemble percolation systems where each site is occupied by a tree with a probability $`\rho `$. (For an introduction to percolation theory, see e.g. .) The following three subsections show how this transition manifests itself in the mean tree density, the fire size distribution, and the probability distribution for the tree density.
### A Lines of constant tree density
First, we measured the mean tree density
$$\overline{\rho }=(1/T)\underset{t=1}{\overset{T}{}}\rho (t)$$
(4)
in the system, averaged over a large number of $`T`$ iterations, for various values of $`L`$ and $`s_0`$. The density $`\rho (t)`$ was always evaluated after the refilling step (ii). Compared to a model where trees grow at a rate $`\dot{\rho }=p(1\rho )`$, the density values in our model are somewhat larger when $`s_0/L^2`$ is not very small. If, for instance, the density is increased from $`\rho \mathrm{\Delta }\rho `$ to $`\rho `$ during the refilling step (ii), the value $`\rho `$ enters the above sum, while an evaluation based on a constant growth rate would give $`1\mathrm{\Delta }\rho /\mathrm{ln}(1+\mathrm{\Delta }\rho /(1\rho ))`$ instead of $`\rho `$.
It it not obvious what the relation between $`s_0`$, $`L^2`$ and $`\overline{\rho }`$ should be if we want to deduce it from an analogy with equilibrium critical systems. We have already mentioned that the temporal fluctuations in $`\rho (t)`$ become larger as the ratio $`s_0/L^2`$ increases. Similarly, temporal fluctuations increase in a critical equilibrium system when the system size becomes smaller. One might therefore expect that decreasing $`L`$ at fixed $`s_0`$ should drive the system toward the critical point, where $`\overline{\rho }=\overline{\rho }_c0.41`$. However, we have argued in the previous section that a given site burns down more often in a large system without finite-size effects than in a smaller system with finite-size effects that has the same value of $`s_0`$. From this, it follows that the mean tree density increases with decreasing $`L`$, when $`s_0`$ is fixed. In the limit $`s_0L^2`$ it must go to one. From this point of view, a system with sufficiently large finite-size effects should rather be compared to an equilibrium system in the ordered phase, for instance to a percolation system beyond the percolation threshold. The analogue of $`s_0`$ in a percolation system is then the mean size of the cluster that a given site belongs to, and it is proportional to $`L^2`$ beyond the percolation threshold . Indeed, we find that the mean size of fires becomes proportional to $`L^2`$ when finite-size effects are strong (see below). However, there is nevertheless a fundamental difference between a percolation system beyond the percolation threshold and our system with a density above $`\rho _c`$: In our model, a tree cluster that spans the system and has a size proportional to $`L^2`$ occurs only rarely when the mean density is only slightly above the critical density, while a percolation system above the percolation threshold has always a system spanning tree cluster.
In Figure 2, lines of constant mean tree density are plotted on a double logarithmic scale in $`s_0`$ and $`L^2`$. This figure shows that there are two qualitatively different regions in the $`s_0`$ vs. $`L^2`$ plane, with a transition region between them. First, there is the region where there are no finite-size effects, $`s_0L^2`$. In this region, the size of the tree clusters is much smaller than the system size, and the global density fluctuations are small. The mean tree density of such a system is smaller than $`\overline{\rho }_c0.41`$. For systems with small density fluctuations the probability that a given empty site is filled with a tree during one time step, is given by $`s_0/L^2`$, and the probability that a given tree is burnt by a fire is $`s_0(1\overline{\rho })/(\overline{\rho }L^2)`$. Since both probabilities decrease as $`1/L^2`$ with increasing $`L`$, dynamics of larger systems are slower than those of smaller systems. Apart from this change of the characteristic time scales, the local dynamics is independent of $`L`$, and consequently correlation functions and cluster size distributions are the same for systems of different sizes (provided that $`L^2s_0`$). This leads to the horizontal slope in the large $`L`$ regime of the curves for $`\overline{\rho }<\overline{\rho }_c`$ in Figure 2.
The curves show deviations from the horizontal behaviour when $`L`$ becomes as small as or smaller than the correlation length $`\xi s_0^\nu `$. These deviation occur in the transition region where finite-size effects begin to become noticeable. The dynamics changes from fires that are not affected by the finite system size to a fire size distribution that includes sometimes events of the order of the system size. Fires of such a large size destroy the patchy structure of the forest described earlier, and cause a more random tree distribution.
In the second region, lines of constant mean density are curves of constant $`s_0/L^2`$. This feature can best be understood when considering the parameter range where $`s_0`$ is of the order of $`L^2`$ or above. In this range, the mean tree density is much larger than $`\overline{\rho }_c`$, and the system contains a spanning cluster after each filling. During the ”burning” step, this cluster is removed with a finite probability, leading to a large change in density in the system. When a ”finite” (i.e. not spanning) cluster is removed during the ”burning” step, the overall density hardly changes for large $`L`$. The time series of the density is therefore determined almost completely by the filling events and by the large burning events. Since the filling events fill a large fraction of empty sites, and since large burning events burn a large fraction of trees, the tree distribution within the system is rather homogeneous. A snapshot of the system at a given time looks therefore similar to a percolation system with a density $`\rho (t)`$. From percolation theory we know that for a given density the fraction of trees sitting in the spanning cluster is independent of $`L`$, and consequently curves of constant large density are curves of constant $`s_0/L^2`$ in Figure 2. Even curves for smaller $`s_0/L^2`$, which correspond to densities only slightly above $`\overline{\rho }_c`$ show for sufficiently large $`L`$ an asymptotic behaviour $`s_0/L^2=`$ const, with the constant vanishing at $`\overline{\rho }_c`$. The reason is again that finite fires do not reduce the density of an infinitely large system, and that the system spanning fires reduce the density by an amount that does not depend on $`L`$, but only on the density itself.
Finally, the critical curve (for the density $`\overline{\rho }=\overline{\rho }_c0.41`$) is obtained from the condition
$$L\xi s_0^\nu ,\text{with}\nu 0.58.$$
This curve is the separatrix between the two regions described above and is indicated in Figure 2 by the bold line. As $`L`$ decreases, more and more curves merge with the separatrix when their correlation length becomes comparable to the system size.
### B The fire size distribution
Since lightning strikes each tree with the same probability, the size distribution of fires is proportional to $`sn(s)`$, with $`n(s)`$ being the size distribution of tree clusters. As mentioned above, conventional scaling would imply a form $`sn(s)s^{1\tau }𝒞(s/s_0^{\mu \nu })`$ for the fire size distribution if the correlation length $`\xi s_0^\nu `$ is smaller than the system size, and a finite-size scaling form $`sn(s)s^{1\tau }𝒞(s/L^\mu )`$ in the opposite case. In both cases, one would obtain a scaling collapse of the curves for different $`s_0`$ or $`L`$.
Figure 3 shows the fire size distribution for parameters such that $`\xi <L`$. While not perfect, the data collapse is good and would not impose the conclusion that simple scaling is violated. The bump near the end of the curves indicates that the cutoff function $`𝒞`$ increases first with increasing argument, before it shows the exponential decay. This bump is believed to contain all the trees that would sit in larger clusters if the system was exactly at the critical point .
Figure 4 shows the fire size distribution for parameter values such that the mean density is ten percent above its critical value. As discussed in the previous subsection, system spanning fires occur, and their size scales as $`L^2`$. These fires are responsible for the peaks in the fire size distribution. Similar peaks occur in equilibrium critical systems in the ordered phase, for example in the cluster size distribution of a percolation system above the percolation threshold. In a percolation system, the occurrence of such peaks implies that the system size is larger than the correlation length, which is identical to the cutoff in the radius of the finite (i.e., non system spanning) clusters. In our system, however, we do not see such an exponential cutoff to the size distribution of the finite clusters. Instead, the curve for $`L=800`$ in figure 4 appears to obey a power law from $`s100`$ up to the point where the peak begins. The explanation for this unusual behaviour must lie in the large temporal fluctuations in the density. The density is only sometimes so large that the large fires, which have a size of the order $`L^2/2`$, occur. At other times, the density values are different and allow for a broad range of other fire sizes.
The transition between critical scaling and $`L^2`$-scaling can be observed when $`L`$ is varied for fixed $`s_0`$, as illustrated in Figure 5. One can see that the shape of the curves changes continuously as $`L`$ is decreased. Clearly, because of this change in shape, finite-size effects do not manifest themselves in a scaling behaviour $`sn(s)s^{1\tau }𝒞(s/L^\mu )`$. It is impossible to generate a scaling collapse of different curves, even if their density is close to the critical density. Furthermore, as mentioned above, the cutoff introduced by the finite system size always scales as $`L^2`$, due to the occurrence of system spanning fires, and not as $`L^\mu `$, as expected for conventional critical systems.
For small system sizes, spanning clusters may already occur for densities below $`\overline{\rho }_c`$, an effect which is clearly visible in the curve for $`L=63`$. The formation of peaks due to finite size effects, was also found in .
As mentioned earlier, for conventional critical systems a system of small size and a small section of a large system are equivalent. We have argued that this is not true for the forest-fire model since the mean tree density and the time interval between fires are different in the two cases. The next two figures show that also the fire size distributions are different. In Figure 6 the fire size distribution of a section of a large system and a corresponding small system is shown. The scaling parts and the form of the bumps near the cutoff are very different. The fire size distribution of a small section of a large system is broader than that of a small system. The reason is that a section of a large system can contain a boundary between a patch of large tree density and a patch of small tree density. This boundary can pass through the section in different ways, and the number of trees in the dense part can take different values. Since large fires only burn the dense part, the size distribution of fires becomes broad.
Figure 7 shows how the fire size distribution changes when smaller and smaller sections of a large system are evaluated. Comparing to Figure 5, one sees again that the fire size distribution of small sections of large systems is different from that of small systems.
To summarize this subsection, the fire size distribution in the presence of finite-size effects does not show the features of finite-size effects in conventional critical systems. We find a continuous change in the shape of the fire size distribution and cutoffs that scale as $`L^2`$, rather than conventional finite-size scaling. Furthermore, the fire size distribution in small sections of large systems is different from small systems. Our results for the fire-size distribution confirm the qualitative transition from a parameter region unaffected by finite-size effects to a region dominated by system spanning fires that we found in the previous subsection.
### C Probability distribution of the density
Finally, we studied the temporal fluctuations in the values of tree density $`\rho (t)`$ by measuring how often a given value of $`\rho `$ occurs within a sufficiently long time series. We denote by $`w(\rho )d\rho `$ the probability that the tree density lies in the interval between $`\rho `$ and $`\rho +d\rho `$. The quantity $`w(\rho )`$ is therefore the probability density for the tree density $`\rho `$. We measured $`\rho `$ always after the trees were refilled, i.e., after step (ii). The results show again a qualitative change as the correlation length becomes smaller than the system size, reflecting the transition from critical to percolation-like behaviour.
In Figures 8 to 11 $`w(\rho )`$ is shown for different values for $`L`$ and $`s_0`$. For large enough and fixed $`s_0/L^2`$, the mean tree density increases with increasing system size, until it reaches its asymptotic value above $`\overline{\rho }_c`$. For fixed $`L`$, the mean tree density increases with increasing $`s_0`$. Apart from this increase in mean tree density, the following other trends are observed: (a) As the mean density approaches $`\overline{\rho }_c`$, the curves for $`w(\rho )`$ become broader (Figure 8). This is because the patches of homogeneous density visible in Figure 1 become larger with increasing $`\overline{\rho }`$, leading to larger global density fluctuations. (b) As the mean tree density increases above $`\overline{\rho }_c`$, the shape of the distribution becomes asymmetric, with the maximum moving from $`\overline{\rho }_c`$ to the percolation threshold $`\rho _{perc}0.59`$ (Figure 9). The reason is that for $`\overline{\rho }>\overline{\rho }_c`$ the patchy structure is replaced by a more homogeneous (percolation like) structure, where the largest fires are system spanning and occur for densities above the percolation threshold. Once the density lies above the percolation threshold the probability that a system spanning fire occurs is very high. This is why densities much higher than the percolation occur seldom, explaining the rapid decrease of $`w(\rho )`$ above the percolation threshold. (c) As the system size increases for fixed $`p=s_0/L^2`$, there occur peaks in the density distribution which become sharper and more numerous for larger $`L`$ (Figures 10 and 11). This can be explained by realizing that the difference between finite and system spanning fires becomes more pronounced as $`L`$ increases. In the limit $`L\mathrm{}`$, finite fires do not affect the density at all, while system spanning fires reduce it to a small value. Subsequent filling events then increase the density to $`1\mathrm{exp}(p)`$, $`1\mathrm{exp}(2p)`$, $`1\mathrm{exp}(3p)`$, etc., until the density is above the percolation threshold and another system spanning fire can occur. These system spanning fires do not always occur at the first instance where the density is above $`\rho _{perc}`$, since lightning might strike and empty a site. Also, the density immediately after a system spanning fire depends slightly on the density before the fire. Therefore, the series of density values given above, becomes slightly shifted, depending on the density just before the last system spanning fire. These shifted series of peaks, in turn, give rise to further possible density values above $`\rho _{perc}`$, leading to an additional series of peaks, etc. This is the mechanism leading to the fractal peak structure that emerges as $`L`$ is increased. For smaller $`L`$, the effect of small fires leads to a larger width of the peaks, which can therefore not be resolved when they are close together.
As mentioned in the introduction, the peaks in $`w(\rho )`$ are due to the fact that lightning can strike the system only between two filling steps. Had we instead performed our simulations using a small tree growth probability $`p`$ and a small lightning probability $`f`$, lightning could strike the system between the growth of any two trees. However, such a simulation would be very slow. In oder to make sure that our choice of the algorithm has no other effect on the results apart from the peaks in $`w(\rho )`$, we performed a test simulation where $`s_0`$ is not the same for each filling step. For each filling step, we chose $`s_0`$ randomly from an exponential distribution $`P(s_0)=(L^2p)^1\mathrm{exp}(s_0/(L^2p))`$. Such an exponential distribution results when lightning can strike the system between the growth of any two trees with the same probability. The mean number of trees growing between two lightning strokes is now smaller than before. The reason is that the majority of filling steps increase the tree number by a value smaller than $`L^2p`$, and that during large filling events the tree density becomes high and most of the sites chosen for filling are already occupied. The mean tree densities evaluated according to Eq. (4) are consequently slightly smaller than before.
The probability density $`w(\rho )`$ for the tree density resulting from this modified algorithm is shown in the insets in Figures 9 and 10. As expected, the peaks have vanished, while the change in shape from a curve with peak around 0.4 to a curve with peak near 0.6 due to finite-size effects is the same as before.
## IV Conclusion
In this paper, we have studied finite-size effects in the SOC forest-fire model. As these effects become stronger, the system rearranges from a structure with patches of different densities to a more homogeneous structure with large density fluctuations in time. This rearrangement is reflected in the structure of the fire size distribution, in the mean tree density, and in the temporal density fluctuations. Qualitatively similar (although quantitatively different) rearrangements are observed when smaller and smaller sections of a large SOC system are studied. Due to these qualitative changes, conventional finite-size scaling does not hold. Our work thus demonstrates that concepts from equilibrium critical phenomena cannot be taken over to the study of SOC systems such as the forest-fire model. Instead, these nonequilibrium critical systems show generically new features unknown in equilibrium. As the scaling ansatz eq. (2) which is based on a single length scale $`\xi s_0^\nu `$ can only be approximately correct, the true asymptotic scaling behaviour of the model is still an open question.
We suggest that the reason for the unconventional behaviour of the SOC forest-fire model is the fact that two qualitatively different types of fires occur: those that burn down a patch of high tree density of fractal dimension 2, and those that burn down a tree cluster of a smaller fractal dimension within a region of a tree density below the percolation threshold. As a consequence, the scaling behaviour of the system cannot be characterized using only one length scale. While the superposition of the two types of fires creates the impression of simple scaling as long as finite-size effects are small, the difference between them becomes clearly visible for smaller system sizes, where system spanning fires receive a larger weight. We suggest that the superposition of the two types of fires is also responsible for the other unconventional features of the SOC forest-fire model listed in the introduction.
Models related to the present one have been studied in and . In these models, the tree density is globally conserved by filling exactly the same number of trees into the system that have been burnt. As long as the density is below the critical value, these models are equivalent and show the critical behaviour of the SOC forest-fire model as the critical density is approached from below. They were introduced for the purpose of studying the SOC forest-fire model beyond the critical point, i.e. for densities larger than the critical density $`\overline{\rho }_c`$. As the density increases beyond the critical density, both models undergo large-scale rearrangements. In , where trees are refilled only after the end of a fire, the new structure consists of a finite number of large domains of different density. In , where each tree is refilled into the system immediately after it is burnt, the new structure has a continuously burning fire, and resembles the forest-fire model without lightning introduced earlier by Bak, Chen, and Tang , which shows spiral-shaped fire fronts . In both these models, the dynamics in the restructured state are dominated by large fires burning forests of a fractal dimension two, similarly to the restructuring due to finite-size effects reported in this paper.
Let us conclude by noting that it is unclear whether the behaviour in higher dimensions resembles that in two dimensions. Clearly, as long as the “patchy” structure with two qualitatively different types of fire occurs, mean-field theory which neglects all spatial structure cannot apply, and the system must be below its upper critical dimension. The recent paper by Bröker and Grassberger on the forest-fire model without lightning indicates that unusual scaling behaviour can occur also in 3 and 4 dimensional forest-fire models. If 6 is the upper critical dimension of the forest-fire model, as suggested in , then the scaling behaviour of the forest-fire model should be conventional above 6 dimensions.
###### Acknowledgements.
This work was supported by EPSRC Grant GR/K79307, and by the EU network project (TMR) “fractal structures and self organization”, EU-contract ERPFMRXCT980183.
|
no-problem/9904/hep-ph9904340.html
|
ar5iv
|
text
|
# FIELD-THEORETICAL TREATMENT OF NEUTRINO OSCILLATIONS
## I Introduction
Neutrino oscillations play a central role in neutrino physics. The most important condition for this phenomenon is given by neutrino mixing which is described by $`\nu _{L\alpha }=_jU_{\alpha j}\nu _{Lj}`$ with $`\alpha =e,\mu ,\tau ,\mathrm{}`$ and $`j=1,2,3,\mathrm{}`$ labelling neutrino flavours (types) and mass eigenfields, respectively. All neutrino oscillation experiments are evaluated with the formula
$$P_{\nu _\alpha \nu _\beta }(L/E_\nu )=\left|\underset{j}{}U_{\beta j}U_{\alpha j}^{}\mathrm{exp}\left(i\frac{m_j^2L}{2E_\nu }\right)\right|^2,$$
(1)
where $`U`$ denotes the unitary mixing matrix, $`L`$ the distance between source and detector and $`E_\nu `$ the neutrino energy. The neutrino masses $`m_j`$ are associated with the mass eigenfields $`\nu _j`$. It has been indicated in several publications that the standard derivation of Eq. (1) raises a number of conceptual questions (see, e.g., Ref. for a clear exposition). Some of these questions are solved by the wave packet approach (see also the review where a list of references can be found), however, the size and form of the wave packet is not determined in this approach and remains a subject to reasonable estimates.
The idea has been put forward to include the neutrino production and detection processes into the consideration of neutrino oscillations. Such an approach can be realized with *quantum mechanics* – in which case the neutrinos with definite mass are unobserved intermediate states between the source and detection processes – or with *quantum field theory* where the massive neutrinos are represented by inner lines in a big Feynman diagram depicting the combined source – detection process. In the following we will discuss the field-theoretical treatment. The aims and hopes of such an approach are the following: 1. The elimination of the arbitrariness associated with the wave packet approach, 2. the description of neutrino oscillations by means of the particles in neutrino production and detection which are really manipulated in an experiment, 3. a more complete and realistic description in order to find possible limitations of formula (1) in specific experimental situations.
Considering laboratory experiments, there are two typical situations for neutrino oscillation experiments. The first one is *decay at rest* (DAR) of the neutrino source. Its corresponding Feyman diagram is depicted in Fig. 1. The wave functions of the source and detector particles are localized (peaked) at $`\stackrel{}{x}_S`$ and $`\stackrel{}{x}_D`$, respectively. The other situation is *decay in flight* (DIF) of the neutrino source as represented in Fig. 2 where it is assumed that a proton hits a target localized at $`\stackrel{}{x}_T`$. The detector particle sits again at $`\stackrel{}{x}_D`$ but the source is not localized. In both situations the distance between source and detection is given by $`L=|\stackrel{}{x}_D\stackrel{}{x}_S|`$. Note that in the Feynman diagrams of Figs. 1 and 2 the neutrinos with definite mass occur as inner lines. In the spirit of our approach, neutrino oscillation probabilities are proportional to the cross sections derived from the amplitudes represented by these diagrams.
## II Assumptions and the resulting amplitude
The further discussion is based on the following assumptions:
1. The wave function $`\varphi _D`$ of the detector particle does not spread with time which amounts to
$$\varphi (\stackrel{}{x},t)=\psi _D(\stackrel{}{x}\stackrel{}{x}_D)e^{iE_{DP}t},$$
(2)
where $`E_{DP}`$ is the sharp energy of the detector particle and $`\psi _D(\stackrel{}{y})`$ is peaked at $`\stackrel{}{y}=\stackrel{}{0}`$.
2. The detector is sensitive to momenta (energies) and possibly to observables commuting with momenta (charges, spin).
3. The usual prescription for the calcuation of the cross section is valid.
With the amplitudes symbolized by Figs. 1 and 2 the oscillation probabilities are obtained by
$$P_{\underset{\alpha }{\overset{()}{\nu }}\underset{\beta }{\overset{()}{\nu }}}_𝒫𝑑P_S_𝒫\frac{d^3p_{D1}^{}}{2E_{D1}^{}}\mathrm{}\frac{d^3p_{Dn_D}^{}}{2E_{Dn_D}^{}}\left|𝒜_{\underset{\alpha }{\overset{()}{\nu }}\underset{\beta }{\overset{()}{\nu }}}\right|^2.$$
(3)
In this equation we have indicated the average over some region $`𝒫`$ in the phase space of the final particle of the detection process. If no final particle of the neutrino production process is measured then one has to integrate over the total phase space of these final states. By definition, at the source (detector) a neutrino $`\underset{\alpha }{\overset{()}{\nu }}`$ ($`\underset{\beta }{\overset{()}{\nu }}`$) is produced (detected) if there is a charged lepton $`\alpha ^\pm `$ ($`\beta ^\pm `$) among the final states.
In perturbation theory with respect to weak interactions, according to the Feynman diagrams Figs. 1 and 2 one has to perform integrations $`d^4x_1`$, $`d^4x_2`$ and $`d^4q`$ corresponding to the Hamiltonian densities for neutrino production and detection and the propagators of the mass eigenfields, respectively. These integrations are non-trivial because $`\psi _D`$ and the source (target) wave functions are not plane waves, but are localized at $`\stackrel{}{x}_D`$ and $`\stackrel{}{x}_S`$ ($`\stackrel{}{x}_T`$), respectively. After having performed the integrations over $`x_{1,2}`$ and $`q^0`$, in the asymptotic limit $`L\mathrm{}`$ only the neutrinos on mass shell contribute to the amplitude which can be written as
$$𝒜_{\nu _\alpha \nu _\beta }^{\mathrm{}}=\underset{j}{}𝒜_j^S𝒜_j^DU_{\beta j}U_{\alpha j}^{}e^{iq_jL}$$
(4)
with
$$E_D=\underset{b=1}{\overset{n_D}{}}E_{Db}^{}E_{DP}\text{and}q_j=\sqrt{E_D^2m_j^2}.$$
(5)
$`𝒜_j^S`$ and $`𝒜_j^D`$ denote the amplitudes for production and detection, respectively, of a neutrino with mass $`m_j`$. Note that $`E_D`$ is the energy on the neutrino line in Figs. 1 and 2 and it is independent of $`m_j`$. This is an immediate consequence of the assumptions in this section. Furthermore, due to the above-mentioned integrations and the asymptotic limit we obtain
$$𝒜_j^D\stackrel{~}{\psi }_D(q_j\stackrel{}{\mathrm{}}+\stackrel{}{p}_D^{})\text{with}\stackrel{}{\mathrm{}}=(\stackrel{}{x}_D\stackrel{}{x}_S)/L\text{and}\stackrel{}{p}_D^{}=\underset{b=1}{\overset{n_D}{}}\stackrel{}{p}_{Db}^{},$$
(6)
where $`\stackrel{~}{\psi }_D`$ is the Fourier transform of $`\psi _D`$.
## III Results
The preceding discussion leads us to the conclusion that with the assumptions stated in Section 2 the neutrino mass eigenstates are characterized by the energy $`E_\nu E_D`$ and momenta $`q_j`$ (5). Thus they have all the same energy determined by the detection process, but the momenta are different. The summation over $`E_D`$ is incoherent, i.e., it occurs in the cross section (see Eq. (3)), not in the amplitude (4). In this sense there are no neutrino wave packets in experiments conforming with our assumptions. Note that it has been pointed out in that a coherent or incoherent neutrino energy spread cannot be distinguished in neutrino oscillation experiments. Since the neutrino energy can in principle be determined with arbitrary precision the coherence length can theoretically be increased solely by detector manipulations. From Eq. (6) it follows that with $`\mathrm{\Delta }m^2|m_j^2m_k^2|`$ the condition
$$|q_jq_k|\frac{\mathrm{\Delta }m^2}{2E_D}\sigma _D\text{or}\sigma _{xD}\frac{1}{4\pi }L^{\mathrm{osc}}$$
(7)
is necessary for neutrino oscillations, where $`\sigma _D`$ and $`\sigma _{xD}`$ are the widths of the wave function of the detector particle in momentum and coordinate space, respectively, and $`L^{\mathrm{osc}}`$ is the oscillation length. In realistic experiments condition (7) holds because $`\sigma _{xD}`$ is a microscopic whereas $`L^{\mathrm{osc}}`$ a macroscopic quantity. For DAR an analoguous condition exists for the width of the neutrino source wave function.
For more details of the field-theoretical approach to neutrino oscillations, for a consideration of the finite lifetime of the neutrino source in the case of DAR and for an application of the results to the LSND and KARMEN experiments we refer the reader to . We have shown that in these experiments effects of the finite liftetime can be neglected. For a discussion of DIF see . Thus in the framework discussed here all corrections to Eq. (1) are negligible. Note that we have not taken into account or discussed the interaction of the final state particles in the source with the environment (“interruption of neutrino emission”), a possible intermediate range of the asymptotic limit $`L\mathrm{}`$ as found in and the possibility that in some cases (e.g., the KARMEN experiment) it is not realistic to use the conventional procedure to calculate the cross section (3) by taking the asymptotic limit of the final time to infinity.
|
no-problem/9904/cond-mat9904403.html
|
ar5iv
|
text
|
# Anomalous Raman shift in the ternary fullerides with 𝑡_{1𝑔} states
## Abstract
Raman spectra have been studied on two kinds of highly doped fullerides of $`A_xBa_3C_{60}`$ (x=0, 3; A=K, Rb) and $`K_xSm_{2.75}C_{60}`$ (x=0, 3.25). It was found that the Raman spectra are essentially identical to each other for all ternary fullerides. The results show a crossover point at the boundary of $`t_{1u}`$ and $`t_{1g}`$ bands for the charge transfer dependence of Raman shift. Particularly, the totally symmetric $`A_g(2)`$ mode in $`t_{1g}`$ fullerides cannot be understood by a simple extrapolation from the low doped $`t_{1u}`$ fullerides, where the $`A_g(2)`$ mode follows a characteristic shift of $`6.3cm^1`$ per elementary charge . The present result shows that the Raman spectra of $`t_{1g}`$ states cannot be explained only by charge transfer.
PACS numbers: 78.30.-j, 72.80.Rj, 74.70.-b
Raman scattering is a useful technique to study the vibrational properties of the $`C_{60}`$ molecule and its doped derivative compounds. Raman scattering is widely used to evaluate the electron-phonon coupling constant $`\lambda `$ for the doped $`C_{60}`$ superconductors basing on the analysis of the linewidths. In addition, the tangential pinch $`A_g(2)`$ mode is of particular interest in the studies of the doped $`C_{60}`$ as it yields a strong and narrow line in the Raman spectrum. The different stages of doping can be easily followed by in-situ Raman-scattering experiments. A continuous change of line intensity for the $`A_g(2)`$ mode of the stable phase is observed and the doping process leads to a characteristic downshift of this line regarding the number of electrons transferred to the $`C_{60}`$ molecule. A down-shift of 6-7 $`cm^1`$ per elementary charge on $`C_{60}`$ independent of the doping ion is observed. The $`AC_{60}`$ phase was discovered by down-shift of the $`A_g`$ pinch mode in in-situ Raman-scattering experiments.
$`A_3Ba_3C_{60}`$ phases (A=K, Rb, Cs) with a formal $`C_{60}^9`$ charge and half-filling of the $`t_{1g}`$ are recently reported. As the intercalation host, $`Ba_3C_{60}`$ is a vacancy-ordered derivative of the bcc $`A_6C_{60}`$ structure with half of the cation sites empty ($`A15`$ structure). Three alkali metal cations are introduced into $`Ba_3C_{60}`$ to form a cation-disordered $`A_3Ba_3C_{60}`$ phase isostructural with $`A_6C_{60}`$. $`K_3Ba_3C_{60}`$ with half-filled $`t_{1g}`$ band exhibits superconductivity at 5.6 K. However, the insertion of large $`A^+`$ cations leads to a decrease in $`T_c`$, contrary to the behavior of the $`A_3C_{60}`$ phases.
To investigate the vibrational properties of ternary fullerides and comparison of physical properties in between $`t_{1u}`$ and $`t_{1g}`$ fullerides, we have carried out a Raman scattering study on the fullerides $`A_xBa_3C_{60}`$ (x=0, 3; A=K, Rb) and $`K_xSm_{2.75}C_{60}`$ (x=0, 3.25). The results show that the Raman spectra are amazingly similar to each other for all ternary fullerides $`A_3Ba_3C_{60}`$ and $`K_{3.25}Sm_{2.75}C_{60}`$. An anomalous Raman shift of the two $`A_g`$ modes is observed when alkali metals are introduced into $`Ba_3C_{60}`$ and $`Sm_{2.75}C_{60}`$. It does not follow the characteristic relation between charge transfer and Raman shift widely observed for the two $`A_g`$ modes in alkali and Ca-doped $`C_{60}`$.
Samples of $`Ba_3C_{60}`$ and $`Sm_{2.75}C_{60}`$ were synthesized by reacting stoichiometric amount of powers of Ba, Sm, and $`C_{60}`$. A quartz tube with mixed powder inside was sealed under high vacuum of about $`2\times 10^6`$ torr, and heated at $`550600`$ $`{}_{}{}^{o}C`$ for three days. Synthesis of $`A_3Ba_3C_{60}`$ and $`K_{3.25}Sm_{2.75}C_{60}`$ was carried out in a similar manner to that of alkali doping into pure $`C_{60}`$, A piece of alkali and $`Ba_3C_{60}`$ or $`Sm_{2.75}C_{60}`$ powders were loaded in a Pyrex tube, which was sealed under $`2\times 10^6`$ torr and calcined at 250 $`{}_{}{}^{o}C`$ for three days. X-ray diffraction analysis was performed by a system equipped with a 4.5 kW rotating molybdenum anode as the x-ray generator and an imaging plate (IP, MAC, Science , DIP320V ) as the detector. X-ray diffraction showed that all samples were single phase, which was also confirmed by the single peak feature of the pentagonal pinch $`A_g(2)`$ mode in the Raman spectra.
Raman scattering experiments were carried out using the 632.8 nm line of a He-Ne laser in the Brewster angle backscattering geometry. The scattering light was detected with a Dilor xy multichannel spectrometer using a spectral resolution of 3 $`cm^1`$. In order to obtain good Raman spectra, the samples were ground and pressed into pellets with pressure of about 20 $`kg/cm^2`$, which were sealed in Pyrex tubes under a high vacuum of $`10^6`$ torr.
As reported in reference 10, X-ray diffraction shows that the structure is changed from $`A15`$ phase to a cation-disordered $`A_3Ba_3C_{60}`$ phase isostructural with bcc $`A_6C_{60}`$ (a=K, Ba) when alkali metal is intercalated into $`Ba_3C_{60}`$. Figure 1 shows the X-ray powder diffraction patterns for the samples $`Sm_{2.75}C_{60}`$ and $`K_{3.25}Sm_{2.75}C_{60}`$. For the pattern of $`Sm_{2.75}C_{60}`$, all peaks can be indexed with an orthorhombic lattice parameters a=28.158 Å, b=28.077 Å, and c=28.270 Å, which is consistent with previous report. It is easily seen in Fig.1 that the diffraction peak is much less in the pattern of $`K_{3.25}Sm_{2.75}C_{60}`$ than in the pattern of $`Sm_{2.75}C_{60}`$. In addition, no peak is observed at $`2\theta `$ below 5 degree. It suggests that the structure of $`K_{3.25}Sm_{2.75}C_{60}`$ should be simple relative to that of $`Sm_{2.75}C_{60}`$. We can index all diffraction peaks using a body-center ed-cubic cell with lattice parameter a=11.093 Å except for the two peaks marked by arrows in Fig.1. It is known that the basic structure of $`Sm_{2.75}C_{60}`$ is fcc, but the cation vacancy ordering in tetrahedral sites leads to a superstructure accompanied with a slight lattice deformation from cubic to orthorhombic. The Sm cations occupying the octahedral and tetrahedral sites experience off-center displacements since one out of every eight tetrahedral sites in the subcell is vacant. X-ray diffraction suggests that introduction of the alkali cations into $`Sm_{2.75}C_{60}`$ leads to the disappearance of cation-vacancy ordering and formation of a cation-disordered phase with composition $`K_{3.25}Sm_{2.75}C_{60}`$, similarly to the case of $`A_3Ba_3C_{60}`$. A detailed analysis on the structure of $`K_{3.25}Sm_{2.75}C_{60}`$ will be reported elsewhere.
Figure 2 shows Raman spectra of $`Ba_3C_{60}`$, $`K_3Ba_3C_{60}`$, and $`Rb_3Ba_3C_{60}`$. The positions ( $`\omega `$ ) and halfwidths ( $`\gamma `$ ) of the Raman modes observed are listed in Table I for all samples. In the $`Ba_3C_{60}`$ spectrum, there are about 13 strong Raman lines observed, some of which are doublets. The low-and high-frequency $`A_g`$ derived modes are located at 506 and 1430.8 $`cm^1`$, respectively. The position of $`A_g(2)`$ pentagonal pinch mode is identical to that of $`K_6C_{60}`$, suggesting that the charge transfer from Ba to $`C_{60}`$ is complete, also being consistent with $`6.3cm^1`$ redshift per electron relative to neutral $`C_{60}`$. However, the up-shift of 13 $`cm^1`$ for the radial $`A_g(1)`$ mode is larger than that for $`K_6C_{60}`$ (9 $`cm^1`$ ). It is to be pointed out that the Raman spectrum of $`Ba_3C_{60}`$ is amazingly similar to that of $`K_6C_{60}`$ except for the relative intensities between $`A_g(2)`$ and $`H_g(2)`$ modes.
In the case of $`K_3Ba_3C_{60}`$ and $`Rb_3Ba_3C_{60}`$, the similar Raman lines to $`Ba_3C_{60}`$ are observed. However, the strongest line is $`A_g(2)`$ mode, similarly to the case of $`K_6C_{60}`$ with bcc structure. It is worth pointing out that the $`K_3Ba_3C_{60}`$ and $`Rb_3Ba_3C_{60}`$ spectra are essentially identical to each other. These spectra are relatively insensitive to the choice of alkali-metal ion species. This behavior is similar to the case of $`A_3C_{60}`$ and $`A_6C_{60}`$ ( A=K, Rb ) compounds, where this behavior is explained both by a weak coupling between $`C_{60}`$ and alkali cations and by a complete charge transfer from the alkali-metals to $`C_{60}`$ molecules. However, an anomalous Raman shift for the two $`A_g`$-derived modes is observed in $`K_3Ba_3C_{60}`$ and $`Rb_3Ba_3C_{60}`$. The first thing to be noted is the $`A_g(2)`$ mode observed at 1425.5 ( 1424.1 ) $`cm^1`$ in $`K_3Ba_3C_{60}`$ ($`Rb_3Ba_3C_{60}`$). The relative down-shift of $`A_g(2)`$ mode measured from that of $`Ba_3C_{60}`$ is only 5.3 and 6.7 $`cm^1`$ for $`K_3Ba_3C_{60}`$ and $`Rb_3Ba_3C_{60}`$ , respectively. These values are much smaller than that expected basing on the established empirical relation between the formal $`C_{60}`$ valence and the Raman shift. Another anomaly is found in the radial $`A_g(1)`$ mode, which shows a downshift of 8.6 and 10.5 $`cm^1`$ relative to $`Ba_3C_{60}`$ for $`K_3Ba_3C_{60}`$ ($`Rb_3Ba_3C_{60}`$) , respectively. Such a large downshift upon alkali metal doping displays a sharp contrast with the case of $`A_xC_{60}`$ (A=K, Rb) where a slight upshift is observed with increasing alkali metal. Low frequency $`H_g`$ modes also show complicated and characteristic behavior. The $`H_g(2)`$ mode of cubic $`Ba_3C_{60}`$ is single peak at 432 $`cm^1`$, while it shows a splitting into two peaks both in $`K_3Ba_3C_{60}`$ and $`Rb_3Ba_3C_{60}`$. The splitting of $`H_g(3)`$ mode is observed in all samples in Fig.2, and the splitting is smaller in the alkali-doped ternaries, while the center of doublet remains unchanged. .
Room temperature Raman spectra of the samples $`Sm_{2.75}C_{60}`$ and $`K_{3.25}Sm_{2.75}C_{60}`$ are shown in Fig.3. For $`Sm_{2.75}C_{60}`$, an anomalously broad distribution of vibrational structures for the low frequency $`H_g`$ modes and around the $`A_g(2)`$ mode is observed, which could be related to the complexity of $`Sm_{2.75}C_{60}`$ structure. The low frequency $`H_g(1)`$ and $`H_g(2)`$ modes are asymmetric. $`H_g(2)`$ mode has to be fitted with four components. It suggests that the degeneracy $`H_g(2)`$ mode is lifted. This splitting may be attributed to the symmetry lowering due to the orthorhombic superstructure of this material. A similar behavior has been observed in single crystal $`K_3C_{60}`$ at 80 K and in $`Ba_4C_{60}`$ and $`Ba_6C_{60}`$ at room temperature. A bunch of lines appears around 700 $`cm^1`$ which are likely to be assigned to the $`H_g(3)`$ and $`H_g(4)`$ modes. It is to be pointed out that the pinch $`A_g(2)`$ mode occurs at 1432.8 $`cm^1`$, indicating that Sm is divalent, and the charge transfer is complete according to 6.3 $`cm^1`$ redshift per electron relative to neutral $`C_{60}`$. The position of the radial $`A_g(1)`$ is also consistent with that of $`A_6C_{60}`$ (A=K, Rb). This definitely confirms divalent Sm by Raman scattering, being consistent with the results of near-edge and extended X-ray absorption fine structure in $`Yb_{2.75}C_{60}`$.
The top of Fig.3 shows the spectrum of $`K_{3.25}Sm_{2.75}C_{60}`$ , which is amazingly similar to $`A_3Ba_3C_{60}`$ although the Raman spectra of the intercalation host materials are different. Table I shows that even the positions of all corresponding modes are almost the same between $`K_{3.25}Sm_{2.75}C_{60}`$ and $`A_3Ba_3C_{60}`$ compounds. This surprising result strongly indicates that Raman spectrum is insensitive to the metal ion species either in the intercalation host or as an intercalation guest, suggesting a weak coupling between the $`C_{60}`$ and the metal ions and a complete charge transfer from metal cations to $`C_{60}`$ molecules.
Raman shift of the radial $`A_g(1)`$ and $`A_g(2)`$ pinch modes as a function of the nominal charge transfer simply derived from the chemical formula for $`Ba_3C_{60}`$, $`Sm_{2.75}C_{60}`$, $`A_3Ba_3C_{60}`$ (A=K, Rb), and $`K_{3.25}Sm_{2.75}C_{60}`$ are plotted in Fig.4. For comparison, the results of $`Ba_xC_{60}`$ (x=4, 6) recently reported by us, $`K_xC_{60}`$ (x=3, 6) reported by Duclos et al., and $`KBa_2C_{60}`$ and $`KBaCs_{60}`$ reported by Yildirim et al. are also plotted in the figure. For the $`A_g(2)`$ mode, there is a boundary at charge transfer of 6 electrons. Below which, the relation between the Raman shift of $`A_g(2)`$ mode and charge transfer is linear. They follow the characteristic shift of $``$6.3 $`cm^1`$ redshift per elementary charge for the all samples including $`Ba_3C_{60}`$ and $`Sm_{2.75}C_{60}`$. However, when the charge transfer exceeds six, the pinch $`A_g(2)`$ mode does not follow the simple characteristic relation, showing a complicated behavior . In other words, although there exist a systematic relation between Raman shift and charge transfer for the ”$`t_{1u}`$ band” fullerides, the ”$`t_{1g}`$ band” fullerides are quite different. The charge transferred from metal to $`C_{60}`$ molecules were -9, -15, and -7 for the $`Ba_4C_{60}`$, $`Ba_6C_{60}`$, and $`K_3Ba_3C_{60}`$ ( $`K_{3.25}Sm_{2.75}C_{60}`$ ), respectively, if charge transfer was derived according to 6.3$`cm^1`$ redshift per elementary charge. This is much larger ( less ) than the nominal charge transfer for $`Ba_xC_{60}`$ (x=4 and 6) ($`K_3Ba_3C_{60}`$ and $`K_{3.25}Sm_{2.75}C_{60}`$). If the charge transfer from Ba to $`C_{60}`$ molecules is complete, the redshift per elementary charge is about 9 $`cm^1`$, being much larger than $``$6.3 $`cm^1`$. In addition, an anomalous Raman shift for the $`A_g`$ derived modes takes place in the ternary $`A_3Ba_3C_{60}`$ (A=K, Rb) and $`K_{3.25}Sm_{2.75}C_{60}`$, the down-shift is only about 6 $`cm^1`$ for the nominal charge transfer of three electrons relative to $`Ba_3C_{60}`$ and $`Sm_{2.75}C_{60}`$.
Let us switch to the arguments on the radial $`A_g(1)`$ mode. A continuous up-shift of the $`A_g(1)`$ mode with increasing charge transfer until six electrons observed as shown in Fig.4. This mode-stiffening has been explained by electrostatic interactions which produces sufficient stiffening to encounter the softening of the mode expected on the basis of charge transfer effects. It is to be noted that the Raman shift of the radial $`A_g(1)`$ for $`Sm_{2.75}C_{60}`$ with $`t_{1u}`$ states falls on the linear line of $`K_xC_{60}`$ with $`t_{1u}`$ states. In contrast , in the $`t_{1g}`$ fullerides, the frequency of the $`A_g(1)`$ mode almost remains unchanged for $`Ba_xC_{60}`$, while it decreases with increasing charge transfer in the ternaries $`A_3Ba_3C_{60}`$ and $`K_{3.25}Sm_{2.75}C_{60}`$. It is worth noting that both $`A_g(1)`$ and $`A_g(2)`$ modes show complicated behavior in the $`t_{1g}`$ bands.
In the case of Ba derived fullerides, a strong hybridization between the Ba atoms and the $`\pi `$-type functions of the $`C_{60}`$ network may be responsible for the behavior of the $`A_g`$ derived modes as discussed in Ref.15. The charge transfer is complete in the intercalation hosts of $`Ba_3C_{60}`$ and $`Sm_{2.75}C_{60}`$, following the characteristic relation between Raman shift and charge transfer, while the introduction of alkali metals leads to an anomalous Raman shift for the two $`A_g`$ derived modes in the ternary $`A_3Ba_3C_{60}`$ and $`K_{3.25}Sm_{2.75}C_{60}`$,. However, the Raman shift of $`A_g(2)`$ pinch mode follows the characteristic shift of $``$6.3 $`cm^1`$ per elementary charge transfer for the co-intercalated $`C_{60}`$ of $`ABa_2C_{60}`$ (A=K, Rb, and Cs) and $`KBaCsC_{60}`$ with $`t_{1u}`$ states. It suggests that the anomalous Raman shift of the $`A_g`$ mode could be related to the ”$`t_{1g}`$” band electrons. In fact, some other differences between the ”$`t_{1u}`$” and ”$`t_{1g}`$” superconductors have been found. In the $`t_{1u}`$ band, superconductivity occurs only at $`(C_{60})^3`$ state which corresponds to the half-filling of the $`t_{1u}`$ band. However, the reported ”$`t_{1g}`$” superconductors are $`B_4C_{60}`$ (B=Ba, Sr), $`A_3Ba_3C_{60}`$, and $`Ca_5C_{60}`$. which have the molecular valences of -8, -9, and -10, respectively. Such tolerance for the molecular valence in $`t_{1g}`$ superconductor makes a striking contrast with the strict constraint for the valence state in the case of ”$`t_{1u}`$” superconductors. Another difference to be pointed out is that the ”$`t_{1u}`$” superconductivity appears only in fcc or related structures. In the case of $`t_{1g}`$ band, superconductivity is observed in various structures, even orthorhombic structure. To completely understand the physical properties and superconducting mechanism of fullerides, it is necessary to further investigate $`t_{1g}`$ fullerides, and to explain the difference between $`t_{1u}`$ and $`t_{1g}`$ fullerides.
In summary, the Raman scattering study has been carried out in the two fulleride families of $`A_xBa_3C_{60}`$ (x=0, 3; A=K, Rb) and $`K_xSm_{2.75}C_{60}`$ (x=0, 3.25). The results definitely show some differences between $`t_{1u}`$ and $`t_{1g}`$ fulleride for Raman shift of the two $`A_g`$ modes. For $`t_{1u}`$ fullerides, the down-shift of the $`A_g(2)`$ mode and the up-shift of $`A_g(1)`$ mode can be explained by charge transfer and the electrostatic interactions upon doping, respectively. In contrast, the $`t_{1g}`$ fullerides show a complicated behavior for Raman shift of the two $`A_g`$ modes. The Raman shifts of the two $`A_g`$ modes , observed in the tenaries $`A_3Ba_3C_{60}`$ and $`K_{3.25}Sm_{2.75}C_{60}`$, significantly differ from that of the binary fullerides $`Ba_4C_{60}`$ and $`Ba_6C_{60}`$. It is also found that Raman spectra of $`A_3Ba_3C_{60}`$ and $`K_{3.25}Sm_{2.75}C_{60}`$ are identical to each other, indicating a weak coupling between the $`C_{60}`$ and metal cations, and the same charge transfer to $`C_{60}`$ molecules. However, the anomalous Raman shift could not be explained only by the charge transfer.
X. H. Chen would like to thank the Inoue Foundation for Science for financial support. X. H. Chen acknowledges Dr. Kitagawa for his help in experiment. This work is partly supported by Grant from the Japan Society for Promotion of Science (RFTF 96P00104, MPCR-363/96-03262) and from the Ministry of Education, Science, Sports, and Culture.
FIGURE CAPTIONS
Figure 1:
Powder X-ray diffraction patterns of $`Sm_{2.75}C_{60}`$ and $`K_{3.25}Sm_{2.75}C_{60}`$. The two peaks marked by arrows can not indexed with body-centered-cubic cell with lattice parameter a=11.093 Å.
Figure 2:
Room temperature Raman spectra of $`Ba_3C_{60}`$, $`K_3Ba_3C_{60}`$, and $`Rb_3Ba_3C_{60}`$.
Figure 3:
Room temperature Raman spectra of $`Sm_{2.75}C_{60}`$ and $`K_{3.25}Sm_{2.75}C_{60}`$.
Figure 4:
Charge transfer-Raman shift relation for the radial $`A_g(1)`$ and $`A_g(2)`$ pinch modes. Squares represent the current experimental results, circles refer to the experimental results of $`Ba_xC_{60}`$ reported by us (Ref.15), up-triangles are from the results of $`K_xC_{60}`$ reported by Duclos et al. (Ref.5), and down-triangles refer to the results of $`KBa_2C_{60}`$ and $`KBaCsC_{60}`$ reported by Yildirim et al. (Ref.16).
|
no-problem/9904/hep-th9904147.html
|
ar5iv
|
text
|
# Improved heavy quark potential at finite temperature from anti-de Sitter supergravity
## Abstract
We improve the heavy quark potential, extracted from the Wilson loop average in the Ads/CFT approach by taking the quantum fluctuation of the only radial coordinate of $`Ads_5`$ which is transverse to the world-sheet of the classical Nambu-Goto string in the static gauge, and obtain the universal Lüscher-Symanzik-Weisz/Lüscher term.
A seminal conjecture by Maldacena states the duality between the large $`N`$ superconformal Yang-Mills theory and the low energy behaviour of the superstring theory in the Anti-de Sitter space back-ground. It is further shown that the correlation functions of the local gauge invariant operators of $`𝒩=4`$ supersymmetric Yang-Mills theory are related to the classical limit of the superstring partition function . Also the vacuum expectation values of the non-local operators like Wilson loops are easily computed using this duality . This facilitates to extract the heavy quark potential for SUSY Yang-Mills theories. The proposal became quite useful after Witten has shown how one can break Supersymmetry with non-zero temperature. This makes a close contact with the pure QCD and as far as possible, several qualitative behaviour of QCD in the strong coupling limit are verified . In this approach the background Euclidean time is compactified on a circle which is related to the equilibrium temperature T. Here one takes the space-time fermion to be anti-periodic in this direction and then all the fermions will be as heavy as the order of the temperature T. In the extreme high temperature limit fermions decouple. Thus the d+1 dimensional theory is described as a d dimensional theory without any fermion. This gives us a d dimensional pure QCD in the zero temperature.
The heavy quark potential of the planar $`QCD_3`$ is extracted from the vacuum expectation value of the Wilson loop operator . The space-like Wilson loop gives a linear potential which correctly signals the area law. However the Lüscher-Symanzik-Weisz (LSW) term or commonly known as Lüscher term is absent from this potential . This is a universal term independent of any coupling constant and is a characterstic of the string generated QCD potential and has been seen and correctly estimated by Lattice simulation . This might cast doubt on the supergravity approach to study planar QCD. However there are some attempts by Greensite and Olesen and by Dorn and Otto to get rid of this problem. So far there is no concrete explanation in this regard. Here we claim to get this LSW term by considering quantum fluctuations of the classical solution which is an improvement over the Nambu-Goto solution taken by authors of Ref. . We will discuss all the points in the sequel.
Here we consider the case of planar $`QCD_3`$ which can be trivially extended to higher dimension. Following Maldacena we put heavy quark and anti-quark at $`x=L/2`$ and $`x=L/2`$ respectively which are attached by a Nambu-Goto string in an $`Ads_5\times S_5`$ black-hole background, where the metric reads
$$ds^2=\alpha ^{}\left\{\frac{U^2}{R^2}\left((1U_T^4/U^4)dt^2+\underset{i}{}dx_i^2\right)+\frac{R^2}{U^2}\frac{dU^2}{1U_T^4/U^4}+R^2d\mathrm{\Omega }_5^2\right\},$$
(1)
Where $`R^2=\sqrt{4\pi g_sN}=\sqrt{4\pi g_{}^{2}{}_{YM}{}^{}N}`$ , and $`U_T=\pi R^2T`$ . Here $`g_s`$ is the string coupling and $`N`$ is the number of coinciding $`D_3`$ branes which gives $`U(N)`$ Yang-Mills theory and $`R`$ is the radius of the anti-de Sitter space. The Nambu-Goto string action in this background is
$$S=\frac{1}{2\pi \alpha }𝑑\tau 𝑑\sigma \sqrt{\text{det}(G_{MN}_\mu x^M_\nu x^N)}.$$
(2)
We look for the Wilson loop which traces area in the space-like surface. The boundaries are accordingly $`x^1=\pm \frac{L}{2},x^0=x^3=0`$ and the line where $`U\mathrm{}`$ runs parallel to $`x^2`$ axis. In the static gauge the world-sheet cordinates are $`\tau =x^2,\sigma =x^1`$ and $`_\tau x^M=\delta _2^M`$ and $`_\sigma x^0=_\sigma x^3=0`$. Thus the action
$$S=\frac{𝒯}{2\pi }_{\frac{L}{2}}^{+\frac{L}{2}}𝑑\sigma \sqrt{\frac{U^4}{R^4}+U^{\mathrm{\hspace{0.17em}2}}\frac{U^4}{U^4U_T^4}}.$$
(3)
The potential is extracted in the $`𝒯\mathrm{}`$ limit as
$$V(L)=Lim_𝒯\mathrm{}\frac{1}{𝒯}S_{cl}.$$
(4)
The $`𝒯\mathrm{}`$ limit is the analouge of $`\mathrm{}0`$ limit and thus the classical solution is an exact solution. However Lüscher term is a purely quantum mechanical term showing the finite size scaling. In this very brief report we show here how this classical potential is extracted and how it can be further improved to accommodate the LSW/Lüscher term.
Using symmetry and the classical equation of motion we get
$$\sqrt{\frac{U^4}{R^4}+U^{\mathrm{\hspace{0.17em}2}}\frac{U^4}{U^4U_T^4}}=C_0\frac{U^4}{R^4},$$
(5)
Where $`C_0`$ is a constant which is fixed from the boundary relation $`U(\sigma _0)=U_0`$ is such that $`U^{}(\sigma _0)=0`$. This gives
$$\frac{L}{2}\sigma =\frac{R^2}{U_0}_{U/U_0}^{\mathrm{}}\frac{dy}{\sqrt{(y^41)(y^41+ϵ)}},$$
(6)
where $`ϵ=1U_T^4/U_0^4`$ and the dimension less variable $`y(\sigma )=\frac{U(\sigma )}{U_0}`$. The classical action is
$$S=\frac{𝒯}{2\pi }U_0_1^{\mathrm{}}𝑑y\frac{y^4}{\sqrt{(y^41)(y^41+ϵ)}},$$
(7)
which gives
$$V(L)=\frac{U_0^2}{2\pi R^2}L+\frac{U_0}{\pi }_1^{\mathrm{}}𝑑y\left(\sqrt{\frac{y^41}{y^41+ϵ}}1\right)+\frac{U_0U_T}{\pi }.$$
(8)
This is a linear potential $`V(L)=\sigma L+\mathrm{}`$ where $`\sigma =\frac{U_0^2}{2\pi R^2}`$ is the string tension. In terms of Yang-Mills coupling this gives $`\sigma =\frac{\pi }{\sqrt{2}}g_{YM}\sqrt{NT^3}`$. However the subleading term $`\frac{c}{L}`$ is missing in the classical approximation. Also $`c`$ is supposed to be universal term which does not contain any physical parameter except the finite size scale $`L`$. This is a purely quantum effect. We take here one-loop quantum effect around the classical solution of $`U`$ as $`U=U_{cl}+\eta `$ where $`\eta `$ is the quantum field. The action in this order is given by
$$S_2=\frac{𝒯}{2\pi }_{\sigma _0}^{L/2}_2$$
(9)
where
$$_2=\frac{^2}{U^2}\eta ^2+2\frac{^2}{UU^{}}\eta \eta +\frac{^2}{U_{}^{}{}_{}{}^{2}}(\eta )^2.$$
(10)
Here $``$ is the Lagrangian given in eq.(3). To integrate over the field $`\eta `$ we redefine
$$\phi =\alpha \eta ,$$
(11)
where $`\alpha `$ is a function of the classical field $`U_{\mathrm{𝑐𝑙}}`$ obeying equations of motion. (Henceforth we denote $`U`$ as $`U_{\mathrm{𝑐𝑙}}`$ and drop the subscript.) Now $`\eta `$ is substituted as $`\frac{\phi }{\alpha }`$ in eq.(10) to give
$$_2=(\phi )^22\phi \phi (\frac{\alpha ^{}}{\alpha }\frac{b}{2\alpha ^2})+(\frac{\alpha _{}^{}{}_{}{}^{2}}{\alpha ^2}b\frac{\alpha ^{}}{\alpha ^3}+\frac{c}{\alpha ^2})\phi ^2+_{measure},$$
(12)
and $`_{measure}`$ is due to the redefinition of the field $`\eta `$ , where
$`\alpha ^2`$ $`=`$ $`{\displaystyle \frac{^2}{U_{}^{}{}_{}{}^{2}}}`$ (13)
$`b`$ $`=`$ $`2{\displaystyle \frac{^2}{UU^{}}}`$
$`c`$ $`=`$ $`{\displaystyle \frac{^2}{U^2}}.`$ (14)
( Here $``$ denotes the derivative with respect to $`\sigma `$. ) Any generic term like $`f\phi \phi `$ can be taken as $`\frac{1}{2}f\phi \phi _{bounadry}`$ and a term like $`\frac{1}{2}f\phi \phi `$ due to integration by parts in the action. We take here the Dirichlet boundary condition for the fluctuating field $`\phi `$ and hence drop the boundary term from the action. This gives
$$_2=((\phi )^2+(\frac{\alpha ^{\prime \prime }}{\alpha }\frac{1}{2}\frac{b^{}}{\alpha ^2}+\frac{c}{\alpha ^2})\phi ^2)+_{measure}.$$
(15)
Taking further derivatives with respect to $`\sigma `$ of the equation of motion
$$\frac{}{U^{}}=\frac{}{U},$$
(16)
one gets
$$\frac{c}{\alpha ^2}\frac{1}{2}\frac{b^{}}{\alpha ^2}=2\frac{U^{\prime \prime }}{U^{}}\frac{\alpha ^{}}{\alpha }+\frac{U^{\prime \prime \prime }}{U^{}}.$$
(17)
Substituting this we get
$$_2=(\phi )^2+m^2(\sigma )\phi ^2$$
(18)
where the effective mass like term
$$m^2(\sigma )=(^2\alpha +2\frac{U^{\prime \prime }}{U^{}}\alpha +\frac{U^{\prime \prime \prime }}{U^{}}\alpha )\frac{1}{\alpha }.$$
(19)
From eq.(13) $`\alpha `$ is given by
$$\alpha =\frac{g(\sigma )}{U^{}(\sigma )}$$
(20)
where
$$g(\sigma )=\frac{U_0}{R}\frac{\sqrt{y^41}}{y^2}.$$
(21)
This shows that
$$m^2(\sigma )=\frac{^2g}{g}.$$
(22)
From eq.(6) one observes that $`y(\sigma )`$ remains infinitesimally closer to 1 and then rapidly grows to infinity. If we take $`y(\sigma )=1+\delta (\sigma )`$ where $`\delta `$ is a very small parameter and keeping up to first order we get
$$\frac{L}{2}\sigma =const.\sqrt{\delta (\sigma )}$$
(23)
and for $`y(\sigma )`$ little bigger we can neglect 1 compare to $`y^4`$ in eq.(6) and get
$$\frac{L}{2}\sigma =const.\frac{1}{y^3}.$$
(24)
For the first case (c.f. eq.(21) and eq.(23))
$$g(\sigma )=p(1\sigma )$$
(25)
where $`p=2U_0^3\frac{\sqrt{ϵ}}{R^3}`$ which gives $`^2g=0`$ and for $`y`$ large $`g(\sigma )`$ is a constant and $`^2g=0`$. This shows that $`m^2(\sigma )`$ is zero for the entire range of integration.
The measure is now changed from $`D\eta D\phi `$ and the contribution of this to the effective action
$$S_{measure}=\frac{1}{L}_{\sigma _0}^{L/2}𝑑\sigma ln\alpha ^2,$$
(26)
This integration can be trivially done which contributes a constant to the potential. We take the quantum fluctuation $`\eta `$ to be zero on the boundary of the Wilson loop i.e. for $`\sigma =0`$ and $`L/2`$. Then the integration over $`\phi `$ gives the most required LSW term $`\frac{\pi }{12L}`$ for the potential which is our main result.
As mentioned earlier the classical action is exact in the static gauge when $`𝒯\mathrm{}`$. We just supplemented this classical action with the quantum fluctuation of the transverse cordinate $`U`$. We do not take the fluctuations of the world-sheet for all the transverse coordinate as has been shown by Greensite and Olesen . In the finite temperature case and the high temperature limit when dimensional reduction is taking place, fermions completely decouple from the spectrum. Since we are in the very low energy limit of extracting the potential we feel it is legitimate not to take any of these fluctuations.
Note Added
This work was completed in December and was presented in the String theory workshop at Puri (9th Dec- 19th Dec 1998). That time I was not aware of the work of Greensite and Olesen and also of Förste et. al. where both the groups have considered the worldsheet fluctuations of both the fermionic and the bosonic coordinates.
Acknowledgements: I would like to thank the organisers and all the participants of the Puri workshop for fruitful discussions.
|
no-problem/9904/math9904035.html
|
ar5iv
|
text
|
# Untitled Document
On the minimal normal compactification
of a polynomial in two variables
Angelo Vistoli<sup>*</sup><sup>*</sup>Partially supported by the University of Bologna, funds for selected research topics.
Dipartimento di Matematica
Università di Bologna
Piazza di Porta San Donato 5
40127 Bologna, Italy
E-mail address: vistoli@dm.unibo.it
1 Introduction
Let $`C`$ be an integral affine curve over a field $`\kappa `$, $`\alpha ,\beta :C𝐀^2`$ two closed embeddings. We say that $`\alpha `$ and $`\beta `$ are equivalent when there is an automorphism $`\varphi `$ of $`𝐀^2`$ with $`\varphi \alpha =\beta `$. It was stated by B. Segre (\[Se\]) and proved by Suzuki in \[Su\] that when $`\kappa =𝐂`$ any embedding of $`𝐀^1`$ into $`𝐀^2`$ is equivalent to the standard embedding $`t(t,0)`$. This was generalized to the case that $`\kappa `$ is arbitrary, and the degree of $`f`$ is prime to the characteristic of $`\kappa `$, by Abhyankar and Moh (\[AM\]). This is what is usually called the Abhyankar–Moh theorem. On the other hand, there are many affine curves with an infinite number of non equivalent embeddings into $`𝐀^2`$: for example, $`𝐀^1\{0\}`$.
Suzuki in \[Su\] also proves a very nice result: if $`C`$ is smooth and has only one branch at infinity (that is, it is the complement of a point in a smooth projective curve) and $`f`$ is a generator of the ideal of $`C`$ in $`𝐀^2`$, then $`C`$ is an ordinary fiber of $`f`$, that is, $`f`$ is a topological fibration in a neighborhood of 0. In their important, and arduous, article \[AS\] Abhyankar and Singh carry the study of this case much further, over arbitrary fields; in particular, for example, such a curve $`C`$ has at most finitely many nonequivalent embeddings, with appropriate conditions on the characteristic of the field.
Another proof of Suzuki’s theorem was given by Artal Bartolo, in \[AB\], based on the results of \[EN\], relating knot theory with the theory of polynomials in two variables.
Now, let $`f:𝐀^2𝐀^1`$ be a polynomial in two variables defined over an algebraically closed field $`\kappa `$. We shall always assume that $`f`$ is primitive, that is, that the generic fiber of $`f`$ is integral. We consider the minimal normal compactification $`\overline{f}:X𝐀^1`$ of $`f`$, namely the only normal irreducible surface $`X`$ containing $`𝐀^2`$ as an open subset, together with a proper morphism $`\overline{f}:X𝐀^1`$ extending $`f`$, with the property that each fiber of $`f`$ is dense in the corresponding fiber of $`\overline{f}`$. It is often singular.
Let $`E_1,\mathrm{},E_r`$ be the horizontal components of $`X𝐀^2`$, namely the irreducible components of $`X𝐀^2`$ that dominate $`𝐀^1`$. By standard results, $`E_1,\mathrm{},E_r`$ are isomorphic to $`𝐀^1`$ and do not intersect (Proposition 1). To each $`E_i`$ we associate two integers. The first is the degree $`e_i`$ of $`E_i`$ over $`𝐀^1`$; one can think of $`e_1,\mathrm{},e_r`$ as the orders of the orbits of the monodromy group acting on the branches at infinity of a general fiber of $`f`$. The second is the least positive integer $`\delta _i`$ such that $`\delta _iE_i`$ is a Cartier divisor on $`X`$; since $`E_i`$ is smooth, we have that $`\delta _i=1`$ if and only if $`X`$ has no singularities along $`E_i`$.
Our result (Theorem 1) says that if the characteristic of $`\kappa `$ is 0, the greatest common divisor of $`\delta _1e_1,\mathrm{},\delta _re_r`$ is 1. In particular, if there only one component $`E_1`$, this maps isomorphically onto $`𝐀^1`$, and $`\overline{f}`$ is smooth along $`E_1`$. So, if one of the fibers of $`f`$ has only one branch at infinity, then there is a simultaneous resolution of singularities at infinity of $`f`$. This easily implies the Suzuki–Abhyankar–Moh embedding theorem.
One can show that the integers $`e_i\delta _i`$ coincide with the integers $`m_i`$ defined by Eisenbud and Neumann (see \[AB\], p. 102). So in characteristic 0 our result follows from \[EN\], section 4, although our proof is shorter.
In characteristic $`p`$ we only get that the greatest common divisor of $`\delta _1e_1,\mathrm{},\delta _re_r`$ is 1 when the degree of the polynomial is prime to $`p`$ (see Theorem 2). This implies the Suzuki–Abhyankar–Moh embedding theorem over a perfect field.
The proof of the Theorem 1 is entirely straightforward, and very short; it uses standard topological methods, plus some elementary facts about rational surface singularities. If one substitutes ordinary topological cohomology with étale cohomology with $`𝐙_{\mathrm{}}`$ coefficients, where $`\mathrm{}`$ is a prime different from the characteristic of $`\kappa `$, one gets a proof of Theorem 2. We do not include the proof of this general case, but anyone who is familiar with étale cohomology will be able to reconstruct the details.
2 Acknowledgments
I am grateful to Pierrette Cassou-Noguès for several helpful comments about the history of the results I quote. In particular she pointed out reference \[AB\] to me.
3 The results
Consider a complex polynomial in two variables, i.e., a morphism $`f:𝐀^2𝐀^1`$ defined over C. We shall always assume that $`f`$ is primitive, that is, that $`f`$ is not constant, and not obtained by composition with a polynomial in one variable of degree greater than 1. This the same as saying that the generic fiber of $`f`$ is integral, or that the subfield $`𝐂(f)`$ is algebraically closed in $`𝐂(x,y)`$.
We consider the minimal normal compactification $`\overline{f}:X𝐀^1`$ of $`f`$, obtained by taking the closure $`\mathrm{\Gamma }`$ of the graph of $`f`$ in $`𝐏^2\times 𝐏^1`$, considering its normalization $`X^{}`$, and then calling $`X`$ the inverse image of $`𝐀^1`$ in $`X^{}`$. Then $`X`$ is a normal integral complex quasiprojective scheme over $`\kappa `$, containing $`𝐀^2`$ as an open subscheme. Furthermore the morphism $`f`$ extends to a morphism $`\overline{f}:X𝐀^1`$, which has the useful property that every fiber of $`f`$ is dense inside the corresponding fiber of $`\overline{f}`$. Let us call $`E_1,\mathrm{},E_r`$ the irreducible components of the complement $`E`$ of $`𝐀^2`$ in $`X`$. Each of the $`E_1,\mathrm{},E_r`$ is an affine integral curve dominating $`𝐀^1`$: we will call $`e_1,\mathrm{},e_r`$ the degrees of $`E_1,\mathrm{},E_r`$ over $`𝐀^1`$.
Furthermore, the divisor class groups of the local rings of $`X`$ are finite, because $`X`$ has rational singularities (\[Li\], Proposition 17.1.) We will call $`\delta _i`$ the least common multiple of the orders of $`E_i`$ in each of the divisor class groups of the local rings of $`X`$ at points of $`E_i`$; clearly $`\delta _iE_i`$ is a Cartier divisor on $`X`$, while $`\delta E_i`$ is not a Cartier divisor for any integer $`\delta `$ with $`0<\delta <\delta _i`$.
Theorem 1.Each of the $`E_1,\mathrm{},E_r`$ is isomorphic to $`𝐀^1`$, and they are pairwise disjoint.
Furthermore the greatest common divisor of the products $`\delta _1e_1,\mathrm{},\delta _re_r`$ is 1.
The first statement in the theorem is quite standard. It has an important consequence; if $`\delta _i`$ is 1, that is, if $`E_i`$ is a Cartier divisor on $`X`$, then $`X`$ is smooth at all point of $`E_i`$.
Corollary 1.Assume that $`X`$ has only one component at infinity. Then all the fibers of $`\overline{f}`$ are integral and smooth at infinity.
In particular, this happens when one of the fibers of $`f`$ has only one branch at infinity.
This follows immediately from the theorem, because the hypothesis implies that $`\delta _1=1`$, i.e., $`X`$ is smooth, and $`e_1=1`$, i.e., $`E_1`$ maps isomorphically onto $`𝐀^1`$.
From the corollary we get a new proof of the renowned Suzuki–Abhyankar–Moh theorem. For this we only need to assume that $`\kappa `$ is perfect.
The Suzuki–Abhyankar–Moh theorem over C.Any embedding of $`𝐀^1`$ into $`𝐀^2`$ defined over C is equivalent to the standard embedding $`t(t,0)`$.
Proof. . Let $`C`$ be a curve in $`𝐀^2`$ isomorphic to $`𝐀^1`$, $`f𝐂[x,y]`$ a generator of the ideal of $`C`$. Because of the corollary, each geometric fiber of $`\overline{f}`$ is isomorphic to $`𝐏^1`$, so $`X`$ is a $`𝐏^1`$-bundle on $`𝐀^1`$. If we call $`E`$ the complement of $`𝐀^2`$ in $`X`$, with its reduced scheme structure, then the projection from $`E`$ onto $`𝐀^1`$ is an isomorphism. Hence there is an isomorphism $`\varphi `$ of $`𝐏^1\times 𝐀^1`$ with $`X`$ carrying $`𝐀^1\times \mathrm{}`$ into $`E`$, and such that $`\overline{f}\varphi :𝐏^1\times 𝐀^1𝐀^1`$ is the second projection. The restriction of $`\varphi `$ to $`𝐀^2`$ carries the line with equation $`y=0`$ into $`C`$, and this proves the theorem. $`\mathrm{}`$
This can be made to work in positive characteristic. Let us fix an algebraically closed field $`\kappa `$, and call $`p`$ be the characteristic exponent of $`\kappa `$, namely the characteristic of $`\kappa `$ if this is positive, and 1 otherwise.
Consider a primitive polynomial in two variables, i.e., a morphism $`f:𝐀^2𝐀^1`$ defined over $`\kappa `$ with integral general fiber; as before, $`f`$ has a minimal normal compactification $`\overline{f}:X𝐀^1`$. Define $`E_1,\mathrm{},E_r`$, $`e_1,\mathrm{},e_r`$ and $`\delta _1,\mathrm{},\delta _r`$ as before. Then we can not conclude that the $`\delta _ie_i`$ are relatively prime; however, we have the following.
Theorem 2.Each of the $`E_1,\mathrm{},E_r`$ is isomorphic to $`𝐀^1`$, and they are pairwise disjoint.
Furthermore the greatest common divisor of the products $`\delta _1e_1,\mathrm{},\delta _re_r`$ is a power of $`p`$ and divides the degree of $`f`$.
We still get the corollary, in the following form.
Corollary 2.Assume that the degree of $`f`$ is prime to the $`p`$, and that $`X`$ has only one component at infinity. Then all the fibers of $`\overline{f}`$ are integral and smooth at infinity.
In particular, this happens when one of the fibers of $`f`$ has only one branch at infinity.
Remarkably, using a different technique one can prove that $`X`$ is smooth when it has only one component at infinity, without assuming that the degree of $`f`$ is prime to $`p`$. Unfortunately, I do not have any interesting application of this.
From Corollary 2 we get a new proof of the Suzuki–Abhyankar–Moh theorem over any perfect field.
The Suzuki–Abhyankar–Moh theorem over a perfect field.Any embedding of $`𝐀^1`$ into $`𝐀^2`$ defined over a perfect field, whose degree is relatively prime to the characteristic of $`\kappa `$ is equivalent to the standard embedding $`t(t,0)`$.
Proof of Theorem 1. Recall that $`\mathrm{\Gamma }`$ is the closure of the graph of $`f`$ in $`𝐏^2\times 𝐏^1`$, $`X^{}`$ its normalization, $`f^{}`$ and $`\pi `$ the projections of $`X^{}`$ onto $`𝐏^1`$ and $`𝐏^2`$, respectively. Let $`L=𝐏^2𝐀^2`$ be the line at infinity, $`L^{}`$ its proper transform in $`X^{}`$.
Let $`E_i^{}`$ be the closure of $`E_i`$ in $`X^{}`$: the first statement of the theorem is a consequence of the following fact.
Lemma 1.The curves $`L^{}`$ and $`E_i^{}`$, for each $`i=1,\mathrm{},r`$, are isomorphic to $`𝐏^1`$, and any two of them do not intersect in more than one point. Furthermore, if $`E_i^{}`$ and $`E_j^{}`$, with $`ij`$, intersect in a closed point $`pX^{}`$, then $`pL^{}`$.
Assuming Lemma 1, and keeping in mind that that $`L^{}`$ is the fiber of $`f^{}`$ over the point at infinity $`\mathrm{}𝐏^1(\kappa )`$, we see that each of the $`E_i^{}`$ can have only one point over $`\mathrm{}`$, and therefore the inverse image $`E_i`$ of $`𝐀^1`$ in $`E_i^{}`$ is isomorphic to $`𝐀^1`$. Also from Lemma 1 we get that the $`E_i`$ do not intersect.
Proof. The natural morphism $`\pi :X^{}𝐏^2`$ is birational and $`𝐏^2`$ is smooth, so $`\mathrm{R}^1\pi _{}𝒪_X^{}=0`$. Let $`\stackrel{~}{L}=\pi ^1(L)_{\mathrm{red}}`$. Since $`𝒪_{\stackrel{~}{L}}`$ is a quotient of $`𝒪_X^{}`$, so $`\mathrm{R}^1\pi _{}𝒪_{\stackrel{~}{L}}=0`$. We have $`\pi _{}𝒪_{\stackrel{~}{L}}=𝒪_L`$, so from the Leray spectral sequence
$$E_2^{ij}=\mathrm{H}^i(𝐏^1,\mathrm{R}^j\pi _{}𝒪_{\stackrel{~}{L}})\mathrm{H}^{i+j}(\stackrel{~}{L},𝒪)$$
we get that $`\mathrm{H}^1(\stackrel{~}{L},𝒪)=0`$. If $`Z`$ is subscheme of $`\stackrel{~}{L}`$, the sheaf $`𝒪_Z`$ is a quotient of $`𝒪_{\pi ^1(L)}`$, and if $``$ is the ideal of $`Z`$ in $`\pi ^1(L)`$ we have $`\mathrm{H}^2(\pi ^1(L),)=0`$, hence $`\mathrm{H}^1(Z,𝒪)=0`$. This in particular applies to any of the curves $`L^{}`$ and $`E_i^{}`$. Any integral projective curve with arithmetic genus 0 is isomorphic to $`𝐏^1`$.
Also, if $`C_1`$ and $`C_2`$ are two of these curve, from the fact that $`\mathrm{H}^1(C_1C_2,𝒪)=0`$ we see that $`C_1`$ and $`C_2`$ have at most one common point. Analogously, the fact that $`\mathrm{H}^1(L^{}E_i^{}E_j^{},𝒪)=0`$ implies that $`E_i^{}`$ and $`E_j^{}`$ cannot meet outside of $`L^{}`$, because $`L^{}`$ meets both $`E_i^{}`$ and $`E_j^{}`$. $`\mathrm{}`$
Now consider the group $`PicX`$ of Cartier divisors on $`X`$, and the natural map $`PicXClX`$ into the group of Weil divisors. Since $`𝐀^2`$ is factorial, and all of its invertible regular functions are constant, it follows that $`ClX`$ is a free abelian group with basis $`E_1`$, …, $`E_r`$. Since the map $`PicXClX`$ is injective, because $`X`$ is normal, this proves the following.
Lemma 2.The group $`PicX`$ is free, with basis $`\delta _1E_1`$, …, $`\delta _rE_r`$.
The fact that the $`\delta _ie_i`$ are relatively prime is easily proved, after having established the following two facts.
Lemma 3.The first Chern class map $`PicX\mathrm{H}^2(X,𝐙)`$ is an isomorphism.
Lemma 4.Let $`C`$ be a general fiber of $`\overline{f}`$. The the restriction map $`\mathrm{H}^2(X,𝐙)\mathrm{H}^2(C,𝐙)`$ is surjective.
In fact, the restriction of $`\delta _iE_i`$ to $`C`$ has degree $`\delta _ie_i`$; the three lemmas together imply that the restriction of the $`\delta _iE_i`$ generate $`\mathrm{H}^2(C,𝐙)=𝐙`$, hence that 1 is a linear combination of the $`\delta _ie_i`$.
There remains to give proofs of the last two lemmas; both are rather formal.
Proof of Lemma 3. Let $`\rho :\stackrel{~}{X}X`$ be a resolution of the singularities of $`X`$, $`F_1,\mathrm{},F_s`$ the exceptional divisors, $`\stackrel{~}{E}_i`$ the proper transforms of the $`E_i`$. Then the complement of the $`F_j`$ and the $`\stackrel{~}{E}_i`$ in $`\stackrel{~}{X}`$ is $`𝐀^2`$; therefore the Picard group of $`\stackrel{~}{X}`$ is freely generated by the $`F_j`$ and the $`\stackrel{~}{E}_i`$. Likewise, $`\mathrm{H}^2(\stackrel{~}{X},𝐙)`$ is freely generated by the cohomology classes of the $`E_i`$ and $`F_j`$; so the first Chern class map $`Pic\stackrel{~}{X}\mathrm{H}^2(\stackrel{~}{X},𝐙)`$ is an isomorphism.
The pullback map $`PicXPic\stackrel{~}{X}`$ is clearly injective, and a divisor class in $`Pic\stackrel{~}{X}`$ is in the image of $`PicX`$ if and only if its restriction to each of the $`F_j`$ has degree 0. The reason is that $`X`$ has rational singularities (\[Li\], Theorem 12.1
Now take cohomology. We have that $`\mathrm{R}^1\rho _{}𝐙_{\stackrel{~}{X}}=0`$, while $`\rho _{}𝐙_{\stackrel{~}{X}}=𝐙`$, and $`\mathrm{R}^2\rho _{}𝐙_{\stackrel{~}{X}}`$ is a sheaf concentrated in the singular points of $`X`$, whose stalk over $`pX`$ is a direct sum of one copy of Z for each exceptional divisor over $`p`$. By considering the Leray spectral sequence of the map $`\rho :\stackrel{~}{X}X`$, one deduces that the restriction map $`\mathrm{H}^2(X,𝐙)\mathrm{H}^2(\stackrel{~}{X},𝐙)`$ is injective, and its image consists exactly of the classes in $`\mathrm{H}^2(\stackrel{~}{X},𝐙)`$ which have degree 0 on each $`F_j`$.
By putting these two statements together, we see that $`PicX`$ and $`\mathrm{H}^2(X,𝐙)`$ are identified with two subgroups of $`Pic\stackrel{~}{X}`$ and $`\mathrm{H}^2(\stackrel{~}{X},𝐙)`$ which correspond under the isomorphism $`Pic\stackrel{~}{X}\mathrm{H}^2(\stackrel{~}{X},𝐙)`$ given by the first Chern class. This proves Lemma 3. $`\mathrm{}`$
Proof of Lemma 4. Consider the Leray spectral sequence
$$E_2^{ij}=\mathrm{H}^i(𝐀^1,\mathrm{R}^j\overline{f}_{}𝐙_X)\mathrm{H}^{i+j}(X,𝐙);$$
since $`\mathrm{H}^2(𝐀^1,\mathrm{R}^1\overline{f}_{}𝐙_X)=0`$, because $`𝐀^1`$ is an affine curve and $`\mathrm{R}^1\overline{f}_{}𝐙_X`$ a constructible sheaf, we get that the map
$$\mathrm{H}^2(X,𝐙)\mathrm{H}^0(𝐀^1,\mathrm{R}^2\overline{f}_{}𝐙_X)$$
is surjective. Now take the trace map
$$tr:\mathrm{R}^2\overline{f}_{}𝐙_X𝐙_{𝐀^1}.$$
Because the general fiber of $`\overline{f}`$ is integral, the trace map is generically an isomorphism. Let $`F=\overline{f}^1(t)`$ be a fiber of $`\overline{f}`$ over a closed point $`t𝐀^1(\kappa )`$, $`F_1,\mathrm{},F_s`$ the irreducible components of $`F`$, $`m_1,\mathrm{},m_s`$ the lengths of the local rings of $`F`$ at $`F_1,\mathrm{},F_s`$. By proper base change the stalk $`(\mathrm{R}^2\overline{f}_{}𝐙_X)_t`$ is canonically isomorphic to
$$\mathrm{H}^2(F,𝐙)\underset{i=1}{\overset{s}{}}\mathrm{H}^2(F_i,𝐙)𝐙^s;$$
with this identification, the trace map on the stalks over $`t𝐀^1(\kappa )`$ is identified with the map from $`𝐙^s`$ to Z that sends $`(k_1,\mathrm{},k_s)`$ to $`k_1m_1+\mathrm{}+k_sm_s`$. But $`m_1,\mathrm{},m_s`$ are relatively prime, because $`f:S𝐀^1`$ does not have multiple fibers, so the trace map is surjective, and its kernel is concentrated on a finite number of points. By taking global sections we see that the global trace map
$$tr:\mathrm{H}^2(X,𝐙)𝐙$$
is surjective. But $`tr:\mathrm{H}^2(X,𝐙)𝐙`$ coincides with the restriction map $`\mathrm{H}^2(X,𝐙)\mathrm{H}^2(C,𝐙)𝐙`$. Hence this restriction map is surjective. This proves the lemma, and hence the theorem. $`\mathrm{}`$
Note. From the spectral sequence of the map $`X𝐀^1`$ one deduces that $`\mathrm{H}^3(X,𝐙)=0`$; furthermore, from the spectral sequence of a resolution $`\stackrel{~}{X}X`$ one sees that the restriction map $`\mathrm{H}^2(\stackrel{~}{X},𝐙)\mathrm{H}^2(F,𝐙)`$ is surjective. From this one can deduce that the class of $`E_i`$ generates the product $`_{pE_i}Cl\widehat{𝒪}_{X,p}`$; this means that $`\delta _i`$ can also be defined as the product of the orders of the group $`Cl\widehat{𝒪}_{X,p}`$ for $`pE_i`$.
To prove Theorem 2 one follows the steps in the proof of Theorem 1, subsituting étale cohomology with $`𝐙_{\mathrm{}}`$ coefficients to classical cohomology, where $`\mathrm{}`$ is a prime different from the characteristic of $`\kappa `$; in this way one shows that $`\mathrm{}`$ does not divide the greatest common divisor of the $`e_i\delta _i`$. We leave the details to the interested reader. The only thing that does not follow is that the greatest common divisor of the $`\delta _ie_i`$ divides the degree $`d`$ of $`f`$.
To show this, call $`C`$ the closure in $`𝐏^2`$ of a general fiber of $`f`$, $`C^{}`$ the proper transform of $`C`$ in $`X^{}`$. Then $`C^{}`$ is a general fiber of $`f^{}`$, hence it is a Cartier divisor on $`X^{}`$; the intersection number $`(C^{}L^{})`$ is 0, and $`(C^{}E_i^{})=e_i`$ for each $`i=1,\mathrm{},r`$. We have a decomposition of $`\pi ^{}(L)`$ as a Weil divisor
$$\pi ^{}[L]=[L^{}]+\underset{i=1}{\overset{r}{}}m_iE_i$$
for certain positive integers $`m_1,\mathrm{},m_r`$. Since the restriction of the $`m_iE_i`$ to $`X`$ must be a Cartier divisor, we see that $`\delta _i`$ divides $`m_i`$, so we write
$$\pi ^{}[L]=[L^{}]+\underset{i=1}{\overset{r}{}}n_i\delta _iE_i.$$
But
$$d=(CL)=(C^{}\pi ^{}[L])=(C^{}L^{})+\underset{i=1}{\overset{r}{}}n_i\delta _i(C^{}E_i^{})=\underset{i=1}{\overset{r}{}}n_i\delta _ie_i,$$
by the projection formula, and this completes the proof. $`\mathrm{}`$
4 References
\[AM\] Abhyankar, S. S., Moh, T. T.: Embeddings of the line in the plane, J. Reine Angew. Math. 276, 148–166 (1975).
\[AS\] Abhyankar, S. S., Singh, B.: Embeddings of certain curves, Am. J. of Math. 100, 99–175 (1978).
\[AB\] Artal Bartolo, E.: Une démonstration géométrique du théorème d’Abhyankar–Moh, J. Reine Angew. Math. 464, 97–108 (1995).
\[EN\] Eisenbud, D., Neumann, W. D.: Three-dimensional link theory and invariants of plane curve singularities, Ann. Math. Studies 101, Princeton University Press, Princeton N.Y. (1985).
\[Li\] Lipman, J.: Rational singularities, with applications to algebraic surfaces and unique factorization, Publications Mathématiques I.H.E.S. 36, 195–280 (1969).
\[Se\] Segre, B.: Forme differenziali e loro integrali, Docet, Roma (1956).
\[Su\] Suzuki, M.: Proprietes topologiques des polynomes de deux variables complexes et automorphismes algébrique de l’espace $`𝐂^2`$, J. Math. Soc. Japan 26, 241–257 (1974).
|
no-problem/9904/astro-ph9904154.html
|
ar5iv
|
text
|
# Diffuse Ionized Gas in Edge-on Spiral Galaxies: Extraplanar and Outer Disk H𝛼 Emission1footnote 11footnote 1Observations made with the Burrell Schmidt of the Warner and Swasey Observatory, Case Western Reserve University.
## 1 Introduction
Spiral galaxies contain a widespread layer of ionized Hydrogen, known as diffuse ionized gas (DIG, also called WIM for warm ionized medium). The properties of this component of the interstellar medium (ISM) are important for many aspects of galactic research, such as the influence of massive stars on the ISM, the porosity of the ISM, and the disk-halo connection. The Reynolds layer, as DIG in the Milky Way is known, has a filling factor of at least 0.2, and accounts for nearly all of the mass of ionized gas, equal to about 30% of the HI mass. DIG is less dense than HII regions ($`n_e`$ $``$ 0.2 cm<sup>-3</sup>, compared to 10<sup>2</sup> to 10<sup>4</sup> cm<sup>-3</sup> in HII regions) but has a similar temperature ( $``$ 8000K). An excellent review of the properties of the Reynolds layer can be found in Reynolds (1990).
The \[SII\] 6717+6731Å to H$`\alpha `$ ratio is higher in the DIG relative to HII regions, while \[OIII\] 5007Å to H$`\alpha `$ is lower. In the Reynolds layer and in M31 the \[NII\] 6548+6584Å to H$`\alpha `$ ratio is about the same as in HII regions (Reynolds 1989; Greenawalt, Walterbos, & Braun 1997), while for NGC 891, the best studied edge-on galaxy, it appears to be higher in the DIG (Rand 1997a). These ratios can result from photoionization by a radiation field with a low ionization parameter $`U`$, the ratio of the photon density to the gas density (Mathis 1986; Domgörgen & Mathis 1994). However, transporting ionizing photons from HII regions to the DIG is still a problem, because of the long pathlengths that ionizing photons have to travel. In order to remain ionized, the Reynolds layer requires at least 15% of the Lyman continuum photons from OB stars in the Galaxy, equal to the amount of energy produced in supernovae. Studies of external galaxies suggest the requirement is even higher, consistently 30 to 50% (Walterbos & Braun 1994; Hoopes, Walterbos, & Greenawalt 1996; Ferguson et al. 1996; Greenawalt et al. 1998). Whether this energy is leaking out of density-bounded HII regions, or can be provided by field OB stars is still an open question. Recent spectroscopic observations have challenged the photoionization models, including lower HeI 5876Å than predicted (Reynolds & Tufte 1995, Rand 1997b) indicating a softer ionizing spectrum than expected, higher \[OIII\] 5007Å in some galaxies than in others (Wang, Heckman, & Lehnert 1997; Martin 1997) and rising \[OIII\]/H$`\alpha `$ ratio with height in the halo of NGC 891 (Rand 1998). Shock ionization by supernovae is another source which likely plays a role at some level and may explain the anomalous \[OIII\] ratios, but it cannot provide enough energy to ionize the bulk of the DIG.
The Reynolds layer is vertically extended, with a scale height of 900 pc (Reynolds 1990). This distribution is quite different from that of OB stars, which have a scale height closer to 100 pc. H$`\alpha `$ imaging of external edge-on galaxies has revealed the presence of DIG with varying properties. The first galaxy to be studied was NGC 891, which has a very thick DIG layer that has been traced as far as z=3.5 kpc in imaging (Rand, Kulkarni, & Hester 1990, hereafter RKH; Dettmar 1990; Pildis, Bregman, & Schombert 1994), and z=5 kpc with spectroscopy (Rand 1997b). Several other galaxies possess smaller DIG layers, and others show very little extraplanar emission (Rand, Kulkarni, & Hester 1992; Rand 1996; Walterbos 1991). Filamentary structure is visible in the extraplanar emission of several galaxies, indicating an active disk-halo connection and illustrating a probable relation between DIG and star formation.
Here we present deep H$`\alpha `$ images of 5 edge-on galaxies. Some properties of these galaxies are given in Table 1. We attempt to address several questions with these images. First we would like to know the extent of the DIG layers in galaxies down to levels fainter than previously studied. The DIG in NGC 891 may extend even farther into the halo at very low surface brightness, and galaxies that do not show a bright DIG layer may in fact have faint extraplanar emission. This is important for understanding the structure of galaxies and their gaseous halos. Second, the correlation of DIG properties with other parameters of galaxies, such as star-formation rate or Hubble type, can be a clue toward the ionization source of the DIG, as well as the mechanism for cycling gas into the halo. The strengths of other emission lines can also provide information on the ionization source through comparison with models (i.e. Mathis 1986; Domgörgen & Mathis 1994). For this we obtained \[SII\] and \[OIII\] images NGC 4631.
Another goal of this project is to investigate the possibility of detecting H$`\alpha `$ emission from the outer disks of galaxies. Such emission is expected to occur when neutral Hydrogen in the outer disk and halo is ionized by the metagalactic radiation field (Silk & Sunyaev 1976). The observation of a sharp cutoff in the HI disks of several galaxies (NGC 3198 by Van Gorkom 1991; M33 by Corbelli, Schneider, & Salpeter 1989) provides indirect evidence that the outer disk gas may be ionized. Maloney (1993), Dove & Shull (1994) and Corbelli & Salpeter (1993) have modeled the situation, and they predict the emission to be very faint, from 0.2 pc cm<sup>-6</sup> in emission measure down to 0.05 pc cm<sup>-6</sup> (Maloney 1993). Fabry-Perot observations have pushed the observational limits close to the theoretical predictions (Bland-Hawthorn, Freeman, & Quinn 1997, Vogel et al. 1995). We obtained very deep images of two edge-ons, NGC 3003 and UGC 9242, for the purpose of searching for emission from the outer disk. Direct detection of ionized gas would allow the determination of the strength of the metagalactic radiation field at wavelengths below 912 Å, and would have important implications for cosmological models, for prospects of measuring rotation curves of galaxies at large radii, and for our knowledge of the structure of spiral galaxies.
The layout of this paper is as follows. In section 2 we detail the observations and data reduction techniques. In section 3 we describe the H$`\alpha `$ morphology of the galaxies in our sample. In section 4 we examine the DIG in these galaxies, including the vertical extent, and contribution to the total H$`\alpha `$ luminosity. In section 5 we discuss the \[OIII\] and \[SII\] images of NGC 4631. In section 6 we discuss the outer disk emission in NGC 3003 and UGC 9242. Section 7 contains a discussion of the implications of our results.
## 2 The Data
### 2.1 Observations and Data Reduction
The images discussed in this paper were obtained during several different observing runs at KPNO. A log of the observations is given in table 2. For all of the datasets we removed the bias level and bias structure using standard methods in IRAF. Twilight flatfields were combined into a super flatfield which was used to remove gain variations. In some cases (noted below) the galaxies were observed using the shift and stare technique, where the telescope is moved between exposures to place the galaxy on different regions of the chip. This allows the use of the image frames to produce a night sky flatfield, if the galaxy is small enough in the field of view of the telescope. We constructed a night sky flatfield by median combining the galaxy images after editing out the galaxy and foreground stars. The resulting image was heavily smoothed and then applied to the original images after flattening with twilight flatfields.
NGC 891 was observed with the Burrell-Schmidt telescope at KPNO in both H$`\alpha `$ and narrow-band continuum filters. Four pointings were observed and combined into a mosaic. The large field of view of the Schmidt (1.15) allowed us to make a night sky flatfield even though NGC 891 is a relatively large galaxy (12). The final flatfielding accuracy is about 1% across the images. NGC 891 was also observed with the 0.9 meter at KPNO, in H$`\alpha `$ and narrow-band continuum. Two pointings (out of 4 planned) were observed. NGC 891 is too big in the 23 field of view of the 0.9 meter to use the images to make a night sky flatfield, but five blank sky images were taken through the H$`\alpha `$ filter, with the telescope moved slightly after each exposure. These were combined to make a night sky flatfield which was then applied to the H$`\alpha `$ images. The continuum image was already flat to better than 1% without the use of a night sky flatfield.
We obtained H$`\alpha `$ and narrow-band continuum images of NGC 3003 and UGC 9242 using the 0.9m telescope at KPNO. All of the images used were taken during photometric conditions. The final mosaics consist of nine pointings for each galaxy. A few of the images of UGC 9242 had to be discarded due to large diffraction spikes from a bright star off of the image. The 23 field of view is large enough compared to these galaxies that the object images could be used to make a night sky flatfield, which provided a flatfielding accuracy for NGC 3003 of about 0.5%, and probably much better than this in the region the galaxy covers. The night-sky flat alone did not produce satisfactory results for the images of UGC 9242, but a second night sky flatfield made from observations of blank sky worked well on the remaining structure. The flatfielding uncertainty after this correction was about 0.5%, and as with NGC 3003 it is better over smaller scales.
NGC 4244 was observed at the KPNO 0.9 meter, during non-photometric conditions. Three 30 minute exposures in the same position were combined to produce the final image. NGC 4631 was also observed at the KPNO 0.9 meter in non-photometric conditions. A different detector was used, which provided a 6.6 field of view. Three pointings were required to cover the 14.3 length of NGC 4631, and these were combined into a mosaic during the reduction. The small field of view and necessity of mosaicing make it difficult to match background levels, hence flatfielding uncertainties may be larger than for the other targets. Each pointing consists of approximately 1.4 hours, with two pointings overlapping in the central field (roughly 7 arcminutes). Images of the central field were also obtained in \[OIII\] and \[SII\] filters.
The 0.9 meter observations of NGC 891, NGC 3003, and UGC 9242 were calibrated using observations of spectrophotometric standard stars. The Schmidt observation of NGC 891 and the 0.9 meter observations of NGC 4244 and NGC 4631 were calibrated using the R-magnitude of the galaxy. This was scaled by the response and transmission of the continuum filter and used to calibrate the continuum image. Then the stars in the line image were scaled to the continuum, taking into account the differences in the response of the filters. The \[OIII\] image of NGC 4631 was calibrated in a similar manner using the V magnitude of the galaxy. For convenience, all of the H$`\alpha `$ images were converted to emission measure as if no \[NII\] was transmitted by the filter. The wide filter used on NGC 4631 contains the \[NII\] lines, but the narrow filters used for the other galaxies transmit very little \[NII\]. The \[NII\]/H$`\alpha `$ ratio has been observed to vary from 0.6 to 1.1 in the DIG of NGC 891 (Rand 1997a). At the highest value, \[NII\] transmitted by the narrow filters contributes only 10$``$20% of the observed H$`\alpha `$ flux. For NGC 4631 we correct for \[NII\] where it is relevant.
### 2.2 Continuum Subtraction
One of the most uncertain procedures in analyzing emission line images is the removal of the underlying stellar continuum. Subtracting too little continuum will leave a faint background that can be interpreted as diffuse line emission, while subtracting too much continuum can remove a layer of real diffuse emission. These concerns can be minimized by observing the continuum close in wavelength to the line, but varying and unknown H$`\alpha `$ absorption produced by the stars in the galaxy makes a perfect subtraction difficult to achieve.
As a first determination of the continuum scale factor, we measured the fluxes of foreground stars in both images, and computed the factor needed to make them equal (for galaxies that were calibrated using the R magnitude, the fluxes are made equal during calibration). This relies on the foreground stars being of similar spectral type as the stars in the galaxy, which is not necessarily the case. We then visually inspected the images to be sure that there are no negative regions, and that there is no obvious component of the stellar continuum left in the image. For all of the galaxies the scale factor derived from foreground stars was satisfactory, except the \[OIII\] image of NGC 4631. For this image we adjusted the scale factor to achieve the best subtraction. To quantify the dependence of our analysis on the continuum subtraction, we vary the scale factor by $`\pm `$ 3%. At this level it is usually obvious that the continuum is incorrectly subtracted. Our results throughout this paper will include the variation in the continuum subtraction in the total uncertainty.
### 2.3 Scattered Light
Another potential problem with analyzing faint emission is the possibility of scattered light. Telescope optics produce halos around point sources, which could be mistaken for diffuse emission. In addition, dust within the galaxy itself may scatter light from HII regions, producing a halo of scattered light around HII regions. Walterbos & Braun (1994) concluded that scattered light from within the galaxy is not a major component, based on the spectrum of the DIG. The \[SII\] to H$`\alpha `$ ratio is high in the DIG compared to HII regions, which would not be expected if the DIG were actually scattered light since the lines are very close in wavelength. Ferrara et al. (1996) modeled the scattering of HII region light by dust in the halo of NGC 891, and found it contributed only 10% of the DIG emission at 600 pc off the plane.
Methods of determining the contribution of scattered light from optics have been explored by Walterbos & Braun (1994) and Hoopes et al. (1996), and methods for correcting for scattered light have been used by RKH. Following these methods, we determined the possible contribution of scattered light from bright foreground stars in all of our images. We determined the radius in which 95% of the energy from the star is encircled, shown in table 2. Scattered light from HII regions will appear as halos around the regions to this radius, with only 5% of the light scattered further. We find that DIG extends further than this in all 5 galaxies, at a much higher level than 5%.
Scattered light can also affect the determination of the scale height of emission above the plane. To correct for this we deconvolved the images with the stellar point spread function (PSF). The PSF was measured using 10 stars in both the line and continuum images (except NGC 4631, where only 5 stars were suitable). The images were then deconvolved with the PSF using the Lucy-Richardson deconvolution algorithm with the task LUCY in IRAF. This was done before calibration and continuum subtraction. Vertical profile fitting was done on the deconvolved images (see section 4.1).
## 3 Results for Individual Galaxies
The H$`\alpha `$ images are shown in figures 1$``$5. The H$`\alpha `$ luminosities of the galaxies are listed in table 3. The correction for Galactic extinction (Burstein & Heiles 1984) is negligible for all of the galaxies except NGC891. The value listed in the table has been corrected; the luminosity before correcting for extinction is 2.8 $`\times `$10<sup>40</sup> erg s<sup>-1</sup>. No correction for internal extinction was made. The uncertainty given in the table is due to varying the continuum subtraction by $`\pm `$ 3%. Our H$`\alpha `$ luminosity for NGC 891 agrees well with Rand et al. (1992). Our flux for NGC 4631 is about 6% higher than that given in Rand et al. (1992), but our image contains \[NII\] emission as well, so our H$`\alpha `$ flux is actually lower. There are no published luminosities of NGC 3003, NGC 4244, and UGC 9242 with which to compare. NGC 4631 and NGC 4244 were observed during non-photometric conditions.
In this section we discuss the general appearance of the H$`\alpha `$ images. The H$`\alpha `$ morphology of several of the galaxies in our sample have been discussed in great detail elsewhere, so we will keep our description brief. These include NGC 891 (RKH; Dettmar 1990; Pildis, Bregman, & Schombert 1994) and NGC 4631 (Rand et al. 1992; Golla, Dettmar, & Domgörgen 1996). NGC 4244 was discussed briefly by Walterbos (1991) and Walterbos & Braun (1996). NGC 3003 and UGC 9242 have not previously been imaged in H$`\alpha `$.
### 3.1 NGC 891
In figure 1 we show both the Schmidt and the 0.9 meter continuum subtracted H$`\alpha `$ images of NGC 891. Vertically extended emission is clearly present. The DIG layer can be traced to more than 3 kpc away from the plane in these figures. The emission seems fairly uniform, but does appear brighter and more extended near two large HII regions on the east side of the disk. In the higher resolution 0.9 meter image the extraplanar emission begins to show filamentary structure, although it is not as pronounced as in NGC 4631 (see below). There is a correlation with the brightest star forming regions, with the brightest DIG near bright HII regions in the inner disk. Rand (1997a) noted that the bright filaments almost always connect to an HII region in the disk.
### 3.2 NGC 3003
The morphology of this galaxy (figure 2) suggests that it may be disturbed, in that the spiral arms appear asymmetric. There is a small dwarf galaxy (or possibly a background galaxy) at the lower right edge of figure 2, but other than this there are no close companions visible in our images, which cover an area of 300 $`\times `$ 300 kpc at the distance of NGC 3003. However, just outside of our field of view is the small spiral NGC 3021. These two galaxies are separated by about 200 kpc in projected distance, and about 50 km s<sup>-1</sup> in velocity (Tully 1988). The total H$`\alpha `$ luminosity of 1.18 $`\times `$ 10<sup>41</sup> erg s<sup>-1</sup> indicates that active star formation is ongoing. This H$`\alpha `$ luminosity is about an order of magnitude higher than for the starbursts NGC 253 and M82. In fact, of the galaxies observed by Young et al. (1996), one of the largest sample of galaxies observed in H$`\alpha `$ in the literature, only six of the 120 spirals have higher observed H$`\alpha `$ luminosities than NGC 3003 (not corrected for extinction). The H$`\alpha `$ image shows a bright nucleus, and many bright HII regions, including a very bright region in the outer western part of the disk. A projected spiral arm protrudes below the galaxy in the image, with a bright HII region at the end. The H$`\alpha `$ emitting disk is about 35 kpc across. The galaxy is not quite edge-on, so it is difficult to make any statements about the vertical extent of the DIG, though there is pervasive diffuse emission in the disk.
### 3.3 NGC 4244
This is a nearby galaxy (3.1 Mpc), so we can resolve a great deal of detail in the ionized disk. The HII regions distribution is extended in the vertical direction, a result perhaps of a weaker disk potential (Olling 1996). There are a few HII regions as far as 700 pc above the midplane. DIG is clearly visible in the disk, spread between several bright HII regions. It is immediately obvious, however, that an extensive DIG layer such as in NGC 891 is absent in this galaxy. Some short filaments exist in the central regions of the galaxy, but overall the DIG layer is confined to the disk. The galaxy appears quiescent in H$`\alpha `$, and the low H$`\alpha `$ luminosity and low FIR surface brightness imply that there is relatively little star formation occurring. Most of the HII regions are small and faint, except for two bright complexes on either end of the disk. The absence of long bright filaments, such as those seen in NGC 891 and NGC 4631, imply that the galaxy also lacks any visible disk-halo interaction.
### 3.4 NGC 4631
The H$`\alpha `$ +\[NII\] image of this galaxy is shown in figure 4. An overlay of H$`\alpha `$ and x-ray emission for this galaxy, based on our data, can be found in Wang et al. (1995). There appear to be significant flatfielding uncertainties on the south side of the disk, near the bottom edge of the image, so we restrict our analysis to the north side. The image shows a disturbed disk, most likely due to an interaction with its companion galaxies NGC 4627 and NGC 4656. The disk appears to be actively forming stars, and the disk-halo interface is also very active in this galaxy. Rand et al. (1992) pointed out two bright vertical “worms” of emission east of the nucleus on the north side of the disk. We note that these worms are connected to longer, fainter filaments, and the western worm appears to curve back around toward the disk. There is an even larger loop of emission which surrounds the two worms (indicated by the arrows in figure 4). It begins just east of the easternmost worm and extends 3.5 kpc into the halo. Unfortunately there is a seam in the image where the loop might reconnect to the disk. The surface brightness at the top of the loop is about 12 pc cm<sup>-6</sup>. The large loop is barely visible in the H$`\alpha `$ image presented by Rand et al. (1992, their figure 5). What appears to be another filament about 3 kpc east of the giant loop in our image is actually another seam. There is significant extraplanar emission in the form of discrete features such as the loops and worms, and a smoother component can be traced up to about 2 kpc on the north side. Donahue, Aldering, & Stocke (1995) detected a faint halo extending 16 kpc from the plane of NGC 4631, with a maximum brightness of 0.69 pc cm<sup>-6</sup> per square arcsecond. We could not confirm the detection because of the smaller field of view of our observations. This emission is probably related to the high star formation and disturbed nature of this galaxy resulting from the tidal interaction, as it is too bright to be caused by the metagalactic ionizing radiation field.
### 3.5 UGC 9242
This galaxy appears to be close to exactly edge on (figure 5). The continuum image shows a very thin and symmetric disk. The H$`\alpha `$ emitting disk is about 35 kpc across. It has a bright nucleus in H$`\alpha `$ and several bright star forming regions. There are two plumes visible on the north side of the nucleus. They can be traced about 1.8 kpc from the midplane, with typical emission measures ranging from 20 just above the disk to about 5 at the highest point. These are probably associated with star formation in the nucleus, but it is not known whether these are two separate outflows, or whether they are the brightened edges of a conical outflow such as that seen in NGC 253 (Heckman, Armus, & Miley 1990). If it is a conical outflow, it is 1200 pc wide at the top. There is also some structure on the other side of the nucleus, which also may be 2 spurs of emission, shorter and weaker than the counterparts on the north side. This suggests a double sided outflow, with a morphology similar to the “H” shaped filaments seen in NGC 3079 (Hester et al. 1990; Veilleux, Cecil, & Bland-Hawthorn 1995) and NGC 4013 (Rand 1996). The bright nucleus is about 1 kpc across, with an H$`\alpha `$ luminosity in a 12<sup>′′</sup>(about 1 kpc) diameter aperture of 4.2 $`\times `$ 10<sup>39</sup> erg s<sup>-1</sup>, comparable to the brightest HII regions in most galaxies (Kennicutt 1988), and in fact equal to the total H$`\alpha `$ luminosity of NGC 4244. Except for these filaments there is very little obvious extraplanar emission, and there is no evidence for disk-halo interaction beyond the nucleus, but like all of the galaxies in our sample the disk is filled with DIG.
## 4 The Diffuse Ionized Gas
### 4.1 Vertical Extent
Walterbos & Braun (1996) compared the appearance of the DIG layers in three of the galaxies in our sample, NGC 891, NGC 4244, and NGC 4631. Figure 6 is a more exact comparison, incorporating UGC 9242. The images are shown on the same spatial scale, brightness scale, and logarithmic stretch. The comparison shows the range of DIG morphologies. The most prominent example is the smooth, bright, extended layer in NGC 891. The patchy, filamentary layer in NGC 4631 may be related to recently enhanced star formation as a result of an encounter. The most common appearance of the DIG may be more similar to the weaker layers in NGC 4244 and UGC 9242. The inclination of NGC 3003 hinders our analysis of the extraplanar emission.
Following RKH and Rand (1996), we have attempted to characterize the DIG as an exponential layer. We fit the vertical profile averaged over the central 10 kpc to increase the signal to noise. In order to avoid including light from HII regions in the fit, we excluded emission from $`|z|`$ 300 pc. NGC 4244 has a thicker disk of HII regions, so we excluded emission from $`|z|`$ 500 pc. We tried fits using a single exponential function as well as fits using two exponentials. The parameters of the best fits for three of the galaxies are given in table 4. The vertical profile of NGC 3003 suggests that it may also possess faint extraplanar emission, but the lower inclination of this galaxy make the detection uncertain, as such emission may arise from the outer part of the disk. We do not address it further here, nor do we attempt the analysis for NGC 4631 due to its disturbed nature and the poor flatfielding of the image. Note that the scale heights given are for the surface brightness (emission measure). The electron scale height is twice the emission measure scale height, assuming the emission is only H$`\alpha `$. If much \[NII\] is included in the filter, a rising \[NII\]/H$`\alpha `$ ratio in the halo such as those observed in NGC 891 (Rand 1998) and NGC 4631 (Golla, Dettmar, & Domgörgen 1996) could make the scale heights appear larger than they really are. The electron scale height for the Reynolds layer is 900 pc (Reynolds 1990).
Previous imaging of NGC 891 (RKH) has traced the extraplanar DIG to as far as 3.5 kpc from the plane of the galaxy. Spectroscopy (Rand 1997b) has traced the emission even further, to at least 5 kpc. Figure 7 shows a vertical profile of the deconvolved Schmidt H$`\alpha `$ image of NGC 891, averaged over the central 10 kpc. The emission can be traced as far as 5 kpc off the plane in both directions. Thus we confirm the spectroscopic detection. There is a suggestion that the emission continues even further, to as far as 7 kpc, on the west side of the disk. However, foreground stars in this part of the image make this uncertain. Although the two fits look quite similar in the linear scaling, the logarithmically scaled plot shows that two exponential functions better fit the faint component of the profile. For comparison, the stellar thin disk of NGC 891 has a scale height of 425 pc, and the stellar thick disk has a scale height of about 1.9 kpc from surface photometry (Morrison 1999; Morrison et al. 1997).
Even averaged over 10 kpc the extraplanar emission in NGC 4244 is very weak (figure 8). Using two exponential functions, we find one component with a very small scale height, about a factor of 2$``$3 lower than that of the Galaxy, and a fainter, broader component. The faint component may be due to a flatfielding problem, such as low level vignetting in the image. If we fit only one exponential the scale heights are closer to that of the Galaxy (450 pc). The two-exponential fit is statistically better, but both functions describe the emission fairly well. NGC 4244 does not appear to possess a thick stellar disk (Morrison 1999).
Figure 9 shows that the extraplanar emission in UGC 9242 cannot be described by a single exponential layer. Using two exponential functions, the bright component of the vertical profile is well described by a relatively low scale-height exponential, as expected from the appearance of the H$`\alpha `$ image. However, there is a faint tail of emission which is clearly visible out to 3$``$4 kpc. The scale height of this component is similar to and perhaps even larger than the scale height of the Reynolds layer. As noted, UGC 9242 has bright filaments expending from the nucleus which may hamper the fit. In figure 10 we show the vertical profile of the central 20 kpc with the central 3 kpc excluded so as to remove the contribution of these filaments. The parameters of the model fits are shown in table 4. A single exponential still cannot fit the emission. The scale heights for the bright component are similar to the 10 kpc fit, but the faint components are even more extended. The high-z tail appears to extend well past 5 kpc.
In figure 11 we show the profiles of NGC 891, UGC 9242, and NGC 4244 overplotted with the best fit model for each. Although NGC 891 has a much brighter DIG layer than UGC 9242, extraplanar emission in both galaxies actually reaches comparable distances above the disk. NGC 4244 clearly has no extraplanar emission comparable to the other two galaxies. The logarithmically scaled plots clearly show that two exponential components make up the extraplanar DIG in NGC 891 and UGC 9242, and possibly also in NGC 4244.
Table 5 shows some parameters derived from the model fits. It is necessary to know the diameter of the DIG cylinder to derive these properties, so we estimate it from the H$`\alpha `$ images. The emission measure in a column perpendicular to the disk is given for each of the two components. We call the lower scale height component “thick disk” DIG (so as not to be confused with the HII region thin disk), and the higher scale height component “halo” DIG. The two components contribute nearly equally to the total perpendicular emission measure in NGC 891, while the halo components is weaker in the other two galaxies. The Galaxy resembles UGC 9242 in terms of total perpendicular emission measure and surface density, but keep in mind that the Galactic DIG parameters are derived for the solar neighborhood, while for the external galaxies these DIG parameters apply closer to the center where it is brighter. Following RKH, a constant filling factor of $`\varphi `$=0.25 is assumed when calculating the surface density, although it may rise in the halo (Kulkarni & Heiles 1988). The surface densities include helium at solar abundance.
### 4.2 Diffuse Fractions
Another way to compare the DIG in different galaxies is through the diffuse fraction, which is the contribution of the DIG to the total H$`\alpha `$ luminosity. We us the same method used by Hoopes et al. (1996) for isolating diffuse emission from HII region emission. This technique compensates for the varying brightness of the DIG layer with radius in the galaxy and galaxy inclination (the reasons a simple isophotal cut at a given surface brightness fails). The method involves subtracting smoothed version of the image from the original to remove the diffuse component, then making a mask on the resulting image which removes pixels greater than a certain value and replaces them with zero. The mask is applied to the original image to remove HII regions emission. The technique was tested by applying it the M31 images of Walterbos & Braun (1992, 1994). We chose a scale of 900 pc for the median box used to smooth the image, and a cut level of 50 pc cm<sup>-6</sup> to make the mask, because these parameters resulted in diffuse fractions similar to those that were found by manually cataloging and removing the HII regions by hand in M31. Note that the cut level of 50 pc cm<sup>-6</sup> in the image after subtracting the smoothed version does not correspond to the same level in the original image.
We applied this technique to the galaxies in our sample. The measured diffuse fractions are given in table 3. Our H$`\alpha `$ flux NGC 4631 is low compared to Rand et al. (1992), so to test the effects of a possible calibration error we multiplied the image by 1.3 and recalculated the diffuse fraction using the same method, which gave 37 $``$ 41%, very similar to the original value. The diffuse fractions for NGC 3003, NGC 4244, NGC 4631, and UGC 9242 all fall within the 30 to 50% range found for face on galaxies (Hoopes et al. 1996; Greenawalt et al. 1998). NGC 891, however, shows a much higher fraction, 83$``$86%. The diffuse fraction for NGC 891 was measured on the 0.9 meter image. Applying the technique to the lower resolution Schmidt image gives a fraction of 90%. The difference between the fractions measured on the Schmidt and 0.9 meter images of NGC 891 is most likely due to unresolved HII regions in the lower resolution Schmidt image being counted as DIG.
Comparison of the diffuse fractions found in edge-on galaxies to those found in face-on systems is not straightforward. Since the HII region layer is confined to the midplane where the dust density is the highest, one might expect to measure a higher diffuse fraction in edge on galaxies, as the HII regions would be affected more by the dust disk than they would in face-on galaxies. The DIG extends well out of the dust disk, especially in NGC 891, so much of it is less obscured. This may explain the high ratio in NGC 891. As the only Sb galaxy in our sample, it may have a higher metal content than the later type spirals, and the higher abundance of dust may obscure more of the disk HII region emission. A comparison of the H$`\alpha `$ surface brightness with the FIR surface brightness supports the idea that NGC 891 contains more dust than the other galaxies in our sample. The FIR luminosity is an upper limit to the star formation rate, as an unknown fraction arises from dust heated by older stars. The H$`\alpha `$ luminosity gives a lower limit to the star formation rate, since it is affected by extinction. Table 3 shows that the L/L<sub>FIR</sub> ratio for NGC 891 is low relative to the other galaxies in the sample, implying higher extinction. Interestingly, NGC 4244 has a high value for this ratio, implying that it may be dust poor.
## 5 Line Ratios in NGC 4631
Figure 12 shows a subsection of the H$`\alpha `$ image of NGC 4631 and the \[OIII\]/H$`\alpha `$ and \[SII\]/H$`\alpha `$ ratio images. The figure shows a portion of the disk east of the bulge of the galaxy. In the \[OIII\]/H$`\alpha `$ image the cores of HII regions show a high ratio, while the DIG shows a lower ratio. Not all HII regions have high \[OIII\]/H$`\alpha `$ ratios, however, and in those that do only the central core has an elevated ratio, except for one region described below. The \[SII\]/H$`\alpha `$ image shows the opposite trend, with the DIG showing a higher \[SII\]/H$`\alpha `$ ratio than HII regions.
Applying the mask made on the H$`\alpha `$ image of NGC 4631 to the \[OIII\] and \[SII\] images allows us to investigate the line strengths in the DIG and in HII regions. The continuum in the bulge did not subtract well from the \[OIII\] image, so we omitted that region and determined the ratio in the disk east and west of the bulge. The global \[SII\]/H$`\alpha `$ and \[OIII\]/H$`\alpha `$ ratios are given in table 6. The \[SII\]/(H$`\alpha `$ +\[NII\]) ratio is elevated in the DIG, while \[OIII\]/(H$`\alpha `$ +\[NII\]) is lower. Note that the H$`\alpha `$ filter also contains a contribution from the nearby \[NII\] lines, which must be taken into account when comparing with other measurements. In the disk \[NII\]/H$`\alpha `$ varies from 0.1 to 0.2 in HII regions, and 0.3 to 0.5 in the DIG (Golla, Dettmar, & Domgörgen 1996). Thus when comparing to measurements made without \[NII\], the observed \[OIII\]/H$`\alpha `$ and \[SII\]/H$`\alpha `$ ratios for NGC 4631 could be 1.1 to 1.2 times higher in HII regions, and 1.3 to 1.5 times higher in the DIG. Another important point is that the images are not corrected for internal extinction. Greenawalt et al. (1997) found that the extinction in HII regions is higher than that in the DIG in M31. If this is also true in NGC 4631 then correcting for extinction would reduce the \[OIII\]/H$`\alpha `$ ratio in HII regions more than in the DIG. The large uncertainties at high z distance, due to low signal in the \[OIII\] and \[SII\] images, prevent us from investigating the behavior of the line ratios away from the plane.
There is an extended region on the SE side of the disk where the \[OIII\]/H$`\alpha `$ ratio reaches over 1.0 (near the bottom of the images in figure 12). This is comparable to the ratios seen in the cores of HII regions, but the high \[OIII\] gas appears more extended than other HII regions, and lies in a region where the H$`\alpha `$ emission is about 180 pc cm<sup>-6</sup>. By contrast, the cores of HII regions which show high \[OIII\] have emission measures of about 1500 pc cm<sup>-6</sup>. This is the same location in the galaxy where Rand & van der Hulst (1993) found a large HI supershell. The shell has been modeled as a collision of a high velocity cloud with the disk of NGC 4631 (Rand & Stone 1996). The region of high \[OIII\] emission is about 650 pc in diameter, but borders on a large HII region which also shows high \[OIII\]/H$`\alpha `$ , so the true extent is difficult to measure. The ratios in a 370 pc wide vertical slice through the region are shown in figure 13. The location of this slice is marked in figure 12. The \[SII\]/H$`\alpha `$ ratio in this region is about 0.2, close to the value in HII regions. The gas here may be shock ionized, which can produce a high \[OIII\]/H$`\alpha `$ ratio (Shull & McKee 1979; Dopita & Sutherland 1995). Shock-ionized gas can also have low \[SII\]/H$`\alpha `$ if the density is high enough to collisionally de-excite S<sup>+</sup> (Dopita & Sutherland (1995). This region was not included in the determination of the global \[OIII\]/H$`\alpha `$ and \[SII\]/H$`\alpha `$ ratios.
## 6 Limits on Emission from the Outer Disk
One of the goals of this project was to detect or set limits on H$`\alpha `$ emission from the outer disk. These observations were driven by the recent discovery of sharp edges to the HI disks in nearby galaxies such as NGC 3198 (van Gorkom 1991). A possible explanation for these edges is that the outer Hydrogen is ionized, making it undetectable to 21 cm observations, as suggested by Silk & Sunyaev (1976). There are no ionizing stars at these large radii, and since the HI typically extends several kpc past the optical or H$`\alpha `$ disk any ionizing radiation from within the galaxy would surely be absorbed by the intervening HI. Therefore the ionizing source must be extragalactic, and is thought to be the metagalactic ionizing radiation field produced by quasars and AGN.
This idea has been modeled (Maloney 1993; Corbelli & Salpeter 1993; Dove & Shull 1994) and the observed cutoff can be reproduced. A crucial test, however, is to directly detect the ionized outer disk. The models indicate that the emission would be extremely faint, as low 0.05 pc cm<sup>-6</sup>, but possibly as high as 0.2 pc cm<sup>-6</sup>, depending on the gas density, clumpiness, and the strength of the ionizing radiation field. This is fainter than imaging studies have reached in the past. Ionized gas was detected in the outer disk of NGC 253 (Bland-Hawthorn et al. 1997) at a level of 0.23 pc cm<sup>-6</sup>, using very sensitive Fabry-Perot observations. The emission would imply a metagalactic radiation field stronger than the current upper limit of 8$`\times `$10<sup>-23</sup>ergs cm<sup>-2</sup> s<sup>-1</sup> Hz<sup>-1</sup> sr<sup>-1</sup> (Vogel et al. 1995), and the authors argue that the gas is photoionized by disk OB stars which can see the warped outer disk.
With extreme care in flatfielding, it may be possible to detect this level of emission through deep narrow-band imaging. Recently Donahue et al. (1995) detected a very faint halo around NGC 4631. The surface brightness of this halo is as high as 0.69 pc cm<sup>-6</sup>, much brighter than that expected from the metagalactic radiation field, so the emission is most likely a result of the star formation activity in the galaxy or related to the gravitational interaction with its neighbors. We used similar imaging techniques in order to optimize the detection of faint emission. Our target galaxies (NGC 3003 and UGC 9242) were chosen to be at high galactic latitude so that they were not affected by many bright foreground stars or emission from the Reynolds layer. The sensitivity of the NGC 3003 image is about 2.0 pc cm<sup>-6</sup> per square arcsecond, and about 2.7 pc cm<sup>-6</sup> per square arcsecond for UGC 9242, much brighter than the strongest expected emission of 0.25 pc cm<sup>-6</sup> found by Maloney (1993). In order to reach fainter levels, we spatially averaged over an ever increasing area to lower the noise until the limit of the flatfielding accuracy was reached and no further increase in S/N with smoothing was apparent. The NGC 3003 image was binned into 25$`\times `$25 pixel (17.3<sup>′′</sup>$`\times `$17.3<sup>′′</sup>) boxes, which increased the sensitivity by a factor of 17.3 to 0.13 pc cm<sup>-6</sup>. The UGC 9242 could be averaged over 30$`\times `$30 pixel (20.7<sup>′′</sup>$`\times `$20.7<sup>′′</sup>) regions, increasing the sensitivity to 0.13 pc cm<sup>-6</sup>. The rms intensity deviation between the background levels in boxes in flat regions of the image was equal to or less than this limit, even for boxes separated by large distances. The spatial scale for these limits are 2 kpc for NGC 3003 and 2.6 kpc for UGC 9242.
The binned images show no obvious outer disk emission. In figure 14 we show the major axis profiles of the two galaxies from the binned images. UGC 9242 has a distinct hole on the west side (positive major axis distance in figure 14). This appears to be a flatfielding error at the edge of one of the images which went into the final mosaic. Aside from this, any outer disk emission is below the limits of our flatfielding accuracy. We also smoothed using a median filter, using the same size median box as the average box used above. The NGC 3003 image shows no evidence for outer disk emission, but the UGC 9242 image has some interesting features (see figure 15). Faint extraplanar emission is visible on both sides of the disk. On the east side of the disk there is a bright star, which may account for the emission in this region. On the west side, however, the stars are fainter and less likely to contribute much scattered light. On this side the emission ranges from about 2 pc cm<sup>-6</sup> at 2 kpc above the midplane to our limit of 0.13 pc cm<sup>-6</sup> at 8.5 pc above the midplane. It appears that the emission is centered towards the nucleus, which may imply a connection to the central starburst, but might also just reflect the distribution of halo gas.
The H$`\alpha `$ disk also appears more extended in the median smoothed image. The H$`\alpha `$ emission extends about 8 kpc past the southern edge of the optical disk (the left side of figure 15). The optical edge is at about -21 Mpc in figure 14. Several factors lead us to question the validity of this feature. The emission ranges in brightness from 0.16 to 0.20 pc cm<sup>-6</sup>, just barely above our detection limit, making this a 1$``$2$`\sigma `$ detection at best, and is not confirmed in the binned image. There is a background galaxy and some faint foreground stars which may affect the flatfielding in this region. There are other variations of similar magnitude near stars in the image. An example of this can be seen just below the extended disk in figure 15, where scattered light from a background galaxy and a group of several small stars created a spot in the smoothed image. Although stars are removed more cleanly during continuum subtraction in the deconvolved images than in the original images, there are still residuals due to effects such as a changing PSF across the field, leading to this further source of uncertainty for faint emission. If this emission were real, it would contradict the limits set by the NGC 3003 image, as well as other established upper limits on the metagalactic ionizing flux (see section 7.2).
## 7 Discussion
### 7.1 Halo Emission
We have detected the DIG layer in NGC 891 out to at least 5 kpc from the plane, and possibly as far as 7 kpc. NGC 891 has the brightest and largest DIG layer known, and it is not surprising to detect emission so far from the plane in a deep image. What is surprising is that UGC 9242, which shows much less extraplanar emission, still has a halo extending as far as 3 to 4 kpc, though significantly fainter than the NGC 891 halo. The image of UGC 9242 is of very high sensitivity, so it is possible that similar faint halos might be seen around other edge-ons even if they do not possess a bright DIG layer. We also confirm the existence of extraplanar emission in NGC 4631, although we could not have verified the halo claimed for NGC 4631 by Donahue et al. (1995). For the other galaxies in our sample, even in our very sensitive images of NGC 3003 and UGC 9242, we do not detect such a halo.
Excluding NGC 3003, which has too low an inclination to properly study the extraplanar emission, NGC 4631 and NGC 891 are the brightest in H$`\alpha `$, and they also show the most extraplanar emission of the sample. Although directly inferred star formation rates from the H$`\alpha `$ luminosity can be very unreliable for edge-on galaxies due to extinction, in a relative sense NGC 891 and NGC 4631 would appear to be the most actively star forming galaxies of the sample (again excluding NGC 3003). The extreme difference between these galaxies and NGC 4244, with its low H$`\alpha `$ luminosity and weak DIG layer, point to a link between active star formation and extraplanar emission, as discussed previously by Rand (1996). This conclusion is supported by the FIR luminosity (Rand 1996), which traces star formation more accurately than H$`\alpha `$ in edge-on galaxies, although some fraction of the FIR emission may stem from dust heated by general starlight, not OB stars. In table 3 we list the FIR luminosity normalized by the square of the disk diameter (D<sub>25</sub>), following Rand (1996). The galaxies with the highest current star formation have the most prominent extraplanar emission. This does not carry over to the diffuse fractions, however. In fact the diffuse fractions seem relatively constant with the exception of NGC 891, and this may be due to higher extinction in that galaxy. NGC 891 is the only Sb in the sample and may have a higher dust content than the rest of the sample (all Sc galaxies). The constancy of the diffuse fraction further reinforces the connection with star formation, since it essentially means that the DIG luminosity scales with the HII region luminosity. It is also interesting that there is not a more pronounced difference between face-on and edge-on diffuse fractions. This might imply that we are seeing most of the disk H$`\alpha `$ emission even in edge-on galaxies, with the exception of NGC 891. Diffuse fractions for more edge-on spirals are necessary to test this idea further.
The vertical profiles of NGC 891, UGC 9242, and possibly NGC 4244 are best described by two distinct exponential components, raising the possibility that more than one mechanism is responsible for creating H$`\alpha `$ halos. NGC 891 possesses a thick stellar disk, while NGC 4244 does not, which may suggest a connection between the mechanism which creates a thick disk and that which is responsible for the H$`\alpha `$ halo. However, a counter-example can be found in NGC 4565, which does possess a thick stellar disk (Morrison 1999), but has little extraplanar H$`\alpha `$ emission (Rand et al. 1992). An x-ray observation of NGC 891 (Bregman & Pildis 1994; Bregman & Houck 1997) revealed a halo of 10<sup>6</sup> K gas, with a distribution similar to the H$`\alpha `$ emission. The idea that hot supernovae-heated gas vented into the halo through chimneys is responsible for this emission led the authors to calculate whether the cooling of this hot gas could be the source of the H$`\alpha `$ emission. They found that the cooled gas mass was an order of magnitude too low to account for the mass of H$`\alpha `$ emitting gas. Rand (1997b) suggested that cooling gas may explain the more-extended halo component of the H$`\alpha `$ emission, while gas ionized by photons from OB stars leaking out of the disk is responsible for the brighter, less-extended component. The H$`\alpha `$ image of the DIG in UGC 9242 revealed a very extended halo component, but the scale height of the disk component is much lower than that of the similar component in NGC 891. Perhaps cooling gas could be responsible for a faint halo such as in UGC 9242, but some mechanism prevents the gas which makes up the disk component from reaching as high as in NGC 891. There is indication of an outflow into the halo from the nuclear filaments on both sides of the disk, which might provide a source of hot halo gas. Table 6 shows that the halo components in both UGC 9242 and NGC 4244 are much less prominent relative to the disk components than is true for NGC 891, where the two components are nearly equal. If cooling gas is responsible for the halo component, NGC 891 must have a much more active chimney mode than UGC 9242 and NGC 4244.
Recently several galaxies have been observed to have high \[OIII\]/H$`\alpha `$ ratios in the DIG, including NGC 891 (Rand 1998). The implication is that a single ionization source cannot be responsible for both the low \[OIII\] and high \[OIII\] emitting gas, so another source of ionization is necessary. As Rand (1998) and Wang & Heckman (1997) point out, shock ionized gas can have high \[OIII\]/H$`\alpha `$, so shocks may be an important mechanism for ionizing this component of the DIG. In NGC 891 the high \[OIII\] values are measured in the very high-z gas, suggesting that shock ionization or some other mechanism is important in the upper halo. In NGC 4631 we measure \[OIII\]/H$`\alpha `$ +\[NII\] ratios consistent with the DIG in several external galaxies (Greenawalt et al. 1997; Wang, Heckman, & Lehnert 1997), but higher than the Milky Way ratios. Photoionization models (i.e. Domgörgen & Mathis 1994) have been constructed which reproduce the Galactic values, so they underpredict the \[OIII\] in NGC 4631. The \[OIII\]/H$`\alpha `$ ratio seen in the DIG of M31 (Greenawalt et al. 1997) is similar to those we find for NGC 4631. Those authors found that a model adjusted so that the ionizing radiation is less diluted on average could reproduce the observed ratios. A similar situation may exist in NGC 4631, where the overall star formation is enhanced in the disk. It then appears that photoionization is the only ionization mechanism necessary for NGC 4631, in the disk at least. The sensitivity at high z in the \[OIII\] and \[SII\] images is not good enough to determine the ratio, so a trend such as that in NGC 891 may well exist.
### 7.2 Outer Disk Emission
Our efforts to detect the outer disk H$`\alpha `$ emission allow us to set an upper limit on the metagalactic ionizing radiation field. Maloney (1993) constructed a model of the ionization of the outer disk of NGC 3198 by the metagalactic radiation field. The model predicts a range of expected H$`\alpha `$ surface brightness for different values of the ionizing flux. An important result from the model was that the predictions are insensitive to galaxy parameters, so we can apply the results for NGC 3198 to the galaxies in our sample. The models assume that the gas is smoothly distributed; if the gas is clumpy the emission will be brighter.
Comparison of our upper limit on the H$`\alpha `$ emission of 0.13 pc cm<sup>-6</sup> for the outer disk of NGC 3003 with the predicted emission measure in Maloney (1993, see their figure 14) shows that the upper limit on the metagalactic ionizing flux J<sub>ν</sub> is between 8 and 12 $`\times `$10<sup>-23</sup>ergs cm<sup>-2</sup> s<sup>-1</sup> Hz<sup>-1</sup> sr<sup>-1</sup>. The calculations in Maloney (1993) are based on the observed sharp cutoff in the HI disk at large radii. The scenario is one in which the HI disk dips below a critical column density where it becomes optically thin to the metagalactic ionizing flux, going from a mostly neutral disk to a mostly ionized disk. Vogel et al. (1995) set a limit on extragalactic H$`\alpha `$ emission using Fabry-Perot observations of an intergalactic Hydrogen cloud. They placed a 2$`\sigma `$ limit on the metagalactic ionizing flux to 8$`\times `$10<sup>-23</sup>ergs cm<sup>-2</sup> s<sup>-1</sup> Hz<sup>-1</sup> sr<sup>-1</sup>, in agreement with the result presented here. Donahue et al. (1995) found that J<sub>ν</sub>$`<`$3.3$`\times `$10<sup>-23</sup>ergs cm<sup>-2</sup> s<sup>-1</sup> Hz<sup>-1</sup> sr<sup>-1</sup>, by imaging intergalactic Hydrogen clouds in H$`\alpha `$, more stringent than our limit.
A simple check of these numbers can be made by calculating the number of ionizing photons required to produce the observed H$`\alpha `$ emission. An emission measure of 0.13 pc cm<sup>-6</sup> requires 8$`\times `$10<sup>4</sup> photons cm<sup>-2</sup> s<sup>-1</sup>, assuming the HI is optically thick to ionizing photons and that 45% of ionizing photons result in H$`\alpha `$ photons (Case B). The current limit is 6$`\times `$10<sup>4</sup> photons cm<sup>-2</sup> s<sup>-1</sup> for two-sided incident ionizing flux (Vogel et al. 1995). This would imply that our limit is a factor of 1.3 higher, so J<sub>ν</sub> $`<`$ 11 $`\times `$10<sup>-23</sup>ergs cm<sup>-2</sup> s<sup>-1</sup> Hz<sup>-1</sup> sr<sup>-1</sup>. We have not quite reached the limits set by Fabry-Perot observations nor that set by H$`\alpha `$ imaging of Hydrogen clouds, so it is not surprising that we do not detect any outer disk emission from NGC 3003. It is important to remember that the conversion from H$`\alpha `$ emission measure to metagalactic ionizing flux assumes that no \[NII\] is present. The filter used here may transmit \[NII\] at the 10$``$20% level. This is normally a small percentage of H$`\alpha `$ , but it could be significant if the \[NII\]/H$`\alpha `$ ratio is very high in the outer disk, as was found by Bland-Hawthorn et al. (1997) for NGC 253.
In UGC 9242 there is an indication of outer disk H$`\alpha `$ emission at the 0.16 pc cm<sup>-6</sup> level. If this emission were a result of the metagalactic ionizing radiation field, it would require that J<sub>ν</sub>=17.0$`\times `$10<sup>-23</sup>ergs cm<sup>-2</sup> s<sup>-1</sup> Hz<sup>-1</sup> sr<sup>-1</sup>, higher than the Vogel et al. (1995) upper limits, and higher than the limit set by NGC 3003. The presence of several foreground stars and a background galaxy at this location in the image lead us to question whether this feature is real. It could be either the result of flatfielding errors, or it could be real emission but powered by some source other than the metagalactic ionizing radiation field.
We thank the KPNO staff for their help during these observing runs. We also thank B. Greenawalt for observing NGC 891, and M. F. Steakley for reducing the NGC 4244 and NGC 4631 images. We thank the referee, James Schombert, for useful comments which improved the presentation of the results. This research was supported by the NSF through grant AST-9617014, and by a Cottrell Scholar Award from Research Corporation. C.G.H. was supported by a grant from the New Mexico Space Grant Consortium.
|
no-problem/9904/astro-ph9904037.html
|
ar5iv
|
text
|
# A Plasma Instability Theory of Gamma-Ray Burst Emission
## 1 Introduction
A plasma instability theory for the prompt emission of gamma-ray bursts is presented. In this theory, a relativistic shell with $`\mathrm{\Gamma }1`$ passes through the interstellar medium. Two plasma instabilities, the filamentation instability and the two-stream instability, generate a magnetic field and heat the electrons to relativistic energies. The heated electrons emit synchrotron radiation in the radio to optical bands and synchrotron self-Compton radiation from the optical to gamma-ray bands.
This theory produces the observed prompt gamma-ray emission seen in all bursts, and the prompt optical emission seen in GRB 990123. The magnetic field generated by the filamentation instability is calculated from first principals. Lower limits on $`\mathrm{\Gamma }`$ and $`n_{ism}`$ arise from the requirement that the model efficiently radiate gamma-rays. The limit on density requires each gamma-ray burst to be surrounded by a medium that is optically thick to Compton attenuation. The limit on $`\mathrm{\Gamma }`$ suggests that there exists a class of transient that produces optical and ultraviolet emission but no gamma-ray emission.
One of the more interesting aspects of the theory is that the plasma instabilities cannot satisfy the Rankine-Hugoniot conditions, so a shock is not produced by these instabilities. As a consequence, the interstellar medium remains in place after passage of the relativistic shell. This permits the interstellar medium to interact with multiple relativistic shells to produce the complex time structure seen in gamma-ray burst light profiles.
## 2 Plasma Instabilities
Two instabilities arise when a plasma streams through a second plasma at a highly relativistic velocity. The first is the filamentation instability, while the second is the two-stream instability. Of these, the former has the higher growth rate.
The growth rate of the filamentation instability as measured in the shell rest frame is
$$\gamma _f^{}\frac{1}{2}\sqrt{\frac{4\pi e^2}{m}}n_{ism}^{\frac{1}{2}},$$
(1)
where $`n_{ism}`$ is the number density of the interstellar medium in the ISM rest frame, and $`m`$ is the mass of the filamenting plasma component of the ISM. The filamentation occurs for wave numbers perpendicular to the velocity vector that obey the inequality
$$k_{}>\omega _{p,e,shell}^{}/c.$$
(2)
In other words, the length scale of the filamentation is set by the electron plasma frequency of the shell as measured in the rest frame of the shell.
The filament is a magnetic pinch, with a toroidal magnetic field that is formed when the particles in the filament collapse to the center of the filament. The filament grows until its growth rate equals the bounce frequency of a particle across the magnetic pinch. This saturation defines the maximum magnetic field that can be generated. For both ions and electrons, the maximum field strength is the same
$$\frac{B^{\mathrm{\hspace{0.17em}2}}}{8\pi }=\frac{m_ec^2n_{ism}^2\mathrm{\Gamma }^2}{n_{shell}^{}}$$
(3)
One finds that for $`n_{shell}^{}/n_{ism}=\mathrm{\Gamma }`$, which is a natural value for the efficient emission of radiation, the magnetic field is $`B^{}=0.14\text{G}`$ when $`\mathrm{\Gamma }=10^3`$ and $`n_{ism}=1\text{cm}^3`$, and $`B^{}=45.4\text{G}`$ when $`\mathrm{\Gamma }=10^3`$ and $`n_{ism}=10^5\text{cm}^3`$. The energy extracted from the interstellar medium and converted into magnetic and thermal energy is small. This is shown in Figure 1. As a consequence, even when the Rankine-Hugoniot condition on $`n_{shell}^{}`$ of $`n_{shell}^{}/n_{ism}=\mathrm{\Gamma }`$ is satisfied, the fraction of kinetic energy that is converted to thermal and magnetic energy is tiny, so that the interstellar medium continues to stream through the shell.
The two-stream instability acts on the electrons to bring them into an equilibrium with the two ion streams. This instability grows at the rate of
$$\gamma _{2s}^{}\frac{1}{2}\sqrt{\frac{4\pi e^2}{m_e}}n_{ism}^{\frac{1}{2}}\mathrm{\Gamma }^1.$$
(4)
The growth rate of the electron two-stream instability is smaller than the ion filamentation instability by the factor of $`m_p^{1/2}/m_e^{1/2}\mathrm{\Gamma }0.04`$. The two stream instability will grow until the electron Lorentz factor is of order $`\mathrm{\Gamma }`$. At this point, the system becomes stable. This instability therefore only converts $`m_e/m_p`$ of the total energy of the interstellar medium into thermal energy.
## 3 Radiative Mechanisms
The two radiative mechanisms at work in this theory are synchrotron emission and synchrotron self-Compton emission. The characteristic energy of each process is given in Figures 2 and 3 for $`n_{ism}=1\text{cm}^3`$ and $`n_{ism}=10^5\text{cm}^3`$. This characteristic energy defines the maximum energy of the continuum as defined by the maximum energy of the electron distribution. The minimum energy of the synchrotron emission is given by the cyclotron energy, which is
$$\nu =2.68\times 10^8\text{Hz}n_{ism}^{\frac{1}{2}}\mathrm{\Gamma }_3^{\frac{3}{2}},$$
(5)
where $`\mathrm{\Gamma }_3=\mathrm{\Gamma }/10^3`$. This value is smaller than the characteristic synchrotron energy by the factor of $`\mathrm{\Gamma }^2`$.
The synchrotron self-Compton energy range is determined by a single scattering, since more than one scattering takes the photons to the characteristic electron energy. The minimum energy is set by $`\mathrm{\Gamma }^2`$ times the cyclotron resonance energy, so that the low end of the synchrotron self-Compton continuum overlaps the high end of the synchrotron continuum. The characteristic Compton scattering energy is a factor of $`\mathrm{\Gamma }^2`$ larger than this. From the figures, one sees that for $`\mathrm{\Gamma }>10^3`$, the energy range of the Compton scattered radiation extends above $`1\text{MeV}`$. For a lower value of $`\mathrm{\Gamma }`$, the cooling occurs predominately at optical and ultraviolet wavelengths. Because gamma-ray bursts are identified by their gamma-ray emission, bursts with $`\mathrm{\Gamma }<10^3`$ will not be observed. This suggests that there exists a class of burst phenomena with optical and ultraviolet emission, but no gamma-ray emission.
## 4 Selection Effects
The emission of gamma-rays defines one selection effect. In order to produce gamma-rays, $`\mathrm{\Gamma }>10^3`$. This provides an explanation for the absence of photon-photon pair creation in gamma-ray bursts, as demonstrated by the absence of gamma-ray bursts with thermal spectra.
A second selection effect that comes into play in this theory concerns the efficient production of gamma-rays. If the density is too low, the rate of radiative cooling is far below the rate at which electrons are thermalized. Such bursts would be dim relative to their high-density counterparts. The most observable bursts are therefore those with a sufficiently high $`n_{ism}`$ to efficiently convert electron thermal energy into gamma-rays. This lower limit on $`n_{ism}`$ can be written as
$$n_{ism}>8.22\times 10^4\text{cm}^3_{27}^{\frac{1}{2}}\mathrm{\Gamma }_3^{\frac{13}{4}}f_{emis}^{\frac{3}{2}}\left(\frac{R}{R_0}\right)^{\frac{3}{4}}$$
(6)
where $`f_{emis}`$ is the fraction of electrons at the characteristic energy, where $`R/R_0\left(m_p/m_e\right)^{1/3}`$ is the distance traveled when $`\mathrm{\Gamma }`$ drops by a factor of 2 relative to the distance traveled when the ISM is swept up, all under the assumption that the heating of the electrons represent the maximum energy loss, and where $`_{27}`$ is the mass of the shell per unit ster radian in units of $`10^{27}\text{gm}`$.
The limit given by this last equation is plotted in Figure 4 as a solid line. A similar curve that gives emission that is 10% effective at radiating the energy converted into thermal energy is plotted as a dotted line. The importance of this limit is that the gamma-ray burst mechanism requires a high density interstellar medium to operate efficiently. When the burst radiates efficiently, the interstellar medium is of sufficient density to attenuate the gamma-ray spectrum through Compton scattering. This provides the plasma instability theory with a mechanism that gives the spectrum a characteristic energy of several hundred keV, despite the broad spectral range of the synchrotron self-Compton continuum, and its strong dependence on $`\mathrm{\Gamma }`$. Such a mechanism is required to correctly reproduce the observed spectra.
## 5 Discussion
The plasma instability theory discussed above is a new mechanism to produce prompt gamma-ray burst emission. It has several unique features.
* The plasma instability theory produces the prompt gamma-ray emission without creating a shock. This implies that additional gamma-ray bursts can occur in the interstellar medium when new relativistic shells are ejected by the source, because the region is not cleared of material.
* The requirement that the mechanism efficiently produce gamma-rays introduces selection effects, so that bursts have $`\mathrm{\Gamma }>10^3`$ and $`n_{ism}>10^5\text{cm}^3`$.
* There exists a class of optical transient that has no gamma-ray emission. These bursts differ from gamma-ray bursts in having $`\mathrm{\Gamma }<10^3`$.
* The lower limit on $`n_{ism}`$ ensures that gamma-ray bursts are always in region in which Compton attenuation by the surrounding interstellar medium occurs.
The next step in developing this theory is to undertake a numerical study of the plasma instabilities. The goal of the study will be to confirm the analytic results discussed above, and to provide a precise calculation of the electron distribution produced by the instabilities. A second aspect of the theory that will be examined is the afterglow radiation produced in this theory; the cooling of the region behind the relativistic shells gives a light curve that differs from the light curve created by the decelerating shell, adding a complexity to the afterglow from this theory that is not present in the shock theories of afterglow radiation. Finally, calculations of model spectra and their comparison to observed spectra is will be undertaken. In particular, a comparison of the optical and gamma-ray spectrum of this model to that of GRB 990123 is planned. This study will test whether the prompt optical and gamma-ray emission are part of a single synchrotron self-Compton continuum.
## References
|
no-problem/9904/cond-mat9904364.html
|
ar5iv
|
text
|
# Analysis of the statistical behavior of genetic cluster-exact approximation
## I Introduction
The finite-dimensional Edwards-Anderson spin glass is a model for disordered systems which has attracted much attention over the last decades. The opinion on its nature, especially for three dimensional systems, is still controversial . Beside trying to address the problem with the help of analytic calculations and simulations at finite temperature, it is possible to investigate the behavior of the model by means of ground-state calculations . Since obtaining spin-glass ground states is computationally hard , the study is restricted to relatively small systems. Recently a new algorithm, the cluster-exact approximation (CEA) was presented, which allows in connection with a special genetic algorithm the calculation of true ground states for moderate system sizes, in three dimensions up to size $`14^3`$. By applying this method it is possible to study the ground-state landscape of systems exhibiting a $`T=0`$ degeneracy . For a thermodynamical correct evaluation it is necessary that each ground state contributes to the results with the same weight, since all ground states have exactly the same energy. Recently it was shown , that the genetic CEA causes a bias on the quantities describing the $`T=0`$ landscape. The aim of this paper is to analyze the algorithm with respect to its ground-state statistics. The reasons for the deviation from the correct behavior are given and an extension of the method is outlined, which guarantees thermodynamical correct results.
In this work, three-dimensional Edwards-Anderson (EA) $`\pm J`$ spin glasses are investigated. They consist of $`N`$ spins $`\sigma _i=\pm 1`$, described by the Hamiltonian
$$H\underset{i,j}{}J_{ij}\sigma _i\sigma _j$$
(1)
The sum runs over all pairs of nearest neighbors. The spins are placed on a three-dimensional (d=3) cubic lattice of linear size $`L`$ with periodic boundary conditions in all directions. Systems with quenched disorder of the interactions (bonds) are considered. Their possible values are $`J_{ij}=\pm 1`$ with equal probability. To reduce the fluctuations, a constraint is imposed, so that $`_{i,j}J_{ij}=0`$.
The article is organized as follows: next a description of the algorithms is presented. Then it is shown for small systems, that the method does not result in a thermodynamical correct distribution of the ground states. In section four, the algorithm and its different variants are analyzed with respect to the ground-state statistics. In the last section a summary is given and an extension of the method is outlined, which should guarantee thermodynamical correct results.
## II Algorithms
The algorithm for the calculation bases on a special genetic algorithm and on cluster-exact approximation . CEA is an optimization method designed specially for spin glasses. Its basic idea is to transform the spin glass in a way that graph-theoretical methods can be applied, which work only for systems exhibiting no bond-frustrations. Now a short sketch of these algorithms is given, because later the influence of different variants on the results is discussed.
Genetic algorithms are biologically motivated. An optimal solution is found by treating many instances of the problem in parallel, keeping only better instances and replacing bad ones by new ones (survival of the fittest). The genetic algorithm starts with an initial population of $`M_i`$ randomly initialized spin configurations (= individuals), which are linearly arranged using an array. The last one is also neighbor of the first one. Then $`n_o\times M_i`$ times two neighbors from the population are taken (called parents) and two new configurations called offspring are created. For that purpose the triadic crossover is used which turned out to be very efficient for spin glasses: a mask is used which is a third randomly chosen (usually distant) member of the population with a fraction of $`0.1`$ of its spins reversed. In a first step the offspring are created as copies of the parents. Then those spins are selected, where the orientations of the first parent and the mask agree . The values of these spins are swapped between the two offspring. Then a mutation with a rate of $`p_m`$ is applied to each offspring, i.e. a randomly chosen fraction $`p_m`$ of the spins is reversed.
Next for both offspring the energy is reduced by applying CEA: The method constructs iteratively and randomly a non-frustrated cluster of spins. During the construction of the cluster a local gauge-transformation of the spin variables is applied so that all interactions between cluster spins become ferromagnetic. Fig. 1 shows an example of how the construction of the cluster works for a small spin-glass system. To increase the performance, spins adjacent to many unsatisfied bonds are more likely to be added to the cluster. This may introduce a bias on the resulting distribution of the ground states. Later this scheme (“BIAS”) is compared to a variant (“SAME”), where all spins may contribute to the cluster with the same probability.
For 3d $`\pm J`$ spin glasses each cluster contains typically 55 percent of all spins. The non-cluster spins remain fixed during the following calculation, they act like local magnetic fields on the cluster spins. Consequently, the ground state of the gauge-transformed cluster is not trivial, although all interactions inside the cluster are ferromagnetic. Since the cluster exhibits no bond-frustration, an energetic minimum state for its spins can be calculated in polynomial time by using graph-theoretical methods : an equivalent network is constructed , the maximum flow is calculated and the spins of the cluster are set to orientations leading to a minimum in energy. Please note, that the ground state of the cluster is degenerate itself, i.e. the spin orientations can be chosen in different ways leading all to the same energy. It is possible to calculate within one single run a special graph, which represents all ground states of the cluster , and select one ground state randomly. This procedure is called “BROAD” here. On the other hand, one can always choose a certain ground state of the cluster directly<sup>*</sup><sup>*</sup>*This ground state has the maximum possible magnetization of the gauge-transformed spins among all cluster ground states.. Usually this variant, which is called “QUICK” here, is applied, because it avoids the construction of the special graph. But this again introduces a certain bias on the resulting distribution of the ground states. Later the influence of the different methods of choosing ground states is discussed.
This CEA minimization step is performed $`n_{\mathrm{min}}`$ times for each offspring. Afterwards each offspring is compared with one of its parents. The offspring/parent pairs are chosen in the way that the sum of the phenotypic differences between them is minimal. The phenotypic difference is defined here as the number of spins where the two configurations differ. Each parent is replaced if its energy is not lower (i.e. not better) than the corresponding offspring. After this whole step is conducted $`n_o\times M_i`$ times, the population is halved: From each pair of neighbors the configuration which has the higher energy is eliminated. If more than 4 individuals remain the process is continued otherwise it is stopped and the best individual is taken as result of the calculation.
The following representation summarizes the algorithm.
| algorithm genetic CEA($`\{J_{ij}\}`$, $`M_i`$, $`n_o`$, $`p_m`$, $`n_{\mathrm{min}}`$) |
| --- |
| begin |
| | create $`M_i`$ configurations randomly |
| | while ($`M_i>4`$) do |
| | begin |
| | | for $`i=1`$ to $`n_o\times M_i`$ do |
| | | begin |
| | | | select two neighbors |
| | | | create two offspring using triadic crossover |
| | | | do mutations with rate $`p_m`$ |
| | | | for both offspring do |
| | | | begin |
| | | | | for $`j=1`$ to $`n_{\mathrm{min}}`$ do |
| | | | | begin |
| | | | | | construct unfrustrated cluster of spins |
| | | | | | construct equivalent network |
| | | | | | calculate maximum flow |
| | | | | | construct minimum cut |
| | | | | | set new orientations of cluster spins |
| | | | | end |
| | | | | if offspring is not worse than related parent |
| | | | | then |
| | | | | | replace parent with offspring |
| | | | end |
| | | end |
| | | half population; $`M_i=M_i/2`$ |
| | end |
| | return one configuration with lowest energy |
| end |
The whole algorithm is performed $`n_R`$ times and all configurations which exhibit the lowest energy are stored, resulting in $`n_G`$ statistically independent ground-state configurations (replicas). A priori nothing about the distribution of ground states raised by the algorithm is known. Thus, it may be possible that for one given realization of the disorder some ground states are more likely to be returned by the procedure than others. Consequently, any quantities which are calculated by averaging over many independent ground states, like the distribution of overlaps, may depend on a bias introduced by the algorithm. For a thermodynamical correct evaluation all ground states have to contribute with the same weight, since they all have exactly the same energy.
For the preceding work, the distribution of the ground states determined by the algorithm was taken. The method was utilized to examine the ground state landscape of two-dimensional and three-dimensional $`\pm J`$ spin glasses by calculating a small number of ground states per realization. Some of these results depend on the statistics of the ground states, as it will be shown in the next section for the $`d=3`$ case.
On the other hand, the main findings of the following investigations are not affected by the bias introduced by genetic CEA: the existence of a spin-glass phase for nonzero temperature was confirmed for the three-dimensional spin glass . The method was applied also to the $`\pm J`$ random-bond model to investigate its $`T=0`$ ferromagnetic to spin-glass transition. Finally, for small sizes up to $`L=8`$ all ground-state valleys were obtained by calculating a huge number of ground states per realization and applying a new method called ballistic search .
## III Numerical evidence
In this section results describing the ground-state landscape of small three-dimensional $`\pm J`$ spin glasses are evaluated. It is shown that the data emerging from the use of raw genetic CEA and from a thermodynamically correct treatment differ substantially.
Several ground states for small systems of size $`N=L^3=3^3,4^3,5^3`$ were calculated. 1000 realizations of the disorder for $`L=3,4`$ and 100 realizations for $`L=5`$ were considered. The parameters ($`M_i`$, $`n_o`$, $`p_m`$, $`n_{\mathrm{min}}`$), for which true ground states are obtained, are shown in . For all calculations the variants BIAS and QUICK were used to obtain maximum performance. The effect of different variants on the results is discussed in the next section.
Two schemes of calculation were applied:
* For each realization $`n_R=40`$ runs of genetic CEA were performed and all states exhibiting the ground-state energy stored. Consequently, this scheme reflects the ground-state statistics which is determined solely by the genetic CEA method. Configurations which have a higher probability of occurrence contribute with a larger weight to the results.
* For each realization the algorithm was run up to $`10^5`$ times. Each particular state was stored only once. For later analysis the number of times each state occurred was recorded. Additionally, a systematic local search was applied to add possibly missing ground states which are related by flips of free spins to states already found. Finally, a $`L=3`$ realization exhibits 25 different ground states on average. For a $`L=4`$ realization on average 240 states were found and 6900 states for $`L=5`$.
For the evaluation of physical quantities every ground state is taken with the same probability in this scheme. Thus, the statistics obtained in this way reflect the true $`T=0`$ thermodynamic behavior.
To analyze the ground-state landscape, the distribution of overlaps is evaluated. For a fixed realization $`J=\{J_{ij}\}`$ of the exchange interactions and two replicas $`\{\sigma _i^\alpha \},\{\sigma _i^\beta \}`$, the overlap is defined as
$$q^{\alpha \beta }\frac{1}{N}\underset{i}{}\sigma _i^\alpha \sigma _i^\beta $$
(2)
The ground state of a given realization is characterized by the probability density $`P_J(q)`$. Averaging over the realizations $`J`$, denoted by $`[]_J`$, results in ($`Z`$ = number of realizations)
$$P(q)[P_J(q)]_J=\frac{1}{Z}\underset{J}{}P_J(q)$$
(3)
Because no external field is present the densities are symmetric: $`P_J(q)=P_J(q)`$ and $`P(q)=P(q)`$. So only $`P(|q|)`$ is relevant.
The result of $`P(|q|)`$ for $`L=5`$ is shown in Fig. 2. For the true thermodynamic result small overlaps occur less frequent than for the data obtained by the application of pure genetic CEA. Large overlap values occur more often. This deviation has an influence on the way the spin glass behavior is interpreted. The main controversy about finite-dimensional spin glasses mentioned at the beginning is about the question whether for the infinite system $`P(|q|)`$ shows a long tail down to $`q=0`$ or not .
To investigate the finite size behavior of $`P(|q|)`$ the fraction $`X_{0.5}`$ of the distribution below $`q_0=0.5`$ is integrated:
$$X_{q_0}_0^{q_0}P(|q|)𝑑q$$
(4)
The development of $`X_{0.5}`$ as a function of system size $`L`$ is shown in Fig. 3. The datapoints for the larger sizes $`L6`$, obtained using pure genetic CEA, are taken from former calculations . These values are more or less independent of the system size, while the correct thermodynamic behavior shows a systematic decrease. Whether for $`L\mathrm{}`$ the long tail of $`P(|q|)`$ persists cannot be concluded from the data, because the systems are too small. Nevertheless, the true $`T=0`$ behavior differs significantly from the former results.
## IV Analysis of genetic CEA
To understand, why genetic CEA fails in producing the thermodynamical correct results, in this section the statistics of the ground states, which is determined by the algorithm, is analyzed directly.
For the case where all ground states were calculated using a huge number of runs, the frequencies each ground state occurred were recorded. In Fig. 4 the result for one sample realization of $`N=5^3`$ is shown. The system has 56 different ground states. For each state the number of times it was returned by the algorithm in $`10^5`$ runs is displayed. Obviously the large deviations from state to state cannot be explained by the presence of statistical fluctuations. Thus, genetic CEA samples different ground states from the same realization with different weights.
To make this statement more precise, the following analysis was performed: Two ground states are called neighbors, if they differ only by the orientation of one spin. All ground states which are accessible from each other through this neighbor-relation are defined to be in the same ground-state valley. That means, two ground states belong to the same valley, if it is possible to move from one state to the other by flipping only free spins, i.e. without changing the energy. For all realizations the valleys were determined using a method presented in , which allows to treat systems efficiently exhibiting a huge number of ground states. Then the frequencies $`h_V`$ for each valley $`V`$ were computed as the sum of all frequencies of the states belonging to $`V`$. In Fig. 5 the result is shown for a sample $`N=5^3`$ realization, which has 15 different ground state valleys. Large valleys are returned by the algorithm more frequently, but $`h_V`$ seems to grow slower than linearly. A strict linear behavior should hold for an algorithm which guarantees the correct $`T=0`$ behavior.
For averaging $`h_V`$ has to be normalized, because the absolute values of the frequency differ strongly from realization to realization, even if the size $`|V|`$ of a valley, i.e. the number of ground states belonging to it, is the same. For each realization, the normalized frequency $`h_V^{}`$ is measured relatively to the average frequency $`\overline{h}_1`$ of all valleys of size 1: $`h_V^{}h_V/\overline{h}_1`$
If a realization does not exhibit a valley consisting only of one ground state, the frequency $`h_{V_s}`$ of the smallest valley $`V_s`$ is taken. It is assumed, that the normalized frequency exhibits a $`h_V^{}=|V|^\alpha `$ dependence, which is justified by the results shown later. Consequently, for the case the size $`|V_s|`$ of the smallest valley is larger than one, $`\overline{h}_1h_{V_s}/|V_s|^\alpha `$ is chosen. The value of $`\alpha `$ is determined self-consistently.
The result for $`L=3`$ of $`h_V^{}`$ as a function of the valley-size $`|V|`$ is presented in Fig. 6. A value of $`\alpha =0.854(3)`$ was determined. Please note, that the fluctuations for larger valleys are higher, because quite often only one valley was available for a given valley-size. The algebraic form is clearly visible, proving that genetic CEA overestimates systematically the importance of small ground-state valleys.
For $`L=4`$ a value of $`\alpha =0.705(3)`$ was obtained, while the $`L=5`$ case resulted in $`\alpha =0.642(5)`$. Consequently, with increasing system size, the algorithm fails more and more to sample configurations from different ground-state valleys according to the size of the valleys. This explains, why the difference of $`X_{0.5}(L)`$ between the correct result and the values obtained in increases with growing system size.
Similar results were obtained for two-dimensional systems. For $`L=5`$ a self-consistent value of $`\alpha =0.650(1)`$ was found, while the treatment of $`L=7`$ systems resulted in $`\alpha =0.659(2)`$. Here only a slight finite-size dependence occurs. This may explain the fact, that the width of the distribution of overlaps, even calculated only by the application of pure genetic CEA, seems to scale to zero .
In the second section of this paper two variants of the algorithm were presented, which may be able to calculate ground states more equally distributed. To investigate this issue, similar ground-state calculations were conducted for $`L=4`$ and again $`h_V^{}`$ was calculated. For the case, were SAME was used instead of BIAS, a value $`\alpha =0.801(2)`$ was determined self-consistently. Using BROAD instead of QUICK resulted in $`\alpha =0.749(3)`$. Finally, by applying SAME and BROAD together, $`\alpha =0.843(3)`$ was obtained. Consequently, applying different variants of the method decreases the tendency of overestimating small valleys, but the correct thermodynamic behavior is not obtained as well. Even worse, BROAD and SAME are considerably slower than the combination of QUICK and BIAS.
So far it was shown, that genetic CEA fails in sampling ground states from different valleys according the size of the valleys. Now we turn to the question, whether at least states belonging to the same valley are calculated with the correct thermodynamic distribution. By investigating the frequencies of different ground states belonging to the same valley it was found again, that these configurations are not equally distributed. But it is possible to study this issue in a more physical way. For that purpose ground states of 100 $`L=10`$ realizations were calculated. Then the valley structure was analyzed. The average distribution of overlaps was evaluated, but only contributions of pairs of states belonging to the same valley were considered. For comparison, for the same realizations a long $`T=0`$ Monte-Carlo (MC) simulation was performed, i.e. randomly spins were selected and flipped if they were free. The ground states were used as starting configurations. Since a MC simulation ensures the correct thermodynamic distribution of the states, all ground states of a valley appear with the same frequency, if the simulation is only long enough. A length of 40 Monte-Carlo steps per spin were found to be sufficient for $`L=10`$. The result for the distribution of overlaps $`P_{\text{valley}}(|q|)`$ restricted to the valleys is displayed in Fig. 7. Significant differences between the datapoints from the pure genetic CEA and the correct $`T=0`$ behavior are visible. Consequently, the algorithm does not sample configurations belonging to the same ground-state valley with the same weight as well.
## V Conclusion
In this work the genetic cluster-exact approximation method is analyzed. The algorithm can be used to calculate many independent true ground states of EA spin glasses. The results from the raw application of the method and from calculations of all ground states for small system sizes were compared. By evaluating the distribution of overlaps is was shown, that genetic CEA imposes some bias on the ground-state statistics. Consequently, the results from the application of the raw method do not represent the true $`T=0`$ thermodynamics.
To elucidate the behavior of the algorithm the statistics of the ground states were evaluated directly. It was shown, that different ground states have dissimilar probabilities of occurrence. To understand this effect better, the ground-state valleys were determined. The genetic CEA method finds configurations from small ground-state valleys relative to the size of the valley more often than configurations from large valleys. Additionally, within a valley the states are not sampled with the same weight as well. It was shown that two variants of the algorithm, which decrease its efficiency, weaken the effect, but it still persists.
Summarizing, two effects are responsible for the biased ground-state sampling of genetic CEA: small valleys are sampled too frequently and the distribution within the valleys is not flat.
For small system sizes it is possible to calculate all ground states, so one can obtain the true thermodynamic average directly. But already for $`L=5`$ there are realizations exhibiting more than $`10^5`$ different ground states. Since the ground-state degeneracy grows exponentially with system size larger systems cannot be treated in this way. The following receipt should overcome these problems and should allow to obtain the true thermodynamic $`T=0`$ behavior for larger systems:
* Calculate several ground states of a realization using genetic CEA.
* Identify the ground states which belong to the same valleys.
* Estimate the size of each valley. This can be done using a variant of ballistic search , which works by flipping free spins sequentially, each spin at most once. The number of spins flipped is a quite accurate measure for the size of a valley.
* Sample from each valley a number of ground states, which is proportional to the size of the valley. This guarantees, that each valley contributes with its proper weight. Each state is obtained by performing a $`T=0`$ MC simulation of sufficient length, starting with true ground-state configurations. Since MC simulations achieve a thermodynamical correct distribution, it is guaranteed that the states within each valley are equally distributed.
Please note, that it is not necessary to calculate all ground states to obtain the true thermodynamic behavior, because it is possible to estimate the size of a valley by analyzing only some sample ground states belonging to it. Furthermore, it is even only necessary to have configurations from the largest valleys available, since they dominate the ground-state behavior. This condition is fulfilled by genetic CEA, because large valleys are sampled more often than small valleys, even if small valleys appear too often relatively.
From the results presented here it is not possible to deduce the correct $`T=0`$ behavior of the infinite system, because the system sizes are too small. Using the scheme outlined above, it is possible to treat system sizes up to $`L=14`$ .
## VI Acknowledgements
The author thanks K. Battacharya and A.W. Sandvik for interesting discussions. The work was supported by the Graduiertenkolleg “Modellierung und Wissenschaftliches Rechnen in Mathematik und Naturwissenschaften” at the Interdisziplinäres Zentrum für Wissenschaftliches Rechnen in Heidelberg and the Paderborn Center for Parallel Computing by the allocation of computer time. The author announces financial support from the DFG (Deutsche Forschungsgemeinschaft).
|
no-problem/9904/astro-ph9904081.html
|
ar5iv
|
text
|
# Evidence for an Expanding Molecular Superbubble in M 82
## 1 Introduction
M 82 is the best studied nearby starburst galaxy ($`\mathrm{D}=\mathrm{\hspace{0.17em}3.25}\mathrm{Mpc}`$). The central few hundred parsecs of this galaxy are heavily obscured by dust and gas which hides the central starburst region against direct observations at optical wavelengths. Evidence for strong star–forming activity in the central region comes from radio (e.g. Kronberg et al. kronberg81 (1981)) and infrared observations (e.g. Telesco et al. telesco91 (1991)) and also from the prominent bipolar outflow visible in $`\mathrm{H}\alpha `$ (e.g. Bland & Tully bland88 (1988), McKeith et al. mckeith95 (1995), Shopbell & Bland–Hawthorn shopbell98 (1998)) and in X-rays (e.g. Bregman et al. bregman (1995)). The massive star formation (SF) is believed to be fuelled by the large amount of molecular gas which is present in the centre of M 82.
On the other hand, SF effects the distribution and kinematics of the surrounding interstellar medium (ISM). Recent millimetre continuum observations (Carlstrom & Kronberg carlstrom90 (1990)) suggested that the HII regions in M 82 have swept up most of the surrounding neutral gas and dust into dense shells. This is in agreement with the standard picture: shells are created by young star–forming regions through strong stellar winds of the most massive stars in a cluster and through subsequent type–II supernovae (e.g. Tenorio–Tagle & Bodenheimer tenorio88 (1991)). These processes are thought to blow huge cavities filled with coronal gas into their ambient ISM (e.g. Cox & Smith cox74 (1974), Weaver et al. weaver77 (1977)). This hot interior is then believed to drive the expansion of the outer shell of swept–up material.
Once superbubbles reach sizes that are comparable to the thickness of a galaxy’s disk, the bubble will eventually break out into the halo. This then leads to an outflow of the hot gas with velocities much higher than the expansion of the shell within the disk of the galaxy. In the following we present evidence for a molecular superbubble in M 82 which already broke out of the disk and seems to contribute significantly to the well–known prominent outflow of M 82.
## 2 Observations
### 2.1 Molecular Lines
For our analysis we used the $`{}_{}{}^{12}\mathrm{CO}(J=10)`$ data cube obtained by Shen & Lo (shen (1995)) with the BIMA array (spatial resolution: $`2.5^{\prime \prime }`$) and the $`{}_{}{}^{13}\mathrm{CO}(J=10)`$ data cube from Neininger et al. (nico13 (1998)) observed with the Platau de Bure interferometer (PdBI) (spatial resolution: $`4.2^{\prime \prime }`$). In addition, we used unpublished PdBI data of the $`{}_{}{}^{12}\mathrm{CO}(J=21)`$ and $`\mathrm{C}^{18}\mathrm{O}(J=10)`$ transitions. These observations where carried out in April 1997 in the CD configuration, resulting in a spatial resolution of $`1.4^{\prime \prime }\times 1.2^{\prime \prime }`$ ($`{}_{}{}^{12}\mathrm{CO}(J=21)`$) and $`3.7^{\prime \prime }\times 3.5^{\prime \prime }`$ ($`\mathrm{C}^{18}\mathrm{O}(J=10)`$), and a velocity resolution of 3.3 km s<sup>-1</sup> and 6.8 km s<sup>-1</sup>, respectively. In order to increase the sensitivity to extended CO emission we combined the $`{}_{}{}^{12}\mathrm{CO}(J=21)`$ data cube with single dish measurements obtained with the IRAM 30m telescope. These observations were carried out in May 1998. The combination is essential for this particular study because the receding part of the superbubble is only marginally visible in the mere interferometer maps. A full account of the data reduction will be given elsewhere.
### 2.2 Evidence for an Expanding Superbubble
Fig. 1 shows the integrated $`{}_{}{}^{12}\mathrm{CO}`$ line emission published by Shen & Lo (shen (1995)). The cross corresponds to the position of the supernova remnant 41.9+58 (SNR 41.9+58) which is the strongest cm continuum point–source in M 82 (Kronberg et al. kronberg81 (1981)). It is considered to be the aftermath of a ’hypernova’, which exhibits a radio luminosity 50–100 times greater than typical for type–II SNe (Wilkinson & de Bruyn wilkinson (1990)). The line along the major axis indicates the orientation of the position–velocity (pv) diagrams shown in Fig. 2.
The pv–cut is orientated in such a way that the signature of the expanding superbubble is visible most distinctly. The angular axis of the pv diagrams correspond to the offset in arcseconds from SNR 41.9+58. Besides a constant velocity gradient (cf. Shen & Lo shen (1995)) the pv–diagrams show an expanding ring–like feature centred on SNR 41.9+58, with a central velocity of $`\mathrm{V}_{\mathrm{lsr}}150\mathrm{km}\mathrm{s}^1`$. The approaching velocity component of the ring is clearly seen in all cubes and is centred at $`100\mathrm{km}\mathrm{s}^1`$. The receding component is only marginally visible in the $`{}_{}{}^{13}\mathrm{CO}(J=10)`$ and $`\mathrm{C}^{18}\mathrm{O}(J=10)`$ lines; its central velocity is about $`190\mathrm{km}\mathrm{s}^1`$. An enlargement of the pv–diagram along the major axis is shown in Fig. 3 (left). From this diagram we estimate the radius of the ring to be ($`65\pm \mathrm{\hspace{0.17em}5}`$) pc and the expansion velocity ($`45\pm \mathrm{\hspace{0.17em}5}`$) km s<sup>-1</sup>. A pv–diagram along the minor axis centred on SNR 41.9+58 is presented in Fig. 3 (right). The orientation of the cut is shown in Fig. 1.
Fig. 3 (right) reveals two emission features centred at the position of SNR 41.9+58, which correspond to the approaching ($`v100\mathrm{km}\mathrm{s}^1`$) and the receding component ($`v190\mathrm{km}\mathrm{s}^1`$) of the expanding superbubble. Hardly any CO emission with a velocity between $`v=100\mathrm{km}\mathrm{s}^1`$ and $`v=190\mathrm{km}\mathrm{s}^1`$ is found south and north of SNR 41.9+58. The pv–diagrams therefore show that the expanding molecular shell has already broken out of the disk and now only shows the signature of an expanding molecular ring. It should be noted that the remaining molecular gas in M 82 shows clear solid–body rotation.
### 2.3 Other wavelengths
#### Radio continuum observations: <br>
Radio continuum observations at 408 MHz by Wills et al. (wills97 (1997)) unveiled a prominent ‘hole’ of approximately 100 pc diameter around SNR 41.9+58 which they attribute to free–free absorption. They propose that this feature is due to absorption by a large H II region that has been photoionized by a cluster of early–type stars of which the progenitor of SNR 41.9+58 was originally a member. It should be noted, however, that SNR 41.9+58 is only about 50 years old (Wilkinson & de Bruyn wilkinson (1990)) and therefore cannot be the source that drives the expansion of the superbubble. But its presence supports the scenario that the expanding molecular superbubble is powered by an interior stellar cluster.
#### Neutral hydrogen (HI): <br>
H I emission line studies show that the central kpc of M 82 is dominated by absorption (Yun et al. yun93 (1993)), which makes it difficult to study the H I kinematics in the central part of the galaxy. Recent H I absorption studies against supernova remnants in M 82 by Wills et al. (wills98 (1998)) disclosed two absorption features against SNR 41.9+58. The velocities of the components are $`87\pm 7\mathrm{km}\mathrm{s}^1`$ and $`200\pm 7\mathrm{km}\mathrm{s}^1`$. The H I component at 87 km s<sup>-1</sup> can be attributed to the approaching component of the expanding superbubble. The absorption feature at 200 km s<sup>-1</sup>, if associated with the receding part, seems to contradict the hypothesis that SNR 41.9+58 is located within the expanding superbubble. However, the velocity resolution of the absorption study was rather poor (26.4 km s<sup>-1</sup>) and other explanations cannot be ruled out at this point.
#### Optical observations: <br>
Optical observations of the centre of M 82 suffer heavily from light absorption by the prominent dust lanes, rendering an optical analysis of this particular region impossible. HST V– and I–band images of the centre of M 82 (O’Connell et al. connell95 (1995)) do not reveal any optical sources in the vicinity of SNR 41.9+58. The prominent $`\mathrm{H}\alpha `$ outflow is visible north and south of the absorbing dust lane (Bland & Tully bland88 (1988), McKeith et al. mckeith95 (1995), Shopbell & Bland–Hawthorn shopbell98 (1998)). The orientation of its filaments suggests that the $`\mathrm{H}\alpha `$ outflow emerges at least partly from the location of the molecular superbubble.
#### X–ray observations: <br>
Fig. 4 shows an overlay of archival ROSAT HRI X–ray data (Bregman et al. bregman (1995)) onto the CO emission. To emphasize the diffuse, extended X–ray outflow we subtracted a model of the brightest X–ray point source at $`\alpha =9^h55^m50^s.4`$, $`\delta =69^{}40^{}47.5^{\prime \prime }`$ (J2000.0) and smoothed the residual image to a $`6^{\prime \prime }`$ resolution. The positions of the three point sources in the field are marked with stars, SNR 41.9+58 is marked by a cross (see also Bregman et al.). The overlay shows that most of the diffuse X–ray emission arises from the vicinity of SNR 41.9+58. The shape of the diffuse emission is elongated, extending several arcminutes along the minor axis of the galaxy. Model calculations by Bregman et al. show that the emission is consistent with an outflow of heated material from the central region coincident with the position of the expanding molecular superbubble.
We estimate an upper limit for the energy still present in the hot gas emerging from the superbubble. For the calculation we adopt a density $`\mathrm{n}=0.2\mathrm{cm}^3`$ and a temperature $`\mathrm{kT}=1\mathrm{keV}`$ for the central region (Bregman et al.) as an upper limit for the mean values in the outflow. We find an upper limit for the energy of $`\mathrm{E}_{\mathrm{gas}}2.7\times 10^{53}`$ergs present in a cylindrical volume with $`r=130\mathrm{pc}`$ and $`z=550\mathrm{pc}`$ centred on the superbubble. From the thickness of the molecular disk of M 82 ($`90\mathrm{pc}`$) and the expansion velocity of the superbubble we estimate a time elapsed since outbreak of $`\tau \mathrm{\hspace{0.17em}3}\times 10^5\mathrm{yr}`$. The required outflow velocity to reach the observed height above the disk therefore is $`\mathrm{v}\mathrm{\hspace{0.17em}1000}\mathrm{km}\mathrm{s}^1`$. This value is comparable with estimates based on particle aging as derived from radio continuum observations (Seaquist et al. seaquist85 (1985)) and measured outflow velocities in $`\mathrm{H}\alpha `$ (McKeith et al. mckeith95 (1995)).
#### Ionized gas: <br>
Most studies of the ionized gas in M 82 unfortunately do not provide sufficient spatial and spectral resolution to reveal details of the region under study. A high–resolution study of the Ne II emission line has been published by Achtermann & Lacy (achtermann (1995)). The integrated line emission in the south–western region of M 82 shows a similar double–peaked feature centred on SNR 41.9+58 as traced by the CO. The pv–cut along the major axis clearly shows a disturbed velocity field in the vicinity of SNR 41.9+58. The approaching side of the expanding superbubble at 95 km s<sup>-1</sup> is clearly visible while the receding side at 190 km s<sup>-1</sup> is only marginally seen. The diameter of the Ne II shell ($``$ 100 pc) seems to be somewhat smaller than the corresponding CO feature, suggesting that the inner part of the expanding superbubble is ionized. Similar features are visible in the distribution and the kinematics of the H41$`\alpha `$ emission line (Seaquist et al. seaquist96 (1996)). The Ne II distribution might represent the transition from the hot coronal gas with a temperature of several $`10^6`$K towards the cold molecular rim of the superbubble.
## 3 Discussion
The picture that emerges from the observations presented above is the following: a major SF event at the centre of what today shows up as the prominent expanding molecular superbubble created a cavity filled with coronal gas by the combined effects of strong stellar winds and SN explosions. The pressure of the hot–gas interior drove the expansion of the shell of swept–up material until it broke out of the disk. The subsequent outflow of hot material today shows up as diffuse X–ray emission and contributes to the prominent $`\mathrm{H}\alpha `$ filaments.
From the observational parameters of the expanding superbubble (expansion velocity and diameter) we estimate a kinematic age of the superbubble of $`1\times 10^6`$ years. The kinematic age is an upper limit for the actual age since the superbubble is most probably decelerating. We use Chevalier’s equation (Chevalier chev74 (1974)) to derive the amount of energy needed to create the expanding superbubble. Since Chevalier’s equation applies to H I shells only we corrected the energy input by a factor of 2 to correct for the difference in mass between H<sub>2</sub> and H. We estimate an ambient H<sub>2</sub> density prior to the creation of the shell of order $`120\mathrm{cm}^3`$. This is done by converting the $`{}_{}{}^{12}\mathrm{CO}(J=10)`$ line integral to H<sub>2</sub> column density using a conversion ratio of N(H<sub>2</sub>)/W(CO) = $`1.2\times 10^{20}`$ cm<sup>-2</sup> K<sup>-1</sup> km<sup>-1</sup>s (Smith et al. smith (1991)) and estimating the volume of the region from which material was swept up to be cylindrical with $`z=100\mathrm{pc}`$ and $`r=65\mathrm{pc}`$. We derive a total energy of order $`2\times 10^{54}`$ergs, which corresponds to an energy equivalent of approximately 1000 type–II SNe and the strong stellar winds of their progenitors. With the estimate for the age of the superbubble this leads to a SN rate of 0.001 SN yr<sup>-1</sup> for the central stellar cluster. This is a reasonable number given the fact that the SN rate in the central part of M 82 was estimated to be about 0.1 SN yr<sup>-1</sup> (Kronberg et al. kronberg81 (1981)). The total H<sub>2</sub> mass of the superbubble is about $`8\times 10^6\mathrm{M}_{}`$. Less than 15% of the total energy is still present in the hot gas which emerges from the superbubble; the fraction of kinetic energy of the molecular superbubble is about 10% ($`\mathrm{E}_{\mathrm{kin}}\mathrm{\hspace{0.17em}1.6}\times 10^{53}`$ergs). It should be noted, however, that the numbers given above are only order of magnitude estimates.
The wealth of observations presented above indicates that the region under study represents an unusually active segment of the starburst in M 82. The finding of an expanding molecular superbubble in this particular region supports this view and provides clear evidence for a footprint of violent SF in the ISM of M 82. Furthermore the analysis of the molecular superbubble presents an alternative tool to investigate the energy release of the central source which drives the X–ray and contributes to the $`\mathrm{H}\alpha `$ outflow.
###### Acknowledgements.
A.W. and F.W. acknowledge DFG grant III GK–GRK 118/2. We thank J. Shen and K.Y. Lo for making available their CO data, J. Kerp for his help on the ROSAT data and the referee, P. Kronberg, for helpful comments. We acknowledge the IRAM staff for carrying out the observations and the help provided during the data reduction.
|
no-problem/9904/astro-ph9904078.html
|
ar5iv
|
text
|
# Variation of the broad X-ray iron line in MCG–6-30-15 during a flare
## 1 Introduction
A major discovery from ASCA has been the discovery of a clear, broad, skewed iron line in the spectrum of the Seyfert galaxy MCG–6-30-15 (Tanaka et al 1995). This emission line has a profile matching that expected from the inner regions, from about 6 to 40 gravitational radii (i.e. 6 – 40 $`GM/c^2`$), of a disk around a black hole (Fabian et al 1989). No simple alternative model is capable of explaining this profile (Fabian et al 1995). Similar skewed lines have since been found in the spectra of many other Seyfert galaxies (Nandra et al 1997; Reynolds 1997). The broad line in MCG–6-30-15 has also been clearly detected with BeppoSAX (Guainazzi et al 1999).
The 1994 ASCA observation of MCG–6-30-15 reported by Tanaka et al (1995) remains however the best example of a broad line due to the good spectral resolution of the detectors used and the long integration time of 4.5 days. Here we report on a similar long ASCA observation of the object made in 1997. We confirm in detail the time-averaged line shape, which only shows a small change in the ‘blue’ horn.
During the 1994 ASCA observation the light curve of the source showed both a flare and a deep minimum (Iwasawa et al 1996). The line profile was seen to alter, being mostly a blue horn during the flare and then showing only an extreme red horn during the minimum. These changes were assumed to be due to changes in the location in the most active regions irradiating the disk (and so producing the iron line), the flare being on the approaching side of the disk and the minimum emission from within the innermost stable orbit of a non-spinning Schwarzschild black hole (Iwasawa et al 1996). This last possibility has been explored further by Dabrowski et al (1997), Reynolds & Begelman (1998); Weaver & Yaqoob (1998) and Young, Ross & Fabian (1998).
The light curve of the source during 1997 also shows flares and dips. We examine in detail here the major flare seen during which both the continuum and line show large changes.
## 2 Observations and data reduction
MCG–6-30-15 was observed with ASCA from 1997 August 3 to 1997 August 10 with a half-day gap in the middle. It was also observed simultaneously with Rossi X-ray Timing Explorer (RXTE) (Lee et al 1999).
The Solid state Imaging Spectrometer (SIS; S0 and S1) was operated in Faint mode throughout the observation, using the standard CCD chips (S0C1 and S1C3). The Gas Imaging Spectrometer (GIS; G2 and G3) was operated in PH mode. We present results mainly from the SIS data in this Letter. The ASCA S0 light curve in the 0.6–10 keV band is shown in Fig. 1.
Data reduction was carried out using FTOOLS version 4.0 and 4.1 with standard calibration provided by the ASCA Guest Observer Facility (GOF). The good exposure time is approximately 231 ks from each SIS detector. The source counts are collected from a region centred at the X-ray peak within 4 arcmin in radius for the SIS and 5 arcmin for the GIS. The background data are taken from a (nearly) source-free region in the same detector with the same observing time. The efficiency of the S1 detector below 1 keV appears to be severely reduced due to the Residual Darkframe Distribution (RDD, Dotani 1998), which the current response matrix (generated from calibratione files in the FTOOLS version 4.1 release) does not take into account for. The RDD effect on the S0 data from 1CCD observations has been found to be very little (Dotani 1998). Therefore, the S1 data below 1 keV were discarded for the spectral analysis presented here. The energy resolution of the SIS at 6.4 keV when the observation was carried out had degraded to $`250`$ eV (FWHM), about twice that attained immediately after launch of the satellite.
## 3 Comparisons with the 1994 long observation
The average count rates in the 0.6–10 keV band from the S0/S1 detectors are 1.16/0.93 ct s<sup>-1</sup>(cf. 1.53/1.25 ct s<sup>-1</sup> during the previous long observation in 1994). The average observed fluxes are $`1.47\times 10^{11}`$erg cm<sup>-2</sup> s<sup>-1</sup> in the 0.5–2 keV band and $`3.41\times 10^{11}`$erg cm<sup>-2</sup> s<sup>-1</sup> in the 2–10 keV band.
### 3.1 Total energy spectrum and warm absorber
The observed 0.6–10 keV X-ray flux during the present observation is lower by 26 per cent than that during the 94 long observation. The 3–10 keV (the iron K band, 4–7.5 keV, excluded) power-law slope is $`\mathrm{\Gamma }=1.94_{0.07}^{+0.06}`$, which is similar to the 94 data. Features of the warm absorber detected are two edges due to OVII at 0.72 keV and OVIII at 0.85 keV (e.g., Otani et al 1996) and one at 1.1 keV, probably due to NeIX and/or Fe L. The 97 spectrum is harder in the low energy band than the 94 spectrum, which may be explained by an increase in absorption. Details will be reported by Matsumoto et al (in prep).
### 3.2 The iron K line
The 3–10 keV data were investigated for iron K line emission. The continuum spectrum was modelled with a power-law reflection model (pexrav, Magdziarz & Zdziarski 1995) modified by cold absorption of $`N_\mathrm{H}`$$`=7\times 10^{20}`$cm<sup>-2</sup>(which has virtually no effect on the 3–10 keV continuum). The parameters of pexrav, apart from photon index and normalization, were matched to the previous measurements of MCG–6-30-15: the cut-off energy, 130 keV (from the BeppoSAX observation by Guainazzi et al 1999); reflection intensity, corresponding to $`\mathrm{\Omega }/2\pi =1`$ (Guainazzi et al 1999; Lee et al 1999); iron abundance of unity (Lee et al 1999); and inclination of the reflecting slab, 30 (Tanaka et al 1995). The line feature is fitted by the diskline model for a Schwarzschild black hole (Fabian et al 1989). The rest energy of the line emission is assumed to be 6.4 keV, appropriate for cold iron. A power-law ($`\mathrm{\Gamma }=1.96_{0.03}^{+0.04}`$) modified by reflection plus a diskline model provide a good fit ($`\chi ^2=724.7`$ for 729 degrees of freedom). The best-fit parameters of the diskline model are shown in Table 1.
The efficiency-corrected line profile<sup>1</sup><sup>1</sup>1This is not an ‘unfolded’ spectrum but obtained from the ratio, the data divided by the power-law model (folded through the detector response) best-fitting the neighbouring continuum, multiplied by the power-law (in original form). The plot is therefore independent from the model used for fitting the line. for the present data set is shown in Fig. 2, along with the one from the previous long observation in 1994 (Tanaka et al 1995). The profile from the present observation appears to be less bright in the blue peak while it shows a slightly more extended red wing. During the previous observation, the bright flare (i-3) data showed a narrow-core-dominated line profile (Iwasawa et al 1996). Such a line shape is not found during the present observation. However, this is not sufficient to explain the difference in the blue horn of the time-averaged line profiles between the two observations. Although the overall line shape is similar between 94 and 97, the steeper radial emissivity index suggests that the mean weight of the line emissivity may be slightly shifted towards the inner part of the accretion disk in 97 as compared to 94.
## 4 The major flare
Changes in the iron line profile were investigated in time sequence, details of which will be reported elsewhere. Here we show the peculiar behaviour of the energy spectrum and iron line during the major flare, which occured around $`1.9\times 10^5`$ s in the light curve (see Fig. 1 and Fig. 3 for a detailed version of the light curve around the flare). The continuum is steeper than usual ($`\mathrm{\Delta }\mathrm{\Gamma }0.17`$), particularly at low energies (1–3 keV band), as shown in Fig. 4.
### 4.1 Iron K line emission with large redshift?
Excess emission above a power-law continuum, which is presumably due to broad iron K line emission, is found in the 4–7 keV range. The ratio plots of the data and the baseline power-law model for the intervals of the flare peak (a) and of the subsequent dip (b) are shown in Fig. 5. The baseline model is obtained by fitting a power-law to the data adjacent to the iron line band.
The line profile of the flare peak (a) shows a sharp decline at $``$5.6 keV, which is far below the rest energy of the line emission of Fe K$`\alpha `$, 6.4 keV, and a red-wing extending down to $``$3.5 keV. We have checked the GIS data which confirm the SIS result. No line emission is detected at 6.4 keV ($`<3\times 10^5`$ph s<sup>-1</sup> cm<sup>-2</sup>, $`EW<60`$ eV, 90 per cent upper limits obtained from the joint fit to the SIS and GIS data). This extremely redshifted line profile cannot be explained by the diskline for a Schwarzschild black hole because of insufficient gravitational redshift (it may arise from infalling gas, as proposed by Reynolds & Begelman 1997, but then there would also be a large absorption edge; Young et al 1998). On the other hand, the diskline model for a Kerr black hole by Laor (1991; which is for a maximally-rotating black hole) gives a good fit. The result of this diskline fit is shown in Table 5, where the rest line energy, the inner radius and inclination of the disk are assumed to be 6.4 keV, 1.235$`r_\mathrm{g}`$, and 30. The outer radius is constrained well at $`(5\pm 1)`$$`r_\mathrm{g}`$ due to the well-defined decline of the redshifted, blue peak. The inferred negative emissivity index (it is poorly constrained) and the well-constrained outer radius perhaps suggest the line emission is concentrated in annuli around $``$ 5$`r_\mathrm{g}`$. A fit with a double-gaussian to the line profile is also given in Table 5. The line intensity of the line is about 3 times larger than that of the time-averaged one. The EW is $`700`$ eV when computed with respect to the continuum at 6.4 keV but $`400`$ eV to the continuum in the energy range of the observed line.
The line profile models available in XSPEC, the fits of which have just been reported above, are for complete disk annuli. It is possible however that just part of an annulus of the disk is irradiated during a flare, i.e. that part immediately below the flare itself. This offers further possible locations for the flare such as on the receding side of the disk where the peak at 5 keV is mostly due to the doppler effect, or on the approaching side much closer in where it is due to gravitational redshift. We show in Fig. 6 the locus of points which cause a 6.4 keV line in the disk frame to appear at 5 keV for an observer seeing the disk around a maximally spinning black hole at an inclination of $`30^{}`$. We have also fit the flare spectrum with model line profiles created from a disk divided into 36 azimuthal sectors and 24 radial bins between 1 and 25 $`r_\mathrm{g}`$. Although acceptable fits at $`3\sigma `$ level are found for most of the points around this locus, the region at small radii ($`2.5`$$`r_\mathrm{g}`$) on the approaching side of the disk is favoured most (see the confidence contours in Fig. 6). This is mainly due to the excess flux around 4 keV shown in Fig 5a and not the choice of grid size.
In the subsequent time-interval b, the averaged 0.6–10 keV count rate dropped by a factor of 2.2, compared with the flare interval (see Fig. 3). The line shape has then recovered to the ordinary one, as seen in the time-averaged spectrum (Fig. 5b). The line intensity, $`(2.2\pm 1.3)\times 10^4`$ph s<sup>-1</sup> cm<sup>-2</sup>, also dropped by a factor of 2.
## 5 Discussion
The long ASCA observation of 1997 has confirmed in detail the broad iron line in the Seyfert galaxy MCG–6-30-15. The time-averaged emission appears to originate from a disk extending between about 6 and 40 $`r_\mathrm{g}`$ of a massive black hole. If the rapid variability of this source is due to flares above the accretion disk, then the long term constancy of the line profile indicates that there are usually several flares at once on the disk and that the distribution of flares is almost constant in a time-averaged sense.
The spectrum of the source, and in particular the line profile, changed dramatically during a bright flare. Unlike the bright blue horn apparent during the 1994 flare, we now see essentially a bright red horn. If this is interpreted in the context of a relativistic diskline, the dominant flare must occur at smaller radii than usual. There are two possible locations of the flare, depending on the mass of the black hole in MCG–6-30-15.
One interesting possibility is a flare localized on the approaching side of the disk at $`2.5`$$`r_\mathrm{g}`$ (see Fig. 6). The duration of the flare is about 1 hour while the Keplerian orbital time around a $`10^7M_7`$M black hole is $`10^4M_7r_1^{3/2}`$s at $`10r_1`$$`r_\mathrm{g}`$. If the flare is confined within, say, 1/6 orbit at 2.5$`r_\mathrm{g}`$ (see Fig. 6), the duration of the flare (and peculiar line shape) requires the black hole mass to be larger than $`20M_7`$M. Therefore this solution is valid only if the black hole in MCG–6-30-15 is more massive than $`10^8`$M.
The spectral fit with the model for azimuthally-averaged line emission (Laor 1991) suggested that the line emission may be produced in a narrow range of radii around 5$`r_\mathrm{g}`$ during the major flare (see Table 2). The duration of the flare corresponds to $`1M_7^1`$ orbital time at 5$`r_\mathrm{g}`$. This is the preferred solution if the black hole mass is significantly smaller than $`10^8`$M.
There are some difficulties for the first interpretation. In order to restrict the line production to part of the disk, the flare must be placed very close to the disk surface. A flare on the approaching side of the disk is generally expected to be amplified due to relativistic beaming (e.g., Karas et al 1992), which appears to be consistent with the observed flux variation. However, at a small radii such as 2.5$`r_\mathrm{g}`$ on a disk inclined at $`30^{}`$, gravitational redshift and frame dragging overwhelm doppler boosting so that the emission reaching a distant observer is suppressed by more than an order of magnitude when the X-ray source is placed at 1$`r_\mathrm{g}`$ above the disk (M. Ruszkowski, priv. comm.). Therefore the flare would have to be intrinsically much more intense than observed. This may be possible if the emitted power increases rapidly towards inner radii around a spinning black hole. The strong light deflection implies that the reflection from the disk should also be enhanced by a factor of $`2`$ at the same time (see also Martocchia & Matt 1996). The closeness of the continuum source may also cause the disk surface to be highly ionized. Although the high energy end of the ASCA data is rather noisy, the 6–10 keV spectrum during the flare ($`\mathrm{\Gamma }=1.5\pm 0.4`$) suggests a possible spectral flattening which could be due to strong reflection.
The interpretations discussed above are, of course, not unique, but both require that the accretion disk extends close to the central black hole and that it spins rapidly, as suggested by Iwasawa et al (1996).
Although the profile of the broad line in MCG–6-30-15 appears to be fairly constant in a time-averaged sense, it does undergo dramatic changes every few days. Such changes offer interesting possibilities with which to probe different parts of the disk and to map the innermost regions about the black hole.
The lack of a narrow 6.4 keV line during the flare (the 90 per cent upper limit of intensity is only 40 per cent of the blue peak intensity of the time-averaged line) confirms the suggestion made by the previous ASCA observation (Iwasawa et al 1996) that there is little line emission from far out in the disk or torus. A narrow 6.4 keV line might be delayed by an hour or so, if it is produced around $`100M_7`$$`r_\mathrm{g}`$. Any line emission from farther out should be more constant because the variability of the line is smeared out. Evidence for such line emission appears to be weak.
## Acknowledgements
We thank all the members of the ASCA team. ACF and KI thank Royal Society and PPARC, respectively, for support. Chris Reynolds is thanked for his useful comments.
|
no-problem/9904/cond-mat9904421.html
|
ar5iv
|
text
|
# SIMPLE MODELS OF PROTEINS WITH REPULSIVE NON-NATIVE CONTACTS
Functionally useful proteins are sequences of amino acids that fold rapidly under appropriate conditions (temperature range, acidity of the water solution etc) into their native states commonly assumed to be their ground state configurations. The dynamics of folding is akin to motion in a rugged free energy landscape and it crucially depends on two factors: the interactions between the amino acids and the target conformations. In this paper, we focus on the role of non-native contact energies in the folding process. We study this issue in a simple model which is an extension of the so called Go model proposed by Go and Abe
Specifically, we consider the standard two-dimensional lattice model of 16 monomers. Its Hamiltonian is as follows
$$H=\underset{i<j}{}\alpha _{ij}\mathrm{\Delta }_{ij},$$
(1)
where $`\mathrm{\Delta }_{ij}=1`$ if monomers $`i`$ and $`j`$ are in contact and $`\mathrm{\Delta }_{ij}=0`$ otherwise (monomers $`i`$ and $`j`$ are considered to be in contact if they are seperated by one lattice bond and $`|ij|1`$). The quantity $`\alpha _{ij}=1`$ if monomers $`i`$ and $`j`$ are in contact in the native conformation and $`\alpha _{ij}=\alpha `$ otherwise. The sequence is thus defined by the native conformation. We allow $`\alpha `$ to be attractive ($`\alpha <0`$) or repulsive ($`\alpha >0`$). In what follows $`\alpha `$ is assumed to be larger than -1.0 so that the native state is guaranteed to be a maximally compact conformation. In the original Go model $`\alpha =0`$. It should be noted that using model (1) one can monitor the effect of non-native contact energies by varying only one parameter $`\alpha `$. Furthermore, the two-dimensional model is simple enough to study the effect of the target conformations on the folding dynamics.
It should be noted that there have already been several studies of models with repulsion in the non-native contacts: a Gaussian model , a designed model , and the so called HP+ model . The generalized Go model that we study here allows one to vary the strength, and sign, of the non-native contacts relative to the native ones.
We focus on 4 target conformations shown in Fig. 1. There are 37 compact conformations (for the two-dimensional 16-monomer chain one has 69 compact $`4\times 4`$ conformations but for the Go-like model only 38 of them remain different due to the end-to-end reversal symmetry and one is not accessible kinetically) which may act as the native conformations. Among these $`S_1`$ and $`S_4`$ shown in Fig. 1 are the two fastest folders at $`T0`$, whereas $`S_2`$ and $`S_3`$, also shown in Fig. 1, have intermediate folding properties.
It should be noted that the repulsive non-native contact interaction are expected to improve the foldicity because they restrict the size of the relevant phase space. In our case, however, this effect becomes so dramatic that for $`\alpha >0.1`$ sequences $`S_1`$ and $`S_3`$ can fold with a finite $`t_{fold}`$ even at $`T=0`$. Such an exotic phenomenon has been also observed in the HP model for some 13-monomer chains so it is not restricted to the Go-like sequences that we study. Here, we study how this arises as a function of $`\alpha `$ and show how do folding characteristics, i.e. characteristic temperatures and folding times, depend on $`\alpha `$. The ability of $`S_1`$ and $`S_3`$ to fold at $`T=0`$ may be partly understood by the fact that for these sequences the repulsive non-native interactions reduce the number of local minima by two orders of magnitude compared to the case of $`\alpha <0`$. Furthermore, repulsive interactions dramatically affect partitioning of the phase space into regions associated with the local energy minima. We demonstrate this by using the disconnectivity graph technique and show, in particular, that connectivities to the folding funnel become simplified significantly.
We have found that the folding temperature, $`T_f`$, increases with $`\alpha `$. This result agrees with that of Camacho for an effectively zero-dimensional model. We obtain it, however, not only by the numerical calculations for the lattice model but also by the analytical argument. The minimum folding time, $`t_{min}`$, defined at the temperature where the folding is fastest has found to decrease with $`\alpha `$ and it gets saturated for $`\alpha \mathrm{}`$.
The folding dynamics of a chain is studied by a Monte Carlo procedure that satisfies the detailed balance condition, and was motivated by the studies presented in Ref.. The dynamics allows for single and two-monomer (crankshaft) moves. For each conformation of the chain one has $`A`$ possible moves and the maximum value of $`A`$, $`A_{max}`$, is equal to $`A_{max}=N+2`$. In our 16-monomer case $`A_{max}=18`$. For a conformation with $`A`$ possible moves, probability to attempt any move is taken to be $`A/A_{max}`$ and probability not to do any attempt is 1-$`A/A_{max}`$ . In addition, probability to do a single move is reduced by the factor of 0.2 and to do the double move - by 0.8 . The attempts are rejected or accepted as in the standard Metropolis method. The folding time is equal to the total number of Monte Carlo attemps divided by $`A_{max}`$.
We have carried out the Monte Carlo simulations to determine the dependence of the median folding time, $`t_{fold}`$, on $`T`$ and $`\alpha `$. The results for $`S_1`$ and $`S_2`$ are shown in Figure 2. For each temperature, $`t_{fold}`$ is obtained based on 200 independent runs starting from random configurations. The results are averaged over 2 - 4 batches, of 200 trajectories each.
For sequence $`S_2`$ the standard U-shape for the temperature dependence of $`t_{fold}`$ is observed for all values of $`\alpha `$. In other words, no qualitative change occurs if the non-native contact energies change from attractive to repulsive. In the case of sequence $`S_1`$, however, for $`\alpha =1`$ the standard U-shape disappears suggesting that the glass transition temperature $`T_g`$ which is operationally defined as the value of the temperature at which the median time is equal to some cut-off value (usually this cut-off value is chosen to be 300000 Monte Carlo steps for the two-dimensional 16-monomer chain) becomes zero. In the standard scenario, at low temperatures the system may get trapped in some local minima and the folding process becomes extremly slow. The folding time is then governed by the Arrhenius law, $`t_{fold}\mathrm{exp}(\delta E/T)`$, where $`\delta E`$ is the energy barrier energy (at $`T=0`$ one has $`t_{fold}\mathrm{}`$). Thus, the absence of the $`U`$-shape dependence suggests that the energy barrier $`\delta E=0`$.
In order to know whether $`\delta E`$ is exactly zero or not one has to study the folding at $`T=0`$. At zero temperature the system gets trapped in some local minimum or in the native state. If the fraction of the trajectories from random conformations fold into the native state is bigger than 50$`\%`$, then the chain is said to be folded and $`\delta E=0`$.
The fraction of folded trajectories is shown in Fig. 3 for $`S_1`$, $`S_2`$, $`S_3`$ and $`S_4`$. For $`S_1`$ and $`S_3`$ this fraction becomes bigger than $`50\%`$ for $`\alpha >\alpha _c`$, where $`\alpha _c0.1`$. Sequences $`S_2`$ and $`S_4`$ have $`\delta E0`$ for any value of $`\alpha `$. It is interesting to mention that $`S_4`$ folds even faster than $`S_1`$ at $`T0`$ but its foldicity becomes much worse at $`T=0`$. Furthermore, the folding rate of $`S_3`$ at $`T0`$ is slower than for the three other sequences and yet $`S_3`$ can fold at $`T=0`$. Thus the geometry of the native targets has a dramatic effect on the folding at $`T=0`$. Among the 37 maximally compact $`4\times 4`$ structures it is only $`S_1`$ and $`S_3`$ that do not obey the Arrhenius law.
We study the dependence of $`T_f`$ on $`\alpha `$ for two typical sequences $`S_1`$ and $`S_2`$ shown in Fig. 1. The total number of conformations of the 16-monomer chain is only 802075 and is amenable to exact enumeration. This allows for an exact evaluation of the equilibrium parameters such as the folding temperature $`T_f`$. The latter is defined as a temperature at which the probability of occupancy of the native state is $`1/2`$. The results for sequences $`S_1`$ and $`S_2`$ are shown in Fig. 4 and 5. Obviously, $`T_f`$ increases with $`\alpha `$ but this dependence gets weaker for larger values of $`\alpha `$.
The increase of $`T_f`$ with $`\alpha `$ may be understood in the following simple way. Let $`P_{\mathrm{\Gamma }_0}(\alpha )`$ be the probability of occupying the native state of conformation $`\mathrm{\Gamma }_0`$. Then
$$P_{\mathrm{\Gamma }_0}(\alpha )=\frac{\mathrm{exp}^{\beta E_{\mathrm{\Gamma }_0}}}{_\mathrm{\Gamma }\mathrm{exp}^{\beta \alpha n\beta E_\mathrm{\Gamma }^{}}},$$
(2)
where $`E_0`$ is the energy of the native state, $`n`$ is the number of non-native contacts and $`E_\mathrm{\Gamma }^{}`$ is the part of energy in conformation $`\mathrm{\Gamma }`$ which corresponds to the native contacts. Then
$$\frac{P_{\mathrm{\Gamma }_0}}{\alpha }\beta <n>_T0,$$
(3)
where $`<n>_T`$ is the average number of non-native bonds at temperature $`T`$. So the probability of being in the native state cannot decrease with $`\alpha `$. $`T_f`$, therefore, should increase with increasing $`\alpha `$ and then become $`\alpha `$-independent.
It should be noted that in Camacho’s model $`T_f`$ was found to increase with $`\alpha `$ ($`\alpha <0`$) linearly. Our results presented in Figures 4 and 5 show that the region of $`\alpha `$ where one can observe the linear dependence is rather narrow. Such region becomes much wider, for example, in the case of the 27-monomer chain in three dimensions (the results are not shown). Overall, the results for $`T_f`$, shown in Fig. 4 and 5, demonstrate that the non-native contact repulsive energies improve both the thermodynamic stability and dynamical characteristics of folding.
We now focus on the $`\alpha `$ dependence of the temperature at which the folding time is minimal, $`T_{min}`$. The results for $`S_1`$ and $`S_2`$ are shown in Fig. 4 and Fig. 5. For positive values of $`\alpha `$, $`T_f`$ and $`T_{min}`$ are comparable for both sequences and they should be good folders. For large negative values of $`\alpha `$, $`T_{min}`$ is bigger than $`T_f`$ but $`S_1`$ and $`S_2`$ remain good folders by the Thirumalai-Camacho criterion \- the peaks of the structural susceptibility and the specific heat coincide for this interval of $`\alpha `$. (For $`\alpha =0`$, the coincidence of the peaks is, in fact, a general feature of the Go models because the proximity to the native state and occurence of rapid changes in an average energy as a function of $`T`$ are both controlled by establishment of the same native contacts).
Fig. 4 and 5 also show the dependence of the minimal folding time, $`t_{min}`$, on $`\alpha `$. The dependence seems to saturate at large values of $`\alpha `$. If one extends the Camacho result to positive $`\alpha `$ then $`t_{min}`$ should decrease with $`\alpha `$ exponentially. Our results suggest that such conclusion is valid only for $`\alpha <0`$ but not for $`\alpha >0`$. Thus, the repulsive non-native contact energies make the polypeptide chain to fold faster which is similar to what has been observed in Ref..
In order to understand why the repulsion improves the folding so much we study the dependence of number of local minima, $`N_{lm}`$, on $`\alpha `$. The results for $`S_1`$ and $`S_2`$ are shown in Fig.6. $`N_{lm}`$ of $`S_1`$ is found to be smaller than for $`S_2`$. Clearly, the number of local minima strongly depends on $`\alpha `$ and for $`\alpha >0`$ it become by 2 orders of magnitude smaller than that for $`\alpha <0`$. So, the positive values of $`\alpha `$ make the energy landscape less rugged and the folding dynamics get faster.
In order to get more insight into the nature of the energy landscapes in the models studied here, we use the disconnectivity graph technique which maps the potential energy surface onto the set of local minima . The technique involves checking what local energy minima are connected by pathways (sets of moves that are allowed kinetically) that do not exceed a given total threshold energy. For each value of this energy the minima are divided into disconnected sets of mutually accessible minima separated by barriers. The local minima which share the lowest energy threshold are joined at a node by lines and are called a basin corresponding to the threshold. The procedure of construction is stoped when one gets only one basin for all of the minima. For $`S_1`$ the number of local minima is equal to 152 and 81 for $`\alpha =0`$ and $`\alpha =1`$, respectively. The construction of the graphs, therefore, may be done exactly. The disconnectivity graphs obtained for $`S_1`$ and $`\alpha =0`$ and $`\alpha =1`$ are shown in Fig. 7 and Fig. 8, respectively. For both values of $`\alpha `$ sequence $`S_1`$ is a good folder and consequently, the structure corresponding to a folding funnel is clearly visible. However, the funnel for the repulsive case of $`\alpha =1`$ has a significantly less complex pattern than for $`\alpha =0`$. Moreover, the repulsive non-native interactions reduce the number of the local energy minima that have links to the native state below any predetermined energy threshold. Thus, the comparison of the disconnectivity graphs also shows that the repulsion may facilitate the folding substantially.
We now address the question of what happens with trapped local minima when $`\alpha `$ changes from negative to positive values. Fig. 9 shows the histogram of the local minima for $`\alpha =0.1`$ and 0.1 for sequence $`S_1`$ and $`S_2`$. Interestingly, for $`S_1`$ none of the local minima obtained for $`\alpha =0.1`$ appears for $`\alpha =0.1`$. The situation changes dramatically for $`S_2`$ for which 19 local minima are common for both $`\alpha =0.1`$ and $`\alpha =0.1`$. Our results suggest that for the sequence with $`\delta E=0`$ the local minima in which the chain is trapped at the negative values of $`\alpha `$ are effectively avoided if $`\alpha `$ is changed to positive values.
In conclusion, we state that for the simple extended Go model $`T_f`$ ($`t_{min}`$) increases (decreases) with the non-native contact energy and it gets saturated for large values of $`\alpha `$. The complexity of the disconnectivity trees becomes reduced on making $`\alpha `$ more and more repulsive so that some two dimensional sequences may even lose the Arrhenius like behavior of $`t_{fold}`$ at low temperatures. It would be interesting to determine whether there are any three-dimensional Go-like sequences that fold even at $`T=0`$. Another important question is what kind of an effective parameter $`\alpha `$, or its conceptual equivalent, characterizes real proteins.
We thank Jayanth R. Banavar, P. Garstecki and T. X. Hoang for many useful discussions. This work was supported by KBN (grant No. 2P03B-025-13).
|
no-problem/9904/cond-mat9904188.html
|
ar5iv
|
text
|
# Phonon mediated drag in double layer two dimensional electron systems
## I Introduction
Electron-phonon interactions directly affect numerous physical properties in solids. It is generally the case, however, that electrons and phonons are considered independently, with each system a weak perturbation on the other. This distinction is usually clear in consideration of collective modes of the electron system. Plasma oscillations, in particular, have been studied extensively by considering only the effects of Coulomb interactions between electrons. Of particular interests to this work is the limit in which the assumption of independent electron and phonon systems is not valid. If the interactions between electrons and phonons are strong enough, phonon mediated electron-electron (e-e) scattering could support a new collective mode.
A recent theoretical calculation proposed this type of electron-phonon collective mode in a double layer two-dimensional electron gas (2DEG) system, where both Coulomb and phonon mediated interactions are important. It was shown that a mode similar to plasmons emerges when the excitation energy $`\omega `$ approaches $`sq`$, where $`s`$ is the sound velocity and $`q`$ is the wavevector. The work predicted that this mode could be detected in measurements of phonon-mediated e-e scattering between two 2DEG layers. This coupled mode was argued to dominate and enhance such scattering when the phonon mean free path, $`l_{ph}`$, is larger than a critical value, $`l_c`$. The unexpectedly large magnitude observed in experiments provides clear motivation for investigating its existence, which would indicate a breakdown of the assumption of independent systems. When $`l_{ph}<l_c`$, this new mode was predicted to be negligible in the scattering process, and electrons and phonons could instead be treated independently. It was also established that the layer spacing and density dependences would show different behavior in the regimes of strongly coupled or independent electron-phonon systems. For small $`l_{ph}`$, i.e. the independent system regime, the scattering decreases logarithmically with increasing layer spacing, $`d`$, until $`dl_{ph}/2k_FL`$, beyond which it decreases exponentially ($`k_F`$ and $`L`$ are the Fermi wavevector and the width of the quantum well respectively). In the coupled mode regime, i.e. large $`l_{ph}`$, the scattering also varies logarithmically with $`d`$ but exhibits a local maximum at $`d\sqrt{l_{ph}/k_F}`$, beyond which it also rapidly declines. The dependence on the density ratio of the two layers is expected to show almost identical behavior in either regime, characterized by a peak at matched density. It was argued that distinct behaviors were present in the dependence on total density with the individual densities matched. For temperatures, $`T`$, between 2 K and 4 K, the scattering is predicted to increase with density in the coupled mode regime, while it remains nearly constant or decreases in the short mean free path regime. In both regimes, the scattering decreases with density at 1 K. The matched density dependence provides the central experimental test for the existence of the coupled collective mode.
In this paper, we report measurements of phonon-mediated drag examining the dependence on temperature, layer spacing, relative density, and matched densities. In addition to testing for the existence of the proposed coupled mode, these measurements explore the general properties of phonon drag in detail. New measurements on remotely spaced layers clearly confirm the fundamental elements of the phonon scattering process which is the basis for phonon drag, including the dominance of $`2k_F`$ scattering. These new measurements include density dependence measurements which are the first performed in samples in which all Coulomb scattering is absent. Additional measurements probe specific aspects of phonon drag, focusing on elements related to the existence of a coupled electron-phonon mode. The layer spacing dependence confirm the logarithmic behavior predicted theoretically for closely spaced layers. The value of $`l_{ph}`$ determined by fitting to the theoretical layer spacing dependence is found to be small and within the independent system regime. The dependence on total matched density for a sample in this regime closely mimics theoretical predictions for $`l_{ph}<l_c`$, providing additional evidence for independent systems. Contradictory evidence is provided by the magnitude of the drag signal, which is substantially larger than predicted for small $`l_{ph}`$, and by the value determined for $`l_{ph}`$ from the density dependence, which is much smaller than that found in thermal conductivity measurements on other samples. These apparent contradictions raise questions regarding phonon drag which have not been addressed in theoretical studies to date. Measurements of phonon drag in samples with the largest layer spacing, 5200 Å, raise additional questions regarding the underlying mechanism of phonon drag. The overall magnitude of the measured signal, in agreement with earlier measurements, lies well below the logarithmic dependence which applies for smaller layer spacings. Measurements of the dependence on matched densities at various temperatures, performed to test whether the deviation may be related to the independent or strongly coupled regime, provides no clear guidance, as the dependence observed is inconsistent with the predictions for either regime.
In the following section, the background of phonon drag and experimental details are presented. General properties of phonon-mediated drag are explored through measurements discussed in the following section. The subsequent section directly examines the possible existence of the coupled electron-phonon mode by measurements of the layer spacing dependence and the dependence on total density, and through comparison of these measurements with existing theoretical calculations. Finally, we present the results for a very remotely spaced layer sample and discuss several features which cannot be explained by current theory.
## II Phonon drag: background and technique
Interactions between electrons are known to have dramatic effects on the properties of 2DEG systems. While direct measurement of these interactions is generally difficult, it was proposed by Pogrebinskii and later by Price that the e-e interactions might be measurable in a double layer, independent, 2D electron system due to the presence of interactions between layers. Such measurements, termed electron drag, have been achieved in the double layer 2DEG systems, as well as in a 2D/3D electron system and the electron-hole double layer system.
In electron drag, a current, $`I`$, is driven through one of two closely spaced but electrically isolated 2DEG layers. Scattering between electrons in opposing layers results in a transfer of momentum from the current carrying layer to the other, effectively dragging the electrons in the second layer. If current is not permitted to flow from the second layer, charge accumulates at one end resulting in a voltage, $`V_D`$. The voltage increases until the force of the electric field due to charge accumulation balances the effective drag force resulting from interlayer interactions. Measurements of the drag resistivity, $`\rho _D`$, the ratio of the induced voltage to the drive current per square, has been shown to be directly related to an interlayer electron-electron scattering rate, $`\tau _D^1`$, through a simple Drude style relation $`\rho _D=m^{}/ne^2\tau _D`$, where $`m^{}`$ is the effective mass and $`n`$ is the electron density. This technique thus provides quantitative evaluation of e-e scattering.
Prior studies have established two distinct contributions to the interlayer e-e interactions: direct Coulomb scattering and phonon exchange scattering. For very closely spaced layers, the Coulomb contribution dominates, resulting in $`\rho _D`$ that varies as $`T^2`$ at low temperatures. This contribution to interlayer scattering has a strong layer spacing dependence, $`d^4`$. This results from a cutoff in the maximum scattering wavevector at the inverse of the layer spacing and modifications to screening efficiency as the layer spacing is changed. The strong layer spacing dependence permits phonon scattering to dominate $`\rho _D`$ for remotely spaced layers. However, the contribution of phonon scattering is still evident even for closely spaced layers through a peak near 2 K when $`\rho _D`$ is scaled by $`T^2`$, emphasizing deviations from the quadratic dependence of Coulomb scattering.
While previous studies clearly established the existence of a phonon mediated e-e scattering process and the important role of $`2k_F`$ scattering, questions regarding details of the underlying mechanism, especially in regard to the magnitude of $`\rho _D`$, are still not conclusively understood. Gramila et al. showed that real phonon exchange, while generating approximately the right temperature dependence, could not explain the magnitude of $`\rho _D`$ and proposed a virtual phonon exchange. Tso et al. included virtual phonon exchange in a calculation of $`\rho _D`$ and obtained reasonable agreement in both magnitude and temperature dependence, but questions have been raised about the electron-phonon coupling used in their work. Zhang et al. used a first-principles approach but found a phonon mediated drag which has the same layer spacing dependence as the Coulomb interaction, inconsistent with prior experimental results. Virtual phonons have also been considered by Badalyan et al.. The existence of the coupled electron-phonon collective mode was proposed by Bønsager et al. as a possible basis for the strong scattering if the phonon mean free path is large. The aim of this work is to firmly establish the essential physical elements of phonon drag and to test for the existence of the coupled electron-phonon collective mode.
The samples used in the study are GaAs/Al<sub>0.3</sub>Ga<sub>0.7</sub>As double quantum well structures grown by Molecular Beam Epitaxy. Two 2DEG layers formed in 200 Å wide quantum wells are separated by a potential barrier. Samples with barrier thicknesses of 225, 500, 2400, and 5000 Å were used, corresponding to well center-to-center spacings, $`d`$, of 425, 700, 2600, and 5200 Å, respectively. The density of each 2DEG layer as grown is approximately $`1.5\times 10^{11}`$ cm<sup>-2</sup> with mobilities $`2\times 10^6`$ cm<sup>2</sup>/V s. An active region, in which interactions are probed, was defined through standard photolithographic techniques in a mesa approximately $`440\mu `$m long and $`40\mu `$m wide. Electrical connection was provided through Au/Ge/Ni ohmic contacts. The ability to separately contact the individual 2DEG’s was provided through a gating technique which uses both front and backside aluminum Schottky gates. The use of an interlayer bias and application of a voltage to an overall top gate allowed the density of the layers to be altered, as calibrated by Shubnikov-de Haas measurements. In the drag measurement itself, low-frequency, low-excitation currents of $``$ 100 nA were used, with the nV or smaller drag signals detected using established drag techniques.
## III General Properties of Phonon Drag
This section presents new measurements of the temperature and density dependences of phonon drag in a sample with $`d=`$ 2600 Å. General features of the phonon scattering process are discussed. These include a temperature dependence which is characteristic of an electron-phonon scattering processes and the dominant role of $`2k_F`$ scattering.
### A Temperature dependence
Measured drag resistivity, $`\rho _D`$, scaled by $`T^2`$ as a function of temperature for the 2600 Å spacing sample is shown in Fig. 1. For this layer spacing, the strong $`d`$ dependence of the Coulomb interaction makes its contribution to $`\rho _D`$ completely negligible. The magnitude of Coulomb scattering can be estimated based on its $`d^4`$ spacing dependence and its magnitude as determined in other samples. The estimate for the Coulomb component of $`\rho _D/T^2`$ ($`0.5\mu \mathrm{\Omega }/\mathrm{}`$K<sup>2</sup>) is three orders of magnitude less than the phonon contribution. The most striking feature in these data is a clear transition which occurs near 2 K from a strong temperature dependence at lower temperatures to a dependence weaker than quadratic at higher temperatures. The general character of the transition is identical to that observed in other samples, for both closely spaced layers where Coulomb scattering is present and for a sample even more remotely spaced. This transition is a general feature of 2DEG phonon scattering and was first observed in the Bloch-Grüneisen transition of acoustic-phonon-limited mobility in 2DEG’s.
There are two essential elements which contribute to this change in temperature dependence. The first is a sharp cutoff for the wavevectors of phonons which can be scattered by the electron system. At low temperatures, only phonons with wavevector $`q<2k_F`$ can be thermally excited. All such phonons can participate in the scattering process. As the temperature increases, progressively larger wavevector phonon states become occupied. The increase in wavevector provides both a larger momentum transfer and access to more of the electron phase space, leading to a strong temperature dependence for phonon scattering. This increase continues until phonons with $`q2k_F`$ become thermally excited. Low energy excitations of the electron system are absent for changes in $`q>2k_F`$, so phonons with these $`q`$’s cannot scatter and conserve both energy and momentum. The transition to a weaker temperature dependence resulting from this cutoff will scale with the size of the Fermi surface.
A second element which contributes to the behavior is the enhanced phase space available for large angle scattering in the electron system, which diverges at zero temperature for a scattering wavevector equal to $`2k_F`$. The divergence allows $`2k_F`$ phonons to dominate scattering, even at temperatures below the energy of such phonons. The important role of $`2k_F`$ scattering is illustrated in Fig. 2. It shows the relative net momentum $`P_Q`$ transfered to the phonon system per unit time calculated for a single current carrying electron layer as a function of wavevector parallel to the 2DEG, $`Q`$. The calculation is directly related to earlier approaches for calculating phonon scattering. At 7 K, the momentum transfer rate is dominated by phonons in the vicinity of $`Q=2k_F`$ for both deformation potential and piezo-electric coupling. As the temperature is reduced to $``$ 3 K, this peak sharpens, with $`2k_F`$ scattering continuing to dominate the momentum transfer process. Even at 1.02 K, a fraction of the energy of a $`2k_F`$ phonon, $`2k_F`$ phonons continue to represent a significant portion of the total momentum, although low $`Q`$ phonons also become important.
This predominance of $`2k_F`$ scattering is critical for determining the temperature which characterizes the change in temperature dependence. If there were simply a cutoff at $`2k_F`$, the transition would be expected to occur when the characteristic phonon wavevector equals $`2k_F`$; at a temperature of $`T_c=2k_F\mathrm{}s/k_B`$, where $`s`$ is the sound velocity. By using the longitudinal acoustic-phonon velocity for $`s`$, the transition temperature $`T_c`$ 7.8 K is obtained; the transverse sound velocity yields $`T_c`$ 4.6 K. However, the strong contribution of $`2k_F`$ scattering still dominates well below this temperature. This dominance permits the scattering rate to be determined primarily by the thermal occupancy of $`2k_F`$ phonons, even at temperatures well below the expected transition. Indeed, the characteristic shape for the drag momentum transfer of Fig. 1 can be reasonably well represented by simply the thermal occupancy of the $`2k_F`$ phonons. At the lowest temperatures, where the contribution of $`2k_F`$ phonons is lost, a characteristic strong power law dependence is recovered.
### B Density dependence
Evidence for the dominance of $`2k_F`$ phonons in the drag process is available in the dependence of $`\rho _D`$ on density. The dependence of $`\rho _D`$ on the relative densities of the two layers is shown in Fig. 3.
In this measurement, the density of one layer is fixed while the density of the other is changed using an overall top gate. Measurements at four different temperatures for the 2600 Å spacing sample are shown. The key aspect of this measurement is a clear maximum in the drag resistivity when the densities of both layers are matched. At matched densities, each electron system has an identical Fermi surface size so the dominant emission of $`2k_F`$ phonons in one layer matches those phonons most likely to be absorbed by the other. When the densities of two layers are mismatched, $`2k_F`$ phonons of the higher density layer have $`q`$’s too large to permit absorption in the other layer. The substantial reduction in phonon exchange at $`q`$’s other than $`2k_F`$ cause a significant reduction in $`\rho _D`$ as compared to matched densities. While the peak at matched density has been observed earlier in a $`d=`$ 425 Å sample at 2.3 K and in a $`d=`$ 500 Å sample at 4.2 K, the behavior for both of those measurements included a significant contribution of Coulomb scattering. The data presented here can be directly compared to calculations of phonon drag without the significant added uncertainty arising from a subtraction of Coulomb scattering. The elimination of Coulomb scattering also permits the observation of a peak in $`\rho _D`$ down to previously unmeasured temperatures, where the signal is well below a nanovolt. This observation of a maximum in $`\rho _D`$ at matched density for this low temperature clearly supports the dominant role of $`2k_F`$ scattering, even for a temperature below the peak in $`\rho _D/T^2`$ of Fig. 1, and well below the temperature corresponding to the energy of a $`2k_F`$ phonon.
Further evidence for phonon scattering across the Fermi surface can be obtained from the temperature dependence measured for various matched layer densities. Measurements of $`\rho _D/T^2`$ are shown in Fig. 4 for three different densities with $`n`$ changing by a factor of 2. This measurement directly tests the assumption that the observed transition temperature is determined by the size of the Fermi surface. Since the size of $`2k_F`$ scales as $`\sqrt{n}`$, decreasing the density should move this transition to lower temperatures. Both the observed direction and magnitude of this change are in agreement with a dominance of phonon drag by $`2k_F`$ phonon exchange.
## IV Coupled Electron-phonon mode
In this section, measurements of layer spacing and matched density dependence are presented and compared to theoretical predictions. Fits to the phonon mean free path determined by this comparison, detailed density dependence measurements, and the overall magnitude of $`\rho _D`$ are considered as tests of the existence of a coupled electron phonon collective mode.
### A Layer spacing dependence
It has been experimentally established that phonon drag has a weak layer spacing dependence. A specific form of the dependence has not been studied in detail, however, until the recent calculation by Bønsager et al.. Their work predicts that in the limit of small phonon mean free path, drag varies as $`ln(d_a/d)`$ until $`d`$ reaches $`d_a`$, which equals $`l_{ph}/2k_FL`$. For $`d>d_a`$, $`\rho _D`$ decreases more abruptly as $`(d_a/d)exp(d/d_a)`$. In the coupled mode regime, which requires large phonon mean free paths, drag was also expected to decrease logarithmically as long as $`d`$ is less than $`d_B=(1+q_{TF}/2k_F)/16k_FC_{DP}`$, where $`q_{TF}`$ is the Thomas-Fermi wavevector and $`C_{DP}`$ is the dimensionless deformation potential coupling constant as defined in eq. (30) of their paper. For the parameters typical of our samples, $`d_B`$ is approximately 5000 Å. For $`d`$ greater than $`d_B`$, it was found that $`\rho _D`$ has a local maximum at $`d\sqrt{l_{ph}/k_F}`$, beyond which the electron-phonon coupled mode, which involves both electron layers, begins to separate into two independent modes, substantially reducing interlayer drag.
Measurements of the layer spacing dependence of $`\rho _D`$ provides a test of both this general theoretical approach, as a logarithmic spacing dependences should be observed, and as an indication of the existence of the electron-phonon collective mode. Measurements of $`\rho _D`$ for five different values of $`d`$ are shown in Fig. 5. Data for one layer spacing, $`d=`$ 375 Å, is reproduced
from the work of Ref. 2; current measurements for $`d=`$ 425, 700, and 5200 Å generally confirm prior experimental results. The 2600 Å spacing has not been previously measured. The data is plotted as $`\rho _D/T^2`$, where Coulomb contributions can be represented by a horizontal line. Visual inspection shows that the phonon contribution to $`\rho _D`$, which can be estimated by the deviation from a quadratic temperature dependence, shows little variation for layer spacing 700 Å or less, but reduces significantly for larger layer spacings. This variation can be quantified after subtraction of the Coulomb contribution of $`\rho _D`$. This contribution for $`d=`$ 700 Å is that determined in Ref. 3 and 4; a $`d^4`$ spacing dependence was used to infer the value for smaller layer spacings. No subtraction is necessary for $`d=`$ 2600 Å and 5200 Å. The resultant phonon scattering for the $`d=`$ 700 Å sample was used to fit the Coulomb adjusted data sets; a single multiplicative constant was applied to obtain a best fit to the other layer spacings. Nearly identical results for the 375 Å and 425 Å spacing samples are obtained if their Coulomb contribution is permitted to be a free fitting parameter. The overall relative magnitude of phonon drag determined through this fitting is shown in Fig. 6. A logarithmic layer spacing dependence is evident for the data with $`d`$ 2600 Å. This behavior does not extend to the 5200 Å spacing sample, the behavior of which will be postponed for a later section; we focus here on the smaller layer spacings.
The logarithmic dependence seen in Fig. 6 can be used to determine a value for the phonon mean free path, assuming that the drag corresponds to the independent system regime where $`\rho _Dln(d_a/d)`$. The corresponding spacing dependence is shown in the figure as a dashed line. The value determined for the phonon mean free path from this fit is 15 $`\mu `$m. This value is substantially less than the critical value which determines the onset of the dominance of the coupled mode regime; that value was shown to be $``$ 200 $`\mu `$m for the parameters of our sample. This small value for $`l_{ph}`$ appears to indicate that the coupled electron-phonon collective mode is absent.
A difficulty with the value obtained for $`l_{ph}`$ is that it differs considerably from the phonon mean free path determined through thermal conductivity measurements on similar samples. Measurements by Eisenstein et al. found $`l_{ph}`$ to be nearly two orders of magnitude larger. It is possible that this difference can be accounted for by the fact that the phonons which scatter from the 2DEG are predominantly $`2k_F`$ phonons. Interactions with the electron system could result in a $`l_{ph}`$ for $`2k_F`$ phonons substantially shorter than for phonons with other wavevectors. The removal of $`2k_F`$ phonons from a thermal conductivity measurement could have a small effect, as a broad range of phonon wavevectors contribute as determined by the Bose distribution. The scale of the difference in $`l_{ph}`$ weakens the use of its value as determined in Fig. 6 as a sole indicator for the absence of an electron-phonon collective mode.
### B Matched density dependence
An important additional test for the existence of the electron-phonon collective mode is the dependence on total matched density at various temperatures. This dependence has been calculated to show distinctly different behavior in the two regimes of $`l_{ph}`$. At 1 K in both regimes, $`\rho _D`$ should decrease monotonically as the density increases. At higher temperatures, a peak in the density dependence emerges for $`l_{ph}<l_c`$, i.e. for the case where no electron-phonon collective mode contributes to drag. The emergence of this peak at 2 to 4 K is accompanied by a change in the background density dependence. The decrease with increasing density seen at 1 K is smaller at 2 K. For 3 K and 4 K, there is only a small overall change with density in the magnitude of $`\rho _D`$. This behavior contrasts with that expected for $`l_{ph}>l_c`$, where the coupled electron-phonon collective mode dominates. In this regime, $`\rho _D`$’s behavior changes from a decrease with increasing density at 1 K to a substantial increase with density for higher temperatures. By 4 K the dependence is nearly proportional to density. These general differences in behavior, which are related to differences in screening and the role of the coupled electron-phonon collective mode, should be readily discernible in experiments.
Measurements of the density dependence of $`\rho _D/T^2`$ for the 2600 Å spacing sample are shown in Fig. 7. The general dependence seen at 1.2, 2.0, 3.0 and 4.0 K is strikingly similar to that predicted in the calculation of Ref. 1 for the small $`l_{ph}`$ regime. For each temperature measured, both the overall dependence on density and the emergence of the peak at high temperatures is in good agreement with predictions based on weak coupling and display clear differences from expectations for the strong coupling regime. These measurements clearly support the assertion that phonon drag in this sample corresponds to the short phonon mean free path regime, that is, the regime for which the coupled electron-phonon collective mode is absent.
A key element which distinguishes these measurements is the confidence that Coulomb scattering can be neglected. Earlier measurements which confirmed the existence of a peak in the relative density dependence at various matched densities, as had been seen in Ref. 4, also explored the dependence of $`\rho _D`$ at 4.2 K on total matched density. Those measurements, however, explore a density regime generally above the theoretical investigations of Ref. 1, and required a subtraction of Coulomb scattering. The relatively small increase of $`\rho _D`$ observed ($``$ 25 %) for densities between 2 and $`3.8\times 10^{11}`$ cm<sup>-2</sup> appear inconsistent with extrapolation of the current 4 K measurements; but this small increase assumes subtraction of a Coulomb contribution with a density dependence that has not been established in experiment. The measurements presented here can be compared to theory without the additional uncertainty introduced by such subtraction.
### C Magnitude of the drag
Although both the layer spacing and the matched density dependence show results clearly consistent with the short mean free path regime, there remains a contradiction as regards the magnitude of $`\rho _D`$. The measured magnitude is inconsistent with calculations assuming a small phonon mean free path, and is in much better agreement with the assumption of a large $`l_{ph}`$. It could be argued that the magnitude of $`\rho _D`$ alone could be considered evidence for the existence of a coupled electron-phonon collective mode. It is this mode which provides the increase in $`\rho _D`$ in Ref. 1 beyond that expected for ordinary phonon exchange, which has been shown to be too small to account for the strength of the phonon based interactions. This argument cannot be considered conclusive, however, as there is considerable uncertainty in the overall magnitude calculated for $`\rho _D`$. This uncertainty arises from the use of the random-phase approximation (RPA) for screening. It is possible that the use of a more complete description of screening would generate a substantially larger magnitude for $`\rho _D`$ where $`l_{ph}<l_c`$, which must increase by an order of magnitude to match the observed $`\rho _D`$. The contradiction between evidence for a short phonon mean free path, as seen in the layer spacing and density dependences, and that for a long phonon mean free path, as seen in the magnitude of $`\rho _D`$, remains. The question of the existence of the coupled electron-phonon collective mode has thus not been conclusively addressed.
## V Remotely Spaced Layer Sample
The discussion above is restricted to those samples for which a logarithmic spacing dependence is observed. Measurements made on a sample with $`d=`$ 5200 Å has a magnitude for $`\rho _D`$ which lies well below that expected from the logarithmic dependence, consistent with prior measurements. A deviation from a logarithmic dependence is expected in the short $`l_{ph}`$ regime for layer spacings larger than $`d_a`$ where $`\rho _D`$ decreases exponentially with $`d`$. Use of the value of $`l_{ph}`$ derived from the spacing dependence measurements yields a value for $`d_a`$ of 3.8 $`\mu `$m. Significant deviations from a logarithmic dependence should not occur in the small $`l_{ph}`$ limit until the spacing is an order of magnitude larger than 5200 Å.
The unexpected reduction in $`\rho _D`$ requires consideration of alternate origins. A possible explanation is that drag occurs in the large $`l_{ph}`$ collective mode limit despite the evidence for a small $`l_{ph}`$. In this regime, a change in characteristic behavior occurs at a length scale much smaller than for the small $`l_{ph}`$ regime. This characteristic layer spacing in our samples corresponds to a length of approximately 5000 Å. If the large $`l_{ph}`$ limit applies, a substantial deviation below the observed logarithmic dependence for $`d=`$ 5200 Å could then be consistent with expectations for a electron-phonon collective mode.
The absence of a clear indication of which regime is appropriate provides motivation for re-examining the dependence on matched densities for the 5200 Å sample. Such a measurement is extremely difficult, requiring accurate detection of signals as small as 250 pV. The central difficulty in measuring such small signals lies in ensuring that potential spurious signals are absent; this was verified here to a level of 30 pV. An additional complication for the measurement is that large layer spacings require a substantial interlayer bias to be applied, limiting the range of densities measured to between 1.1 and $`1.7\times 10^{11}`$ cm<sup>-2</sup> per layer.
The results of these measurements are shown in Fig. 8. A reduction in $`\rho _D`$ with increasing density at 1.2 K is consistent with the behavior of the 2600 Å spacing sample. For higher temperatures, however, there are distinct differences from the earlier measurements. A clear reduction in magnitude with increasing density is observed for all measured temperatures. There is no sign of a peak in the density dependence, as was clearly evident for the 2600 Å spacing sample. These data are inconsistent with both the small $`l_{ph}`$ limit, which well describes the density dependence for the 2600 Å spacing sample, and with the large $`l_{ph}`$ limit. In either case a peak in the density dependence can be discerned, but no peak is apparent in the 5200 Å spacing measurements. In both cases, the theoretical dependence at high temperatures makes a clear departure from the decreasing dependence typical at $``$ 1 K: to a weak density dependence for small $`l_{ph}`$ and to a strongly increasing density dependence for large $`l_{ph}`$. The 5200 Å spacing sample, by contrast, retains a decreasing density dependance as the temperature is increased, clearly contradicting either theoretical prediction.
The unusual behavior seen for this sample is not reflected in measurements exploring more fundamental elements of the electron-phonon scattering process, such as the presense of a cutoff at $`2k_F`$. This can be seen in the data of Fig. 9 which shows the temperature dependence of $`\rho _D/T^2`$ for two densities. The data show the change in $`T`$ dependence characteristic of this cutoff, and a shift in the transition temperature corresponding to the change in the size of the Fermi surface. Quantitative information about this transition temperature is somewhat uncertain, as it was not possible to verify the lack of spurious signals to a level below 30 pV. Adding a small offset to the measured signal can change the apparent peak position. The general confirmation of the properties related to the $`2k_F`$ wavevector cutoff in the 5200 Å spacing sample make its magnitude and its dependence on matched densities more puzzling. The only free parameter in the theoretical calculations is $`l_{ph}`$, which determines whether the electron and phonon systems are weakly or strongly coupled. We would expect this parameter not to vary between samples, as they are all essentially indentical apart from the size of the barrier. Yet the 5200 Å spacing sample is clearly distinct in behavior from the more closely spaced layer samples, and from theoretical predictions for either $`l_{ph}`$ regime. The measurements for this sample strongly suggest that understanding of interlayer phonon mediated electron-electron scattering is not yet complete.
## VI Conclusions
Extensive measurements of phonon mediated electron-electron scattering have been performed using electron drag on double layer 2DEG samples. Measurements for a sample in which Coulomb scattering is absent explored fundamental elements of this scattering process. These include the presence of a cutoff for scattering of phonons having wavevectors greater than twice the Fermi wavevector of the electron system and the dominance of $`2k_F`$ scattering even at relatively low temperatures. These properties of phonon drag, which are directly related to elements of electron-phonon scattering in single layers, have been verified in temperature and density dependence measurements.
Additional measurements explore the existence of a coupled electron-phonon collective mode, as has been recently proposed for 2DEG systems. Contradictory evidence concerning the existence of this mode has been observed. Both the dependence of the scattering on layer spacing below 3000 Å and its dependence on matched layer densities at various temperatures show behavior consistent with the absence of this mode. The overall magnitude of the scattering, however, suggests the presense of the collective mode. The phonon mean free path, as determined by fits to the spacing dependence assuming the absense of the mode, also differs by orders of magnitude with previous measurements of this length via thermal conductivity. Phonon drag measurements for a sample with a 5200 Å layer spacing raise further questions about details of the scattering process. Although the fundamental elements of scattering related to the size of the Fermi surface are confirmed in this sample, its scattering strength is well below the logarithmic spacing dependence found for more closely spaced layers. The sample further shows a decrease in scattering strength as the density of both layers is increased, a behavior generally inconsistent with theoretical predictions assuming either the absence or presense of the collective mode.
In general, while basic properties of phonon drag are well established, these measurements show that a full understanding of the process, especially with respect to the magnitude of the scattering and the presense of a coupled electron-phonon collective mode, requires further investigations.
###### Acknowledgements.
Discussions with A. H. MacDonald and M. C. Bønsager are gratefully acknowledged. We are indebted to Jim Eisenstein for his contributions to prior investigations of phonon drag and for his considerable efforts in the growth of the 2600 Å spacing sample. This work was supported by the NSF through grant DMR-9503080 and DMR-9802109, by the Alfred P. Sloan Foundation, and by the Research Corporations Cottrell Scholar program.
|
no-problem/9904/cond-mat9904171.html
|
ar5iv
|
text
|
# Collisional Cooling of a Charged Granular Medium
## I Introduction
The particles in most granular materials carry a net electrical charge. This charge emerges naturally due to contact electrification during transport or is artificially induced in industrial processes. It is well known , for instance, that particles always charge when transported through a pipe. In industry, contact electrification is used for dry separation of different plastic materials or salts , which tend to get oppositely charged and hence are deflected into opposite directions when falling through a condenser. Another application is powder varnishing, where uniformly charged pigment particles are blown towards the object to be painted, which is oppositely charged.
Whereas the dynamics of electrically neutral grains have been studied in great detail, little is known about what will change, if the grains are charged. In this paper we present the answer for collisional cooling, a basic phenomenon, which is responsible for many of the remarkable properties of dilute granular media. By collisional cooling one means that the relative motion of the grains, which lets them collide and can be compared to the thermal motion of molecules in a gas, becomes weaker with every collision, because energy is irreversibly transferred to the internal degrees of freedom of the grains.
In 1983 Haff showed, that the rate, at which the kinetic energy of the relative motion of the grains is dissipated in a homogeneous granular gas, is proportional to $`T^{3/2}`$, where $`T`$ is the so called granular temperature. It is defined as the mean square fluctuation of the grain velocities divided by the space dimension:
$$T=\stackrel{}{v}^2\stackrel{}{v}^2/3.$$
(1)
A consequence of this dissipation rate is that the granular temperature of a freely cooling granular gas decays with time as $`t^2`$. We shall discuss, how these laws change, if the particles are uniformly charged.
Due to the irreversible particle interactions large scale patterns form in granular media, such as planetary rings or the cellular patterns in vertically vibrated granular layers . This happens even without external driving , where one can distinguish a kinetic, a shearing and a clustering regime. The regimes depend on the density, the system size and on the restitution coefficient $`e_\mathrm{n}=v_\mathrm{n}^{}/v_\mathrm{n}`$, which is the ratio of the normal components of the relative velocities before and after a collision between two spherical grains. The $`T^{3/2}`$ cooling law holds, provided the restitution coefficient may be regarded as independent of $`v_\mathrm{n}`$ , and the system remains approximately homogeneous . The latter condition defines the kinetic regime, which is observed for the highest values of the restitution coefficient, whereas the two other regimes are more complicated because of the inhomogeneities. Such inhomogeneities can only occur as transients, if all particles are equally charged, because the Coulomb repulsion will homogenise the system again.
In order to avoid additional dissipation mechanisms due to eddy currents within the grains we consider only insulating materials. Unfortunately, up to now, no consistent microscopic theory for contact electrification of insulators exists . In powder processing two types of charge distribution are observed : A bipolar charging, where the charges of the particles in the powder can have opposite sign and the whole powder is almost neutral. The other case is monopolar charging, for which the particles tend to carry charges of the same sign and the countercharge is transferred to the container walls. It depends largely on the type of processing, whether one observes bipolar or monopolar charging, which means, that the material of the container, the material of the powder and other more ambiguous things, like air humidity or room temperature are important .
The outline of this article is as follows: The next section specifies the model we are considering. A simple derivation of the dissipation rate in dilute charged granular media based on kinetic gas theory is given in section III. We find, that the dissipation rate is essentially the one known from uncharged granular media multiplied with a Boltzmann factor. Section IV compares the analytic results with computer simulations. We find that in non-dilute systems the Coulomb repulsion is effectively reduced. This reduction will be explained, and we determine its dependence on the solid fraction of the granular gas. In the appendix we discuss the new simulation method we developed for this investigation. It is a molecular dynamics method, that avoids the so called brake-failure .
## II The model
In this paper, we consider monopolar charging, which is the usual case if insulators are transported through a metal pipe . For simplicity we assume, that all particles have the same point charge $`q`$ centred in a sphere of diameter $`d`$ and mass $`m`$. No polarisation and no charge transfer during contact will be considered. The particle velocities are assumed to be much smaller than the velocity of light, so that relativistic effects (retardation and magnetic fields due to the particle motion) can be neglected. The electrodynamic interaction between the particles can then be approximated by the Coulomb potential:
$$\mathrm{\Phi }_{ij}=q^2/r_{ij},$$
(2)
where $`r_{ij}`$ is the distance between the centers of particles $`i`$ and $`j`$.
We consider the collisions as being instantaneous, which is a good approximation for the dilute granular gas, where the time between collisions is much longer than the duration of the contact between two particles. As the incomplete restitution ($`e_\mathrm{n}<1`$) is the main dissipation mechanism in granular gases, Coulomb friction will be neglected in this paper. Also, the dependence of the restitution coefficient on the relative velocity will be ignored, so that the constant $`e_\mathrm{n}`$ is the only material parameter in our model.
The particles are confined to a volume $`V=L^3`$ with periodic boundary conditions in all three directions. The periodic volume can be thought of as a sufficiently homogeneous subpart of a larger system, which is kept from expanding by reflecting walls. For vanishing particle diameter this model corresponds to the One Component Plasma (OCP) . In the OCP a classical plasma is modelled by positive point charges (the ions) acting via the Coulomb potential, whereas the electrons are considered to be homogeneously smeared out over the whole system. In the OCP the electron background cannot be polarised, i.e. Debye screening does not exist, as is the case in our model, too.
## III Analytical results for dilute systems
In this section we derive an approximate expression for the dissipation rate in a dilute system of equally charged granular spheres, neglecting particle correlations. Basically we apply the kinetic gas theory, but include inelastic collisions. Using the analytic form of the dissipation rate in the dilute limit derived here, we will discuss the dissipation in a non-dilute system, where correlations are important, in the next chapter.
We start with calculating the collision frequency of a fixed particle $`i`$ with any of the other particles $`j`$. If they were not charged, two particles would collide provided the relative velocity $`\stackrel{}{u}`$ points into the direction of the distance vector $`\stackrel{}{r}=\stackrel{}{r}_j\stackrel{}{r}_i`$ connecting the particle centers, $`\stackrel{}{u}\stackrel{}{r}>0`$, and the impact parameter $`b=|\stackrel{}{r}\times \stackrel{}{u}|/u`$ is smaller than the sum of the particle radii, $`bb_{\mathrm{max}}=d`$. If the particles carry the charge $`q`$, they repel each other and the maximum impact parameter $`b_{\mathrm{max}}`$ becomes smaller than $`d`$ (see Fig.1). By the conservation laws for angular momentum and for energy one gets:
$$b_{\mathrm{max}}^2=d^2\left(1\frac{2E_\mathrm{q}}{\mu u^2}\right)$$
(3)
where $`\mu =m/2`$ is the reduced mass. $`E_\mathrm{q}=q^2/d`$ denotes the energy barrier which must be overcome to let two particles collide in the dilute limit. It is the difference of the potential energies at contact and when they are infinitely far apart. Eq. (3) is independent of the actual form of the potential, as long as it has radial symmetry. (Note that energy is conserved as long as the particles do not touch each other.)
Imagine a beam of particles, all having the same asymptotic velocity $`\stackrel{}{u}`$ far away from particle $`j`$. All particles within an asymptotic cylinder of radius $`b_{\mathrm{max}}`$ around the axis through the center of $`j`$ with the direction of $`\stackrel{}{u}`$ will collide with particle $`j`$. There will be $`\pi b_{\mathrm{max}}^2un`$ such collisions per unit time, where $`n=N/V`$ is the number density. Integrating over all relative velocities $`\stackrel{}{u}`$ gives the collision frequency of a single particle in the granular gas in mean field approximation:
$$f=\pi n\underset{uu_0}{}d^3uub_{\mathrm{max}}^2(u)p(u).$$
(4)
$`u_0=\sqrt{2E_\mathrm{q}/\mu }`$ is the minimal relative velocity at infinity, for which a collision can occur overcoming the repulsive interaction. We assume that the particle velocity distribution is Gaussian with variance $`3T`$ (see (1)), so that the relative velocity will have a Gaussian distribution as well, with
$$u^2=6T$$
(5)
Hence, the total number of binary collisions per unit time and per unit volume is given by:
$$\dot{N}_g=1/2fn=2\sqrt{\pi }n^2d^2\sqrt{T}\mathrm{exp}\left(\frac{E_\mathrm{q}}{mT}\right)$$
(6)
The factor $`1/2`$ avoids double counting of collisions. This corresponds to textbook physics for chemical reaction rates as can be found for example in Present.
Now we calculate the dissipation rate: The energy loss due to a single inelastic collision is:
$$\delta E(u,b)=\frac{\mu }{2}\left(1e_\mathrm{n}^2\right)u_{\mathrm{n}}^{}{}_{}{}^{2}$$
(7)
where $`u_\mathrm{n}^{}`$ means the normal component of the relative velocity $`\stackrel{}{u}^{}`$ at the collision. It can be calculated easily from $`u_{\mathrm{n}}^{}{}_{}{}^{2}=u_{}^{}{}_{}{}^{2}u_{\mathrm{t}}^{}{}_{}{}^{2}`$: The tangential component is determined by angular momentum conservation,
$$\mu ub=\mu u_\mathrm{t}^{}d,$$
(8)
and energy conservation gives
$$u_{}^{}{}_{}{}^{2}=u^2\left(1\frac{2E_\mathrm{q}}{\mu u^2}\right).$$
(9)
This yields
$$u_{\mathrm{n}}^{}{}_{}{}^{2}=u^2\left(1\frac{b^2}{d^2}\frac{2E_\mathrm{q}}{\mu u^2}\right)$$
(10)
The energy loss in one collision is therefore:
$$\delta E(u,b)=\frac{\mu }{2}\left(1e_\mathrm{n}^2\right)u^2\left(1\frac{b^2}{d^2}\frac{2E_\mathrm{q}}{\mu u^2}\right)$$
(11)
Assuming a homogeneous distribution of particles, we eliminate the $`b`$-dependence by averaging over the area $`\pi b_{\mathrm{max}}^2`$ (see Fig.1):
$`\delta E(u)`$ $`=`$ $`{\displaystyle \frac{1}{\pi b_{\mathrm{max}}^2}}{\displaystyle \underset{0}{\overset{b_{\mathrm{max}}}{}}}𝑑b\mathrm{\hspace{0.17em}2}\pi b\delta E(u,b)`$ (12)
$`=`$ $`{\displaystyle \frac{\mu }{4}}u^2\left(1e_\mathrm{n}^2\right)\left(1{\displaystyle \frac{2E_\mathrm{q}}{\mu u^2}}\right)`$ (13)
The dissipated energy per unit time due to collisions with relative velocity $`u`$ is then the number of such collisions per unit volume, $`1/2n^2\pi b_{\mathrm{max}}^2u`$, times the energy loss $`\delta E`$, Eq. (13).
Finally we get the dissipation rate per unit volume in the dilute limit ($`\nu 0`$) by integration over the relative velocity distribution:
$`\gamma `$ $`=`$ $`{\displaystyle \frac{\pi }{2}}n^2{\displaystyle \underset{uu_0}{}}d^3ub_{\mathrm{max}}^2u\delta E(u)p(u)`$ (14)
$`=`$ $`2\sqrt{\pi }n^2d^2m\left(1e_\mathrm{n}^2\right)T^{3/2}\mathrm{exp}\left({\displaystyle \frac{E_\mathrm{q}}{mT}}\right)`$ (15)
The dissipation rate of an uncharged granular system in the dilute limit in the kinetic regime is given by :
$$\gamma _0=2\sqrt{\pi }n^2d^2m\left(1e_\mathrm{n}^2\right)T^{3/2}$$
(16)
Thus the dissipation rate (15) in a monopolar charged dilute granular gas and the one for the uncharged case differ only by a Boltzmann factor, $`\gamma =\gamma _0\mathrm{exp}\left(E_\mathrm{q}/mT\right)`$. This is the main result of the analytic treatment in this section. It remains valid for any repulsive pair interaction between the grains that has rotational symmetry.
## IV Dissipation rate for dense systems
In order to discuss the dissipation rate $`\gamma `$ in a non-dilute system of charged granular matter, let us recall the analytic form of $`\gamma `$ in an uncharged non-dilute system. The derivation is basically done by using the Enskog expansion of the velocity distribution function for dense gases . One gets the dissipation rate for a non dilute uncharged system:
$$\gamma =\gamma _0g_{\mathrm{hs}}(\nu )$$
(17)
where $`\gamma _0`$ is given by Eq. (16) and $`g_{\mathrm{hs}}(\nu )`$ is the equilibrium pair distribution function of the non-dissipative hard-sphere fluid at contact. It only depends on the solid fraction $`\nu =\pi nd^3/6`$:
$$g_{\mathrm{hs}}(\nu )=\frac{2\nu }{2(1\nu )^3}$$
(18)
(Carnahan and Starling , Jenkins and Richman ).
Our system consists of dissipative charged hard-spheres (CHS). The Boltzmann factor in Eq. (15) is just the equilibrium pair distribution function at contact in the dilute limit for a CHS-fluid, $`lim_{\nu 0}g_{\mathrm{chs}}(\nu ,q)=\mathrm{exp}\left(E_\mathrm{q}/mT\right)`$. So it is plausible, that the dissipation rate for a dense system of dissipative CHS is:
$$\gamma =\gamma _0g_{\mathrm{chs}}(\nu ,q)$$
(19)
Unfortunately the literature is lacking a satisfying analytic expression for $`g_{\mathrm{chs}}`$. In 1972 Palmer and Weeks did a mean spherical model for the CHS and derived an analytic expression for $`g_{\mathrm{chs}}`$, but this approximation is poor for low densities. Many methods give $`g_{\mathrm{chs}}`$ as a result of integral equations, that can be solved numerically. We do not use those approximations, but make the following ansatz for $`g_{\mathrm{chs}}`$:
$$g_{\mathrm{chs}}(\nu ,q)g_{\mathrm{hs}}(\nu )\mathrm{exp}\left(\frac{E_{\mathrm{eff}}(\nu )}{mT}\right)$$
(20)
As in the dilute case we assume that the long range Coulomb repulsion modifies the pair correlation function of the uncharged hard sphere gas by a Boltzmann factor. Note that the granular temperature enters the pair correlation function only through this Boltzmann factor. The hard core repulsion is not connected with any energy scale, so that the pair correlation function $`g_{\mathrm{hs}}`$ cannot depend on $`T`$. The effective energy barrier $`E_{\mathrm{eff}}`$ must approach $`E_\mathrm{q}`$ in the dilute limit. Hence the ansatz (20) contains both the uncharged and the dilute limit, (17) respectively (15).
In order to check the ansatz (20) we did computer simulations using the MD algorithm as described in the appendix. Test systems of varying solid fraction $`\nu `$ and particle number ranging from $`N=256`$ to $`N=1024`$ were prepared at a starting temperature $`T_0`$. As soon as the simulation starts, the granular temperature drops because of the inelastic collisions. We measured the dissipation rate $`\gamma `$ and the granular temperature during this evolution. According to Eq. (20) and Eq. (19) the dissipation rate is $`\gamma =\gamma _0g_{\mathrm{hs}}(\nu )\mathrm{exp}(E_{\mathrm{eff}}(\nu )/mT)`$. An Arrhenius plot ($`\mathrm{ln}(\gamma /\gamma _0g_{\mathrm{hs}})`$ versus $`E_\mathrm{q}/mT`$) should give a straight line whose negative slope is the effective energy barrier $`E_{\mathrm{eff}}`$. Fig. 2 shows two examples of these simulations. The Arrhenius plots are linear to a very good approximation. This confirms the ansatz (20). Systems with high densities show slight deviations from linearity.
The negative slopes $`E_{\mathrm{eff}}/E_\mathrm{q}`$ in Fig. 2 are smaller than $`1`$, which means, that the effective energy barrier is smaller than in the dilute system. The explanation is that two particles which are about to collide not only repel each other but are also pushed together by being repelled from all the other charged particles in the system.
For dimensional reasons the effective energy barrier to be overcome, when two particles collide, must be of the form
$$E_{\mathrm{eff}}=\frac{q^2}{d}\frac{q^2}{\mathrm{}}f(d/\mathrm{}),$$
(21)
where $`\mathrm{}>d`$ is the typical distance between the charged particles and $`f`$ is a dimensionless function. The first term is the Coulomb interaction $`E_\mathrm{q}`$ of the collision partners at contact. The second term takes the interaction with all other particles in the system into account. It is negative, because the energy barrier for the collision is reduced in dense systems.
Obviously, for a dense packing, $`\mathrm{}d`$, the energy barrier for a collision must vanish, i.e. $`E_{\mathrm{eff}}|_{d=\mathrm{}}=0`$. Moreover, if one takes a dense packing and reduces the radii of all particles infinitesimally, keeping their centers in place, all particles should be force free for symmetry reasons. Therefore, the energy barrier must vanish at least quadratically in $`(\mathrm{}d)`$, i.e. $`E_{\mathrm{eff}}/d|_{d=\mathrm{}}=0`$. For the function $`f`$ this implies
$$f(1)=1\text{and}\frac{\mathrm{d}f(x)}{\mathrm{d}x}|_{x=1}=1.$$
(22)
If the particle diameter $`d`$ is much smaller than the typical distance $`\mathrm{}`$ between the particles, the function $`f(d/\mathrm{})`$ may be expanded to linear order,
$$f(x)=c_0+c_1x+\mathrm{}$$
(23)
In linear approximation the coefficients are determined by (22): $`c_0=2`$ and $`c_1=1`$. This determines the energy barrier (21).
In 1969 Salpeter and Van Horn pointed out, that inside a strongly coupled OCP a short-range body centered cubic (BCC) ordering will emerge. In the BCC lattice the nearest neighbour distance $`\mathrm{}`$ is related to the volume fraction $`\nu `$ by
$$\frac{d}{\mathrm{}}=\frac{2}{\sqrt{3}}\left(\frac{3}{\pi }\nu \right)^{1/3}1.14\nu ^{1/3}.$$
(24)
Assuming a BCC structure and using the linear approximation for $`f(x)`$ in (21), the effective energy barrier is therefore given by
$$E_{\mathrm{eff}}=E_\mathrm{q}\left(12.27\nu ^{1/3}+1.29\nu ^{2/3}\right)$$
(25)
To test Eq. (25) we simulated systems with densities ranging from $`\nu =0.001`$ to $`\nu =0.216`$ and determined the ratio $`E_{\mathrm{eff}}(\nu )/E_\mathrm{q}`$ as in Fig. 2. The results are plotted in Fig. 3. The agreement of the theoretical formula (25) with the simulations is excellent. One can see, that in the dilute limit the effective energy barrier extrapolates to $`E_\mathrm{q}`$. We cannot simulate low density systems, because collisions are too unlikely.
For the highest densities one cannot expect that the linear approximation (23) remains valid. Also, the dense packing of spheres is achieved with an FCC (face centered cubic) rather than a BCC ordering. This may be responsible for the systematic slight deviation from the theoretical curve in Fig. 3.
The reduction of the Coulomb repulsion was also found in the OCP, when it was applied to dense stars . There the analogue of the second term in (21) is called the “screening potential” (somewhat misleadingly, as there is no polarizable counter charge and hence no screening). Monte Carlo simulations of the OCP were interpreted in terms of a linear “screening potential” , which corresponds to (23), and the analogue of the conditions (22) also occurs in the plasma context , although based on a different physical reasoning. Corrections to the linear approximation are the subject of current research . However, applying these more sophisticated forms of the “screening potential” of the OCP model to dense charged granular gases seems arguable as for higher densities the influence of the hard spheres become more and more important and so the analogy to the OCP model, which uses point charges, does no longer hold.
## V Discussion
We derived the dissipation rate of a charged granular gas in the dilute limit. Compared to the uncharged case it is exponentially suppressed by a Boltzmann factor depending on the ratio between the Coulomb barrier and the granular temperature. This result was obtained assuming a Gaussian velocity distribution, although it is known that in the uncharged case deviations from a Gaussian behaviour emerge due to the inelastic collisions . These deviations, however, were shown to have little effect on the dissipation rate . As the system becomes less dissipative in our case, it is reasonable to expect that the effect of deviations from a Gaussian velocity distribution will be even weaker. One may say that a dilute granular gas with monopolar charging is more similar to a hard sphere gas in thermal equilibrium than a neutral one.
In a dense system particle correlations enter the collision statistics and hence the dissipation rate in two ways: First there is the well known Enskog correction as in the uncharged case. It describes that the excluded volume of the other particles enhances the probability that two particles are in contact. Second the Coulomb barrier which colliding particles must overcome is reduced and will vanish in the limit of a dense packing.
In our simulation we did not observe the shearing or the clustering instability, probably because our systems were rather small. It is reasonable to expect, however, that shearing or clustering instabilities may at most exist as transients in the presence of monopolar charging, because the Coulomb repulsion will homogenise the system in the long run.
## Acknowledgements
We thank Lothar Brendel, Alexander C. Schindler and Hendrik Meyer for useful comments. We gratefully acknowledge support by the Deutsche Forschungsgemeinschaft through grant No. Wo 577/1-2.
## A Computer Simulation Method
Distinct element (or molecular dynamics (MD)) simulations are usually done with time step driven or event driven algorithms . None of them is well suited for investigating a charged granular gas. Therefore we developed a new simulation scheme, which combines the virtues of both and will be described in this section.
We use a “brute force” MD algorithm, which is simple and sufficient for our problem. More sophisticated ways of dealing with the long range interactions, such as the multipolar expansion , the particle-particle-particle-mesh or the hypersystolic algorithms should be used, if larger systems need to be studied.
The event driven method for simulating the motion of all particles in the granular gas can be applied, whenever the particle trajectories between collisions can be calculated analytically, so that the time interval between one collision event and the next can be skipped in the simulation. Obviously this is impossible in a system with long range Coulomb interactions. However, the idea to avoid the detailed resolution of a collision event in time is still applicable. So the velocities of the collision partners are simply changed instantaneously to the new values predicted by momentum and angular momentum conservation and an energy loss determined by the restitution coefficient. We shall keep this feature of event driven simulations.
In the time step driven simulation method the equations of motion of all particles in the granular gas are discretized using a fixed time step, which is small compared to the duration of a collision. Hence each collision, which is modelled as an overlap between particles, is temporally resolved in detail. This has the advantage, that the formation of long lasting contacts between particles can in principle be simulated realistically. If the particles carry equal charges, however, this will not happen, so that the collisions may be approximated as being instantaneous like in event driven simulations. Apart from being more efficient, this automatically avoids the so called brake-failure artifact , which hampers time-step driven molecular dynamics simulations with rapid relative motion. On the other hand, we need a time discretization of the particle trajectories between collisions, in order to take the changing electrostatic interactions properly into account.
Because of the long-range nature of the Coulomb potential, we have to include the interactions with the periodic images of the particles in the basic cell. One way to do this is by Ewald summation. Details of this method can be found in . Another method is the minimum image convention: Only the nearest periodic image is taken into account for the calculation of the interactions. The minimum image method has the advantage, that it is much faster than the Ewald summation. We checked the validity of the minimum image method compared to the Ewald summation and found, that as long as $`E_\mathrm{q}/mT<10`$ both methods yield indistinguishable results. This upper limit for the coupling has been found before in Monte-Carlo simulations of the OCP . As our systems all satisfy this condition, we used the minimum image convention.
|
no-problem/9904/astro-ph9904065.html
|
ar5iv
|
text
|
# A Test of the Standard Hypothesis for the Origin of the H I Holes in Holmberg II
## 1 Introduction
High spatial and spectral resolution observations of the neutral hydrogen gas in nearby dwarf and Magellanic irregular galaxies have revealed remarkably intricate and complex structures (Puche et al. (1992), hereafter P92; Puche & Westpfahl (1994); Westpfahl & Puche (1994); Staveley-Smith et al. (1997), Kim et al. (1998)). Numerous holes are visible in the H I distributions of these galaxies, surrounded by shells of higher density. In Holmberg II (Ho II, DDO 50) for example, P92 identify 51 holes, and show that in many cases expansion of the gas surrounding the holes is directly detectable. Typical expansion velocities of 4 – 10 km s<sup>-1</sup> have been measured, and holes as large as 1600 pc across are present. This hole–shell morphology for the H I gas appears to be quite common, being present in essentially all of the nearby dwarfs studied to date with adequate spatial and velocity resolution.
Previous studies of the H I distributions in nearby spiral galaxies such as M31 and M33 (Brinks & Bajaja (1986); Deul & den Hartog (1990)) also reveal the presence of H I holes. The origin of these features has generally been attributed to stellar winds and supernova explosions (SNe) from young stellar associations and clusters. This previous work naturally suggested that the holes in Ho II and other dwarfs were caused by a similar process. P92 put forth the hypothesis that the holes in Ho II were due to the combined action of hot stellar winds from O and B stars, plus the SNe shocks from the same massive stars after they exploded at the end of their lifetimes. Based on their measured expansion rates and hole sizes, they determined that as much as $``$2 $`\times `$ 10<sup>53</sup> ergs of kinetic energy, the equivalent of $``$200 SNe, was required to produce the largest holes. The more typical holes require a few to several dozen SNe.
Although the SNe hypothesis for explaining the H I holes in Ho II is an appealing one, it is not without problems. Several of the H I holes are located at large distances from the center of the galaxy, in regions of both low optical surface brightness and low H I column density. At least 18 holes are located beyond the Holmberg radius of the galaxy (the radius where the $`B`$-band surface brightness drops below 26.6 magnitudes/square arcsec), where it is unlikely that large numbers of massive stars could have formed. The H I column densities at these radii are well below the canonical threshold density of 10<sup>21</sup> cm<sup>-2</sup> (Kennicutt (1989)), a further indication that star formation is unlikely to occur at these locations in the galaxy. Therefore, it seemed appropriate to investigate further the SNe hypothesis of P92, to determine whether it was possible to corroborate the general picture or, alternatively, to rule it out.
The supernova scenario described in P92 does in fact provide us with a direct observational test. Those authors found that many of the expanding holes in Ho II require the kinetic energy input of $``$10 to as many as 200 SNe each. Such multiple supernova events should only occur in massive clusters or OB associations. Age estimates of the H I holes found in Ho II have been assigned based on their sizes and expansion velocities, and typical ages fall in the range of 10<sup>7</sup> to 10<sup>8</sup> years. Consequently — if the ages are indeed accurate — the upper main sequence stars (late-B, A and F) should still be present in the clusters which produced the SN precursors. During the brief time period since these clusters formed, they will not have dispersed significantly, and should still be observable as blue sources at the centers of the H I holes. Since the surface brightness level associated with the underlying old population of stars in Ho II is fairly low, the upper main sequence population in these clusters should readily stand out. If these young star clusters produced, for example, 50 OB stars which became SNe, then a Salpeter IMF would predict the existence of at least 300 upper main sequence stars which would still be present after 10<sup>8</sup> years. Although this population of stars would not be resolved from the ground (a single main sequence A0 star would have m<sub>V</sub> = 28 and 1 arc second corresponds to 15 parsecs at the distance of Ho II), the total cluster brightness should be m<sub>V</sub> $``$ 22, which would be readily detectable in deep CCD images.
In order to look for the young clusters which would provide evidence for the SN scenario, we decided to carry out a deep, multi-color imaging study of Ho II. Our main goal was to obtain accurate optical magnitudes and colors of all objects appearing in and around the locations of the H I holes found in the VLA maps. The information tabulated in P92 regarding the ages and energy requirements of the holes can be used to calculate the magnitudes and colors of the clusters that should be present in the holes if the SN scenario is correct. Direct comparison of these calculations and our imaging photometry allows us to carry out a sensitive test of the SN scenario.
In the following section, details of the optical observations carried out for this study are discussed, along with the reduction steps executed on the data, and a complete description of the photometric measurements of the H I hole regions and of individual sources located in or around the holes. Section 3 describes calculations of the sizes and brightnesses of the putative clusters made using the information tabulated in P92. The last two sections of the paper consist of a discussion of our findings and their implications for the scenario proposed in P92, followed by a summary and final remarks.
## 2 Observations
### 2.1 Description and Preliminary Reductions
Observations of Ho II were obtained in February 1994 and April 1995 with the 0.9-meter telescope at Kitt Peak National Observatory. The galaxy was imaged in three broad-band filters ($`BVR`$) during both observing runs, and in narrow-band H$`\alpha `$ during the February 1994 observing run. The detector used was a Tektronix 2048$`\times `$2048 CCD (T2KA), formatted to read out only the central 1280$`\times `$1280 pixels. Each pixel subtended 0.68″ on the sky, resulting in a total field-of-view 14.5′ on a side. Multiple exposures were obtained through each filter, and the telescope was dithered between exposures, to facilitate removal of particle events (cosmic rays) in the images. Total integration times were 2400 s in $`B`$, 1800 s in $`V`$, and 1200 s in $`R`$ for the 1994 data, and 2700 s in $`B`$, 1800 s in $`V`$, and 1800 s in $`R`$ for the 1995 data. Images were taken under clear observing conditions. Photometric standards (Landolt (1983, 1992)) were also observed for use as calibration sources. Observations of additional dwarf galaxies were obtained during these runs (Ho I, K 73, M81dwA, IC 2574, Leo A, Sex A, Gr8, DDO 147); the results for these objects will be presented in a subsequent paper (Salzer et al. (1999)).
Images taken during the 1994 and 1995 observing runs were reduced separately. Preliminary reductions (overscan level subtraction, bias image subtraction, flat field division) were carried out following standard practices. Multiple images taken in sequence through a particular filter were aligned with the middle image in the sequence. Sky subtraction was accomplished by defining a series of rectangular regions surrounding the galaxy, measuring the mean flux level in those regions (after masking stars and cosmic rays), and subtracting that flux level from the image. Multiple images in a given filter were scaled to a common flux level (to preserve photometric integrity) and combined into a single image, using a pixel-rejection algorithm to eliminate cosmic rays. The combined $`B`$, $`V`$, and $`R`$ images were rotated to a north-up, east-left orientation, and their central coordinates were determined using an astrometry routine which measures the positions of Guide Star Catalog (Lasker et al. (1990)) objects appearing in the field.
A total of six standard stars (Landolt (1983, 1992)) were used to calibrate the broad-band images of Ho II from the February 1994 run, while 21 were available for the April 1995 data. In all cases, the photometric zero-point constants were determined with an accuracy of $``$0.01 magnitude, i.e., the nights were photometric.
In order to create the deepest possible $`BVR`$ images with which to do photometry of the hole regions, the images from the two runs were convolved to a common resolution, then scaled, aligned, and combined to create a single, deep image in each filter. The fluxes were scaled to the April 1995 values, since the photometric calibration was deemed to be of slightly higher quality for that run. Photometry was carried out on faint point sources in the resultant images to quantify the detection limit. The 4-$`\sigma `$ limit on the brightness of any point source is $`B`$ $`=`$ 23. The resolution (PSF FWHM) of each of the final combined $`BVR`$ images is 2.4″.
A continuum-subtracted H$`\alpha `$ image was created using the images taken in February 1994 through the on-band ($`\lambda _o`$ = 6569 Å, $`\mathrm{\Delta }\lambda `$ = 89 Å) and off-band ($`\lambda _o`$ = 6409 Å, $`\mathrm{\Delta }\lambda `$ = 88 Å) filters. Both narrow-band images were aligned with the composite $`BVR`$ frames and convolved to a common resolution. The off-band image was scaled to the on-band image by comparing the fluxes for several bright stars in the two images, after which it was subtracted from the on-band image to produce the final H$`\alpha `$ frame. Since the narrow-band images were acquired under non-photometric conditions, no flux calibration was attempted.
The high-resolution (natural weight) H I map of Ho II from P92 was used for this study to provide information about the hole locations. To facilitate comparison between the optical and radio data, the format of the H I map was modified to match that of the optical images. Specifically, the original H I map (2″/pixel) was rescaled to the optical image scale (0.68″/pixel). The section of the radio map corresponding to the area covered by the optical frames — i.e., having the same size and central coordinates as the optical images — was then extracted and used to create a new image. Figure A Test of the Standard Hypothesis for the Origin of the H I Holes in Holmberg II shows the high-resolution H I map after it has been scaled and aligned with the optical image, and Figure A Test of the Standard Hypothesis for the Origin of the H I Holes in Holmberg II shows a composite $`BVR`$ image, created by combining the $`B`$, $`V`$, and $`R`$-band data taken in April 1995.
### 2.2 Photometry
#### 2.2.1 H I Holes
To search for evidence of star clusters at the centers of the H I holes, photometric measurements were carried out on the combined $`B`$, $`V`$, and $`R`$ images created by merging the data from the February 1994 and April 1995 observing runs. These combined frames provided us with the deepest images possible for our search.
The photometry was performed using the following strategy. The hole positions given in Table 5 of P92 were used as an initial guess for locations of our synthetic apertures. These apertures were overlaid on the scaled H I image, and the positions adjusted slightly when required to center the apertures within the H I holes. This was necessary since the reported accuracy of the P92 positions was $`\pm `$0.1′. Next, we divided the holes into four categories according to size (again based on the data from P92), and assigned aperture sizes accordingly. The apertures adopted for the four size groups are listed in Table 1. The decision to use variable aperture sizes for the photometry was motivated by several factors: (1) the central positions of the holes become increasingly uncertain for larger holes; (2) there is no a priori reason to expect the putative star clusters to be precisely in the hole centers; (3) the larger holes tend to be older so that diffusion would tend to increase the size of the clusters in these holes. The apertures used ensure that we did not miss the clusters.
Because substantial galaxian background is present in many hole locations, we measured the flux in both a circular aperture corresponding to the hole location and in a concentric annular ring surrounding the aperture. This allows us to compare the fluxes and colors in the two regions. The area of each annular region was chosen to be approximately equal to that of the corresponding circular aperture to ensure comparable signal-to-noise for the two measurements. The test for the presence of a star cluster within the H I holes then involves comparing the fluxes within each aperture/annulus pair to look for an excess of flux (expected to be blue in color) coming from the central circular aperture.
In several instances, bright foreground stars were located inside the aperture or annular regions we had marked for photometry. In such cases we either adjusted the positions of the hole centers slightly, if possible, to avoid contamination from these objects, or excluded those holes from the sample. Photometry was successfully executed on 44 of the 51 hole regions identified in P92. The locations of the apertures and annuli are shown on the radio map (Figure A Test of the Standard Hypothesis for the Origin of the H I Holes in Holmberg II) and on the composite $`BVR`$ optical image (Figures A Test of the Standard Hypothesis for the Origin of the H I Holes in Holmberg II). Results from photometry of the aperture/annular regions are given in Table 2. The equatorial coordinates of our aperture centers are given for each hole, along with the aperture (inner diameter) and annular (outer diameter) sizes in arc seconds, plus the $`B`$ band magnitude, $`B`$$``$$`V`$ and $`B`$$``$$`R`$ colors, and average $`B`$-band surface brightness of the integrated light within both the aperture and annulus. Throughout this paper we use 1950 coordinates, for consistency with P92. Note that the background-subtracted flux measured in the annulus of hole #15, as well as for both the aperture and annulus of hole #27, was slightly negative, resulting in indeterminate magnitudes and colors. These two holes contain no measurable light from either foreground objects or from Ho II itself.
#### 2.2.2 Point Sources in the HI Hole Regions
A number of the holes in Ho II had faint objects which appeared within the region delimited by our circular apertures or in the surrounding annulus. Since these objects could be the remnant star clusters which produced the sequential SN explosions described in P92, we wished to measure accurate magnitudes and colors for them. Photometry was performed on 29 such objects using a 5″ diameter aperture and a background annulus with an inner diameter of 8″ and an outer diameter of 20″.
Because we were measuring magnitudes and colors of individual sources rather than an entire hole region, we decided to use the $`BVR`$ images from the April 1995 observing run alone for this step, rather than using the images merged from both observing runs. The April 1995 data were of better image quality compared to those from February 1994, and this improvement in resolution almost completely compensated for the added depth of the combined images with regard to point-source detection. The resolution (PSF FWHM) of the April 1995 broad-band images is $`1.71.8`$″, and the 4$`\sigma `$ limit on the brightness of a point source is $`B=23`$. Results from the point source photometry are given in Table 3. The first column lists an object designation reflecting the hole number in which the object is located, the next two columns give the position of the object, and the $`B`$ magnitude and colors of the source are given in columns 4 – 6. We note that the flux from any object within the apertures or annuli was measured, regardless of the object’s brightness or location. Therefore, this list should not be taken as a list of possible candidate star clusters. In many cases the objects listed have magnitudes and colors consistent with their being faint foreground stars.
### 2.3 Results of Photometry
We summarize the results of our imaging and photometric analysis in Table 4. In this table, each of the 51 H I holes is characterized with regard to the following key question: is there evidence for a star cluster at or near the center of the H I hole? The categories into which each hole has been assigned are: (1) Empty Hole – No source within the central aperture exceeding a flux level of 3$`\sigma `$ above the annular flux level. (2) Galaxian Background – Source(s) present in the hole, but with the characteristics/appearance of general galaxian background light rather than the putative star clusters. (3) Possible Star Cluster – Source(s) within the hole with the correct characteristics (color, brightness, appearance) of a genuine star cluster. (4) Possible Photoionization Region – The hole is coincident with an H II region and is most likely a cavity of photoionized gas rather than a wind/SNe-blown hole. (5) Faint Foreground Star – Photometry reveals that the object in the hole has the brightness and colors consistent with it being a foreground star not associated with Ho II. (6) Contaminated/No Photometry – A bright foreground star is present in the hole; no photometry was attempted.
It should be stressed that the assignment of a given hole into one of the six categories is by no means unambiguous in all cases. For example, assigning objects to category 2 (galaxian background) as opposed to category 5 (faint foreground star) was at times subjective. Further, we emphasize that objects in category 3 (possible star cluster) are assigned that designation even if the color and brightness of the source is only broadly consistent with the expected values. An object classified as category 3 cannot be interpreted as having been shown to be a young star cluster within Ho II. Rather, it is consistent with that hypothesis, but could just as likely be a foreground star (e.g., a halo white dwarf). Two holes (numbers 1 and 8) have point source photometry listed in Table 3, but the sources in question are located in the outer annuli. Hence, these two holes are classified as category 1.
The main results from Table 4 that we want to emphasize are: (1) nearly one third of the holes have no obvious optical source within our photometric apertures, and (2) only a minority of holes (6 of 44 for which photometry was obtained) have optical sources with the colors and brightnesses consistent with the expected star cluster. One hole (#43) is located in a complex of bright H II regions, and may represent a hole which is created by photoionization of the H I gas by the O and B stars. The status of the remaining holes (category 2 and 5) are uncertain, owing to the foreground or galaxian light they contain. However, in most cases this light is inconsistent with the properties of the putative star clusters, as quantified in the following section.
We note that the photometry listed in Tables 2 and 3 has not been corrected for reddening, either due to the Milky Way or internal to Ho II. Burstein & Heiles (1984) list a color excess of $`E`$($`B`$$``$$`V`$) $`=`$ 0.03 due to foreground reddening in the direction of Ho II. With regard to intrinsic absorption in Ho II, spectra of four H II regions by Hunter & Gallagher (1985) yield Balmer decrements (H$`\alpha `$/H$`\beta `$) consistent with zero or modest reddening. Hence, we believe that dust is not significantly affecting our photometry. This should be especially true in the outermost regions of Ho II, where our photometry provides the most sensitive test of the SNe hypothesis.
We will return to the photometry results in more detail in section 4.
## 3 Modeling Analysis
P92 catalogued the observed properties of the H I holes in Ho II, and from these, derived quantities such as the ages of the holes and the kinetic energies required to create them. Some of these observed and derived quantities (such as the radial expansion of the holes, and their energy requirements) were interpreted as evidence in favor of the stellar wind/multiple SNe scenario for the origin of the holes. In order to compare directly our observations with predictions that arise from the SN scenario, we have used the hole properties tabulated in P92 to derive the observable characteristics of the clusters which should exist if the SN hypothesis is correct.
A number of the quantities tabulated in P92 are distance-dependent; for example, the kinetic energies they calculated are given by the equation (Chevalier (1974)):
$$E=5.3\times 10^{43}n_{HI}^{1.12}R^{3.12}V^{1.4}\mathrm{ergs}$$
(1)
where $`n_{HI}`$ is the volume density of the surrounding medium in particles per cubic centimeter, $`R`$ is the radius of the hole in parsecs, and $`V`$ is the expansion velocity of the hole in km s<sup>-1</sup>. P92 adopted a distance of 3.2 Mpc for Ho II. A more recent distance determination has been done by Hoessel et al. (1998) using Cepheid variable stars in Ho II. We have adopted their revised distance modulus for our calculations: $`(mM)=27.42`$, which yields a distance of 3.05 Mpc. Therefore we have used the data tabulated in P92 but have re-scaled it in accordance with this revised distance. Some of the relevant quantities that changed with the new distance have been included in Table 5. For most holes, the revised distance resulted in computed hole energies that are $``$14% less than those tabulated by P92. Note, however, that a typographical error in P92 caused their published energy value for hole #43 to be underestimated by more than an order of magnitude. Our table reflects the corrected energy for this hole.
To derive the characteristics that should be observable if the remnant clusters are located inside the H I holes, we proceeded as follows: using the kinetic energies of the holes, the number of supernova explosions required to create each hole was calculated. The energy imparted to the ISM by one supernova explosion was taken to be 10<sup>51</sup> ergs (cf. McCray & Kafatos (1987)). If the number of supernova explosions required to create a hole was $`<`$1, that hole was omitted from the rest of the calculations. For each remaining hole, a Salpeter IMF (Salpeter (1955)) was used to calculate the mass distribution for a model star cluster. This distribution was then scaled so that the number of stars of mass $``$ 7 $`M_{\mathrm{}}`$ was equal to the number of supernovae required to create that particular hole. (Stars with masses $``$ 7 $`M_{\mathrm{}}`$ were assumed to be of sufficient mass to end their lives as Type II supernovae.) Next, stars which would have evolved off the main sequence over a time scale equal to the age of the hole were removed from the distribution. Finally, composite magnitudes and colors were calculated for each of the synthetic clusters, for comparison with our observations.
The same calculations were repeated for a Miller-Scalo IMF (Miller & Scalo (1979)) and using a limit of $``$ 8 $`M_{\mathrm{}}`$ for the lowest mass stars which end their lives as Type II supernovae. Performing the calculation with these two sets of parameters provides us with a reasonable range of predicted brightnesses for the putative clusters, and reflects the current levels of uncertainty in both the slope of the IMF and lower mass limit for Type II supernova precursors.
We note that in this simple calculation we make no effort to take into account the light that the post-main-sequence stars would contribute to the cluster total. For clusters in the age range considered here, red supergiants might make a significant contribution to the total light. By ignoring the evolved stars, our models are providing only lower limits to the total cluster brightness. This must be kept in mind when comparisons are made with the observations.
The final results of the modeling analysis are given in Table 5. The hole energies and ages listed in columns 2 and 3 differ slightly from those tabulated by P92, for the reasons mentioned above. The predicted number of SNe for each hole is one-tenth of the number given in column 2 (i.e., the hole energy divided by 10<sup>51</sup>). Only those holes with energies in excess of 10<sup>51</sup> ergs (i.e., those which require at least one SN) are listed in the table. The $`B`$ magnitude and $`B`$$``$$`V`$ color listed under the heading “Model I” are the composite values for the model clusters computed using the Salpeter IMF and a Type II SN mass limit of 7 $`M_{\mathrm{}}`$. The corresponding values listed under “Model II” are for models using a Miller-Scalo IMF and a mass limit of 8 $`M_{\mathrm{}}`$. In general, Model II predicts brighter clusters.
For many of the holes, the predicted cluster brightnesses are well below our observational limits, and hence we are unable to say anything definitive regarding the presence or absence of a young stellar population. However, for holes requiring energies in excess of 10<sup>52</sup> ergs (or $`>`$ 10 SNe) the expected brightness of the putative star cluster is at or above the limits set by our data. These holes will receive careful scrutiny in the next section.
## 4 Discussion
### 4.1 Results
The main results of our study are summarized in Table 6, which combines data from Tables 2, 3, 4, and 5 for specific holes. In Section 2.1 we established that a point source with $`B`$ $`=`$ 23.0 would be detected at the 4$`\sigma `$ level in our images. We list in Table 6 all H I holes that are predicted to contain star clusters as bright or brighter than $`B`$ $`=`$ 23.0 in one or both of the models presented in Section 3. The entries are sorted by hole category, as taken from Table 4.
A number of holes characterized in Table 4 as category 1 (no objects visible in the hole) are seen to have limiting magnitudes significantly fainter than the expected brightness of the putative clusters. In other words, if the SNe hypothesis for the origin of the holes is correct, one should definitely be able to see the clusters in a number of cases. Specific holes for which this test fails include 10, 13, 47, 49, and 50. In all these cases, the observational limits are more than one magnitude fainter than the predicted values for the case of Model II (8 M mass limit for SN). If the light from the evolved cluster stars is taken into account, the differences between the predicted brightnesses and the observed limits becomes even greater. It is interesting to note that these five holes are among the most energetic in Ho II, each requiring between 32 and 65 supernovae to create them in the P92 scenario.
For the specific H I holes mentioned above, we can clearly rule out the presence of the predicted star clusters at the expected levels. Therefore, it seems extremely unlikely that these holes were created by multiple SNe as hypothesized by P92. This is perhaps no great surprise, since the holes in question are located in regions of extremely low surface brightness. In fact, most lie outside of the Holmberg radius of Ho II. At these large distances from the galaxian center, the H I gas column densities are quite low, well below the empirical limits suggested by Kennicutt (1989) necessary for star formation to occur. In general, little star formation should occur beyond the Holmberg radius, so that the existence of multiple massive star clusters in the outer regions of Ho II, as predicted by P92, would be at odds with what we know about star formation in other galaxies.
The limits we can place on the presence of star clusters in the inner H I holes are less interesting for two reasons. First, the galaxian background is much higher there, making it easier to hide the presence of any cluster light. Second, the inner holes tend to be smaller in size and hence require fewer SNe to create them. Consequently the brightnesses predicted for these holes tend to be fainter than our observational limits. Therefore, we cannot rule out the possibility that some of these inner H I holes are produced by SNe. There is of course no compelling reason for believing that all the holes have the same origin; we can only adequately test the SNe hypothesis for the outer H I holes. A few holes for which the observations may actually support the SNe hypothesis of P92 are numbers 21, 36, 44, and 48. These are listed in Table 6. Holes 21 and 48 are the most likely candidates, although hole 21 is in a very crowded region with many point sources in the optical images. The sources in holes 36 and 44 are both significantly brighter than the model predictions.
Further evidence against the SNe hypothesis comes from the H$`\alpha `$ image, which is shown in Figure 3 with the H I hole apertures superposed. Given the predicted ages and energetics of the holes, one might expect that at least some of the holes would exhibit diffuse H$`\alpha `$ emission. No such emission is seen in any of the holes, with the possible exception of the few holes that are coincident with H II regions (e.g., holes 16, 20, and 43). The locations of H$`\alpha `$ emission trace out the regions of high H I column density seen in Figure 1, indicating that the current star formation is occurring at local density maxima in the neutral gas distributions. P92 interpreted this as SNe-induced star formation caused by the compression of swept-up gas. We note, however, that this interpretation does not appear to be consistent with most of the holes, since only the minority have H$`\alpha `$ emission nearby.
ROSAT observations of Ho II have failed to detect the presence of any diffuse X-ray gas inside the HI cavities (F. Walter & J. Kerp, private communication). Such X-ray emission would be expected if the holes were filled with hot coronal gas from the SNe explosions. Furthermore, Stewart et al. (1997) analyzed far-ultraviolet (FUV) images of Ho II taken with the Ultraviolet Imaging Telescope and found no bright FUV knots located within the H I hole locations identified by P92. They found instead that bright FUV emission, if present, was likely to occur outside the hole boundaries.
### 4.2 Alternative Explanations
The observational evidence presented here strongly suggests that at least some of the H I holes in Ho II are not caused by multiple SN explosions in the manner envisioned by P92. Here we discuss possible alternative explanations for the presence of these features.
#### 4.2.1 Modified Hole Energetics
One possibility is that the SNe hypothesis is correct, but that the numbers published by P92 for the energetics of the H I holes are systematically overestimated. The energies and ages derived by P92 depend critically on the observed expansion velocities of the holes. These are difficult to measure precisely, and in some cases the evidence for the expansion is weak at best. If the measured expansion velocities given in P92 are systematically too high, then the holes might actually be substantially older, and the true required energies would be significantly reduced. This being the case, it might be possible to reduce the expected number of SNe, and hence the brightness of the remnant star clusters, below the observational limits found from our data. However, for the largest holes, a reduction in the total number of SNe of more than a factor of $``$5 is required. This would then imply an overestimation of the expansion velocities by a factor of more than three. Such a large error in the expansion velocities seems unlikely. Furthermore, the additional factors mentioned above regarding star formation beyond the Holmberg radius and the lack of H$`\alpha `$ and X-ray gas could still pose problems for this interpretation.
Another possibility is that the energetics of the holes have been overestimated not because of errors in the the expansion velocities, but because of uncertainties in the energy calculation itself. Estimates for the amount of energy imparted to the ISM by stellar winds and supernovae are based on our knowledge of the efficiencies of these processes, which are not well-constrained. P92 calculated the energy associated with each of the H I holes in Ho II using an expression derived from a hydrodynamical model by Chevalier (1974), which describes the evolution of a single spherically-symmetric supernova remnant in a uniform medium. It seems possible that using the results from such a model to calculate the total energies associated with H I holes in the non-uniform ISM of a galaxy like Ho II could introduce uncertainties of at least a factor of a few. As explained in the previous paragraph, if the energies of the holes have been overestimated, then fewer SNe may be needed to create the H I holes, which could push the putative cluster brightnesses below our observational limits. As before, however, this explanation still requires that some of the H I holes be created by supernovae occurring well beyond the Holmberg radius, in regions of the galaxy where star formation does not appear to have occurred in the past.
#### 4.2.2 A Non-Standard IMF
Another way to retain the SNe hypothesis would be to invoke an unusual IMF for the putative clusters. This possibility was actually suggested by P92. A top-heavy IMF, rich in massive stars but poor in low-mass ones, could explain the observations. However, there is as yet no real evidence for significant variations in the IMF as measured in different environments in the Milky Way and other nearby galaxies (Leitherer (1998)). Furthermore, since the stars which would be providing the expected cluster signature are themselves fairly massive, the IMF slope required to produce the requisite number of SNe would have to be quite severe, perhaps even inverted, in order to not produce a detectable population of B and A stars. Invoking such an unusual IMF is not a very compelling explanation.
Since the SNe hypothesis has significant problems, even when allowances for the above variants are made, we are forced to consider the alternative that the outer H I holes are not produced by the action of stellar winds and SNe. We again stress that our analysis does not exclude the possibility that the inner holes are produced by SNe/winds.
#### 4.2.3 Gamma-Ray Bursts
One possible explanation that has gained attention recently is the suggestion that the holes are remnants of Gamma-Ray Burst events (GRBs). Recent work by Loeb & Perna (1998) and Efremov et al. (1998) proposes that some of the HI supershells and hole features seen in nearby galaxies (such as the dwarfs in our sample) are remnants of GRBs. These authors suggest that GRBs might be associated with the release of gravitational binding energy during, for example, the collapse of a single, massive star to a black hole. They argue that such an event could produce a blast wave with energy comparable to the multiple-SN events thought to be necessary to produce the H I hole features. Since this proposal does not necessarily require that a large star cluster be left behind after the explosive event, the observational test applied here does not rule it out. Because only a single star can account for the GRB, one could hypothesize that the explosions which create the H I holes do not occur in massive star clusters, but rather in smaller associations which we would have no hope of detecting in our data. However, like the SN scenario, the GRB scenario requires that massive stars be present at the centers of the H I holes, in order to produce the expanding blast wave that creates the hole. As we have noted, many of the holes in Ho II occur in extremely LSB regions which show no indication of recent massive star formation. In addition, the GRB hypothesis would still predict the presence of hot X-ray emitting gas within the H I cavity. Hence, although the GRB hypothesis is an attractive alternative to the multiple-SNe scenario, and may well explain some of the H I holes seen in Ho II, we consider it unlikely to be correct for the outer holes, i.e., the same ones for which our current study has ruled out the SNe hypothesis.
#### 4.2.4 Impacts from High-Velocity Clouds
If stellar energy sources are ruled out, the next two most likely explanations for the H I holes are large-scale dynamical effects and ionization. The dynamical effects of a collision between an infalling neutral gas cloud and a galactic disk, resulting in an H I hole, were modeled by Tenorio-Tagle (1980, 1981). Clear associations between H I holes and high-velocity clouds are known in the Milky Way (Heiles 1985), M101 (van der Hulst & Sancisi 1988; Kamphuis, Sancisi, & van der Hulst 1991), NGC 628 (Kamphuis & Briggs 1992), NGC 6946 (Kamphuis & Sancisi 1993), and NGC 5668 (Schulman et al. 1996). The high-velocity clouds could be primordial material or the result of a galactic fountain.
This mechanism is attractive for explaining the holes at large galactocentric radii in Ho II, where there is very little starlight and very little likelihood for star formation. One of us (DJW) has re-examined the data cubes from P92 to search for candidate high-velocity clouds. The data cube with the greatest sensitivity, with a synthesized beam of 28 $`\times `$ 27 arc seconds and pixels of 10 arc seconds, was examined in the velocity range 64 to 244 km s<sup>-1</sup>. The root-mean-square (RMS) signal in an apparently line-free channel was 1.88 mJy beam<sup>-1</sup>. Candidate clouds were found by first making a statistical search for bright pixels, then searching for extended bright areas around those pixels. A candidate was required to have extended, superimposed signal in at least three adjacent channels.
Only one candidate cloud met our requirements. It is located at 08<sup>h</sup> 16<sup>m</sup> 37<sup>s</sup> 70° 42′ 36″, with a central velocity of 223.5 km s<sup>-1</sup> and a total velocity width of 7.7 km s<sup>-1</sup>. It is 17′ from the center of Ho II, giving a projected separation of about 15 kpc at the assumed distance of 3.05 Mpc. Pixels brighter than the background RMS are found in three channels. The peak intensity is 9.81 mJy beam<sup>-1</sup> (5.2 times the RMS) in the raw maps, and 19.63 mJy beam<sup>-1</sup> after correction for the primary beam. The maximum brightness temperature is 2.4 K. The cloud’s integrated H I line signal is 4.9 Jansky km s<sup>-1</sup>, giving a total H I mass of 1.2 $`\times `$ 10$`{}_{}{}^{7}M_{\mathrm{}}^{}`$. Its total angular extent is 92″ or 1.4 kpc. The properties of this candidate cloud are similar to those of the Milky Way high-velocity clouds described by Wakker & van Woerden (1997).
This cloud is not easily visible in the data cube, so no mention was made of it in P92. We wish to emphasize that this is a marginal detection at best, and must be confirmed by an independent observation before it can be considered anything but a candidate cloud. More than two million apparently empty pixels were searched for local peaks which might be candidate clouds. Only two pixels were found to be as bright as 5.2 times the RMS, and of these, only one had extended signal in three channels. In a sample of this size, two pixels as bright as 5.2 times the RMS are expected if the noise is Gaussian. There is a good possibility that we have simply found a noise peak rather than a real cloud. Even so, the candidate cloud can be used to set limits on the presence of high-velocity clouds within the observed velocity range.
The cloud has a velocity of 65 km s<sup>-1</sup> relative to Ho II, which has a systemic velocity of 158 km s<sup>-1</sup>. Its kinetic energy in the frame of Ho II is 5 $`\times `$ 10<sup>46</sup> Joules or 5 $`\times `$ 10<sup>53</sup> ergs, large enough to cause the largest holes in Ho II. The possible detection of a single candidate cloud with enough energy to cause a hole does not prove or disprove the hypothesis that infall caused the holes. Infall could be episodic, or there could be clouds outside the observed velocity range. It does show the observational difficulty in identifying high-velocity clouds in galaxies at the distance of the M81 group — even in observations as deep as those of P92, candidate clouds are nearly indistinguishable from 5-$`\sigma `$ noise peaks.
#### 4.2.5 Large-Scale Turbulence
Holes might be unavoidable due to the nature of the interstellar medium. It is well known that molecular clouds are fractal — see Beech (1987), Bazell & Désert (1988), Falgarone (1989), Scalo (1990), Falgarone, Phillips, & Walker (1991), and Elmegreen & Falgarone (1996). It is becoming clear that H I is fractal as well — see Vogelaar & Wakker (1994) and Westpfahl et al. (1999). The fractal dimension of the ISM is similar to that of structures seen in laboratory turbulence, which has led Elmegreen & Efremov (1999) and others to conclude that interstellar clouds form by processes related to turbulence. The processes which cause fractal structure, including turbulence, usually produce an internal distribution of holes, characterized as lacunarity by Mandelbrot (1983). If H I clouds are produced by processes related to turbulence, the holes may be a manifestation of the formation process.
We note that the fractal nature of H I distributions may change the energy required to form a hole via supernovae and stellar winds. A fractal structure with significant lacunarity may provide natural chimneys through which supernova ejecta can flow, thus significantly increasing the amount of energy injection needed to form an expanding shell.
#### 4.2.6 Ionization
Holes might also be formed by ionization. A source of ionizing photons in the outer regions of Ho II might be the intergalactic UV field. The observed column densities in the HI gas at the locations of these outer holes is of order a few times 10<sup>20</sup> cm<sup>-2</sup>, significantly above the densities at which the UV radiation field can keep a large fraction of the HI ionized for an HI disk of normal thickness. However, P92 argue convincingly that the scale height of the gas in Ho II is significantly larger than in a typical spiral disk. If this is correct, the actual volume density of the gas in the outer regions would be significantly less than that for gas in a spiral disk with the same measured column density. In this case, it might be possible for lower density pockets of gas to approach the threshold volume density below which the ionization fraction of the gas remains high in steady state due to the ionizing UV photons. In other words, once a pocket of low density gas is created in the outer parts of a puffed-up disk as is envisioned for Ho II, it could become ionized by the intergalactic UV radiation field and remain highly ionized for a long time. Eventually the holes would be destroyed by dynamical processes (e.g., rotational shear), but in the outer parts of the galaxy they might well maintain their integrity for a substantial length of time due to the slow rotation speeds in low-mass galaxies like Ho II.
#### 4.2.7 An Unresolved Question
The actual origin of the H I holes in the outer parts of Ho II remains an open question. In the present study, we suggest strongly that at least some of the holes are not produced by the combined action of stellar winds and SNe explosions. In many ways the SNe hypothesis is a natural and sensible explanation for the holes, which was perhaps why it was almost universally accepted on face value. However, in several instances it clearly fails the direct observational test which we have applied. Additional work needs to be done on this problem in order for a clearer picture of the nature of these large-scale features to be developed. Our current lack of understanding leaves open a number of questions regarding the evolution of the ISM in irregular galaxies, and in particular the actual role of feedback from massive stars in shaping the ISM.
## 5 Summary & Conclusions
We have carried out a deep, multi-color imaging study of Ho II, a dwarf galaxy in the M81 group which has been shown to contain a large number of expanding holes in its neutral hydrogen distribution. The formation of the H I holes in Ho II and other galaxies like it has been attributed to multiple SNe occurring within wind-blown shells around young, massive star clusters. To search for evidence of the clusters, we have compared our optical images with the published H I maps, and have measured accurate magnitudes and colors of all objects in and around the H I holes. Photometry of 44 hole regions in Ho II reveals that at least 16 holes contain no detectable point sources brighter than $`B`$ $`=`$ 23.0. Ten of these holes are located beyond the Holmberg radius. An additional 21 holes contain only red ($`B`$$``$$`V`$ $`>`$ 1.0) sources, which are most likely either faint foreground stars or diffuse galaxian background emission from Ho II. Only 6 holes contain sources which could be interpreted as being young clusters of stars with the requisite brightness and color.
Comparison of models which predict the brightness of the putative star clusters with the observational limits obtained from our imaging data appear to rule out the SN scenario as being the cause of at least several of the most substantial H I holes. While convincing arguments cannot be made against the SNe scenario for the majority of the holes due to the lack of depth of our images coupled with severe crowding in the central portions of the galaxy, the fact that at least several holes appear to require an alternative explanation for their origin raises doubts about the SNe scenario in general. The lack of diffuse H$`\alpha `$ and X-ray emission from any of the holes further supports the possibility that the SNe scenario may be incorrect. Recent suggestions that the H I holes in galaxies like Ho II are caused by the events which create Gamma-ray bursts are also not favored by the current findings, although such scenarios are more difficult to rule out with the optical data since they require only a single massive star. A number of other alternative explanations for the existence of the H I holes are explored, including errors in the hole energetics, non-standard IMFs, dynamical processes such as large-scale turbulence or impacts from high-velocity clouds, and ionization. None of these alternatives is clearly favored at this time, and the origin of the H I holes remains an open question.
There is no doubt that energy input from massive stars plays a major role in shaping the ISM in galaxies. The current study, however, suggests that one must interpret the observational evidence for such influence carefully. Although the scenario proposed by P92 appears sensible, it makes a direct observational prediction which is not verified by the current study. The precise role that winds from massive stars and SNe shocks play in sculpting the gaseous distribution in galaxies remains an open question, calling for continued careful work on both the observational and theoretical fronts.
We are grateful for the professional support of the staff of Kitt Peak National Observatory during our two observing trips. This research has made use of the NASA/IPAC Extragalactic Database (NED) which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. KLR, JJS and LAR acknowledge with gratitude financial support from Wesleyan University, the Keck Northeast Astronomy Consortium, the Research Corporation, and the National Science Foundation, all of whom provided partial support for this project. DJW gratefully acknowledges financial support from the New Mexico Space Grant Consortium, the Kahlmeyer Foundation, and the Research and Economic Development division of New Mexico Tech. We have benefited from discussions with many colleagues, including D. Puche, R. Larson, S. Van Dyk, F. Walter, and E. Brinks. JJS would like to recognize L. van Zee and NRAO-Socorro for their hospitality during the preparation of portions of this paper. We would also like to thank the anonymous referee for useful comments.
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.